MACROECON 1n

I Sorensen-Whitta-Jacobsen:
Introducing Advanced
I Part 5- The Building
Blocks for the Short-run
I 14. The economy in the
short run: Some facts about
I © The McGraw-Hill
Companies. 2005
I e
Book
The Short Run

Economic Fluctuations,
· Short-run Unemployment and
Stabilization Policy
395
0 I Sorensen- Whitta- Jacobsen:
Part 5- The Building
Blocks for the Short- run
14. The economy in the
© The McGraw-Hill
Companies. 2005
Macroeconomics Model business cycles
Sorensen-Whitta-Jacobsen: I Part 5- The Building 14. The economy in the © The McGraw-Hill
Introducing Advanced Blocks for the Short-run short run: Some facts about Companies. 2005
Chapter
The economy in the

short run
Some facts about business cycles
S
ince the time of the Industrial Revolution the Western world has experienced a
tremendous growth of total output. In Book One we focused on this long-run aspect
of economic development. But history tells us that economic growth has been far
from steady. In the short and medium term the growth rate fluctuates considerably, as
you can see from Fig. 14.1 which plots quarterly data for the logarithm of real GDP for a
number of Western countries. If the growth rate of the economy were constant, the log of
GDP would follow the straight lines in Fig. 14.1. The fact th at the graphs lor the log of
actual GDP are sometimes steeper and sometimes flatter than the straight lines reflects
th at periods of rapid growth tend to alternate with in tervals of slow growth . Indeed, the
graph has frequently had a negative slope, indicating that the growth rate sometimes
even becomes negative. Note that the data underlying Fig. 14.1 are seasonally adjusted,
so the fluctuations do not reflect the regular changes in business activity occurring with
the changing seasons of the year, for example, the seasonal s\>Vings in construction
activity due to changing weather conditions. Thus there are more fundamental forces
causing an uneven pace of economic growth.
The rest of this book studies these short-term fluctuations in economic activity,
co nun only known as business cycles. How can we explain that the state of the economy
repeatedly alternates between business cycle expansions characterized by rapid growth.
and business cycle co11trnctions or recessions characterized by declining econ omic activity?
To answer this question is one of the basic challenges of macroeconomic theory.
Wh y understan ding business cycles is in1portant

................................................................................................................................................................................
The fact that economic gro\ovth is repeatedly interrupted by recessions is a major source of
concern for economic policy makers and the general public. since recessions bring consid-
erable economic hardship to workers who lose their jobs. to entrepreneurs and home-
m.vners who go bankrupt, and to ordinary consumers who suffer capital losses on their
assets. Even for those who are not directly allected by layoffs and bankruptcies. recessions
may cause a decline in well-being by generating fears ofj ob losses and offuture reductions
397
eI Sorensen-Whitta- Jacobsen:
© The McGraw-Hill
Companies. 2005
398 INTRODUC ING ADVANCE D MACROECONOM ICS
log of GOP United Kingdom log of GOP USA

12.6 9.5
12.4
9
12.2
12 8.5
1 1.8 8
1 1.6
7.5
1 1.4
1 1.2 7 ,...
<D
10 ""
~
00
<D
,...
~ 0
00
<D
00
('ol
0>
CX)
0> ~
<D
10
10
<D
,...
~ (')
00
('ol
0> 00
Ol 0:· Ol Ol Ol Ol 0> Ol Ol Ol Ol Ol Ol 0>
~ ('ol
log of GOP Denmark log of GOP Finland

7.1 10.5
10.4
7
10.3
6.9
1U.:l
6.8 10.1
10
6.7
9.9
6.6
9.8
6.5
;:
,...
,... (') 0> 10 ;;
9.7
00 a;
,... ,...
10 ~
,...
CX)
0 (') <D Ol ('ol
CX) CX) Ol 00 Ol Ol Ol Ol 0
Ol Ol 0> 0> Ol 0 Ol Ol 0> Ol 0> Ol Ol Ol Ol 0
('ol
""
log of GOP Belgium log of GOP Netherlands
11.1 11.5
11.4
11
11.3
10.9
11.2
10.8 11 .1
11
10.7
10.9
10.6
10.8
10.5 10.7 ,...
0
00
(')
00
<D
00
Ol
00 ""
Ol
10
Ol
CX)
0> 00 ,... 0
CX)
(')
00
<D
00
Ol
00
('ol
0>
10
Ol
00
Ol ;;
Ol Ol Ol Ol 0> 0
~
Ol 0> Ol Ol 0> Ol Ol Ol Ol
('ol
""
Figure 14.1: The evolution of real GDP in some W estern countries
Source: See Table 14.2.
in income and wealth. Understanding business cycles is therefore not only of academic
interest; it may also help the economist to oller advice to policy makers on the possibility
of reducing business fluctuations through macroeconomic stabilization policy, that is,
active monetary and tiscal policy. At the very least an insight into the workings of the
business cycle may enable the economist to suggest how policy makers can avoid
amplifying the business cycle through misguided macroeconomic policies.
Introducing Advanced Blocks for the Short- run short run: Some facts about Companies. 2005
14 TH E EC ONOMY IN TH E SHORT RUN 399
On several occasions in history, recessions have developed into severe economic

depressions paving the way for social and political disaster. A glance at Fig. 14.2 should
convince you why it is important to understand the causes of depressions and the means
to avoid them. The figure shows a striking correlation between the unemployment rate in
interwar Germany and the share of total votes in Reichstag elections going to Adolf
Hitler's Nazi Party. In the 1928 election, when unemployment stood at the low level of
2.8 per cent. the Nazi Party captured only 2 .6 per cent of the votes and were not consid-
ered a serious political force. But as the democratic system of the Weimar Republic proved
unable to prevent the mass unemployment and human suffering caused by the Great
Depression, a rapidly growing number of Gennans became receptive to Hitler's radical
critique oftbe parliamentary system. By 19 33 . with unemployment close to 30 per cent of
the labour force. Hitler obtained a vote share of almost 44 per cent in the last free election
before he established his Nazi dictatorship and steered Germany towards the Second
World War. Although there were several other factors explaining Hitler's rise to power,
there is no doubt that the economic depression made it easier for him to gather support.
The Great Depression of the 1930s was exceptional in its severity and in its social and
political consequences. Nevertheless we have several recent examples of economic down-
turns wh!ch have caused social upheaval, including the South-east Asian crisis of
199 7- 9 8 which brought down the Indonesian government and led to serious civil
unrest, and the economic crisis in Argentina in 2001 which forced the government to
resign after riots in the streets. By studying business cycles we will not only learn more
about the workings of a market economy; we will also improve our understanding of the
general course of social and political events.
The reason why it mal<es sense to theorize about business cycles is that, even though
no two business cycles are identical. they usually have some important features in
%
50
40
30
20
10
0+---.--.---.---.---.--.---.--.,--.---.--~
1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934
Year
Figure 14.2: Unemployment and political extremism in Germany, 1924-1 933

Note : Nazi vote share in 1932 is an average across two elections.
Sources: Unemployment rate from B .R. Mitchell, International Historical Statistics, Europe 1750- 1988, Macmillan,
New Yorl<, 1993. Nazi vot e share from Richard F. Hamilton, Who Voted for Hitler?, Princeton University Press, 1982.
eI Sorensen-Whitta-Jacobsen:
© The McGraw-Hill
Companies. 2005
400 IN TRODUCING ADVANCED MACROECONOMICS
common. Nobel laureate Robert Lucas of the University of Chicago made this point in the
following way:
Though there is absolutely no theoretical reason to anticipate it, one is led by the facts to con-
clude that, w ith respect to the qualitative behavior of comovements among series (economic
variables), business cycles are all alike. To theoretically inclined economists, this conclusion
should be attractive and challenging, for it suggests the possibility of a unified explanation of
business cycles, grounded in the genera/ laws governing market economies, rather than in
political or institutional characteristics specific to particular countries or periods. 1
In the rest of this chapter we will describe some of those co-movements of economic
variables which are characteristic of business cycles. Before we start theorizing about
business cycles, we want to get some idea of the phenomenon which our theory is
supposed to explain. We will begin in the next section by restating a definition of business
cycles which has become familiar in the literature. We will then move on to the question
how we can measure business cycles in quantitative terms. That is, bow can we separate
short-term business cycle fluctuations in economic activity from the long-term economic
growth trend? Following this. we '.viii be ready to describe in quantitative terms the
co-movements of important economic variables during a 'typical' business cycle.
14.2 Defining business cycles

•.............................................................................................................................................................................. .
In a famous book which became a milestone in empirical business cycle research, the
American economists Arthur Burns and Wesley Mitchell gave the lollmoVing classic
definition of business cycles:
Business cycles are a type of fluctuations found in the aggregate economic activity of nations
that organize their wo rk mainly in business enterprises: a cycle consists of expansions
occurring at about the same time in many economic activities, followed by similarly general
recessions, contractions, and revivals wh ich merge into the expansion phase of the next
cycle; this sequence of changes is recurrent but not periodic; in duration business cycles vary
from more than one year to ten or twelve years; they are not divisible into shorter cycles of
similar character with amplitudes approximating their own. 2
This definition of business cycles emphasizes several points.
• Aggregate economic activity: Business cycles are characterized by a co-movement of a

large number of economic activities and not just by movements in a single variable like
real GDP.
• Organization in business enterprises: Business cycles are a phenomenon occurring in
decentralized market economies. Although they had several other economic problems.
the former socialist economies of Eastern Europe did not go through business cycles of
the type known in the Western world.
1. Robert E. Lucas, Jr., 'Understanding Business Cycles', in K. Brunner and AH. Meltzer, eds., Carnegie-Rochester
Conference Series on Public Policy, 5, Autumn 1977, p. 10.
2. Arthur F. Bums and Wesley C. Mitchell, Measuring Business Cycles, National Bureau of Economic Research
(NBER), New York, 1946, p. 1. Mitchell was a principal founder of the NBER, and Burns was a student of his who
later became chairman of the US Federal Reserve Board from 1970 until 1918.
Sorensen- Whitta- Jacobsen: I Part 5- The Building 14. The economy in the © The McGraw-Hill
• Expansions and contractions: Business cycles are characterized by periods ofexpansion of

economic activity followed by periods of contraction in which activity declines.
• Duration of more than a year (persistence): A lull business cycle lasts for more than a year.
Fluctuations of shorter duration do not have the features characteristic of business
cycles. This means that purely seasonal variations in activity within a year are not
business cycles. We may also say that business cycle movements display persistence:
once an expansion gets going, it usually lasts for some time during which the expan-
sionary forces tend to be self-reinforcing, and once a contraction sets in, it tends to
breed further contraction for a while.
• Recurrent but not periodic: Although business cycles repeat themselves. they are far from
being strictly periodic. The duration of cycles has varied from slightly more than a year
to 10- 12 years. and the severity of recessions has also varied considerably, with
recessions sometimes (but not always) turning into depressions.
j 4.3 .I:>..CI.~i.~~--~-l:l.S..i~~S.S..9'<::~~S. ............................................................................................................. .

The contribution of Burns and Mitchell was to document the movements over time of a
large number of economic variables. Through their work it became possible to ident({y the
turning points in economic activity and hence to offer a dating of business cycles. Since the
movements of the dillerent economic variables are not perfectly synchronized, it is a
matter of judgement to identity the point in time at which the business cycle reaches its
peak and moves from expansion into contraction, and to determine when it reaches its
trough. moving from contraction (recession) to expansion. Building on the tradition
established by Burns and Mitchell. the US National Bureau of Economic Research (NBER)
has for many years had a Business Cycle Dating Cormnittee consisting of experts in
empirical business cycle research. The NBER committee detlnes a recession (contraction)
as a period of signiHcant decline in total output, income. employment and trade. usually
lasting from six months to a year. and marked by widespread contractions in many sectors
of the economy. On this basis the conunittee has arrived at a dating of US business cycles
between 1854 and 2001. This dating is reproduced in Table 14.1.
The length of the business cycle is measured from trough to trough. and the last
column of the table measures the duration of the expansion phase relative to the duration
of the contraction phase. Table 14.1 illustrates the point stressed by Burns and Mitchell:
business cycles are far from regular and periodic. The duration of the cycle varies greatly,
and though the expansion phase usually lasts longer than the contraction phase -
reflecting the economy's long-term potential for growth - there are also examples of cycles
where the contraction has lasted longer than the previous expansion. At the bottom of the
Great Depression in March 1933. the US economy had been contracting tor 43 months.
During this economic nightmare, real GDP fell by almost 30 per cent. and u nemployment
rose to 25 per cent. Notice, however, that the contraction of the 18 70s lasted considerably
longer than the Great Depression, alth ough economic historians tell us that the decline in
activity was less catastrophic.
The dates in Table 14.1 indicate that business cycle expansions have tended to last
longer and that contractions have on average been shorter after the Second World War
I Part 5- The Buil ding
Companies. 2005
402 IN TRODUC ING ADVANCED MACROECONOMICS
Table 14.1: US business cycle expansions and contractions

Business cycle reference dates Duration in months
Trough Peak Trough 1. Expansion 2. Contraction Cycle 1 1./2.
December 1854 June 1857 December 1858 30 18 48 1.67

December 1 858 October 1860 June1861 22 8 30 2.75
June1861 April1865 December 1867 46 32 78 1.44
December 1867 June 1869 December 1870 18 18 36 1.00
December 1870 October 1873 March 1879 34 65 99 0.52
March 1879 March 1882 May 1885 36 38 74 0.95
May 1885 March 1887 April1888 22 13 35 1.69
April1888 July 1890 May 1891 27 10 37 2.70
May 1891 January 1893 June 1894 20 17 37 1.18
June 1894 December 1895 June 1897 18 18 36 1.00
June 1897 June 1899 December 1900 24 18 42 1.33
December 1 900 September 1902 August 1904 21 23 44 0.91
August 1904 May 1907 June 1908 33 13 46 2.54
June 1908 January 191 0 January 191 2 19 24 43 0.79
January 1912 January 1913 December 1914 12 23 35 0.52
December 1 914 August 1918 March 1919 44 7 51 6.29
March 1919 January 1920 July1921 10 18 28 0.56
July1921 May 1923 July 1924 22 14 36 1.57
July 1924 October 1926 November 1927 27 13 40 2.08
November 1927 August 1929 March 1933 21 43 64 0.49
March 1933 May 1937 June 1938 50 13 63 3.85
June 1938 February 1945 October 1945 80 8 88 10.0
October 1945 November 1 948 October 1949 37 11 48 3.36
October 1949 July 1953 May 1954 45 10 55 4.50
May 1954 August 1957 April1958 39 8 47 4.88
April1958 April1960 February 19 61 24 10 34 2.40
February 1961 December 1969 November 1970 106 11 117 9.64
November 1970 November 1973 March 1975 36 16 52 2.25
March 1975 January 1980 July 1980 58 6 64 9.67
July 1980 July1981 November 1982 12 16 28 0.75
November 1982 July 1990 March 1991 92 8 100 11.5
March 1991 March 2001 November 2001 120 8 128 15.0
Average for pre-First World War period (15 cycles) 2 25 23 48 1.10
Average for interwar period (5 cycles) 3 26 20 46 1.30
Average for post-Second World War period (1 0 cycles) 4 57 10 67 5.50
1. The duration of the full business cycle is measured from trough to trough.
2. December 1854 to December 19 14.
3. M arch 1919 to June 1938.
4. October 1945 to November 2001 .
Source : 1\ational Bureau of Economic Research.

than before that time. Until recently. the expansion from February 19 61 to December
1969 was the longest economic boom in US history. That record was beaten by the ten-
year long expansion starting in March 1991. Although the subsequent recession lasted
only eight months. the latest cycle was the longest business cycle on record in the US.
The long US boom of the 1990s inspired many commentators to speculate about the
arrival of a 'New Economy' in which the expansionary forces stemming from innovations
in information technology were so strong that serious recessions would be a thing of the
past. at least in the US. Interestingly, the long boom of the 1960s gave rise to a similar
unfounded optimism. In 1967. several leading US economists gathered for a conference
asking: 'Is the Business Cy cle Obsolete?'. 3 But the recession in the US and in many other
countries at the beginning of this decade shows that the business cycle is still with us.
Let us therefore move from the dating of business cycles to the problem of quantitative
measurement of business fluctuations.
14.4 .~.~.~~~~.~.~.~. ~~.~~~~~~. ~Y..~~·~·~································ ..·································································

Most economic time series fluctuate around a growing time trend. The grm>Vth trend
reflects the forces described in the theory of economic growth. while the task of business
cycle theory is to explain the lluctuations around that trend. For example. if Y, is real GDP
in period t. it is useful to think of Y, as the product of a growth component Yf indicating the
trend value which Y 1 would assume if the economy were always on its long-term growth
path, and a cyclical component Y~ which fluctuates around a lon g-run mean value of 1:
Y, = Yf · Y~. (1)
Our assumption on the mean valueof Y~ implies that Y1 = Yf on average. Equation (1)
also implies that as long as the amplitude of the fluctuations in the cyclical component y;·
remains constant, the absolute amplitude ofthe business cycle fluctuations in real GDP will
rise in proportion to the trend level of output so th at the percentage deviations of actual
output from trend output over the business cycle will tend to stay constant over time.
It will be convenient to work with the natural logarithms of the various variables
rath er than with the variables themselves. because ch anges in the log of some variable
X approximate percentage changes in X. Taking Jogs on both sides of equation (1) and
detlning y, =In Yl' g, = ln Y f. and c, = ln Y~. we get:
y, = g:+c,. (2)
In this section we will discuss how we can estimate the trend component g, and the
cyclical component c1 separately. given that we only have observations of y 1• Let us start
by going back to Fig. 14.1. The straight lines in that figure are regression lines with a slope
equal to the average growth rate over the period of observation. Technically, the intercept
and the slope of the regression line are chosen so as to minimize the sum of the squared
deviations between the observed values of the log of GDP and the points on the line.
3. The conference participants tended to answer the question in the affirmative. The conference papers were published
in Martin Bronfenbrenner (ed .), Is the Business Cycle Obsolete?Wiley, New York, 1969.
© The McGraw-Hill
Companies. 2005
404 INTRODUC ING A DVANCE D M ACROECONOM ICS
It might be tempting to let the straight regression line represent the trend value of
output and to measure the cyclical component of GDP as the deviation from that hypo-
thetical steady growth path. But a moment's reflection should convince you that this is
not a satisfactory procedure. Recall that along the straight regression line the economy's
real growth rate is constant. If we take the regression line to represent the trend growth
path, we are therefore postulating that the economy would always be in a steady state
equilibrium 'oVith a constant growth rate if it were not disturbed by business fluctuations.
However. the theory of economic growth gives no reason to believe that the economy
is always in a steady state. Conventional growth theory tells us that the economy's
growth rate will decline over time if it starts out with a capital- labour ratio below the
steady-state level. and vice versa. Moreover. the modern theory of endogenous growth
suggests that the rate of technical progress may vary with the endogenous innovation
activity oflirms. Indeed, even if major technological innovations are exogenous to the eco-
nomic system, they are unlil<ely to arrive at an equal pace over time, and this is suft1cient
reason to discard the assumption of a constant long-term growth rate. An inspection of
Fig. 14.1 also suggests that the longer-term movement of the economy does not follow a
straight line, even if we abstract from the short-term ups and downs of the graphs lor real
GDP.
Hence we need a more sophisticated method for separating the growth trend from the
cyclical component of a variable like GDP. Since we do not perfectly understand how the
economy works. we cannot claim that there is a single objectively correct way of separat-
ing c, from g ,. Still, our reasoning above suggests that we need a method which allows for
variation over time in the underlying growth trend, but which nevertheless ensures that
the short-term fluctuations are categorized as temporary cyclical deviations from trend.
One such method which has become 'oVid ely used in recent years is the so-called Hodrick-
Prescott filter, named after American economists Robert Rodrick and Edward Prescott
who popularized its use. 4 Under this method of 'filtering' (that is, detrending) an economic
time series lil<e(y,);~ 1 lor the log ofGDP, the growth component g 1 is determined by mini-
mizing the magnitude:
2
1' c, T_ 1 [ change in trend growth rate l
HP = L (H) 2 +A L from period t to period 1 + 1 {3)
t-L 1-2 (gl+ l _ gl) _ (g, _ g, _I )
with respect to all of the g1• where observations are available for the time periods t = 1,
2, ... , T, and where A is a parameter which is chosen by the observer. Note that since y 1 is
measured in logarithms, the magnitudes gI+ 1 - g, and g1 - g 1 _ 1 are approximately the
perce11tage growth rntes of the trend value of real GDP in periods t + 1 and t. respectively.
The term in the square bracket in (3) thus measures the change in the estimated trend
growth rate from one period to the next. Note also that. by dellnition, the term y 1 - g 1 in ( 3)
measures the cyclical component c, of log(GDP) in period t. Minimizing the expression in
(3) therefore forces us to compromise between two objectives. On the one h and. we want
to choose the g,'s so as to minimize ch anges in the estimated trend grmvth rate over time.
4. Robert J. Hodrick and Edward C. Prescott, 'Post·war U.S. Business Cycles: An Empirical Investigation', Journal of
Money, Credit, and Banking, 29, 1997, pp. 1-16 (originally published as a worl<ing paper in 1980).
Sorensen-Whitta-Jacobsen:
© The McGraw-Hill
Companies. 2005
lCD
14 THE ECONOMY IN THE SHORT RUN 405
since this will minimize the expression in the second term in (3). On the other hand, we
want to bring g 1 as close as possible to (the log oO actual output y 1 so as to minimize the
first sum in (3). The relative weight placed on each of these conflicting objectives depends
on our choice of il.
Consider first the extreme case where we set A= 0. In this special case we will mini-
mize HP by simply choosing g, = y, for all t = 1, 2 . ... , T. since HP will then attain its lowest
possible value of zero. But this amounts to postulating that all observed fluctuations in y,
reflect changes in the underlying growth trend! This clearly does not make sense, unless
we want to deny the existence of business fluctuations.
Consider next the opposite extreme where we let/. tend to infinity. In that case the
first sum in (3) does not carry any weight, and HP will then attain its lowest possible value
of zero if we choose t he g,'s to ensure that the estimated trend growth rate is constant
throughout the period of observation, thatis g, + 1 - g, = g 1 - g, _ 1 lor all t = 2. 3 ... . . T - 1.
This would give us the straight lines in Fig. 14.1, but we have already seen that it is
unreasonable to assume that the trend growth rate is a constant.
Clearly. then, il should be positive but fmite. The greater the value of il. the more we
will try to avoid variation over time in the estimated trend growth rate, that is. the
smoother will be our estimated growth trend (the closer it will be to a straight line). On the
other hand. the smaller the value of il. the smaller \<Viii be the deviation between our esti-
mated g 1 and the actual value ofoutputy 1, that is, the greater is that part of the movement
in actual output which we ascribe to changes in the underlying growth trend.
Among business cycle researchers using C JI.Iarterly data, it is customary to set
il = 1600. This is basically a convention. based on a consensus that this value of A pro-
duces a fitted growth trend that a 'reasonable' student of business cycles would draw
through a time plot of quarterly data for (the log of) real GDP. Figure 14.3 shows the fltted
growth trend for the countries included in Fig. 14.1 when the trend is estimated via the
HP iilter using il = 1600. We see that the HP lllter does indeed seem to capture the gradual
changes in the growth trend which are apparent to the naked eye.
Once we have fitted a growth trend using the HP filter, we immediately obtain an esti-
mate of the cyclical component of the (log of) quarterly real GDP by rearranging equation
(2) to get c, = y, - g 1• In Fig. 14.4 we have plotted the resulting estimates of business
cycles, that is, the values of all the c ,'s for our group of countries. The graphs conllm1 what
we have already emphasized: periods of expansion and periods of contraction tend to
alternate, but the business cycles are far from periodic and regular, and even within each
longer cycle there are erratic quarterly fluctuations in activity.
It is tempting to oller a dating of recent business cycles based on Fig. 14.4, lor
although business cycles involve the co-movement of many economic variables (as we
shall see in the next section), GDP is, after aU. the most common single indicator of aggre-
gate economic activity. We have inserted vertical lines in Fig. 14.4 to indicate business
cycles measured from trough to trough. The identilication of business cycle troughs and
peal<.~ is based on the following simple rules of thumb:
1. A trough must be followed by a peak, and a peak must be followed by a trough.

2. The expansion phase as well as the contraction phase must last lor a minimum of two
quarters.
3. A business cycle must span a minimum of five quarters.
© The McGraw-Hill
Companies. 2005
log of GOP United Kingdom log of GOP USA

12.6 9.5
12.4
9
12.2
12 8.5
11.8
8
11.6
7.5
11.4
11.2
<:11 10
;:: ....
.... (') <:11 10 7
0 <0 .... U'>
10 <D
<:11 <:11
(J) co <:11 U'>
<:11
U'> <I> .... <D
co <:11
<:11 <:11 <:11 <:11 <:11 <:11 <:11 0> <:11 <:11 <:11
log of GOP Denmark log of GOP Finland

7.1 10.5
10.4
7
10.3
6.9
10.2
6.8 10.1
10
6.7
9.9
6.6
9.8
'<t ....
6.5 9.7
.... ....
<:11 <:11
0
co
<:11
(')
co
<:11
<D
(J)
<:11
<:11
co
<:11
"
<:11
<:11
U'>
<:11
<:11
co
<:11
<:11
....<:11co a;<:11 '<t
co
<:11
....
co
<:11
0
<:11
<:11
(')
<:11
<:11
<D
<:11
<:11
<:11
<:11
<:11
log of GOP Belgium log of GOP Netherlands

11.1 11.5
11.4
l
11
11.3
10.9
11.2
10.8
11.1
10.7
11
10.6
10.9
10.5 10.8
(')
co
<0 <:11 U'> (J)
0 10 co
<:11
co
0>
co
<:11
"
<:11
<:11
<:11
<:11
<:11
<:11
co
<:11
(')
co
<:11
<D
(J)
<:11
<:11
co
<:11
"
<:11
<:11
<:11
<:11
<:11
<:11
Figure 14.3: Real GDP and its underlying trend

Note: Quarterly data. The growth trend has been estimated by using the Hodrick·Prescott fiUer with A= 1600.
Source: See Table 14.2.
If the tlrst criterion is not met. it does not make sense to speak of a 'cycle'. The second
criterion implies that we require a minimum degree of persistence in the movement of
economic activity before we can speak of a systematic tendency for activity to expand or
contract. The last criterion reflects the convention that tluctuations spanning only a year
or less do not count as business cycles. 5
Sorensen-Whitta- Jacobsen: I Part 5- The Building 14. The economy in the I © The McGraw-Hill
Yt - Yt United Kingdom
0.06
0.05
0.04
0.03 -
0.02
.r.... . .
1\i.A Ii jl / I
0.01 II
0 "" Wt U\ I A .I \ (v..., ,;., If
- 0.01 "" r 'vv\I 1
1 \(1 1 IV \ /1 1\j(' \ I
- 0.02
j v LN \ I
- 0.03
\ ~
- 0.04
1959 1965 1971 1977 1983 1989 1995
Yt - Yt USA
0.05
0.04
0.03
A .,A
~
A I~
0.02
0.01
lln II11 1\ t.A I, ril f'l
0
l Ill
, A, I~ 1"'\ If M lr
I{V I I'J Ul""
- 0.01 l fV 1..
~~
- 0.02
- 0.03
v v
- 0.04
- 0.05
1950 1959 1968 1977 1986 1995
Yt - Yt Denmark
0.04
0.03
M /~J
"
0.02
0.01
j
I VI 11
1\ I 1
A A l.i r \r-.1\ fvly
v
0 \
~
\~tvtiLY_ V-\JJv '\;.' \ "'
~
- 0.01
- 0.02 '
\ \
- 0.03
- 0.04
- 0.05
-t '
- 0.06
1974 1977 1980 1983 1986 1989 1992 1995 1998
Figure 14.4a: Cycles in real GOP in the United Kingdom , United States and Denmark
Source: Own calculations, based on the sources underlying Table 14.2.
5. The computer algorit hm used to dat e the business cycles was developed and kindly provided by Jesper Linaa from
the University of Copenhagen.
CD I Sorensen-Whitta- Jacobsen:
© The McGraw-Hill
Companies. 2005
408 INTRODUCING ADVANCED MACROECONOMICS
Yt-Yt Finland
0.08
0.06
A
!\
( \
0.04
A
rJ \
0.02
t/V~ A (~V
['j_\ NV
0
YY\14
- 0.02
I \
lf'/\J \
- 0.04
\. f
11
- 0.06
1978 1981 1984 1987 1990 1993 1996 1999
y,- y, Belgium
0.03 - , - - - - - - - - . - - - - - - - . - - - - . - - - - - - , - - - - . - - - - - ,
-0.03+-----~-----~--~---~---L--_J
1983 1986 1989 1992 1995 1998
Yt - Yt Netherlands
0.04
0.03
0.02 I A
0.01
-~\A A rl\N\ r-
0
AN tJ \ 1\_f
- 0.01
" NV" JC\~_ \ ~p::-'L\!6/
\
lA I
- 0.02
- 0.03
- 0.04
1980 1983 1986 1989 1992 1995 1998
Figure 14.4b: Cycles in real GDP in Finland, Belgium and the Netherlands
Source: Own calculations, based on the sources underlying Table 14.2.
Sorensen-Whitta- Jacobsen: I Part 5- The Building 14. The economy in the © The McGraw-Hill
We see from Fig. 14.4 that all the countries for which data are available experienced
a serious contraction in the mid-1970s. All countries also fell into recession in the early
1980s and in the early 1990s. In Exercise 1 we test your knowledge of recent economic
history by asking you to mention some factors or 'shocks' wh ich might help explain these
downturn~ in economic activity.
14.5 yy~~.~. ~.~P..P.~.~.~. ~~.~~~&..~.~.~.~~.~~.~y~~~~................................................................

The HP filter is a useful method of separating the growth trend from the cyclical compo-
nent of an economic time series. In this section we will study the statistical properties of
the estimated cyclical components of a number oftime series lor our group of countries. By
doing so we will get more information on w hat happens during business cycles.
Our study is based on seasonally adjusted quarterly data going back as far as such
data are available for our six countries. The data sets include the series lor real GDP which
we have already considered. The components of national income and product are
measured in fixed prices. and all variables have been detrended by means of the HP 111ter.
using the A.-value of 1600 which is customary lor business cycle analysis of quarterly
data. Variables displaying a growing time trend have been transformed into natural
logarithms before detrending. Researchers have found that the HP tilter tends to give
imprecise estimates of the trend at the end-points of the time series. For this reason our
statistical analysis excludes the first and the last 12 estimated cyclical components of each
quarterly time series. in line with common practice.
Volatility
The first question we will ask is: what is the magnitude of the variability in dillerent eco-
nomic variables during a 'typical' business cycle? We may quantify this variability by cal-
culating the standard deviations of the estimated cyclical components of the various time
series. The empirical standard deviation s x of a series of observations of variable :1.~ 1 over the
time interval t = 1, 2, ... . T is defined as:
1
sx= . j--I 1
T- 1 1= 1
T
cx~- :\Y, "' = - '\'
A · -
T
L .."" t'
T 1= 1
(4)
where x is the empirical mean value of the :1.~ /s. The empirical standard deviation mea-
sures the 'average' deviation of x 1 from its mean over the period of observation. It is thus a
natural indicator of the degree of volatility of the economic variable x 1•
In the middle columns of Table 14.2 we use the absolute standard deviation as a
measure of the volatility of the cyclical components of some important macroeconomic
variables. Recall that when variables are measured in logarith ms, the absolute standard
deviation of some variable x =In X roughly indicates the average percentage deviation of X
from its mean. For example, the first figure in the middle column for the United Kingdom
indicates that. on average over the UK business cycle, the cyclical component of real
quarterly GDP deviates about 1. 5 per cent from its mean value.
CD I Sorensen-Whitta-Jacobsen:
© The McGraw-Hill
Companies. 2005
Table 14.2a: Macroeconomic volatility in the United Kingdom and the United States
Average share Absolute standard Relative standard
United Kingdom in GOP(%) deviation (%) deviation 1
GDP 100 1.51 1.00

Private consumption 61 1.68 1.11
Domestic investment 2 15 6.34 4.21
Public consumption 23 1.30 0.86
Exports 18 2.98 1.98
Imports 18 3.78 2.51
Employment 1.15 0.76
1
S tandard deviation relative to standard deviatiln of GOP.
2
1ncludes both private and public investment.
Not e: Based on quarterly data from 19560 1 to 200304. 24 end·point observations excluded.
Sources: UK Office for Naticnal Statistics, Ecowin and Oatastrearn.

United States in GOP(%) deviation (%) deviation 1
GDP 100 1.66 1.00

Private investment 16 7.61 4.58
Public investment 4 6.02 3.62
Exports 7 5.43 3.26
Imports 8 5.15 3.10
1
Standard deviation relative to standard deviatiln of GOP.
Not e: Based on quarterly data from 194 70 1 to 200304. 24 end·point observations excluded.
Sources: Bureau of Economic Analysis, Bureau of labor S tatistics, Federal Reserve Bank of St. Louis.
In the last columns in Table 14.2 we have measured the standard deviations of the
various variables relative to the standard deviation of GDP. A figure in excess of (smaller
than) 1 means that the variable considered tends to be more (less) volatile than GDP. One
striking feature of Table 14.2 is that investment is between three and eight times as
volatile as GDP. The volumes of exports and imports also fluctuate a lot more than GDP.
indicating that the foreign trade sector is relatively unstable. On the other hand. employ-
ment fluctuates considerably less than GDP. Private consumption is sometimes more and
sometimes less volatile than GDP.
The main messages from Table 14.2 may be summarized as three stylized facts
regarding business cycles.
Stylized business Investment is many times more volatile over the business cycle than GDP. It is the most
cycle fact 1 unstable component of aggregate demand.
Stylized business Foreign trade volumes are typically two to three times as volatile as GDP.
cycle fact 2
Table 14.2b: Macroeconomic volatility in Denmark and Finland

Denmark in GOP (o/o) deviation (o/o) deviation 1
GOP 100 1.6 7 1.00

Public consumption 26 1.19 0.7 1
Public investment 2 7.91 4.73
Exports 31 2.65 1.58
Imports 28 4.53 2.7 1
1
Standard deviation relative to standard deviation of GDP.
Note: Based on quarterly data from 197101 to 200302. 24 end·point observations excluded.
Sources: Danmarks Nationalbank and Ecowin.

Finland in GOP(%) deviation (% ) deviation 1
GOP 100 2.29 1.00

Private fixed investment 20 8.49 3.7 1
Public fixed investment 3 5.72 2.50
Exports 27 4.46 1.95
Imports 25 5.97 2.61
1
Standard deviation relative to standard deviation of GDP.
Note: Based on quarterly data from 19750 1 to 200301. 24 end·point observations excluded.
Sources: S tatistics Finland and Bank of Finland.
Stylized bus iness Employment and u11employment are consideralJly Jess vokrtile over the busi11ess cycle tha11
cycle fact 3 GDP.
Correlation, leads and lags
Table 14.2 tells us how much dillerent variables fluctuate relative to the fluctuations in
GDP. But v1re are also interested in studying whether and to what extent the cyclical com-
ponent. x 1, of some economic variable moves in the same direction as or opposite to the
cyclical component of real GDP. c 1• For this purpose we introduce the empirical covarial'lce
between x, and c 1, deHned as:
1 r 1 T
Sxc = - -
T- 1
L (x, - x)(c,- c),
t=l
c=-
T t= l
L:c 1, (5)
where cis, of course. the mean value of the c,·s. The covariance measures the degree to
which x and c move together, but its magnitude '<Viii depend on our choice of the units in
which we measure x and c. To obtain an indicator which is independent of the choice of
© The McGraw-Hill
Companies. 2005
Table 14.2c: Macroeconomic volatility in Belgium and the Netherlands

Belgium in GOP(%) deviation (%) deviation 1
GDP 100 1.06 1.00

Public consu mption 23 1.07 1.01
Public fixed investment 2 7.1 5 6 .72
Exports 69 2.63 2.47
Imports 67 2.69 2.52
Unemployment 0.58 0.54
1
S tandard deviation relative to standard deviatiln of GOP.
Not e: Based on quarterly data from 19800 1 to 200304. 24 end·point observations excluded .
Sources: Belgostat, Banque Nationale de Belg ique and Ecowin.

Netherlands in GOP(%) deviation (% ) deviation 1
GDP 100 1.22 1.00

Private fixed investment 18 4.05 3 .32
Public consu mption 24 0 .96 0 .79
Public fixed investment 3 3.76 3 .08
Exports 50 2.25 1.85
Imports 47 2.67 2.19
Unemployment 0 .66 0 .54
1
Standard deviation relative to standard deviatiln of GOP.
Not e: Based on quarterly data from 19770 1 to 200304. 24 end·point observations excluded.
Sources: Statistics Netherlands, De Nederfandse Bank and Ecowin.
units. it is preferable to normalize the observations of x,- x and c,- c by the respective
standard deviations s, and s,. and to study the covariation of the normalized deviations.
(x 1 - x)/s, and (c 1 - c)/s,. Following this procedure. we obtain the coejficient of correlation
between x and c, which we have already encountered in Chapter 11, and which is defined
as:
(6)
It can be shown that the coell1cient of correlation '>Viii always assume a value some-
where in the interval from - 1 to +1. If p(x .. c,) is equal to 1 we say that x, and c, are per-
fectly positively correlated. and if p (x ,. c ,) equals - 1 we say that the two variables are
perfectly negatively correlated. In both cases there is a strict linear relationship between
the two variables. If p(x 1, c,) is positive but less than 1, x and c will tend to move in the
same direction, 'o\Tith the co-movement being more systematic the smaller the deviation of
p(,.,:,, c1) from 1. On the other hand, a negative value of p(x,, c,) indicates that the two
variables tend to move in opposite directions. Clearly, if'p(x 1, c,) is (close to) 0 , there is no
systematic relationship between x and c.
In the present context where c1 represents the cyclical component of real GDP, we say
that x varies procyclical/y when p(x ,, c1) is substantially greater than 0 . since the positive
correlation indicates that x tends to rise and fall with GDP. By analogy, if p(x" c 1) is
negative, we say that x moves in a cozmtercyclical fashion because x tends to move in the
opposite direction ofGDP.
As already noted, the co-movements of the di!lerent economic variables are not
always synchronized over the business cycle: some variables may reach their turning
point before others do. To investigate whether a variable x moves out of sync with real
GDP, we may measure the coefficient of correlationp(x 1_ , , c1) between c1 and the value of
x observed n periods earlier (x 1 _ , ) , and the correlation coefficient p(x 1+"' c,) between c1
and the value of x observed 11 periods later (x 1 + 11) . U'p(x 1 _ , c,) is significantly dillerent from
0 and numerically greater than p(x" c,), we say that x, is a leading indicator, because a
change in x observed n periods earlier tends to be associated with a change in GDP in the
current period. In other words, movements in x tend to lead movements in aggregate
output, so a turnaround in x indicates a later turnaround in c. Similarly, we say that xis a
lagging variable if p(x ,+,. c 1) is significantly dillerent from 0 and numerically greater than
p(x ,. c,). since this is an indication th at x tends to reach its peaks and troughs later than c.
In Table 14.3 we show coefficients of correlation between the logs of various vari-
ables and the log ofGDP, including leads and lags. The middle columns headed '0' show
the contemporaneous correlation coelllcients, p(x,. c,), where x, is the relevant variable
indicated in the left part of the table. The llrst two columns for each country show correla-
tions between current GDP (c 1) and x 1 _ 2 and x 1_ 1 , respectively, whereas the last two
columns show correlations between y 1 and x 1 + 1 and ,.:1 + 2 , respectively. Recall that the
length of a time period is one quarter. From the second row for the United Kingdom we
therefore infer that the coelllcient of correlation between current GDP and private con-
sumption two quarters earlier is 0.55, whereas the correlation between current GDP and
private consumption one quarter later is 0 .60.
Table 14.3 shows that whereas public consumption seems to be more or less
uncorrelated with GDP. private consumption, private investment and imports are all
procyclical, displaying a clear positive correlation with aggregate output. In particular, we
see that private consumption and investment are strongly correlated with contempo-
raneous GDP.
Not surprisingly, we see th at employment varies procyclically, since an increase in
output requires an increase in labour input. The mirror image is that unemployment is
countercyclical, as shown by the data for Belgium and the Netherlands. We also see that
employment and unemployment are lagging variables, since they are more strongly corre-
lated with the GDP of the previous quarter than '.vith contemporaneous GDP.
While employment displays a strong positive correlation with output, the correlation
between real wages and real GDP is seen to be much weaker and less systematic. Average
labour productivity (output per working hour) tends to be positively correlated with con-
temporaneous GDP, suggesting either that workers tend to work harder when the
demand for output is high, or that productivity shocks are of some importance for the
fluctuations in GDP. This may help to explain why employment is a lagging variable: if
0 I Sorensen-Whitta-Jacobsen: I Part 5- The Building
Introducing Advanced Blocks for the Short-run
© The McGraw-Hill
Companies. 2005
Table 14.3a: Macroeconomic correlations, leads and lags in the United Kingdom and the
United States
Coefficient of correlation between GOP and Xr
United Kingdom Quarterly leads and lags
-2 - 1 0 1 2
Xr (real variables)
GOP 0.61 0.78 1.00 0.78 0.61
Private consumption 0.55 0.66 0.76 0.60 0.51
Gross domestic investment 0.46 0.64 0.78 0.66 0.49
Public consumption -0.23 -0.19 -0.16 -0.22 -0.14
Exports 0.14 0.27 0.45 0.33 0.34
Imports 0.41 0.61 0.69 0.61 0.54
Employment 0.32 0.50 0.66 0.74 0.77
Real wage -0.06 0.01 0.11 0.07 0.07
Labour productivity 0.51 0.52 0.60 0.25 0.00
Xt (Nominal variables)
Inflation rate (CPI) -0.03 0.02 0.27 0.27 0.38
Short-term nominal interest rate -0.30 -0.12 0.05 0.27 0.38
Long-term nominal interest rate -0.33 -0.16 -0.03 0.11 0.24
Note: Based on quarterly data from 195601 to 200304. 24 end-point observations excluded.
Sources: See Table 14.2.
Coefficient of correlation between GOP and Xr

United States Quarterly leads and lags
-2 -1 0 1 2
Xr (real variables)
GOP 0.59 0.84 1.00 0.84 0.59
Private investment 0.57 0.75 0.84 0.65 0.39
Public consumption -0.09 -0.01 0.12 0.21 0.27
Public investment 0.02 0.12 0.23 0.25 0.26
Exports -0.07 0.15 0.36 0.45 0.45
Imports 0.46 0.62 0.68 0.60 0.38
Employment 0.26 0.57 0.81 0.89 0.82
Real wage 0.20 0.23 0.21 0.16 0.09
Labour productivity 0.56 0.56 0.51 0.17 -0.14
Xr (Nominal variables)
Inflation rate (CPI) 0.16 0.29 0.39 0.39 0.35
Short-term nominal interest rate -0.13 0.14 0.35 0.44 0.45
Long-term nominal interest rate -0.29 -0.10 0.05 0.11 0.12
Note: Based on quarterly data from 194701 to 200304. 24 end-point observations excluded.

Sorensen-Whitta- Jacobsen:
© The McGraw-Hill
Companies. 2005
lCD
Table 14.3b: Macroeconomic correlations, leads and lags in Denmark and Finland
Coefficient of correlation between GOP and X 1
Denmark Quarterly leads and lags
-2 -1 0 1 2
X 1 (real variables)
GOP 0.51 0.73 1.00 0.73 0.51
Public consumption -0.01 0.10 0.21 0.25 0.24
Public investment 0.24 0.33 0.38 0.40 0.39
Exports -0.1 3 -0.12 0.07 -0.06 -0.02
Imports 0.51 0.66 0.72 0.51 0.26
Employment 0.35 0.55 0.69 0.72 0.65
Real wage -0.22 -0.18 -0.04 0.10 0.21
Labour productivity 0.06 0.13 0.47 0.1 2 -0.07
X 1 (Nominal variables)
Inflation rate {consumption deflator) -0.1 7 -0.01 0.12 0.08 -0.02
Short-term nominal interest rate -0.37 -0.26 -0.07 0.11 0.28
Long-term nominal interest rate -0.45 -0.36 -0.22 -0.16 -0.16
No te: Based on quarterly data from 197 101 to 200302. 2 4 end·point o bservations excluded.
Coefficient of correlation between GOP and X 1

Finland Quarter!~ leads and lags
-2 -1 0 1 2
X 1 (real variables)
GOP 0.77 0.86 1.00 0.86 0.77
Private gross fixed investment 0.60 0.72 0.88 0.83 0.82
Public consumption 0.07 0.22 0.36 0.49 0.62
Public gross fixed investment 0.06 0.17 0.24 0.25 0.27
Exports 0.38 0.30 0.29 0.13 -0.03
Imports 0.62 0.66 0.73 0.64 0.51
Employment 0.43 0.60 0.75 0.85 0.90
Real wage -0.19 -0.12 -0.09 0.04 0.10
Labour productivity 0.34 0.29 0.29 0.13 0.04
X 1 (Nominal variables)
Inflation rate (Cost-of-Living Index) 0.28 0.33 0.36 0.32 0.27
Short-term nominal interest rate -0.26 -0.09 0.09 0.20 0.30
Long-term nominal interest rate -0.1 8 -0.02 0.11 0.20 0.27
No te: Based on quarterly data from 197401 to 200301. 2 4 end·point observations excluded.

0 I Sorensen-Whitta- Jacobsen: I Part 5- The Building
Introducing Advanced Blocks for the Short- run
© The McGraw-Hill
Companies. 2005
Table 14.3c: Macroeconomic correlations, leads and lags in Belgium and the Netherlands
Coefficient of correlation between GOP and X r
Belgium Quarterly leads and lags
-2 -1 0 1 2
xt (real variables)
GDP 0.61 0.81 1.00 0.81 0.6 1
Private consumption 0.33 0.52 0.68 0.69 0.6 1
Public consumption -0.05 0.02 -0.01 -0.06 -0.04
Public fixed investment 0.07 0.01 -0.08 -0.20 -0.30
Exports 0.53 0.60 0.67 0.65 0.47
Imports 0.46 0.60 0.69 0.66 0.54
Unemployment -0.37 -0.52 -0.65 -0.71 -0.73
Real wage -0.22 -0.36 -0.46 -0.48 -0.43
X r (Nominal variables)
Short term nominal interest rate 0.10 0.26 0.45 0.56 0.57
Long-term nominal interest rate 0.32 0.48 0.61 0.60 0.51
Note: Based on q uarterly d ata from 198001 to 2 00 3 04. 24 end -po int observations excluded.
Coefficient of correlation between GOP and X r

Netherlands Quarterly leads and lags
-2 - 1 0 1 2
xt (real variables)
GDP 0.57 0.71 1.00 0.71 0.57
Domestic investment 0.49 0.50 0.73 0.55 0.37
Public consumption -0.18 -0.14 -0.05 0.02 -0.02
Exports 0.42 0.58 0.55 0.43 0.35
Imports 0.59 0.61 0.59 0.55 0.39
Unemployment -0.42 -0.53 -0.65 -0.72 -0.74
Real wage -0.09 -0.15 -0.21 -0.25 -0.18
X r (Nominal variables)
Short-term nominal interest rate 0.20 0.37 0.59 0.66 0.70
Long-term nominal interest rate 0.10 0.27 0.44 0.45 0.45
Note: Based on quarterly d ata fro m 197 70 1 to 2 00 3 04. 24 end -po int observations exclud ed .

labour productivity rises as soon as output goes up, there is less need for Hrms to add to
their stock of employees right from the start of a business cycle upswing.
While employment is clearly a lagging variable. there are only a few examples of
leading variables in Table 14.3. In Finland and the Netherlands. the volume of exports
seems to be a leading indicator, having a stronger positive correlation with GDP in the
subsequent quarter(s) than with contemporaneous GOP. However. this pattern is not
found in the other countries considered, so in general business tluctuations do not seem to
be initiated by changes in export demand.
All the variables discussed so far have been deHned in real terms. The bottom rows in
Table 14.3 show the correlation between real output and some important nominal vari-
ables. In almost all countries the rate of inflation tends to be positively correlated with
GDP, althoug h the correlation is not particularly strong. As our theoretical analysis in
later chapters will make clear, a positive correlation between output and inflation indi-
cates that business cycles are driven mainly by shocks to aggregate demand. We also see
that in all countries the short-term nominal interest rate tends to go up in the two quar-
ters following a rise in GDP, reflecting that central banks typically tighten monetary policy
in reaction to higher economic activity.
Let us sum up the main lessons from Table 14.3.
Stylized business Private consumption, investment and imports are strongly positively correlated with GDP.
cyclefact4
Stylized business Employment (unemployment) is procyc/icnl (countercyclical) nnd more strongly con·elated
cycle fact 5 with GDP than real wages and labour productivity. Lnbour productivity tends to be procyc/ical.
whereas real wages teud to be very wenkly correlated with GDP.
Stylized business Tn most coulltries inflation teuds to be positively correlated 1vitlz GDP. although tire correlation
cycle fact 6 is not very strong.
Stylized business Employment is a lngging varinble; short-term rwminal iuterest rates also tend to lie lagging
cycle fact 7 variables.
Persistence
Another interesting property of an economic variable ls its degree of persistence. As you

recall. one characteristic of business cycles L~ that, once the economy moves into an
expansion or a contraction, it tends to stay there for a while. Persistence in some variable
x means that the observed value of x in period t, x,. is not independent of the value. x 1 _ ,.
of x in some previous period t - n, where n ~ 1. In other words. if:~.: assumed a high (low)
value in previous period t - H. there is a greater chance that it will also assume a high
(low) value in the current period t. We can measure such persistence in a time series
(x,)[=1 by calculating the coetncient of correlation between x 1 and its own lagged value
x,_,, for 11 = 1, 2, ... This particular correlation coelllcientp(x 1• x 1 _,) is called thecoejjlcient
ofnutocorrelation and was already introduced in Chapter 11 . If p(x,. x,_,) is signiHcantly
above zero for several positive values of n, there is a high degree of persistence: once x has
0 I Sorensen-Whitta- Jacobsen:
© The McGraw- Hill
Companies. 2005
Table 14.4a: Macroeconomic persistence in the United Kingdom and the United States
Coefficient of autocorrelation
1-quarter 2-quarter 3-quarter 4-quarter

United Kingdom lag lag lag lag
Real variables
GDP 0.78 0.61 0.45 0.26
Private consumption 0.78 0.64 0.50 0.31
Gross domestic investment 0.70 0.49 0.30 0.13
Public consumption 0.59 0.45 0.39 0.20
Exports 0.40 0.29 0.06 -0.11
Imports 0.68 0.43 0.22 0.00
Employment 0.94 0.81 0.64 0.47
Real wage 0.74 0.51 0.26 0.07
Labour productivity 0.73 0.53 0.37 0.19
Nominal variables
Inflation rate (CPI) 0.44 0.26 0.01 -0.07
Short term nominal interest rate 0.79 0.59 0.38 0.20
Long-term nominal interest rate 0.78 0.49 0.29 0.11
Not e: Based on quarterly data from 19560 1 to 20030 4. 24 end·point observations excluded.

United States lag lag lag lag
Real variables
GDP 0.84 0.59 0.33 0.1 0
Private investment 0.79 0.55 0.28 0.04
Public investment 0.80 0.61 0.40 0.26
Exports 0.70 0.54 0.30 0.05
Imports 0.7 1 0.45 0.27 0.07
Employment 0.91 0.70 0.44 0.19
Real wage 0.81 0.61 0.42 0.21
Labour productivity 0.7 1 0.45 0.19 -0.01
Nominal variables
Inflation rate (CPI) 0.49 0.27 0.33 0.11
Short-term nominal interest rate 0.80 0.53 0.39 0.22
Not e: Based on quarterly d ata from 194 701 to 20030 4. 24 end·point observations excluded .
Sources: See Table 14. 2.

Table 14.4b: Macroeconomic persistence in Denmark and Finland


Denmark lag lag lag lag
Real variables
GOP 0.73 0.5 1 0.27 0.1 1
Private consumption 0.61 0.47 0.3 1 0.14
Private investment 0.65 0.45 0.23 0.06
Public investment 0.68 0.55 0.38 0.32
Exports 0.45 0.46 0.18 -0.02
Imports 0.70 0.35 -0.01 -0.28
Employment 0.86 0.71 0.55 0.41
Real wage 0.66 0.50 0.36 0.22
Labour productivity 0.23 0.05 -0.05 -0.02
Nominal variables
Inflation rate {consumption deflator) -0.02 -0.10 -0.15 -0.15
Short-term nominal interest rate 0.53 0.26 0.05 -0.25
Note: Based on quarterly data from 1971 0 1 to 200302. 24 end·point observations excluded.
S ources: See Table 14.2.

Finland lag lag lag lag
Real variables
GOP 0.86 0.77 0.67 0.52
Private gross fixed investment 0.82 0.75 0.65 0.53
Public gross fixed investment 0.65 0.27 0.08 0.07
Exports 0.37 0.30 0.18 -0.06
Imports 0.52 0.49 0.34 0.08
Employment 0.95 0.86 0.74 0.58
Real wage 0.35 0.31 0.13 0.12
Labour productivity 0.06 0.01 0.03 0.09
Nominal variables
Inflation rate {Cost-of-Living Index) 0.26 -0.17 0.08 0.52
Long-term nominal interest rate 0.82 0.49 0.18 -0.04
Note: Based on quarterly data from 19750 1 to 20030 1. 24 end-point observations excluded.

© The McGraw-Hill
Companies. 2005
Table 14.4c: Macroeconomic persistence in Belgium and the Netherlands


Belgium lag lag lag lag
Real variables
GDP 0.81 0.61 0.37 0.12
Private investment 0.35 0.34 0.21 -0.05
Public consumption 0.74 0.41 0.17 -0.07
Public fixed investment 0.92 0.73 0.46 0.20
Exports 0.73 0.55 0.36 0.13
Imports 0.60 0.49 0.37 0.19
Unemployment 0.94 0.82 0.65 0.46
Real wage 0.90 0.71 0.45 0.17
Nominal variables
Short term nominal interest rate 0.76 0.49 0.34 0.29
Not e: Based on quarterly data from 198001 t o 200304. 24 end·point observations excluded.

Netherlands lag lag lag lag
Real variables
GDP 0.71 0.57 0.47 0.32
Domestic investment 0.53 0.43 0.12 0.04
Public fixed investment 0.29 -0.13 0.07 0.15
Exports 0.62 0.50 0.20 0.01
Imports 0.75 0.55 0.34 0.13
Unemployment 0.96 0.88 0.74 0.56
Real wage 0.89 0.69 0.44 0.17
Nominal variables
Not e: Based on quarterly d ata from 197 701 to 200304. 24 end·point observations excluded .
been pushed above or below its mean value, it tends to continue to be above or below its
mean lor a long time.
Table 14.4 measures the persistence of business fluctuations by the coefficients of
autocorrelation. The first figure in the first column lor the UK (0. 78) means that if
real GDP goes up by one percentage point in the current quarter, then on average 0 .78
percentage points of that increase \>Viii remain in the next quarter if the economy is not
exposed to new shocks. We see from Table 14.4 that there is considerable persistence in
the movements of GDP and of private consumption . but the most persistent variables are
employment and unemployment. The high persistence of employment may retlect that
fums are reluctant to hire and Hre workers because hiring and firing is costly.
To sum up. we have
Stylized bus iness There is cm1siderable persistence in GDP a11d about the same degree of persistence in private
cycle fact 8 COIISUIIIplion.
Stylized business Employment and unemployment are even more persiste11t than GDP.
cycle fact 9
14.6 Measuring and decomposing the output gap: the production

r~~~.~.~g~ ~r..P.~?..~~?.. . . . .................................................. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
The percentage dill'erence between real GDP and its trend value is usually termed the
output gap. Economic policy makers and business cycle researchers take great interest in
th is variable, since a signiHcant positive output gap means that the economy is in a boom,
whereas a substantial negative output gap indicates that the economy is in recession, or
at least that resources are being underutilized.
As we have seen, the output gap may be measured by detrending the time series for
the log of real GDP by means of the HP filter. However, there is another and more elabo-
rate way of estimating the output gap which allows an interesting decomposition of the
gap. We shall now sketch this method which makes use of the concept of the aggregate
production function known from the theory of economic grm>Vth.
Specitlcally. suppose that real GDP is given by the following Cobb- Douglas
production function:
O<o.<l. (7)
where K, is the aggregate capital stock. L1 is the aggregate number of hours worked. and
B, is the 'total factor productivity' measuring the combined productivity of capital and
labour. By definition. total working hours are given as:
L, = (1 - u)N,H,. (8)
where u 1 is the unemployment rate, N 1 is the total labour force. and H, is the average
number of working hours per person employed. Hence we can specify GDP as:
(9)
Suppose now that Y,, B,. ''~' N, and H1 all tend to fluctuate around some long-run
trend levels denoted by Y, . B,. rr, and N,, respectively. By analogy to (9), we may then write
trend output (also referred to as potential output) as:
Y, = B,K;'[(l - rr,)N,H,] 1 -". (10)

© The McGraw-Hill
Companies. 2005
Note that this specification does not distinguish between the actual capital stock and its
trend level, since we assume for simplicity that the capital stock is always fully utilized.
The output gap may be approximated by y,- [1~' where y,=ln Y, andy, = In Y1•
Taking logs on both sides of (9) and (10). subtracting the resulting expressions from one
another. and using the approximation ln(1 - u) "" - u, we get:
y, - y, "" In B, - ln B, + (1 - a) [(In N,- ln N,) + (In H,- ln H,) - (u, - u)]. (11)
Thus the output gap may be found as the cyclical component of total factor productivity,
In B, - In B,. plus 1 - a times the cyclical component of total labour input (the term
within the square brackets). The latter may in turn be decomposed into the cyclical com-
ponent of the labour Ioree, In N,- In N,. the cyclical component of average working
hours. In H, - In H,, and the amount of cyclical unemployment. u, - u,.
Equation (11) may be used to estimate the output gap provided one has access to data
on real GDP, the labour force. average working hours, unemployment. and the total
capital stock. The cyclical components of labour supply and unemployment may be
estimated by detrending the time series for In N,, In H, and u! by means of the HP fllter. To
estimate the cyclical component of (the log of) total factor productivity. one may proceed
as follows. First. take logs on both sides of ( 7) and rearrange to find:
In B, = ln Y, - a In K, - (1 - <x)ln L,. (12)
where L, may be calculated from (8) (or where data on L1 may be directly available). From
the theory and empirics of economic growth we know that the magnitude 1 - (l should
correspond roughly to the labour income share in total GDP which is close to ~ in most
countries. Hence we may set a = !. Plugging the data for Y,. K, and L, into equation (12),
then, gives an estimate of total factor productivity. 6 In a second step. one may detrend the
estimated time series lor In B, through HP nltering to obtain an estimate of the cyclical
component of total factor productivity. The estimates for In B, - In B,, In N, - In N,.
In H, - In H, and u, - u, may finally be inserted into (11) along with a = ~ to give an
estimated time series for the output gap.
Figure 14.5 shows an estimate and decomposition of the output gap in the United
States. based on the method just described and using annual data. 7 The upper part of
Fig. 14.5 shows the part of the fluctuations in the output gap that can be traced to cyclical
fluctuations in total labour input. It follows from (11) that the vertical distance between
the curve for the output gap and the curve (1 - a)(ln L, - In L,) measures the contribution
of cyclical fluctuations in total factor productivity to the total gap. Note that if there are
variations in the degree of utilization of the capital stock. this will tend to generate cyclical
fluctuations in our measure of total factor productivity (TFP). Similarly, if work intensity
varies over the business cycle, say, because people tend to work harder when there is more
work to do, this will also be captured in the cyclical component ofTFP. Hence the fluctua-
tions in In B, - In B1 do not only reflect instability in the rate of technical progress: they
also capture variations in the intensity \>\lith which factors or production are utilized.
6. If you have already studied Chapter 5, you will note that this way of estimating total factor productivity is closely
related to the method of 'growth accounting' used to identify the sources of economic growth.
7. The value of the parameter ,tin the HP filter has been set to 100, which is common practice in business cycle analysis
of annual data. Since the HP filter gives imprecise estimat es of the trend at the end points, we have excluded the first
and the last three estimates of the cyclic al components at the end points of each time series.
0.06
0.04
0.02
0
- 0.02
- 0.04
(1 - a)(ln L1- ln [ 1)
- 0.06
1963 1968 1973 1978 1983 1988 1993 1998
0.06
0.04
0.02
0
- 0.02
- 0.04
(1- a)(ii1 - u1)
- 0.06
1963 1968 1973 1978 1983 1988 1993 1998
0.06
0.04
0.02
0
- 0.02
- 0.0 4
(1- a)(ln H1- ln R1)
- 0.06
1963 1968 1973 1978 1983 1988 1993 1998
0.06
0.04
0.02
0
- 0.02
- 0.04
(1 - a)(ln N1- ln N,)
- 0.06
1963 1968 1973 1978 1983 1988 1993 1998
Figure 14.5: The US output gap and its components estimated by the production function method
Source: OECO, Economic Outlook Database.
We see from the top diagram in Fig. 14.5 that a large part of the variation in the
output gap stems from the cyclical variation in total labour input. At the same time the
cyclical component ofTFP accounts for a large fraction of the output gap at business cycle
peaks and troughs, presumably reflecting that work intensity and capacity utilization are
unusually low in recessions and unusually high in boom periods.
The other diagrams in Fig. 14.5 show the separate contributions of the cyclical
swings in unemployment, average working hours and the labour force to the swings in
the output gap. Changes in cyclical unemployment account for the greatest part of the
© The McGraw-Hill
Companies. 2005
424 I NTRODU C ING A DVA NCE D M AC ROECO N OM ICS
variation in labour input. but average working hours also tend to fluctuate in a pro-
cyclical manner, suggesting that employees tend to work longer hours when there is more
work to be done. The labour Ioree does not seem to respond very much to the business
cycle, although there is a tendency lor the workforce to vary in a slightly procyclical
manner (while lagging a bit behind the output gap) from the early 1970s and onwards. s
We may sum up this analysis of the output gap as follows:
Stylized business Totaljactor productivity, TFP, varies procyclically and the cyclical component ofTFP accounts
cy cle fact l 0 for a large fraction of the total output gap at business cycle peaks and troughs.
Stylized business Most of the cyclical variation in total labour input stems from fluctuations in cyclical
cycle fact ll unemployment, but average working hours, and to some extent tile total labour.force, also vary
procyclically.
It is of interest to compare the estimate of the output gap based on the production
function method to the estimate obtained by simply HP filtering the time series lor the log
of GDP. This comparison is made in Fig. 14.6. It is reassuring to see that the output gaps
found from the two alternative methods are quite close to each other. Hence HP filtering
of the real GDP series may be a quick and easy way of obtaining a llrst estimate of the
output gap, but if one wants to identil)r the various contributions to the gap, one will have
to use some version oft he production function method, as international organizations like
the OECD, the IMP and the European Commission actually do.
o.o8-.-- - - - - - - - - - - - - - - - - - - .
Production function method
0.06 -1-- -1- - - - - - - - - - - - - - - - - 1
0.04 +----._,_- - - - - - - - - - - - - - - - - 1
0.02
0~~-~-4~-~~~-r---~-~.¥-1
- 0.02 -1--,:-:-::'-..,---1-L---1\- 1 -- - \1
- 0.04 +-- - - - - - - - ' 1 ' - - -__,.,.-1-- - - - - - - - - 1
- 0.06 -1-- - - - - - - - - f - - - ---------1

- 0.08 -1-----.---.----.-----.---.-----.---.-----'
1963 1968 1973 1978 1983 1988 1993 1998
Figure 14.6: A ternative measures of the US output gap: the simple HP filter versus the production
fu nction method
Source: O EC D, Economic Outlook Database.
8. The slightly negative contribution of the workforce to the output gap in the booming 1960s could reflect that many
potential members of the labour force during that period were called up for military service in the Vietnam War.
Sorensen- Whitta- Jacobsen:
© The McGraw-Hill
Companies. 2005
lCD
14 TH E EC ONO MY IN TH E SHOR T RU N 425
HP filtering: a final word of caution
Even more sophisticated versions of the production function method typically rely on the
use of the HP fllter, say, for the purpose of estimating the trend in TFP. Although conve-
nient. the method of HP ftltering is not unproblematic. We have already mentioned that
the HP trend tends to be imprecisely estimated at the end points of the time series. Hence
the seriou~ researcher should disregard some end-point observations, as we have done
above. However, this is unfortunate since one is often particularly interested in estimating
the output gap lor the most recent periods in order to e\·aluate the need for active macro-
economic policy to smooth the business cycle.
Another problem is that there is no objectively correct value of the parameter A which
determines the estimated HP trend. In Fig. 14.7 we have shown two different estimates of
the US output gap. using the HP filter with A= 100 and with A= 1000. respectively. We
see that the difference between the two measures is non-negligible in the 1960s and early
1970s, so the basic arbitrariness in the choice of A. adds another element of uncertainty to
measures of cyclicaliluctuations based on the HP niter.
A further problem is that the HP niter cannot capture stmctural breaks in the trends
of economic time series. For example, if a labour market reform leads to a significant
one-time shift in the level of the natural unemployment rate. this change in structural
unemployment will only be slowly and gradually picked up by the estimated HP trend in
unemployment.
Thus there is considerable uncertainty about the 'true' output gap, reflecting our
imperfect knowledge of the way the economy works. As we shall discuss at length in
Chapter 22, the uncertainty regarding the size (and sometimes even the sign) of the
output gap creates difficulties when policy makers try to reduce the short-mn fluctuations
in output and employment through active llscal and monetary policy.
o.o6 ,-- - - - - - - - - - - - - - - - - - - - - ,
A= 1000
- 0.06 - 1 - - - - - - - - - - -r- - - - - - - - - - - 1
- 0.08 -t-----,,.------.---.----..---.,----.--- -.--'

1963 1968 1973 1978 1983 1988 1993 1998
Figure 14.7: The US output gap estimated by the HP filter: the importance of t he c hoice of A
Source: OECD, Economic Outlook Database.
© The McGraw-Hill
Companies. 2005
426 INTRODUC ING A DVANCE D M ACROECONOM ICS
14.7 A look ahead

We have now described some facts about business cycles. In the rest of this book our main
goal will be to constmct an economic model which may explain short-mn fluctuations in
aggregate output, employment and inflation. SpeciHcally, we will gradually build up a
model. which may be summarized in the manner depicted in Fig. 14.8. where real output
and the rate or inflation are determined by the intersection of an upward-sloping aggre-
gate supply curve and a downward-sloping aggregate demand curve. We will then use
this model to study how the economy reacts over time to various sh ocks to the aggregate
supply curve and the aggregate demand curve.
Our goal is to construct a model in which the efl'ects of aggregate demand and supply
shocks tend to persist over time through so-called propagation mechanisrns arising from the
links between central macroeconomic variables. In this view of the world business cycles
are initiated by random shocks to the economy such as an unanticipated change in the oil
price or a change in business confidence due, say, to unexpected political events.
However. the cumulative a11d systematic character of business fluctuations documented in
this chapter will be explained endogenously by the properties of our macroeconomic
model. We hasten to add that we caru1ot promise to explain all the features of business
cycles- economists are still far from having a perfect understanding of this complex phe-
nomenon -but our model economy will at least have the property that random shocks
tend to generate irregular cycles displaying a certain persistence.
As indicated in Fig. 14.8. the aggregate demand curve and the aggregate supply
curve are the central building blocks of our model. To construct the aggregate demand
curve we must build a theory of private investment and private consumption. This is our
agenda for Chapters 15 and 16. In Chapter 1 7 we combine the insights from Chapters 15
and 16 with a study of monetary policy to derive the aggregate demand curve. Chapter 18
then analyses the relation between inflation and unemployment as a basis for construct-
ing the aggregate supply curve. This enables us to set up our basic AS-AD model in
Chapter 19 where we use the model to reproduce some of the stylized facts about business
cycles. The subsequent Chapters 20 to 22 extend the basic model in order to analyse the
problems of macroeconomic stabilization policy. In the last four chapters of the book we
Inflation rate
AS
AD
' - - - - - - - - - - - - - - - - - Output
Figure 14.8: A~gregate supply and aggregate demand

extend the AS-AD model to the open economy. studying regimes with llxed as well as
flexible exchange rates. and end with a discussion of the choice of exchange rate regime.
14.8 §. ~~.~~Y.. ..................................................................................................................................................

1. Business cycles are periods of expansion of aggregate economic activity followed by periods
of contraction in which activity declines. These fluctuations are recurrent but not periodic; in
duration business cycles have varied from a little more than a year to 10 - 12 years. The
severity of recessions has also varied considerably, with recessions sometimes turning into
depressions where output and employment fall dramatically.
2. The expansion phase of the business cycle usually lasts considerably longer than the con·
traction phase, reflecting the economy's capacity for long-term growth. In the period since the
Second World War business cycle expansions have tended to last longer, and recessions
have tended to be shorter and milder than was the case before the war. For the post·Second
World War period the average duration of the US business cycle expansions has been
57 months, while recessions have on average lasted 10 months. By contrast, the average
recession lasted about 22 months before the war.
3. The business cycle fluctuation in a macroeconomic time series may be measured as the
deviation of the actual time series from its long·term trer d. The trend may be estimated by
means of the HP filter which allows for smooth changes in the underlying (growth) trend in a
series.
4. The volatility of the cyclical component in a macroeconomic time series may be measured by
its standard deviation. By this measure, investment is a lot more volatile over the business
cycle than GOP, whereas employment is considerably less volatile.
5. The co-movements in different economic variables over the business cycle may be measured
by their coefficients of correlation with GOP. Private investment, consumption and imports all
have a strong positive correlation with GOP. Employment also displays a clear positive corre·
lation with GOP, but it is a lagging variable, being more strongly correlated with the GOP of
the two previous quarters than with current GOP. Labour productivity tends to vary positively
with GOP, and so does inflation in most countries, although the latter correlation is not very
strong. The nominal short·term interest rate tends to go up in the two quarters following a rise
in GOP, reflecting a tightening of monetary policy.
6. The degree of persistence in a macroeconomic variable may be measured by its coefficients

of correlation with its own lagged values, the so-called coefficients of autocorrelation. There
is considerable persistence in GOP and private consumption, and even more persistence in
employment. This means that once these variables start to move in one direction, they will
continue to move in the same direction for a while, unless they are disturbed by new
significant shocks.
7. The output gap is the percentage difference between actual GOP and trend GOP. The output
gap may be estimated by means of the production function method which allows a decom·
position of the gap into contributions from cyclical variations in unemployment, average
© The McGraw-Hill
Companies. 2005
acroeconomics Model business cycles
working hours, the total labour force, and total factor productivity (TFP). Such a decomposi-
tion shows that fluctuations in TFP - capturing cyclical swings in work intensity and capacity
utilization as well as an uneven pace of technical progress - account for a relatively large
share of the fluctuations in output at business cycle peaks and troughs. The largest part of the
cyclical variation in labour input comes from fluctuations in cyclical unemployment, but
average working hours (and to a minor extent the labour force) also tend to vary positively with
the output gap, reflecting that labour supply tends to increase when there is more work to do.
8. The method of detrending an economic time series by means of the HP filter should be used
with care, because (i) the HP filter gives imprecise estimates of the trend at the end-points of
the time series; (ii) the filter relies on an arbitrary choice of the ,1, -parameter which determines
the smoothness of the estimated trend, and (iii) the HP filter cannot capture structural breaks
in the data series. For these and other reasons there is considerable uncertainty associated
with the measurement of business cycles.
14.9 Exercises
Exercise 1. Accounting for recent recessions
1. All of the countries included in Fig. 14.4 experienced serious recessions in the mid-1 970s,
in the early 1980s and in the early 1990s. Try to give a brief account of the factors and exo-
genous 'shocks' which are likely to have generated these contractions in economic activity.
2. According to Table 14.1 the US economy fell into recession in 2001. The same thing
happened in most other countries in the world. Mention some factors and 'shocks' which in
your view contributed to this recent economic downturn.
Exercise 2. Explaining the turning points in the US business cycle
In Table 14.1 we summarized the NBER datings of US business cycles. In an NBER Working
Paper (no. 6692, August 1998) entitled 'The Causes of American Business Cycles: An Essay
in Economic Historiography', economic historian Peter Temin attempts to explain the turning
points in the chronology of US business cycles by identifying various demand and supply
shocks to the US economy. Using Peter Temin's NBER Working Paper and your knowledge
of the basic AS-AD model from introductory macro, explain some of the most spectacular
turning points of the US business cycle in the twentieth century.
Exercise 3. Measuring the output gap by means of the HP filter
At the internet address www.econ.ku.dk/ pbs/ courses/mode/s&data.htm you may get access
to a computer facility performing HP filtering of economic time series (you will also find a brief
guide on how to use this facility).
1. Find annual data for real GOP for your country, going back in time as far as consistent data
are available (you may want to use the Penn World Table or some other easily accessible
source). Transform the time series into natural logarithms and use the HP filter to estimate a
series for the trend in the log of output and for the output gap, In Y,- In Y,, setting A,= 1 00
(because of the imprecise end-point estimates produced by the HP filter, you should discard
your estimates for the first and last three years in your estimated series). Perform the same
exercise with?. = 1000. Are there significant differences in the estimated output gaps with the
two alternative values of A?
2. Use your estimated series for the output gap to offer a rough dating of recent business cycles
in your country. Are the major peaks and troughs in your estimated output gap in line with your
prior expectations, given your knowledge of the recent economic history of your country?
Exercise 4. Measuring and decomposing the output gap
In Section 5 of this chapter we explained how one may estimate and decompose the output
gap by means of the production function method. In this exercise we ask you to use this
method to estimate a time series for the output gap and its components using annual data for
your own country (or for some other country of your choice for which data are available).
The production function method described in Section 5 requires data on real GOP (Y),
the total capital stock (K) , the total labour force (N), the average number of (annual) working
hours per person employed (H), and the unemployment rate (u) . One practical problem is that
the available data on (some of) the variables K, N, H or u, do not always cover the entire
economy, but sometimes only, say, the private business sector. However, under certain
assumptions the data K•, N•, H • and u• covering only part of the economy may be used as
substitutes for the missing economy· wide variables K, N, Hand u without generating a bias
in the estimate of the output gap and its components. First, define the economy-wide rate o f
employment as E"' 1 - u and the corresponding rate of employment recorded in the statistics
as E• e 1 - u•. Then suppose that:
(13)
where c K, eN, cH and cE are all constant over time. If Cx= 1 we have the fortunate situation
where the data for variable X covers the entire economy, but if Cx< 1 the available data for
variable X only covers a subsector of the macroeconomy. W ith X denoting the trend value of
variable X, it follows from (13) that:
(14)
Furthermore, define the following indicators of labour input measured in hours:
L,e E,N,H,, r,"' E,l'iJ,R, (15)
1. The cyclical component of the macroeconomic variable X! is defined as In X,- In X,. Use Eqs
(13) - (15) to show that the cyclical components of L ~, N~, H~ and E~ are unbiased estimates
of the cyclical components of the corresponding economy-wide variables L,, N, H, and E,.
Using the approximation ln( 1 - x,) "' -x, (which is valid for 'small' values of x,), show that the
magnitude D~ - u~ is a good approximation of In E, - In E,.
Given the assumption that the capital stock is always fully utilized (so that there is no
need to distinguish between K, and 'R,), it follows from Eq. (12) in the main text that:
In B ,=ln Y, - a InK, - (1- a)ln L,,
In B, =ln Y', - a InK, - (1- a)ln [, , (16)
where 8 1 and B, are the actual and the trend level of total factor productivity (TFP),
© The McGraw-Hill
Companies. 2005
acroeconomics Model business cycles
respectively. By analogy, if TFP is estimated on the basis of the available data, we have:
In 87= In Y1 - a In K~ -(1- a)ln L: ,

In B7 =In Y,- a InK~- (1 - a)ln r:. (17)
2. Show by means of (13) - (1 7) that the magnitude In a:- In B~ is an unbiased estimate of the
'true' cyclical component of TFP, In 8 1 -In B, .
3. Try to find annual data on Y, K, N, Hand u for your country (as a default, you may also use the
data for Denmark available at www.econ.ku.dk/ pbsjcoursesjmode/s&data.htm) and apply the
production function method described in Section 5 to these data to estimate the output gap
and its components. If the available data for one or several of the variables K, N, H or u do not
cover the entire economy, you may assume that assumptions (13) and (14) are satisfied (with
cK, eN, cH, and cE all being constant over time) and use the available data K•, N•, H" or u•,
since you have shown in questions 1 and 2 that this will yield unbiased estimates of the true
gaps. You may get access to a computer facility performing HP filtering at the internet
address: www.econ.ku.dk/pbsjcoursesjmode/s&data.htm. To ensure comparability with
Fig. 14.5, set A,= 100.
Construct d iagrams showing the evolution of the output gap and its components (because
of the imprecise end-point estimates produced by the HP filter, you should d iscard your esti-
mates for the first and last three years in each of you r estimated series). Investigate whether
the Stylized business cycle facts 10 and 1 1 derived from US data (see Section 5) also hold
for your country.
If your data for K, N, H or u do not cover the who le economy, discuss briefly whether there
is reason to believe that any of the parameters cK, eN, cH and cE may not be constant over
time. If they are not, is this necessarily a problem?
4. Estimate a time series for the output gap for your country by running a HP filter through the
time series for the natural log of real GOP with }.= 100 (you may also reuse your HP estimate
from Exerc ise 3, assuming it is based on the same time series for real GOP as the one used
in this exercise). Draw your estimate in a diagram and compare it to your estimate of the
output gap based on the production function method. Are the two estimates as c lose as was
the case fo r the US in Fig. 14.6?
I Sorensen- Whitta- Jacobsen: I Part 5- The Building I 15. Investment and asset I © The McGraw-Hill
Introducing Advanced Blocks for the Short- run prices Companies. 2005
•
•
•
•
•
•
Part •
•
h\he Building
•
~ Blocl<:s for the

•
Short-run Model
15 .~J?:Y~~.~~~~~..~J?:.~ . ~~·~·~·~.P~.i.~.~.~...................................................................................... .
16 .g9..~.~.~!l?:P~~9..~.~J~~9..~~..~.~.~..~~~.~.~~....................................................................
17 .~?..~.~.~~EY..P..?..~~~Y...~~~..~.~~E.~.~~.~~..~~~~.~.~.................................................. .
18 .~J?:P.~.~i.?.~.~..~.~~.I?.P.~?..Y~~J?:!.~.~~. ~~~!.~.~~!~. ~~PP.~Y.. ............................ .
431
Part 5- The Building
15. Investment and asset
prices
© The McGraw-Hill
Companies. 2005
Macroeconomics Model
Sorensen-Whitta-Jacobsen: I Part 5- The Building 15. Investment and asset © The McGraw-Hill
Chapter
In vestment and asset

•
prices
T
he previous chapter showed that private investment is often the most volatile
component of aggregate demand and that it is highly correlated with total output.
Understanding the forces driving private investment is therefore crucial lor under-
standing business cycles. In this chapter we present a theory of business investment as
well as a theory of h ousing investment. This will give us an opportunity to study two of the
most important asset markets in the economy: the stock market and the market lor
owner-occupied housing. As we shall see, there is a systematic link between stock prices
and business investment, and a similar systematic impact of housing prices on housing
investment. To w1derstand investment. we must therefore study how asset prices are
formed.
A glance at Fig. 15.1 should make clear why we are interested in asset prices. The
figure is constructed from data for the 16 most important industrial corm tries and shows
the link between the evolution of housing prices and stock prices and the evolution of the
output gap. There is a close relationship betwem asset price.fluctuations and output .fluctuations
and a clear tendency .for stock price movements to lead movements in output. Thus the llgure
suggests that an increase in stock prices or in housing prices will trigger an increase in
economic activity, whereas a signillcant drop in asset prices may be a signal of a future
economic downturn. As we will show in this chapter and the next one, Fig. 15.1 reflects
th at higher asset prices tend to stim ulate private consumption and investment. In par-
ticular, the present chapter will explain why higher stock prices tend to be followed by
higher business investment, and why higher housing prices provide a boost to housing
investment.
The basic idea underlying our theory of investment may be most easily illustrated by
looking at the housing market. At any point in time there is a certain market price lor
houses of a given size and quality. This price may well exceed the cost of constructing a
new house of similar size and quality (the replacement cost). The more the market price
exceeds the replacement cost. the more prolltable it will be for construction firms to build
and sell new houses. Hence we will observe a higher level of housing investment the
greater the discrepancy between the market price and the replacement cost of housing.
Note that the market price can deviate from the replacement cost for a long time. since it
takes time for new construction to produce a significant increase in the existing housing
433
prices
© The McGraw-Hill
Companies. 2005
434 PART 5: T HE BUI LDING BLOCKS FOR THE SHORT-RUN MODEL
3 20
2 15
10
0 5
,g
0
.g.
-1 0
-2 -5
-3 - 10
-4 - 15
1980 1982 1984 1986 1988 1990 1992 1994 1996 1998
Year
3 Equity price c hange

2 25
(right scale}
2 20
15
.g. 0 10 .g.
-1 5
-2 Output gap 0
(left scale}
-3 -5
-4 - 10
1980 1982 1984 1986 1988 1990 1992 1994 1996 1998
Year
Figure 15.1: Output gap, real property prices, and real equity prices in industrial countries 1
1
Arithmetic average of the respective variables in 16 industrial countries, excluding Portugal prior to 1989 owing
to lack of data
2
Three·year movirg average.
Source: IMF, World Economic Outlook, May 2000.
stock, and since it is time-consuming to shift economic resources into the construction
industry if construction activity becomes more prolltable due to a rise in the market price
of housing.
For business investment a similar basic principle applies. The market price of the
business assets owned by a corporation is rellected in the market price of the shares in the
llrm. The replacement cost of the firm's assets is given by the price at which it can acquire
machinery, etc., from its suppliers of capital goods. If the stock market value of the llrm's
assets is higher than their replacement cost, the llrm can increase the wealth of its share-
holders by purchasing additional capital goods. that is. by investing. The higher the stock
price relative to the replacement cost, the greater is the incentive to invest, so the higher
the level of investment will be. One might think that the firm would instantaneously
adjust its capital stock so that any discrepancy between the stock market value and the
replacement cost of its assets is immediately eliminated. However. this is not realistic.
prices
© The McGraw-Hill
Companies. 2005
lCD
15 INVESTMENT AND ASSET PRICES 435
since in practice the firm will incur various costs of adjusting the capital stock, and these
costs are likely to increase more than proportionally to the level of investment. Hence it
will be more profitable to allow a gradual adjustment of the capital stock. and during this
(potentially long) adjustment period the stock market \'alue of the firm will deviate from
the replacement cost of its assets.
In the following sections we will explain this theory of the link between asset prices
and investment in more detail. starting with the theory of the stock market and business
investment.
15.1 The stock market

A few facts about the stock market
It is well known that stock prices are highly volatile and sometimes experience dramatic
swings. For example, on19 October 198 7 the US Dow Jones index fell by 22.6 per centin
a single day. This was even more apocalyptic than the notorious crash in the 'Black
October· of 19 29 when the Wall Street stock market dropped by 23 per cent ln the course
of two days (however, the recovery of stock prices after October 1987 was much faster, so
the macroeconomic effects of the crash of 19 2 9 were much more serious) .
Figure 15.2 illustrates the long-term trends in the US stock market. documenting the
evolution of real (inflation-adjusted) stock prices and the so-called price- earnings ratio,
defined as the market value of shares relative to the profits of the companies w hich have
issued the shares. The curve for the real stock price index highlights the enormous stock
market boom of the 1990s which was followed by a sh arp downturn after the turn of the
new century. However. the graph for the price- earnings ratio shows that even after this
2500 45
40
2000
35
Price-earnings ratio
~ 1500 30
0
·~
25 ·~
a:
LLI
""
~ 1000 20
(ij
& 15
500
10
Figure 15.2: Real stock price index and the price·earnings ratio in the United States
Source: Calculated by Robert S chiller on t he basis of Standard and Poor's Composite Stock Price Index, the US
Consumer Frice Index, and a 1O·year moving average of the earnings of the companies included in the S&P stock
index.
prices
© The McGraw-Hill
Companies. 2005
436 PART 5: T HE B UILDING BLO C KS FO R THE S HO RT-RUN MO DEL
dramatic adjustment of share prices. the price of stock relative to the earnings capacity of
companies remains very high by historical standards.
Mainly as a consequence of rising stock prices, the market value of outstanding
shares (the 'stock market capitalization') as a percentage ofGDP rose sharply during the
1990s. as shown in Fig. 1 5.3. In this way stocks became a much more important
component of total financial wealth.
In many countries the booming stock market motivated a growing proportion of
households to invest in shares, and hall· the adult US population now owns stocks, as
indicated in Fig. 15.4. In a country like Sweden. the corresponding proportion is about
one third, whereas only about 15 per cent of the adult population in Denmark hold
shares.
The stock market is actually more important lor househ olds than Fig. 1 5.4 might
suggest, since the numbers in the figure only include households who are direct owners of
stocks. Households also channel a large part of their savings through pension funds, lile
insurance companies and other financial intermediaries which, in turn. invest a substan-
tial part of their funds in shares. Hence the performance of the stock market directly or
indirectly determines the return to a large fraction of household savings. In this way stock
market developments may determine when people feel they can afford to retire from the
labour market or when they can afford to buy new consumer durables. Moreover. the
evolution of stock prices may have an important impact on the level of output and
employment, because of its in 11uence on consumption and business investment. In short.
the stock market is important lor individual consumers and for the macro economy, so it
is worthwhile investing some ellort in understanding how it works, and how it allects
investment decisions.
160
140
Jl
120
r
100
.g. 80 I
60
40
20
~
0
(') It) ['- <»
['- ['- ['- ['-
<» <» <» <»
Figure 15.3: Stock market capitalization as a percentage of GOP

Source: Datastream.
Introducing Advanced Blocks for the Short-run prices Companies. 2005
United States UK Germany France Japan
Figure 15.4: Percentage of adult population owning stocks

Source: Hall Edison and Torsten Sl0k, 'Wealt h Effects and t he New Economy', IMF Working Paper, WP/01/77,
2001.
15.2 T..~.~. P..r.~~~. ?..f..S..t.?.~~~ ............................................................................................................................ .

The value of a firm and the fundamental stock price
The starting point for our analysis is the assumption that business investment is guided by
a desire to maximize the wealth of the owners of the jirrn. In modern Western economies
where the bulk of business activity is carried out by firms organized as joint stock compa-
nies (corporations). maximization of the wealth of the firm's m>\T!lers is equivalent to max-
imization of the market value of the outstanding shares in the corporation. This is the
reason for our focus on the stock market. However. our theory of investment will also be
relevant for unincorporated firms or for corporations which are not quoted in an olllcial
stock exchange. As we shall see, our theory of the stock market implies that the market
value of shares equals the discounted value of the expected jiJture cash }low from the .firm to its
owners. But this is exactly how a rational outside investor would also value an unquoted
or unincorporated Hrm if he were contemplating buying or investing in such a Hrm. If the
owner of an unincorporated firm wants to maximize the market value of his business
assets, his investment behaviour will therefore be similar to the investment behaviour of a
corporation whose shares are traded in a public stock exchange. as we shall explain in
more detail later.
You may wonder why we assume that the objective of the firm is to maximize the
wealth of its owners? The answer is that maximization of the current market value of the
Hrm will also maximize the consumption possibilities of its owners. This will become clear in
the next chapter where we show that a person·s potential present and future consumption
is constrained by the sum of his financial wealth and his current and discounted future
labour income. Therefore. if a firm can change its operations so as to increase the market
value of it~ assets, it will increase the financial wealth of its owners and enable them to
CD I Sorensen- Whitta- Jacobsen:
prices
© The McGraw-Hill
Companies. 2005
438 PART 5: T HE B UILDING BLO C KS FO R THE SH O RT-RUN MO DEL
increase their consumption either now or in the future. In both cases the owners will
obviously be better ofl'. In your basic microeconomics course you may have learned that
firms maximize their profits rather than the market value of their assets. Fortunately. there
is no contradiction between these two goals. A firm that maximizes its discounted stream of
profits over time will also maximize its market value, as we shall demonstrate below.
If a corporation plans its investment with the purpose of maximizing the market
value of its shares. we must base our theory of investment on a theory of the value of the
firm . Our starting point for such a theory is an arbitrnge condition which says that the
market value of the shares in the firm must adjust to ensure that the holding of shares is
equally attractive as the holding of bonds. Suppose that, at the beginning of period t, the
shareholders in the firm expect to receive a dividend D~ at the end of the period. and that
they expect the market value of their shares at thestartofperiod t + 1 to be v~+ l ' IfV, is the
actual market value of shares in the firm at the beginning of period t. shareholders thus
expect to earn a capital gain equal to V~+ 1 - V, during period t. Hence the total expected
return on shareholding is D~ + (V~+ L- V,). composed of the expected dividend plus the
expected capital gain.
In a capital market equilibrium this expected return must equal the 'required' return
on shares. The required return consists of the opportunity cost of holding shares rather
than bonds, plus an appropriate risk premium. If r is the market rate of interest on bonds
(assumed for simplicity to be constant over time). the opportunity cost of shareholding is
rV,. since this is the interest income which the shareholder could have earned during
period t if he had sold his shares at the inital market value V, and invested the corre-
sponding amount in bonds. Furthermore. since stock prices and dividends are generally
more volatile than bond prices and interest payments, shares are a riskier investment than
bonds. Because investors are risk averse, the expected rate of return on shares must
therefore include a risk premium ton top of the market interest rate to ensure that share-
holding is considered just as attractive as the holding of bonds. Hence the total required
return on the shares is (r + t )V,, and the arbitrage condition for capital market equilibrium
may then be written as:
required expected
~ ~
(r +t)V, = ...___
D~ + V~+ 1 - V,.
___.. (1)
total expected
return on shares
If the current market value V 1 is so high that the required return on the left-hand sideof (1)
exceeds the expected return on the right-hand side, financial investors will sell o!I' their
shares in the firm in order to buy bonds. and the market value V 1 will drop. On the other
hand. if the current share price (and hence V,) is so low that shares in the firm promise a
total rate of return in excess of r + t , investors will shift from bonds to shares, thereby
driving up the current market value V, . Hence the stock market can only be in equilibrium
when the arbitrage condition (1) is met. Since investors derive utility from their real
consumption possibilities, we are measuring all variables in equation (1) in real (inflation-
adjusted) terms. so r is the real rate of interest.
We may rearrange (1) to get:
(2)
15 INVE STMENT A ND ASS ET PRICES 439
This is a very important relationship in our analysis below. It says that the value of the
fum at the beginning of any period equals the present value of that period's expected divi-
dend plus the expected market value at the end of the period. We see that the rate at which
future values are discounted includes the market interest rate r and the required risk
premium<..
As we have argued above, the llrm will choose an investment plan which maximizes
V1• To characterize the firm's optimal investment policy we must therefore study how V~+ l
and D~ and hence V1 depend on the firm's planned investment. This is the purpose of the
following analysis.
Since arbitrage conditions similar to ( 2) must hold for all subsequent periods, rational
financial investors will expect that future stock prices will satisfy the relationships:
etc. (3)
By successive substitutions of the expressions in (3) into (2). we llnd that
(4)
It is reasonable to assume that investors do not expect future real stock prices V~+r, to rise
indefinitely at a rate faster than r + t, lor if the opposite were the case. the current stock
price V1 would become infinitely high according to (4). Hence we assume that
v~+,
lim --"'"-!!..- 0. (5)
,_,~ (1 + r+ t)"
Uwe continue the successive substitutions indicated in (4) and use our assumption ( 5). we
end up with:
D' D~+ I D;·+2

v= I + + + .. . (6)
1
]+r+t (l + r+t) 2 (l+r+t) 3
Equation !6) is our llrst important result, stating that the market value of the shares in a
llrm equals the present discounted value of the expected future dividends paid out by the
flrm. This is sometimes referred to as tbe.fundamentnl share price, because it is a price based
on a 'fundamental' condition. namely the llrm's ability to generate future cash flows to its
ml\l'ners.
Note that there must be a close correlation between a firm 's dividends and its proilts,
since the latter are the source of the former. This observation is the basis for our earlier
claim that maximization of (the present value of) prollts is roughly equivalent to maxi-
mization of market value.
prices
© The McGraw-Hill
Companies. 2005
440 PART 5: T HE B UILDING BLO C KS FO R THE S HO RT-RUN MO DEL
Why are stock prices so volatile?
As illustrated in Fig. 15.2, stock prices tluctuate quite a lot. Equation (6) suggests three
possible explanations for the observed volatility of stock prices:
1. Fluctuations in (the growth rate of) expected future real dividends

2. Fluctuations in the real interest rater
3. Fluctuations in the required risk premium on shares. t: .
The great stockmarket boom of the 1990s seems to have been driven mainly by more
optimistic expectations regarding future real dividends, as 11nancial investors came to
believe that the rapid innovations in information technology would create a 'New
Economy' characterized by a signillcantly higher real growth rate in output and business
prollts. Perhaps changes in the required risk premium on shares have also contributed to
the recent turbulence on the stock market. There is some evidence th at investors fre-
quently change their attitude towards risk. This is illustrated by Fig. 15.5 which plots
recent values of a 'risk appetite index' for the major industrial and emerging market
economies, constmcted by the International Monetary Fund.
The idea behind the index is that if investors become willing to bear more risks. they
will bid up the price of assets th at have been risky in the past. and if they become more
risk averse they will drive down the price of risky assets by selling them. Hence Fig. 15.5
assumes that movements in current asset prices which are systematically correlated
with the observed past riskiness (the past volatility of returns) of those assets indicate a
ch ange in investor appetite lor risk. The 11gure suggests that the terrorist attacks on
11 September 2001 caused a sharp temporary drop in investor appetite for risk.
Following a recovery, risk appetite again began to weaken towards the end of 2001,
perhaps as a reaction to the worsening news about the state of the world economy
announced at th at time.
0.8
0.6
0.4
0.2
0 +--+~~r--+~~--------~----~U-------~~
- 0.2
- 0.4
- 0.6
- 0.8 +----------.--------,--------.---------,---------,
15·May·01 25·Jun·0 1 5·Aug-01 15·Sep-0 1 26·0ct·01 6·Dec·0 1
Date
Figure 15.5: Monthly risk appetite index

Source: IMF, World Economic Outlook, December 2001.
15 INVESTMENT A ND ASS ET PRICES 441
It is often asked whether the observed movements of the stock market are really
consistent with rational investor behaviour. In the short run the stock market sometimes
seems to overreact to news, showing signs of 'herd behaviour' . One may also wonder how
the large longer-term stock market fluctuations observed during the last decade can be
reconciled with realistic changes in the long-term earnings potential of firms . 1 Notice.
however, that the theory presented above is compatible with the observation of frequent
changes in stock prices, if investors frequently revise their forecasts of future dividends
and their appetite for risk, and if they are often faced with unanticipated changes in real
interest rates. Notice also that our theory only assumes a weak form of rationality: all we
assume is that investors require the holding of shares to be just as attractive as the holding
of interest-bearing assets. We have not excluded the possibility that financial investors
may at times hold unduly optimistic or pessimistic expectations about future dividends,
and that they may sometimes require 'unreasonably' high or low risk premia due to an
inability to make reali~tic forecasts of the riskiness of business investment. In short, Eq. ( 6)
makes no assumptions regarding the formation of expectations and risk premia; it only
assumes that the expected return on shares is systematically related to the return obtain-
able on bonds. According to Fig. 15.6 this does not appear to be a bad assumption lor the
longer run. since the realized returns on bonds and stocks in Denmark do indeed seem to
move together when the rates of return are calculated over a five-year or a ten-year
holding period.
Although Fig. 15.6 suggests that the price of stocks is in fact linked to fundamentals
in the long run, many observers of the stock market believe that stock prices can some-
times deviate from the fundamental value of firms . During such periods of 'speculative
bubbles' . stocks become objects of pure short-term speculation. and their prices cease to be
pinned down by the discounted value of expected future dividends. We do not deny that
speculative bubbles may sometimes occur, but in this chapter we will assume that stock
prices reflect fundamental values and show th at the theory of the stock market outlined
above can take us a long way towards understanding business investment.
Business investment
Stock prices and investment
Our working hypothesis is that the fll'm chooses its level of investment with the purpose of
maximizing its market value V,. From Eq. (2) we see that maximization ofV, is equivalent
to maximizing the sum of its owners' expected dividends and expected end-of-period
wealth. o; + V~+ 1• since the individual firm has no influence on rand r.. The question is:
what level of investment will maximizeD~+ V~+ 1 ?
To answer this. let us introduce the variable q to indicate the ratio between the market
value a11d the replacement value of the firm's capital stock. By definition, we thus have
=
V, q 1 K" where K, is the real capital stock, and where the acquisition price of a unit of
1. For a fascinating account of the less rational aspects of stock market behaviour, placing the bull market of the 1990s
in hist orical perspective, see Robert J. Shiller, Irrational Exuberance, New York, B roadway Books, 200 1 (paperback
edition).
prices
© The McGraw-Hill
Companies. 2005
0.4
5-year holding period
0.3
eli
ci 0.2
c:
:;
Q;
0::: 0.1
- 0.1
1925 1935 1945 1955 1965 1975 1985 1995
Year
0.3
10 -year holding period
0.25
0.2
eli
ci
E 0.1 5
::;)
Q;
a:
0.1
0.05
0
1925 1935 1945 1955 1965 1975 1985 1995
Year
Figure 15.6: Rates of return on stocks and bonds in Denrnark

Source: Jan Overgaard Olesen, 'Empirical Studies of Price Behavior in the Danish Stock Market', PhD dissertation,
PhD serie 2.2001, Copenhagen Business School.
capital has been normalized to 1 so that the replacement value of the capital stock is
simply K,. Note the direct link between stock prices and our q-variable: if the market price
of shares in the firm goes up. the value of q increases correspondingly. The advantage of
specilymg our theory of investment in terms of q is that this variable can be measured
empirically (since stock market values as well as replacement values of business assets can
in principle be observed), whereas the expected future dividends underlying V and q are
very hard to measure. Introducing q, therefore, helps to make our theory of investment
empirically testable.
Assuming that the fim1 communicates its investment plans lor the current period to
its owners, the shareholders will know the size of the firm's capital stock at the start of the
next period (K,+ 1), but they cannot know for sure what the level of the stock price q,+ 1 at
that time '>\Till be. However, assume they expect the share price per unit of capital one
period from now to be the same as the current share price so that q~+l = q,. We then
have:
(7)
We have now expressed V~+l in a form that will turn out to be convenient. Let us next
consider the other determinant of current market value V, = (D~ + V~+ 1)/(1 + r + t:), that is,
the expected dividend D~ for period t. Suppose for the moment that the llrm iinances all
of its current investment spending I, via retained proiits (we shall consider external
iinancing later on). Furthermore, suppose realistically that increases in the linn's capital
stock imply adj11stment costs, including costs of installing new machinery, costs of training
workers to use the new equipment. and possibly costs of changing the firm's organization.
For convenience. all such costs will be called 'installation costs' and will be denoted by c(l,)
to indicate that they are a function of investment spending.
With these assumptions, the expected dividend lor period t '0\lill be equal to the
expected prollt in period t, IT ~ (measured before deduction of installation costs), minus that
part ofpront which is retained in order to llnance the expenditure r, + c(l,) associated with
new investment:
D~ =II~- I, - c(T;), c(O) = 0, c! > 0 . (8)
It seems reasonable to assume that installation costs will rise more than proportionately
with investment spending. If investment is low. the changes in the capital stock are small
and can be accommodated without signillcant changes in the firm's organization. But
when investment is high, the firm may have to undertake signillcant organizational
changes and extensive training of employees, and the attention of managers will be
diverted from the llrm's day-to-day business. Such an organizational overhaul is typically
very expensive. A simple installation cost function capturing this assumption is:
c(T,) =
a
r,.
2
(9)
2
where a is a positive constant. Equation (9) implies that dc/dl, = al" that is, the marginal
installation cost increases in proportion to the level of investment, reflecting th at large
changes in the capital stock are disproportionately more costly than small changes.
To derive our investment schedule, we llnally need the bookkeeping identity:
(10)
stating that the capital stock at the beginning of period t + 1 equals the capital stock
existing at the beginning of the previous period plus the level of investment during period
t. For simplicity, (10) abstracts from depreciation of the existing capital stock, but as you
will learn from Exercise 2 . our theory of investment can easily be generalized to allow for
depreciation.
Starting from (2), and inserting (7)-(10) into (2), we now llnd that the llrm's market
value at the start of period t can be written as follows:
D~ V"
--------
a- - t+ 1
..----..
rr~- r, -
V = _D_,_~_+_V_,~c:.+_,_1 2 r~ + q,(K, + r,) (11)
1
1 + r+ t 1+r+f
prices
© The McGraw-Hill
Companies. 2005
The ftrm chooses its level of gross investment I 1 so as to maximize the initial wealth of its
owners. V1• taking the stock market's valuation ofa unit of capital (q 1) as given. The first-order
condition oV/i'JI, = 0 lor the solution to this maximization problem yields:
forgone
expected dividend
__..,__,_
capit~gain
de
q, 1+-.
dl,
and then. using dc/dl, = al1:
q- 1
I = - '- - (12)
' a
The investment rule q 1 = 1 + dc/dl1 may be explained as follows: to finance the acquisition
and installation of an extra unit of capital in period t, the firm must reduce its dividend
payout in period t by an amount equal to the acquisition cost of a unit of capital - which
we have set equal to 1 - plus the marginal installation cost dc/dl1• This forgone dividend
1 + de/dl1 is the shareholder's marginal opportunity cost of allowing the firm to undertake
an extra unit of Investment. The shareholder's marginal benellt from Investment Is the
gain q1 in the value of shares resulting from the installation of an extra unit of capital. At
the optimal level of investment, the marginal dividend forgone is just compensated by the
extra capital gain on shares. Clearly. the higher the market valuation q, of an extra unit of
capital. the further the flrm can push its level of investment before the marginal install-
ation cost reaches the threshold where the shareholder's additional capital gain is ollset by
the extra dividend forgone. Hence we obtain the simple investment schedule in (12)
which says that investment will be higher the higher the level of the stock price q.
Equation (12) also shows that high marginal installation costs (reflected in a high value of
the cost parameter a) reduce the optimal level of investment. as one would expect.
This is also clear from the graphical illustration of the determination of investment in
Fig. 15.7: a higher value of a increases the slope of the curve 1 + dc/dl, = 1 + al1 and
thereby reduces the value-maximizing level of investment where 1 + de/dl, = q 1•
The investment schedule (12) also holds when investment outlays are financed by
issuing new debt or new shares rather than by retaining prollts. Regardless of the mode of
Hnance, the installation of an extra unit of capital will increase the expected market value
of the Hnn's assets by the amount q., still assuming that the current stock price gives an
indication of the expected value. If the cost of buying and installing the extra unit of
capital (1 + dcfdl1) is financed by an increase in the ftrm's outstanding debt or by the issue
of new shares. the expected increase in the market value of the shares owned by the Hrm's
existing shareholders will be equal to the rise in total market valueq 1 minus the value of the
newly issued debt or equity , 1 + dc/dl1• Of course it is optimal for existing shareholders to
let the Hnn expand its investment until the expected marginal gain in the value of their
shares is driven down to 0. that is, until q1 - (1 + dc/dl,) = 0 . But this is exactly the invest-
ment rule leading to the investment schedule in (12)! Hence we obtain the important
result that the investment function I1 = (q 1 - 1)/a is valid regardless of the method of invest-
ment finance.
The theory outlined above is known as Tobin's q-theory of investment, named after
Nobel Laureate James Tobin who was the first economist to give a systematic formal
prices
© The McGraw-Hill
Companies. 2005
lCD
15 INVESTMEN T AND ASSE T PRICES 445
qt
1 + dc/d/1
qt
It
Figure 15.7: The optimal level of investment
account of the link between stock prices and business investment. 2 Figure 15.8 plots total
fixed business investment in the US against an estimate of Tobin's q. dellned as the ratio of
the stock market value of llrms to the replacement value of their capital stock, as above.
We see that although total investment and the q-ratio do tend to move together over
the longer run, they do not always change in the same direction in the short run. Part of
the problem may be that. in practice. stock prices rellect many 'intangible' business assets
besides physical capital. for example patents and know-how. Another part of the explana-
tion for the sometimes weak relationship between investment and Tobin's q may be that
the estimated value of q reflects the average ratio between the total market value and the
total replacement value of the capital stock, whereas in theory, investment decisions
depend on the margirwl value of q. that is, on the increase in market value relative to the
acquisition price of an additional unit of capital. In our analysis above, the marginal
and average values of q were identical, because our simplifying assumptions implied
proportionality betv.reen the frrm's future capital stock and expected future prollts. But if
this proportionality breal<s dmoVll, the marginal q will no longer coincide with the average
q, and since only the latter can be measured empirically, this may make it dillicult to test
the q-theory of investment.
The role of interest rates, profits and sales
How does the q-theory of investment square with the conventional assumption that
investment depends negatively on the real interest rate( The claim that investment varies
2 . The classi: stat ement of t he theory was given in James Tobin, 'A General Equilibrium Approach to Monetary Theory',
Journal o/ Money, Credit, and Banking, 1, 1969, pp. 15 -29. The theory was later refined and extended by Fumio
Hayashi, 'Tobin's Marginal q and Average q: A Neoclassic al Int erpretation', Econometrica, 50, 1982, pp. 2 13-224.
- I Sorensen-Whitta-Jacobsen: I Part 5- The Building 15. Investment and asset © The McGraw-Hill
3 1600
1400
2.5
1200
2
1000
1.5 BOO
600
400
0.5
200
o +-~~-r-r~~~~-r-r~~~-T-r,-~~~-r-r~ 0
1975 1980 1985 1990 1995 2000
Year
Figure 15.8: Investment and Tobin's q in the United States

' Real private non-residential fixed capital formation measured in 1996 US dollars.
Source: Board of Governors of Federal Reserve System (1999), Aow of Funds Accounts of the United States:
Flows and Outstanding Stocks (fourth quarter 1998).
positively with stock prices is fully consistent with the hypothesis that it varies negatively
with the real interest rate.
To see this, let us go back to Eq. (6) and let us assume lor simplicity (since this will not
ailect our qualitative conclusion) that real dividends are expected to stay constant at the
level D~ from period t and onwards. Equation ( 6) then becomes:
V, = D; 1 + 1 ,+ 1 + ... ]. (13)
[ 1+r+e (1 + r+ e)- (1 + r+e) 3
If we multiply both sides of(13) by 1 + r + e and subtract (13) from the resulting equation,
we get:
D''
V = -~ (14)
t r +e
Equation (14) is just a special version of the general formula stating that the value of the
11rm equals the present discounted value of expected future dividends. Now recall that, by
dellnition, V, = q,K,. From this relationship and (14) it follows that:
D~/K,
q,=- -. (15)
r +c
According to ( 15) the market value of a unit of the firm· s capital stock (q,) equals the dis-
counted value of the expected future dividends per unit of capital. Hence a rise in the real
interest rate r will. ceteris paribus, reduce the stock price q,, and this will reduce invest-
ment. From Figs 15.6 and 15.8 we have seen that stock prices do tend to adjust to keep the
return on stocks in line with the return on bonds, and that investment tends to move in
line with stock prices. This is indirect evidence of a negative impact of interest rates on
investment.
Figure 1 5.9 shows direct evidence of the relationship between the bond interest rate
and the ratio of business investment to output in Denmark. 3 We see that there has been a
clear negative relationship between the interest rate and investment in the second half of
the 1990s. In other periods the negative correlation has been less clear, but this is not sur-
prising. since our Eq. (1 5) implies that q will fluctuate not only with interest rates, but also
with changes in the risk premium f. and in the expected dividend ratio, D'/K. Let us take a
closer look at the likely determinants of the latter variable.
It seems reasonable to assume that expected dividends are positively related to the
observed current protlts of the firm. IT,. For concreteness, suppose shareholders expect that
the firm will pay out a fraction eofits profits as dividend~ at the end of the period, soD~ = eiT 1•
In that case the numerator of (15) may be \oVTitten as 8(II/ K,). where II/K, is the firm's
current rate ofprotlt (the profit to capital ratio). From (15) we would then expect to observe
a positive correlation between (changes in) the current profit rate and (changes in) current
investment. Figure 15.10 suggests that such a positive relationship does in fact exist.
We would also expect a positive relationship between the protlt rate and the output-
capital ratio. For example, we know from growth theory that if output Y is given by the
Cobb- Douglas production function Y = AK" L1 - " (where Lis labour input). and if markets
are competitive, total profits will be equal to a Y (see Chapter 3, Section 1). In that case the
rate of profit is a Y/K which is directly proportional to the output- capital ratio Y/K. Even if
markets are not competitive, it is still reasonable to assume that the more firms can
produce and sell on the basis of a given capital stock. the higher their profit rate will be.
Figure 15.11 roughly confirms this expectation.
25 0.25
23 Ratio of business
investment to output
21 (right axis) 0.2
19
17 0.15
0
'$. 15 ·-;
0::
13 0.1
11
9 0.05
7
5
-
.....
C»
(")
.....
C»
..... .....
It)
C»
..... .....
C»
C»
C»
Cii
C»
(")
00
C»
It)
00
C»
.....
00
C»
C»
00
C»
-
(1)
C»
(')
(J)
(J)
It)
C»
C»
.....
C»
C»
C»
C»
C»
0
0
0
"'
Figure 15.9: Investment ratio and bond interest rate in Denmark
Source: MONA database, Danmarks Nationalbank.
3. Ideally, Fig. 15.9 should plot the real interest rate, but since the expected inflation rate is unobservable, we have
pragmatically chosen to show the nominal interest rate as a proxy for the real rate.
prices
© The McGraw-Hill
Companies. 2005
140 20
120 lii
Change in profit to ·a.
capital ratio 15 ~
'0 100
.£
~~
E o
gj ~
80
60
~ 10 ""e
Q.
>a> o~
·~ ~ 40 5 0~
·~2--
~~
c c
20
~
~
0 .;;
~g 0 .S
0~ ~
-~ m
-5 ~
~0 6
-60 +-~r-~~~~-r-r-r,_~~r-~~~-r~- 1 0
0 00 0 00 (J:) 0
(J:) (J:) (I) 00 (1) 0
(1) (1) (1) (1) (1) 0
C'<
Year
Figure 15.10: Changes in investment and profit in the United States 1960-2000
Source: National Income and Products Account s (Bureau of Economic Analysis).
8 Change in profit to 12
~ capital ratio
lii 6
·a. lii
·a.
"' 4 7
0
E E
"'
0
:; 2
Q.
"5
0
:; 0 2 a.
'15~ o~
.Q~ - 2 ·~~
'§
-4
=
-3
=
~
~
.S -6 .S
~
~
-8 "'c
0
"'c
"'
..c
- 10
~ Change in output to -8
0"'
..c
capital ratio
- 12 - 13
0 '<I' (I) C'< (J:) 0 '<I' 00 C'< (J:) 0
(J:) (J:) (J:) t'- t'- 00 00 (I) (1) (1) 0
(1) (1) (1) (1) (1) (1) (1) (1) (1) (1) 0
C'<
Year
Figure 15.11: Changes in the ratio of output to capital and profit to capital in the United States
1960-2000
Source: National Income and Products Accounts (Bureau of Economic Analysis).
Although a higher current prollt is likely to boost expected future dividends. it is too
primitive to expect a mechanical one-to-one impact of the former on the latter variable.
Firms and investors may sometimes have good reason to expect that future prolltability will
deviate from realized current prollts. Indeed. the fact that the economy moves up and down
in cycles suggests that intelligent investors \<Viii not mechanically extrapolate current
Introducing Advanced Blocks for the Short-run prices Companies. 2005
15 INVES TMENT AND ASSET PRICES 449
earnings into the future. Instead they will revise their expectations regarding future sales
and profits as they receive new information on relevant economic and political events.
Most advanced countries publish indices of 'business confidence' to measure business
expectations regarding the near future. Usually these indices build on survey data where a
sample of business managers report current and expected movements in their future
output, sales, employment. investment, etc. Figure 15 .12 shows the evolution of one such
index of business conlldence in the US. We see that business confidence can fluctuate quite
a lot and that it is sometimes signiflcantly aflected by unanticipated political events.
Hence. to understand the high volatility of investment spending, we have to allow for
changes in expectations regarding the future. Our earlier discussion of Fig. 15.5 suggests
that part of the impact of a change in expectations on investment may be caused by a
change in the required risk premium e.
To summarize. we have seen that the q-theory of investment is quite consistent with
the hypothesis that business investment varies positively with the level of output (sales),
whereas the existing capital stock and the real interest rate both have a negative impact
on investment. Specifically, if E is an index of the 'state or confidence'. we have argued that
the expected dividend ratio D~/K, will be given by a function like g(YJK,. E1), where both
of the first derivatives of the g()-function are positive. Using this relationship along with
(12) and (15), we may thunvrite our investment function as:
I, =(!_)(D~/K, _1)= (!_)(g(Yt /K 1, E,) _ 1)·

a r +e a r +e
or, in more general form , and dropping time subscripts for convenience:
I=f\~· <~'(~)'(¥)), (16)
where the signs below the variables indicate the signs of the corresponding partial
derivatives ofthe.f()-fimction . In terms of the q-theory, an increase in Yor E will stimulate
50 US Embassies n Kenya
New York Stock and Tanzania attacked
40 Exchange crash Aug.1998-
Oct. 1987
30
20
10
\
0+---~J--------+-----+--~~--~----------~~
- 10 ; World Trade Center and

- 20 Iraq invades Pentagon attacked
Kuwait Sept. 2001
- 30 Aug. 1990
- 40
- so +--.-.-,,-.-.--.-.--.-.--.-.--,-.--.-.-,--
1985 1987 1989 199 1 1993 1995 1997 1999 2001
Year
Figure 15.12: Business confidence in the United States

Source: IMF, World Economic Outlook, December 2001.
prices
© The McGraw-Hill
Companies. 2005
0.08. -- - - - - - - - - - - - - - - - - - - -
Actual investment
-0.02
co co co
CD
<»
0 co
""<» "'""<» ""<» ""<» ""<»
<Q' CD 0
co
<»
"'co<» <Q'
co
<»
CD
co
(J)
co
<»
0
<»
<»
"'<»<»
<Q'
<»
<»
CD
<»
<»
<»
<»
Year
Figure 15.13: Actual and predicted values of net investment ratio in Denmark 1
1
Net business investment in machinery and equipment relative to value added.
Source: S MEC database, S ecretariat of the Danish Economic Council.
investment by rmsmg q 1 through an increase in the expected dividend ratio

D~/K1 = fJg(YJK,, E,) (and possibly through a fall in the risk premium t:). An increase in the
current capital stock K reduces investment by driving down D~/K1 • and an increase in the
real interest rater likewise discourages investment via a negative impact on q,.
In the case of firms whose market value is not directly observable because they are
not quoted on the stock exchange. it is inappropriate to interpret Eq. (16) literally
in terms of the q-theory. Nevertheless, as Exercise 3 ''\Till make clear. the investment
behaviour of such firms may still be described by an equation like (16) if they invest with
the purpose of maximizing the present value of the net cash flow to their owner (thereby
maximizing his wealth). Equation (16) therefore summarizes our general theory of
business investment.
Econometric research has confirmed that changes in Y. K and r influence investment in
the manner indicated in (16 ). However, researchers have also found that it is quite dillicult
to explain liilly all of the observed movements in investment. To illustrate, Fig. 15.13 plots
actual investment against the predicted level of investment estimated on the basis of a
sophisticated version of the investment function (1 6). To a large extent, the dilliculties of
predicting investment undoubtedly stem from the dilllculty of finding reliable quantifiable
proxies for the state of confidence', E. Because expectations are so hard to measure and may
sometimes change abruptly, it is inherently dill1cult to forecast investment.
15.4 The housing market and housing investment

················································································································································································
A q·theory of housing investment
Housing investment is an important component of total private investment. and as indi-

cated in Table 15 .1. it is even more volatile than business investment. Hence fluctuations
prices
© The McGraw-Hill
Companies. 2005
lCD
15 INVESTMENT A ND ASS ET PRICES 451
Table 15.1: The volatility of housing investment

Average share
of total Coefficient of Absolute Volatility relative Volatility
private correlation with volatility to volatility of total relative to
investment contemporaneous (standard private fixed volatility of
(%) GOP (same quarter) deviation) investment GOP
Belgium 29 0.47 6.15 1.37 5.78

Denmark 28 0.70 9.57 1.20 5.72
Netherlands 32 0.22 5.95 1.47 4.87
United Kingdom 25 0.47 8.78 1.82 5.81
United States 31 0.54 10.19 2.12 6.13
Sources: UK Office for National S tatistics, Ecowin, Datastream, Bureau of Economic Analysis,
Bureau of Labor Stati stics, Federal Reserve Bank of St. Louis, Danmarks Nationalbank, Belgostat,
Banque Nationale de Belgique, S tatistics Netherlands and De Nederlandse Bank.
in residential investment often play an important role during business cycles. A basic
factor contributing to the volatility of housing investment is the fact that housing capital
is highly durable. In any year the construction of new housing is only a very small fraction
of the existing housing stock. To accommodate even small percentage changes in con-
sumer demand for housing capital, construction activity may therefore have to undergo
large relative changes.
In this section we will show that housing investment may be explained along lines
which are similar to the q-theory of business investment. The present section may there-
fore be seen as an illustrative special version of the q-theory, adapted to fit the housing
market. A~ a byproduct of our theory of housing investment, we will develop a theory of
the forma don of housing prices and identify the factors which may cause fluctuations in
the market value of the housing stock. Since the stock of housing capital is an important
component in total household wealth, and since the next chapter will show that private
consumption depends on private wealth, the theory of the housing market developed
below will also help us to w1derstand fluctuations in private consumption.
We start by considering the production function of the construction sector. For con-
creteness, suppose th at the constmction of new housing, [H, is given by th e production
function:
111 = A · Xfl. 0 < (3 < 1, (17)
where X is a composite input factor (to be specilled below), and A is a constant which
depends on the productive capacity of the constmction sector. The assumption that the
parameter {3 is Jess than 1 implies that. over the time horizon we are considering,
production is subject to diminishing returns to scale.
For simplicity, we assume that construction llrms combine labour L and building
materials Qin llxed proportions. Specifically, each unit of the composite input X includes a
units oflabour and b units of materials:
L= aX, Q = bX. (18)
If Wis the wage rate and p<l is the price of materials, it follows from (18) that the price P of
prices
© The McGraw-Hill
Companies. 2005
a unit of the composite input X is equal to

(19)
We will refer to Pas 'the construction cost index'. If pH is the market price of a unit of
housing, the sales revenue of the representative construction firm will be p 11 I 11 , and its
profits, rr. will be
IT= p11 l 11 - PX =p 11 l 11 - P(I 11/A) 11P. (20)
In deriving the second equality in (20). we have solved (17) for X and substituted the
solution into the expression for pronts. Taking the housing price p 11 and the input price P
as given. the construction llrm chooses its level of activity l 11 with the purpose of maxi-
mizing its profit. (We might also assume that the flrm maximizes its market value. This
would give the same results. but via a more cumbersome procedure.) According to (20) .
the first-order condition for profit maximization, diT/dl 11 = 0. implies:
d(PXl/til'' =marginal
construction cost
..----..-......
PH -.!_ (IH)(l-p)/{J = 0
{3A A
[~'~ = k. - (PpH)fl/(1-p) ' k =f3PI<J -P> A 1/(1-{J) . (21)
Equation (21) is the supply curve for the construction sector. It is seen to be derived from
the fact that prollt-maximizing construction flrms will push construction activity to the
point where the marginal construction cost equals the market price of a housing unit. The
relative price variablep 11/P is an analogue of Tobin's q. Thus, since 0 < f3 < 1. Eq. (21) says
95 1.5
Price of housing relative to 1.4
85 construction costs
(right axis) 1.3 -0 c0
0:::
::.:: 75 1.2 Q) ·~
() ()
0 ·~ 2
10
65 1.1 a. -
0> ---~
(/) 0
0>
1 Q) ()
.g (ij
(/)
c 55 o.g a. ·~
0 Q) c
iii 45 0.8 0~~
(/)
J:~
investment 0.7
35 (left axis)
0.6
25 0.5
-.,.. .,..
(')
.,.. .,...,.. .,..
10 0>
-
(I)
(')
(I)
10
(I)
.,..
(I)
0>
(I) -
0>
(')
0>
10
0>
.,..
0>
0>
0> -
0
0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0
"'
Figure 15.14: Residential investment and the price of housing relative to construction costs in
Denmark
Source: MONA database, Danmarks Nationalbank.
th at housing investment IH will be larger the higher theq-ratio of the housing price to the
construction cost index is.4 Figure 15.14 shows that this theory of housing investment
fits the facts very well.
Housing investment, interest rates and income
Like the q-theory of business investment, our theory of housing investment is consistent
with the hypothesis that investment varies negatively with interest rates and positively
with total income. To demonstrate this, we will now develop a theory of housing demand
in order to explain the housing price p 11•
Consider a representative consumer who has borrowed to acquire a housing stock H
at the going market price p 11 per unit of housing. Suppose the consumer has to spend an
amount <5p 11 H on repair and maintenance each period to maintain the value of his house,
and suppose the interest rate on mortgage debt is r. The consumer's total cost of housing
consumption will then be (r + b)p 11 H. The consumer also consumes an amount C of non-
durable goods. IfhL~ income is Y, and if we abstract fi:'om savings (which will not allect our
qualitative results), the consumer's budget constraint may then be written as:
C+ (r + o)p 11 H = Y, (22)
where we have set the price of non-durables equal to 1.

The consumer wishes to allocate his total consumption between housing and non-
durables so as to maximize his utility U which we assume to be given by the Cobb- Douglas
function:
0<17<1. (2 3)
In practice. the consumer will derive utility fi:'om the housing sen1ice flowing from the
housing stock H, and not from the housing stock as such. The specification in (23) just
assumes that the housing service is proportional to the housing stock. Using the budget
constrain t (22) to eliminate C from (23), we get:
(24)
The consumer's optimal level of housing demand is found by maximizing the utility li.mc-
tion (24) with respect to H. The first-order condition dU/dH = 0 for the solution to this
problem is:
oU/iJH oU/ilC
-------'~ ~~-------
;;H'I -1[ Y - (r + b)p11 H) 1 ~- (r + <~)p 116 - 1])H'i [Y - (r + <5)pHH]~ = 0. (25)
or
()Uj ()H N
()Uj ()C = (r + c~)p . (26)
Equation 126) says that. in the consumer's optimum situation, the marginal rate of sub-
stitution between housing and non-durables (the left-band side) must equal the relative
4. In this theory of housing investment the assumption of diminishing returns to the composite input X has taken the
place of the installation costs which we included in our model of business investment.
eI Sorensen- Whitta- Jacobsen:
prices
© The McGraw-Hill
Companies. 2005
price of housing, (r + b)pN. If we solve (2 5) lor H, we get the demarrd for housing, now
denoted as H' 1:
'YfY
H" = . (27)
(r +o)p''
The term (r +<~)pH in the denominator of (2 7) is sometimes referred to as the user-cost
of housing, reflecting the financial cost, r, as well as the costs of maintenance, captured by
the parameter c) wh ich may be seen as a depreciation rate for housing capital. We see from
(2 7) that housing demand varies positively with income and negatively with the user-cost
of housing. Note that even if the consumer has financed the acquisition of his house by his
own past savings. the user-cost should still include the interest rate r as an opportunity
cost. since this is the income the consumer forgoes by investing his savings in a house
rather than in interest-bearing assets. 5
While (2 7) gives the demand lor housing. the aggregate supply of housing is fixed in
the short run where the housing stock is predetermined by the accumulated historical
levels of housing investment. In other words. at the start of each period there is a given
predetermined housing stock, since the current construction activity determined by (21)
does not add to the housing stock until the start of the next period. In the short run. the
market price of houses must therefore adjust to bring the demand lor housing. H'1, in line
with the existing supply. H. Inserting the equilibrium condition H' 1= H into (2 7) and
solving for p 11 , we get the market-clearing price of houses:
H 'I]Y
p = (28)
(r + c\)H
Figure 15.15 illustrates how the equilibrium price of houses is determined in the
short run where the supply of housing is fixed at the level H•. Ceteris paribus, a higher pre-
existing housing stock will imply a lower current housing price. We also see from (28) that
the housing price will be lower the higher the real interest rater and the lower the level of
income. Y.
Since we know from (21) that current construction activity varies positively '>\lith the
housing price. we may combine (21) and (28) to get a housing investment function of the
form
IN = k.
'I]Y r/(1-{J)
[ (r + o)PH '
or more generally:
I 11 = h(Y. H,r ) . (29)

(+) (-) (-)
The negative impact of the interest rate on housing investment in (29) is based on the
theory that a higher interest rate '>Viii, ceteris paribus, reduce the market price of housing.
The negatively-sloped regression line in Fig. 15.16 confirms that housing prices do in fact
tend to fall when the bond interest rate goes up. and vice versa.
5. In practice, the tax system also affects the user·cost of housing (see Exerc ise 4). Moreover, if the consumer expect s
a capital gain on his house due to a rise in p H, this gain should be subtracted from the total user cost Here we
abstract from (ant ic ipated) capital gains for simplicity.
prices
© The McGraw-Hill
Companies. 2005
lCD
15 INVESTMEN T AND ASSET PRICES 455
Hd = 17Yf(r + <l}H
Ho H
Figure 15.15: Short-run equilibrium in the housing market
10
1/)
Q)
0
•
·~
Q)
1/)
::J 5
0
~
.!:::
Q)
0)
c
~
"'
0 0
(ij
::J
c:
c
"'
Q)
5
0)
c"' ~
~
(;;
a.
- 10 •
- 25 - 20 - 15 - 10 -5 0 5 10 15 20
Percentage an nual change in bond interest rate
Figure 15.16: The long-term bond interest rate and house prices in Denmark
Note: Percentage changes are nonnalized by subtracting the average change from all observations. Quarterly data,
1975- 1998.
Source: MONA database, Oanmarks Nationalbank.
prices
© The McGraw-Hill
Companies. 2005
acroeconomics Model
456 PART 5: T HE B UILDING BLOCKS FOR THE SHO RT-RU N MODEL
Housing market dynamics
At the aggregate level, part of the current construction activity, IH. serves to compensate
for the depreciation of the existing housing stock, ()H. The housing stocks in period t and
in period t + 1 are therefore linked by the identity
(30)
Equations (21), (28) and (30) constitute a simple dynamic model of the housing market.
For given values of Y and r. the predetermined housing stock, H1, determines the housing
price for period t via (28) . Given the value of P. Eq. (21) then determines the current level
r:
of housing investment 1 which subsequently determines the next period's housing stock
H1+ 1 via (30) . We then get a new housing price v:~ 1 via (28) which enables us to deter-
mine r;~ 1 by use of (21), giving a new housing stock H1 _ 2 via (30). and so on . This
dynamic process will continue until the housing price has reached a level where con-
struction actiYity is just suflicient to compensate for the depreciation of the existing
housing stock so that the stock of housing remains constant. Thus, whereas an upward
shift in housing demand is fully absorbed by a rise in house prices in the short run, over the
longer run it will cause an increase in the housing stock which will dampen the initial
price increase. Exercise 4 asks you to explore these dynamics of the housing market in an
extended model allowing for property taxes and income taxes.
15.5 Summary
................................................................................................................................................................................
1. Empirically, changes in stock prices and in housing prices tend to be followed by changes in
output in the same direction. In part this reflects that higher asset prices lead to higher
investment. This chapter explains the links between asset prices and investment.
2. A firm seeking to maximize the wealth of its owners will choose an investment plan which
maximizes the market value of the firm's assets. The value of the firm, referred to as the
fundamental stock value, is the present discounted value of the expected future dividends
paid out by the firm. This follows from the shareholder's arbitrage condition which says that
the expected retum to shareholding, consisting of d ividends an d cap ital gains on shares,
must equal the return to bondholding plus an appropriate risk premium.
3. When share prices reflect the fundamental value of firms, there are three possible reasons for
the observed volatility of stock prices: (i) fluctuations in (the growth rate of) expected future
dividends, (ii) fluctuations in the real interest rate, and (iii) fluctuations in the required risk
premium on shares. There is indirect evidence that the requ ired risk premium fluctuates quite
a lot.
4. The evidence suggests that the rate of return on stocks is tied to the rate of return on bonds
over the long term. This accords with the view that stock prices reflect the fundamental value
of firms. However, many observers believe that stock prices can sometimes deviate from
fundamentals. The analysis in this ch apter abstracts from such 'bubbles' in stock prices.
5. Increases in the firm's capital stock imply adjustm ent costs (ins1allation costs), including costs
of installing new machinery, costs of training workers to use the new equipment, and perhaps
15 INVESTMEN T AND ASSE T PRICES 457
costs of adapting the firm's organization. These installation costs will typically increase more
than proportionally to the firm's level of investment.
6. The value-m aximizing firm will push its investment to the point where the shareholder's capital
gain from a unit increase in the firm's capital stock is just offset by the dividend he must forgo
to enable the firm to purchase and install an extra unit of capital. Because the marginal instal-
lation cost is increasing in the volume of investment, this investment ru le implies that the firm's
optimal level of investment will be higher, the higher a unit increase in the firm's capital stock
is valued by the stock market. An increase in the ratio of stock prices to the rep lacement cost
of the firm's assets will therefore stimulate its investment.
7. The market value of stocks relative to the replacement value of the underlying business assets
is referred to as Tobin's q. Our theory of investment may be summarized by saying that
business investment is an increasing function of Tobin's q.
8. Stock prices reflect expected future dividends which tend to be positively affected by a rise
in current profits. The value of Tobin's q therefore tends to vary positively with current profits,
which in turn vary positively with the output- capital ratio. Hence investment is an increasing
funr.tion of r.urrP.nt outrut !lncl !l dP.r.rP.!lsing funr.tion of thP. P.xisting r.!lrit!ll stor.k.
9. Ceteris paribus, a rise in the real interest rate implies that expected future dividends are
discounted more heavily, leading to a fall in Tobin' s q via lower stock prices. Thus a higher real
interest rate tends to depress investment. A rise in the required risk premium on shares,
generated by more uncertainty about the future, will have a similar negative impact on
investment.
10. A version of the q-theory can explain investment in owner-occupied housing. When the
market price of residential property increases relative to the cost of housing construction, it
becomes profitable for firms in the construction sector to increase the supply of new housing
units. As a consequence, housing investment (construction activity) goes up. There is strong
empirical evidence in favour of this hypothesis.
11. In the short run, the market price of housing varies positively with current income and nega-
tively with the real interest rate and with the existing housing stock. Since construction
increases with the market price of housing, it follows that housing investment is an increasing
function of income and wealth and a decreasing function of the real interest rate and the
current housing stock.
15.6 Exercises
Exercise 1. Stock market valuation and 'fundamentals'
The purpose of this exercise is to illustrate how our equation (6) for the fundamental stock
price may be used to evaluate whether stock prices are unrealistically high or low, that is,
whether the stock market is 'overvalued' or 'undervalued'. Suppose that real dividends are
expected to grow at the constant rate g•. If the actual real dividend for period t is 0~ the
expected real dividend for future period n will then be given by:
0~ = 0 1( 1 + g •)n- t for n = t + 1, t + 2, ... (3 1)

prices
© The McGraw-Hill
Companies. 2005
acroeconomics Model
458 PART 5: THE BUI LDING BLOCKS FOR THE SHORT-RUN MODEL
1. Use Eq. (6) in the text to demonstrate that, when expected future dividends are given by {31 ),
the value of shares w ill be given by:
V= D, {32)
1
r+ e -g•
Stock market analysts often focus on the so-called 'trailing dividend yield', 0 1/ V" defined as
the current d ividend relative to the current market value of shares. We w ill simply refer to this
ratio as the 'dividend yield'.
2. Suppose you have information on the current dividend yield, D,I V~ the real interest rate, r, and
the required risk premium on shares, e. Use {32) to solve for the value of the expected real
growth rate g• wh ich is necessary to justify current stock prices.
3. Suppose alternatively that, in addition to information on the current d ividend yield and the
current real interest rate, you have somehow obtained information on the expected future
growth of real dividends whereas you do not know the required risk premium on shares. Use
{32) to derive the value of the risk premium which will justify current stock prices.
4. In the US in 1999 the average dividend yield was 1 .2 per cent and the real interest rate on
{approximately) risk-free ten-year government bonds was 3.4 per cent. Moreover, the average
historical risk premium on shares in the period 1980- 99 was 2.8 per cent per annum {all of
these figures are taken from the IMF's World Economic Outlook, May 2000). What was the
annual growth rate of real dividends wh ich US financial investors expected in 1999 if they
required a risk premium equal to the historical average? On the basis of your result, would you
say that the US stock market was overvalued or undervalued in 1999? Justify your answer.
5. Over the period 1980-1 999 the average growth rate of US real GOP was 3 .0 per cent.
Suppose now that US investors in 1999 expected future real d ividends to grow in line w ith
historical GOP growth so that g• = 0.03 {discuss whether this might be a reasonable assump·
tion). G iven the other pieces of information in the previous question, what was the risk
premium on shares required by US investors in 1999? Would you say that this risk premium
was 'reasonable'? Can you imagine any reasons why US investors in 1999 should require a
lower or a higher risk premium than the average historical premium?
Exercise 2. A generalized q-theory of investment
Suppose that instead of Eq. {9) in the main text of th is chapter, the installation cost function
takes the more general form:
a , q+ 1
c(l) = __ YJ > 0. {33)
I 1]+ 1 I '
1. Derive an expression for investment as a function of q 1 (a generalized version of Eq. {12)).

Give a brief verbal explanation of the idea and economic mechanisms underlying the q·theory
of investment. Explain why the q·theory is consistent w ith the well-documented fact that
investment tends to vary negatively w ith the real interest rate and positively w ith economic
activity.
o
Suppose that during each period, a fraction of the capital stock has to be scrapped
because of wear, tear and technical obsolescence, so the change in the capital stock is given
by:
Kl+ ,- K, = I,- oK, (34)
where 11 indicates gross investment, including the replacement investment serving to com-
pensate for depreciation. Suppose that only net additions to the capital stock generate
adjustment costs. In that case installation costs will be determined by net investment 11 - oK1
so that (33) must be replaced by:
c(l) = - a-(1 -oK)''+,. (35)

t I] + 1 t t
2. Discuss whether it is reasonable to assume that only net investment (but not replacement
investment) generates adjustment costs. Derive a revised expression for gross investment lb
assuming that installation costs are given by (35).
In a stationary economy with no long-run growth, a long·run equilibrium requires the capital
stor.k to hP. r.onst~nt ovP.r timP.. In sur.h ~ situ~tion whP.rP. nP.t invP.stmP.nt is 7P.ro thP.rP. is no
need for firms to retain any part of their net profit. Hence all net profits will be paid out as
dividends. According to Eq. (15) in the text, this implies q,= (IT 1/K 1)/(r+ c).
3. What is the ratio of the market value to the replacement value of the firm's capital stock in a
stationary long-run equilibrium? Furthermore, what is the relationship between the profit rate
IT ,/K1 and the requ ired return on shares in such a long-run equilibrium? Try to provide some
economic intuition for your results.
Exercise 3. Tax policy and investment
This exercise serves two purposes. First, you are asked to demonstrate that the investment
behaviour of unincorporated firms is similar to the investment behaviour of corporations which
are quoted on the stock market. Second, you are invited to study how various forms of capital
income ta)ation will influence investment.
We consider an entrepreneur who owns a private unincorporated business firm. We
d ivide the entrepreneur's time horizon into two periods which may be thought of as 'the
present' (period 1) and 'the future' (period 2). At the beginning of period 1 the entrepreneur
has accumulated a predetermined capital stock, K,, which is invested in his firm. During
period 1 he incurs gross investment expenditure, I, with the purpose of maintaining and
increasing his capital stock. In each of the two periods, the capital stock depreciates at the
rate o, so at the beginning of period 2 the entrepreneur's capital stock will be given by:
O <o< 1. (36)
The entrepreneur undertakes investment and employs labour with the purpose of maximizing
the present value V of the net cash flows IT 1 and IT 2 withdrawn from the firm during periods
1 and 2, respectively. In other words, the entrepreneur wants to maximize:
(37)
At the end of period 2 the entrepreneur p lans to liquidate the firm and sell its remaining assets
prices
© The McGraw-Hill
Companies. 2005
acroeconomics Model
at their replacement value (1 - o)K?. For simplic ity, we assume that there are no adjustment
costs associated with changing the firm's capital stock. If Y, is the firm's output, L, is labour
input, and w, is the real wage rate, the cash flows from the firm to the owner during the two
periods w ill then be:
n, = Y1 - w 1 L 1 -I, (38)
ll 2 = Y2 - w2 L2 + ( 1 - o)K2 , (39)
where (39) includes the revenue from the sale of the firm's rerraining capital stock at the end
of period 2. Output in the two periods is given by the Cobb - Douglas production function:
Y,= A , KfL~, 0 <a< 1, 0< P< 1, 0< a+,8<;; 1, t = 1, 2. (40)
1. Demonstrate that the entrepreneur's optimal gross investment during period 1 can be written
as:
aY2
1= - -<1 -a)K1 . (41)
r+O
(Hint: use (36) to eliminate K2 before you derive your first-order condition.) Does the invest·
ment function {41) have the same qualitative properties as the business investment function
derived in the main text of the chapter? (Hint: note that the variable Y2 must be interpreted as
expected output in period 2.)
We now invite you to study the effects of a profits tax. In accordance w ith existing tax rules,
suppose that the firm is allowed to deduct its labour costs and the depreciation on its capital
stock from taxable profits. If the profits tax rate is r , the firm's tax bill Tthen becomes:
T,=r(Y,- w, L,-oK,), 0 <r< 1, t = 1' 2, (42)
and the after-tax cash flows from the firm to the entrepreneur become equal to:
n, = Y1 -w1 L1 - / - T11 (43)

ll 2 = Y2 -w2 L2 +(1-(5)K2 - T2 . (44)
2. Derive the analogue of the firm's investment function (41) in the presence of the profits tax.
Explain how the profits tax affects investment.
We have ~o far a~~umed that i nve~tment i ~ finam;ed by retained profits. Con~ider now the
alternative case where only replacement investment oK is financed by retained earnings
whereas net investment expenditure /- (5K is financed by debt. In that case the firm' s stock of
debt 8 will always be equal to its capital stock, that is:
B, = K, (= 112, (45)
and the firm's revenue tJ.8 1 from new borrowing during period 1 will be equal to its net invest·
ment during that period:
t.B, = ,_ (w,. (46)
Using (45) and (46) and noting that the firm's expenses on interest payments will be rB = rK,
we may then write the net cash flows to the entrepreneur as:
ll 1 = Y1 - w 1 L1 - r8 1 - T1 -I + tl8 1
= Y1 -w1 L 1 -(r+o)K1 - T11 (47)
l12 = Y2 -w2 L 2 +(1 -o)K2 - T2 -(1 +r)B

= Y2 - w2 L2 - (r+ o)K2 - T2 , (48)
assuming that the entrepreneur must repay all of his debt with interest at the end of the
second period. Since the tax code allows interest payments as well as depreciation to be
deducted from taxable profits, the tax bills for the two periods will be
T, = r [Y,- w,L,- (r + o)K,], 0< r< 1, t = 1' 2. (49)
3. Derive the firm's investment function on the assumption that net investment is fully financed
by debt. (Hint: remember to use (36) and (40).) Does the profits tax affect investment? Does
it yield any revenue? Is the tax system neutral towards the firm's choice of financing method?
Explain your resu lts.
Exercise 4. Tax policy and the housing market
In this exercise we will extend our model of the housing market to allow for taxes. You are then
asked to study how the housing market reacts to tax policy in the short run and the long run.
We assume that the government levies a proportional property tax at the rate r on the
current value pH H of the consumer's housing stock H. We also assume that the government
imposes a proportional income tax at the rate m but that it allows a fraction d of interest
expenses to be deducted from taxable income, where 0 ~ d ~ 1. Finally, we assume that the
consumer has to spend an amount <'ipH H on repair and maintenance during each period to
maintain t~ e value of his house. The parameter <'i may thus be interpreted as the depreciation
rate for housing capital. Expenses on repair and maintenance are assumed not to be
deductible from taxable income.
We consider a young consumer who starts out with zero financial wealth at the begin-
ning of period t and who must therefore borrow an amount p~H, to acquire the housing stock
H, at the going market price p~. For simplicity we assume that the consumer's planned net
saving is 0, so his current spending on non-durable consumption goods C, is given by the
budget constraint:
Z,=( 1 -m)Y, -[r( 1 -dm) + (5+r]p~H,, 0<m < 1, (50)
where Yis pre-tax labour income and where the term [r{1 - dm) + <'i + r ]pHH reflects net inter-
est payments on mortgage debt plus expenses on housing repair and property tax. Note that
because interest expenses are deductible, it is the after-tax interest rate r{1 - dm) which
appears in the budget constraint (50).
The consumer's preferences are given by the Cobb- Douglas utility function:
0 < 1] < 1. (5 1)
1. Demonstrate that the consumer's housing demand will be given by:
H = 1J( 1 - m)Y,
(52)
' [r( 1-dm) + <'i+r]p ~'
and give your comments on this expression.
We will now set up a complete partial equilibrium model of the housing market. We start
out by rearranging (52) to get an expression for the housing price:
prices
© The McGraw-Hill
Companies. 2005
acroeconomics Model
H q(1 - m)Y,
p = {53)
' [!(1 - dm) + +r]H,o
The supply side of the housing market is modelled in the same manner as in Section 4 in
the main text. The construction of new houses IH is therefore g iven by Eq. (21) which we
repeat here for convenience:
0< /1< 1, (5 4)
where we remember that k is an exogenous constant. From Section 4 we also recall that the
evolution of the housing stock is given by:
H,. , =(1 -o)H, + I~, (55)
Equations (53)- (55) constitute our model of the housing market where the exogenous vari·
abies and parameters are Y, P, r, q, /J, k, o, d, m and r . At the beginning of each period the
existing housing stock H, is predetermined by the accumulated historical levels of housing
investment. Given the initial housing stock, (53) may therefore be used to find the short-run
equilibrium housing price p~ which may then be inserted into {54) to give the current level of
housing investment 1~. Once I~ is known, we may insert its value into (55) along w ith the pre·
determined value of H, to find the housing stock H,., at the beginning of the next period,
which then determines p~ 1 via (53), and so on.
2. Condense the housing market model (53)- (55) into a single non-linear first-order difference
equation in the housing stock H (for convenience you may set P, = 1, s ince we treat P as an
exogenous constant). State the condition under which th is difference equation is locally
stable so that the housing market will converge on a long-run equilibrium characterized by a
constant housing stock. (Hint: the d ifference equation will be locally stable if dH,. 1/dH, is
numerically smaller than 1. When you derive and simplify your expression for dH,. 1/ dH,, use
Eqs (53) and (54) plus the fact that, in a local stability analysis, the derivative dH,. ,jdH, is
calcu lated in the long-run equilibrium point where I~ = oH, initially, because the housing stock
is constant in long-run equilibrium.) Discuss whether the housing market is likely to be locally
stable for plausible parameter values.
3. In a long-run equilibrium the housing stock is constant, H ,+ 1 = H 1, imp lying from {55) that
IH = oH. Insert this condition for long-run equilibrium into (54) and invert the resulting equation
to get a long-run supply curve for the housing market. Construct a diagram (with H along the
horizontal axis and pH along the vertical axis) in wh ich you draw th is long-run supply curve
along w ith the housing demand curve given by (53). Identify the long -run equilibrium of the
housing market and denote the long-run equilibrium housing stock by H*. Insert a short-run
housing supply curve for period 0 in your diagram, assuming that the housing market starts
out in period 0 with a housing stock H0 < H*. (Hint: remember that in the short run the housing
stock is predetermined. What does th is imply for the slope of the short-run housing supply
curve?) Provide a graphical illustration of the adjustment to a long-run housing market equi·
librium. G ive an intuitive verbal description of the adjustment process. How does the level of
housing investment I~ evolve during the adjustment to long-run equilibrium?
4. Use the model (53) - (55) to derive expressions for the long-run equilibrium values of the
housing price and the housing stock. Comment on the expressions.
5. Suppose now that the government permanently raises the property tax rate r . Use your
expressions from Question 4 to derive the long-run effects on pH and H of a marginal increase
in r . Use a d iagram like the one you constructed in Question 3 to illustrate the effects of the
property tax increase in the short run (where the housing stock is predetermined) and in the
long run when the housing stock has fully adjusted to the tax increase. Illustrate the gradual
adjustment of the market to the higher property tax and provide a verbal explanation of the
adjustment process.
6. Suppose next that the government permanently reduces the income tax rate m. How will this
affect the housing market in the short run and in the long run? G ive a graphical illustration and
explain your results. In particu lar, explain how the effects depend on the policy parameter d.
7. Consider again a cut in the income tax rate m, but suppose now that the tax code allows prop-
erty taxes as well as expenses on housing repair and all interest expenses to be deducted
from taxable income. In these c ircumstances, how w ill the income tax cut affect the housing
market? Explain your finding. (Hint: start by considering how the consumer budget constraint
must be modified to allow for deductibility of expenses on property taxes and housing repair.
Then use the revised budget constraint to derive a revised expression for housing demand
and the housing price.)
You are now invited to simulate the housing market model consisting of Eqs (53) - (55) on
the computer (you may use an Excel spreadsheet, fo r example). Your first sheet should allow
you to choose the values of the parameters I'J, {3, k and <l. We suggest that you set I'J = 0.3,
{3= 0.9, k = 1 and (5 = 0.02. Your second sheet should contain specified time paths for the
exogenous variables P1, Y1, m 1, r 11 d 1 and r 1, say from period 0 to period 100. We propose that
you choose P = 1, r = 0.05 and Y = 1 00 for all periods and that you assume m 0 = 0.4, d 0 = 1,
and r 0 = 0.015 in period 0. You should assume that the housing market starts out in a long-
run equilibrium in period 0 (and calcu late the corresponding value of H_1 = H0 ) . Your subse-
quent sheets should draw d iagrams of the simulated values of pH, JH and H (and possibly their
respective relative rates of change, for further information).
8. Use your calibrated simulation model of the housing market to s imulate the effects of the tax
policy experiments defined in Questions 5 and 6 above over 100 periods. Check whether
your simulation analysis accords w ith your previous theoretical and qualitative analysis.
Exercise 5. The housing market and the business cycle
To answer Question 8 in the previous exercise, you implemented a simulation model of the
housing market on your computer (if you haven't gone through Exercise 4, you can still use the
instructions above to construct your model and proceed w ith the present exercise). As you
recall, the model consisted of the following equations:
P H){J/(1 -{J)
Construction: jH = k, __!__ 0 </3< 1, (56)
t ( P, '
H 17( 1 - m,)Y1
Housing demand: Pr = [r ( 1 - d m ) + (5 + r ]H ' 0<17 < 1. (57)
1 1 1 1 1
Dynamics of housing stock: O <o< L (58)

prices
© The McGraw-Hill
Companies. 2005
acroeconomics Model
464 PART 5 : THE BUI LDING BLOCKS FOR THE SHORT-RUN MODEL
You are now asked to use your calibrated simulation model as a basis for discussing how
fluctuations in aggregate economic activity, Y,, are likely to affect the housing market.
1. Suppose that q = 0.3, /1 = 0.9, k = 1, 0 =0.02, m = 0.4, d = 1 and r = 0.015 in all periods.
Furthermore, let the subscript ' o' denote the initial period and suppose P0 = 1, Y0 = 100 and
r0 = 0.05. Do you find that this calibration of the model produces a plausible initial equili·
brium? (Hint: what is a plausible ratio of housing investment, JH, to total GOP, Y? What is a
plausible share of housing costs in consumer budgets?)
2. This question asks you to simulate the effects on the housing market of a stylized business
cycle, using the parameter values specified in Question 1. For the moment we model the
business cycle as a cycle in Y,, keeping all other exogenous variables in the model constant
at the levels assumed for the initial period. Starting with the initial value of aggregate income
Y0 = 100, suppose that total income assumes the following values over the subsequent
periods: Y1 = 101, Y2 = 103, Y3 = 105, Y4 = 103, Y5 = 101 , Y6 = 99, Y7 = 97, Y8 = 95, Y9 = 97,
Y10 = 99, Y11 = 100 and Y, = 100 for all t ;;> 12. Describe and comment on the way the housing
market reacts to the business cycle in your simulation. Is housing investment a lot more
volatile than GOP (Y) as the empirics in Table 15. 1 indicate?
3. Central banks usually try to drive up real interest rates in response to a rise in economic
activity, and vice versa. Suppose therefore that, as a result of the time path of Y, assumed in
Question 2, the real interest rate displays the following tirre path: r 1 = 0.05, r2 = 0.051,
r 3 = 0.052, r 4 = 0.051 , r 5 = 0.05, r 6 = 0.049, r7 =0.048, r8 = 0.046, r 9 = 0.047, r 10 = 0.049,
r 11 = 0.05, r 1 = 0.05 for t;;> 12. Investigate and explain how these interest rate dynamics
moderate the effects of the cycle in Y, on the housing market.
4. In practice construction costs usually move in a procyclical manner. Discuss why this is likely
to be the case. Against this background, suppose for concreteness that the construction cost
P, varies in the following way, reflecting that P, lags a little behind the cycle in Y,: P1 = 1,
P2 = 1.0 1, P3 = 1.02, P4 = 1.02, P5 = 1.01 , P6 = 1, P7 = 0.99, P8 = 0.98, P9 = 0.98, P10 = 0.99,
P,, = 100, P,= 100 fort;;> 12. To focus on the role of the dynamics of construction costs, let
the real interest rate be constant over time at its initial level r=0.05. Plug the time series for
P, into the model and investigate how it modifies the effects of the cycle in Y, on the housing
market. Explain your findings.
16. Consumption, income
and wealth
© The McGraw-Hill
Companies. 2005
lCD
Chapter
Consumption, income
and wealth
rivate consumption is by far the largest component of aggregate demand for goods
P and services. Although Chapter 14 showed that consumption is less volatile than
investment. changes in the propensity to consume are often the dominant source
of changes in total demand. simply because private consumption is typically around three
times as large as private investment. A theory of private consumption is therefore an
essential building block in any theory of aggregate demand.
Studying private consumption '<Viii not only help us to build a short-run model of the
business cycle. Since consumption is a basic determinant of economic welfare. our theory
of the link between consumption and other macroeconomic variables will also help us
understand how business cycles and economic policies aflect consumer welfare.
Moreover, our theory of consumption '<Viii imply a theory of saving, because saving is
equal to income minus consumption . Since saving is the basis lor capital accumulation.
our analysis of consumption is also relevant for the theory of economic growth presented
in Book One.
In the section on housing demand in the previous chapter we studied how the con-
sumer allocates his total consumption between housing consumption and consumption
of non-durables within a given time period. This ch apter complements the previous one by
analysing how the consumer will wish to allocate his total consumption over time. Since
we now wish to explain aggregate consumption, this chapter '<\Till not elaborate on the pre-
vious chapter's analysis of the compositiol'l of consumption. Thus we will treat consump-
tion as a single aggregate which. of course, must be thought of as a bundle of commodities.
including housing services.
We 11vill start this chapter by briefly restating and discussing the simple Keynesian
theory of private consumption. We '<Viii then introduce a richer model of consumption to
illustrate how consumption is linked to income, wealth and interest rates. In the final part
of the chapter we will show how our theory of consumption can be used to analyse the
ellects of the government's tax and debt policies on aggregate demand. 1
1. This chap1er borrows heavily from the t eaching note by Henrik Jensen, 'Mikrofundament for konventionelle makro·
adlaerdsrelationer', Kebenhavns Universitets 0konomiske lnstitut, Marts 1996. We are grateful to Henrik for the
inspiration, but of course he should not be held responsible for any shortcomings in our exposition.
465
Sorensen-Whitta-Jacobsen: I Part 5- The Building 16. Consumption, income © The McGraw-Hill
Introducing Advanced Blocks for the Short-run and wealth Companies. 2005
466 PART 5: THE B UILDING BLOCKS FOR THE SHORT- RUN MODEL
16.1 T.?..~. ~?..~.~.~1.~P~~?..~..f~~~~~.~.?.~..................................................................................................

The simple Keynesian consumption function
In his famous General Theory of Employrne11t, Interest and Money. published in 19 36. John
Maynard Keynes \>VTote that' ... the propensity to consume is a fairly stable function so
that, as a rule, the amount of aggregate consumption mainly depends on aggregate
income .. .'.In other words, Keynes argued that real private consumption during period t,
denoted by C1, is mainly determined by real disposable income Y;1 during that period. In
formal terms. C1 = C(Y;1) . Keynes went on to argue that:
The fundamental psychological law, upon which we are entitled to depend with great con·
fidence both a priori from our knowledge of human nature and from the detailed facts of
experience, is that men are disposed, as a rule and on the average, to increase their con·
sumption as income increases, but not by as much as the increase in their income.
The claim made by Keynes in this passage is that the marginal propensity to consume,
=
C' dC,/dYf, is positive but less than 1. In a subsequent passage he asserted that
... it is also obvious that a higher absolute level of income will tend, as a rule, to widen the gap
between income and consumption. For the satisfaction of the immediate primary needs of a
man and his family is usually a stronger motive than the motives towards accumulation, which
only acquire effective sway when a margin of comfort has been attained. These reasons will
lead, as a rule, to a greater proportion of income being saved as real income increases. 2
Thus Keynes believed that the average propensity to consume, C,/Y;'. \o\Jill decrease with
the level of income. In other words, the rich are assumed to have a higher average savings
rate than the poor.
A consumption function with all of these Keynesian properties is the simple linear
one:
C1 = a + bY;'. n > 0, O<b<l. (1)

In this consumption function the marginal propensity to consume is the constant b. which
is less than 1, and the average propensity to consume is CJY1 = b + n/Y;' which is
obviously decreasing \oVith income.
The consumption function (1) has the virtue of being simple, but there are at least
two problems with it. The Hrst problem is theoretical: although it seems plausible that
consumption is positively related to current income, it is not clear why the current con-
sumption of an optimizing consumer should depend only on current income and not also
on. say. expected future income and the real rate of interest. Thus one must doubt
whether a consumption function like (1) is consistent \oVith optimizing behaviour.
The second problem is empirical: although microeconomic cross-section data on the
relationship between consumption and income for different families within a given period
2. This quotation and the two previous ones can be found on pp. 96-97 of John Maynard Keynes, The General Theory
of Employment, Interest and Money, London and Basingstoke, Macmillan Press, 1936. These brief quotations do not
do full justice to Keynes' theory of consumption. He did in fact discuss a host of other factors likely to influence
private consumption. Still, it is fair to say that as a first approximation, he belie,ed that current consumption depends
mainly on current income.
and wealth
© The McGraw-Hill
Companies. 2005
lCD
16 CONSU MPTIO I\ , INCO ME AN D W EALTH 467
c
0
·a
E
::J
"'c
8
Di sposable income
Figure 16.1a: Stylized relationship between income and consumption in microeconomic cross-
section data
Disposable income
Figure 16.1b: Stylized relationship between income and consumption in macroeconomic time
series data
do indicate that the rich save a larger fraction of their current income than the poor.
macroeconomic time series data for most countries indicate th at the ratio of aggregate
consumption to aggregate income is roughly constant over the long run . These appar-
ently contradictory stylized facts are illustrated in Fig. 16.1, where Fig. 16.la indicates
the relationship between income and consumption for d~{ferent households within a given
period such as a year. while Fig. 16.1 b indicates the relation between aggregate household
income and aggregate consumption over tirne.
The rough long-run constancy of the average propensity to consume is illustrated for
the USA and Denmark in Fig. 16.2 . Apart from a temporary drop due to consumption
rationing during the Second World War. the average propensity to consume in the US has
Introducing Advanced Blocks for the Short- run and wealth Companies. 2005
468 PA RT 5 : THE B UIL D ING BLOC KS FO R THE S H O RT- RU N MO DEL
1.2
Q)
<U
QlQ)
~ E
OlO
QlU
.S
"'
--o 0.8
go
-~~
E
a.Ul
=>
::J 0 0.6
Ul ~
c
0~
U.D
Q) Ul
<U 0 "' 0.4
Ol a.
Q) Ul
0,:.0
Ol
<( 0.2
0
0>
"'
0>
(')
(')
0>
"'
(')
0>
;
a>
"''<I'a> 0>
'<I'
0>
(')
"'a> "'"'
0>
- "'
(Jj
0>
(Jj
0>
0>
(Jj
0>
(')
"' "'"'
0> 0>
-
(I)
0>
"'
(I)
0>
0>
(I)
0>
(')
0>
0>
"'
0>
0>
0
0
Year "'
Figure 16.2: The average propensity to consume in USA and Denmark
Source: National Income Accounts, Bureau of Economic Analysis and A DAM database, S tatistics Denmark.
been remarkably stable over the long run, despite the tremendous growth in income since
1929. The Danish propensity to consume has been more volatile. but without any
systematic trend. Figure 16.2 clearly contradicts Eq. (1) which implies th at the ratio of
consumption to income should decline over time as income grows. Instead ofEq. (1). we
therefore need a theory of consumption which bas a solid theoretical foundation and
which is able to explain why we observe dill'erent relationships between consumption and
income in microeconomic cross-section data and in macroeconomic time series data. In
the rest of th is chapter we shall try to build such a theory.
Consumer preferences
The starting point for a micro-based theory of consumption is a specification of consumer

preferences. Consider a consumer who plans for a certain finite time h orizon. We wUI
divide this time interval into two periods which may be thought of as 'the present' (the
current period 1) and 'the future' (period 2). The limitation to only two periods is just a
simplification: one can show that all our qualitative conclusions will continue to hold in a
setting with many periods.
In each period t (t = 1, 2) the consumer derives utility u(C1) from consumption.
However, because the consumer is 'impatient', he or she prefers a unit of utility today to a
unit of utility tomorrow. When evaluated at the beginning of period 1. the consumer's
l!fetime utility lJ is therefore given by:
u(C, )
U = u(C ) +- -- 11
1
> 0. ll" < 0, ¢ > 0. (2)
J 1 +¢ '
The consumer's impatience is captured by the parameter¢ which is referred to as the rat.e
I Part 5 - The Building
and wealth
© The McGraw-Hill
Companies. 2005
lCD
16 CONSUMPTIO I\ , INCO ME AN D W EALTH 469
of time pre.ference. The positive rate of time preference means that a given amount of con-
sumption today is valued more highly than a similar amount of consumption tomorrow.
On the other hand. as long as cp is not infinitely high. the consumer is not indifferent about
the future. and he or she will then have to decide how to allocate his or her consumption
optimally over time. The assumptions u' > 0 and u" < 0 reflect that the marginal utility of
consumption in any period is positive but decreasing. An increase in consumption in any
period will thus reduce the marginal utility gain from a further consumption increase in
that period.
Our theory of consumption will be derived from the assumption that the consumer
trades off present against future consumption so as to maximize the lifetime utility func-
tion (2). The terms of this trade-off will depend on the consumer's intertemporal budget
constraint to which we now turn.
The intertemporal budget constraint
When specifying the consumer's budget constraints for the two periods. we will assume
that capital markets are pe1j'ect. This means that the consumer can freely lend and borrow
as much as he or she likes at the going market rate of interest. In practice some consumers
may face credit constraints preventing them from borrowing as much as they would have
preferred at the going interest rate. We will discuss credit constraints later in the chapter,
and Exercise 3 asks you to consider their implications in detail. but for the moment we will
assume perfect capital markets.
We specify the consumer's budget constraints in real terms. At the beginning of
period 1 the consumer is endowed with a predetermined stock of real financial wealth V1 .
During period 1 he or she earns real labour income Y~ , pays the real amount of taxes T1 ,
and spends the real amount C1 on consumption. For convenience we assume that all
payments are made at the beginning of the period. 3 Aller having received this income
and incurred this expenditure on taxes and consumption. the consumer has an amount
V1 + Y~- T1 - C1 left over for investment in interest-bearing financial assets. If the real
interest rate is r, the consumer will therefore end up with a real stock of financial wealth
V2 = (1 + r)(V 1 + Y\-- T1 - C1) at the beginning of period 2. Hence the budget constraint
for period 1 is:
(3)
Note that V2 may well be negative. In that case the consumer is a net borrower during
period 1. Since be or she does not plan any consumption beyond period 2, the consumer
will simply spend all his or her resources during that period, including the financial
wealth accumulated during period 1. Using the same notation as before. we may therefore
write the budget constraint for period 2 as:
(4)
3. If some or all payments were made at the end of the period inst ead, we would obtain a consumption function w ith the
s ame qualitative properties as those described below. However, the assumption that payments take place at the start
of each period leads to slightly more elegant analytical expressions. Whenever we divide the time axis into discret e
finit e inter•tals, there is no objectively 'correct' assumption on the timi'lg of payments (beginning·of·period versus
end ·of·period). Hence we are free to choose the assumption on timing that is most convenient for analytical
purposes.
470 PA RT 5 : THE B UIL DING BLOC KS FO R THE S H O RT- RUN MO DEL
Equation (4) states that the consumer will spend his or her initial financial wealth plus the
after-tax labour income on consumption during period 2. Notice that if he or she bas
borrowed during period 1 so that V2 < 0, the consumer must reserve part of the period 2
labour income lor debt repayment.
It will turn out to be convenient to consolidate (3) and (4) into a single constraint. We
therefore use (3) to eliminate V1 from (4). divide through by (1 + r) on both sides of the
resulting equation and rearrange to get:
(5)
Equation (5) is the consumer's intertempora/ budget constraint. It states that the present
value of the consumer's lifetime consumption (the left-hand side) must equal the present
value of his or her after-tax labour income plus the initial financial wealth (the right-hand
side).4 In other words. with a perfect capital market current consumption does not have to
equal current income, but over the life cycle the consumer cannot spend any more than
his or her total resources. These resources consist of labour income and the initial
llnancial wealth.
We can write (5) in a simpler form by introducing
H = yt _ T yi. - T,
+ -z__- (6)
1 1 t
1+r
The variable H1 is the present value of the consumer's disposable lifetime labour income,
and it is referred to as human wealth or hwnan capital because it measures his or her
capitalized earnings potential in the labour market. Note that H 1 carries the time subscript
1 because it includes labour income from period 1 and onwards. Inserting (6) into (5), we
get:
(7)
Hence the consumer's intertemporal budget constraint simply states that the present
value of real lifetime consumption is constrained by total real initial wealth, consisting of
lhc:: sum of Dmuu;ial wealth aud h uman wc::alllJ .
We shall now study how consumers will actually wish to allocate consumption over
time.
The allocation of consumption over time
The consumer chooses hL~ or her time path of consumption so as to maximize the lifetime
utility function (2) subject to the intertemporal budget constraint (7). The consumer's
wage rate is taken as given, and we assume that his or her working hours are institution-
ally determined. say, by collective bargaining agreements or by law. This means that the
4. O f course, the consumer could choose to consume less than his or her t otalli'etime resources, but since an increase
in consumption today or tomorrow w ill always inc rease lifetime utility, he or she w ill always choose to consume as
much as the budget constraint permits. This is why (5) is w ritten with an equality sign rather than an inequality sign.
and wealth
© The McGraw-Hill
Companies. 2005
lCD
16 CONSUMPTIOI\ , INCOME AND WEALTH 471
labour incomes Y~ and Y~ and thereby H1 are exogenously given to the consumer. 5 Since
the market value of financial assets is also beyond the control of the individual consumer,
it follows that the consumer's total initial wealth, V1 + H1, can be taken as given when
optimizing consumption. Because we are mainly interested in studying the determinants
of current consumption C1 , we will use the intertemporal budget constraint (7) to
eliminate C2 from the lifetime utility function (2). Doing that. we obtain:
(8)
The consumer's problem then boils down to choosing C1 so as to maximize (8). The first-
order condition for a maximum is:
which we can write as:
(9)
or as:
(10)
Equations (9) and (10) are just two alternative ways of writing the consumer's optimum
condition. Consider first the interpretation of (9). If the consumer increases consumption
by one unit in period 1. his or her liletime utility will increase by u'( Ct). If he or she chooses
instead to save an extra unit and invest the timds in the capital market to earn the real
interest rater, the consumer will be able to increase consumption in period 2 by 1 + r units.
This will generate an increase in lifetime utility equal to 11 : ; u'(C 2) . According to (9) these
hvo alternatives must be equally attractive. In other words, in optimum the consumer
must be indillerent between consuming an extra unit today and saving an extra unit today.
Equation (10) is a version of the usual optimum condition that the consumer's
marginal rate of substitution, MRS. between any two goods must equal the price ratio
between the two goods. In this case the two goods are 'present consumption·. C1 , and
'future consumption', C2 • The optimum condition (10) is illustrated in Fig. 16.3 .
To interpret the diagram. note that along any consumer indifference curve liJ'etime
utility is constant. According to the lifetime utility function (2) a constant utility level
implies:
or:
_ dC2 = u'(C 1)
(11)
dC 1 u'(C 2)/(1 + ¢>) ·
5. In Exercise 2 you w ill be asked to study the det ermination of consumption in the more complicated case where the
consumer can choose his or her working hours and hence labour incone.
Equation (11) shows that the numerical slope of the indi!Terence curve is equal to the
marginal rate of substitution between present and future consumption. MRS(C2 : C1),
defined as the ratio of the marginal lifetime utility of present consumption u'(C 1) to the
marginal lifetime utility of future consumption u'( C2 )/(1 +¢). Figure 16.3 also shows that
the numerical slope of the consumer's lifetime budget constraint equals 1 + r: if the con-
sumer gives up one unit of present consumption, 1 + r additional units of future con-
sumption will result because his or her saving earns the real interest rater. Hence we can
say that 1 + r is the relative price of present conswnptio11, since it measures the amount of
future consumption which must be given up to enable the consumer to increase present
consumption by one unit. When the consumer attains the highest possible level ofliletime
utility consistent with his or her lifetime budget constraint, we see from Fig. 16.3 that the
slope of the indifference curve must equal the slope of the budget constraint, given by the
relative price of present consumption. This is exactly what Eq. (10) says.
The optimum consumption rule (10) is called the Key11es-Ramsey rule after its dis-
coverers. Although formally derived by Cambridge economist Frank Ramsey in 1928. 6
the rule had been verbally anticipated by the very same John Maynard Keynes who later
(in his General Theory of 1936) came to believe in a consumption function like (1) ! As we
will discuss later in this chapter {and in Exercises 1 and 3), lor some households current
consumption is indeed likely to depend only on current income, as postulated in the
consumption function (1). But we shall also see that the consumption function implied by
the Keynes-Ramsey rule (10) is consistent with several empirical observations on con-
sumption which cannot be reconciled with the simple Keynesian consumption function.
One important implication of (10) is that consumers will typically want to use the
capital market to smooth their consumption over time. This is seen most clearly in the
O ptimal intertemporal
allocation of consumption
saving
v, + yf - T1 V1 +H1 c,
Figure 16.3: The consumer's optimal intertemporal allocation of consumption
6. Frank Ramsey, 'A Mathematical Theory of Saving', Economic Journal, 38, 1928, pp. 543-559. This was one of two
path-breaking articles published by the unusually gifted Frank Ramsey before his premature death at the age of 25.
benchmark case where the real interest rate equals the rate of time preference (r = ¢) so
that the consumer's impatience¢ is exactly oflset by the capital market reward r for post-
poning consumption. When r = ¢. condition (1 0) can only be met if C1 = C2 , that is if
consumption is constant over the consumer's life cycle. Suppose for simplicity that the
consumer starts out in period 1 INith zero financial wealth. V1 = 0 . Unless the consumer's
labour income happens to be the same in periods 1 and 2, he or she will then h ave to
engage in financial saving in one period and financial dissaving in the other period to keep
consumption constant over time.
Figure 16.4a shows the case where labour income in period 1 L~ lower than labour
income in period 2. The consumer will then want to borrow in period 1 to smooth con-
sumption over the life cycle even though this means having to reserve part of the higher
future labour income for payment of interest. Figure 16.4b shows the alternative case
where future labour income is lower than current labour income (perhaps because the
consumer plans to retire some time in period 2). In this case the consumer will want to
save part of his or her current labour income and will partly finance period 2 consumption
out of preYious savings.
If there were no capital market allowing saving and dissaving, consumption would
have to equal income within each period. If income L~ low in one period and high in
another, the marginal utility of consumption would then difTer across periods, because
marginal utility decreases with increasing consumption. The consumer 10\lill therefore
enjoy a welfare gain to the extent that he or she is able to smooth consumption by bor-
rowing or saving through the capital market. Capital markets enable consumers to decou-
ple current consumption from current income, and utility-maximizing consumers will
typically want to take advantage of that in order to spread their consumption more evenly
over time. Although total lifetime consumption is constrained by total lifetime resources
(wealth), we should not necessarily expect to observe a close link between current
consumption and current income.
------------------
Saving in period 2
Borrowing in period 1
Figure 16.4a: Consumption smoothing for a consumer with relatively low income during period 1
{V1 = 0)
Saving in period 1
Dissaving in period 2
----------------- -+-----------1
0 2
Figure 16.4b: Consumption smoothing for a consumer w ith relatively high income d uring period 1
(V, =0)
I
Do consumers actually smooth their consumption over their working life? A glance
at the upper two curves in Fig. 16.5 might suggest that the answer is 'No!'. The figure
tracks data for the average consumption-age proHle and the average income-age prollle
for a cohort of British married couples. The figures for income and consumption have
been deflated by the consumer price index and transfonned into logarithms. The age
of the household is identified using the age of the female household member. We see
that both consumption and income follow a hump-shaped pattern over the life cycle.
peaking a little before the age of 50. In particular, current consumption seems to follow
current income fairly closely. in apparent contrast to the hypothesis of consumption-
smoothing.
However. Martin Browning and Mette Ejrnres of Copenhagen University have
demonstrated that if one corrects the consumption figures for the systematic impact of dif·
ferences in the number of children across households, consumers do actually tend to
smooth their consumption over time. 7 This is illustrated by the bottom curve in Fig. 16.5
which shows 'adjusted consumption', that is, the level of consumption adjusted for the
estimated impact of the numbers and ages of children on the consumption needs of the
household. We see that the adjusted consumption-age prollle is much flatter than the
income profile, suggesting that consumers do indeed prefer to smooth consumption per
household member. in accordance with our theory.
The determinants of current consumption
The Keynes-Ramsey rule (10) only provides an implicit solution to the consumer's
problem. To derive an explicit analytical solution for current consumption Cl' we need to
7. See Martin B rowning and Mette Ejrnaes, 'Consumption and Children' , Worl<ing Paper, Institute of Economics,
University of Copenhagen, February 2002.
- 1.6 - - Income
Consumption
- - Adj usted
consumption
- 1.8
-2
- 2.2
~-------------r------------,-------------,-
30 40 50 60
Age
Figure 16.6: C onaump tion and income profilea for houaeholda in the Un ited Kingdom (born
1935-1939)
Source: Data are from the UK Family Expenditure Survey, kindly provided by Mette Ejrnaes.
specify the form of the consumer's utility function . Onespecillcation which has often been
used in economic research is the following:
u(C) = _!!_ c<u- 1)/n

, a- 1 , tor o > o. * 1. (12)
tor (1 = 1. (13)
As you may easily verify, this specil1cation satistles the assumptions in (2) that u' > 0 and
u" < 0 . To interpret the parameter <1 in (12), we introduce the concept of the intertemporal
elasticity ofsubstitution in consumption. IES, detlned as the percentage change in the ratio of
future to present consumption (C2 /C 1) implied by a one per cent change in the consumer's
marginal rate of substitution. MRS(C2 : C1) :
(14)
The IES measures the degree to which the consumer is willing to substitute future for
present consumption . This is illustrated in Fig. 16.6 where we recall that MRS(C2 : C1)
measures the numerical slope of the consumer's indillerence curve. The slope of the
straight line from the origin through point E0 measures the consumption ratio (C~/CV)
prevailing when the marginal rate ofsubstitution equals MRS( C~ : C~). As we move up the
indillerence curve from point E0 to point Ep thereby raising the consumption ratio from
(C~/C?) to (CI/CD. the consumer becomes less willing to trade present consumption for
future consumption. This is reflected by the increase in the marginal rate of substitution,
that is, by the fact that the numerical slope of the indillerence curve is steeper at point E 1
than at Ell' It is obvious that, for any given increase in MRS(C2 : C1) . the rise in the
476 PART 5 : THE B UIL DING BLOCKS FO R THE SHORT-RUN MODEL
c~
cg -------
Figure 16.6: The relation between the consumption ratio (C 2/ C 1) and the marginal rate of
substitution
consumption ratio (CiC1) will be greater the flatter the indilrerence curve. that is. the
easier it is to substitute future for present consumption. In other words. the more the
consumer is willing to engage in intertemporal substitution in consumption. the greater
will be the value ofiES defined in (14).
Since u'(C,) = c;tfo, we may now write the MRS. u'(C1)/(u'(C2)/(l +¢)),as:
MRS(C 2 : C1) = (1 + ¢)
(c: ,
c )11"
and hence:
from which:
(15)
The utility function (12) thus has the property that the intertemporal elasticity of sub-
stitution is constant and equal to o. By varying <1 we may therefore study the ellects of
variations in the consumer's willingness to substitute consumption over time. As o tends
to zero. the consumer becomes quite unwilling to trade present for future consumption,
and his or her indillerence curves become rectangular, as shown in Fig. 16.7. On the
Indifference curves
o =O
I 0 <a< 1
c,
Figure 16.7: The relation between the shape of the indifference curve and the intertemporal
substitution elasticity
other hand, as a tends to infinity. substitution possibilities also become infinite, and the
indillerence curves converge to straight lines.
We are now ready to derive the consumption function. Substituting the expression
(1 + ¢)(C2/C1) 1111 lor MRS(C2 : C1) into the condition 1\11RS(C2 : C1) = 1 + r for an optimal
intertemporal allocation of consumption, we get:
C, = (1 + r)" cl . (16)
- 1+¢
This may be inserted into the lifetime budget constraint C1 + C2/(1 + r) = V1 + H 1 to give
c)+ (1 + ly- L(l + t/>) -"CI - VL +HI. implying:
1
0<8= I <1. (17)
1+(1+r)" (1+¢t"
Equation (17) is seen to be rather diflerent from the simple Keynesian consumption
function (l). According to (17) current consumptiOI'I is proportional to current wealth . The
propensity to consume out of current wealth (B) is positive but less than 1. and its mag-
nitude will depend on the real rate ofinterest (in a manner to be studied in detail below).
16.2 T.P:.~. P~?.P~.~~·~·~·~·gf..~9.:~.gg~~~~P~.~g~. f.~.~g~~~.~................................................. .

Consumption and income
Let us now explore the implications of (17) for the relationship between current
=
consumption C1 and current disposable labour income. Y'i Y\'- T1 • If we insert our
definition of human wealth (6) into (17), we get:
t = 1, 2. (18)
This may be rewritten as:
cl = liY1 . (19)
l+r =--=.
y~
e=e(l +~+vl)· R yd 'I
(20)
whereB measures the propensity to consume (out of) current income. R is the ratio of
future to current income, and v 1 is the current wealth- income ratio. On the basis of (19)
and (20) we may now explain the empirical puzzle that the average propensity to
consume current income seems to decrease with income when we consider micro-
economic cross-section data whereas it seems to be roughly constant when we consider
macroeconomic time series data.
Consider tirst the observed cross-section relationship between individual household
income and individual consumption suggesting that the rich h ave a higher average
savings rate than the poor. Within any given time period. some of those individuals who
record a high current income level cannot expect to earn similar high incomes in the
future. One important reason is that workers must some day retire from the labour market
in which case they will no longer earn income from labour. Moreover, for self-employed
individuals a high current income will sometimes reflect that business has been unusually
good during that year. Thus, if a high current income level is expected to be ten1porary. the
ratio of future to current income (R) will be relatively low, and according to (20) this will
imply a low average propensity to consume current income. In a similar way, some
consumers with low current incomes may have good reason to expect that they will earn
more in the future. This will typically be the case for university students and other young
people who have not yet realized their earnings potential in the labour market. and for
entrepreneurs who have just started up a business or who experience unusually bad
business during the current year. For such individuals the ratio R in (20) will be high, and
hence they will have a high propensity to consume out of their low current income. If
some of the observed variation in income levels in a cross-section of consumers retlects
such temporary factors, Eq. (20) may thus explain why the average propensity to
consume appears to fall as income goes up.
In a celebrated article. Franco Modigliani and Richard Brumberg provided the first
statement of the so-called life cycle theory of consumption according to which consumers
in di!Jerent stages of the lile cycle will have different propensities to consume current
income because of their desire to smooth consumption over time. 8 In a similarly famous
contribution, Milton Friedman developed the so-called permane11t income hypothesis which
says that transitory changes in income will mainly lead to temporary changes in savings,
whereas current consumption will depend on the consumer's permanent income, that is,
8. See Franco Modigliani and Richard Brumberg, ' Utility Analysis and the Conw mption Function: An Interpretation of
Cross-Section Data', in Kenneth K. Kurihara, ed., Post-Keynesian Economics, pp. 388-436, New Brunswick, NJ,
Rutgers University Press, 1954.
his expected long-run average income (which is proportional to his total wealth).9 Thus
these \<\TI'iters pointed out that some of the high incomes observed during a given year are
only temporarily high and hence will not induce high levels of current consumption, and
some of the low incomes recorded in a given period are just temporarily low and thus will
not cause similarly low levels of consumption. Because of this, a cross-section analysis of
the relationship between current consumption and current income in any given year will
give the impression that high-income people have a lower average propensity to consume
than low-income people.
Let us next see how Eqs (19) and (20) may explain the empirical observation that the
average propensity to consume is roughly constant in the long run when we look at
aggregate time series data. If we denote the growth rate of real income by g. we have
=
Y1 (1 + g) Y'l which may by inserted into (20) to give:
. 1+g
8 = 1 + - - + v 1. (21)
1+r
While the growth rate g fluctuates in the short run, on average it h as been fairly constant
over the long run, as we saw in Fig. 2.8. Moreover, the real interest rate r shows no
systematic long-run trend as we also saw in Fig. 2.10. and over the long run Chapter 2
found that the capital stock tends to grow at the same rate as income so that the long-run
=
wealth- income ratio v 1 V 1/Y1 is roughly constant. It then fo llows from (21) that the
long-run average propensity to consume will also tend to be constant, as we do indeed
observe.
To sum up. our reconciliation of the microeconomic cross-section data and the
macroeconomic time series data runs as follows: a temporary increase in income for an
individual consumer will reduce the expected values of the parameters R and g in (20) and
(21), and will probably also reduce his or her short-run wealth- income ratio v 1• For these
reasons the average propensity to consume will tend to fall with rising income levels in a
cross-section of consumers. Over the long run, the average growth rate of income across
all consumers is roughly constant, and wealth moves roughly in line with income. From
(21) this implies a constant long-run average propensity to consume at the macro level.
Consumption and wealth
While the average propensity to consume current income tends to be constant in the long
run, it fluctuates quite a lot in the short nm. as indicated in Fig. 16.8. Equation (21) sug-
gests three possible reasons lor this:
1. short-run changes in the expected growth rate of income (g).
2. short-run changes in the wealth- income ratio (v 1) . and
3. short-run changes in the real rate of interest.
According to (21) a higher expected income growth or a rise in the market value of
flnancial wealth relative to current income will increase the propensity to consume. Note
9. Milton Friedman, A Theory of the Consumption Function, Princeton, NJ, Princeton University Press, 1957. Not e that
t he consumer's stock of total wealth equals the discounted value of hi5 expected future income. Permanent income
is that hypothetic al const ant level of income w hich has the same present value as the consumer's expected future
income stream.
480 PART 5: THE B UILDING BLOCKS FO R THE SHORT- RUN MODEL
1.1 6.5
Average propensity to consume
(left axis)
Q)
1.05
E
::;)
6
"'
c N
0
8 ·~
£ 5.5 Q)
.~ E
0
"'c 0.95 0
Q) c
a. 5 T
c. £
0
0.9 (;j
~
Q)
Ol
"'Q; 0.85
4.5
~
Figure 16.8: Private consumption and wealth in Denmark

I Raliu or privale t:un~urnpliun to dispusabl~ iru;um~.
2
Ratio of private wealth to disposable income.
Source: ADAM database, S tatistics Denmark
that changes in g and v 1 will often go hand-in-hand. since a higher expected income
growth is likely to drive up the market prices of stocks and owner-occupied housing,
thereby increasing V1 , because a higher expected growth rate will tend to raise expected
future corporate dividends and to drive up the demand for housing.
Figure 16.8 shows that there is indeed a fairly close empirical relationship between
the wealth-income ratio and the average propensity to consume. as our theory of
consumption would lead us to expect. 10
Consumption and interest rates
The third determinant of the propensity to consume current income is the real interest
rate r. The Interest rate afl'ects consumption through three different channels. First of all,
it influences the propensity B to consume out of wealth. Second. it affects the market value
ofHnancial wealth. Third. the real interest rate also affects the value of human wealth. We
will consider each of these channels in turn.
In (17) the propensity to consume wealth was deHned as:
1
e= . (22)
1 + (1 + r)"- 1(1 + </>t"
1 0 . It is worth not1ng that even though he did not include financial wealth as an explanatory variable in his basic
consumption function, Keynes was fully aware of the potential influence of wealth on private consumption. On
pp. 92-93 of his General Theory he wrote: 'W indfall changes in capital-values ... are of much more importance in
modifying the propensity to consume, since they will bear no stable or regular relationship to the amount of income.
The consumpti:>n of the wealth·owning c lass may be extremely susceptible to unforeseen changes in the money·
value of its wealth. This should be classified amongst the major factors capable of causing short·period changes in
the propensity to consume'.
Clearly this propensity increases with r if a< 1, and decreases if o > 1. and we cam1ot say
a priori which direction is correct as o can be either smaller or larger than 1. The indeter-
minacy in how r afiects the propensity to consume is due to offsetting substitution and
income effects. On the one hand an increase in the real interest rate raises the relative
price 1 + r of current consumption. Ceteris pari/Jus, this \>Viii induce the consumer to sub-
stitute future for present consumption through an increase in current saving. Obviously
this substiwtion effect will tend to reduce the propensity to consume in the current period.
At the same time the higher interest rate also increases the amount of future consumption
generated by a given amount of current saving. Hence the consumer can atlord a higher
level of current consumption without having to reduce future consumption. Other things
equal. this income ej)'ect of the rise in the interest rate works in favour of a rise in current as
well as future consumption and thus tends to increase 8.
Since a measures the strength of the substitution effect. it is not surprising that the
sign of the derivative d8/dr depends on the magnitude of o. We see that, for an inter-
temporal substitution elasticity equal to 1, the income and substitution effects will exactly
ollset each other, and the propensity to consume \>Viii be unaffected. Over the years many
researchers have tried to estimate the empirical magnitude of the intertemporal substitu-
tion elasticity, <1. Most auth ors have found values of <1 well below 1, suggesting that the
propensity to consume wealth \>Viii rise when the real interest rate goes up.
However, a higher interest rate will also influence the level of wealth itself. As already
mentioned, the stock of financial wealth, VL. includes the value of stock and of housing
capital. In Chapter 15 we saw that both of these important wealth components will be
negatively allected by a rise in the real interest rate. Thus, a rise in r means that expected
future dividends are discounted more heavily. leading to a fall in share prices. Moreover,
by raising the user cost of moVUer-occupied housing. an increase in r reduces housing
demand which in turn drives down the market value of the existing housing stock. For
these reasons a rise in the real interest rate will reduce the wealth- income ratio, v 1•
In addition . the higher interest rate means that expected future labour income is
discounted more heavily. As a consequence. the value of human wealth specified in ( 6)
goes down. Intuitively. a higher real interest rate mal<es future labour income less
valuable by making it easier to attain a given level of future consumption through saving
out of current income.
The fall in human and ilnancial wealth induced by a rise in the real interest rate
clearly tends to reduce the propensity to consume current income. but since we have seen
that there may be an o!Isetting increase in the propensity to consume wealth (8), the net
ellect of a rise in ron the ratio of current consumption to current income (B) is ambiguous.
Against this background it is not surprising that empirical research has found it dill1cult to
document a strong effect of real interest rates on consumption. although the dominant
view is that a rise in r tends to reduce consumption because of the negative impact on
wealth.
16.3 Consumption, taxation and public debt

................................................................................................................................................................................
We will now show how our theory of consumption can be used to derive the effects of the
government's tax and debt policies on private consumption demand. This will give us an
opportunity to highlight the importance of the way in which consumers form their
expectations about the future . It will also enable us to illustrate the important distinction
between temporary and permanent changes in tax policy.
Temporary versus permanent tax cuts
To focus on tax policy, we start by rewriting the consumption function (18) as:
(23)
Suppose now that the government wishes to stimulate consumption demand by cutting
net taxes T1 during period 1 . say. because the economy is in recession. For the moment, we
assume that consumers expect the tax cut to be temporary, perhaps because the govern-
ment has announced that it may have to raise taxes again once the recession is over. In
that case consumers \<Viii expect T2 to be unch anged. and according to (23) the immediate
impact of the temporary tax cut will then be given by the derivative:
(24)
In other words, if the government cuts taxes temporarily by one unit, current consump-
tion will go up by 8 units. 1 1 Thus the temporary tax cut will indeed succeed in stimulating
current consumption, but recalling from (17) that e < 1. we also see that part of the
increase in disposable income will be saved for future consumption. Of course. this saving
reflects the consumer's desire to smooth consumption over time.
For comparison. suppose instead that consumers expect the tax cut to be pennanent so
that T2 goes down by the same amount as T1• From (23) we then lind the effect on current
consumption to be:
oC + oC
1 1 = _ 8(1+ _1_ )· (25)
oT1 oT2 1 +r
Not surprisingly. we see from (24) and (25) that a permanent tm.: cut will have a stronger
impact on current consumption than a temporary tax cut. In the benchmark case where the
real interest r ote r equals the rate of time preference rp so that the consumer wants to
consume equal amounts in the two periods, you may easily verily from (17) that
e = ( 1 + r)/(2 + r). In that case it follows from (2 5) that:
for r=¢. (26)
In other words, when the consumer wants to smooth consumption perfectly over time, he
or she \<Viii consume all of the period 1 tax cut during that period. There is no need to save
any part of the increase in current disposable income to smooth consumption. since his or
her expected future disposable income has gone up by a similar amount as h is or her
current net income.
1 1. Note that this is a partial analysis. If the tax cut succeeds in stimulating economic activity, it w ill ind irectly cause a
further rise in consumption by raising aggregate labour income Y\. However, our focus here is on the immediate
d irect impact on consumption, for a given level of activity.
The government budget constraint
In the analysis above we have not specified whether the tax cuts are associated \>\lith
changes in public spending. As we shall argue below, the effects of a tax cut on private
consumption may depend on whether or not it reflects a cut in public expenditure. To see
this, we must make a slight detour to study the government's budget constraint. Let us
assume for the moment that the government plans for the same time horizon as the
household sector (later we wUl discuss the implications of relaxing this assumption). As
before, we will divide this time horizon into two periods representing the 'present' (period
1) and the 'future' (period 2). At the beginning of period 1 the government starts out \>\lith
a given real stock of public debt D1 which is predetermined by the accumulated historical
government budget deficits. In period 1 (at the beginning) the government spends a real
amount G1 on public consumption and collects a real net tax revenue Tp where T1
measures taxes net of government transfer payments (transfers are simply treated as
negative taxes) . U public spending exceeds tax revenues, the government must issue new
government bonds in the real amount ti D = G1 - T1 in period 1. 12 Since debt carries
interest. the government will then end up with a total amount of real debt D2 = (1 + r)
(D 1 + t.. D) at th e beginning of period 2 . Hence the government budget constraint for period
1 may be written as:
(27)
In period 2 the government must run a budget surplus T1 - G2 wh ich is suflicient to enable
it to repay the public debt accumulated during the previous period. The government
budget constraint lor period 2 is therefore given by:
(28)
Inserting (2 7) into (28) and dividing th rough by 1 + r, we obtain the intertempora/ govem-
ment budget constraint:
(29)
Equation (29) says that the present value of current and future tax revenues must cover
the present value of current and future government spending plus the initial government
debt. As we shall now see. the intertemporal government budget constraint may have
profound implications for the effects of tax policy.
Tax finance versus debt finance of government spending: the Ricardian

Equivalence Theorem
It follows directly from (29) that if the government does not reduce its current or planned
future spending - that is, if G 1 and G2 are unchanged - a tax cut dT1 < 0 in the current
period must be followed by a future increase dT2 > 0 in taxes so that:
dT2
dT1 + - - = 0, (30)
1+r
12. We abstract from money printing (seigniorage) as a way of financing government consumption, since this plays very
little role in developed coun tries.
or dT2 = - (1 + r)dT1 • To understand this relationship between current and future taxes,
consider (27) and (28) once again. According to (27). if the government cuts current
taxes by an amount IdT1 Iwithout reducing current public spending. G 1• the stock of real
public debt will have risen by the amount dD2 = (1 + r) IdT1 Iat the beginning of period 2.
If the government does not cut back on its period 2 consumption . G2 • it follows from (28)
that it will have to raise taxes in period 2 by the amount dT2 = (1 + r) IdT1 I to pay for the
principal and interest on the extra debt created by the tax cut in period 1.
Suppose now th at consumers have rational expectations in the sense that they look
forward and understand the implications of the intertemporal government budget con-
straint. In that case they will realize that if the government reduces current taxes without
reducing current or planned future public spending, the present value of future taxes '<\Till
have to increase by as much as the cut in current taxes. It then follows from (2 3) and (30)
that:
dT-
dC 1 = - 8 dT 1 + - 2 )
=0 (31)
( 1+r
The implication of(31) is striking: a cut in current taxes which is not accompan ied by a cut
in present or planned future public spending will h ave no effect on private consumption!
In other words, a switch from tax finance to debt finance of current public spending leaves
private consumption unatlected, since consumers realize that the tax cut today will be
offset by the future tax increase needed to service the additional government debt. Indeed.
because the present value of their lifetime tax burden is unchanged, consumers know that
their net human wealth is also unchanged, and hence they feel unable to afl'ord an
increase in current consumption. Instead they '<Viii save the entire current tax cut and
invest it in the capital market. This increase in current private saving will raise consumer
cash receipts in period 2 by the amount (1 + r) IdT1 I which is just sufficient to pay for the
higher future tax bill. Hence future private consumption is also unchanged. The increased
supply of government bonds in period 1 is thus matched by a similar increase in private
demand for bonds. and the higher period 2 taxes are matched by an increase in private
sector income from bond holdings.
The observation that tax llnance and debt fmance of government spending are in
principle equivalent was first made as early as 1820 by the classical Dritish economist
David Ricardo. In a treatise on public debt Ricardo discussed three alternative ways of
11nancing a war. For concreteness, he assumed that the war would generate military
expenditure of£20 million per year. One financing option would be to impose additional
taxes amoun ting to £20 million per year until the end of the war. Alternatively, the
government could borrow £20 million every year during the war and increase tax
collections by just £1 million each year to cover the interest payments on a f20 million
loan. assuming an interest rate of 5 per cent. In this case the public debt would continue
to rise until the war was over and would never be repaid, and taxes would be permanently
higher. The third possibility considered by Ricardo was one where the war was mainly
debt-11nanced, but where taxes would be raised by £1.2 million per year for every £20
million borrowed. enabling the government to pay oft' its debt in 45 years. Under the
assumptions made, the present value of tax payments would be the same under the three
modes of finance. As Ricardo put it: 'In point of economy, there is no real ditlerence in
and wealth
© The McGraw-Hill
Companies. 2005
lCD
either of the modes: for twenty millions in one payment, one million per annum for ever,
or 1.200.000 pounds for 45 years, are precisely of the same value'. 13
Because of this statement by Ricardo, the claim that taxes and debt are equivalent
methods of public finance is referred to as the Ricardian Equivalence Theorem. Before you
dismiss the theorem as being utterly unrealistic, take a look at Fig. 16.9. In that figure we
have plotted private and public saving in Denmark as a percentage of GDP against time.
As we have seen above, the Ricardian Equivalence Theorem implies that whenever the
public sector reduces its saving, thereby increasing new issues of public debt (or reducing
the rate at which public debt is retired). we should observe an o!Jsetting increase in private
saving, and vice versa. Figure 16.9 shows that there is indeed a clear tendency for private
and public saving to move in opposite directions, suggesting that Ricardian equivalence
may not be that unrealistic after all.
However, while Fig. 16.9 does not seem inconsistent with Ricardian equivalence, it
does not 'prove' that the theorem is correct, since there may be other explanations lor the
observed relationship between private and public saving. One potential Keynesian expla-
nation L~ that Fig. 16.9 simply reflects the so-called automatic stabilizers built into the
public budget: if the private propensity to save goes up for some reason. the resulting fall
in consumption demand will tend to generate a fall in economic activity which in turn will
increase the public budget deficit by automatically reducing government revenue from
taxes on income and consumption and by automatically increasing public expenses on
unemployment benefits. Alternatively, if the economy is hit by some other contractionary
25 ~-------------------------------------------
Ap.
20 +---------------------------~r-----
~~--~~~------
,.._
~~~-r-
A.--
l v_
at_e_s_a_
vl_n_
g _ _ _ ___
~ ~....::::::::::
0.. 15
0
0
-10
0 N '<t co 0 N '<t <0 co 0 N '<t <0 co 0
co co co co co
<0
lj) lj) lj) lj) lj) 0
"' "' "' "' "'
lj) lj) lj) lj) lj) lj) lj) lj) lj) lj) lj) lj) lj) lj) lj) 0
N
Year
Figure 16.9: Private and public saving in Denmark

Note: Private saving is gross saving of companies and households relative to GOP. Public saving is gross public
saving relative to GOP.
Source: ADAM database, S tatistics Denmark.
13. See Ricardo's paper on the 'Funding System', The Works and Correspondence of David Ricardo, voi. IV, Piero
Sraffa, ed., Cambridge University Press, 1g51, p. 186. The subtle phrase: 'In point of economy ...' has been inter·
preted to mean something like: 'From a rational economic perspective ...'.
486 PART 5: THE B UILDING BLOCKS FO R THE SHORT-RUN MODEL
1.1 0 .94
Employment
----- @1Et ___ ------------------------------------- 0 .93
1.05
0 .92
0 .9 1
0 .90
0.95
0 .89
0.9
0 .88
0.85 0 .87
1978 1980 1982 1984 1986 1988 1990 1992 1994 1996
Figure 16.10: The rate of employment and the average propensity to consume in Denmark
Source: SMEC database, Danish Economic Council.
shock, say. a drop in investment. the resulting fall in economic activity may increase the
private propensity to save at the same time as it increases the budget deficit via the auto-
matic stabilizers. Thus. Fig. 16.10 suggests that reductions in the rate of employment are
associated with reductions in the average propensity to consume, perhaps because higher
unemployment reduces expected future income growth or increases uncertainty about
the future. 14
16.4 !.,().~.~r.ds..~.~?..r..~..r..eal~~.~i.<;..t.J:l.~c:>.r.J.' of <;().I.J:~.l!l?..P.~.i()n ................................... .
Why Ricardian equivalence is likely to fail
Allhough Fig. 16.9 suggc::sts that the:: Rkanlian Equivalc::nce Thc::orem cannot be dis-
missed so easily, most economists remain sceptical of the theorem in its strong form.
Indeed, Ricardo himself did not believe in the practical relevance of his theorem. Right
after having explained that his three alternative methods of war finance would imply the
same present value of taxes (d. the quotation above). he proceeded to write:
... but the people who pay the taxes never so estimate them, and therefore do not manage
their private affairs accordingly. We are too apt to think that the war is burdensome only in
proportion to what we are at the moment called to pay for it in taxes, without reflecting on the
probable duration of such taxes. 15
1 4. Of course, the fact that the employment rate and the propensity to consume move together does not necessarily
mean that movements in employment cause movements in consumption. Causality could also run in the opposite
d irection, or the two time series could be driven by a common third factor not included in the figure.
15. Ricardo, op. ci:., pp. 186- 18 7.
What Ricardo essentially says here is that ordinary taxpayers simply do not have the fore-
sight or sophistication to calculate the present value of their expected future taxes. Hence
they do not realize that lower taxes today (more government debt) must mean higher
taxes tomorrow as long as the time path of public consumption is unchanged. Casual
observation suggests that many people may indeed be myopic in this sense. But even if
consumers are not irrational or myopic. there may still be several reasons why taxes and
debt are not fully equivalent modes of public finance. Let us briefly consider these reasons.
Finite horizons nnd intergenemtiom!l distribution ejfects

Our analysis above assumed that the private and public sectors have identical planning
horizons. In practice, the state will continue to exist beyond the finite lifetime of the
individual consumer. Many current taxpayers (especially the elderly) may therefore ratio-
nally expect that some of the future taxes needed to service the existing public debt will be
levied on future generations and not on those currently alive. In that case a shift trom tax
finance to debt finance is a way of shilling (part of) the burden of paying lor current
govermnent spending onto future generations. A debt-llnanced tax cut will therefore
increase the human wealth and consumption of the currently living generations.
This argument against Ricardian equivalence may sound plausible, but it was met by
an ingenious objection from the American economist Robert Barro. He pointed out that
parents care about their children. If the government tries to shift part of the tax burden
from current to future generations through debt Hnance, parents may use their tax savings
to increase their bequests to compensate their children for the higher future tax burden.
When parents internalize the welfare of their children who in turn care about the welfare
of their children and so on, the current generation will ellectively behave as if it has an infi-
nite time horizon. Any attempt by the government to redistribute resources across genera-
tions will then be neutralized by of[~etting private intergenerational transfers, according to
Barro. 16 However. this assumes that all parents actually plan to leave bequests to their chil-
dren. In reality, some parents may not do so, for example if they believe that their children
will be much richer than themselves. Moreover, the Barro argument abstracts from popu-
lation growth. If population is growing, the future tax burden will be spread out across a
larger number of taxpayers, and parents will not have to pass on all of their current tax
savings to their children to compensate the latter for the higher future taxes needed to
service the higher public debt. Still. Barro's analysis is an important reminder that human
mortality as such is not a sufllcient reason to dismiss Ricardian equivalence.
Intragenerationnl redistribution
The macro analysis of the preceding section hides the fact that most taxes are redistri-
butive in nature. For the individual taxpayer or the individual family dynasty , a tax cut
today may not be matched by an equivalent present-value tax increase tomorrow. even if
16. Robert J. Barro, 'Are Government Bonds Net Wealth?', Journal of Politic<lf Economy, 82, 1974, pp. 1095-111 7.
This seminal article is the authoritative modern stat ement of the Ricardian Equivalence Theorem. Int erestingly,
even Ba.ro's sophisticated reasoning was antic ipated by Ricardo, fer our p revious quotation from Ricardo's text
continues: ' It would be d ifficult to convince a man possessed of 20,000 pounds, or any other sum, that a perpetual
payment of 50 pound per annum was equally burdensome w ith a sin£le tax of 1000 pounds. He would have some
vague notion that the 50 pounds per annum would be paid by posterity, and would not be paid by him; but if he
leaves his fortune to his son, and leaves it charged with this perpetual lax, where is the difference whether he leaves
him 20,000 pounds with the tax, or 19,000 pounds w ithout it?'
the government obeys its aggregate intertemporal budget constraint. If a shift in the tax
burden over time also involves a shift in the lifetime tax burdens across different indi-
viduals or families, the resulting intragenerntional redistribution across agents belonging
to the same generation may have macroeconomic ellects. The reason is that different
individuals or families have dillerent characteristics and hence different propensities to
consume so that redistribution of income does not necessarily leave aggregate consump-
tion unchanged. However. intragenerational redistribution does not mean that a switch
from tax finance to debt finance will necessarily raise current consumption. As a counter-
example. if a lot of people fear that they will have to bear a disproportionate share of the
higher future tax burden implied by an increase in public debt. a debt-financed cut in
current taxes may actually reduce current private consumption by reducing expected
aggregate human wealth.
Distortionary taxes
The Ricardian Equivalence Theorem assumes that taxes take the form of lump sum pay-
ments which are unrelated to individual economic behaviour. In practice. a person's tax
bill is typically linked to his or her income or consumption. Taxes on income and con-
sumption may discourage labour supply and savings and induce consumers to change
their pattern of consumption. In particular, if a debt -ilnanced cut in current income taxes
generates expectations of higher future tax rates to service the higher public debt, con-
sumers may want to increase their current labour supply and reduce their fhture labour
supply to take advantage of the fact that marginal tax rates are lower today than they will
be tomorrow. By inducing such intertemporal substitution in labour supply . the government
may be able to stimulate current economic activity through a s\>Vitch from tax finance to
debt finance of current public spending.
Credit constraints
Our derivation of the Ricardian Equivalence Theorem relied on our assumption that
capital markets are perfect so that no consumers are credit-constrained. In reality some
consumers (such as university students!) may be unable to borrow as much as they would
like at the going market interest rate. For example, a person may expect that he will earn
much more in the future th an today, but if his bank does not have the information neces-
sary to estimate his future earnings potential, it may be reluctant to grant him a credit
against his expected future labour income if he cannot produce any collateral. Such a
credit-constrained consumer will \>\!ish to spend all of his current disposable income here
and now, since he would prefer to increase his current consumption at the expense of
future consumption if he could only obtain more credit. If the government implements a
debt-financed tax cut, credit-constrained consumers will therefore increase their current
consumption even if they realize that they will face higher taxes in the future. By means of
the debt-tlnanced cut in present taxes the government is using its access to the capital
market to shift the consumption possibilities of credit-constrained consumers from the
future to the present, thereby ollsetting the imperfections in the private capital market. In
this scenario Ricardian equivalence breaks down. as you are invited to demonstrate in
Exercise 3.
Empirical research has found th at current consumption tends to react more strongly
to changes in current income than one would expect if consumption were governed only
16 CONSU MPTIOI\ , INCO ME AN D W EALTH 489
by expected lifetime income (of which current annual or quarterly income is usually only
a small fraction) . This so-called 'excess sensitivity' of current consumption to current
income may reflect that many consumers are indeed credit-constrained and will hence
wish to consume all of an increase in their current income. or it may reflect that they are
simply short-sighted.
The generalized consumption function
Let us end this chapter by summarizing our theory of private consumption. Althoug h we
have just argued that the consumption of liquidity-constrained consumers will depend
only on current disposable income. it is also realistic to assume that many consumers will
not be credit-constrained in any particular year. For example, people with positive net
financial assets are unlikely to be liquidity-constrained. since they can always sell oil' some
of their assets if they want to increase their present consumption relative to planned future
consumption. For these individuals we have seen that consumption depends on expected
future income as well as on current income. In other words, aggregate consumption
depends on current disposable labour income, Y1. and on the expected rate of income
growth (g) for those consumers who are not credit-constrained. We have also seen that
consumption is affected by the real rate of interest. r, and by the market value of initial
private wealth. V1. We may therefore sum up our theory of consumption in the following
generalized consumption function C(·):
C1 = C(Y1 .g . r,V 1) . (32)

(+) (+) (i) (+)
The signs below the variables in (32) indicate the signs of the partial derivatives implied by
our theory. Recall that. because of ollsetting income and substitution effects, we cannot
say for sure whether a rise in the real interest rate will raise or lower consumption.
However, once we recognize that the value of financial and human wealth varies nega-
tively with the interest rate. it becomes more likely that a higher real interest rate will
cause consumption to fall, in accordance with popular beliefs.
Notice also how expectations feed into consumption. When consumers are optimistic
about the future. they will expect a relatively high rate of income growth. g. This will have
a direct positive ellect on current consumption . A high expected growth rate will also tend
to imply high stock prices and high prices of owner-occupied housing. This will further
stimulate consumption by a positive impact on private wealth V1•
In the next chapter we shall see how our consumption function (32) may be com-
bined with our investment function from Chapter 15 to give a theory of aggregate
demand for goods and services.
§.~ ~~-~~Y....................................................................................................................................................
1. Private consumption is by far the largest component of the aggregate demand for goods and
services. A satisfacto ry theory of private consumption must explain the paradoxical stylized
facts that the average propensity to consume is a decreasing function of disposable income
in microeconomic cross-section data, whereas it is roughly constant in long-run macro-
economic time series data.
lacroeconomics Model
490 PART 5: THE B UILDING BLOCKS FOR THE SHORT-RUN MODEL
2. The properties of the aggregate consumption function may be derived by studying the
behaviour of a representative consumer who must allocate his consumption optimally over
time, subject to his intertemporal budget constraint. W ith perfect capital markets, this con·
straint implies that the present value of lifetime consumption cannot exceed the sum of the
consumer's initial financial and human wealth. Human wealth is the present value of current
and future disposable labour income.
3. In the consumer's optimum, the marginal rate of substitution between present and future con·
sumption equals the relative price of future consumption, given by one plus the real rate of
interest. When disposable income varies over time, the optimizing consumer will want to
smooth the time path of consumption relative to the time path of income. There is evidence
that such consumption smoothing does indeed take place.
4. Given the assumption of perfect capital markets, the optimal intertemporal allocation of con·
sumption implies that current consumption is proportional to total current wealth (the sum of
financial and human wealth). The propensity to consume current wealth depends on the real
interest rate, the consumer' s rate of time preference, and on his intertemporal elasticity of
substitution, defined as the percentage change in the ratio of futu re to present consumption
implied by a 1 per cent change in the consumer' s marginal rate of substitution.
5. A rise in the real interest rate will have offsetting income and substitution effects on the
propensity to consume current wealth. If the intertemporal substitution elasticity is greater
than 1, reflecting a strong willingness of consumers to substitute future for present con·
sumption, the substitution effect will dominate. A rise in the real interest rate will then reduce
the propensity to consume current wealth. The opposite will happen if the intertemporal sub·
stitution elasticity is smaller than 1. Even if a change in the interest rate does not significantly
affect the propensity to consume a given amount of wealth, it may reduce current consump·
tion by reducing the present value of future labour income, that is, by reducing human wealth,
and by reducing the market value of the consumer's stockholdings and housing wealth.
6. For an optimizing consumer, the average propensity to consume current income will vary
positively with the ratio of current financial wealth to current income. There is strong empirical
evidence that such a positive relationship exists. The rough long· run constancy of the average
propensity to consume observed in macroeconomic time series data may be explained by the
fact that the wealth -income ratio, the growth rate of real income, and the real rate of interest
tend to be roughly constant over the long run.
7. The negative correlation between income and the average propensity to consume observed
in microeconomic cross-section data may be explained by the fact that, in any given period,
many consumers will have a relatively low current income relative to their average income over
the life cycle. Such consumers will therefore have a high level of current consumption relative
to their current income, because they expect higher future incomes, or because they have
accumulated wealth by saving out of higher past incomes.
8. A tax cut which is expected to be permanent will have a stronger positive impact on current
consumption than a tax cut which is believed to be temporary. When the real interest rate
equals the rate of time preference, a permanent tax cut will induce a corresponding rise in
current consumption.
and wealth
© The McGraw-Hill
Companies. 2005
lCD
9. The government's intertemporal budget constraint implies that the present value of current
and future taxes must be sufficient to cover the present value of current and future govern-
ment spending plus the initial stock of government debt. For given levels of current and future
government spending, a tax cut today must therefore be offset by a future tax increase of
equal present value.
10. If consumers have rational expectations they will realize the implications of the intertemporal
government budget constraint. This means that a cut in current (lump sum) taxes which is not
accompanied by a cut in present or future public spending w ill have no effect on private con-
sumption: consumers will save all of the current tax cut to be able to finance the higher future
taxes w ithout having to reduce future consumption. This equivalence between tax finance and
debt finance of current public spending is referred to as Ricardian equivalence.
11. In practice, consumers are unlikely to save the full amount of a current tax cut, even if they
realize that lower taxes today must imply higher taxes in the future. First of all, consumers may
believe that some of the future taxes will be levied on future generations. Second, some con-
sumers may be credit-constrained. A switch from current to futu re taxes w ill help these indi-
viduals to achieve a desired rise in current consumption at the expense of futu re consumption.
The use of red istributive and distortionary taxes also means that a switch from tax finance to
debt finance of current public spending is likely to have real effects on current consumption
and labour supply.
12. The theory of private consumption is summarized in the generalized consumption function
which states that aggregate consumption is an increasing function of current disposable
income, of the expected futu re growth rate of income, and of the current ratio of financial
wealth to income. A rise in the real interest rate has a theoretically ambiguous effect, although
it is likely to reduce current consumption due to its nega1ive impact on human and financial
wealth.
Exercises
Exercise 1. Important concepts and results in the theory of private
consumption
1. The theory of consumption presented in this chapter says that private consumption is pro-
portional to private wealth. Explain the definition of wealth, including the concept of 'human
wealth'. Explain the factors which determine the propensity to consume current wealth.
2. Explain the assumptions underlying our theory of consumption. Which assumptions do you
find most important and problematic?
3. Empirical evidence suggests that many consumers tend to spend all of their current d isposable
income immediately. Is this irrational? Discuss.
4. Explain why we cannot say whether an increase in the real (after-tax) interest rate w ill raise or
lower the propensity to consume. Explain the concept of the intertemporal elasticity of sub-
stitution in consumption and its role in determining the effect of a change in the interest rate
on consumption. On balance, what do you consider the most likely effect of a rise in the real
interest rate on consumption (positive or negative?). Justify your answer.
492 PART 5: THE BUILDING BLOCKS FOR THE SHORT-RUN MODEL
5. Explain the conflicting evidence on the relationship between the average propensity to
consume and disposable income found in microeconomic cross-section data and in macro·
economic time series data. Explain how the theory of consumption presented in this chapter
helps to resolve the apparent inconsistency between the two types of evidence.
6. Explain the consumer's intertemporal budget constraint and the government's intertemporal
budget constraint and how the two constraints are linked, assuming that the representative
consumer and 1he government have the same two-period planning horizon. {Hint: try to elimi·
nate the consumer's lump sum tax payments from his budget constraint by using the govern·
ment's intertemporal budget constraint.) What does your finding imply for the relationship
between private and public spending? Explain and discuss.
Exercise 2. The consumption function with endogenous labour supply
In the main text of this chapter we made the simplifying assumption that the consumer's
labour income is exogenously given to him, say, because wage rates as well as working hours
are regulated by collective bargaining agreements. This exercise invites you to derive the
consumption function when the rep resentative consumer may freely choose his preferred
number of working hours, h, while still taking the real wage rate was given by the market.
To simplify, we will assume that the consumer is retired from the labour market during the
second period of his life, earning labour income only during the first period. In real terms, his
budget constraints for the two periods of life may then be written as:
V2 = (1 + r)[V 1 + w(1 - r)h- C,] (33)
c2= v2 (34)
where r is a proportional labour income tax so that w(1 - r) h is disposable labour income. For
concreteness, we assume that the consumer's lifetime utility is given by the function:
InC
U = In C 1 + ,81n( 1- h) + - -2 (35)
1 +1/>
where ¢ is the rate of time preference. In {35) we have assumed that the total time available
to the consumer in the first period is equal to 1. Hence the magnitude 1 - h is the amount of
leisure enjoyed during period 1, and ,8 is a parameter indicating the strength of the con·
sumer's preference for leisure. Furthermore, the after· tax real wage rate w( 1 - r ) may be
called the consumer's potentia/ labour income (or the market value of his time endowment),
since it measures the amount of wage income he could earn if he worked all the time. Note
that w{1 - r) may also be seen as the 'price' of leisure, since it measures the net income
forgone by the consumer if he chooses to consume one more unit of leisure.
1. Derive and interpret the consumer's intertemporal budget constraint. Does lifetime consump·
tion depend on actual or on potential labour income? (Hint: for later purposes it may be useful
for you to rewrite the lifetime budget constraint so that the value of the consumption of leisure,
w(1 - r) (1 -h), appears on the left·hand side as part of total lifetime consumption.)
2. Rewrite the utility function (35) and the intertemporal budget constraint by introducing the
variable F "' 1 - h, where F is leisure, and use the intertemporal budget constraint to eliminate
C 2 from the utility function. Then show that the utility-maximizing choices of C , and F w ill be
given by:
- 1+¢
C1= i.l[V, + w(1 - r)], (J = - - - - ' - - - (36)
- 1 +( 1 +¢)( 1 +/3) '
,BC 1
(37)
F = w(1-r) '
Compare your consumption function to the consumption function (1 7) derived in the main text
of the chapter and comment on the similarities and differences.
3. Derive an expression for the consumer's optimal labour supply h during period 1 (recall that
h = 1 -F). How does labour supply depend on the labour income tax rate r ? G ive an intuitive
explanation.
4. Suppose now that the consumer's desired working hours h exceed the amount of hours li that
he is actually allowed to work as a member of his trade union. In that case he w ill work the
maximum hours allowed, and his d isposable labour income in period 1 will be w(1 - r)li.
Derive his optimal period 1 consumption level in this case and compare w ith the consumption
function derived in Question 2. Comment on the d ifference.
Exercise 3. Fiscal policy and consumption with credit-constrained

consumers
In Section 3 of this chapter we saw that the Ricardian Equivalence Theorem rests on the
assumption of perfect capital markets. In this exercise you are asked to analyse the effects of
a switch from tax finance to debt finance of public consumption when some consumers are
credit-constrained.
For concreteness, suppose that all consumers in the economy earn the same labour
income and pay the same amount of taxes in each period, but that one group of consumers
('the poor') enters the economy w ith a zero level of initial wealth at the beginning of period 1,
whereas the remaining consumers ('the rich') start out with a level of initial wealth equal to V1 •
Moreover, suppose that disposable labour income in period 1 is so low that the poor would
like to borrow during that period, but that the banks are afraid of lending them money because
they cannot provide any collateral. In that case the poor w ill be credit-constrained during
period 1, and the consumption of a poor person during that period, C~, w ill then be given by
the budget constraint:
C ~= Y~- T1 (38)
A rich consumer does not face any borrowing constraint, and his optimal consumption in
period 1, C~, w ill therefore be given by the consumption function (16) from the main text, that
is:
c I = 8 v, + T i -
, ( ,A_ Y~-
T, + - -- ' T2) 0 <8< 1 (39)
1+r
Suppose that the total population size is equal to 1 (we can always normalize population size
in this way by appropriate choice of our units of measurement). Suppose further that a fraction
lt of the total population is 'poor' in the sense of having no initial wealth.
494 PART 5: THE B UILDING BLOCKS FOR THE SHORT-RUN MODEL
1. Derive the economy's aggregate consumption function for period 1, that is, derive an expres·
sion for total consumption C 1 = 1tC~ + ( 1 -~t)C; . Derive an expression for the economy's mar·
ginal propensity to consume current disposable income, dC 1/ il(Yt- T1). Compare this
expression with the value of the marginal propensity to consume in an economy with no
credit·constraired consumers. Explain the difference.
Suppose now that the government enacts a debt-financed reduction of current taxes T1 by
one unit, without cutting current or planned future public consumption. Assume further that
the government and the private sector have the same planning horizons, and that the private
sector understands that the tax cut today will have to be matched by a tax hike tomorrow of
the same present value, due to the intertemporal government budget constraint. In other
words, suppose that all consumers realize that:
dT2
dT1 + - - =0 (40)
1+r
2. Derive the effect of the switch from tax finance to debt finance on aggregate private con·
sumption in period 1. Compare with the situation with no credit-constrained consumers and
explain the difference. Discuss whether the increase in public debt in period 1 will improve the
lifetime welfare of consumers.
Instead of fir ancing the tax cut in period 1 by debt, the government may finance the tax
reduction by cutting public consumption. In the question below we will distinguish between
two scenarios. In the first scenario the fall in public consumption is expected to be temporary,
that is, dG 1 = dT1 and dG2 = dT2 = 0 . In the second scenario the cut in public consumption is
expected to be permanent so that dG 2 = dG 1 = dT1 = dT2 •
3. Derive the effect on current private consumption C 1 of a temporary tax cut financed by a
temporary cut in public consumption. Compare this to the effect on C 1 of a permanent tax cut
financed by a permanent cut in public consumption. (In the latter case you may assume that
r=¢ .) Explain the difference between your expressions. Does it make any difference for the
effects of temporary and permanent tax cuts whether or not consumers are credit-constrained?
Exercise 4. Ricardian Equivalence?

1. Explain the content of the Ricardian Equivalence Theorem and its assumptions. Discuss
whether the theorem is likely to hold in practice.
2. Try to find data for private saving and public sector saving as ratios to GOP in your country
and plot the two series against each other. (If you are a Danish student, try to update Fig. 16.9
in the text.) Do the movements in the two time series seem consistent with the Ricardian
Equivalence Theorem? Explain.
3. In many official statistics public sector saving is simply measured by the balance on the public
sector budget (the negative of the budget deficit), defined as T- C9-/9- rD, where T is net
taxes (taxes net of transfers), C9 is public consmption, J9 is public investment, and rD is the
net interest payment on government debt. Discuss whether the ideal measure of public sector
saving for the purpose of testing Ricardian Equivalence should include public investment, J9?
(Hint: if the budget deficit increases due to an increase in J9, and if public investment yields a
positive return, how will this affect the private sector's future tax liability?) Try to check if your
empirical data for public sector savings include or exclude the public sector's net investment.
Exercise 5. Wage taxes versus consumption taxes versus wealth taxes
In the main text of the chapter we assumed for simplicity that taxes took the form of lump sum
payments which were unrelated to the consumer's behaviour. Now we assume instead that
the consumer must pay a proportional tax, r w, on his wage ncome, a proportional value-added
tax, r 0 , on all his consumption expenditure, and potentially also a one-time tax, r", on his initial
wealth. This exercise asks you to consider some similarities and d ifferences between these
taxes.
We assume that the consumer's working hours are exogenously g iven to him and equal
to 1 so that his total pre-tax labour income in period 1 is equal to the real wage rate w. We
also assume that he is retired from the labour market during period 2. We begin w ith a situa-
tion w ithout any taxes on wealth. The consumer budget constraints for the two periods of life
are then given by:
V2 = (1 +r) [ V1 +(1- r~w - ( 1 +r")C 1], (41 )

( 1 + r") C 2 = V2 (42)
1. Derive the consumer's intertemporal budget constraint and comment on your expression.
From the consumer's perspective, what is the similarity and the d ifference between the wage
tax and the consumption tax?
The consumer has the lifetime utility function:
In C 2
U = lnC 1 + - - , (43)
1 +¢
2. Show that the consumer's optimal consumption is given by:
C1 = (~)[ V1 + (1 - 0r"') w]· (44)

2 + </> 1 + 7:
C2 = (~)[ V1+ (1 - 'I:~W]· (45)

2 + </> 1 + 7: 0
Comment on these expressions.
Suppose now that the government needs to raise more to.x revenue to finance additional
public spending. The government presents two alternative proposals in parliament. The first
proposal imp lies that the consumption tax rate (the VAD will be raised from 20 per cent to
25 per cent while the wage tax rate will be kept unchanged at 50 per cent. The second
proposal implies that the consumption tax rate is maintained at 20 per cent, wh ile the wage
tax rate is raised from 50 per cent to 52 per cent. In addition, the second proposal includes a
one-time proportional tax of 4 per cent on existing initial wealth V1 • The government stresses
that this is a once-and-for-all wealth tax which w ill not be imposed on future wealth V2 •
3. Would consumers prefer one of the government's proposals to the other one? (Hint: how do
the two alternative tax plans affect the consumer's total wealth, including his human wealth?)
What d ifference does it make to your answer whether the wealth tax is a one-time levy or a
permanent tax?
17. Monetary policy and
aggregate demand
© The McGraw-Hill
Companies. 2005
Chapter ••
•
•
~ Monetary policy and
•
aggregate demand
•
••
•
: According to the classical macroeconomic theory which dominated economic
• thought before the Great Depression of the 1930s, the total level of output and
: employment is determined from the economy's supply side. In the classical world,
wages and prices adjust to ensure that the available supplies of labour and capital are
utilized at their 'natural' rates determined by the structure oflabour and product markets.
In Chapter 1 we argued that this is a useful working assumption when we analyse the
long-run economic phenomena which are the subject matter of the theory of economic
growth. But in the short and medium term economic activity often deviates from its long-
run growth trend. To understand these short-run macroeconomic fluctuations, we must
explain why the aggregate demand for goods and services does not necessarily correspond
to the aggregate supply which is forthcoming when all resources are utilized at their
natural rates. Building on the previous two chapters, the present chapter therefore
develops a theory of aggregate demand.
~~Y..~.~.~.:..~~~..9~.~ ~.~.~~.~. ~~4. ~~.~. 9t~.~~. P..~.P.~~~.~.~g.!:.......................................... .

The classical economists did not literally claim that a capitalist market economy could
never deviate from its natural rate of employment and output. but they did believe that if
only market forces were allowed to work. such disturbances would be temporary and
quite short-lived. The classical economists therefore saw no need for the government to
engage in macroeconomic stabilization policy. In their view, the only role of monetary
policy was to secure price stability. and the task ofllscal policy was to avoid budgetdeflcits
which would crowd out private capital formation and thereby hamper economic growth.
Winston Churchill. who was Secretary of the Treasury in Britain for several years during
the 1920s. was very much in agreement with this classical view when he explained h is
approach to fiscal policy as follows: 'It is the orthodox Treasury dogma, steadfastly held,
that whatever might be the political or social advantages, very little employment can. in
fact, as a general rule. be created by state borrowing and state expenditure. ' 1
1. Quoted in Richard T. Froyen, Macroeconomics - Theories and Policies, 6th edition, Prentice-Hall, Inc., 1999, p. 68.
496
Sorensen-Whitta- Jacobsen: I Part 5- The Building 17. Monetary policy and © The McGraw-Hill
Introducing Advanced Blocks for the Short- run aggregate demand Companies. 2005
17 MONETARY POLICY AND AGGREGA TE DEMAND 497
When the Great Depression of the 1930s struck the Western world, this classical
laissez-faire position came under heavy attack. The Great Depression was an economic
earthquake. Due to a catastrophic combination of negative shocks and macroeconomic
policy failures. output in several countries fell by 25- 30 per cent between 1929 and
1932- 33, with disastrous consequences for employment (see Fig. 17.1). For a country
% USA % UK
25 16
14
20
12
15 10
8
10 6
4
5
2
0 0
0 (') CD Ol ('ol t.O (X) 0 (') <0 Ol ('ol t.O (X)
('ol ('ol ('ol ('ol (') (') (') ('ol ('ol (") (") (')
Ol Ol Ol Ol Ol Ol Ol Ol Ol "'~ C'l
Ol Ol Ol Ol
% Germany % Canada
18 25
1\
16
14 I \ 20
12
I \
10 I \ 15
8
I \
6
1\ I \ 10
4
/\I\ I \
I v 5
2 ""-..../
0
0
('ol
Ol
(')
('ol
Ol
CD
"'
Ol
"'-
(")
('ol
Ol
<0
"'~
Ol
('ol
Ol
('ol
(")
Ol
t.O
(')
Ol
(X)
(')
Ol
o/o Netherlands % Denmark

14 25
12
20
10
8 15
6 10
4
5
2
00 (") C'l t.O (X)
00 (") t.O (X)
('ol
Ol
('ol
Ol
CD
"'
Ol
Ol
C'l
Ol
(")
Ol
(")
Ol ~
('ol
Ol "'
Ol
CD
C'l
Ol
Ol
('ol
Ol
"'
(")
Ol
(')
Ol
(")
Ol
Figure 17.1: Unemployment rates 1920-39

Note: The unemployment rate is defined as the number of unemployed persons as a percentage of the total labour
force.
Source: Macroeconomic dat abase constructed by Jacob Brechner Madsen, University of Copenhagen. The under-
lying sources are document ed in: Jacob Br0chner Madsen, 'Agricultural Crises and the International Transmission
of the Great Depression' , Journal of Economic History, 61 (2), 2001, pp. 327-365.
aggregate demand
© The McGraw-Hill
Companies. 2005
498 PA RT 5 : TH E B UIL DING BLOCKS FO R THE S HO RT- RUN MO DEL
like the United States, it took almost a decade for output to return to its pre-1929 peale a
whole decade of economic growth lost!
Against this background the British economist John Maynard Keynes and several
others attacked the classical view that resource utilization at natural rates is the normal
state of a!Tairs. Indeed. Keynes challenged the time-honoured dellnition of economics as
;the study of the allocation of scarce resources to satisfy competing ends'. Keynes' point
was that quite often resources are not scarce. they are merely underutilized due to a lack
of demand. In such circumstances the government will be able to raise total employment
and output through a llscal or monetary policy which stimulates aggregate demand.
These ideas were laid out in 1936 in Keynes' famous book, The General Theory of
Employment. Interest and Money . That book is often considered to mark the birth of modern
macroeconomics, because it revolutionized the way economists thought about the
problem of business cycles.
Today most macroeconomists believe that economic activity in the short and
medium run is determined by the interaction of aggregate demand and aggregate supply.
In the long run the forces of aggregate supply stressed by the classical economists carry
the day. but in the short run aggregate demand plays a key role in the determination of
output and employment. As a step on the way to constructing a model of short-run
macroeconomic fluctuations, we must therefore develop a theory of aggregate demand.
We have already seen that private investment as well as private consumption are
inlluenced by the real rate of interest. In the long run the equilibrium real interest rate-
the so-called natural rate ofinterest- is determined by the forces of productivity and th rift,
as we explained in Chapter 3. However, in the short run monetary policy can have a
signiHcant impact on the real interest rate. Hence much of this chapter will focus on the
conduct of monetary policy and how it affects aggregate demand.
We start our analysis by specifying the equilibrium condition for the goods market,
drawing on the theory of consumption and investment developed in Chapters 15 and 16.
We then move on to a study of the monetary sector and the conduct of monetary policy.
Incorporating our specification of monetary policy into the equilibrium condition lor the
goods market. we then end up deriving a systematic link between the level of output and
the rate of inllation which must hold whenever the goods market clears. This link is called
the aggregate demand curve. and it will be one of the two central building blocks of the
short-run macroeconomics model we set up in Chapter 19.
17.2 T.?..~.~.<?.?..~.~ ..~.~r..~~f?.~............................................................................................................................

Goods market equilibrium
For the product market to clear, the aggregate demand lor goods must be equal to total
output, Y. In this chapter we will locus on a closed economy (we will consider the open
economy in later chapters). Aggregate demand for goods then consists of the sum of real
private consumption, C. real private investment. I, and real government demand for goods
and services, G. Hence goods market equilibrium requires:
Y = C+I+G. (1)
Sorensen-Whitta-Jacobsen: I Part 5- The Building 17. Monetary policy and © The McGraw-Hill
Introducing Advanced Blocks for the Short-run aggregate demand Companies. 2005
17 M ONETARY PO LICY AN D AG G RE G ATE DEMA ND 499
In Chapter 15 we saw that private investment behaviour can be summarized in an invest-

ment function of the form I = I(Y, r, K, c), where r is the real interest rate, K is the pre-
determined capital stock existing at the beginning of the current period. and e is a
parameter capturing the 'state of confidence' , reflecting the expected growth of income
and demand. For the purpose of short-run analysis, 'Ne may treat the predetermined
capital stock as a constant and leave it out of our behavioural equations. 2 We may then
write private investment demand as:
oi > 0, oi oi
I = I(Y, r, c), Iy = -
av I=
r
-or < 0 • I= -
t oc > 0 ' (2)
where the signs oft he partial derivatives of the investment function follow from the theory
developed in Chapter 15. Thus, investment increases with current output and with
growth expectations, t, whereas it decreases with the real interest rate.
Our theory of private consumption presented in Chapter 16 implies a consumption
function of the form C = C(Y - T, r, V, c) . where T denotes total tax payments so that Y - T
is current disposable income, and Vis non-human wealth. We assume that the future
income growth expected by consumers equals the growth expectations of business firms,
since Hnns are owned by consumers. Our analysis in Chapter 16 showed that the market
value of non-human wealth is a decreasing function of r, since a rise in the real interest
rate will, ceteris paribus, drive down stock prices as well the value of the housing stock. In
other words, V = V(r) and dV/dr < 0 . To simplify exposition, we will use this relationship to
eliminate V from the consumption function and simply wTite:
ac ac
C = C(Y - T,r,e) , 0 < C,, =o(Y - T) < 1, c, =a;;:;; o.
The signs of the partial derivatives were explained in Chapter 16. From that chapter we
recall that the real interest rate has an ambiguous effect on consumption, due to ollsetting
income and substitution eflects, although the negative impact of a higher interest rate on
private wealth suggests that the net eflect on consumption is likely to be negative. The
analysis in Chapter 16 also implied that the marginal propensity to consume current
income is generally less than 1, as we assume above.
=
Let us denote total private demand by D C +I. To avoid complications arising from
the dynamics of government debt accumulation, we will assume that the government
balances its budget so that T = G. It then follows from (2) and (3) that the goods market
equilibriwn condition (1) may be stated in the form:
Y = D(Y. G. r. e)+ G. (4)
Properties of the private demand function
We will now consider the signs and magnitudes of the partial derivatives of the
private demand function D(Y, G. r, e). Since D= C+ I, it follows from (2) and (3) that
2. The dynamics of capital accumulation were dealt w ith in Book One. To include it here as well would complicate the
fonnal analysis considerably, given that we also want to study the dynamics of output and inflation. We therefore
leave the inclusion of capital stock adjustment in the economy's short·run dynamics for a more advanced macro-
economics course.
aggregate demand
© The McGraw-Hill
Companies. 2005
=
Dv ()D/()Y = Cv + Iv > 0 . The derivative Dv is the marginal private propensity to spend,
defined as the increase in total private demand induced by a unit increase in income. We
will assume that the marginal spending propensity is less than 1 so that:
()D
0 < D1, = - = Cy+ l y < 1. (5)
()Y
=
The assumption that Dy < 1 guarantees that the Keynesial! multiplier ri1 1/(1 - Dy) is
positive. Recall from your basic macroeconomics course that the Keynesian multiplier
measures the total increase in aggregate demand for goods generated by a unit increase in
some exogenous demand component, provided that interest rates and prices stay con-
stant. The Keynesian multiplier captures the phenomenon that once economic activity
goes up. the resulting rise in output and income induces a further increase in private con-
sumption and investment. which generates an additional rise in output and income that
in turn causes a new round of private spending increase, and so on. Below we shall return
to the role played by the Keynesian multiplier in our theory of aggregate demand.
Since T = G, we see from (3) that:
()D ()C
De; = ()G = - ()(Y - T) = - c,, < 0 . (6)
Given that Cy < 1, it follows th at the net eflect of a unit increase in government demand on
aggregate (private plus public) demand '>Viii be 1 + De; = 1 - C1, > 0. In other words. a fully
tax -financed increase in public consumption will only be partially offset by a fall in private
consumption, so the net eilect on aggregate demand will be positive. This assumes that at
least part of the increase in taxes is expected to be temporary, lor as we saw in Chapter 16,
a permanent tax increase '>Viii tend to generate an equivalent fall in private consumption.
The ellect of a rise in the real interest rate on private demand is given by
=
D, ()Dj()r = C, + l ,. The derivative D, measures the effect of a rise in the real interest rate
on the private sector savings surplus. The private sector savings surplus is defined as
SS = S - l, where private saving is given by S = Y - T - C. Hence we have ()SS/()r =
- C, - l , = -(C, +I,)= - D,. There is strong empirical evidence that a higher real interest
rate raises the private sector savings surplus. For example. Fig. 17.2 illustrates a clear
positive correlation between SS and a measure of the real interest rate in Denmark. Even
though economic theory does not unambiguously determine the sign of the derivative C, ,
we may therefore safely assume that:
()D
D, = - = C,+ I, < 0. (7)
dr
Finally we see from (2) and (3) that theeflect on private demand of more optimistic growth
expectations is:
(R)
Restating the condition for goods market equilibrium
It will be convenient to re,.vrite the goods market equilibrium condition (4) such that
output, government spending and the confidence variable, t:, appear as percentage
17 MO NETARY POLICY AN D AG G RE G ATE D EMA ND 501
7.------------------------------------------------.s
- - Private sector savings surplus (left axis)
- Real interest rate (right axis)
41---------------~~~--------~-,~~~--------r 3
a.
Cl
~
0
.,
0> -2
c.,"'
0
Q;
a.
-5 ---------------r-3
- 11 +---~-.---.--.-.-.-~----~------~----~.---.-r-+ - 7
.....
lt) .....
..... (') lt) lt) ..... <»
..... co co co
<» -<» (j) <» <»
<» <» <» <» <» <» <» <» <» <»
Year
Figure 17.2: The real interest rate and the private sector savings surplus in Denmark, 1971-2000
Note: The real interest rat e is measured as the after-tax nominal interest rat e on ten·year government bonds minus
an estimat ed trend rate of inflation which includes the rate of increase of housing prices.
Source: Erik Haller Pedersen, 'Udvikling i og mating af realrent en', Dan11arks Nationalbank, Kvartalsoversigt, 3.
kvartal, 2001, Figure 6.
deviations from their trend values. We begin from an initial situation in which the
economy is on its long-run growth trend so that initial output is equal to Y. We then
consider a small deviation from trend. Taking a tirst-order linear approximation of the
goods market equilibrium condition (4), remembering that DG= Dr = - C1, and denoting
the initial trend values by bars. we get:
Y - Y = ri1(1 - Cv)(G - G)+ l'hD,(r - f) + ir1D.(e - c), ih =---.

1
1 - Dy
(9)
Our next step is to rewrite (9) in terms of relative changes in Y, U and t:
Y-Y=
----y- li1(1 - Cy) (~(G
Y} G - ~(IT! (D)i (r - f) +lrl (W)(e-c)
v' [ . (10)
In the !Ina! step we use the fact that the change in the log of some variable is approxi-
mately equal to the relative change in that variable. Defining
y =In Y, y =h1 Y, g = In G. g= In G.
we may then write (10) in the form:
y - [1 = a.l(g - g) - o.2(r - f) +v, (11)
where
-(D')
o.2= - my' (12)
aggregate demand
© The McGraw-Hill
Companies. 2005
The magnitudes G, f and E: are the values of G. rand t prevailing in a long-run equilibrium
where output is at its trend level. Thus Eq. (11) says that the percentage deviation of
output from trend (the output gap) can be approximated by a linear function of the
percentage de\iations of G and f. from their trend values and of the absolute deviation of r
from its trend level. 3 Of course, ( 11) is just a particular way of stating that the aggregate
demand for goods varies negatively with the real interest rate and positively with govern-
ment spending and with expected income growth. Note that the long-run equilibrium real
interest rate f can be found from the following condition for long-run goods market
equilibrium:
Y = D(Y, G. r, c) + G. (13)
=
Notice also the role played by the Keynesian multiplier 1h 1/(1 - Dy) in the definitions of
the coell1cients a 1 and a 2 given in (12). For example, if taxes are raised by one unit to
finance a unit increase in government consumption, the immediate impact is a net
increase in aggregate demand equal to 1 - Cy. But when the Keynesian multiplier ellect is
accounted for. the total increase in demand adds up to ii-1(1 - C1,) . Therefore, if public
consumption increases by 1 per cent, the resulting percentage increase in total demand
will be l'h(1 - C1,)(G/Y). given that the initial ratio of public consumption to total output is
G/Y. This explains the coeflicient a 1 on the percentage increase in government consump-
tion. g - g. in (11). Similarly, if the real interest rate goes up by one percentage point, the
resulting perce11tage drop in total demand is D,./Y. When this initial fall in demand is
magniHed by the Keynesian multiplier. the total percentage fall in demand adds up to
- 1iz(D,/Y). as shown by the expression lor a 2 in (12). Thus the familiar Keynesian
multiplier theory is built into our theory of aggregate demand.
Equation (11) is our preliminary version of the economy's aggregate demand curve.
Below we will show that {11) implies a systematic link between output and inflation. once
one allows lor the way monetary policy is typically conducted. To understand this link, we
must study the relationship between inflation and the real interest rate, and that requires
tal<ing a closer look at the money market and the behaviour of central banks.
17.3 T.?..~. !:?.?..~.~Y...~.~~~~.~.~..~.~~..~.?..~.~.~~~Y...P?.~~~Y............................................................

The money market
From your basic macroeconomics course you wUI recall that equilibrium in the money
market is obtained when
M ol
-p = L(Y• i) . L. =- < 0, (14)
J ()j
3 . Using the linear approximation ( 11) implies that we are treating the coefficients a 1 and a 2 as constants. To justify the
assumption that a 2 = -m(O,/Y) remains roughly constant despite the fact that Y is growing over time, we must
assume that the derivative 0, takes the fonn 0 , = f(r) Y, f ' < 0. This is equivalent to assuming that the investment ratio,
1/ Y, w ill stay constant as long as the real interest rat e stays constant. Thi s assJmption is warranted by our analysis of
economic g rowth in B ook One where we found that once the real interest rat e has settled down to its long·run
equilibrium value, the level of investment t ends to grow at the same rate as output.
where L(Y, i) is the real demand for money, i is the nominal interest rate, M is the nominal
money supply, and Pis the price level. The left-hand side of (14) is the supply of real money
balances which must be equal to real money demand in equilibrium. Real money demand
varies positively with income. since a rise in income leads to more transactions which in
turn requires more liquidity. At the same time money demand varies negatively with the
nominal interest rate, because a higher interest rate raises the opportunity cost of holding
money rather than interest-bearing assets, inducing agents to economize on their money
balances so as to be able to invest a larger share of their wealth in interest-bearing
financial instruments. For concreteness, we will assume that the demand for real money
balances can be approximated by a function of the form:
k > 0, rt> 0, fbO . (15)
where e is the exponential function . TJ is the income elasticity of money demand. and {3 is
the semi-elasticity of money demand with respect to the interest rate. 4 Notice that the
interest rate i appearing in the money demand function should be interpreted as a short-
term interest rate, since the closest substitutes for money are the most liquid interest-
bearing assets with a short term to maturity.
The constant money growth rule
To llnd the link between output and inflation on the economy's demand side. we need to
know how the real interest rate r appearing in (11) is related to these two variables. This
depends on the way monetary policy is conducted. i\'lonetary policy regimes vary across
time and space. Here we sh all focus on two bench mark monetary policy rules which have
received widespread attention in the literature. A monetary policy rule is a rule or prin-
ciple prescribing how the monetary policy instrument of the central bank should be
chosen. In practice, the main monetary policy instrument of the central bank is its short-
term interest rate charged or oflered vis-a-vis the commercial banking sector. Through
their control of the central bani< interest rate, monetary policy makers can roughly
control the level of short-term interest rates prevailing in the interbank market. The inter-
bank market is the market for short-term credit where commercial banks with a tem-
porary surplus of liquidity meet other commercial banks with a temporary liquidity
shortage. The interbank interest rate in turn heavily influences the le\rel of market interest
rates on all types of short-term credit.
Under the constant money growth rule for the conduct of monetary policy the central
bank adjusts its short-term interest rate to ensure that the forthcoming money demand
results in a constant growth rnte of the nominal rnonetary base. Assuming a constant money
multiplier (that is. a constant ratio between the broader money supply and the monetary
base), this will also ensure a constant growth rate of the broader money supply which
includes bank deposits as well as base money. In an influential book published in 19 60. the
American economist Milton Friedman argued that a constant money supply growth rate
would in practice ensure the highest degree of macroeconomic stability which could real-
istically be achieved, since it would imply a stable increase in aggregate nominal income. 5
4. The semi·elastic ity p measures the percentage drop in real money demand induced by a one percentage point
increase in the interest rate.
5. See Milton Friedman, A Program for Monetary Stability, New York, Fordham University Press, 1960.
aggregate demand
© The McGraw-Hill
Companies. 2005
This argument was based on Friedman's belief in a stable money demand function with a
low interest rate elasticity. To see his point most clearly, suppose for a moment that our
parameter fJ in (15) is close to 0 . and that the income elasticity of money demand 'YJ is equal
to 1. Money market equilibrium then roughly requires M = kPY. where k is a constant.
Hence aggregate nominal income PY must grow roughly in proportion to the nominal
money supply M. Securing a stable growth rate of M will then secure a stable growth rate
of nominal income.
Friedman pointed out that we have only limited knowledge of the way the economy
works. His studies of American monetary history also suggested that monetary policy
tends to affect the real economy with long and variable lags.6 Friedman therefore argued
that the central bank may often end up destabilizing the economy if it attempts to manage
aggregate demand through activist monetary policy by constantly varying the growth
rate of money supply in response to changing economic conditions. Moreover, according
to Friedman, the self-regulating market forces are sufl1ciently strong to ensure that real
output and employment will be pulled fairly quickly towards their ;natural' rates follow-
ing an economic disturbance. Given that activist monetary policy may fail to stabilize the
economy, and that the need for stabilization is limited anyway. Friedman concluded that
his constant money supply growth rule would be the best way to conduct monetary
policy.
Friedman's arguments did not go unchallenged, but they had a substantial impact on
many central banks. In particular. the German Bundesbank adopted stable target growth
rates lor the money supply from the 19 70s, and after the formation of the European
Monetary Union the European Central Bank has maintained a target for the evolution of
the money supply to support its target for (low) inflation.
What does the constant money grm.vth rule imply for the formation of interest rates?
To investigate this, suppose that the central bank knows the structure of the money
market sufficiently well to be able to implement its desired constant growth rate fl of the
nominal money supply. Using (14) and (15). and denoting the rate of inflation by nso that
=
P (1 + n)P_I' we may then write the condition for money market equilibrium as:
(1 +ft)M_1
(16)
(1 + n)P - :
where M_1 and P_1 are the nominal money supply and the price level prevailing in the
previous period, respectively. We want to study how the economy behaves when it is not
;too' far off its long-run trend. We therefore assume that the economy was in long-run
equilibrium in the previous period. Ignoring grm.vth for simplicity, a long-run equilibrium
requires that the real money supply be constant, since the variables in the money demand
function. Y and i , must be constant in a long-run equilibrium without secular growth.
Constancy of the real money supply means that the inflation rate n must equal the mone-
tary growth rate It· 7 As we shall explain more carefully in Section 4. the approximate link
6. M ilton Friedman and Anna Schwartz, A Monetary History of the United States, 1867 - 1960, Princet on, NJ, Princeton
University Press, 1963. In Chapter 22 we shall discuss the lags in monetary policy in more detail.
7. If we assume that trend output Y grows at the constant rate x, the real money supply would have to g row at the rate
YJX in a long·run equilibrium w ith a constant interest rate. This in turn would imply an equilibrium rate of inflation n •
equal to , • =Jt-1/X. Nevertheless, for constant values of 'I and x, one can show that the nominal interest rate would
still react to changes in inflation and output in accordance with our Eq. (20) derived below (see Exerc ise 2).
betv.reen the nominal and the real interest rate is i = r +:n, so in a long-run equilibrium
where :rc = )1 and r = f, we have i = f +It· If we denote the long-run value of the real money
stock by L*, our assumption that the money market was in long-run equilibrium in the
previous period then implies that:
M_, =L* = kYqe-il<'+~'). (17)

P_,
Taking natural logs of (16). remembering that M_tfP_1 = L*, and using the approxi-
mations ln(1 + fl) "" ft and ln(1 + :n) "":n, we get:
ft.- :n + In L * = In k + rJY - {Ji. (18)
where (17) implies:
ln L* = In k + r([l - {J(r + )1). (19)
By inserting (19) into (18) and rearranging, you may verily that:
i = f 1 - {J) (:n - fl) + (t]) (y - y).

+ :rc + ( -(3- p (20)
Equation (20) shows how the short-term nominal interest rate i will react to changes in
inllation and output if monetary policy aims at securing a constant growth rate )1 of the
nominal money supply. Since t'J and {J are both positive, 'Ne see that the interest rate varies
positively with the output gap, y - [!. If the numerical semi-elasticity {J of money demand
with respect to the interest rate is not too high ({J < 1), as Friedman assumed. we also see
that the nominal interest rate will increase more than one-to-one with the rate of infla-
tion, implying an increase in the real interest rate. Note that since the long-term equilib-
rium inflation rate equals the monetary growth rate, our parameter fl may be interpreted
as the central bank's target injlntion rate.8
The Taylor rule
As we have mentioned. some central banks have occasionally dellned targets for the
growth rate of the nominal money supply, in accordance with Milton Friedman's recom-
mendation. However, in an influential article. American economist John Taylor argued
8. In a provocative essay Milton Friedman argued that the target inflation rate I' ought to be negative and numerically
equal to the equilibrium real interest rate so that the nominal interest rate i = r + I' becomes zero. Friedman's
argument was that the marginal social cost of supplying money to the public is roughly zero, since printing money is
virtually costless. To induce people to hold the socially optimal amount of money balances, the marginal private
opportunity cost of money·holding - given by the nominal interest rate - should therefore also be zero. If the nominal
interest ra:e is positive, people w ill economize on their money balances to hold more of their wealth in the form of
interest·bearing assets. The resulting inconvenience of having to exchange interest·bearing assets for money more
often t o handle the daily transactions w ill yield a utility loss. According :o Friedman this welfare loss can be avoided
at zero social cost by driving the nominal interest rate to zero so that people are no longer induced to economize on
their money balances. This recommendation of a steady rate of deflation to ensure a zero nominal interest rate is
sometimes referred to as the ' Friedman Rule' . See Milton Friedman, 'The Optimum Quantity of Money', in The
Optimum Quantity of Money and Other Essays, pp. 1 -50. Chicago, Aldine Publishing, 1969.
Many economists consider Friedman's recommendation t o be theoretically interesting, but dangerous in practice.
They argue that a policy of deflation can trigger a destabilizing wave ot bankruptcies if debtors have not fully antic i·
paled the future fall in prices and the resulting increase in their real debt burdens. In Chapter 20 we shall consider
some furtl'er reasons why a negative inflation target may be undesirable.
aggregate demand
© The McGraw-Hill
Companies. 2005
that rather than worrying too much about the evolution of the money supply as such. the
central bank might as well simply adjust the short-term interest rate in reaction to
observed deviations of inflation and output from their targets.9 Assuming that policy
makers wish to stabilize output around its trend level, and denoting the inflation target by
.n:*, we may then specif)r the monetary policy rule proposed by Taylor as:
i = f + .n: + h(.n: - n *) + IJ(y - [J), I!> 0. IJ > 0 . (21)
Equation (21) is the famous Taylor mle. Recalling that the monetary growth rate ,u may be
interpreted as an inllation target. we see from (20) and (21) that the nominal interest rate
follows an equation of the same form under the constant money growth rule and under
the Taylor rule. Yet there is an important difference. Under the constant money growth
m le the coelllcients in the equation lor the interest rate depend on the parameters 'I'J and P
in the money demand function . In contrast. under the Taylor rule the parameters h and lJ
in (21) are chosen directly by policy makers, depending on their aversion to inflation and
output instability. According to Taylor it is important that the value of h is positive so that
the real interest rate goes up when inflation increases. If 1 + h is less than 1, a rise in
inflation will drive down the real interest rate i - n. and this in turn will further feed infla-
tion by stimulating aggregate demand lor goods, leading to economic instability . In fact,
Taylor suggested that the parameter values I! = 0. 5 and IJ = 0. 5 would lead to good
economic performance. given the stmcture of the US economy.
Empirical studies have found that although central bankers never mechanically
follow a simple policy rule, central bank interest rates do in fact tend to be set in accord-
ance with equations of the general form given in (21). As we have seen, such interest rate
behaviour is consistent with the constant money growth rule as well as the Taylor rule.
However. one problem with the former rule is that a constant monetary growth rate may
not succeed in stabilizing the evolution of nominal aggregate demand if the parameters of
the money demand function are changing over time in an unpredictable fashion. Such
unanticipated shitts in the money demand function may occur when new financial
instruments and methods of payment emerge as a result of llnancial innovation.
In part because of this problem with the constant money growth rule, monetary
policy has increasingly come to be discussed in terms of the Taylor rule in recent years.
Table 17.1 shows econometric estimates of the 'Taylor' coefllcients hand IJ in some large
OECD economies where interest rate policies have not been significantly constrained by a
target for the foreign exchange rate. According to the studies from which these estimates
are taken, all coelllcients are statistically significant and are estimated '>\lith considerable
accuracy. 10 We see from the table that in recent years all of the four central banks have
followed Taylor's recommendation that h should be substantially above zero to ensure a
rise in the real interest rate in response to a rise in inflation.
Figure 17.3a shows the estimated interest rate reaction function of the European
Central Bank compared to the actual three-month interest rate in the Euro area. In
Fig. 17.3b we compare the evolution of the actual short-term interest rate to the interest
9. See John B. Tayl or, ' Discretion versus Policy Rules in Practice', Carnegie·R?chester Conference Series on Public
Policy, 39, 1993 , pp. 195-214.
10. The estimated ·1alues of h and b are coefficients in an equation like (21) which determines the cent ral bank's target
(desired) interest rate; the estimation m ethod assumes t hat the act ual interest rate adjusts gradually to this target
Table 17.1: Estimated interest rate reaction functions of four central banks
Estimate of
h b Estimation period
German Bundesbank 1 0.31 0.25 1979:3-1 993: 12 (monthly data)

Bank of Japan 1 1.04 0.08 1979:4 -1 994:12 (monthly data)
US Federal Reserve Bank 1 0.83 0.56 1982:10- 1994:12 (monthly data)
European Central Bank 2 0.74 0.82 1999:1-2003: 1 (quarterly data)
1. Source: Richard Clarida, Jordi Gali and Mark Gertler, 'Monetary Policy Rules in Practice -
Some International Evidence' , European Economic Review, 42, 1998, pp. 1033- 1067.
2. Source: Centre for European Policy Studies, Adjusting to Leaner Times, 5 th Annual Report
of the CEPS Macroeconomic Policy Group, B russels, July 2003.
rate which would have prevailed if the ECB had simply followed a Taylor rule of the form
(21) withcoefficientsh = 0.4, b = 0.6,.n* = 1.5 percent and f = 2 percent. Weseethatin
both cases the Taylor rule gives a fairly good description of actual monetary policy in the
Euro area. In Chapter 2 2 we shall discuss whether the Taylor Rule is in fact also an optimnl
monetary policy.
Monetary policy and long-term interest rates: the yield curve
The central bank can control the current short-term interest rate via the choice of its own
borrowing and lending rate. However, the incentive to invest in a real asset depends on
the expected cost of capital over the entire useful life of the asset. This lifetime may be
many years if the asset is, say, a house, a truck, or a piece of machinery. The crucial
5.5
5.0
4.5
4.0
3.5
3.0 - 3 ·month rate

- Estimated Taylor rate
2.5
2.0 + - - - - - , . . . . - - - - - , . . - - - - - - - r - - - - - - , -
1999 2000 2001 2002 2003
Figure 17.3a: Three·month rate and the estimated Taylor rate in the Euro area
Source: Centre for European Policy S tudies, Adjusting to Leaner Times, 5th Annual Report of the C EPS
Macroeconomic Policy Group, B russels, July 2003, p. 53.
aggregate demand
© The McGraw-Hill
Companies. 2005
508 PA RT 5 : TH E B UIL DING BLOCKS FO R THE S HO RT- RUN M O DEL
5.5
5.0
4.5
4.0
3.5
3.0 - 3 -month rate

- Calibrated Taylor rate
2.5
2.0+-----.-----.-------.-----.-
1999 2000 2001 2002 2003
Figure 17.3b: Th ree-month rate and the calibrated Taylor rate in the Euro area
Source: Centre for European Policy Studies, Adjusting to Leaner Times, 5th Annual Report of the CEPS
Macroeconomic p,)licy Group, Brussels, July 2003, p. 51.
question is: to what extent can monetary policy aiTect the incentive to acquire the long-
lived assets which make up the bulk of investment?
Consider a firm which is contemplating investment in a real asset with an expected
lifetime of n periods. Suppose first that the firm plans to finance the investment with short
term debt which is 'rolled over' in each period so that the interest rate varies with the
movements in the short-term interest rate. For simplicity, suppose further that the llrm
does not need to pay any interest until time t + n when the entire Joan is paid back with
interest. If one euro of debt is incurred in period t. the expected amount A s to be repaid at
time t + n will be:
(22)
where it is the current short-term (one-period) interest rate which is known at the time the
debt is incurred. and i~+i is the future short-term interest rate expected to prevail in period
t +j .
As an alternative, the firm may finance the investment by a long-term loan with n
terms to maturity (i.e., a loan lasting for r·r periods) where the interest rate per period, i 1, is
llxed at timet when the debt is incurred. Assuming again that no interest is paid until the
loan expires at timet+ n, the amount A 1 to be repaid at that time will then be:
(23)
If the fu-m is risk-neutral, it will not worry about the uncertainty pertaining to the future
short-term interest rates but will simply choose the mode of llnance with the lowest
expected cost. Thus, if A1 >A •, the tlrm will choose to llnance the investment by a variable-
interest rate loan, but if A 1 <As it will prefer the long-term loan with a fixed interest rate.
In the latter case it would seem that monetary policy has no influence at all on the cost of
investment tinance. since the short-term interest rate controlled by the central bank does
not enter the expression for A 1• However, this ignores that the arbitrage behaviour of
tinancial investors creates a link between short-term and long-term interest rates.
We will now explore this link. Suppose that llnancial investors consider short-term
debt instruments (with one term to maturity) and long-term debt instruments (\>\lith n
terms to maturity) to be perfect substitutes for each other. Since short-term and long-term
instmments have different risk characteristics. perfect substitutability requires that
investors be risk neutral. In that case the eflective interest rate on the long-term instru-
ment must adjust to ensure that the expected returns on slwrt-tenn and long-tem1s instnt-
ments are equalized. In a tinancial market equilibrium investors must thus expect to end up
with the same stock of wealth at timet+ rz whether they buy a long-term instrument and
hold it until maturity. or whether they mal<e a series of 11 short-term investments.
reinvesting in short-term instruments every time the instrument bought in the previous
period matures. At the beginning of period t we therefore have the tinancial arbitrage
condition:
(24)
The term on the left-hand side of {24) is the investor's wealth at timet+ n if he invests in
the long-term instrument at timet and holds on to his investment. The right-hand side of
(24) measures the wealth he expects to accumulate iJ he makes a series of short-term
investments, reinvesting his principal plus interest in each period until time t + 11. In
equilibrium the two investment strategies must be equally attractive, given the perfect
substitutability of short-term and long-term financial in~truments.
Equations (22)-(24) obviously imply that As = A 1• so under risk neutrality the cost of
long-term finance is identical to the (expected) cost of short-term finance. This means th at
the long-term interest rate is influenced by the short-term interest rate controlled by the
central banl<. More precisely, according to (24) the current lm1g-term interest rate depends 011
the current and the expectedjitture short-term interest rates. This is referred to as the expect-
ations hypothesis. If the length of our period is, say, a year, a quarter, or a month , the
interest rates appearing in (24) will not be far above 0, and the approximation ln{l + i) ""i
will be fairly accurate. Taking logs on both sides of {24) and dividing through by 11, we
then get:
(25)
Equation (2 5) says that the current long-term interest rate is a simple average of the current
and the expectedjitture short-term interest rates. This relationship assumes that investors are
risk neutral. If they are risk averse, one must add a risk premium to the right-hand side of
Eq. (2 5) to account for the greater riskiness oflong-term bonds whose prices are more sen-
sitive to changes in the market rate of interest and hence more volatile.
We have so far considered only two dillerent debt instruments. In reality a large
number of securities \>\lith many different terms to maturity are traded in llnancial
markets. But the reasoning which led to Eq. (2 5) is valid for any 11;;. 2, so (2 5) determines
the entire term structure a,{ interest rates, that is, the relationship between the interest rates
on securities \>\lith dillerent terms to maturity (dillerent values of n).
aggregate demand
© The McGraw-Hill
Companies. 2005
510 PA RT 5 : TH E B UIL DING BLOCKS FO R THE S HO RT- RUN M O DEL
From the term structure equation (2 5) one can derive the so-called yield curve which
shows the ellective interest rates on instruments of different maturities at a given point in
time. According to (25) we have:
if and only if for all j = 1. 2, ... , n - 1. (26)
In other words, if Hnancial investors happen to expect no changes in future short-term

interest rates - a situation sometimes described as 'static expectations'- the interest rates
on long-term and short-term instruments will coincide, and the yield curve will be quite
flat. Figure 17.4 shows that the yield curve in Denmark did in fact look this way in the
beginning of January 2001 .
As we move from left to right on the horizontal axis. we consider instruments with
increasing terms to maturity. The Hrst point on the yield curve shows the market interest
rate on interbank credit with 14 days until maturity. This interest rate is almost perfectly
controlled by the interest rate policy of the Danish central bank (Danmarks
Nationalbank). The last point on the yield curve plots the effective market interest rate on
30-year Danish govermnent bonds. The flatness of the yield curve suggests that investors
in Denmark roughly expected constant short-term interest rates at the beginning of 2001.
A rather flat yield curve is often considered to represent a 'normal' situation w here
investors have no particular reason to believe that tomorrow will be much different from
today. But sometimes the situation is not normal. Figure 17.4 shows that short-term
interest rates were far above long-term rates on 2 August 1993. Around that date
Denmark and many other European countries sullered from the speculative attack on the
European Monetary System, the Hxed exchange rate system that existed before the forma-
tion of the European Monetary Union. To stem the capital outflow generated by fears of a
14
12
l::g 10
Q)
·:;:.. 8
Q)
>
·n 6
$
w 2 January 2001
4
0
.c:
c0
(/) (/) (/) (/)
~
.c:
c0
.c:
c0 ~>. (;;
Q)
"0 >.
<:t E
E E
(') <0 "'
Term to maturity (logarithms)
Figure 17.4: The term structure of interest rates in Denmark

Source: Oanmarks Nationalbank.
devaluation of the Danish krone. Danmarks Nationalbank drove up the 14-day interbank
interest rate to the exorbitant height of 45 per cent p.a.! The fact that long-term interest
rates remained much lower indicates that investors did not expect the extreme situation at
the short end of the market to last long.
In contrast. the yield curve had an unusually steep upward slope on 1 August 1996.
as illustrated in Fig. 17.4 . At that time it was widely expected that the pace of growth in
the European economy was about to increase significantly. Market participants therefore
expected future monetary policy to be tightened to counteract inflationary pressures, and
the expectation of higher future short-term interest rates drove current long-term rates
signillcantly above the current short rate. 11
Implications for monetary policy
What does all this imply for monetary policy? Recall that the left-hand side of (2 5) reflects
the cost of flnancing investment by long-term debt while the right-hand side represents
the cost offlnancing investment throug h a sequence of short-term loans. Moreover, even
if a real in vestment is flnanced by equity, the cost of finance is still represented by either of
the two sides of (2 5), since the opportunity cost of equity tlnance Is the rate of Interest
which the ovvners of the flrm could have earned if they had chosen instead to invest their
wealth in the capital market. Regardless of the mode of flnance, (25) thus implies that
monetary policy can only have a sign(ficmrt impact on the incentive to invest in long-lived real
assets ifit affects expectatiol'ls about.fitture short-term interest rates. For example, if the central
bank engineers a unit increase in the current short rate it which the market considers to
be purely temporary. the expected future interest rates appearing on the right-hand side of
(2 5) will be unaffected. and the interest rate on long-term debt with n periods to maturity
will only increase by 1/rr. If the short-term rate applies to an instrument with a term of one
month, and the long-term rate relates to a 30-year bond. n will be equal to 12 x 30 = 3 60.
In that case a one-percentage point increase in the short term interest rate will only raise
the long-term bond rate by a negligible 0 .0028 percentage points. i.e., less than 0 .3 basis
points! Thus there is h ardly any impact on the incentive to invest in long-lived real assets.
regardless of whether the investment is financed by long-term bond issues or by a
sequence of short-term loans. At the other end of the spectrum L~ the situation where a
change in the current short-term interest rate is expected to be permanent. According to
(2 5) the long-term interest rate will then rise by the full amount of the increase in the
short rate. This corresponds to the assumption of static expectations in (2 6) .
The difficulties of controlling the cost of long-term investment flnance through
central bank interest rate policy are illustrated in Fig. 17.5. Despite the many successive
cuts in the target short-term interest rate of the US Federal Reserve Bank (the Federal
funds target rate) undertaken during 2001 in reaction to economic recession, the long-
term interest rate refused to come down signillcantly. This suggests that market partici-
pants expected a quick economic recovery which would induce the Federal Reserve Bank
to raise its interest rate again.
11 . As mentioned earlier, in practice risk averse investors will require some risk premium on long·term bonds to com·
pensate 'or the fact that their market prices are more volatile than the prices of short·term bonds. In an equilibrium
where the market does not expect any changes in f uture monetary policy, the yield curve w ill therefore typically have
a slightly positive slope.
aggregate demand
© The McGraw-Hill
Companies. 2005
512 PA RT 5: THE BUILDING BLOCKS FO R T HE SHORT - RUN MODEL
4
"'ci
'$- 3 target rate
Jan Feb Mar Apr May Jun Jul Aug Sep Ocl Nov Dec Jan Feb
Figure 17.5: The decoupling of short-term and long·term interes: rates in the United States,
2001-2002
Source: Danmarks Nationalbank.
The fact that monetary policy works to a large extent through its impact on market
expectations explains why central banks care so much about their communication
strategies. and why market analysts scrutinize every statement by central bankers to fmd
hints about future monetary policy. In any given situation, the transmission from a
change in the central bank interest rate to the change in long-term market interest rates
will depend on market expectations. These in turn will depend on context and historical
circumstances.
14
12
10
8
'$-
6
00
<»
<»
-
<»
<»
"'<»<»
(')
<»
<»
'<t
<»
<» "'<»<» <D
<»
<»
""<»<» co
<»
<»
<»
<»
C1>
0
0
0
0
0
"' "'
- 1O·year government bond yield
- 'Signalling' interest rate of the central bank
Figure 17.6: The 'signalling' interest rate of the central bank and the ten-year government bond
yield in Denmark
Source: Danmarks Nationalbank.
In the analysis below we will ignore the complication that long-term interest rates do
not always move in line with the current short-term rate controlled by monetary policy.
In fact, we will assume that financial investors have static expectations so that i: = i 1• in
accordance with (26). As the preceding analysis makes clear. this is a strong simplifi-
cation. Yet we should not exaggerate the loss of generality implied by the assumption of
static interest rate expectations. Figure 17.6 shows that the long-term market rate and
the central bank interest rate do tend to move in tandem over the longer run, even though
they may be out of line in the short nm. Moreover. since a part of household saving is
invested in short-term interest-bearing assets, and since some business credit is short term
in nature, the short-term interest rate has a direct impact on some components of private
aggregate demand.
17.4 T.?..~. ~.~~.r..~.~~.~~.~~!?.~P.~..~~.Y.~...........................................................................................

The real interest rate: ex ante versus ex post
We are now ready to derive the relationship between the intlation rate and the aggregate
demand for goods and services. This relationship. called the aggregate demand (AD)
curve. will be one of the two cornerstones of our model of the macro economy.
The first step in our derivation of the AD curve is the specillcation of the relationship
benveen the nominal interest rate. the real interest rate and inllation. We have previously
used the popular defmition according to which the real interest rate is given by r = i - :n:.
but now we need to be more precise. For a saver or a borrower the actual real interest rate
r" earned or paid between the current period and the next one is given by:
1+i
1+r"= - - -. (27)
1 +.n:+l
The reasoning behind (2 7) is this: if the current price level is P, giving up one unit of
consumpt!on today will enable you to invest the amount Pin the capital market. Your
nominal wealth one year from now will then be P(l + i). With an inllation rate .n:+1
between the current and the next period, a unit of consumption tomorrow will cost
you P(1 +.n:+Jl, so the purchasing power of your wealth one year from now will be
only P(1 + i)/P(1 + .n:+ 1) = (1 + i)/(1 + :n:+1). Thus your real rate of return is r" = {1 + i)/
{1 + .n:+1) - 1, which is just another way of writing (2 7).
The variable r" is called the e:x post real interest rate. because it measures the real
interest rate implied by the actual rate of inllation, measured after the relevant time period
has passed ('ex post'). However, since saving and investment decisions must be made 'ex
ante', before the future price level is known with certainty. the real interest rate atl'ecting
aggregate demand for goods is the so-called ex nnte real interest rate {r) which is based on
the rate ofinllation .n:: 1 expected to prevail over the next period:
1 +i
1+r = - - -. (28)
1 +:n:: l
aggregate demand
© The McGraw-Hill
Companies. 2005
You may easily verifY that:
(29)
where the latter approximation holds as long as n:,

does not deviate too much from zero.
In the special case of static in flation expectations where agents assume that the rate of
price increase over the next period will correspond to the rate of inflation experienced
n:
between the previous and the current period. we have 1 = n . It then follows from (29)
that the ex ante real interest rate may be proxied by r =i - n, corresponding to the popular
definition of the real interest rate. Still, you should keep in mind that the more correct
specification of the real interest rate influencing saving and investment decisions is given
by (29).
Deriving the aggregate demand curve
In many countries consumer and/or business surveys provide an estimate of the rate of
inflation expected by the private sector. Several countries also have markets for indexed
bonds whose principal is automatically adjusted in accordance with the change in some
index of the general price level. For such bonds the interest rate does not have to include
an inflation premium to compensate the creditor for the erosion of his real wealth caused
by inflation. By comparing the interest rate on indexed bonds to that on conventional
non-indexed bonds of similar maturity, one may therefore obtain an estimate of the
expected rate of inflation.
In one of these ways the central bank will usually be able to measure the private
sector's expected rate of inflation. We will therefore assume that the central bank can
n:
observe the expected inflation rate 1• Alternatively, we might assume that the central
bank forms its own estimate of the future inflation rate and uses this as a proxy lor the
private sector's expected inflation rate. If the central bank and the private sector are using
n:
the same information. they will arrive at rough ly the same value of 1 . Furthermore.
since private demand depends on the ex ante real interest rater = i - .n: 1, we will assume
that the central bank sets the short-term nominal interest rate in accordance with the
following slightly modified version of the Taylor rule:
i = f+ n : 1 + lr(.n - n*) + b(y - y), (30)
implying:
r = f + h(n - rr*) + b(y - [J) . (31)
Equation (3 1) shows that if the central bank has a good estimate of the expected inflation
rate. it can control the ex ante real interest rate in the short run. We may now insert the
monetary policy rule (3 1) into the log-linearized version of the goods market equilibrium
condition (11). We then get:
,. - r
which is equivalent to the aggregate demand C!1n1e:
!J - [J = a(rr* - .n:) + z, (32)

(1 =-a h 2
- - > 0,
1+a b
(33)
1
We see from (32) and (3 3) that tile aggregate demand curve is dowm.vard-sloping in the (y, n)
space: a higher rate of inflation is associated with lower aggregate demand for output, as
illustrated in Fig. 17.7 where the aggregate demand curve is denoted by AD. The reason
for the negative slope is that higher inflation induces monetary policy makers to raise the
real interest rate. given that the parameter h in the central bank's reaction function (31)
is positive. The higher real interest rate in turn dampens aggregate private demand for
goods and services.
To identify the determinants of the position and the slope of the AD curve in the (!J, n)
plane. it is convenient to rearrange (32) as:
n = n* + (1/a)z - (1/a)(!J - y). (34)
The variable z on the right-hand side of ( 34) captures aggregate de1nand shocks. From the
definition ofz given in {33) we see that aggregate demand shocks may come from changes
in fiscal policy, reflected in g. or from changes in private sector confidence allecting the
variable v(see the dellnition ofv in {12)). A more expansionary llscal policy (a r ise in g), or
more optimistic growth expectations in the private sector (a rise in t) will shift the aggre-
gate demand curve upwards in the (y, n ) plane. Given our definitions ofvand z in (12) and
(33), the value of z will be zero under 'normal' conditions where public spending and
private sector growth expectations are at their trend levels.
According to (34) the position of the aggregate demand curve is also aii'ected by the
central bank's inllation targetn*. If the central bank becomes more 'hawkish ' in lighting
inllation (if n* falls). the aggregate demand curve will shift dow11wards.
Monetary policy influences the slope of the aggregate demand curve (1/a) as well as
its position. If the central banl< puts strong emphasis on lighting inflation and little
emphasis on stabilizing output, the parameter h in the Taylor rule will be high, and the
parameter lJ will be low. Since a =a 2 1!/(1 +a 2 b). this means that the aggregate demand
AD(h low, b high)
Figure 17.7: The aggregate demand curve

aggregate demand
© The McGraw-Hill
Companies. 2005
516 PART 5 : THE BUILDING BLOCKS FOR THE SHORT-RUN MODEL
curve will be flat (a w ill be high). On the other hand. if monetary policy reacts strongly to
the output gap and only weakly to inflation. we have a low value of h and a high value of
b. generating a steep aggregate demand curve. These results are illustrated in Fig. 17.7.
The aggregate demand curve is one of the two key relationships in our short-run
model of the macro economy. To identify the point on the AD curve in which the economy
will settle down, we need to bring in the aggregate supply side. This is the topic of the next
chapter.
17.5 .~.~.~.~:.~~Y...................................................................................................................................................
1. The aggregate demand curve (the AD curve) is derived by combining the aggregate con·
sumption and investment functions with the goods market equilibrium condition that total
output must equal the total demand for output consisting of private consumption, private
investment and government demand for goods and services. Goods market equilibrium
implies that aggregate saving equals aggregate investment. The AD curve assumes that the
private sector savings surplus {savings minus investment) is an increasing function of the real
rate of interest. The evidence supports this assumption.
2. Because aggregate demand depends on the real rate of interest, it is crucially influenced by
the interest rate policy of the central bank. Historically some central banks have followed
Milton Friedman's suggested constant money growth rule, setting the short-term interest rate
with the purpose of attaining a steady growth rate of the nominal money supply. More
recently, the interest rate policy of many important central banks has tended to follow the rule
suggested by John Taylor according to which the central bank should raise the short-term real
interest rate when faced with a rise in the rate of inflation or a rise in output. If the money
demand function is stable, the constant money growth rule has similar qualitative implications
for central bank interest rate policy as the Taylor rule.
3. The central bank can control the short-term interest rate, but not the long-term interest rate.
The expectations hypothesis states that the long-term interest rate is a simple average of the
current and expected future short-term interest rates. If a change in the short-term interest
rate has little elfect on expected futu re short-term rates, it will also have little effect on the
long-term interest rate. The ability of the central bank to influence the long-term interest rate
therefore depends very much on its ability to affect market expectations.
4. The incentive to invest in a real asset depends on the expected cost of finance over the life·
time of the asset. Under debt fin ance a long-lived asset may be financed by a long-term loan
or by a sequence of short-term loans. Risk neutral investors will choose the mode of finance
which has the lowest expected cost. When the expectations hypothesis holds, the expected
cost of fin ance is the same whether real investment is financed by equity, by long-term debt,
or by a sequence of short-term loans. As a consequence, the ability of the central bank to
influence incentives fo r long-term real investment depends on its ability to influence the long·
term interest rate which in turn hinges on its ability to affect market expectations of future
short-term rates.
5. When expectations are static, the expected future short-term interest rates are equal to the
current short-term rate. A change in the current short-term rate will then cause a correspond-
ing change in the long-term interest rate, and the yield curve showing the interest rates on
bonds with different terms to maturity w ill be completely flat. The AD curve is derived on the
simp lifying assumption that expectations are static so that the central bank can control long-
term interest rates through its control over the short-term rate.
6. Because of its empirical relevance, our theory of the aggregate demand curve also assumes
that monetary policy follows the Taylor ru le which implies that the central bank raises the real
interest rate when the rate of inflation goes up. A higher rate of inflation w ill therefore be
accompanied by a fall in aggregate demand, so the AD curve w ill be downward-sloping in
(y,.n) space. The AD curve will shift down if the central bank lowers its target rate of inflation
or if the economy is hit by a negative demand shock, due to a tightening of fiscal policy or a
fall in private sector confidence.
17.6 Exercises
Exercise 1. Topics in the theory of aggregate demand
1. Explain the concepts of the ex post real interest rate and the ex ante real interest rate. Which
of these measures is most relevant as an indicator of the incentive to save and invest? Which
of the two measures is most relevant for judging how inflation affects the distribution of
income between borrowers and lenders? How are the two measures of the real interest rate
related to the popular measure i 1 - .n 1?
2. Why does the AD curve slope downwards? Which factors can cause the AD curve to shift?
Which factors determine the slope of the AD curve? In particular, explain in economic terms
why a higher value of the parameter b in the Taylor rule makes the AD curve steeper.
Exercise 2. Interest rate setting under a constant money growth rule
When we derived the central bank's interest rate reaction function (20) under the constant
money growth rule, we assumed for simplicity that there was no growth in trend output. We
w ill now assume instead that trend output grows at the constant rate x so that:
Y"' In Y, (35)
where we have used the approximation ln(1 + x) "' x. Following Section 3 and the notation
used there, we specify the demand fo r real money balances as:
(36)
1. Show by means of (35) and (36) that the growth rate of t~e demand for real money balances
in a long-run equilibrium is approximately equal to IJX. (Hint: approximate the growth rate in real
money demand by In L- In L _1.) Show that in long -run equilibrium, the rate of inflation w ill be
given by .n = It -I]X, where It is the constant growth rate of the nominal money supply, defined
by M = (1 - f1)M_1. (Hint: you may approximate the growth rate of the real money supply by
ln(M/ PJ - ln(M_1/ P_1 ), using the fact that p , (1 + .n)P_1 . )
aggregate demand
© The McGraw-Hill
Companies. 2005
518 PART 5 : THE BUILDING BLOCKS FOR THE SHORT-RUN MODEL
We now invite you to derive an equation showing how the central bank should set the
short-term interest rate if it w ishes to maintain a constant growth rate It of the nominal money
supply, in accordance w ith Milton Friedman's recommendation. You may assume that the
economy is in long-run equilibrium in the previous period, with a nominal interest rate equal to
i_1 = r + 11 -I]X. Given that the money market c lears in every period, we then have:
M/ P 1 +p kY'1e -p;
M_/ P_1 "' 1 + .n = k'?~ 1 e-fJ<r+wqx) . (37)
2. Use (35) and (37) to show that Friedman's constant money growth rule requ ires the central
bank to set the nominal interest rate (approximately) in accordance w ith the rule:
(38)
(Hint: take logs on both sides of (37) and use the approximation (1 +!t)/ (1 + .n) ""f! - .n.)
Comment on the similarities and differences between (38) and Eq. (20) in the main text.
3. Friedman has argued that the interest sensitivity of money demand {fi) is very low (although
positive). What does this imply for the evolution over time in the nominal and real interest rate
if monetary policy makers follow the constant money growth rule? Do you see any problem in
this?
Exercise 3. Nominal GOP targeting
In the main text of this chapter we discussed the constant money growth rule wh ich is
intended to ensure a stable evolution of aggregate nominal income. Given this goal, some
economists have proposed that the central bank should not focus on the evolution of the
nominal money supply as such but rather adopt a target growth rate for nominal GOP. Such
a rule wou ld allow real GOP to g row faster when inflation falls and would require real growth
to be dampened when inflation rises. In formal terms, if the target growth rate of nominal GOP
is p , the central bank must adjust the interest rate to ensure that:
nominal GDP growt h
y-y_l+.n=f..t, (39)
where y is the log of GOP so that y- y _1 is the growth rate of real GOP. Ignoring fluctuations
in confidence and government spending (z = 0), and assuming static inflation expectations
so that the ex ante real interest rate becomes equal to i- .n, we may write the goods market
equilibrium cond ition as:
y- y= - aii- .n-1). (40)
Finally, suppose that trend output grows at the constant rate x so that:
y"" Y-1+X. (41)
1. Derive the policy rule for interest rate setting under nominal GOP targeting. How does the
interest rate react to inflation? How does it react to the lagged output gap y_1 - y _1? Explain
in economic terms why and how the parameter a 2 affects the central bank's interest rate
response to changes in the rate of inflation and in the lagged output gap.
2. Compare interest rate setting under nominal GOP targeting to interest rate setting under the
Taylor rule and under the constant money growth ru le. Explain similarities and differences. Do
you see any advantages of nominal GOP targeting compared to Friedman's constant money
growth rule? G ive reasons for your answer.
Exercise4. Topics in monetary policy
1. Explain and discuss the arguments underlying the constant money growth rule for monetary
policy. Explain the similarities and differences between the constant money growth rule and
the Taylor rule. What could be the argument for choosing a Taylor rule rather than a constant
money growth rule? Is it possible to determine by empirical analysis whether a central bank
follows a constant money growth ru le or a Taylor rule?
2. Explain the expectations hypothesis of the link between short-term interest rates and long-
term interest rates. What is the crucial assumption underlying the expectations hypothesis? Is
this assumption reasonable? Discuss the central bank's possibility of controlling long-term
interest rates through its control of the short-term interest rate.
3_ Discuss why financial market analysts study the official statements of central bankers so care-
fully and why central bankers seem to be so careful about what they say. Most observers
argue that central banks should be as open and transparent about their analysis of the
economic situation and their policy intentions as possible, but some argue that complete
openness may not be the optimal policy. Try to think of arguments for and against maximum
transparency of central banks. (As a source of inspiration, you may want to consult the article
'It's not always good to talk', The Economist, 24-30 July 2004, p. 65.)
18. Inflation, unemployment
and aggregate supply
Companies. 2005
Chapter
Inflation,
unemployment and
aggregate supply
••
•I
••
:
nflation and unemployment are two of the most important macroeconomic problems.
TnclP-P.n. thP- milin goills o f milc:roP-r.onomic: stflhili?:iltion polic:y ilrP. to figh t c:yc:l ir.i! l
• unemployment and to avoid high inllation. In this chapter we explore the relationship
between inllation and unemployment. As we sh all see, understanding the link between
these two variables is crucial for understanding how the supply side of the economy works
and how the economy reacts to shocks. In short. studying the relationship between
inllation and unemployment is fundamental lor understanding business fluctuations.
18.1 Background: a brief history of the Phillips curve

················································································································································································
For many years after the Second World War most economists and policy makers believed
that there was an inescapable trade-oil' between inllation and unemployment: u·you want
less inflation, you have to live with permanently higher unemployment. and vice versa.
Figures 18.1a and 18.1b. taken from a famous article published in 1958 by the New
Zealand-born economist A.W. Phillips, suggest why most observers came to believe in a
permanent unemployment- inllation trade-off. Figure 18.1a reproduces the curve which
Phillips lltted to describe the relationsh ip between unemployment and the rate of money
wage infl ation in the United Kingdom in the period 1861- 1913. We see that he found a
clear (although non-linear) negative correlation between the two variables. Phillips then
showed that the curve fltted to the 1861- 1913 data was able to explain the relationsh ip
between UK unemployment and wage inllation in the much later period 1948- 1957.
shown in Fig. 18.1b. Apparently Phillips had discovered a very stable and fundamental
trade-off. This trade-oil' was therefore quickly incorporated into macroeconomic models
under the name of the Phillips curve.
As illustrated in Fig. 18.2a which is based on US data on unemployment and the rate
of consumer price inllation, the Phillips curve trade-oil also seemed to exist th roughout
most of the 1960s. However, in the 19 70s the relationship broke down completely (see
Fig. 18.2b). Many times during the 1970s the US experienced a simultaneous rise in
inllation and unemployment. much to the perplexity and frustration of economic policy
520
Sorensen-Whitta-Jacobsen: I Part 5- The Building 18. Inflation, unemployment © The McGraw-Hill
Introducing Advanced Blocks for the Short- run and aggregate supply Companies. 2005
18 INFLATION, UNEMPLOYMEN T AND AGGREGA TE S UPPLY 521
~10
"'
Q)
>,
~ 8
£
"'
Q) 6
~
Q)
a 4
"'3
>, • • •
••
Q)
c
0 2
•
E
0
•
• •• • • •• •• •
Q)
a 0
c
"' •
••
.<:
~ -2 •
.,
Q)
•
c:r - 4
0 2 3 4 5 6 7 8 0 10 11
Unemployment (%)
Figure 18.1a: The Phillips Curve in the United Kingdom, 1861 - 1913
Source: Figure 1 of A.W. Phillips, 'The Relation between Unemployment and t he Rate of Change of Money Wage
Rates in the United Kingdom, 1861-1957', Economica, New Series, 25 (1 00), B lackwell Publishing, (Nov., 1958),
pp. 283-29a.
11
10
"<:'
"'
Q)
>, 9 - Curve fitted to 1861 - 1913 data
~
£.,
8
2 7
~
Q) • 52
0>
6
"'3
,.,
Q)
c 5
0
E
0 4
Q)
0>
c
"'
.<:
(.)
3
0
Q)
2
'"
c:r
0
0 2 3 4 5
Unemployment (%)
Figure 18.1b: The Phill ips Curve in the United Kingdom, 1948-1957
Source: Figure 10 of A.W. Phillips, 'The Relation between Unemployment and the Rate of Change of Money Wage
Rates in the United Kingdom, 1861- 1957', Economica, New Series, 25 (1 00), B lackwell Publishing, (Nov., 1958),
pp. 283 -29a.
Companies. 2005
522 PART 5: TH E BUILDING BLOCKS FOR THE SHORT-RUN MODEL
0' 4
~
c
0
·-; 3
£
2
1961
0+-----.-----.-----.-----.-----.----..----.-----.
0 3 4 5 6 7 8
Unemployment rate (%)
Figure 18.2a: The Phillips curve in the United States of the 1960s
Source: R.B. Mitchell, International Historical Statistics, Macmillian, 1998; and Bureau of labor Statistics.
14
12
10
~ 1975
c 8
0
·~
] 6 1969
4 1968
2 1963
1965 1964 1962 1961
0
0 3 4 5 6 7 8 9
Figure 18.2b: The breakdown of the simple Phillips curve in the United States
Source: R.B. Mitchell, International Historical Statistics, Macmillian, 1998; and Bureau of labor Statistics.
makers. The same thing happened in practically all OECD countries during that decade.
What was going on?
In this chapter we develop a theory of inflation and unemployment which offers an
explanation for the apparently stable Phillips curve trade-off before the 19 70s as well as
the relationship between unemployment and inflation in the more recent decades.
Our theory of wage and price formation will be consistent with the theory of structural
unemployment presented in Part 4 . As we shall see, this framework can explain the
Sorensen-Whitta-Jacobsen: I Part 5- The Building 18. Inflation, unemployment I © The McGraw-Hill
18 IN FLATIO N, UN EMPLO YM EN T AND AG G RE GATE S UPPLY 523
short run link between inflation and unemployment as well as th e factors determin ing the
long-run equilibrium rate of unemployment, the 'natural' rate. The relationship we shall
arrive at is the so-called expectations-augrnented Phillips curve,
n = n' + a(u - u). (l > 0. (1)

where n is the actual rate of inflation. n ' is the expected inflation rate, u is the actual rate
of unemployment, and u is the natural unemployment rate.
Many roads lead to the expectations-augmented Phillips curve. This chapter will take
you down some of these roads. In Section 2 we offer a theory of the expectations-aug-
mented Phillips curve in line with the theory of trade union behaviour introduced in
Chapter 13. 1 However, the same qualitative results may be obtained from the theory of
elllciency wages presented in Chapter 12, as we shall see in Chapter 2 3.
The model of inilation and unemployment presented in Section 2 assumes that
nominal wages are rigid in the short run. In Section 3 we show that the expectations-aug-
mented Phillips curve may also be derived from a model of a competitive labour market
with fully flexible wages and prices. By comparing the models in Sections 2 and 3, we are
able to high light how nominal rigidities exacerbate the employment fluctuations which
occur when economic agents underestimate or overestimate the rate of inflation.
18•2 Nominal rigidities, expectational errors and employment

fluctuations
Inflation is a continuous rise in the general price level. A theory of inflation therefore
requires a theory of price formation. Since prices depend on the cost of inputs, and since
labour is the most important input, our theory of price formation will build on a theory of
wage formation. The theory will allow lor imperfect competition in the markets lor goods
as well as labour. Introducing imperfect competition in output markets complicates the
analysis, but in return it enables us to illustrate bow stmctural changes in product
markets allect inflation and the natural rate of unemployment.
In Book One where we focused on the long run, we assumed that agents had correct
expectations about the general level of wages and prices, as must be the case in any long-
nm equilibrium. By contrast, in the present short-run context we assume that people do
not have perfect information about the current general price level. As we shall see, this
means that employment and output may deviate from their long-run equilibrium levels.
This section assumes that wages are 'sticky' in the short run, being set by trade unions.
We will therefore start with a description of trade union behaviour and wage formation.
The trade union's objective
We consider an economy which is divided into a number of dillerent sectors each pro-
ducing a dillerentiated product. Workers in each sector are organized in a trade union
which monopolizes the supply oflabour to all Hrms in the sector. Because of its monopoly
1 . The exposition in this chapter does not assume that you have already studied C hapter 13, so you should still be able
to understand all parts of the present chapter even if you have not had the opportunity to s:~o th rous:~ h Book One.
Companies. 2005
524 PA RT 5: TH E BUILDI NG BLOCKS FOR TH E SHO RT-RUN MODEL
position, the trade union in each sector may dictate the nominal wage rate to be paid by
employers in that sector, but employers have the ;right to manage', that is. they can freely
choose the level of employment. For simplicity, we assume that the number of working
hours lor the individual worker is fixed. so total labour input is proportional to the number
of workers employed.
Workers in sector i are educated and trained to work in that particular sector. so they
cannot move to another sector to look lor a job. If a worker fails to Hnd a job in his sector.
he therefore becomes unemployed. His real income will then be equal to the real rate of
unemployment benefit b. 2 An employed worker in sector i earns the real wage w i Tfi;/P. =
where V\'; is the sectoral money wage and Pis (an index of) the general price level, so his
net income gain from being employed is wi - b. The trade union for sector i cares about this
real income gain for its employed members, but it also cares about the total number of jobs
Li secured for the membership. We formalize th is by assuming that the union sets the
nominal wage rate with the purpose of maximizing a utility function Q of the form:
17 > 0. (2)
The parameter IJ reflects the weight which the union attaches to high employment
relative to the goal of a high real wage for employed union members. The more the union
is concerned about employment relative to wages. the higher is the value of IJ. In the
benchmark case where IJ = 1 (corresponding to the cases analysed in Chapters 1 and 13 ).
the union is simply interested in the aggregate net income gain obtained by employed
members.
When setrtng the wage rate, the union must account for the fact that a h igher real
wage will lower the employer's demand for labour. Our next step is to derive this
constraint on the union's optimization problem.
Price setting and labour demand
The representative employer in sector i uses a technology described by the production

function:
0 < a < 1, (3)
where Yi is the volume of real output produced and sold in sector i, and B is a productivity
parameter. Since we are concentrating on the short run where the capital stock is tixed.
we have not included capital explicitly in the production function . 3 According to (3) . the
marginal product of labour. iVrPL. is:
MPLi =dY/dLi = (1 - <l)BLj". (4)
which is seen to diminish as labour input increases. due to the fixity of the capital stock.
2. In Chapters 12 and 13 and in Section 4 of Chapter 1, where we focused on the long run, we assumed that workers
who fail to find a job in their initial sector have time to retrain so that they can nove into other sectors to look for alter-
native emp loyment opportunities. In that case the expected real income obtainable by a sector i worker w ho loses his
initial job is " = (1 - u)w+ ub, where u is the general unemployment rate, w is the average real wage outside sector i,
and where the employment rat e 1 - u represents the probability of finding a jo~ outside sector i. In Exercise 1 you are
asked to consider such a case with intersectoral labour mobility and to show !hat this case leads to the expectations-
augmented Phill ps curve as well.
3. In B ook One we worked w ith the Cobb - Douglas production function Y= BK• L •- •. Equation (3) is just a version of
this produc tion function w here we have fixed the capital stock Kat unity.
The employer representing industry i produces a dill'erentiated product and therefore

has some monopoly power, so we assume that he faces a downward-sloping demand
curve of the form:
P;)-" y
( -
Y I = -p II'
(J > 1. (5)
This demand curve has a constant numerical price elasticity of demand equal to
cJ = - (dY;/dP1)(PjY1), where P1 is the price charged per unit of ¥1• The variable Y is total
GDP. and n is the number of dillerent sectors in the economy. Aggregate output Y is a
measure of the total size of the national market. and Y/n is the market share captured by
each industry if they all charge the same prices (so that P1 = P). 4 The total revenue of llrm
=
i is TR 1 P; ¥ 1• so according to (5) its marginal revenue (the increase in total revenue from
selling an extra unit of output) will be:
(6)
From microeconomic theory we know that a prollt maximizing llrm will expand output to
the point where marginal revenue equals marginal cost. MR 1 = MC1• Because labour is the
only variable factor of production . marginal cost is equal to the price of an extra unit of
labour - the nominal wage rate, W, - divided by labour's marginal product, MPL 1• since
MPL1 measures the additional units of output produced by an extra unit of labour. Thus
MC1 = WJ MPL 1• From (4) and (6), the necessary condition for maximization of profits,
MR1 = MC,. therefore becomes:
which is equivalent to:
~
P.1 = lll p . wI ) mP =__!!__ > 1. (7)
((1 - a)BL;" ' o- 1
Equation 17) shows that the profit-maximizing representative firm in sector i will set its
price as a mark-up over its marginal cost. Our previous assumption cr > 1 guarantees that
the mark-up factor mt' is positive and greater than one. The price elasticity. o, is a measure
of the strength of product market competition. The higher the elasticity, the greater is the
fall in demand induced by a higher price (the Hatter is the demand curve), and the lower is
the mark-up of price over marginal cost. In the limiting case where the price elasticity
tends to infinity, the price is driven down to the level of marginal cost (a~ oo ~ mP ~ 1),
corresponding to perfect competition.
We can now derive the labour demand curve of sector i. showing the relationship
between the real wage VI';/P claimed by the union in sector i and the level of employment
in that sector. Dividing by P on both sides of (7) gives the relative price, PJP, of sector i's
4. As we mentioned in Chapter 12, the demand curve (5) may be derived from the solution to the consumer's problem
of utility maximization if utility functions are of the CES form. In that case the paramet er a is the representative
consumer's elastic ity of substitution between good i and any other good. See Exercise 3 in Chapter 1 2.
Companies. 2005
product. Inserting this rjr into ( 5) t hen gives production, 1';· in sector i. Pinally. we can
use (3) to compute how much employment. L;. is needed to produce that level of output.
Performing these operations. we end up with:
L; (~)'i"(B(l - a))'(W;)-~
=
nB m~' P
c = 1+a(a -
(1
1)
> 0. (8)
The numerical real wage elasticity of labour demand at the sectoral level (dellned as
- (dL;/d(W;/P))((W;/P)/LJ) is equal to the constant c . From the expression for c you may
veri!)' that a higher price elasticity of product demand (tougher competition in product
markets) increases the wage elasticity of sectoral labour demand. ThL~ is intuitive: a rise in
the wage rate will drive up the output price by raising the llrm's marginal cost. The higher
the price elasticity of output demand. the greater is the fall in sales and output. so the
greater is the resulting fall in labour demand.
Wage setting
Th e labour demand curve (8) implies that employment in sector i is a declining function.
L;(w;). of the real wage. W; =W;/P. The union's utility function (2) may therefore be
written as:
Q(w) = (w; - h)[L;(w;)]'1. (9)
Suppose for the moment that the union has perfect infom1ation about the current price
=
level so that it may perfectly control the real wage W; W;/P via its control of the money
wage W;. The union will then choose W; so as to maximize Q (w;). The necessary condition
lor a maximum, dQ(w;)/dw; = 0 . is L7+ (w; - h)T]Lr 1 (dL;/dw'J = 0, which is equivalent to:
Using the fact that (dLJdw;)(w;/L;) = - c. we may rewrite this expression as:
T]€
W; = m "' · h. mw= - - . (10)
rtt: - l
According to (10) the union's target real wage is a mark-up over the opportunity cost of
employment. The opportunity cost of employment is the rate of unemployment beneilt h.
since this is the income a worker forgoes by being employed rather than unemployed. To
secure that {10) actually implies a positive real wage. we assume that rJ€ > 1. It then
follows that the wage mark-up factor, m ... is greater than 1.
Equation 110) implies that the union's real wage claim will be lower the greater the
weight it attaches to the goal of high employment, i.e .. the higher the value of '7· It also
follows from (10) that the target real wage will be lower the higher the elasticity oflabour
demand, c . The reason is that a higher labour demand elasticity increases the loss of jobs
resulting from any given increase in the real wage. Finally. we see from {10) that a higher
rate of unemployment benellt drives up the target real wage because it reduces the income
loss incurred by those union members who lose their jobs when the union charges a
higher wage rate.
We hnve so far assumed that the union has perfect information about th e current
price level and therefore perfectly controls the real wage V.,7;/P through its control of the
money wage rate, H';. However, in practice, nominal wage rates are almost always pre-set
lor a certain period of time, that is, in the short run the nominal wage rate is rigid.
Moreover, at the start of the period when wages are set, trade union leaders cannot
perfectly foresee the price level wh ich will prevail over the period during which the
nominal wage rate will be lb'ed by the wage contract. A trade union setting the wage rate
at the start of the current period must therefore base its money wage claim on its expect-
atimz of the price level which will prevail over the coming period. Given that the union
strives to obtain the real wage specified in (10), it will then set the morzey wage rate so as
to achieve an expected real wage equal to the target real wage m"' b. 5 If the expected price
level lor the current period is p •·, the nominal wage rate set by the union at the start of the
period will thus be:
(11)
Having developed a theory of wage and price setting as well as a theory of labour
demand. we are now ready to derive the link between inllation and unemployment.
The expectations-augmented Phillips curve
Equation {11) implies that the actual real wage may be written as H';/P = (P'/P)m wfJ.
Inserting this expression into the labour demand curve (R) and rearranging, we obtain the
level of employment in sector i:
y)</a(B(1 -a) P)"

L; = (;i3 m 'm wb P' ·
1
(12)
The higher the actual price level relative to the expected price level. P/P', that is, the more
the trade union underestimates the price level. the lower is its nominal wage claim relative
to the actual price level, so the lower is the real wage and the higher is the level of sectoral
employment. as we see from (12).
We will now show that a similar qualitative relationship between employment and
the ratio of actual to expected prices will prevail at the aggregate level. In doing so we will
assume that all sectors in the economy are symmetric so that output and employment in
each sector are given by Eqs (3) and (12), respectively, where all the parameters as well as
the ratio P/P' are the same across sectors. Total employment (L) will then beL = HL;. and
total GDP will beY = 11 Y; = rzBL/ -'". Substituting the latter expression into (12) and com-
puting L = nL;. we get
B(1 - a) P)
(
L/a
L=nL =n · ·- (13)
' mPm'" b P' '
where we have used thedellnition of € given in (8) according to which 1 - €(1 - o.)/rr = 0.€ .
5 . We assume for simp lic ity that the union has a correct estimate of the le\'el of b. For example, we may assume that the
nominal rate of unemp loyment benefit is automatically indexed t o t he current price level so as to protect its real value.
The union will then be able to forecast the level of t he real rate of unemployment benefit even if it cannot perfectly
foresee ttB price level. In Exercise 1 you are asked to consider the alternative case w here the union does not have
perfect information about the real value of the nominal rate of unemplovment benefit.
Cl l Sorensen-Whitta-Jacobsen:
Companies. 2005
Note that since the real wage w = W/P is the same in all sectors and equal to (P'/P)m'"IJ, we
may also write (13) as:
L
=
II
(B(1-a)) /"(w)-
Ill
1'
1
p
1
/"
. (14)
This expression shows that at the aggregate level the numerical real wage elasticity of
labour demand is 1/a, whereas at the level of the individual sector we found it to be equal
to c = a/[1 + a((1 - 1)]. In Exercise 2 we ask you to provide an intuitive explanation for
this difference between the labour demand elasticities at the macro and the micro levels.
In a long-run equilibrium expectations must be fulfilled. Inserting P' = P into (13 ), we
therefore obtain the long-run equilibrium level of aggregate employment, [, also called
the 'natural' level of employment:
_ (B(1
L = r1 -a))
m ' mwb 1
1
'"
• (15)
Equation (15) gives the level of employment which will prevail when price expectations
are correct so that trade unions actually obtain their target real wage. Dividing (13) by
(15). we get a simple relationship between the actual and the natural level of employment:
(16)
If the aggregate labour Ioree is Nand the unemployment rate is 11 . it follows by definition
=
that L (1 - u)N. Similarly, the 'natural' unemployment rate, fi, is defined by the relation-
=
ship [ (1 - fi) N. Substitution of these identities into (16) giYes (1 - u)/(1 - fi) = (P/P') 1'" ·
Tal<ing natural logarithms on both sides and using the approximations h1(l - u)"" - u and
ln(1 - fi) "" - u. we get:
p = p' + a(u - u), p =In P, p' = In P".
Subtracting p _1 =In P_ on both sides finally gives:

1
n = n " + a(u - u), n=p - p_" n " =p'- P-1' (17)
where the subscript '- 1' indicates that the variable in question refers to the previous time
period. Recalling that the change in the log of some variable roughly equals the relative
change in that variable, it follows that n is the actual rate of inflation whereas n • is the
expected rate of inflation, assuming th at agents know the previous period's price level p_1
when they form their expectation about the current price level.
Equation (17) is a key macroeconomic relationship called the expectations-augmented
Phillips cun1e, 6 and it provides the link between inflation and unemployment we have
been looking for. It shows that for any given expected rate of inflation, a lower level of
unemployment is associated with a higher actual rate of inllation, and vice versa. More
precisely. we see from (17) that unal'lticipated inflation (n > n ') will drive unemployment
6. The theory of the expectations-augmented Phillips curve was developed alnost simultaneously by the US econo-
mists Milton Friedman and Edmund Phelps. See Milton Friedman, 'The Role of Monetary Policy', American Economic
Review, 58, 1968, pp. 1-17, and Edmund S. Phelps, 'Money-Wage Dynamics and Labor Market Equilibrium',
Journal of Political Economy, 76, 1968, pp. 678-71 1.
below its natural rate. The reason is that an unexpected rise in the rate of inflation causes
the real value of the pre-set money wage rate to fall below the target real wage of trade
unions, thereby inducing firms to expand employment beyond the natural level.
The simple versus the expectations-augmented Phillips curve
If we set the expected inilation rate in (17) equal to 0, we obtain a version of the simple
Phillips curve presented in Section 1, describing the unemployment- inllation trade-oil'
discovered by Phillips:
n = a(n - u). (18)
We may now oller an explanation why the simple unemployment- inflation trade-on·esti-
mated by Phillips broke down in the US and elsewhere in the OECD from around 19 70.
Over the long historical period considered by Phillips - from around 18 60 to the 19 50s-
there was no systematic tendency for prices to rise lor extended periods of time. as you can
see from Fig. 18.3. Because of this long experience of approximate price stability, it was
natural for economic agents to expect prices to be roughly constant. In such circum-
stances where n ' = 0, Eq. (17) does indeed predict that a lower unemployment rate will
always be associated with a higher inflation rate. and vice versa.
However, towards the end of the 19 60s inflation had been systematically positive and
gradually rising for several years, so people started to consider a positive inflation rate as a
normal state of affairs. As a consequence. the expected inflation rate started to increase.
According to (17) thls tended to drive up the actual rate of inflation associated with any
given level of unemployment, just as portrayed in Fig. 18.2b which showed that many
years during the 19 70s were characterized by simultaneous increases in inflation and
unemployment. There were also other reasons for these developments, such as dramatic
1860 = 100
3500
3000
2500
I
2000
I
1500 L
1000
I
500 _________., J
........
0
g R 0 0 0 0 0 0 0 0 0 0 0
co co
co
co
0>
co
0
0> 0> "'0> (')
0>
'<t
0>
I!)
0>
<0
0>
co
0>
0>
0>
Figure 18.3: The consumer price index in the United Kingdom, 1860-1 993
Source: B.R. Mitchell, /ntemational Historical Statistics: Europe 1750-1993, Macmillan Press, 1998.
Companies. 2005
increases in the price of oil due to turmoil in the Middle Bast. but rising inllation expect
ations probably played an important role in the breakdown of the simple Phillips curve
from the end of the 19 60s.
The implication of all this is that the simple negative Phillips curve relationship
between inflation and unemployment is a sl!ort-run trade-ofJ which will hold only as long
as the expected rate of inflation stays constant. For this reason the simple downward-
sloping Phillips curve (dellned for a given expected rate of inflation) may also be called the
slwrt-nm Phillips curve. Whenever the expected inflation rate n ' increases, the short-run
Phillips curve will shift upwards, as illustrated in Fig. 18 .4 which shows three dillerent
short-run Phillips curves, each corresponding to dilferent levels of expected inflation. In a
long-run equilibrium the expected inflation rate equals the actual inflation rate. n ' = n .
According to (17) this means that only one unemployment rate - the natural rate u- is
compatible with long-run equilibrium. We may say that the long-nm Phillips curve is
vertical. passing through u = U, as indicated in Fig. 18.4 . Hence there is no permanent
trade-ofl' behveen in flation and unemployment.
Outside long-run equilibrium, the expected inflation rate dillers from the actual intla-
tion rate. If such expectational errors persist. it is natural to assume that economic agents
will gradually revise their expectations as they observe that their inflation forecasts turn
out to be wrong. One simple hypothesis encow1tered in the previous chapter is that people
have static expectatio11s, expecting that this period's inflation rate will correspond to the
rate of inflation observed during the previous period:
(19)
Figure 18.4: The expectations-augmented Phillips c urve

Introducing Advanced Blocks for the Short-run and aggregate supply Companies. 2005
This hypothesis means that agents will change their inflation forecasts whenever they
observe a change in last period's inflation rate. Substitution of (19) into (17) gives:
&r =JT. - n _1 = (l(i7 - u). (20)
which shows that inflation will accelerate when unemployment is below its natural rate,
and decelerate when unemployment is above the natural level. To prevent inflation from
accelerating (or decelerating). unemployment will thus have to be kept at its natural rate.
For this reason the natural rate is sometimes called the 'Non-Accelerating-Inflation-Rate-
oi~Unemployment', or just the NAIRU, for short.
Another important implication of the expectations-augmented Phillips curve (17) is
that there is nominal inertia: if unemployment is at its natural rate, the infl ation prevailing
today will automatically continue tomorrow because it is built into expectations. To bring
down inflation. it is necessary to push the actual unemployment rate above the NAIRU for
a while.
What determines the natural rate of unemployment?
It is obvious that the natural rate of unemployment plays an important role in our theory
of inflation. given that inflation will tend to rise if unemployment is pushed below that
level. But what determines the natural rate? Our expression (15) for the natural level of
employment provides the key to an answer. Recall that L = (1 - i7)N. For simplicity, let us
set the number of workers in each sector equal to 1 so that the total labour force becomes
N = n. implying'[= (1 - u)n. Inserting (15) into this expression and rearranging. we get:
B(1 - (l)) l/a

i7 = 1 - . (21)
( m1'mwb
Equation (21) shows that the natural unemployment rate depends on the level of the real
unemployment benefit. b. among other things. It is reasonable to assume that the govern-
ment allows the rate of unemployment benellt to grow in line with real income per capita,
at least over the longer run. From Book One we know that the long-run growth rate of per-
capita income will equal the growth rate of total factor productivity B. We will therefore
assume that the level of unemployment benefits is tied to the level of productivity so that
b = cB. where c> 0 is a parameter reflecting the generosity of the system of unemployment
compensation. Substituting cB for b in (21), we get the following expression for the
natural rate of unemployment, where we assume that the combination of parameter
values ensures a positive value of II:
Il = 1 - - - -
1_a )I/«. (22)
( mPmwc
According to (22) the natural unemployment rate is higher the higher the mark-ups in
wage and price setting. and the more generous the level of unemployment benellts (the
higher the value of c) . A rise in m1' = o/(a - 1) reflects a fall in the representative llrm's
price elasticity of demand (a) which means that it tal<es a larger cut in the firm 's relative
price P/P to obtain a given increase in sales. To sell the extra output produced by an extra
Companies. 2005
worker. the firm must therefore accept a larger price cut the lower the value of cJ. For any
given wage level, the pront-maximizing level of employment will th us be lower the lower
the value of a. This is why the natural unemployment rate will be higher the higher the
mark-up factor mP.
A fall in a will also increase the wage mark-up, since the sectoral labour demand elas-
=
ticity E' a/[ 1 + a( cJ - 1)] is increasing in a. and since m w = IJ E' /( IJ E' - 1) is decreasing in E' .
The intuition for this rise in the wage mark-up is that a lower price elasticity of demand for
the output of the representative firm reduces the drop in sales and employment occurring
when a higher union wage claim drives up the llrm's marginal cost and price. Hence it
becomes less costly (in terms of jobs lost) for the union to push up the wage rate. and this
invites more aggressive wage claims.
The representative fum's price elasticity of demand reflects the degree of competition
in product markets. The greater the number of competing llrms in each market. and the
greater the substitutability of the products of dillerent firms, the tougher competition will
be. and the greater will be the price elasticity of demand faced by the individual firm or
industry. Thus our analysis shows that a lower degree of competition in product markets
(n lowP.r rr) wi ll spil l ovP.r to thP. In hour mnrkP.t nnit rnisP. thP. nnturnl rntP. of tmP.mploy-
ment, partly because it lowers labour demand, and partly because it induces more aggres-
sive wage claims. This is an interesting example of how imperfections in some markets
may exacerbate imperfections in other markets.
It is worth noting bNo more points from (22) . First, a greater union concern about
employment, reflected in a higher value of the parameter TJ. will reduce the natural rate of
unemployment by lowering the wage mark-up m"' = TJE' /(IJE' - 1). Second. the level of pro-
ductivity B does not allect the natural rate of unemployment. This prediction is in line with
empirical observations. As illustrated by Fig. 11.1 in Chapter 11, the unemployment rate
tends to fluctuate around a constant level over the very long run despite the fact that pro-
ductivity is steadily growing over time. However, as we shall see later in this chapter, short-
run fluctuations in productivity growth do affect the short-run unemployment-inflation
trade-off.
Nominal price rigidity
For simplicity. our model of wage and price setting assumes that while nominal wages are
rigid in the short run, output prices adjust immediately to changes in demand and
marginal costs. This is a way of capturing the stylized fact that nominal wages tend to be
fixed for longer periods of time than most goods prices. But in reality many output prices
are also held constant for considerable periods. as we noted in Chapter 1. In that chapter
we also saw that small 'menu costs' of price adjustment- such as the costs of printing new
price catalogues and communicating new prices to customers- may make it suboptimal
for firms to adjust prices too frequently . When you interpret our theory of inflation and
unemployment. you should therefore keep in mind that a sluggish adjustment of inflation
- and hence a sluggish adjustment of unemployment to its natural level - may not only be
due to a slow adjustment of nominal wage rates. It may also reflect that it takes time for
changes in expectations to feed into the actual price level because it is costly for firms to
change prices too frequently.
The expectations-augmented Phillips curve in a competitive

labour market
The theory of inflation and unemployment presented above included two elements which
are typically used to explain how it is possible for economic activity to deviate from its
long-run equilibrium level: expectational errors (erroneous price expectations) and
nominal rigidity. We allowed for nominal rigidity by assuming that nominal wage rates
are pre-set at the start of each period and do not adjust within the period even if labour
demand changes. In other words, within each period the nominal wage rate is Hxed,
although it adjusts between periods as price expectations change.
In this section we '>Vill try to deepen your understanding of the importance of expec-
tationals errors and nominal rigidity for explaining fluctuations in employment. We '>Viii
show that while expectational errors are both necessary and sulllcient to generate devia-
tions of unemployment from its natural rate. nominal rigidity is not necessary but '>Viii
amplify the tluctuations in employment caused by expectational errors. To demonstrate
this. we '"'ill analyse the response of employment to unanticipated in flation in a model
w ith fully flexible nominal wages and compare th is model with th e one developed above
where nominal wages are ;sticky' in the short run.
The link between inflation and employment in a competitive labour market:

the worker·misperception model
To high light the role of nominal rigidity, it is instructive to consider the link between infla-
tion and employment which would prevail if the labour market were cmnpetitive, that is, if
nominal and real wages were fully flexible, adjusting instantaneously to balance labour
supply and labour demand.
In a competitive labour market there are no trade unions. Our wage setting equation
(11) specifying union wage claims is therefore replaced by a labour supply curve showing
how workers adjust their labour supply in response to changes in the expected real wage.
To facilitate comparison with the trade union model considered above, we continue to
assume that each employed worker works a ftxed number of hours which we may denote
by H. Changes in aggregate labour supply will then take the form of more workers
entering the labour market or some workers exiting the market. Suppose th at worker j
requires a minimum real wage IV; to be willing to sacrifice H hours ofleisure by taking a
job. In that case he will only enter the labour market if the expected real wage, W/P', is at
least equal to W;. Suppose further that dillerent workers have dillerent valuations of
leisure. with some requiring only a low real wage to be willing to accept a job, while others
require a high real wage to be willing to enter the labour market. If there is a continuum
of required min imum real wages. the number of workers entering the labour market. L'',
will rise continuously as the expected real wage increases, implying an aggregate labour
supply function of the form L" = f (V.,TjP') . .f' > 0 . For simplicity, let us assume that the
distribution of the taste for leisure across workers (the distribution of required real wages)
is such that the functionf(W/P') has a constant elasticity¢ with respect to the expected
real wage. We then get the aggregate labour supply function:
(P'W)"' (
p )"'
Ls= z - = Zw·-
P" '
z > o. (2 3)
Companies. 2005
where Z is a constant reflecting the size of the population, and where. you recall, th at
=
w HT/P is the actual real wage. Equation (23) makes the reasonable assumption that the
worker knows his nominal wage rate W when he accepts a job, but he does not have
perfect information on the current price level when he makes his labour supply decision,
so he must base his decision on his expectation of the current price level.
Aggregate labour demand Ldis still given by Eq. (14) which may be written as:
B(l - a))l/a
X= 11 - - - (24)
( mP
In a competitive labour market, the real wage wwill adjust to balance supply and demand,
Ls = Ld, implying:
IV = ( ~) 1/ (&+ 1/ a) (:e)-</>/(</>+ 1/a).

(25)
The equilibrium real wage found in Eq. (2 5) may be inserted into (24) to give the level of
employment in the competitive labour market:
p ) (4>/a)/(</> + 1/ a)
L= x<~>M+ I/a) z<1/ a)/ (<t> + 1/ a) _ • (26)
( p··
Figure 18.5 illustrates how the equilibrium levels of w and L are determined by the
intersection of the aggregate labour supply curve (23) and the aggregate labour demand
curve (24). The natural employment level [ is found at the equilibrium pointE where
w
L·= zvl'
L' =Z(w/.'t
wo
Ld = Xw- 1/a
Lo L
Figure 18.5: Labour market equilibrium in the worker-misperception model

price expectations are correct (P'' = P) so that th e labour supply curve (2 3) collapses to
L' Zw''.
=
In the equilibrium E0 employment is above the natural level because workers under-
estimate the price level (P > P'). Whenever there is a change in the ratio of the actual to
the expected price level, P/P' , the labour supply curve will shift, generating new short-run
equilibrium levels of the real wage and employment. This model of the labour market is
sometimes called 'the worker-misperception model' because it postulates that employ-
ment fluctuations are driven by workers' misperceptions of the price level. th at is, by fluc-
tuations in P/P''.
Natural employment may be found from (26) by setting P' = P. We then get:
(27)
Dividing (26) by (27) gives:
~ = ( ; , r </>/a)/(</>+ l/a).
(28)
We sec th at (28) has exactly the same form as our earlier (16) which was derived from the
model with union wage setting. The only difference is that the coef1lcient 1/a h as now
been replaced by (¢/a)/(¢ + 1/a). By taking logs on both sides of (28) and using the
approximations ln(1 - u) "" - u and ln(1 - IT) ~ - u, we can still derive an expectations-
augmented Phillips curve of the fonn :n = :n'' + fl(fi - u), where& is a constant. This shows
that the expectations-augmented Phillips curve is quite a general relationship which does
not assume a particular market structure. But as we shall demonstrate below, it is quite
important for the qunntitative relationship between employment and unanticipated
inflation whether the labour market is competitive or not.
The competitive versus the unionized labour market
To see how unanticipated inflation all'ects employment in the competitive and in the
unionized labour market, we take natural logs on both sides of (28) and (16) and use the
dellnitions of the actual and the expected inflation rate, :n ln P - In P_1 and =
:n' =In P"- h1 P_1• We then obtain the follm>Ving results:
Competitive labour market: _=(

ln L- In L 1/a ) (:n - :n'), (29)
1 + 1/a¢
Unionized labour market: In L - In L= (l/a)(:n - :n'') . (30)
Since (1/a)/(1 + 1/a¢) < 1/a, these equations show that for any given amount of
unanticipated inflation, :n - :n', the percentage deviation of employment from its natural
level. In L- In L, \>Viii be larger in the unionized than in the competitive labour market.
This may be explained as follows. Suppose that, due to some unanticipated positive shock
to aggregate demand for goods, the actual price level rises unexpectedly, driving :n above
:n' . This increase in the price level will erode the real wage. W/P, thereby stimulating the
demand lor labour. In the competitive labour market the increase in labour demand
immediately drives up the flexible money wage, generating an increase in the expected
real wage, W / P' , which induces workers to supply more labour. Because of the rise in W,
Companies. 2005
the fall in the actual real wage generated by the rise in P will be dampened, which in turn
dampens the increase in labour demand and employment However, in the unionized
labour market the nominal wage is fixed at the start of the period, and consequently there
is no dampening effect on labour demand stemming from an increase in W. Hence the
short-run increase in employment generated by any given amount of unanticipated
inflation is greater in the unionized than in the competitive labour market. 7
We may also say that individual labour supply in the unionized labour market is infi-
nitely elastic. Because there is involuntary unemployment, workers are willing to increase
their labour supply in response to an increase in labour demand even if the expected real
wage stays constant or falls (as long as the expected real wage does not fall too much). By
contrast, in the competitive labour market there is no involuntary unemployment, so
labour supply will only increase if the expected real wage goes up. This difl'erence in the
short-run flexibility oflabour supply explains why employment fluctuates more strongly
in the competitive than in the unionized labour market. To convince yourself of this. note
from (29) and (30) that ifindividual labour supply in the competitive labour market were
infin itely elastic so that¢~ oo, the two equations would be equivalent, implying the same
employment response to any given amount of unanticipated inflation.
Equations (29) and (30) also demonstrate the basic point that expectational errors
*
(n n '') are both necessary and sulllcient to cause deviations between the actual and the
natural level of employment. In other words, it is not really necessary to assume nominal
rigidities in order to explain why employment sometimes deviates from its trend level.
However, our analysis shows that once expectational errors occur. nominal rigidities will
ampl!fy the resulting fluctuations in employment. In the real world money wage rates are
in fact typically flxed by wage contracts for a certain period of time even in non-unionized
labour markets. It is therefore realistic to assume some amount of short-run nominal
wage rigidity, so we '>Viii continue to work with our model with a pre-set nominal wage
rate.
§.~P.P~Y...~.~.<?.~.~~~..................................................................................................................................... ..
Our expectations-augmented Phillips curve (20) postulates a strict deterministic relation
between the unemployment rate and the change in the rate of inflation. In this section we
shall see that the link between these two variables is not really that tight. The reason is
that the labour market is frequently hit by shocks which generate 'noise' in the relation-
ship between unemployment and inflation. These so-called supply shocks may take the
form of short-run fluctuations in our parameters m P, mw and B around their long-nm
trend levels. Below we will extend our model of unemployment and inflation to account
for supply shocks.
In Section 2 we assumed th at the rate of unemployment benefit is tied to the level of
B (since productivity determines long-run income per capita). In that case we should
observe substantial short-run fluctuations in unemployment benetlts as B oscillates
7. To put it another way, the short·run aggregate supply curve to be derived in Section 6 is flatter w hen nominal wages
are rigid than when they are flexible.
Sorensen- Whitta- Jacobsen: I Part 5- The Building 18. Inflation, unempl oyment I © The McGraw-Hill
around its long run growth trend. However, in practice benetlts do not move up and down
in this way. It is reasonable to assume that benefits are instead linked to the underlying
trend level of productivity, denoted by B. which evolves gradually and smoothly over time.
This is equivalent to assuming that benefits are allowed to rise in line with the underlying
trend growth in per-capita income.
The magnitude m"' b in the denominator on the right-hand side of (13) is the
representative trade union's target real wage. Given our new assumption that b = cB, the
target real wage becomes m"' cB. Inserting this in (13 ), we lind that the actual level of
employment is given by:
B(l - a) P ) 1/"
L= n· ( m~'m"'cB. P" · (31)
Our next step is to redefine the natural level of employment. Specifically. we now define
natural employment as the level of employment which will prevail when expectations are
lullilled and when productivity as well as the wage and price mark-ups are all at their
'normal' long-run trend levels. Denoting the normal mark-ups by mP and m"' and
remembering th at b = cil it follows that the natural employment level previously stated in
(15) now modifies to:
1
_ ( 1- a ) 1"
L= n · . (3 2)
ffi P • lfi w · C
Dividing (31) by (32) and usingthelacts thatL= (1 - z.I)N and [ = (1 - U)N . weget: 8
1- (B ·
u= ffi P • ffi w . !._) 1'".
1
(33)
1- U Bm1'm"' P'
Taking logs on both sides of (33) and using the approximations ln(1 - u) ""-u and
ln(1 - fl) "' - u plus the definitions :n = In P - In P_1 and :n' = In p •· - In P_1, we end up
with:
n = n' + a(IT - u) + s. s =In (-mP

mP
)+ In (m"'
- .) -
m"
('BB)
In = . (34)
Equation (34) is an expectations-augmented Phillips curve, extended to allow for

supply shocks. The speciHcation of the supply shock variable. s. shows th at a positive
shock to inl1ation occurs if the wage mark-up or the price mark-up rises above its
normal level. whereas a negative shock to inflation occurs if productivity rises above its
trend level. By construction, swill lluctuate around a mean value ofO, since if!P and mw
are the average values of m'' and m"', respectively, and since B ison average on its trend
growth path B.
We are now ready to confront our theory of inflation and unemployment with some
data.
8. For simplicity, we do not distinguish between the current labour force N and its t rend level TV, implic itly assuming
N = TiJ. In Chapter 14 we saw that the cyclic al fluct uations in the labour force tend to be quite modest.
Companies. 2005
18.5 ~~~.~~.~~~.~ . ~.~.~. f..~~.~~~~.P.~ . ~.~.~Y.~..~~~~?..~Y .................................................................................

Does the Phillips curve theory fit the data?
If we insert our earlier assumption of static inflation expectations, :rr.'' = n _1, into (34), we
obtain an equation of the form
E[sJ = o. (35)
where a 0 and a 1 are positive constants. and E[·) is the expectations operator. Thus our
theory implies that the change in the rate of inflation should be negatively related to the
rate of unemployment. If we have data for inflation and unemployment, we can use
econometric techniques (regression analysis) to estimate the magnitude of the parameters
a 0 and a 1 . In this way we can check whether the estimated parameter values have the
'correct' positive sign expected from theory, and we can test whether they are significantly
dillerent from zero in a statistical sense.
Figure 18.6 shows observations of the unemployment rate and the annual change in
the rate of consumer price inflation in the US in the period 1962-1 995 . The downward-
sloping straight line in the ilgure is a regression line indicating the 'average' relationship
between unemployment and the change in inflation. The regression line has the following
quantitative properties, where the figures in brackets indicate the standard errors of
the estimated coelllcients, and where R 2 is the so-called coelllcient of determination
measuring the share of the variation in 6 :rr. which is explained by our estimated regression
line:
6 n = 4 .467 - 0.723 · u. (36)

(se=I.USI) (se=O. I 72)
5
4 •
3 • •
£c:
2
0
·-;
~ 0
.S
Q) -1
Ol
c:
-2
"' •
••
.£:
()
-3
-4 •
-5
3 4 5 6 7 8 9 10
Unemployment rate(%)
Figure 18.6: Relation between unemployment and the change in inflation in the United States,
1962-1 995
Source: R.B. Mitchell, International Historical S tatistics, Macmillian 1998; and Bureau of l abor Statistics.
The coel11cients in (36) do indeed have the signs we would expect from theory. They are
also significantly dillerent from 0 in a statistical sense.9 Figure 18.6 shows a fairly clear
tendency for inflation to fall as unemployment goes up.
Note that the estimated coellicients in (36) enable us to oller an estimate of the natural
rate of unemployment in the US. According to (34) and (3 5) we have !:ln = c1(u - u) =
a0 - a 1 u if we sets equal to its mean value of 0 . Since 8n = 0 when u = u. it follows that
u = ajnL. Inserting the estimated parameter values from (36). we find that i7 = 4 .467/
0. 723 "" 6.2. This implies that the natural unemployment rate in the US averaged around
6.2 percent in the estimation period 1962- 1995.
In summary. the theory of the expectations-augmented Phillips curve seems roughly
consistent with the US data. At the same time we also see that the observed change in
inflation has often deviated quite a lot from the estimated regression line. Indeed, the value
of R 2 suggests that variations in unemployment can only explain a little over one-third of
the variation of inflation. According to our theory, the rest of the variation must be
accounted for by the exogenous shocks incorporated in our supply shock variables.
Given the strong simplifying assumptions we have made. it is not really surprising
that our regression equation leaves a lot of the variation in inflation unexplained. Our
assumption that inflation expectations are static is rather mechanical. For example, in
periods where the llscal or monetary authorities announce a signillcant change in
economic policies. the private sector may have good reasons to believe that tomorrow's
inflation rate will not simply equal the current rate ofinllation. 10 As another example. our
simple production function (3) abstracts from the fact that production requires inputs of
raw materials as well as labour input. Hence our Phillips curve does not capture ch anges
in inflation which are driven by changes in the international price of important raw
materials such as oil.
Productivity growth, the Phillips curve and the 'New Economy'
Despite these weaknesses. the important message from Eq. (36) is that there seems to
be a systematic and statistically highly significant negative relationship between the
level of unemployment and the change in the rate of inflation. However, in the second half
of the 1990s many observers began to question this relationship. The reason was the
remarkable performance of the US economy during that period. As you can see from
Fig. 18.7, having been located to the far northeast of the unemployment - inflation scatter
diagram. during the 1990s the short-run Phillips curve seemed to shift all the way back to
the favourable position it had occupied in the 19 60s. Apparently this shift was not a
simple consequence of a fall in expected inflation generated by the observed fall in actual
inflation since the early 1980s. This point is illustrated in Fig. 18.8 . The figure compares
the actual inflation rate with the rate of inflation predicted from our Phillips curve, (36),
which was estimated from data for 1962- 199 5. We see that from 1996 and onwards, the
rate of inflation predicted from the historical link between unemployment and inflation
9 . As a rule of thumb, the estimated coefficient should be numerically at laast twice as big as its standard deviation to
be statistically significant. This condition is easily met in our estimat ed equation (36). For the benefit of readers who
are familiar with regression analysis, the value of the Durbin-Wat son statistic is 1.515 which indicates that there are
no serious problems of autocorrelation in our regression.
10. In C hapter 21 we shall d iscuss the formation of private sector expectations in more detail.
Companies. 2005
540 PAR T 5: THE BU ILD ING BLOCKS FOR THE SHORT-RUN MODEL
14
12
10
~ 8
c
0
·~
6 82
]
4
83
2
0
0 3 4 5 6 7 8 9 10
Figure 18.7: The shifting short-run Phillips curve in the United States
Source: R.B. M itchell, International Historical Statistics, Macmillian, 1998; and Bureau of Labor Statistics.
14
- Actual - Predicted
12
10
~ 8
c
·~ 6
~
4
o +---. ---. ---.---.---. ---.---.---.---. ---.---

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000
Year
Figure 18.8: Actual and predicted inflation in the United States

Note: The predicted inflation rate was found from Eq. (39).
Source: R.B. Mitchell, International Historical Statistics, Macmillian, 1998 ; and Bureau of labor Statistics.
systematically overshoots the actual inflation rate. For example. based on the behaviour
observed during the 1962- 1995 period, one would have expected to see a US inllation
rate of 8.5 per cent in 2000, but the actual inllation rate remained subdued at a level of
3.3 per cent. despite the low rate of unemployment. In other words, it seemed that a struc-
tural shijt took place in the US economy around the mid 1990s, causing a breakdown of
Sorensen-Whitta- Jacobsen: I Part 5- The Buil ding 18. Inflation, unempl oyment I © The McGraw-Hill
the expectations augmented Phillips curve which had 11tted the data reasonably well up
until1995 .
At the same time as unemployment fell 'oVithout driving up the rate of inflation. the
growth rate of US labour productivity started to pick up. This is shown in Fig. 18.9 . As
indicated by the horizontal lines, the average growth rate of labour productivity during
the period of the prolonged productivity slowdown from 1974 to 1995 was only 1.35 per
cent per year, whereas the average productivity grmoVth rate rose to 2.42 per cent per year
during 1996- 2001 .
Impressed by these developments, many commentators argued that a 'New Economy'
had arrived which did not obey the 'old rules of the game' . Some participants in the
economic policy debate even claimed that it was time to scrap the established macro-
economic theory which apparently could no longer explain what was going on. These
proponents of the 'New Economy' paradigm argued that it was now possible to maintain
a much lower unemployment rate in the US without provoking high and accelerating
inflation.
However, with a slight reinterpretation our theory of wage formation actually oilers
an explanation for recent US developments. Recall that the 'normal' productivity level B
enters our supply shock variable, s. because the target real wage of workers (which we
may denote by w* ) is tied to normal productivity:
w* = mwcB. (37)
We may interpret the target real wage as a wnge norm reflecting worker perceptions of
what a 'fair' or 'normal' real wage ought to be. Worker aspirations regarding the growth
rate of real wages are presumably influenced by the rate of real wage growth experienced
in the past. which is determined by the historical growth in productivity. In (3 7) the
growth over time in the variable B captures the speed with which workers believe that
their real wages 'ought' to grow. During a period like the late 1990s when productivity
8
7
6
5
4
3
.g.
~
v. \
2
1
0
-1
-2
1945 1955 1965 1975 1985 1995 2005
Year
Figure 18.9: Annual growth rate of labour productivity in the United States
Note: Growth rate of output per hour in non·farm business sector. The horizontal lines are average growth rates
over t he subperiods indicated.
Source: B ureau of Labor Statistics.
Companies. 2005
growth was accelerating after a long period of slow growth , the growth in the target real
wage (the growth in B) will probably lag behind the growth in actual productivity B. since
workers emerging from a long period of relatively low productivity grm<Vth are still
accustomed to a relatively low growth rate in their real earnings. In a time of accelerating
productivity growth, the magnitude ln(B/B) = In B - In B included in our supply shock
variable s will thus tend to be positive. According to (34) this will reduce the rate of
inflation associated with any given level of unemployment and expected inflation, in line
with US experience since 19 9 5.
The idea that accelerating productivity grm<Vth may explain the favourable shift in
the US Philips curve has received empirical support in a recent study by American econo-
mists Laurence Ball and Robert Moflltt (2001). 1 1 They show that a Phillips curve which
assumes that the real wage growth demanded by workers is a geometrically weighted
average of realized past rates of real wage growth (which depend on past productivity
growth) can fully account for the favourable US unemployment- inflation trade-off in the
late 1990s.
According to the theory ofBall and Mollltt. an accelerntion of productivity growth can
temporarily improve the short-run unemployment-inflation trade-oil. However. the
theory also implies that once real wage expectations catch up with the higher rate of
productivity growth, the rate of inflation will pick up again, restoring the old relationship
between inflation and unemployment.
18.6 The aggregate supply curve

................................................................................................................................................................................
Our theory of aggregate demand presented in Chapter 17 implied a systematic link

between the output gap (the percentage deviation of output from trend) and the rate of
inflation. We shall now show that our theory of inflation and unemployment implies
another systematic link between these two variables.
Recall that in a symmetric general equilibrium, total GDP is Y = nY;. and total
employment is L = nL;. From (3) we then have:
T.)l-a
( = n" BL
1
Y = nB-;; - ". (38)
Takin g logs on both sides of (38) and using L =(1 - u)N plus ln(1 - u)"" - u, we get:
y =In y = In n" + In B + (1 - a)ln[(1 - u)N]
"" ln n" + ln B + (1 - a)ln N - (1 - a)u *'*
In n" + ln B - y
u = InN + . (39)
1- a
Let us now define 'natural' output, Y. as the volume of output produced when employ-
mentis at its natural level and productivity is at its trend level:
(40)
11 . laurence Ball and Robert Moffitt, 'Prod uctivity Growth and the Phillips Curve', National Bureau of Economic
Research, NBER Working Paper 8421. August 2001.
In other words. natural output sometimes a lso referred to as potential output is the
level of production prevailing when the economy is on it~ long-run growth trend. Defin ing
[I = In Y and using [ = (1 - u)N plus ln(1 - fi) "" - u. we take logs in (40) and lind an
expression analogous to (39):
Inn"
InN+ _ _+ _In_B- y
_.:;,.
IT = (41)
1- a
Substituting (39) and (41) into our expectations-augmented Phillips curve (17) and
collecting terms, we end up with the economy's short-run aggregate supply (SRAS) curve:
n = n' + y(y - [I) + s. (42)
(1
m 1
(mw)
ln(B/B)
y =--> 0.
1- a ml' (
s =ln-' ) + In - - -
m'"
- -.
1- a
The magnitude y - [I is the percentage deviation of output from trend. referred to as the
output gap. From (42) we see that. ceteris paribus, the rate of inllation varies positively with
the output gap. The reason is that a rise in output requires a rise in employment. and
because of diminishing marginal productivity oflabour, higher employment generates an
increase in marginal cost which is translated into an increase in prices via the mark-up
pricing behaviour of tlrms. Equation ( 42) also implies that the actual rate of inflation
varies positively with the expected rate of inflation and with the supply shock variable. s,
capturing shocks to mark-ups and to productivity.
The short-run aggregate supply curve in (42) summarizes the supply side of the
economy. Because the expected inllation rate is here taken as given, the curve is a short-
run relationship. Over time, the expected intlation rate will gradually adjust in reaction to
previous inflation forecast errors. When n ' changes. it follows from (42) that the short-
nm aggregate supply curve '.viii shift upwards or downwards. This is illustrated in
Fig. 18.10 which shows three dillerent SRAS curves corresponding to three different
levels of expected inllation rate. In long-run equilibrium, when expected inflation equals
actual inllation and there are no shocks (s = 0), we see from (42) that output must be
equal to its ;natural' level. [1 . The natural rate of output is independent of the rate of infla-
tion. since the natural unemployment rate IT is independent of n. The long-run aggregate
supply (LRAS) curve is therefore vertical. as shown in Fig. 18 .10.
Apart from depending on the expected rate of inllation, the position of the short-run
aggregate supply curve also depends on the supply shock variable, s. From (42) we see
that the SRAS curve will shift upwards in the case of a positive shock to one of the mark-
ups mwor m v, or in the case of a negative shock to productivity (B < B). Note that several
types of supply shocks may be modelled as productivity shocks. For example, a loss of
output due to industrial conflict may be interpreted as a temporary fall in labour produc-
tivity. An unusually bad harvest due to bad weather conditions may likewise be seen as a
temporary drop in productivity. An exogenous increase in the real price of imported raw
materials such as oil will also work very much like a negative productivity shock. If the
price of oil increases relative to the general price level. an economy dependent on imported
oil will have to reserve a greater fraction of domestic output for exports to maintain a
Companies. 2005
544 PART 5: TH E BUILDING BLOCKS FOR THE SHOR T- RUN MODEL
LRAS
SRAS (nil
SRAS (n~)
SRAS (n8)
Figure 18.10: Aggregate supply in the short run (SRAS) and in the long run (LRAS)
given volume of oil imports. Thus. for given inputs of domestic labour and capital, a lower
amount of domestic output will be available for domestic consumption. just as if factor
productivity had declined. More generally, any exogenous change in the economy's inter-
national terms of trade (a shift in import prices relative to export prices) may be modelled
as a productivity shock in our AS-AD model.
Over the last three decades, the real price of energy inputs has fluctuated consider-
ably. as illustrated in Fig. 18.11. For example. following political upheaval in the Middle
400
350
M
0
0
- 250
300
I vv\
II
~ 200
I
~ 150
~I A.
£
~.~ f\N'v
100
\...J ~
\j
50
0
1971 1974 1977 1980 1983 1986 1989 1992 1995 1998 2001 2004
Figure 18.11: The real price of fuel imports in Denmark

Source: MONA database.
East. the OPEC cartel of oil-exporting countries was able to raise the real price of oil quite
dramatically in 1973- 74 and again in 1979- 80. Because most OECD economies were
large net importers of oil at the time, these oil price shocks worked like a signiilcant
negative productivity shock for the OECD area. On the other hand. the collapse of oil
prices from around 1985 tended to boost real incomes in the OECD, just like a positive
productivity shock.
This completes our theory of the aggregate supply side. In the next chapter we shall
see how aggregate supply interacts with aggregate demand to determine total GDP as well
as the rate of inflation.
18.7 Summary
................................................................................................................................................................................
1. The link between inflation and unemployment determines how the supply side of the economy
works. Some decades ago most economists and policy makers believed in the simple Phillips
curve which postulates a permanent trade-off between inflation and unemployment: a perma·
nent reduction in the rate of unemployment can be only achieved by accepting a permanent
increase in the rate of inflation, and vice versa.
2. Empirically the simple Phillips curve broke down in the stagflation of the 1970s. This led to the
theory of the expectations-augmented Phillips curve which says that the simple Phillips curve
is just a short-run trade-off between inflation and unemployment, existing only as long as the
expected rate of inflation is constant. When the expected inflation rate goes up, the actual
inflation rate increases by a corresponding amount, other things equal.
3. The expectations-augmented Phillips curve implies the existence of a 'natural' rate of unem·
ployment, defined as the level of unemployment which will prevail in a long-run equilibrium
where the expected inflation rate equals the actual inflation rate. Since any fully anticipated
rate of inflation is compatible with long-run equilibrium, the long-run Phillips curve is vertical.
When the actual inflation rate exceeds the expected inflation rate, the actual unemployment
rate falls below the natural unemployment rate, and vice versa.
4. Several different theories of wage and price formation lead to the expectations-augmented
Phillips curve. One such theory is the 'sticky-wage model' in which nominal wage rates are
pre-set at the start of each period. In the sticky wage model presented in this chapter, money
wages are dictated by trade unions seeking to achieve a certain target real wage, given their
expectations of the price level which will prevail over the next period. G iven the wage rate set
by unions, profit-maximizing monopolistically competitive firms set their prices as a mark-up
over marginal costs and choose a level of employment which is declining in the actual real
wage. According to this model, employment increases above its natural rate when the actual
price level exceeds the expected price level, and vice versa. The model also implies that, in
general, there is some amount of involuntary unemployment.
5. In the sticky-wage model the target real wage is a mark-up over the opportunity cost of
employment which is given by the rate of unemployment benefit. The wage mark-up factor -
and hence the target real wage - is higher the lower the wage elasticity of labour demand,
Companies. 2005
and the lower lhe weight the union attaches to the goal of high employment relative to the
goal of a high real wage. The mark-up of prices over marginal costs is higher the lower the
price elastic ity of demand for the output of the representative firm . The natural rate of unem-
ployment is higher the higher the wage and price mark-ups and the more generous the level
of unemployment benefits.
6. Another theory leading to the expectations-augmented Phillips curve is the 'worker-

misperception model' which assumes a competitive c learing labour market w ith fully flexible
wages. Labour demand is a declining function of the actual real wage, wh ile labour supply is
an increasing function of the expected real wage, s ince workers are imperfectly informed
about the current general price level. This model also implies that employment rises above
the natural level when the actual price level exceeds the expected price level. However, for
any given amount of unantic ipated inflation, the increase in employment is smaller in the
worker-misperception model than in the sticky wage model where nominal wage rates are
fixed in the short run. Even in the absence of nominal rigid ities, unanticipated inflation will thus
generate deviations of employment from the natural rate, but nominal rigidities w ill amplify the
fluctuations in employment.
7. According to the hypothesis of static expectations, the expected inflation rate for the current
period equals the actual inflation rate observed during the previous period. Combined w ith
the expectations-augmented Phillips curve, the assumption of static expectations implies that
the rate of inflation w ill keep on accelerating (decelerating) when actual unemployment is
below (above) the natural unemployment rate.
8. So-called supply shocks in the form of fluctuations in productivity and in the wage and price
mark-ups create ' noise' in the relationship between inflation and unemployment. An
unfavourable supply shock implies an increase in the actual rate of inflation for any given levels
of unemployment and expected inflation. In the presence of supply shocks the natural
unemployment rate is defined as the rate of unemployment prevailing when inflation expect-
ations are fu lfilled and productivity as well as the wage and pr ce mark-ups are at their trend
levels.
9. An expectations-augmented Phillips curve w ith static inflation expectations is consistent w ith

US data on inflation and unemployment in the period from the early 1 960s to the mid 1990s.
In the ' New Economy' of the late 1ggos inflation was surprisingly low, given the low rate of
unemployment prevailing during th at period. This experience may be seen as a result of a
favourable supply shock arising from the fact that target real wages were lagging behind the
accelerating rate of productivity growth.
10. The economy's short-run aggregate supply curve (the SRAS curve) implies a positive link
between the output gap and the actual rate of inflation, given the expected rate of inflation. The
SRAS curve may be derived from the expectations-augmented Phillips curve, using the pro-
duction function which links the unemployment rate to the level of output. The SRAS curve
shifts upwards when the expected inflation rate goes up, or when the economy is hit by an
unfavourable supply shock. When there are no supply shocks and expected inflation equals
actual inflation, the economy is on its long-run aggregate supply curve (the LRAS curve) wh ich
is vertical at the natural level of output corresponding to the natural rate of unemployment.
18 INFLATION, UNEMPLOYMEN T AND AGGREGA TE SUPPLY 547
18.8 Exercises
Exercise 1. lntersectorallabour mobility and the expectations-augmented

Phillips curve
In Section 2 we abstracted from intersectoral labour mobility by assuming that individual

workers are educated and trained to work in a particular sector. In that case a worker's
outside option (the income he could expect to earn if he were not employed by his current
employer) is simp ly equal to the rate of unemployment benefit. In this exercise we ask you to
show that one can still derive the expectations-augmented Phillips curve from a trade union
model of wage setting even if unemployed workers can move between sectors in their search
for jobs, as we assumed in our analysis of the labour market in Chapters 1, 1 2 and 13.
To simplify matters a bit, we now set our productivity parameter 8 = 1 and work with a
linear production function with constant marginal productivity of labour, corresponding to
a = 0 in Eq. (3) in the main text. For a = 0 and 8 = 1, it follows from (7) in Section 2 that the
representative firm will set its product price as:
m" ::- 1. (43)
As in Section 2, an optimizing monopoly union for workers in firm (sector) i will set the nominal
wage rate IN; to attain an expected real wage which is a mark-up over the expected real value
of its members' outside option, denoted by v•. The union expects the current price level to be
p•, Hence it will set the wage rate:
(44)
This equaton is a parallel to Eq. (11 ) in Section 2 where we assumed that v• is simply equal
to the real rate of unemployment benefit, b. Here we assume instead that workers who are ini-
tially members of the trade union for sector i are qualified to apply for a job in other sectors if
they fai l to find one in sector i. For an average job seeker, the probability of finding work is
equal to the rate of employment, 1 - u, where u is the unemployment rate which gives the
probability that an average job seeker will remain unemployed. If the ratio of unemployment
benefits to the average wage level is equal to the constant c, we thus have:
v"= (1-u)w" + ub"= (1-u+uc)w•, 0 < C< 1, (45)
where w• is the expected average level of real wages, and b = cw• is the expected real rate
8
of unemployment benefit.
Since the outside option v • is the same across all sectors and all unions are assumed to
hold the same price expectations, it follows from (44) and (43) that all unions will charge the
same nominal wage rate, W, and that all firms will charge the same output price, P:
W;= W, P;= P. (46)
According to (43) the average real wage will then be W/P = 1/rrf. Let us suppose that union
wage setters have a realistic estimate of the average real wage so that:
w•= 1/ mP. (47)
1. Show that the relationship between the actual and the expected price level may be written as:
y"' 1 - ('; > 0. (48)
Companies. 2005
2. Usc (48) to derive an cxpccto.tions-o.ugmcntcd Phillips curve of the so.mc form as equation
(17) in Section 2. (Hint: use the approximation ln(1 - yu) "'-yu and define D "' In mw/y.)
Explain intuitively what determines the natural unemployment rate, V. Explain intuitively what
determines the slope (y) of the expectations-augmented Phillips curve.
3. Does the theory embodied in (43) - (48) assume short-run nominal wage rigidity?
4. The specification of the outside option (45) assumes that unions know the current aggregate
unemployment rate when they set the wage rate. Suppose instead that unemployment
statistics are published w ith a lag so that unions base their estimate of the outside option on
last period's recorded unemployment rate, u_1 :
(49)
Derive the expectations-augmented Phillips curve on the assumption that the perceived
outside option is given by (49) rather than (45). Do we now have nominal wage rig idity in the
short run?
Exercise 2. Wage setting, labour demand and unemployment
1. Explain why union wage claims are moderated by a higher price elasticity in the representative
firm's product demand curve.
2. In the text we found that the wage elasticity of labour demand is higher at the sectoral level
than at the aggregate level. Explain why th is is so.
3. 'Tougher product market competition will reduce structural unemployment'. Explain th is state-
ment. Discuss what the government could do in practice to promote fiercer product market
competition.
Exercise 3. The Phillips curve with endogenous price mark-ups
In the main text we assumed for simplic ity that the representative firm's mark-up mP of price
over marginal cost was an exogenous constant. However, empirical research for the United
States suggests that price mark-up factors in that country tend to move in a countercyclical
fo.sh ion. In other words, the mo.rk-up tends to fall during business cycle expansions o.nd to rise
during recessions. There are several potential reasons for this countercyclical behaviour of
mark-ups, including the possibility that during booms when the demand pressure is high,
more new firms find it profitable to enter the market, thereby increasing the degree of compe-
tition and forcing existing firms to reduce their profit margins.
Since the rate of unemployment moves countercyclically, the countercyclical variation of
the price mark-up means that mPwill tend to move in the same direction as the unemployment
rate. For concreteness, suppose this relationship takes the form:
ih > 1' cp;;. 0, (50)
where e is the exponential function, and cp is a parameter measuring the sensitivity of the mark-
up to changes in the unemployment rate, u. In the main text of the chapter we focused on the
case of cp = 0, corresponding to a constant mark-up. In this exercise we ask you to study the
implications of assuming cp > 0 which is more in line w ith the empirical evidence for the US.
As in Exercise 1, we consider an economy with intersectoral labour mobility, union wage

setting and mark-up price setting. Hence we have (see Exercise 1 in case you need further
explanation):
(51)
(52)
v•=( 1 -u)w"+ ub 0 = ( 1 -u+uc)w•, 0 < c < 1, (53)
where w• s the expected average level of real wages, b• = cw• is the expected real rate of
unemployment benefit, and v• is the individual worker's expected outside option. The outside
option is t~e same across all sectors, so all unions charge the same nominal wage rate and all
firms charge the same price:
P;= P. (54)
Union wage setters assume that the average real wage is:
w•= 1/m, m> 1, (55)
where the constant m is the expected 'normal' price mark·up.

1. Show that the model consisting of (50)- (55) implies an expectations-augmented Phillips
curve of the form:
lnm+ In m w- 1n m
JT =.n:•+(1-C- ({))(D -u), D e - - - - - -- (56)
1 - C - ({)
where .n: "'p - p _1, .n:• "' p• - p ~ l' and P "' In P, and where you may assume that 1 - c- ({)> 0 to
ensure a positive solution for the natural unemployment rate, D. Explain intuitively how the
parameter ({) affects the sensitivity of inflation to unemployment. Explain and discuss the
various factors determining the natural unemployment rate.
2. Suppose that union wage setters expect the 'normal' price mark-up m appearing in (55) to be
(57)
Discuss whether the assumption made in (57) is reasonable. Derive a new expression for the
natural unemployment rateD on the assumption that (57) holds. Compare the new expression
to the expression for D given in (56) and comment on the differences.
Exercise 4. The Phillips curve and active labour market policy
The model with intersectorallabour mobility presented in Exercise 1 assumes that all workers
are competing on equal terms and with equal intensity for the available jobs. In that case it
seems reasonable that any individual worker's probability of fin ding a job is simply equal to the
overall rate of employment, 1 - u, as we assumed when specifying a worker's outside option.
In the present exercise we assume instead that some of the workers recorded in the
unemployment statistics are not fully 'effective' in competing for jobs, perhaps because their
skills do not fully match the qualifications demanded by employers, or perhaps because they
are not actively searching for a job all the time. We may model this in a simple way by
assuming !hat only a fraction, s, of the registered unemployed workers contribute fully to the
Companies. 2005
available labour supply in the sense of being immediately ready and able to accept the jobs
available. Thus we may specify the 'effective' labour supply from the pool of unemployed
workers as su. In the following, we will refer to s as the Job-Search-and-Matching-Efficiency
parameter (the JSME parameter). For the moment, we will assume that s is an exogenous
constant.
If we normalize the total labour force to equal 1, there are thus (1 - s)u unemployed
workers who are not really competing with their fellow workers for a job. Hence we may
measure the 'effective' labour force as 1 - (1 - s)u. For an average member of the effective
labour force (a qualified person who is ready to accept the available j obs), the probability pu
of ending up in the unemployment pool is the ratio of the 'effective' number of unemployed to
the 'effective' labour force, p" = su/ [1 - (1 - s)u]. By implication, a qualified person's prob-
ability of finding a j ob in the labour market is 1 - pu. If the expected average real wage is w•,
and if the ratio of the unemployment benefit to the average wage level is c, we may therefore
specify the real value of a qualified worker's expected outside option as:
unemployment
su ) ( su ) b~it ( 1 - (1 - cs)u)
v• = 1 - w• + cw• = w• 0<c < 1. (58)
( 1 - ( 1 - s)u 1 - ( 1 - s)u 1 - (1 - s)u '
The trade union for sector i expects the current price level to be p • and sets its nominal wage
rate \1\tj to attain an expected real wage W/ P" which is a mark-up over the outside option of
the employable union members:
w
~= m wv• (59)
pc '
The representative firm uses a linear production function with a = 0 and 8 = 1, so according
to Eq. (7) in the main text it sets its price P;as:
mP > 1. (60)
Since v• is the same across all sectors, it follows from (59) and (60) that all unions will set the
same wage rate and that all firms will charge the same price:
W;= W, P;= P. (61)
In accordance with (60) and (61), the representative union expects the average real wage to
be:
w•= 1/ mP. (62)
1. Demonstrate through a logarithmic approximation that the model (58)- (62) leads to an
expectations-augmented Phillips curve of the form
lnmw
:rc=:rc• +sy(D -u), y"' 1- c > 0, D"'--, (63)
sy
where :rc "' p- p _1, :rc• "' p• - p~ ., and p "' In P. (Hint: use the approximations ln[1 - (1 - cs)u]"'
-(1 - cs)u and ln[1 - (1 - s)u] "'-(1 - s)u.) Explain in economic terms how the JSME
parameters affects the sensitivity of inflation to unemployment.
Let us now analyse the effects of active labour market policy. Suppose that a fraction I of
the unemployed workers is enrolled in public education and training programmes aimed at
improving their qualifications for the available jobs. Such programmes may increase the
JSME parameter s partly by improving the match between the qualifications demanded by
employers and the skills possessed by the unemployed, and partly by increasing workers'
motivation to look for jobs (say, because it makes more attractive jobs available to them).
Hence we assume that s is an increasing function of 1:
S= S · 1'1, 0 <5< 1, '7 > 0, 0<1< 1. (64)
The elasticity 17 is a parameter measuring the degree to which the labour market programmes
succeed in actually upgrading the skills and motivation of the unemployed. However, (64)
does not capture all of the effect of active labour market programmes. When people are
enrolled in such a programme, they will often not be immediately available for a job should
they receive a job offer, or they may not have the time to look for a job. For simplicity, let us
assume that only that fraction 1 -I of the unemployed which is not currently engaged in
education and training is able to take a job. Moreover, let us assume that these job seekers
have benefited from previous training so that their Job-Search-and-Matching-Efficiency corre-
sponds to the value of s specified in (64). The 'effective' labour supply coming from the pool
of unemployed workers is then given by:
'Effective' unemployment rate = s( 1 - l)u, (65)
where s is determined by (64). The effective labour fo rce consists of those unemployed
workers who are effectively available for work, s(1 - nu, plus those who are already employed,
1- u. Thus a qualified worker's probability of being unemployed is:
s( 1 -l)u s(1 - l)u
p" = - - - - - - (66)
1 -u+s(1-l)u 1 - [1 - s( 1 -l)]u ·
In the questions below we w ill assume that I is a policy instrument controlled by the makers of
labour market policy. Furthermore, we assume that workers enrolled in active labour market
programmes receive the same unemployment benefits and enjoy the same utility as unem-
ployed workers who are not enrolled in programmes. This implies that active labour market
policy has no effect on the outside option other than the effect working through the impact of
I on pu.
2. Following the same procedure as in Question 1, demonstrate through a logarithmic approxi-

mation that when effective unemployment is given by (65) and the JSME parameter is given
by (64), we obtain an expectations-augmented Phillips curve of the form:
lnmw
n = n • +Y(D- u), y "' (1 - c)sl'1(1 - 1), li " ' -_- . (67)
y
(Hints: start by using (66) to respecify v 8 • Later on, when you take logs, use the approxi-
mations ln{ 1 - [1- cs(1 -/))u} ""-[1- cs(1- J)]u and ln{1- [ 1 - s( 1 - l)]u)"" -[1 - s( 1 - J))u.)
3. How does the natural unemployment rate react to an increase in the proportion of the
unemployed enrolled in active labour market programmes? (Hint: derive ()uj()f_) Explain the
offsetting effects of an increase in I.
4. Suppose that the government w ishes to minimize the natural rate of unemployment through
its active labour market programmes. Derive the value of I which w ill achieve this goal.
Companies. 2005
(Hint: remember that a necessary condition for minimization of ii is CJu/'dl = 0 .) G ive an intuitive
interpretation of your resu lt. Discuss briefly whether the government should necessarily push
active labour market policy to the point implied by your formu la. (Hint: are there any costs and
benefits of active labour market policy wh ich we have not included in our analysis?)
5. Suppose that unemployed workers prefer not to be enrolled in active labour market p ro-
grammes, say, because they consider enrolment to be stigmatizing, or because it reduces
their leisure time. Discuss whether th is wou ld make active labour market policy more or less
effective as a means of reducing structural unemployment.
Exercise 5. The Phillips curve with a time-varying NAIRU
Empirical estimates of the natural unemployment rate (the NAIRU) typically find that the evo-
lution of the NAIRU tends to track the evolution of the actual rate of unemployment, at least in
the short and medium run. This exercise extends our theory of the Phillips curve in order to
explain why the NAIRU tends to move in the same d irection as the actual unemployment rate
in the shorter run.
WP. c:on~ioP.r thP. following moc!P.I w ith intP.r~P.c:tor!l l Ill hour mohility, whP.rP. wP. ilrrly thP.
same notation as in Exercise 1, and where p uis an average worker's probability of remaining
out of work if he fails to find a job in his original sector:
Price formation: P;= mPWP mP > 1, (68)
Wage claim of the representative union: ~ = mwv•, mw> 1, (69)
The 'outside option' of union members: v• = ( 1 - p")w• + pub•, (70)
Expected average real wage: w• = 1/ mP, (7 1)

Expected real rate of unemployment benefit: b• = cw•, 0 < c < 1, (72)
Probability of remaining unemployed in case of job loss:
e;;. o, pu.;. 1 (73)
The variables u and u _1 are the unemployment rates in the current and in the previous period,
respectively. The new feature of the model above is Eq. (73) wh ich says that, ceteris paribus,
an unemployed worker has a smaller chance of finding a job if unemployment is rising than if
unemployment is fall ing.
1. Discuss briefly whether the specification in (73) is plausible.
2. Use the model (68)- (73) to derive an expectations-augmented Phillips curve of the form:
y"' 1 - c > 0, (74)
where :n"' p-p_1, :n•"' p • -p~ 10 and P "'ln P. (You may use the usual approximation
ln(1 - x) "' -x which is valid as long as x is not too far from zero.) Comment on the expression
in (74) and compare w ith the expectations-augmented Phillips curve derived in the main text
of the chapter.
In the following we assume that inflation expectations are static so that n• = n _1 •
3. Define the long-run NAIRU as the rate of unemployment Dwh ich will be realized when the
rate of inflation as well as the rate of unemployment are constant over time, that is when
n: = n: _1 and u =u_1• Derive an equation for the long-run NAIRU and use this expression to
explain the factors which determine the equilibrium rate of unemployment in the long run.
4. Define the short-run NAIRU as the rate of unemployment D. which will be compatib le with a
constant inflation rate in the short run where we do not necessarily have u = u_1 • (At the short-
run NAIRU we thus have :n:=:n: _1 but not necessarily u = u _1.) Derive an expression for the
short-run NAIRU and show that D. may be written as a weighted average of the long-run
NAIRU and last period's actual rate of unemployment u_1 • Which parameter determines how
much the short-run NAIRU is affected by last period's actual unemployment rate?
5. Assume that in period 0 unemployment increases from the long-run NAIRU (D) to the level
D + t:J.u0 . W ill it be possible to return to the unemp loyment rate D in period 1 without creating
higher inflation? G ive reasons for your answer.
Exercise 6. Estimating the time-varying NAIRU
In Section 5 we saw that the natural unemployment rate in the US seems to have varied over
time. In this exercise we invite you to estimate the level and variation in the NAIRU and to
investigate whether changes in the underlying rate of productivity growth may help to explain
the evolution of the NAIRU.
If inflation expectations are static, the expectations-augmented Phillips curve allowing for
supply shocks (s) takes the form:
:n:-:n:_1 =y(D-u)+s
which may be rearranged to give:
D +sh = u + t:,.:n;fy, (75)
Thus we may see the movements in the magnitude u + Mr./y on the right-hand side of (75) to
be a resu lt of gradual movements in the NAIRU, D, as well as a resu lt of the shorter-term and
more erratic supply shocks captured by s. If we have somehow obtained an estimate of the
parameter y, we may construct an estimate of D +sfy by calculating u + t:J.:n:fy, using available
data on unemployment and inflation. We may then use the HP filter introduced in Chapter 14
to split our estimate of D +s/y into a smooth underlying trend, which we interpret as an esti-
mate of D, and a residual term which we take to reflect siy. This is the methodology we ask
you to follow below.
1. The first step is to obtain an estimate of our parameter y. At the internet address
www.econ.ku.dk/pbs/courses/models&data.htm you will find annual data on the unemploy-
ment rate and the rate of inflation in the US for the period 1959- 2003. Using these data,
estimate a standard expectations-augmented Phillips curve of the form:
11_-rr. = a0 + a 1 u + S, (76)
by performing an O LS regression analysis for the period 1960- 2003. (You may assume that
s is norm ally distributed with a zero mean and a constant variance.) Is your estimate of a 1 sta-
tistically significant and does it have the expected sign? Assuming you can answer in the
affirmative, you may use your estimate of the numerical value of a 1 as an estimate of y in your
further analysis.
Companies. 2005
2. Armed w ith your estimate of y and your data set, you can now calculate a time series for the
magnitude u + D.n/y. In this way you obtain an estimated time series for D +s/y for the period
1960- 2003. At the internet address www.econ.ku.dk/pbs/courses/mode/s&data.htm you
will also find a link to a software facility enabling you to estimate a trend by means of the HP
filter (plus a b rief guide on how to use this program). Use th is facility to estimate an HP trend
in your time series for D +s/y, setting the A, parameter equal to 1000 to obtain a quite smooth
trend. Interpret you r HP trend as an estimate of D and construct a d iagram in which you plot
your estimated time series for the NAIRU. What is the range w ithin wh ich the NAIRU has
varied?
In Section 5 we argued that the trade-off between the level of unemployment and the
change in the rate of inflation will tend to improve in periods of accelerating productivity
growth, and vice versa. According to (75), a longer-lasting improvement in the trade-off
between u and D.n (that is, a fall in u + D. n/y) must reflect a fall in the NAIRU, D. Thus the
NAIRU should tend to fall when the underlying rate of productivity growth accelerates,
whereas D should tend to rise when underlying productivity growth slows down. The next
question asks you to explore this relationship.
3. The internet address www.econ.ku.dk/pbsfcoursesfmodels&data.htm contains annual data

on the g rowth rate in output per hour worked in the US for the period 1959 - 2003. Use the
HP filter to estimate an underlying tren d in this productivity growth rate, setting A.= 1000. Plot
the resulting estimate of trend productivity g rowth against your estimate of the NAIRU. Do the
two time series tend to move in opposite directions, as our theory predicts? Explain the theo-
retical reasons why accelerating productivity growth may be expected to reduce the NAIRU,
at least temporarily.
(Postscript: if you wou ld like to know more about the likely reasons for the variations in
the US NAIRU, you may want to consult the following readable artic le from wh ich the idea for
th is exercise was taken: Laurence Ball and N. Gregory Mankiw, 'The NAIRU in Theory and
Practice', Journal of Economic Perspectives, 16 (4), Fall 2002, pp. 115 -1 36.)
I Sorensen-Whitta-Jacobsen: I Part 6- The Short-run I 19. Explaining business I © The McGraw-Hill
Introducing Advanced Model for the Closed cycles: Aggregate supply Companies. 2005
The Short-run
odel for the
Closed Economy
19 Explaining business cycles: aggregate supply and

.~.~~.~.~~~~. ~~~~r.l.~. ~r.l. . ~~~~?..r.l. ....................................................................................
2 0 .~.~~?,g~.~~~~?..~..P~~~~Y..: . ~?.Y...~r.l.~..?.?..~?.. ...............................................................
21 .~.~~?.~~i.~~~~?.~.. P?.~~~Y...~~~.~..~.~~.i.?.J?:~.l..~~.l?.~~~.~~.~.?.r.l.~................................ .
22 The limits to stabilization policy: credibility, uncertainty
.~.r.l.~. ~~~~.~.~~.~ ................................................................................................................................
555
Part 6- The Short-run
Model for the Closed
19. Explaining business
cycles: Aggregate supply
© The McGraw-Hill
Companies. 2005
Macroeconomics Economy and aggregate demand
Sorensen-Whitta-Jacobsen: I Part 6- The Short-run 19. Explaining business © The McGraw-Hill
Chapter
Explaining business
cycles
Aggregate supply and aggregate
demand in action
T
hroughout economic history, the capitalist market economies of the world have
gone through recurrent periods of boom and bust. This is the fascinating phenom-
enon of business cycles described in Chapter 14. Although long periods of high
economic growth have sometimes led people to believe that the business cycle was
dead, Figs 19.1 and 19.2 show that it is still alive and well: economic activity continues to
fluctuate in an irregular cyclical manner around its long-run growth trend. and at the
start of the present decade the growth rate of real GDP per capita turned negative in all of
the three largest OECD economies. A fundamental challenge for macroeconomic theory is
to explain why the economy goes through these cyclical movements rather than evolving
smoothly over time.
The two previous chapters derived the economy's aggregate supply curve and its
aggregate demand curve. In this chapter we bring the two curves together in a complete
5.-------------------------------------------------.
4+---------------------------------------------__,
- US GOP - US inflation
3 +---------/ \---------------------------------------~
2 +-------~--+---------------~-.--------------__,
~ QT;-T,~~~~~Hr~~~~~~~~~~~~~~~~~~~
-1+-lr--~+------+--l
- 2 +-+-~-----------+---~------------------------------~
-3
-4+--V-------------4-l----------------------------__,
- s+---.---.---.---.-~.---.---.---.---.---.---.---.--
1974 1976 1978 1980 1982 198 4 1986 1988 1990 1992 1994 1996 1998
Figure 19.1: The cycl ical components of real GOP and domestic inflation in the United States,
1974-98 (quarterly data)
Source: B ureau of Economic Analysis.
557
I Part 6- The Short-run
© The McGraw-Hill
Companies. 2005
558 PART 6: TH E SHO RT- RUN MO DEL FOR TH E C LO S ED EC ON OM Y
4 .-------------------------------------------------~
USA
- 2 +------r------r------r------r------r-----,----~
1996 1997 1998 1999 2000 2001 2002 2003
Figure 19.2: Growth in real per capita GOP in USA, Japan and Germany
Source: Table 4, p. 19 2 of World Economic Outlook, April 2004, International Monetary Fund.
macro model which enables us to determine the levels of total output and inflation in the
short run. This model allows us to investigate the causes of the fluctuations in economic
activity which we observe in the real world. We will illustrate how business fluctuations
may be seen as the economy's reaction to various shocks which tend to shift the aggregate
supply and demand curves. We will also study the extent to which our model is able to
reproduce the most important stylized facts of the business cycle.
The perspective on business cycles adopted here is sometimes referred to as the Frisch-
Slutsky paradigm, named after the Norwegian economist and Nobel Prize winner Ragnar
Frisch and the Italian statistician Eugen Slutzky who first introduced this way of inter-
preting business cycles. 1 The Frisch-Siutzky paradigm distinguishes between the impulse
which initiates a movement in economic activity. and the propagation mechanism which
subsequently transmits the shock through the economic system over time. In our model
framework. the impulse is a sudden exogenous change in one of the 'shock' variables
determining the position of the aggregate supply and demand curves. The propagation
mechanism is the endogenous economic mechanism which converts the impulse into per-
sistent business fluctuations. The propagation mechanism reflects the structure of the
economy and determines the manner in which it reacts to shocks and how long it takes for
it to adjust to a shock. Ragnar Frisch stressed th at even though shocks to the economy
may follow an unsystematic pattern. the structure of the economy may imply that it
reacts to disturbances in a systematic way which is very diiierent from the pattern of the
shocks themselves. Frisch was inspired by the famous Swedish economist Knut Wicksell
who used the following metaphor to explain the dillerence bea.veen the unsystematic
impulse to the economy and the systematic business cycle response implied by the
propagation mechanism: 'If you hit a wooden rocking chair with a club, the movement
1. See Ragnar Frisch, 'Propagation Problems and Impulse Problems in Dynamic Economics' , in Economic Essays in
Honour of Gustav Cassel, London, Allen and Unwin, 1933; and Eugen S lutzky, 'The Summation of Random Causes
as the Source of Cyclic Processes', Econometrica, 5, 1937, pp. 105- 146.
19 EXPLAINI NG BU SI NES S C YCLES 559
of the chair will be more or less regular because of its form, even if the hits are quite
irregular'. 2
In short. this chapter raises two basic questions:
1. Why do movements in economic activity display persistence?
2. Why do these movements tend to follow a cyclical pattern?
We start out in Section 1 by setting up our model of aggregate supply and aggregate
demand, termed the AS-AD model. In Section 2 we then use the model to illustrate how
the economy reacts to demand and supply shocks in a so-called deterministic world. In
th is deterministic version of our AS-AD model, the demand and supply shocks are non-
random, occurring either within a limited time span, or representing a permanent level
shift in some exogenous variable. Follov.ring a qualitative graph ical analysis, we v.rill set up
a quantitative version of the determin istic AS-AD model to study the impu/se-respm1se func-
tions which show how the economy responds to various shocks over time. We lind that the
deterministic AS-AD model is capable of explaining the observed persistence of the move-
ments in economic activity following a shock, but it cannot really explain why business
fluctuations tend to follow a cyclical pattern . To deal with this problem, Section 3 sets up a
stochastic version of the AS-AD model in which the exogenous demand and supply shock
variables are rnndom variables. As we shall see. this model is able to reproduce the most
important stylized business cycle facts reasonably well. In Section 4 we discuss an alter-
native approach to business cycle analysis, the theory of 'real' business cycles, which
claims that business fluctuations are caused by the irregular evolution of the technical
progress underlying the process oflong-term growth.
19.1 The model of aggregate supply and aggregate demand

................................................................................................................................................................................
Restating the AS· AD model
To prepare the ground for our analysis of business cycles, the first part of this chapter
assembles the building blocks from the two previous chapters to form a model of aggregate
supply and aggregate demand - the AS-AD model - which allows us to determine the
short-run levels of output and inflation.
In extensive form. our AS-AD model may be stated as follows, where the subscript t
indicates that we consider period t:
(1)
(2)
(3)
(4)
(5)
2. This statement by W icksell was made in a discussion at a meeting of the Swedish Economic Association in Uppsala
in 1924. See 'Nationalekonomiska Foreningens Forhandlingar 1924', Uppsala, 1925.
© The McGraw-Hill
Companies. 2005
560 PART 6: THE SHO RT-RUN MO DEL FOR TH E C LO S ED ECON O MY
As you recall from Chapter 17, Eq. (1) is a log-linearized version of the condition for
goods market equilibrium, where v1 captures shifts in private sector conHdence, and g, - iJ
reflects shocks to public expenditure. Equation (2) is the defmition of the ex ante real
interest rate. and (3) is the Taylor rule describing the interest rate policy of a central
bank which pursues the inflation target n * . All of these equations were explained in
Chapter 17. Equation (4) is the short-run aggregate supply curve derived in Chapter 18.
The variable s: captures supply shocks such as fluctuations in wage and price mark-ups
and shifts in productivity. The Hnal equation (5) restates the assumption of static expecta-
tions according to which the expected current inflation rate equals last period's observed
inflation rate.
Chapter 17 showed how (2), (3) and (5) may be inserted into (1) to give the economy's
aggregate demand curve. Repeating this operation. we get the AD curve.
y ,- [I = a(n, - n *) + z,.
all!
a = --- , (6)
1 +a2b
which may be rearranged as:
AD: n, = n*- (:)(y 1- y - z,). ( 7)
Furthermore, we may write the short-run aggregate supply curve in the following
form by merging {4) and (5):
SRAS: :n1 = n,_, + y(y,-Ji) + s,. (8)
Short-run macroeconomic equilibrium
Equations (7) and (8) summarize the AS-AD model in compact form. It is worth recalling
the mechanisms underlying the AD and SRAS curves. The AD curve is downward-sloping
in (y ,. n,) space because a fall in the rate of inflation induces the central bank to cut the
real interest rate, thereby stimulating aggregate demand. The SRAS curve slopes upwards
in (y ,. n 1) space because a rise in output requires a rise in employment which drives down
the marginal productivity oflabour, thereby increasing marginal costs and inducing Hrms
to raise their prices at a faster rate.
At the start of the current period t. last period's inflation rate n 1_ 1 is a predetermined
variable which may be taken as given. The shock variables z1 and s 1 are also treated as
exogenous. The current levels of output and inflation are then determined by the two
equations ( 7) and (8). In graphical terms. the solution to the model is found at point E0 in
Fig. 19.3 where the SRAS curve intersects with the AD curve, that is. where llrms· desired
aggregate supply of goods and services matches aggregate demand.
The adjustment to long-run equilibrium
In Fig. 19.3 we have also included the long-run aggregate supply curve (LRAS). This
curve shows the volume of output wh ich is supplied in a long-run equilibrium where there
are no shocks (s 1 = z1 = 0) and where inflation is stable (n 1 = n 1 _ 1 ) . It lollows from (8) that
y, = [I in lon g-run equilibrium. so the position of the LRAS curve is determined by the
economy's natural rate of output [1, as we explained in Chapter 18.
Sorensen-Whitta- Jacobsen: I Part 6- The Short-run 19. Expl aining business © The McGraw-Hill
LRAS
AD
Yo y y
Figure 19.3: Short-ru n macroeconomic equilibrium with cyclical unemployment
The short-nm equilibrium E0 illustrated in Fig. L9.3 is characterized by cyclical

unemployment. since actual output y 0 falls short of natural output. The question is: will
the economy be able to work itself out of this recession? To focus on the economy's adjust-
ment mechanism. suppose there are no further demand and supply shocks following the
initial shock which created the recession. 3 Furthermore. remember that the position of
the SRAS curve depends on the expected rate of inflation n:~ which is equal to last period's
actual inflation rate n:, _ 1• Given that s, = 0, and since y, = [I when the economy is on its
LRAS curve, it follows from (8) that the SRAS curve for period t must cut the LRAS curve
at an inflation rate equal to last period's inflation n:, _ 1•
Using this insight, Fig. 19.4 traces the economy's adjustment over time. In the initial
time period t = 0 when the economy is in recession at point E0 • the inflation rate is n: 0 • In
the next period t = 1 people expect the previous period's level of inflation to continue, so
the SRAS curve for period 1 cuts the LRAS curve at the inflation rate n: 0 . Hence the SRAS
curve shifts down from SRAS0 to SRAS 1 between period 0 and period 1, generating a new
short-run equilibrium at point E1 where inflation is lower and output is higher than
before. When the economy enters period 2, agents expect an inflation rate n: 1 < n: 0 , so the
SRAS curve sh ills down again, cutting the LRAS curve at the inflation rate n: 1 • This
creates a new short-run equilibrium at E2 where inflation is even lower and output has
risen once more. As illustrated in Fig. 19 .4, this process of successive downward shifts in
3. Below we analyse more carefully how t he economy may be pushed into a recession in the first place. Here we just
take the economy's starting point as g iven, since we are currently focusing on its dynamic adjustment mechanisms.
© The McGraw-Hill
Companies. 2005
562 PART 6: TH E SHO RT-RUN MO DEL FOR TH E C LO S ED EC ON OM Y
LRAS
SRAS 1 (n:• = no}
SRAS2 (n• = JT 1}
SRAS3 (n• = n2}
AD
Yo Y1 Y2 Y3 y
Figure 19.4: The adjustment to long-run macroeconomic equilibrium
the SRAS curve '<\Till continue as long as y1 < [1. Thus the economy will gradually move
down along the AD curve until it finally settles in the long-run equilibrium E where
output is at its natural rate [J . In this long-run equilibrium in 11ation is at its target level.
since it follows from (7) thatn 1 = n * when y 1 = [I and z1 = 0.
The economics underlying this macroeconomic adjustment mechanism may be
explained as follows. When the economy is in recession, wage setters tend to overestimate
the rate ofinllation. This will motivate them to reduce their inflation forecasts over time, and
as a consequence they ~II lower their required rate of increase in money wages. Hence firms
will experience a lower rate of increase in marginal costs. and via mark-up price setting this
will translate into a lower rate of price in11ation. As inflation goes down, the central ban!<
cuts the real interest rate. thereby ensuring that weaker inflationary pressure increases
aggregate demand and output. As long as output and employment remain below their
natural rates, wage setters will continue to overestimate the rate of inflation (although to a
falling degree) and will therefore revL~e their inflation forecasts downwards, resulting in a
lower rate of wage and price inflation which triggers another interest rate cut and generates
yet another tall in expected in11ation, and so on. In this way the gradual reduction in actual
and expected inflation paves the way for successive reductions in the rate of interest which
stimulate aggregate demand and pull the economy out of the recession.
Note the crucial importance of central bank behaviour for this macroeconomic
adjustment process: to make sure that falling inflation actually increases aggregate
demand. the central bank must cut the nominal interest rate by more than one percentage
point lor each percentage point drop in inflation, that is. our parameter h in the Taylor
Sorensen-Whitta- Jacobsen: I Part 6- The Short-run 19. Explaining business © The McGraw-Hill
rule must be positive. 4 Otherwise falling inflation will cause the real interest rate to rise.
exacerbating the initial recession. The stronger monetary policy reacts to falling inllation
(the greater the value of 1!), the flatter is the aggregate demand curve, and the faster is the
convergence of output to its natural rate.
The need to keep the monetary policy parameter h positive has been termed the
'Taylor Principle' because it was stressed by John Taylor h imself. While adherence to this
principle is crucial lor the stability of the macro economy. our specific assumption of static
expectations (.n:~ = n 1_ 1) is less important. The only thing needed to ensure convergence
towards long-run equilibrium is that the expected inflation rate will fall over time when
people learn that they have overestimated the rate of inflation, and that expected inllation
will increase when people llnd out that they have underestimated the inflation rate. In
that case the SRAS curve will gradually shift down when the economy is in recession,
whereas it will shill up over time when the economy is in a boom (where y , > [i) .
Calibrating the model: how long is the long run?
According to the analysis above, the changes in the inflation rate generated by market
forces will automatically pull the economy out of a recession as long as the central bank
follows the Taylor Principle. In the long run the rates of output and employment will thus
converge on their natural rates. But lor workers who cannot flnd a job and for entre-
preneurs struggling to avoid bankruptcy it may not be very interesting to know that the
economy will eventually recover from a recession if the process of recovery is very slow. As
John Maynard Keynes once said: 'In the long run we are all dead'. When deciding whether
political action such as a tax cut or an increase in public spending is needed to llght a
recession, it is obviously very important to know whether the economy's automatic con-
vergence to the natural rate is relatively slow or relatively fast. In the former case flscal
policy intervention may be needed; in the latter case it may not.
To study the question 'How long is the long run?', we may consider a cJual1titative
version of our AS-AD model. By assigning plausible values to the parameters of the model,
we can investigate how fast the economy is likely to move from its short-run to its long-
run equilibrium after having been hit by a shock. For th is purpose we maintain the
assumption that there are no further supply or demand shocks following the initial shock
which created the recession. In formal terms this means that s, = z, = 0 for t;;. 0 . Let
=
[J, y, - !J denote the relative deviation of output from trend (the output gap). and let
ic, =n, -.n:* indicate the deviation ofinllation from the target inflation rate (the inllation
gap). Settings, = z 1 = 0 . we may then restate our AS-AD model (7) and (8) as follows:
(9)
(12h
a=---,
1 + (12b
(10)
4. Inserting r.~. 1 = n 1into (3), we get
i, = r+ n 1 +h(JT1 -n*) + b(y1 - .P)
which shows that a one percentage point increase in the current inflat on rate will trigger a more than one percent·
age point 1ncrease in the nominal interest rate when h > 0.
© The McGraw-Hill
Companies. 2005
From (9) we haven, = - (1/a)[J ,. which may be inserted into (10) along with (9) to give:
1
{3=- . (11)
1 + ay
It also follows from (9) that [J 1,. 1 =-an,,. 1 • which may be substituted into (10) to yield:
(12)
The linear first-order dilrerence equations in (11) and (12) have the solutions:
t = 0,1.2 . ... (13)
t = 0,1,2, ... (14)
where g0 and .ii:0 are the initial values ofy and n. respectively. 5 According to the definition
=
given in (11), P 1/(1 + ay) is less than 1, so the term {3 1 on the right-hand sides of (13)
and (14) will tend to 0 as timet tends to infinity. In other words, y 1 \<Viii converge on !:J and
n, will converge on n*. This proves th at the economy is stable in the sense that it tends
towards its long run equilibrium.
Literally speaking. it will take infinitely long for the economy to reach the long-run
equilibrium, but we may ask how long it will take before, say. half the adjustment to equi-
librium has been completed. Let t1, denote the number of time periods which must elapse
before half of the initial gap [1 0 betv.reen actual output and long-run equilibrium output
has been closed. According to (13), the value oft 1, may be found from the equation:
1 1
fl, = Yof3'• =2 Yo *'* {3'• = 2 *'* t1, ln {3 = ln(1/2) *'*
In 2 0 .693 1
t,, = - In f3 = - ~' {3=- . (1 5)
1 +ay
Hence the economy's speed of adjustment is uniquely determined by the value of the para-
meter f3 which in turn depends on the values of y and (L If one time period corresponds to
one quarter of a year, a value of y around 0 .05 is usually considered to be realistic.~>
According to Eq. (12) in Chapter 17. the parameter a 2 entering the expression for a in (9)
can be written as:
a2 =
- Dr 1- r ) - D,
- Y(l - Dy) ( 1 - Dy IJ. IJ = Y(l - r) ' (Hi)
where D, is the marginal effect of a rise in the real interest rate on private goods demand,
Dy is the marginal private propensity to spend income on consumption and investment
goods, and r is the net tax rate (taxes net of transfers) levied on the private sector. The
5. To see that ( 13) is indeed the solution to ( 11), note that ( 13) implies
In a similar way you may verify that (14) represents the solution to (12).
6. This is roughly one-quarter of the value y = 0 .18 which we estimated in Chapter 18 on the basis of annual data for
the US economy.
Sorensen-Whitta- Jacobsen: I Part 6- The Short-run 19. Expl aining business © The McGraw-Hill
parameter '17 indicates the ellect of a one percentage point rise in the real interest rate on
the private sector's savings surplus (savings minus investment) . measured relative to
private disposable income. For Denmark, this parameter has been estimated to be roughly
3 .6, 7 while plausible values for r and D1, would be r = 0.2 and Dy = 0.8, implying
a 2 = (0.8/0.2) x 3.6 = 14.4. If we use this value of c1 2 and set the monetary policy para-
meters I! and b appearing in (9) equal to 0.5 as proposed by John Taylor,~ we get
(l = 0.878, implying a value of {3 equal to 0 .958 . Inserting this into (15), we obtain
t,."' 16. In other words. for reasonable parameter values our model implies that it will take
roughly 16 quarters(= 4 years) for the economy to complete half of the adjustment to its
new long-run equilibrium after it has been hit by a shock. In a similar way one can show
th at it will take a little less than 13 years before the economy has completed 90 per cent
of the total adjustment to long-run equilibrium. Thus our model implies that it will
take quite a long time for the output gap to be closed if the economy is exposed to a
permanent shock. This is just another way of saying that there is considerable persistence
in the deviations of output from trend. The reason for thls persistence is that actual and
expected inflation adjust only slowly over time, so in the short and medium run output
and employment have to bear a large part of the burden of adjusting to a shock.
Note from the definitions of a and (3 in (9) and (15) that a high value of the monetary
policy parameter fz 'viii push {3 towards 0, thereby speeding up the economy's con-
vergence to long-run equilibrium. This confirms our earlier conclusion that the stronger
the interest rate response to a change in the rate of inflation, the faster is the adjustment of
output to its natural rate. It also follows from (9) and (15) that a lower value of the policy
parameter b will tend to raise the speed of convergence. Th is might seem to suggest that
the central bank should choose the highest possible value of I! and the lowest possible
value of b to minimize deviations from long-run equilibrium. However, through their
influence on the slope of the AD curve. the magnitude of I! and b also determine how far
output and inflation are pushed away initially from their long-run equilibrium levels
when the economy is hit by demand or supply shocks. Choosing the optimal values of fz
and b is therefore a complicated matter to wh ich we return in Chapter 20.
19.2 Business fluctuations in a deterministic world

The analysis above studied the economy's convergence on long-run equilibrium, but
without explain ing how output and inflation came to deviate from long-run equilibrium
in the tlrstplace. In this section we will illustrate how the long-run equilibrium may be dis-
turbed by shocks to aggregate supply and aggregate demand, and how the economy will
react on impact and over time to such shocks. Althoug h in practice the economy is
exposed to new shocks all the time, we can learn about the workings of the economy
by tracing the isolated ellects of a single shock, so this is wh at we 'viii do in the
7. See Erik Hailer Pedersen, ' Development in and Measurement of the Real Rate of Interest', Danmarks Nationalbank,
Monetary Review, 3rd Quarter, 40, (3), 200 1, pp. 7 1-90.
8. See John B. Tayl or, 'Discretion versus Policy Rules in Practice', Carnegie-Rochester Conference Series on Public
Policy, 39, 1993, pp. 195-214. In that article Taylor argued that b = t: = 0.5 was a reasonably good description of
actual US monetary policy since the early 1980s.
© The McGraw-Hill
Companies. 2005
566 PART 6: THE SHO RT-RUN MO DEL FOR TH E C LO S ED ECON OM Y
present section. SpeciHcally, we will treat a shock as a one-time disturbance which hits
the economy in a single period. In this way we are able to highlight how the economy's
reaction to the shock will generate persistent (long-lasting) deviations of output and infla-
tion from trend even if the shock itself is purely temporary. We assume that. after the
initial shock. there are no further subsequent shocks. The economy will then evolve in a
deterministic manner over time. that is, we can use our AS-AD model to calculate the
exact values of the output gap and the inflation gap in each period following the shock. In
Section 3 we will extend the model by assuming that the economy is repeatedly hit by
random shocks which turn our AS-AD model into a stochastic system.
A temporary negative supply shock
We start by studying the economy's reaction to a temporary negative supply shock such
as an industrial conflict or a bad harvest. Let us assume that in period 0, before the shock.
the economy i~ in the long-run equilibrium illustrated by pointE in Fig. 19.5. Suppose
then that the economy is hit by a temporary negative supply shock in period 1, causing
the value of our shock variable s to rise from 0 to some positive number s 1 in period 1.
Because of the temporary nature of the shock, it will not affect the long-run aggregate
supply curve, but according to (8) the short-run aggregate supply curve will move up by
the vertical distance s 1 from SRAS0 to SRAS 1 . The new short-run equilibrium for period 1
is therefore given by point E 1 . In moving from E to E1 we see that the economy goes
LRAS
Y2 y
Figure 19.5: Effects of a temporary negative supply shock

through a period of stagj1atiol'l, defined as the simultaneous occurrence of rising inflation

and falling output. The fact that intlation goes up is not surprising. since a negative supply
shock is essentially an exogenous increase in production costs. The reason why output
falls is that the central bank reacts to the rise in inflation by raising the interest rate,
thereby depressing aggregate demand.
Suppose next that the source of the supply shock disappears from period 2 onwards so
that s returns to its original value of 0. One might then think that the SRAS curve would
shift down to its original position SRAS0 , thereby restoring long-run equilibrium already in
period 2. But since the intlation rate has risen to the level n 1 in period 1, our assumption of
static expectations implies that the expected inflation rate lor period 2 will ben 1 which L~
higher than the expected inflation rate n* corresponding to the original SRAS curve.
Because of this rise in expected inJ1ation, the short-nm aggregate supply curve only shifts
down to the level SRAS 2 in period 2, even though s drops back dmm to 0. The result is the
new short-run equilibrium E2 illustrated in Fig. 19.5 where intlation is still above the target
rate and output is still below its natural rate. Yet, since the actual inflation rate falls between
periods 1 and 2, the expected intlation rate lor period 3 also falls, generating a further dmm-
ward shift in the SRAS curve in period 3, which in turn leads to a new short-nm equilibrium
with lower inflation, causing another dmmward shift in the SRAS curve. and so on. Thus
the continued downward revision of the expected inflation rate enables the economy to
move gradually down the AD curve back to the original long-run equilibrium E. However,
the important point is that because of the dynamic adjustment of expectations, even a
short-lived supply shock has a long-lasting effect on output and intlation.
A temporary negative demand shock
Let us now consider the efiects of a shock to aggregate demand. Suppose that, after having
been in long-run equilibrium in period 0 , the economy is hit by a temporary negative
demand shock in period 1, say, because private agents temporarily become more pes-
simistic about the economy's growth potential. Fig. 19.6 illustrates how the economy will
react to the temporary weakening of private sector con1ldence. Between period 0 and
period 1 the demand shock variable z changes from 0 to some negative number z 1•
According to (7) the AD curve therefore shifts down by the distance Iz,/a. l. li·om ADn to
AD 1 in Fig. 19.6 . This drives the economy from the initial long-run equilibrium E to the
new short-run equilibrium E 1 where output as well as inflation are lower.
However, in period 2 private sector conlldence is restored, pushing the aggregate
demand curve back to its original position AD0 as the variable z1 in (7) returns to its initial
value of 0. You might think that this would immediately pull the economy back to its
initial equilibrium E. Yet thL~ is not what happens, since the observed fall in inllation
during period 1 causes a fall in expected inflation from n* to the lower level n 1 as the
economy moves from period 1 to period 2. Hence the short-run aggregate supply curve
shifts down to SRAS 2 in period 2, generating a new short-run equilibrium at point E2 .
Remarkably, we see that output in period 2 overshoots its long-run equilibrium value y.
Real GDP will only gradually return to its normal trend level as the above-normal level of
activity drives up actual and expected intlation. As expected inflation goes up, the SRAS
curve '>\Till gradually shill: back towards its original position SRAS0 , and the economy will
move back along the AD curve to the initial long-run equilibrium E. The interesting point
© The McGraw-Hill
Companies. 2005
LRAS
SRAS0 = SRAS1
Y2 y
Figure 19.6: E'fects of a temporary negative demand shock
is that the initial recession generated by the temporary demand shock is followed by an
extended economic boom. This shows how the economy's propagation mechanism may
generate a pattern of adjustment which is rather dillerent from the time pattern of the
driving shock itself.
Impulse-response functions
In Section 1 we set up a quantified version of our model to study how long it will take th e
economy to converge towards long-run equilibrium. In a similar way, we may use a quan-
titative version of our AS-AD model to investigate how strongly the economy is lil<ely to
react to demand and supply shocks, assuming plausible values of the key parameters. For
this purpose we must of course allow our shock variables z1 and s, to deviate from 0. Using
=
our previous notation lor the output gap. y1 !J 1 - y, and our definition of the inllation
=
gap, ft n 1 - n*, we may write the AD curve (7) and the SRAS curve (R) in the form:
(17)
(1R)
From (17) we have .n,_ 1 = (1/a)(z 1_ 1 - y1_ 1) . Inserting this plus ( 17) into (1R), we lind:
(19)
In a similar way we may use (17) to eliminate,!}, from (1 R). yielding:
fr.l = f3icl - l + yf3zl + f3st. (20)
Using plausible parameter values, we may now simulate Eqs (19) and (20) from
period 0 and onwards to obtain so-called impu/serresponse .fimctions showing bow output
and inflation react over time to various shocks. Since output y 1 is measured in logarithms.
a unit increase in our demand shock variable z1 reflects an exogenous increase in demand
corresponding to 1 per cent of initial GDP. Moreover. a unit increase in our supply shock
variables 1 corresponds to an exogenous 1 percentage point increase in the rate of inflation.
In Fig. 19.7 we show the impulse-response functions lor the output gap and lor the
inflation gap generated by a temporary increase in our supply shock variables1 from 0 to
1 occurring in period 1. We use the same parameter values as before. except that we now
assume that the length of the period is one year. For this reason we have set the parame-
ter y equal to 0.2, corresponding roughly to our estimate in Chapter 1R based on annual
US data. Not surprisingly, Fig. 19.7 confirms the qualitative analysis in Fig. 19.5: in the
short and medium run, the temporary negative supply shock generates a positive inflation
gap and a negative output gap. In period 1. the inflation rate does not quite rise by 1 per-
centage point. because the drop in output and employment reduces inflationary pressure
by driving down the marginal costs of production. The sh ort-run decrease in output is
quite substantial, amounting to roughly 0 .7 5 per cent.
Figure 19.8 shows the impulse-response functions emerging when our demand
shock variable z1 temporarily falls from 0 to 1 in period 1. In the short run output falls by
less than 1 per cent, since the drop in inflation and economic activity induces the central
bank to cut the real interest rate to limit the fall in demand. When z1 returns to its normal
level of 0 from period 2 onwards, we see that the output gap changes from negative to
positive and then gradually falls back towards 0. This is quite in accordance with the
qualitative analysis of a temporary demand shock in Fig. 19 .6.
1
- y- y - Ji-JT*
0.8
0.6
""" ~ ......____
--
0.4
0.2
'$. 0
- 0.2
- 0.4
- 0.6
/
/
~ --
- 0.8
-1
3 5 7 9 11 13 15 17 19 21 23 25 27 29
Year
Figure 19.7: The adjustment to a temporary negative supply shock

(parameter values: y= 0.2, ~ =0.2, D,= 0.8, tJ= 3.6, h= b = 0.5)
© The McGraw-Hill
Companies. 2005
0.2 r-
r--- - y- y - JC - :n*
o~--~~~~~~~~----~
~~
- 0.2 +'-1- - - - - - - - - - - - - - - - - - - - - - - 1
$ -0.4 +-r-------------------------------------------------1
- 0.6-H- - - - - - - - - - - - - - - - - - - - - - - 1
- 0.8 ·1+- - - - - - - - - - - - - - - - - - - - - - - 1
3 5 7 9 11 13 15 17 19 21 23 25 27 29
Year
Figure 19.8: The adjustment to a temporary negative demand shock

(parameter values: y =0.2, r = 0.2, Dr= 0.8, 17 = 3.6, h =b = 0.5)
Permanent shocks
We have so far considered only temporary shocks. but sometimes a shock may be per-
manent. For example, a tax reform may permanently change the propensities to consume
and invest, and a liberalization of international trade may permanently allect the mark-
ups of prices over marginal costs by changing the degree of product market competition.
Analysing the effects of permanent shocks is more complicated since such shocks will
change the equilibrium real interest rate which enters into the central bank's monetary
policy rule. Moreover, a permanent suppl!J shock will change the economy's natural rate
of output. We will now show how the economy's long-run equilibrium is affected by
permanent shocks. Let f 0 and y0 denote. respectively. the real interest rate and the
level of output prevailing in the initial long-run equilibrium before the economy is hit by
a permanent shock. Since we have linearized the model around the initial long-run
equilibrium, the goods market equilibrium condition (1) may then be written as
!J, - Yo = v, - air,- fo), (21)
and the SRAS curve becomes:

(22)
In the initial equilibrium we have v1 = s, = 0. Now suppose that the economy experiences
a supply shock which permanently changes s, from 0 to some constant s * 0. The new
long-run equilibrium level of output. y, may then be found from (22) by inserting
.n: 1 = :n 1 _ 1 , !J, = y. and s, = s. and solving for y to get:
(23)
Recall from Chapter 17 that the equilibrium real interest rate (the 'natural' interest rate)
is the level of interest ensuring that the goods market clears at the natural rate of output.
The new equilibrium real interest rate, f, may therefore be found fr'om (21) by setting
actual output y 1 equal to the new natural rate of output [J = [1 0 - (s/y), and by setting r 1 = f.
Assuming that no permanent demand shocks have occurred so that u1 = 0 . it then follows
from (21) that:
s
f = f 0 + -. (24)
ya2
The intuition lor (24) is straightforward: a negative supply shock (s > 0) which reduces
the natural rate of output requires a fall in aggregate demand to maintain long-run equi-
librium in the goods market. To curb aggregate demand. the real interest rate must go up,
as stated in (24).
If the economy is instead hit by a permanent demar~d shock changing v, fi·om 0 to
some constant u* 0, we see from (23) that there is no ellect on the natural rate of output
(given that s = 0). We may therefore llnd the new equilibrium real interest rate by setting
[I = flo and u, = jj in (21) and solving for r, = f to obtain:
v
f = f0 + - . (25)
a7
This result is also intuitive: since natural output is unchanged. an autonomous increase
in aggregate demand (u > 0) must be ollset by a rise in the real interest rate which reduces
the interest-sensitive components of demand so that total demand does not exceed the
constant long-run equilibrium level of output.
To achieve its objectives of keeping inflation close to its target rate and avoiding large
output gaps, the central bank's estimate of the variables f and [J entering the monetary
policy rule (3) must try to account for any permanent demand and supply shocks oc-
curring along the way. Otherwise the central bank will not be able to achieve its goals. For
example. if a negative supply shock has permanently shifted the long-run aggregate
supply curve from the position LRAS 0 to the position LRAS 1 in Fig. 19.9. and if the central
bank does not account lor the fact that the equilibrium real interest rate has increased, the
economy will end up in the long-run equilibrium E1 where the inflation rate permanently
exceeds the inflation target n *. To prevent this systematic deviation of in llation from
target, the central bank must revise its estimate off upwards in accordance with (24).
When this revision occurs, the AD curve will shift down to the level AD' in Fig. 19.9 so
that the new long run equilibrium E' becomes consistent with the inflation target.
Of course, requiring that the central bank should update its estimate of the equi-
librium real interest rate to account lor permanent shocks is much easier said than
done. since it may take quite a while before it is possible to judge w hether some shock is
temporary or permanent in nature. When a permanent shock occurs. the economy's
adjustment over time will depend on how long it tal<es lor the central bank to discover the
permanent character of the shock. In Exercise 2 we invite you to study the effects of
permanent demand and supply shocks in more detail.
19.3 ?..~.~~~~~.~. ~Y..~~.~~..i.~ . ~ . ~~.~~~.~~.~~g ~g~'~'~'''"'''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''

••
As we have seen, our deterministic AS-AD model does quite a good job in accounting lor
th e observed persisten ce in t h e movement of output over time. But the deterministic model
© The McGraw-Hill
Companies. 2005
LRAS0
AD
AD'
Yo y
Figure 19.9: The long-run effects of a permanent negative supply shock
does not really explain the crucial feature of business cycles that economic booms tend
repeatedly to be followed by recessions. and vice versa. A satisfactory model of the busi-
ness cycle must be able to replicate the recurrent jluctuations in output and inflation, like
those in Fig. 19.1 which illustrated the evolution of the cyclical components of real GDP
and the rate of inflation in the US in the last part of the twentieth century.
To explain the cyclical pattern of output and intlation, we will now set up a stochastic
version of our AS-AD model in which our demand and supply shock variables z and s are
assumed to be random variables. In taking this step. we are building on a remarkable
discovery made in 1937 by the Italian economist-statistician Eugen Slutzky (see the
reference in Footnote 1 ). Slutzky discovered that if one adds a stochastic term with a zero
mean and a constant variance to a llrst-order linear diflerence equation, and if the coetn-
cient on the lagged endogenous variable is not too far below unity, the resulting stochastic
difference equation will generate a time series which looks very much like the irregular
cyclical pattern of output displayed in Fig. 19 .1!
To illustrate Slutzky's fundamental insight, suppose the variable a 1 evolves according
to the diflerence equation:
(26)
where e, is a random variable follm>Ving the standard normal distribution \>\lith a zero mean
and a unit variance. and being independently distributed over time. Drawing a sample from
the standard normal distribution to obtain a sequence of e1, and assuming an initial value
of a, equal to zero. we simulated Eq. (26) over 100 periods, here thought of as quarters. In
this way we obtained the time path for a, displayed in Fig. 19.10. Comparing this to the
graph lor the evolution of the US output gap in Fig. 19 .1. we see that the simple diflerence
6-.-- - - - - - - - - - - - - - - - - - - - - - - - - ,
5+-------------------------------------------~~~
4 +---------------------~------------------/
0 ~~----~--~~-~~r--+~--+---~
- 1+---------------------~~------4~-------4-+4H·------~
-2+---------~~---+----v-----~
-3+--r~--~-r~--~-r~--r-,-~~r-~-r~--~-r~--~
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 8 1 86 9 1 96
Quarter
Figure 19.10: Illustrating Slutzky's discovery: simulation of Eq. (26) when e1 follows the standard
normal distribution
equation (2 6) is able to generate a time series which qualitatively looks remarkably like the
recurrent business cycles observed in the real world.
Slutzky's discovery suggests that it may be fruitful to treat our shock variables z1 and
s1 as stochastic processes. By nature, the shocks to demand and supply which we have
been discussing are very hard to predict. Recall that supply side shocks include phenom-
ena such as industrial conflicts. fluctuations in agricultural output due to changing
weather conditions. oil price shocks due perhaps to military conflict or political unrest
in oil-producing countries. changes in productivity stemming from irregularly arriving
technological breakthroughs, etc. On the demand side, shocks may occur due to sudden
shifts in market psychology, or due to political regime shifts involving significant changes
in Hscal policies, among other things. Whether events such as these occur with deter-
ministic necessity - that is, whether they were unavoidable, given the way things had
developed- or whether they are fundamentally unpredictable, just like the outcome of the
toss of a coin, is a deep scientiHc question. But as long as our understanding of the causes
of such events - and hence our ability to predict them - is limited, it seems to make sense
to treat the supply and demand shocks in macroeconomic models as random variables. In
doing so, we admit that we can only predict what demand and supply will be on average,
while acknowledging th at the actual levels of demand and supply may deviate from their
average positions in a way we cannot anticipate. Let us therefore investigate how far a
stochastic version of our AS-AD model can take us towards explaining the stylized facts of
business cycles.
The stochastic AS· AD model
The stylized business cycle facts which we would like our model to explain are sum-
marized in the bottom row ofTable 19.1 . These ligures are based on quarterly data for the
US from 19 55 to 2001. We have chosen to focus on the relatively closed US economy
.....
Cll 0
~
a;
~=(I)
-o ~
0cD ... "'
)> n cn en
CD
;:o 0 -· =
~ ~ ~
.. ...."'" ........
-l
0>
= ,. ,..
n· ~ =
I
-l
... g
..
:I:
m
~
(f)
:I: ?!
0
Table 19.1: The stochastic AS-AD model and stylized business cycle facts ;:o g'3:"'V
0 0 ..
-;-i
=
"' ... ::l
;:o
~ i
c
..
.,..,..
Correlation oe; 0
z ;::. ....
Standard deviation (%) between Autocorrelation in output Autocorrelation in inflation
s:: -,..
:....~
output and n"'
0
Out put Inflation inflation t- 1 t- 2 t- 3 t- 4 t- 1 t- 2 t- 3 t- 4 0
m 2
r
-n
"'
AS·AD model with static expectations
0
and no supply shocks 1 1.62 0.52 0.08 0.81 0.66 0.47 0.37 0.99 0.96 0.91 0.85 ;:o
.. "
-l ~ ~ !-'t
~ ~ .5'
AS-AD model with static expectations :I:
m
and no demand shocks 2 1.67 1.90 - 1.00 0.92 0.86 0.79 0.73 0.92 0.86 0.79 0.73 0 ; > ;'
.....~;......
cc ca - ·
.. ...~·
r
AS·AD model with adaptive expectations 0
(f)
and a combination of demand and m
0
= .. ..
..
G)
-
~
c
-·
"' c "'
supply shocks 3 1.66 0.30 0.15 0.82 0.68 0.50 0.38 0.47 0.33 0.24 0.32 m Q.~;
4
0 <
The US economy, 1955:1-2001:1V 1.66 0.29 0.10 0.86 0.65 0.41 0.18 0.50 0.29 0.24 0.17 0
z
1. ¢J . 0, q c - 0, q :1 . 1, 0. 0. 75, 0 0
(J) -
1> = 0, (1 < = 0. 75, (1' = 0, Q= 0, (l/ = 0

s::
2. -<
3 . .p = o.92, o. = o.25, o, = 1, o= o. 75, w = o.25
4. The cyclical components of out put and inflation have been estimated v ia detrending of quarterly data using the HP filter w ith A- 1600.
Common parameter values in all AS-AD simulations: y = 0 .05, r = 0.2, O y= 0.8, 'I= 3.6, h = b = 0.5
~@
3 .....
-g iif
2. ~
"'"'g "'"OJ~
N
U"~
Sorensen- Whitta- Jacobsen: I Part 6- The Short-run 19. Explaining business © The McGraw-Hill
because we still have not extended our AS-AD model to allow lor international trade in
goods and capital (we will do so in Chapter 23). The data in Table 19.1 show the degree of
volatility and persistence in output and inllation. measured by the standard deviations
and the coefficients of autocorrelation. respectively. In addition, the third column indi-
cates the degree to which output and inflation move together, measured by the coefficient
of correlation. Ideally, simulations of our stochastic AS-AD model should be able to
reproduce closely these statistical measures of the US business cycle.
Economists have often debated whether demand shocks or supply shocks are the
most important type of disturbances driving the business cycle. One way of resolving this
issue is to investigate whether a model driven by demand shocks is better at replicating the
stylized business cycle facts than a model driven by supply shocks, or vice versa. In the Hrst
row of Table 19.1 we consider a version of our AS-AD model with static expectations
which includes only demand shocks. Thus we have sets, equal to 0 in all time periods. The
demand shocks are assumed to evolve according to the following Hrst-order autoregres-
sive stochastic process:
(27)
The notation x, - N(O, (1~). x , i.i.d. means that x , is assumed to follow a normal distribu-
tion with a zero mean value and a constant finite variance(J~ . and that it is identically and
independently distributed over time (i.i.d.). Hence the probability distribution of x , is the
same in all time periods. and the realized value of x, in any period tis independent of the
realized value of X; in any other time period j. 9 A stochastic process x , '>\lith these 'i.i.d.·
properties is called 'wh ite noise'. Note from (27) that, by allowing the parameter (5 to be
positive. we allow for the possibility that a demand shock occurring in a given quarter
may not die out entirely within that same quarter, but may be felt partly in subsequent
quarters. At the same time the restriction d < 1 implies that demand shocks do not last
forever.
To arrive at the numbers shown in the top row of Table 19 .1. we go through the
following steps:
1. Insert !2 7) into our AS-AD model given byEqs (19) and (20).
2. Set [J 0 = iT 0 = z0 = 0 fort = 0 and s, = 0 for all t.
3. Let the computer pick a sample of 100 observations from the standardized normal dis-
tribution N(O, 1).
4. Use these observations as realizations of x , and teed these values of x , into Eqs (19) and
(20), thus simulating the AS-AD model over the interval t = 1. 2. 3 ..... 100.
5. Use the resulting simulated values of y, and it, to calculate the standard deviations.
cross-correlation, and coefficients of autocorrelation for the two endogenous variables
over the 100 time periods.
In the simulations we use the same parameter values as those used to generate the
impulse-response functions in the deterministic AS-AD model. except that we set y = 0.05
9. Formally, E[x,x,.] = 0 for all t* t', where E(·) is the expectations operator.
0 I Sorensen-Whitta-Jacobsen: I Part 6- The Short-run
Introducing Advanced Model for the Closed
© The McGraw-Hill
Companies. 2005
576 PART 6: T H E SH O RT- RUN MO DEL FOR TH E C LO S ED EC ON OM Y
because we are now trying to replicate quarterly (rather than annual) data. In addition.
we assume the standard deviation a.,. ofx 1 to be equal to unity. and we set the value of the
o
autocorrelation coelllcient in (2 7) equal to 0. 75 . These values were chosen such that the
model simulation generates a standard deviation of output roughly equal to the one
observed in the US economy. Comparing the top and bottom rows in Table 19.1, we see
that our simulation without supply shocks reproduces the observed correlation between
output and inflation and the persistence (autocorrelation) in output reasonably well.
However, the model simulation exaggerates the volatility (standard deviation) of inflation
and particularly the degree of persistence (autocorrelation) in inflation. This impression is
confirmed by a glance at Fig. 19.11 which plots the simulated values of [J 1 and fi 1 in
our scenario without supply shocks. Comparing this figure to the actual US business
cycle shown in Fig. 19.1. we see that while our model generates a reasonably realistic
cyclical variability of output. it produces much too sluggish movements in the inflation
rate. This suggests that an AS-AD model driven solely by demand shocks cannot give a
fully satisfactory account of the business cycle.
In the second row of Table 19.1 we therefore focus on the opposite benchmark case
where the stochastic disturbances occur only on the economy's supply side. By analogy to
(2 7). we assume that the supply shocks follow a stochastic process of the form:
0 .,; {JJ < 1. (28)
To derive the ligures in the second row of the table, we have followed the same steps
as those explained above, except that we now set the demand shock variable z = 0 for all t.
The standard deviation a .. of the white noise variable c1 was chosen to ensure a simulated
standard deviation of output in line with the empirical standard deviation . The 'per-
sistence' parameter w in (28) was set to 0, since positive values of w generate an even
poorer fit to the data than th at displayed in the second row of Table 19 .1. Even so. we see
that the purely supply-driven AS-AD model is inconsistent with the stylized business cycle
5
:;r - n * - y - ji
4
3
2
1
'$. 0
-1
-2
-3
-4
-5
1 11 21 31 41 51 61 71 81 91
Quarter
Figure 19.11: Simulation of the stochastic AS·AD model with static expectations and no supply
shocks
Parameter values: See notes to Table 19.1.
facts in the bottom row. The model generates far too much persistence in output and
particularly in inllation. far too much volatility of inllation. and a counterfactual perfect
negative correlation between output and inllation. Of course. this negative correlation is
not surprising. since we have previously seen that a positive supply shock. which shifts the
SRAS curve dm>VIlwards, will drive down inllation at the same time as it raises output.
By contrast, in the US economy the correlation between output and inllation has been
positive in recent decades. indicating that supply shocks cannot have been the only driver
of the business cycle.
It should be stressed that the simulation results reported in Table 19.1 are smnple
spec(fic, relying on particular samples from the normal distribution. If we feed different
samples of ,.~ , or c, into the model, we get somewhat dillerent sample statistics, but the
general picture remains that neither a purely demand-driven nor a purely supply-driven
AS-AD model can account fully lor the stylized facts of the business cycle. In the next
section we will therefore consider an extended model allowing both types of shocks to
occur at the same time.
The stochastic AS-AD model with adaptive expectations
Apart from allowing for simultaneous demand and supply shocks, we will also generalize
our description of the formation of expectations. since this will improve the ability of our
AS-AD model to reproduce the empirical business cycle. We have so far assumed that
expectations of inllation are static, meaning that this period's expected inllation rate
is equal to last period's observed inflation rate. .n:~ = .n:,_ 1 • Figure 19.12 indicates that
this assumption may be too simplistic. The fat bold curve in the figure shows the expected
inflation rate lor the current quarter, .n:~. calculated as an average of the inflation forecasts
made by a number of prolessional economic forecasters in the US. The forecasts were
14 ------------------------------------------------,
- Expected inflation rate for the current quarter (n()
12 ----------------------------------------------~
- Actual inflation rate during the previous quarter (n 1_ 1)
1 0 ~----------------------------------------------~
8-41-------------------.--------------------------~
6 --~w~-.----------~~~~--------------------------~
2 --~H-----~~+-------------~~~k-~~b~~---~~~~
0 ----~-----·~------------------------------~~~
-2 ----------~--------------------------------~
-4 ------,------r-----r----~-----.------r-----~--J
1981-111 1984- 111 1987- 111 1990- 111 1993- 111 1996- 111 1999- 111 2002- 111
Figure 19.12: Expected current inflation and lagged actual inflation in the United States
Sources: Federal Reserve B ank of Philadelphia, Survey of Professional Forecasters and B ureau of Economic
Analysis. Inflation is measured by the c onsumer price index.
0 1Sorensen-Whitta-Jacobsen:
© The McGraw-Hill
Companies. 2005
reported before the middle ofthe current quarter, at a time when the forecasters knew the
actual inllation rate in the previous quarter, :rc 1 _ 1. The latter variable is drm.vn as the grey
line in Fig. 19.12. If expectations were entirely static (.n:~ = n 1_ 1), the 1:\ovo curves in the
diagram would thus coincide completely. In reality, this is obviously not the case. Instead,
we see from Fig. 19.12 that the expected inflation rate tends to fluctuate less than the
lagged actual inflation rate.
This suggests that our AS-AD model will become more realistic if we replace
the assumption of static expectations by a hypothesis which implies that the expected
inflation rate iluctuates less than last period's actual inflation rate. We will therefore
assume that expectations are adaptive, adjusting in accordance with the formula:
revlsion of expected last period·s inOation
inllation rate
~ ____......._.._
for<..>cast error
:rc'; - n~- L =(1 - ¢ )(nr- 1- n~-1), 0~¢< 1. (29)
Equation (29) says that the expected inflation rate is adjusted upwards (downwards) over
time if last period's actual inflation rate exceeded (fell short of) its expected level. We also
see that a change in last period's actual inflation rate is not fully translated into a
corresponding change in the expected inflation rate, provided¢> 0. From (29) we get:
n~ = ¢n~_ 1 + (1 - ¢)n1_ L (30)
Jt~- 1 = ¢n;_2 + (1 - ¢):rc,_1 (31)
.n:~-2 = ¢n;_3 + (1 - </J):rcl-3 (32)
and so on. From (30) we see that this period's expected inflation rate is a weighted average
of last period's expected and actual inflation rates. We also see that static expectations
(.n:~ = .n:1_ 1) is that special case of adaptive expectations where the parameter¢ is equal to
0. If we include the adaptive expectations hypothesis (29) in our AS-AD model, we can
therefore easily reproduce all our previous results by simply setting¢ = 0 . Note that¢ is a
measure of the 'stickiness' of expectations: a relatively high value of¢ means th at people
tend to be conservative in their expectations formation, being reluctant to revise their
expected inflation rate in response to previous inllation forecast errors. We can gain
further insight into the implications of adaptive expectations if we use the expressions for
:rc ~_ 1 , :rc~_ 1 . etc., to eliminate n1_ 1 from the right-hand side of (30). Using such a series of
successive substitutions we obtain:
n~ = ¢ 1 n~_ 1 + (1 - ¢):rc 1_ 1+ ¢(1 - ¢):rcr-2

= ¢ 3 n~-3 + (1 - ¢ ):rcr-1 + ¢(1 - ¢ ):rcr-2 + ¢ 2 (1 - ¢):rcH
= ¢ ~~n~-~~ + (1 - ¢):rcr-1 + ¢(1 - ¢)n,_2

+ ¢2(1 - ¢):rcl-3 + ... + ¢ 11-1 (1 - ¢):rct-u.
Since¢< 1. the term </J 11

:rc~_ 11 will vanish as we let n tend to infin ity. Hence we get:
n~ = L- ¢ "- \1 - ¢)nt-rr• 0~¢<1. (3 3)

11= 1
Equation 133) shows that the expected inflation rate for the current period is a weighted
average of all inflation rates observed in the past. with geometrically declining weights as
we move further back into history. Thus adaptive expectations put more weight on the
experience of the recent past than on the more distant past. But unlike the special case of
static expectations, adaptive expectations imply that people do not base their expectations
on only the experience of the most recent period. The higher the value of¢. the longer are
people's memories, that is, the greater is the impact of the more distant inflation history on
current expectations.
Our AS-AD model with adaptive expectations may now be summarized as follows: 10
AD: !J r - [J = e<(.n:* - n r) + z,, (34)
SRAS: .n:, = lt~ + y(y, - [J) + s,, (3 5)
Expectations: .n:~ = ¢n~_ 1 + (1 - ¢)n,_ 1 . (3 6)
Moving the SRAS curve (35) one period back in time and rearranging, we get:
(37)
which may be inserted into (3 6) to give:
(38)
Substituting (38) into (3 5) and using our previous definitions i'lt = y, - [J and ft 1 = .n:, - .n:*,
we may state the AS-AD model with adaptive expectations in the compact form:
(39)
(40)
where (39) is the AD curve and (40) is a restatement of the SRAS curve. As you may
check, (39) and (40) imply:
D, = afft_1 + {3(z 1 - z,_1) - a{3s, + a{J¢s, _l' (41)
ft, = aft1 _ 1 + y{Jz, - y{J¢2 1_ 1 + {3s 1 - {3¢s, _1 , (42)
1 + e<y¢ 1
a= < 1, fJ=--<1. (4 3)
1 + ay 1 + ay
In the third row of Table 19.1 we show the business cycle statistics generated by a
simulation of the model (41) and (42). We assume that z and s are uncorrelated and that
they follow the stochastic processes (2 7) and (28), respectively, '.vith c~ = 0. 75, w = 0.25,
a., = 1 and (1< = 0.2 5. To obtain realizations of the white noise variables x, and c,, we take
the same samples from the standardized normal distribution as we used in the first
and second columns of the table. The parameter¢ has been set at 0.9 2, while the other
parameter values are the same as those used in the AS-AD model with static expectations.
10. You may wonder if the AD curve is unaffected by the switch from static to adaptive expectations. The answer is
' Yes', provided the central bank has a good estimate of the expected inflation rate (such an estimate may be
obtained through consumer and business surveys, or by comparing the market interest rates on indexed and non·
r,=i,-n:.,
indexed bonds). In that case the central bank can control the real interest rate regardless of the w ay
inflation expectations are formed.
© The McGraw-Hill
Companies. 2005
580 PART 6: THE SHOR T-RUN MODEL FO R THE CLOSED ECONOMY
Comparing the third and fourth rows of Table 19.1, we see that the extended AS-AD
model with adaptive expectations and simultaneous demand and supply shocks fits the
empirical business cycle data fairly well, given the parameter values we have chosen
(of course. we chose the parameters to match the data as well as possible) . The simu-
lated volatility of output and inflation and the correlation between the two variables are
realistic. but the model tends to overestimate the degree of persistence in output and
inflation. due to its simpWled dynamic structure.
The model-generated time series for g, and ii 1 are plotted in Fig. 19.13a. For con-
venience we have also reproduced Fig. 19.1 as Fig.19 .3b. This shows the actual US business
cycle lor a recent time interval covering 100 quarters. The reader can thus compare the
model-generated data with reality. Of course, since the timing of the random shocks
5
- y- y - ;r - :rr*
4
3
2
1
'$. 0
-1
-2
-3
-4
-5
1 7 13 19 25 31 37 43 49 55 61 67 73 79 85 91 97
Quarter
Figure 19.13a: Simulation of the stochastic AS-AD model w ith ada,:>tive expectations and a com-
bination of demand and supply shocks
Parameter values: See notes to Table 19. 1.
5 ~------------------------------------------~
4 +-------------------------------------------~
3
2 -1-- --1
'$. o~-+~~~~~+1~~~~~7r~~~~~~~~~~~-q~~
-1
- 2+-~~--------~~1--------------------------~
-3+-++----------~~--------------------------~
-4+-4L------------~~--------------------------~
- 5 +---.---.--,,--.~-.---.---,---.--.---.---.-~
1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998
Year
Figure 19.13b: The cyclical components of real GOP and domestic inflation in the United States,
1974-98
Source: Bureau of Economic Analysis.
hitting our model economy does not coincide with the timing of the historical shocks to
the US economy, our model cannot be expected to reproduce the historical turning points
of the American business cycle. But a glance at Figs 19.13a and 19.13b suggests that the
variance and persistence of output and inflation in our calibrated AS-AD model with
adaptive expectations is fairly realistic.
We have tried to show that a simple stochastic AS-AD model allowing for shocks to
supply as well as demand can provide a reasonably good account of the cyclical move-
ments in output and inllation. In so doing, we have also tried to illustrate the basic
methodology of modern business cycle analysis, showing how macroeconomists build
dynamic stochastic general equilibrium models and calibrate these models to reproduce
the stylized statistical facts of the business cycle.
We must emphasize once again that the simulation reported in the third row of
Table 19.1 has the character of a numerical example, serving to illustrate that our AS-AD
model may be able to lit the data lor an appropriate choice of parameter values. But the
example was based on two particular samples from the normal distribution. To analyse
the model's ability to lit the data more systematically. one should either simulate the
model over a very large number of periods. or one should run a very large number of sim-
ulations based on a correspondingly large number of realizations from the probability dis-
tributions assumed lor the stoch astic shocks. In Table 19.2 we have tal<en the latter route.
The upper row in the table shows the mean values of the results from 1000 simulations of
our stochastic AS-AD model with adaptive expectations, where each simulation covers
100 time periods. The standard deviations (that is. the average deviation fi:'om the mean)
are indicated in brackets below the mean values. The parameter values in the simulation
model are exactly the same as those assumed lor the corresponding model in Table 19 .1.
For comparison, the bottom row in Table 19.2 repeats the stylized facts of the US business
cycle between 19 55 and 2001. We see that our AS-AD model seems to underestimate
slightly the standard deviation and the autocorrelation of output when the model is simu-
lated a large number of times with the parameter values that we used previously.
Nevertheless, the overall impression remains that the modelnts the data reasonably well,
once we consider how simple it really is compared to the staggering complexity of the real
world economy.
Yet we must keep in mind that although our AS-AD model with adaptive expec-
tations seems roughly consistent with the data on output and inflation. this does not
imply that we have found the explanation for business cycles. It is possible to construct
other types of models which match the data on output and inflation equally well. Hence
we cannor claim to h ave found the only correct theory of the business cycle. All we can
say is that our theory does not seem to be clearly rejected by the data.
19.4 A different perspective: real business cycle theory

............................................................................................................................................................................... .
The theory of the business cycle oll'ered by our AS-AD model emphasizes the role of
expectational errors and sluggish wage adjustment. and the microfoundation for the
SRAS curve developed in Chapter 18 implies th at business fluctuations are associated
with fluctuations in involuntary unemployment.
Cll
CCI
N
e
a;
~=(I)
-o ~
0cD ...
c "'
)> n n
en
;:o
CD
0 -· =
~ ~ ~
.. ...."'" ........
-l
0>
= ,. ,..
n· ~ =
I
-l
... g
..
:I:
m
~
(f)
:I: ?!
0
;:o g'3:"'V
0 0 ..
-;-i
=
"' ... ::t.
;:o
~ i
c
..
.,..,..
oe; 0
Table 19.2: The stochastic AS-AD model and the stylized business cycle facts z ;::. ....
s:: -,..
:....~
Correlation n"'
0
Standard deviation (%) between Autocorrelation in output Autocorrelation in inflation 0
m 2
r
output and -n
"'
Output Inflation inflation t- 1 t- 2 t- 3 t- 4 t- 1 t- 2 t- 3 t- 4 0
;:o
AS·AD model with adaptive expectations
-l
:I:
.. "
~ ~ !-'t
and a combination 1.41 0.26 0.09 0.70 0.49 0.34 0.23 0.33 0.18 0.14 0.13
m ~ ~ .5'
0 ; > ;'
..~;......~·
cc ca - ·
of demand and supply shocks 1
........
(0.18) (0.03) (0.14) (0.07) (0.11) (0.14) (0.15) (0.13) (0.1 5) (0.15) (0.15) r
0
(mean values of 1000 simulations, (f)
m
0
= .. ..
..
G)
-
~
c
-·
standard deviations in brackets) "' c "'
m Q.~;
The US economy, 1955:1-2001:1V 2 1.66 0.29 0.10 0.86 0.65 0.41 0.18 0.50 0.29 0.24 0. 17 0 <
0
z
1. )' = 0.05, T = 0.2 , Dy= 0.8, 11 = 3 .6 , h = b = 0.5, t/J = 0 .92, IJ x= 1, IJ c= 0.25, 0 = 0.75, (J) = 0 .25 0
2. The cyclical components of output and inflation have been estimated by detrending quarterly data using the HP filter with A= 1600.
s::
-<
= iif--1
~@
-g
2. ~
"'"'g "'"OJ~
N
U"~
During the 19 80s a group of researchers developed a very different approach known
as 'real business cycle theory' . henceforth termed RBC theory. u This school of thought
seeks to explain the business cycle by fluctuations in the rate of technological progress. In
the basic version ofRBC theory. the impulse initiating the business cycle is a shock to pro-
ductivity which is then propagated through the economy via its impact on capital accu-
mulation and the resulting ellect on productive capacity. According to this view the
employment fluctuations observed during business cycles reflect voluntary movements
along individual labour supply curves, as workers choose to enter the labour market or to
work extra hours when real wages are unusually high due to a high level of productivity,
while reducing their labour supply when productivity and real wages are unusually low
relative to their underlying growth trends. Thus markets are assumed to clear all the time,
and the business cycle is seen as the economy's optimal response to the changing teclmo-
logical opportunities available to economic agents.
As we shall discuss later, this benign view of the business cycle is problematic in
several respects, but its appeal is that it fully integrates business cycle theory with growth
theory. To illustrate this. we \.viii show how a slightly modified version of the Solow model
of economic grO\.vth with tech nological progress can generate 'real' business cycles if we
allow lor stochastic shocks to the rate of productivity growth.
Technology
Suppose that aggregate output is given by the Cobb-Douglas production function:

O<a<l. (44)
where Kt is the capital stock at the start of period t, Lt is total labour input during th at
period (measured in hours worked), and the parameter At captures so-called labour-
augmenting technical progress increasing the productivity of labour over time. 12 The
underlying growth rate of At contains a constant trend component, g, but productivity is
also allected by stochastic factors so th at:
In At = gt+ st' (45)
where the stochastic variables t now measures the relative deviation of productivity from
trend. When we simulate the model, we will assume thats, follows the stochastic process,
s t + 1 = ws 1 + ct + 1• specilled in ( 28). In the following. we shall also need the concept of the
trend level of productivity. At' defined by:
In A1 = gt. (46)
If St denotes total gross saving which is invested in capital, and f~ is the rate of depreciation
of the exL~ting capital stock. we have the book-keeping identity that Kt = (1- b)K, _ L + s l- 1'
11 . Important early contributions to RBC theory were made by Finn Kydland and Edward C . Prescott, 'Time to Build and
Aggregate Fluctuations', Econometrica, SO, 1982, pp. 1345-1370, and by John B . Long and Charles I. Plosser,
' Real Business Cycles', Journal of Political Economy, 91 , 1983, pp. 39-69. A good non·t echnical (but critical)
introduction to the ideas underlying RBC theory can be found in N. Gr~ory Mankiw, 'Real Business Cycles: A New
Keynesian Perspective', Journal of Economic Perspectives, 3, 1989, pp. 79-90.
B, Ai
12. By defining = -",we can easily rewrite the production funct ion as:
Y,= B,K~L:-tt,
where B. is the total fact or productivity w hose cyclical properties were studied in Chapter 14.
© The McGraw-Hill
Companies. 2005
For convenience, we will now assume that all of the pre-existing capital stock depreciates
within the course of one period. that is (5 = 1. We then have:
(47)
The unrealistic assumption that the pre-existing capital stock fully depreciates within one
period is by no means crucial for RBC theory, but in our context it greatly simplifies the
formal analysis without affecting the qualitative conclusions.
Economic behaviour
At the research frontier. RBC theorists are careful to derive their behavioural equations
from explicit intertemporal optimization of the objective functions of economic agents.
However, since advanced RBC models require some sophisticated mathematics which go
beyond the present book, we will make a short-cut and simply assume some behavioural
relationships which yield a model wh ose fundamental properties are characteristic of the
archetypical RBC model. Our first assumption is the standard Solow assumption that total
saving is a constant fraction. s. of total income:
S1 =sY1• O<s<l. (48)
Equations (44) and (48) are responsible for the propagatio11 mechanism in the basic RBC
model. When output and real income rise due to a positive productivity shock (a rise in A 1),
total saving increases. This feeds into an increase in next period's capital stock, which in
turn causes next period's output to remain above trend even if the positive technology
shock has already died out. The high level of output keeps saving and capital accumulation
above their normal levels. paving the way for yet another increase in output, and so on. In
this way the initial impulse (the stochastic productivity shock) gets propagated through the
process of saving and capital accumulation, generating persistence in total output and other
macroeconomic variables. This is very different from our AS-AD model where persistence
was generated by the sluggish adjustment of actual and expected inflation.
The next behavioural assumption in our RBC model is that aggregate labour supply.
L:.is given by:
L'I = _!
w )' ' t > 0. (49)
( WI
-
Here w 1 is the current actual real wage. and w1 is the 'normal' trend real wage. For conve-
nience. we have ch osen our units such that the normal labour supply forthcoming when
the real wage is on trend is equal to 1, so all the variables in the model may be interpreted
as per-capita magnitudes. Equation (49) captures the important idea in RBC theory that
there is intertemporal substitution ill labour supply: when the real wage is u nusually low
(w1 < w1), workers react by working less and consuming more leisure. but when the wage
is unusually high (w1 > w,). workers choose to work more than normal to take advantage
of the extraordinarily favourable labour market opportunities. The elasticity t reflects the
strength of these intertemporal substitution effects. At the same time (49) implies that
over the long run, where w, and w, grow at the same rate. there is no time trend in labour
supply. 13
Maximization of profits under perfect competition implies that the real wage will be
equal to the marginal productivity of labour. From (44) we therefore have:
CIY, ( K, )a (1 - a)Y,
w, = - = (1 - a) - - A, = . (50)
i)L, A,L, L,
In Chapter 5 it was shown that the magnitude K,/(A,L,) will be equal to a constant k*
when the economy is on its steady-state growth path .14 Using (SO). we may therefore
specify the trend real wage as:
w, = (1 - a)cA,. C =(k*)a. (51)
The labour market is assumed to be competitive, so the fully flexible real wage adjusts
to ensure that aggregate labour supply equals the amount of labour demanded, L,. Hence
we have the labour market clearing condition:
L, = L~. (52)
This completes the description of our modified Solow model of growth and business
cycles. We will now reduce the model to study the cyclical beh aviour of output and
employment.
The dynamics of output and employment
Our first step is to insert (49)-(51) into (52) and solve for L, to get:
(~
[;
L, = cA, )"· '1']=--<1. (53)

l+c
Next we substitute (47), (48) and (53) in to (44) and lind:

y )1}(1-a)
y = (sY Y' A,1-a -c.A,' • (54)
, 1- 1 (
Taking logs on both sides of (51), inserting (15) and (16), and isolating y , "' In Y 1 on the
left-hand side, we obtain :
_ aln s - 'I'J(l - a)ln c ( a ) (1 - a)[(l - 'I'J)gt + s,]

y, - + y, I+ , (55)
1 - 1](1 - a) l - 17(1 - a) - 1 - '1'](1 - a)
We now want to focus on the cyclicnl component of output. [i ,. For this purpose we define:
[J, =Do+ gt, (56)
where [1 1 represents the economy's steady state growth trend. i.e .. the level of (log) output
which would prevail if there were no productivity shocks and the economy had reached
13. We are thus implicitly assuming that the income effect and the substitution effect of a higher real wage on labour
supply cancel out when real wages grow at the normal rate.
14. The fact that we now allow for short·run variations in labour supply does not change the condition that K,/(A, L,)
must be constant in the steady state.
© The McGraw-Hill
Companies. 2005
586 PART 6: THE SHORT-RUN MODEL FOR THE CLOSED ECONOMY
=
its steady state. The variable flo ln Y0 is the log of the output level Y0 which would have
materialized in period 0 if the economy had been in steady state at that time. Since wt = wt
and hence Lt = 1 in a steady state. it follows from (44) and our definition ofc that a steady
state in period 0 would have implied YofA 0 = (K0 /A0 )" = (KofA 0 L0 )'" = c so that:
y0 =lnc, (57)
since we have In A0 = 0 according to (46 ). In a steady state without productivity shocks and
population growth, we also know from Chapter 5 that output is growing at the rate g. Hence
we have Y _1 = Yo/(1 +g). Setting t = 0 and A 0 = A0 in (54), inserting Y_ 1 = Yo/(1 +g),
tal<ing logs, and using (57) plus the facts that In A 0 = 0 and ln(1 +g) "' g. we lind:
Yo= (- a- )(ln s- g). (58)

1- a
Exploiting (56!. (57) and (58). you may verily that ( 55) simplifies to:
• =(
lft a )·
Yt 1 + ( 1 - a ) st. (59)
' 1 - 11(1 - a) - 1 - 17(1 - a)
=
Since TJ c/( 1 + t) is positive but less than 1. we see that the coelllcient on fi t_ 1 in (59) is
also positive but less than 1. In the absence of productivity shocks. the economy would
thus converge monotonically towards its steady state equilibrium where [J, = [J t-l = 0,
and where output (per capita) L~ growing at the constant rme g. exactly as implied by the
Solow model with technological progress described in Chapter 5.
Consider next the cyclical component of labour input, defined as Lt ln(L/ L), =
where [is the trend level of labour supply. Since L= 1, it follows from (53) that
Lt =In L, = 17(y, - (Inc+ ln At)). Since (46), (5 6) and (57) imply that Inc + In At = !]0 + gt =
Yt· this result may be written in the simple form:
Lt =1Jfi,. (60)
Equations (59) and (60) summarize our stochastic Solow model of real business
cycles. In mathematical tenns. (59) looks very much like the linear llrst-order dillerence
equation for the output gap in our AS-AD model. but the underlying economic mecha-
nisms are very different in the two models. In the AS-AD model the persistence in output
and employment is generated by the sluggish adjustment of actual and expected inflation,
and fluctuations in the output gap are associated with fluctuations in involuntary unem-
ployment. In our RBC model persistence stems from the dynamics of capital accumula-
tion, and employment fluctuations reflect that workers voluntarily change their labour
supply in response to shifts in productive opportunities.
In the previous section we saw that the AS-AD model does a reasonably good job of
reproducing the most important stylized facts of the business cycle. Is our RBC model
equally good at matching the data?
Taking the RBC model to the data
As the term indicates, the RBC model is indeed 'real', containing no nominal variables.
Hence the model cannot be used to study the dynamics of inflation and the interaction of
19 EXPLAINI NG B USI NES S C YCLES 587
inflation and output, but it does enable us to simulate fluctuations in the cyclical compo-
nents of output and labour input. In the bottom row ofTable 19.3 we report summary sta-
tistics on the behaviour of these variables in the United States in the period 1960- 2003.
based on annual data on real private sector GDP and total hours worked in the private
sector. We now use annual rather than quarterly data, since our simple RBC model is not
designed to reproduce the slight lag in the correlation between output and total hours
worked that one observes in quarterly data. 15 Because our theoretical RBC model mea-
sures the output and employment gaps as deviations from a constant linear growth trend,
we have detrended the logs of output and hours worked by estimating a linear trend using
Ordinary Least Squares regression analysis.
The top row in Table 19.3 sh ows the results of a simulation of our RBC model. using
the parameter values stated in the notes to the table. The stochastic component in our
productivity variable In A, is assumed to follow a first-order autoregressive process of
the form in Eq. (28), and the simulation is based on a sample for c, from the normal
distribution. spanning 100 periods. Our simple RBC model is tightly specilled, leaving us
with the choice of only four parameters a. IJ, w and o,.. We know from the theory and
empirics of growth th at the parameter o. equals the capital income share of GDP which is
roughly! in the United States. We therefore set o. = 0.33 . Moreover. as indicated in ( 60).
the parameter IJ determines the ratio of the volatility of labour input to the volatility of
output. so we set 'YJ = 0.83 to reproduce the observed ratio between the standard deviation
of total hours worked and the standard deviation of output. Given these choices. we have
set the values of wand o .. to ensure that the simulated volatility a nd persistence of output
matches the data as well as possible.
We see that, measured by the coelllcients of autocorrelation, our simple RBC model
produces too much persistence in output and labour input when we move more than one
year back in time, but otherwise themodel llts the data on output and hours worked quite
well. In addition, we have seen in Chapter 5 that the Solow model does a good job of mim-
icking the most important stylized facts of long-term growth . Amended '.vith stochastic
Table 19.3: The RBC model and the observed fluctuations in output and hours worked
Standard Standard deviation
deviation (% ) of output Autocorrelation Autocorrelation
relative to in output in hours worked
Hours standard deviation
Output worked of hours worked t- 1 t- 2 t- 3 t- 1 t- 2 t- 3
RBC model 1 3.42 2.84 0.83 0.75 0.50 0.23 0.75 0.50 0.23
The US economy 2 3.47 2.88 0.83 0.76 0.38 0.08 0.73 0.29 0.06
1. a= 0.33, t} = 0.83, w = 0.1, a 0 = 0.015.

2 . Annual data for the business sector.
Not e: The cyclical components of output and employment have been estimated by linear O LS det rending of annual data.
Source: Economic Outlook Database, OEC D.
15. This lag is documented in the highly readable article by Finn E. Kydland and Edward C. Prescott, ' Business Cycles:
Real Facts and Monetary Myth', Federal Reserve Bank of Minneapoli5, Ouarter/y Review, Spring 1990.
© The McGraw-Hill
Companies. 2005
productivity growth. this basic workhorse model thus seems able to describe the economy
in the short run as well as in the long run.
Some problems with basic real business cycle theory
From a theoretical perspective the great attraction ofRBC theory is that it offers a unilled
and parsimonious explanation lor growth and business cycles. According to the RBC
model above, the economy's short-run fluctuations as well as its long-term growth are
driven by the evolution of technology and the associated accumulation of capital. Thus it
is not necessary to develop separate theories for the long run and lor the short run .
Furthermore. the basic RBC story of the business cycle is very simple. relying only on
technology shocks: in our RBC model there was no need to postulate expectational errors
and/or nominal rigidities to generate business fluctuations. 16
Yet there are serious problems with the basic version ofRBC theory presented above.
despite the fact that it reproduces the data on output and labour input reasonably well.
One issue is whether exogenous technology shocks really are the main driver of business
cycles? Certainly we saw in Fig. 14.5 that the cyclical component of total factor produc-
tivity does account lor a sizeable fraction of the total output gap at business cycle peaks
and troughs. But are recessions really caused by periods of general technological regress?
Why should Hrms and workers suddenly lose the ability to use the existing technologies
and machinery as effectively as before?
One answer might be that negative shocks to productivity could reflect non-techno-
logical factors such as bad weather conditions affecting agricultural output. or sharp
increases in the real price of imported raw materials such as oil (which were seen in
Chapter 18 to work very much like a negative productivity shock). However, agriculture
only accounts for a minor share of GDP in developed economies. and although some inter-
national recessions have in fact been preceded by dramatic increases in oil prices, most
recessions seem to have been triggered by other events.
But if the observed fluctuations in total factor productivity do not reflect a highly
uneven pace of technical progress, and if they only occasionally reflect commodity price
shocks. how can they be accounted for? A plausible explanation is that most of the cycli-
cal variation in factor productivity stems from changes in the degree of utilization of
capital and labour caused by fluctuations in aggregate demand. In a recession ilrms may
keep unnecessary and undemtilized labour to be able to expand output quickly as soon as
demand recovers. When demand picks up, the underutilized labour starts to put in more
eflort. resulting in higher productivity and enabling Hrms to increase output lor a while
without increasing their hirings. This straightforward mechanism may explain the pro-
cyclical variation in productivity as well as the fact that the number of persons employed
lags behind the evolution of output. as documented in Chapter 14.
A second problem with basic RBC theory concerns the interpretation of the observed
lluctuations in employment. The otllcial unemployment statistics record substantial
variations in then umber of unemployed persons over the business cycle. According to our
RBC model these shifts in recorded unemployment reflect that workers voluntarily opt for
16. In more advanced RBC models expectational errors do occur, but they are not systematic and persistent, since
agents are assumed to have rational expectations. The next c hapter explains the concept of rational expectations.
unemployment when real wages are relatively low, preferring to consume more leisure
when work is not so well rewarded. Thus. while workers may be frustrated that market
wages have fallen (relative to the expected normal growth trend), they are not involun-
tarily unemployed. This denial of the phenomenon of involuntary unemployment seems
hard to reconcile with casual observation from serious recessions, in particular with
episodes of mass unemployment like the Great Depression of the 1930s. As some critics
have put it, if the standard RBC interpretation of unemployment were true, the Great
Depression should be renamed as the 'Great Vacation': hardly a convincing idea!
More generally, the idea th at all employment fluctuations reflect intertemporal sub-
stitution in labour supply fits poorly with empirical labour market research. To explain the
procyclical variation of employment as a result of voluntary movements along individual
labour supply curves. the real wage must be procycHcal. Table 14.3 showed that the real
wage is indeed positively correlated with output in the United States, but the correlation is
fairly low. To generate the large procydical pattern of employment despite the very
modest cyclical movements in the real wage, the wage elasticity of labour supply has to be
very high. But empirical research suggests that labour supply elasticities are in fact quite
low, at least for the core groups in the labour market.
=
To illustrate the problem, recall that we had to set our parameter 'YJ e/(1 + t) equal
to 0.83 for our RBC model to reproduce the observed volatility of employment relative to
the volatility of output. This implies that the wage elasticity of labour supply, e, has to be
as high as 4 .9. The empirical labour supply elasticities estimated for the various groups in
the labour market are much lower than th is, even if one accounts for labour market entry
and exit as well as variations in working hours for those who are permanently in the
market. Moreover, while the US real wage is in fact (slightly) procyclical. as predicted
by RBC theory, Table 14.3 documented that the real wage tends to be (slightly) coullter-
cyc/ical in several European countries, in direct contrast to the RBC story.
Our AS-AD model does not have this problem in dealing \Vith the labour market,
since its microloundations imply that there is involuntary unemployment at the natural
employment rate. As we explained in Section 3 of Chapter 18, this means that (some)
workers are forced ojftheir individual voluntary labour supply curves. and that aggregate
labour supply in ellect becomes infinitely elastic up to the point where the pool of involun-
tarily unemployed workers is depleted.
The lasting contribution of real business cycle theory
Of course real business cycle theorists are not unaware of the problems mentioned above.
In trying to reconcile their models with the stylized facts of the business cycle, RBC theo-
rists have gradually modiHed and extended their models to account lor various market
frictions. in some cases including nominal rigidities. At the same time practitioners of the
more comentional AS-AD approach to business cycle analysis have tried to meet the chal-
lenge from RBC theorists by paying more attention to the potential role of supply side
shocks and to the propagation mechanisms stressed by RBC theorists. They have also tried
to provide more convincing microeconomic foundations for the various frictions causing
output and employment to deviate from their natural rates.
As a result, recent years have seen a convergence of the different approaches to
business cycle analysis. Today most researchers recognize that RBC theorists have made
© The McGraw-Hill
Companies. 2005
acroeconomics Economy and aggregate demand
590 PART 6: THE SHORT-RUN MODEL FO R THE CLOSED ECONOMY
valuable methodological contributions to the study of business cycles. First of all. RBC
theorists introduced the practice of constructing and simulating stochastic dynamic
general equilibrium models whose predictions could be compared with the stylized facts of
the business cycle. In this way they introduced more rigorous theoretical and empirical
standards for judging whether a particular theory of the business cycle is 'good' or 'bad' .
The present book, as well as most modern research, is much inspired by this approach to
business cycle analysis. Second, RBC theory has drawn attention to the fact that many
mechanisms on the economy's supply side, such as the dynamics of capital accumulation.
may help to explain the persistence observed in macroeconomic time series.
Thus real business cycle theorists have made important lasting contributions at the
methodological level even if their specific early models were not very convincing.
19.5 Summary
················································································································································································
1. The AS-AD model determines the short-run equilibrium values of output and inflation as the
point of intersection between the upward-sloping short-run aggregate supply curve (the
SRAS curve) and the downward-sloping aggregate demand curve (the AD curve). The model
also determines the long-run equilibrium levels of output and inflation as the point of inter-
section between the vertical long-run aggregate supply curve (the LRAS curve) and the AD
curve.
2. When expectations are static, the expected inflation rate for the current period equals the
actual inflation rate observed during the previous period. Under this assumption the AS-AD
model is globally stable, converging gradually towards the long-run equilibrium where output
is at its natural rate and inflation is at its target rate. The adjustment to long-run equilibrium
takes place through successive shifts in the SRAS curve, as economic agents gradually
revise their inflation expectations in reaction to observed changes in the actual inflation rate.
During the adjustment process the economy moves along the AD curve, as the central bank
gradually adjusts the real interest rate in reaction to the changes in the inflation rate.
3. W ith plausible parameter values the AS-AD model suggests that it will take about four years for
the economy to complete half of the adjustment towards the steady state and about 13 years to
complete 90 per cent of the adjustment. Hence output and inflat on may deviate from their trend
values for quite a long time, once the long-run equilibrium has been disturbed by a shock.
4. The AS-AD model may be specified in deterministic terms or in stochastic terms. In the deter-
ministic versior of the model, the demand and supply shock variables are non-stochastic. In
the stochastic AS-AD model, the shocks to demand and supply are treated as random
variables.
5. The deterministic model may be used to study the isolated effects of a single temporary or
permanent shock to supply or demand. The stochastic AS-AD model may be used to gener-
ate simulated time series for output and inflation, allowing a calculation of the variance,
covariance and autocorrelation in these variables. These statistics may then be compared to
the statistical properties of empirical time series to investigate how well the stochastic AS-AD
model is able to reproduce the stylized facts of the business cycle.
19 EXPLAINING BUSINESS CYCLES 591
6. In the short run, a temporary negative supply shock generates stagflation, defined as an
increase in inflation combined with a fall in output. Even after the negative supply shock has
disappeared, inflation will remain above the target rate and output w ill remain below the
natural rate for many successive periods, because it takes time for inflation expectations to
adjust back to the target inflation rate.
7. A temporary negative demand shock generates a bust-boom adjustment pattern in output. In

the period when the negative demand shock occurs, there is a drop in output as well as infla-
tion. But when the negative demand impulse d isappears, output rises above its natural rate
because the SRAS curve shifts down, due to a fall in expected inflation. In the subsequent
periods, output gradually falls back to its trend level, and inflation gradually rises towards the
target rate.
8. By quantifying the parameter values, the deterministic AS-AD model may be used to generate
impulse-response functions showing how output and inflation will evolve over time in
response to an impulse such as a temporary or permanent demand or supply shock.
9. The deterministic AS-AD model can explain the observed persistence (autocorrelation) in
P.r.onomir. limP. SP.riP.S, lh~t is, it r.~n P.Xf'll~ i n why P.VP.n ~ IP.m['lOr~ry shor.k QP.nP.r~IP.S rrotr~r.tP.cl,
long-lasting deviations of output and inflation from their trend levels. However, the determi-
nistic model cannot explain the observed recurrent cyclical fluctuations in macroeconomic
variables. But if the demand and supply shock variables are treated as stochastic processes,
the AS-AD model is able to generate irregu lar, cyclical fluctuations in output and inflation.
10. A calibrated stochastic AS-AD model with static expectations where all shocks take the form
of demand shocks is reasonably good at reproducing the statistical properties of the empiri-
cal time series for output, but such a model generates an unrealistically high degree of per-
sistence ir the rate of inflation. A stochastic AS-AD model with static expectations where all
shocks originate from the supply side is incapable of reproducing the statistical properties of
the time series for output as well as inflation. This suggests that a model intended to explain
the business cycle must allow for demand shocks as well as supply shocks and that the
assumption of static inflation expectations may be too simple.
11. The hypothesis of adaptive expectations says that the expected inflation rate for the current
period is a weighted average of all inflation rates observed in the past, w ith more weight being
put on the experience of the recent past than on the more distant past. Static expectations are
a special case of adaptive expectations where the weight given to last period's observed
inflation rate is 100 per cent.
12. A calibrated stochastic AS-AD model with adaptive expectations allowing for demand shocks
as well as supply shocks is able to reproduce the statistical properties of the time series for
US output and inflation reasonably well. The stochastic AS-AD model offers an explanation for
the business cycle in line w ith the Frisch-Siutzky paradigm. In this paradigm, business fluctu-
ations are initiated by random demand or supply shocks which are then propagated through
the economic system in a way that generates persistence in macroeconomic variables. In our
AS-AD model, the persistence in the macroeconomic time series stems from the fact that the
expected and actual inflation rates adjust sluggishly over time. However, one can construct
other macroeconomic models w ith other persistence mechanisms, so our AS-AD model is
not the only possible explanation for business cycles.
© The McGraw-Hill
Companies. 2005
13. Real business cycle (RBC) theory seeks to explain business cycles by fluctuations in the rate
of technological progress which is also driving the process of long-term economic growth. A
Solow model in wh ich the rate of productivity growth is stochastic does a fairly good job of
reproducing the short-run dynamics of output and total hours worked as well as mimicking the
long-run facts of economic growth. The model generates endogenous persistence, since a
positive productivity shock increases output which in turn increases saving, feeding into an
increase in the capital stock which generates another round of increase in output and saving,
and so on.
14. Despite its theoretical attraction, basic RBC theory has a problem explaining the observed
fluctuations in aggregate employment as the outcome of intertemporal substitution in labour
supply. The theory postulates that workers voluntarily choose to work less when real wages
are relatively low, and vice versa. For this theory to fit the data, the elastic ity of labour supply
must be much higher than prevailing empirical estimates. Yet RBC theory has made an impor-
tant methodological contribution by introducing the practice of setting up stochastic dynamic
general equilibrium models wh ich can be compared with the stylized facts of the business
cycle.
Exercises
Exercise 1. Temporary shocks in the deterministic AS-AD model
1. Use the AS-AD model with static expectations to undertake a graphical analysis of the effects
of a positive supply shock (a fall in s) wh ich lasts for one period. You may assume that the
economy starts out in long-run equilibrium in period 0 and that the temporary shock occurs in
period 1. Expla n the short-run effects as well as the economy's adjustment over time.
2. Suppose now that the positive supply shock emerging in period 1 also persists throughout
period 2 before it goes away from period 3 onwards. Illustrate the effects of this two-period
shock graphically and explain the difference compared to the one-period shock analysed in
Question 1. Be careful to explain exactly how the SRAS curves shift in the two scenarios.
(Hint: indicate precisely where the SRAS curves cut the LRAS curve in the various periods.)
3 . Now use the AS-AD model w ith static expectations to perform a graphical analysis of the
effects of a positive demand shock occurring in period 1 and dying out from period 2
onwards. Explain the mechanisms underlying the evolution of output and inflation over time.
4. Assume instead that the positive demand shock emerging in period 1 lasts for two periods
before it d ies out. Illustrate the effects graphically. W ill the fluctuations in output and inflation
be larger or smaller than those occurring when the shock lasts for only one period? Explain.
Exercise 2. Permanent shocks in the deterministic AS-AD model
1. Suppose that the economy is hit by a permanent negative demand shock, say, because a tax
reform permanently raises the private sector's propensity to save. Illustrate by a graphical
analysis that such a shock will lead to a permanent violation of the inflation target if the central
bank does not adjust its estimate of the equilibrium real interest rate, l .
19 EXPLAINING B USINESS CYCLES 593
2. Now suppose that the economy is in long -run equilibrium in period 0 and that the perma-
nent negative demand shock hits the economy from period 1 onwards. Suppose further that
it takes the central bank only one period to realize that the shock is permanent. From
period 2 onwards, the bank thus revises its estimate of the equilibrium real interest rate
from the original level 70 to the new level g iven by Eq. (25) in the text. Analyse mathemati-
cally and graphically how these developments w ill shift the AD curve in periods 1
and 2. Illustrate and explain the economy's adjustment over time. Compare your graphical
analysis w ith the analysis of a purely temporary negative demand shock in Fig. 19 .6 . Is there
any difference?
3. Suppose alternatively that it takes until period 3 before the central bank realizes the perma-
nency of the negative demand shock and revises its estimate of the equilibrium real interest
rate. Perform a graphical analysis of the economy's adjustment from period 0 onwards and
compare with the analysis in Question 2. (Be precise in your indication of the shifts in the
AD and SRAS curves.) How does the delayed adjustment of the estimate for r affect the
magnitude of the fluctuations in output and inflation? Explain.
4. Suppose a!=jain that the economy starts out in lon!=j-run equilibrium in period 0 and assume
that a permanent negative supply shock (a rise in s) occurs from period 1 onwards, reducing
natural output from Yo to the new lower level y =Yo- (s/y). Suppose further that the central
bank realizes the permanency of the shock already in period 2 and hence revises its estimate
of the equilibrium real interest rate in accordance w ith Eq. (24) from period 2 onwards so that
the Taylor ru le becomes:
r, = 70 + ~
ya 2
+ h(:n:, - n *) + b[y, - (ro-~)]
y
for t ;;. 2.
Derive the equation for the new AD curve which will hold for t;;. 2 (where will the new AD
curve cut the new LRAS curve?). Illustrate the new short-run equilibria in periods 1 and 2.
(Hint: note that from period 2 onwards, the SRAS curves w ill cut the new LRAS curve at the
level of the previous period's inflation rate.) Illustrate by arrows along the new AD curve how
the economy w ill evolve after period 2 and explain the economic adjustment mechanisms.
Compare your results with the analysis of a purely temporary negative supply shock in
Fig. 19.5.
Exercise 3. Interest rate smoothing in the AS-AD model
We have so far assumed that the interest rate adjusts immediately to the target level g iven by
the Taylor ru le. However, empirical research has found that central banks tend to adjust their
interest rates only gradually towards the target level because they do not like to change the
interest rate too abruptly. We may model such 'interest rate smoothing' by assuming that:
r1 =(1-A-)r7 +..tr,_1 = r,-r, _1 =(1-..t)(r7 -r1_ 1 ) , 0 .;; J < 1, (6 1)
where J is a parameter indicating the sluggishness of interest rate adjustment, and where r7
is the target real interest rate g iven by the Taylor ru le:
r7 = 7+ h(n, - :n:*) + b(y, - y). (62)

© The McGraw-Hill
Companies. 2005
In the main text we have analysed the special case where J = 0, but here we will focus on the
general case where J > 0. The goods market equilibrium cond ition may still be written as:
y 1- Ji = a 1(g- g)- a 2(r, -1) + v,, (63)
and the economy's supply side is still g iven by:
n 1 = :n:~ + y(y1 - Ji) + s 1• (64)
As a new element, we will assume that economic agents are sufficiently sophisticated to
realize that on average over the long run, the inflation rate must equal the central bank's infla-
tion target. Hence we assume that the expected inflation rate is:
for all t. (65)
This completes the description of our revised AS-AD model.
1. Discuss whether the assumption made in (65) is reasonable.
2. Define y1 "" y1 - Ji and fc 1 "" :n: 1 - n* and show that the AD curve is g iven by the equation:
ji1 =py 1_ 1 - ail: 1 +z1 - ) z 1_ 1,

(66)
3 . Show that the model can be reduced to the d ifference equation:
1
ji, = fJpji,_1 + fJ(z,- J z 1_ 1 ) - a{Js,, {Je - , (67)
1 +ya
and prove that, in the absence of shocks, the economy w ill converge on a long-run equilibrium
where y1= fc 1= 0 . How does the parameter A affect the economy's speed of adjustment?
Explain.
4. Suppose that the economy starts out in a short-run equilibrium in period 0 where Yo < y. G ive
a graphical illustration of the initial short-run equilibrium and of the economy' s adjustment to
long-run equilibrium, assuming there are no further shocks. (Hint: note that the adjustment
now takes place th rough successive shifts in the AD curve rather than through shifts in the
SRAS curve.) Explain the economic mechanism wh ich ensures convergence on long-run
equilibrium.
5. Now assume that the economy is in long-run equilibrium in period 0 but is hit by a temporary
negative supply shock (a rise in s 1) in period 1. The shock van ishes again from period 2
onwards. Give a graphical illustration of the new short-run equilibria in periods 1 and 2. Could
y 2 be lower than y 1 ? (Hint: use (67) to justify your answer.) Indicate by arrows (along the
SRAS curve) how the economy w ill evolve after period 2. Explain your findings.
6. Suppose again that the economy is in long-run equilibrium in period 0, but assume now that
it is hit by a temporary negative demand shock wh ich occurs only in period 1. Illustrate graph-
ically how th is will affect the economy in periods 1 and 2 and use arrows to indicate the
economy's adjustment after period 2. Explain your findings. (Hint: you may want to use Eqs
(66) and (67) to show that the AD curve fo r period 2 w ill lie above the AD curve for period 0.)
Exercise 4. Developing and implementing a stochastic AS-AD model

In this exercise you are asked to set up a stochastic AS-AD model, implement it on the com-
puter, and undertake some simulations to illustrate the effects of monetary policy. In this way
you w ill become fam iliar w ith the modern methodology for business cycle analysis wh ich was
described in Section 3 of th is chapter.
Our starting point is a generalized version of the short-run aggregate supply curve. Many
econometric ians studying the labour market have found that wage inflation is moderated not
only by the level of unemployment (u1 ), but also by the increase in the unemployment rate
between the previous and the current period (u,- u, _,). The reason is that, ceteris paribus, it
is more difficult for a d ismissed worker to find an alternative job when unemployment is rising
than when it is fall ing. When labour is mobile across sectors, a rising unemployment rate thus
reduces the value of the representative worker's outside option. Assuming static inflation
expectations (n~ = n 1_ 1 ), we therefore get the following generalized version of the expec-
tations-augmented Phillips curve:
n 1 =n1_ 1 +y(D-u 1)-yB(u1 - u 1_ 1 ), y > o, e> o (68)
where D is the constant natural rate of unemployment, and the parameter e indicates the
degree to which the wage claims of workers are moderated by rising unemployment. We
assume that current output is g iven by a simple linear production function of the form Y, = a ,L v
where a , is the exogenous current average labour productivity, and L, is current employment.
If the constant labourforce isN, we have L, "' (1 - u,)N. Using the approximation ln(1 - u)"' u 1
and the definition y,"' In Yv we thus have:
y, =In a, + InN- u,. (69)
Trend output (natural output) is Y=a[= a(1 - D)N, where [is natural employment and a is the
'normal' (trend) level of productivity. Hence we have:
y, =Ina+ InN- D. (70)
Let us define the supply shock variable:
s, "' In a -In a,. (7 1)
Note that s, is positive when productivity is unusually low, that is, when unit labour costs are
above their normal level.
1. Use (68) - (7 1) to demonstrate that the economy's short-run aggregate supply curve may be
written as:
it,= .n,_, + y(1 + B)y, - yey,_, + y(1 + B)s,- yes,_,, (72)
y, "'y, -y, n- 1 "'n 1 - n * .

How does Eq. (72) deviate from the SRAS curve in the model in the main text? Explain
briefly why the lagged output gap y,_1 appears w ith a negative coefficient on the right-hand
side of (72). Explain in economic terms how the parameter e affects the sensitivity of inflation
to the current output gap.
As usual, the economy's aggregate demand curve is g iven by:
y,= z,- an,, (73)
where n* is the central bank's inflation target.

© The McGraw-Hill
Companies. 2005
2. Show that the solutions for the output gap and the inflation gap take the form:
{74)
{75)
where
1
fJ "' ' a 1 "'{J(1 + yo.fJ), a2 "' {Jya.( 1 + 8), {76)
1 +ya(1 +8)
c1 "' {Jy(1 + 8), {77)
{Hints: write Eq. (72) for period t + 1, insert {73) into the resulting expression and collect
terms to get (74). Then use {73) to write (74) in terms of ii rather than y and collect terms to
get (75).)
The demand and supply shock variables are assumed to follow stochastic processes of
the form:
o .;; o < 1, x, - N(o, a;), x, i.i.d. {78)
s1• 1 = ws 1-ct+ 1 , 0 .;; w < 1, c, - N(O, a~), c 1 i.i.d. {79)
We also remember that:
1](1 -r) a h
a2 =--o-·
2
(J.=--- (80)
1- y 1 +a.2b'
where the parameters are as defined and explained in the main text of this chapter.
3. Implementing the model. In order to undertake simulation exercises, you are now asked to
program the model consisting of (74)- {80) on the computer, for example by using Microsoft
Excel. From the internet address www.econ.ku.dk/pbs/courses/mode/s&data.htm you can
download an Excel spreadsheet with two different 1 00-period samples taken from the stan-
dardized normal distribution. You may take the first sample to represent the stochastic shock
variable x1, and the second sample to represent the shock variable c 1• In your first Excel
spreadsheet you should list the parameters of the model as well as the standard deviations of
the output and inflation gaps, the coefficient of correlation between the two gaps, and the
coefficients of autocorrelation emerging from your simulations. It will also be useful to include
diagrams illustrating the simulated values of the output gap and the inflation gap. We suggest
that you use the parameter values:
y= 0.05, r = 0.2, I] = 3.6, Oy = 0 .8,

h = b = 0.5, f)= 0.5.
{all of these values except the one for 8 were used in the main text). To calibrate the mag-
nitude of the demand and supply shocks, x 1 and k 1, you must choose their respective standard
deviations a x and a c and multip ly the samples taken from the standardized normal distribution
by these standard deviations. As a starting point, you may simply choose:
You also have to choose the value of the parameters oand w. For a start, just set:
0 = 0.5, w= 0.5.
Finally, you must choose the initial values of the endogenous variables in period 0. We
assume that the economy is in a long-run equilibrium in period 0 so that:
.Yo =iro = Zo= So= O fort = 0 .
Now program this model in Excel and undertake a simulation over 1 00 periods. Compare the
model-generated statistics on the standard deviations and the coefficients of correlation and
autocorrelation of output and inflation with the corresponding statistics for the US given in the
bottom row of Table 19. 1 in the main text. Comment on the differences.
4. Experiment with alternative constellations of the parameters a., a co a and w until you find a
constellation which enables your AS-AD model to reproduce the US business cycle statistics
for output and for the correlation between output and inflation reasonably well. {You should
not bother too much about the behaviour of inflation implied by the model, since we know
from the text that a model w ith static inflation expectations is not very good at reproducing the
statistical behaviour of inflation.) State a set of values for a x• a c • c5 and w which in your view
gives a reasonable account of the behaviour of output and of the correlation between output
and inflation. What does your choice of parameter values imply for the relative importance of
demand and supply shocks? Comment.
5. The simulations above use the monetary policy parameter values suggested by John Taylor,
that is, h = b = 0.5. Now suppose that the central bank decides to react more aggressively to
changes in the rate of inflation by raising the value of h from 0.5 to 1.0. Simulate your model
to investigate the effects of such a policy change. Try to explain the effects. Discuss whether
the policy change is desirable.
6. Suppose again that the central bank decides to raise h from 0.5 to 1.0, but suppose also that
supply shocks are very important so that a c = 5. Is the policy change now desirable? Discuss.
Exercise 5. Evaluating real business cycle theory
Discuss the main ideas and assumptions underlying the basic theory of real business cycles.
Explain and discuss the d ifferences between the basic RBC model and our AS-AD model of
the business cycle. Explain and discuss the arguments for and against the real business cycle
approach. {Hint: apart from the paper by Mankiw mentioned in Footnote 1 1 of this chapter,
you may also want to consult the article by Charles I. Plosser, 'Understanding Real Business
Cycles', Journal of Economic Perspectives, 3, 1989, pp. 51 -77.)
0 1 Sorensen-Whitta-Jacobsen:
20. Stabilization policy:
Why and how?
© The McGraw-Hill
Companies. 2005
Macroeconomics Economy
Chapter
Stabilization policy
Whyandhow?
'------·•
••
L--------·
•
re economic recessions and depressions a social evil which economic policy
A makers should try to prevent, or do they reflect the economy's optimal response to
exogenous shocks? During the 1980s and early 1990s several economists
actually defended the latter view. as we saw in the previous chapter. This school of real
business cycle theorists claimed that shocks to productivity are the main driver of business
cycles. and that the economy is reasonably well described by the general equilibrium
model of a perfectly competitive economy. In such an economy Pareto-optimality always
prevails, so even though policy makers may regret that economic activity falls in response
to a negative productivity shock. there is nothing they can do to improve the situation.
given that the economy always makes efficient use of the available resources and
technologies.
Although th is view of the business cycle still has its advocates. most economists
nowadays believe that the competitive real business cycle model is an implausible theory
of short-run business fluctuations . for the reasons given at the end of the previous chapter.
The dominant view is that business cycles reflect market failures and that social wellare
could be raised if business fluctuations could be dampened. A fundamental issue for
macroeconomics is whether policy makers can indeed smooth the business cycle through
macroeconomic stabilization policy.
Stabilization policy involves the use of monetary and fiscal policy to dampen fluctua-
tions in output and inflation. In this chapter we ofler an introduction to the problems of
macroeconomic stabilization policy. The llrst section of the chapter discusses the goals of
stabilization policy, asking why business fluctuations cause welfare losses, and how the
goals of stabilization policy should be dellned. In Section 2 we use our stochastic AS-AD
model from Chapter 19 to characterize the optimal rnonetary stabilization policy under
alternative assumptions regarding the goals of policy makers and the type of shocks
hitting the economy. Section 3 then analyses the optimaljlsca/ stabilization policy.
The formal analysis in th is chapter maintains our previous assumption that expecta-
tions are backlvard-looking. i.e.. that the expected future rate of inflation is determined
solely by the actual rates of inflation observed in the past. In Chapters 21 and 2 2 we shall
analyse the scope lor stabilization policy when expectations are .forward-looking in the
sense that agents try to form the best possible estimate of future inflation, utilizing all the
598
Sorensen-Whitta-Jacobsen: I Part 6- The Short-run 20. Stabilization policy: © The McGraw-Hill
Introducing Advanced Model for the Closed Why and how? Companies. 2005
20 STAB ILIZ ATIO N PO LICY 599
relevant information available to them at the time expectations are formed, including
knowledge of the way the economy works.
The present chapter also assumes that policy makers can inunediately observe and
react to fluctuations in the current levels of output inflation. The complications arising
when policy can only react to movements in output and inflation with a lag will be dis-
cussed in Chapter 22. That chapter will also consider the challenges for stabilization
policy emerging when there is uncertainty in the measurement of the output and inflation
gaps, and when the authorities have difllculties establishing the credibility of their com-
mitment to a low and stable inflation rate.
20.1 The goals of stabilization policy and the welfare costs of

?..~~.i?.~.S..S. . ~.Y.~.l~.S. ................................................................................................................................... .
The social loss function
The case for stabilization policy rests on the hypothesis that policy makers - and
ultimately their constituency. the general public - are averse to fluctuations in output,
employment and inflation. Given that employment and output tend to move together, this
supposed aversion to fluctuations is sometimes formalized in the 'social loss function·:
SL = o~ + K(r;. I(> 0, ( 1)
where E[>~ indicates the mean value of X. Equation (1) postulates that society's welfare
loss from macroeconomic instability, SL, increases with the variance cJ~ of the deviations
of output from its trend level y and with the variance (J; of the inflation rate around
its target level n *. The parameter K measures the degree to wh ich society values stable
inflation relative to output stability.
This social loss function raises a host of issues:
1. Why should society be concerned about the variability and not just the average values
of output and intlation?
2. Should output really be stabilized around its trend level regardless of the type of shocks
hitting the economy?
3. What is the appropriate target level of inflation?
4. Will a macroeconomic policy which reduces the variance of output also reduce
the variance of inflation, or could there be a trade-off between output stability and
in flation stability?
5. What are the properties of the optimal stabilization policy rules if policy makers wish
to minimize the social loss function (1)? In particular, should stabilization policy be
countercyclical, implying a loosening of monetary and fiscal policy when output falls,
and vice versa?
In this section we consider the first three of these challenging questions. The remain-
ing L~sues will be dealt with in Sections 2 and 3.
Why and how?
© The McGraw-Hill
Companies. 2005
600 PART 6: TH E SH O RT -RUN MO DEL FO R TH E C LO SED ECON OM Y
The welfare costs of consumption fluctuations
Let us start by considering why lluctuations in output around its long-run growth trend is
a social problem. One reason why this is so is that lluctuations in real output generate lluc-
tuations in real income and consumption. Since the marginal utility of consumption
declines with increasing consumption. consumers put more value on avoiding low levels
of consumption than on attaining very high consumption levels. Hence they prefer a
smooth path of consumption to an unstable path, so if stabilization policy can even out the
time path of consumption by smoothing lluctuations in output, a consumer welfare gain
\<\Till result.
One objection to this reasoning might be that consumers can use the private capital
market to smooth consumption over time, as we saw in Chapter 16. However, as we
mentioned in that chapter, capital markets are hardly perfect. so some consumers may be
subject to credit constraints. In particular, during a serious recession the market values of
many assets may fall considerably and destroy much of the collateral that consumers
normally use to obtain credit.
In addition to borrowing in the capital market. people can insure themselves against
temporary income losses by ta king out insurance in the private market, say, insurance
against unemployment. But here again we run into problems with the functioning of
markets. A person who is fully insured against any income loss from unemployment has
little incentive to avoid joblessness. If he values leisure more than the social interaction
obtained through his job, he may therefore choose to stay out of work for long periods of
time. Because of this so-called moral hazard problem it is suboptimal for private insurance
companies (as well as for the government) to oller full insurance against income losses
from unemployment.
On top of this, the demand for unemployment insurance is likely to come mainly from
those who face the highest risks of unemployment. If the premium lor unemployment
insurance is set to cover the average income loss from unemployment across a large group
of insured workers, individuals in very risky jobs h ave a strong incentive to buy insurance.
because the present value of their expected unemployment benefits will exceed the present
value of the insurance premium. But workers facing a relatively low risk of losing their
jobs may find that the premium is too high relative to their expected unemployment ben-
d iLs. T herefore:: Lh ey may deciue Lo drop o ul of llte insurance schem e. Because of Lltis so-
called adverse selection problem, private insurance companies may fear that those who
want to take out insurance represent only the highest risks, and companies may therefore
decide that ofJering insurance will not be profitable. Hence the adverse selection problem
may cause too little unemployment insurance to be supplied by the market.
Thus the problems of moral hazard and adverse selection limit the scope for
consumption smoothing through private credit and insurance markets. The moral hazard
problem also makes it suboptimal lor the government to oller full insurance against
private income losses through the system of public transfer payments. A successful stabi-
lization policy may therefore raise consumer welfare by evening out the time path of real
income. We may say that ellective stabilization policy provides a type of income insurance
which cannot be delivered through the private market or the public transfer system.
Economists differ in their perceptions of the severity of the market imperfections men-
tioned here. so it is not surprising that they also differ in their perceptions of the potential
Sorensen-Whitta- Jacobsen: I Part 6- The Short-run 20. Stabilization policy: © The McGraw-Hill
gain from eliminating fluctuations in consumption. Some economists, like the Nobel Prize
winner Robert E. Lucas, have argued that the quantitative gains from consumption
smoothing are very small for realistic values of the intertemporal elasticity of substitution
in consumption. 1 Others have argued th at Lucas underestimates the degree of consumer
risk aversion and hence the value that people attach to consumption stability. In support
of their argument. these economists point out that the big diflerence between the market
rates of return on stocks and bonds (the equity premium) observed historically is very
hard to explain unless consumers are highly risk averse.
The welfare costs of employment fluctuations
Apart from causing instability in consumption, tluctuations in total output also generate
fluctuations in aggregate employment. Such employment volatility generates welfare
losses on top of those arising from consumption fluctuations. To demonstrate this, we may
use the 'sticky-wage' model of a unionized labour market developed in Chapter 18.
Consider Fig. 20.1 where the horizontal axis measures the level of employment, L, while
the vertical axis measures the real wage, defined as the nominal wage rate, W, divided by
the price !eve!, P. The curve labelled MPL shows how the marginal product of labour
declines with the level of employment.
w
p
MRS
[ L' L
Figure 20.1: The natural versus the effic ient level of employment
1. See Robert E. Lucas, Jr., Models of Business Cycles, Oxford, Basil Blackwell, 1987.
Model for the Cl osed
20. Stabilizati on policy:
Why and how?
© The McGraw-Hill
Companies. 2005
602 PART 6 : THE SHORT-R UN MODEL FOR THE CLOSED ECONOMY
Recall from Chapter 18 that when labour is the only variable factor of production, the
marginal cost of the representative firm is MC = W/MPL. With monopolistic competition in
product markets, the representative firm will set its price as a mark-up over marginal cost. If
we denote the mark-up factor by rn''· we thus have P = mP MC = mPW /MPL, implying
W/P = MPLjmP.The curve MPL/m~' in Fig. 20.1 indicates the demand lor labour arising from
the profit-maximizing behaviour of the representative llnn. Since the mark-up factor 1n~' is
greater than 1. this labour demand curve lies below labour's marginal productivity cunre.
The wedge between the marginal product curve and the labour demand curve - which is
equivalent to the wedge between price and marginal cost - may be termed the product
market distortion, PMD, since it arises from imperfect competition in product markets.
The curve labelled MRS in Fig. 20.1 is a labour supply curve reflecting the represen-
tative worker's marginal rate of substitution between income and work effort. The MRS
measures the additional income necessary to compensate the worker for the loss ofleisure
implied by an extra hour of work. 2 We assume that the MRS is an increasing function of
employment, because the marginal value of leisure increases as employment rises and
consumption of leisure falls. A utility-maximizing worker will want to work up to the
point where his MRS is just equal to the real wage. If W/P> MRS. the worker can earn
more from an extra hour of work than the amount necessary to compensate him lor the
loss of an extra hour ofleisure. On the other hand, if W/P <MRS. the earnings from an
additional hour of work are not worth the utility cost of giving up more leisure. Hence a
utility maximum requires MRS = W /P, and since MRS is an increasing function of L, the
MRS curve indicates how the worker's desired labour supply responds to the real wage.
If all workers competed atomistically for work and the nominal wage rate were fully
flexible, the real wage would adjust to ensure that workers were always on their in-
dividual labour supply curves. The aggregate level of employment would then be given by
the pointE' ofintersection between the labour demand curve MPL/m1' and the MRS curve
in Fig. 20.1. However. we assume that trade unions have the market power to fix the
nominal wage rate so as to keep the real wage at some target level w above the individual
workers' MRS. provided union wage setters correctly anticipate the price level prevailing
over the period for which the money wage rate is set. Thus union power implies a labour
market distortion. LMD. which may be measured by the wedge between the real wage and
the representative worker's MRS. In Fig. 20.1 the level of employment therefore ends up
at L when unions correctly foresee the price level. At this employment level at least some
workers are forced ojJ their individual labour supply curves into a state of involuntary
unemployment where they would lil<e to work more at the going real wage w, if only more
jobs were available.
The employment level L is the naturnl level of employment defined as the level of
employment prevailing when the expected price level equals the actual price level so that the
actual real wage coincides with the target real wage. Figure 20.1 also shows the efficient
level of employment, L' . given by the intersection point E' between the individual labour
supply curve and labour's marginal productivity curve. This is the Pareto-optimal employ-
ment level which would prevail in a perfectly competitive economy. Whenever actual
2. If working hours are fixed, the MRS should be interpreted as the amount of income necessary to induce the 'marginal'
worker (the worl<er who has the highest valuat ion of leisure among all the workers employed) to enter the labour
market. With t his interpretation all the qualitative conclusions drawn below will still be valid.
Sorensen-Whitta- Jacobsen: I Part 6- The Short-run 20. Stabilizati on policy: © The McGraw-Hill
Introducing Advanced Model for the Cl osed Why and how? Companies. 2005
employment is below L•. the additional output generated by an extra hour of work. MPL.
exceeds the amount necessary to compensate the worker lor the loss of an extra hour of
leisure, MRS. Hence an increase in employment will generate a surplus above the amount
necessary to keep the worker's utility constant. From a social viewpoint it is therefore desir-
able to expand employment up to the point where MPL = MRS. that is. up to the level L'.
However, we see from Fig. 20.1 that the product market distortion (PMD) and the
labour market distortion (LMD) prevent the realization of the ef11cient employment level.
When employment is at the natural level. society sutlers a loss of economic efflciency
because [ falls short of the efficient employment level L' . The magnitude of this welfare
loss is shown by the h atched area in Fig. 20.1. This area is the dillerence between the area
below the MPL curve and the area below the MRS curve in the employment interval
between [ and L". The area below the MRS curve measures the total amount of real
income which is necessary to compensate workers for the loss of leisure implied by a rise
in work ellort from [to L ' . The area below the MPL curve gives the total increase in output
and real income generated by such a rise in work ellort. Thus the dillerence (the hatched
area) measures the net social gain from the hypothetical rise in employment from [to L' .
We are now able to illustrate the welfare costs of employment lluctuations. At the
start of each period. the union sets the money wage to obtain an expected real wage equal
to its real wage target, implying W = wP', where P' is the expected price level. Starting out
from a long-run equilibrium where P' = P so that WI P = wand L = [.suppose that some
positive aggregate demand shock causes unanticipated inllation. Because the nominal
wage rate is fixed by the union contract in the short run, the unexpected rise in the
price level will cause the real wage W/ P = wP'I P to fall below w , thereby increasing
employment from [ to a point such as L 1 in Fig. 20.2, as the representative linn moves
down along the labour demand curve. Since the wedge MPL - MRS = PMD + LMD
between marginal productivity and the marginal rate of substitution is positive at the
natural employment level [, the increase in employment generates a welfare gain equal to
the sum of the areas A and B. Thus an economic boom is welfare-improving whenever
initial employment is below the ellkient level.
However, suppose that the boom is followed by a recession w hich pushes employ-
ment down to the level L2 in Fig. 20.2. As the llgure is drawn, the fall in employment from
[ to L2 is of the same magnitude as the earlier rise in employment from [ to L 1•
Nevertheless, we see from the figure that the welfare loss from the recession - the sum of
the areas C and D- is considerably larger th an the welfare gain from the preceding boom
(A + B). Clearly the reason is that the wedge between MPL and MRS increases when
employment falls whereas it decreases when employment rises. The important point is
th at society would be better oll'if business lluctuations could be avoided, that is, if employ-
ment cou ld be stabilized at the natural level [instead of lluctuating around an average
level of [ . To be sure, a boom where L > [is always to be welcomed, other things equal,
but if booms tend to be followed by recessions where L < [, as experience suggests, eco-
nomic welfare would be higher if the fluctuations of employment around the natural rate
could be reduced in a synunetric fashion. 3 This provides a rationale for macroeconomic
stabilization policy.
3. This conclusion also holds if t he fluctuations in employment and out put are generated by fluctuations in the t rade
union's wage mark·up (fluctuations in the target real wage w) or by fluctuations in the price mark·up mP.
8 1Sorensen-Whitta- Jacobsen:
Why and how?
© The McGraw-Hill
Companies. 2005
Figure 20.2: The welfare effects of employment fluctuations
As you can see from Fig. 20.2. the welfare gain from a boom and the welfare loss from
a recession \Viii be greater the larger the magnitude of the wedge MPL - MRS = PMD +
LMD at the natural employment level [If one makes explicit assumptions about the forms
of production and utility functions, it is possible to estimate the wedge and the welfare loss
from business cycles on the basis of observable data. For example. if the production ti.mc-
tion tal<es the form Y = BL 1 -".where B is a constant productivity parameter and 0 <a< 1,
the product market distortion is MPL - HT/P = (1 - a)BL _,- W/P = (1 - a)(Y/L) - W/P.
Data on average labour productivity Y/L and on the average real wage W/P are directly
available, and the parameter o. can be estimated. 4 In this way an estimated time series for
PMD can be obtained.
Similarly, as a simplified example, suppose the utility function U = U(Y, L) of the
representative worker takes the particular quasi-linear form:
L~+l
U = Y - --, 1J > 0, (2)
IJ+ 1
where total real income Y consists of labour income (W/P)L plus the profits IT paid out by
the Hrms. In this case the marginal rate of substitution between income and work is simply
MRS= - UJUy = L~. so the labour market distortion becomes W/P - MRS = W/P - U .
Hence, LMD can be estimated on the basis of data for real wages and employment, given
an estimate of the parametern. Since a utility-maximizing worker will want to expand his
4. Recall that P=m• MC=m• WjMPL =m•W/((1- a)(Y/L)], implying WL/ PY=(1 -a)f m•. Data for t he labour income
share, WL/PY, are readily available, and the mark·up factor m• = P/ MC may ~e estimated provided one has access
to data for prices and marginal costs. An estimat e for a is then obtained as a= 1 - m•(WL/ PY).
Why and how?
© The McGraw-Hill
Companies. 2005
lCD
labour supply up to the point where the MRS and the real wage coincide. that is. L'1 = W/P
or L = (W/P) If~. the parameter 1J is the inverse of the real wage elasticity of individual
labour supply. This parameter has been estimated in numerous econometric studies.
Working with a more general utility li.mction than (2) . economists Jordi Gali, Mark
Gertler and Jose Lopez-Salida recently estimated the fluctuations in the PMD and LMD
wedges in the US economy. They then used their results to estimate the welfare costs of US
business cycles over the period fi·om the late 1960s to the mid 1990s. 5 Their 11ndings are
reported in Table 20.1. The ftrst three columns of the table show the year and quarter of
the start. the turning point and the end of the four business cycles completed during this
period, with the 11rst column indicating the start of the upswing, and the third column
indicating the end of the subsequent recession . The fourth column shows the estimated
average welfare gain enjoyed during the boom periods. while the fifth column gives the
estimated average welfare loss sutlered during the subsequent recessions. The gains and
losses are expressed as a percentage oftotal consumption. Finally the sixth column reports
the net welfare loss resulting from the fluctuations of employment over a full cycle. We see
th at the net losses are substantial. despite the fact that the business lluctuations during
th is period were less dramatic than those experienced during the interwar years and
earlier.
It should be recalled that the numbers in Table 20.1 do not capture all of the welfare
costs of recessions. Even if employment could somehow be kept constant, consumers
would still incur a utility loss when lluctuating incomes generate fluctuations in con-
sumption , as we explained in the previous section. Moreover, economic research has doc-
umented that the income losses caused by recessions are very unevenly distributed.
Manual workers- and in particular low-paid unskilled workers and young workers - tend
to sutler a much larger decline in consumption than other people during times of reces-
sion. 6 For these groups who already earn low incomes while they are employed, recurring
recessions and the associated spells of unemployment can imply a serious welfare loss.
Hence a society concerned about inequality should also be concerned about instability of
output and employment.
Table 20.1: The benefits(+) of booms and the costs(- ) of recession

Benefits and costs in% of one year's consumption
Time Period Boom Recession Net
Start Turning Point End

68:2 70:2 72:3 +6.50 - 9.4 - 2.9
72:4 74:4 77 :3 +7.30 - 22.2 -14.9
77:4 80:2 83:4 +8.70 - 16.8 - 8.1
87:4 90:4 94:1 +9.30 - 14.8 - 5.5
Source: Jordi Gali, Mark Gertler, and Jose David L6pez-Salido, ' Markups, Gaps and the Welfare Costs of Business Fluctuations' ,
C EPR Discussion Paper No. 3212, 2002, Table 4.
5. See Jord i Gali, Mark Gertler and Jose David Lopez-Salido, 'Markups, Gaps and the Welfare Costs of Business
Fluctuations', C EPR Discussion Paper No. 3212, 2002.
6. See Kenn~th Clark, Derek Leslie and Elizabeth Symons, 'The Cost of Recessions', Economic Journal, 104, 1994,
pp. 20-36.
Why and how?
© The McGraw-Hill
Companies. 2005
606 PART 6: THE SH O RT-RUN MO DEL FO R THE C LO SED ECON OM Y
Choosing the output target
By assuming that macroeconomic policy makers wish to minimize a social loss function
like (1), we are assuming that they seek to stabilize actual output around the trend level of
output, [J, corresponding to the natural employment level[. But wouldn't it be more rea-
sonable to assume that policy makers strive to keep actual output close to the ejjicient level
of output. y ' , corresponding to the elllcient employment level L"? After all, the ellkient
output level reflects a Pareto-optimal allocation of resources and therefore seems the most
natural candidate for the output target.
The trouble is that with imperfect competition in product and labour markets, trend
output [J is lower than the ellkient output y' , because prices are driven above marginal
costs, and because the real wage is driven above workers' marginal rate of substitution
between work and leisure. The fact that efficient output exceeds trend output raises a
problem lor any stabilization policy choosing y' as the output target. To see the problem,
recall from the previous chapters that the short-run aggregate supply curve in an
economy with static expectations is:
(3)
When y ' is systematically greater than 'fj. it follows from (3) that a policy which succeeds
in keeping actual output y 1 at the elllcient level y ' will tend w generate ever-accelerating
inflation. :rr. 1 > ::r , _ 1 , given that the average value of the supply shock variable s 1 is zero. At
some point such accelerating infl ation '<\Till clearly become intolerable, since it will drive
inflation ever further away from its target rate :n:*.
This line of reasoning assumes static inflation expectations, but even if expectation
formation is more sophisticated, reflecting a full understanding of h ow the economy
works. a stabilization policy pursuing an output target I( > y 'viii still imply a bias
towards inflat!on, as we shall explain in detail in Chapter 22 . Therefore it is usually
argued that stabilization policy should aim at stabilizing output around its trend level, since
this is consistent with maintaining a low and stable long-run rate of inflation. The gov-
ernment should then use structural policies such as labour market policies, tax policies and
competition policies to alleviate the imperfections in labour and product markets. thereby
pushing trend output closer to the elllcient output level.
Au o Llu:r pragm a tic arg urmml is lh al lh e !Jypolhelical couslrucl of lh e 'ellk ieul'
output level is very hard to estimate and hence cannot be used as an operational target for
stabilization policy.
The case for stable inflation
Why do fluctuations in the rate of inflation imply a social welfare loss, as postulated in (1)?
The answer may seem straightforward: in our AS-AD model a fluctuating rate of inflation
means that people generally fail to anticipate the rate of in flation correctly. When house-
holds and llrms underestimate or overestimate inflation, they end up with dillerent
real rates of return to their labour and capital than they expected. Due to these mis-
calculations. they will regret some of the economic decisions they made and 'viii therefore
suffer a welfare loss. Moreover. unanticipated inflation (an unexpectedly low real rate of
interest) implies an arbitrary redistribution of wealth from creditors to debtors, while an
unexpectedly low rate of inflation generates a sim ilarly arbitrary redistribution in the
opposite direction. Such unintended redistribution may threaten the stability of society if
it occurs on a large scale. This may be one reason why large fluctuations in inflation
create a disproportionately greater welfare loss than small fluctuations, as we implicitly
assume by including the varia11ce (rather than, say, the standard deviation) of inflation in
the social loss function (1).
Still, one could argue that these problems of unanticipated inflation could be avoided
if all nominal contracts were systematically indexed to inflation. For example, labour con-
tracts could state that nominal wage rates should rise in line with the general consumer
price index over the contract period, and debt contracts could stipulate that the principal
amount of debt be indexed to consumer prices. With such systematic indexation, un-
anticipated inflation and the associated welfare costs would basically disappear, and there
would be no reason for policy makers to worry about fluctuations in inflation.
However, despite its apparent advantages, indexation occurs only on a lim ited scale.
Presumably this is because the two parties to a con tract often worry about different prices.
For example, in a highly open economy the prices of imported consumer goods enter with
a heavy weight in the general consumer price index. If the prices of imports rise, workers
would want to be compensated through a cost-of-living indexation clause in their con-
tracts. But since the prices of domestic goods do not automatically follow the prices of
imports when the two types of goods are imperfect substitutes, domestic employers would
not want to compensate fully workers for the higher price of imports, since this would cut
into domestic profit margins. Thus employers focus on the evolution of the prices of domes-
tically produced goods relative to nominal wage rates, whereas workers focus on a weighted
avemge of1/w prices of domestic and imported goods relative to their nominal wage. Hence the
two parties would llnd it difllcult to agree on the dellnition ofthe relevant price index, even
if they agreed in principle on the desirability of indexation. In a similar way, an individual
employer may be reluctant to oller full indexation of his employees' wages to some general
price index if he fears that he will not always be able to raise his own price in line with the
general index. Wh atever the cause of the lack of widespread indexation, it seems legiti-
mate lor policy makers to worry about fluctuations in (he rate of inflation as long as we
have so little indexation.
Choosing the inflation target
Given that policy makers want a stable rate ofinflation, what is the appropriate target rate
n:* at which the inflation rate should be stabilized? It is tempting to answer that the target
inflation rate should be zero, since inllation generates social costs even if it is stable and
fully anticipated. These costs may be summarized as follows:
• 'Shoe/eather costs'. A higher inflation rate implies a higher nominal interest rate which
induces households and llrms to economize on their money balances. When people
hold a lower average stock of real money balances, they have to make more frequent
trips to the bank to withdraw money needed for transactions purposes. The resulting
'shoeleather costs' (which include the value of the time spent on trips to the bank. the
resources spent on cash management in firms, as well as other costs of exchanging
non-liquid assets for money) are part of the costs of inflation.
Why and how?
© The McGraw-Hill
Companies. 2005
608 PART 6 : THE SH O RT-RUN MO DEL FO R THE C LO SED ECON OM Y
• 'Menu costs'. Ifthe rate of inflation rises,llrms have to change their nominal prices more
frequently to avoid an erosion of their real prices and prollts. Price changes involve
resource costs which include the costs of printing new price lists and catalogues (or
menu cards. in the case of a restaurant) plus any other costs of communicating the new
prices to the market.
• Relative price distortions. Prices of individual products are only changed at discrete time
intervals. and the price changes of individual firms are not synchronized. At any point
in time some llrms have recently adjusted their prices, whereas others are still stuck
with prices that were set some time ago. The fact that difierent llrms have not caught
up with inflation to the same degree means that consumer choices are distorted: even
llrms with the same level of marginal costs may charge different prices on any given
day, simply because they do not adjust their prices at the same time. This relative price
distortion will tend to be more severe, the higher the rate of inflation .
• Distortions due to an unindexed tax system. A positive rate of inllation drives a wedge
between the nominal and the real rates of return on saving and investment. The
income tax is typically levied on the entire nominal interest rate, including that part
which the investor must set aside to preserve the real value of his nominal asset. As
a result. inllation may cause the return to nominal assets to be overta'<ed relative to
real assets such as land and buildings. This will tend to distort savings and investment
decisions.
Although many economists believe that all of these costs of anticipated inllation are
relatively small as long as inflation is moderate. the fact that they are positive nevertheless
suggests that the target inflation rate sh ould be zero. However. a zero inflation rate may
impair the central bank's ability to stabilize the economy. With zero inflation, the short-
term nominal interest rate will typically be rather low. In that situation the central bank
will not be able to counter a severe recession by a large cut in the interest rate. since the
nominal interest rate cannot tall below zero. If policy makers want interest rate policy to
be an ell'ective tool of stabilization policy, they may therefore have to accept a positive
average rate of inflation to preserve room for substantial interest rate cuts in times of
recession.
The inability to cut the nominal interest rate at very low rates of infl ation is not just a
theoretical possibility. In Japan, which has experienced economic stagnation and genuine
deflation in recent years. the short-term nominal interest rate has in fact been driven
almost all the way to zero without putting an end to the economic malaise. as illustrated
in Fig. 20.3. At a time when a boost to aggregate demand is badly needed, the central
bank ofJapan has thus lost its ability to stimulate demand through its interest rate policy.
As a further argument against a zero inflation target, several economists have
pointed out that nominal wage rates tend to be particu larly sticky in the downward direc-
tion. If the economy is hit by a negative shock which calls lor a fall in real wages, this
adjustment is therefore easier to achieve through price inflation than through a fall in
nominal wages. This is one additional reason why it may be easier to stabilize the real
economy if there is a moderate positive rate of inflation.
Finally, olllcial price indices tend to overestimate the true rate of inllation since they
cannot fully account for the fact that some price increases reflect improvements in
product quality rather than genuine inflation.
20 S TA B ILIZ A TIO N PO LI CY 609
8
Unemployment rate
4
'$.
2
0
-~--- - --- - ___ ,,,'
Consumer price inflation
-2
1990 1992 1994 1996 1998 2000 2002
Figure 20.3: Economic stagnation in Japan

Sources: Eo)nomic Outlook Database, OECO. Ecowin.
For these reasons countries '>\lith explicit otllcial inflation targets such as Australia,
Canada, I\ew Zealand, Norway. Sweden, Switzerland and the UK. have typically accepted
a positive target inllation rate of 2- 2.5 per cent per annum. The European Central Bank
also accepts an inflation rate of up to 2 per cent as being consistent '>\lith its goal of price
stability.
Rules versus discretion
Apart from choosing their targets lor output and inllation. policy makers must also
decide whether they wish to follow a fixed policy ntle. or whether they prefer to be left with
discretio11 in their policy choices.
Under the rules approach, stabilization policy is essentially automatic, since the
policy rule prescribes how the policy instruments should be set in any given situation. The
Taylor rule discussed in Chapter 1 7 is an example of a fixed monetary policy rule specify-
ing h ow the central bank interest rate should be set, given the observed state of the
economy. The Friedman rule prescribing a constant growth rate of the nominal money
supply is another example of a fixed monetary policy rule. In an open economy a Hxed
exchange rate regime can also be seen as a monetary policy mle which requires the
central bank interest rate to be set so as to stabilize the foreign exchange rate.
By contrast. under discretion policy mal<ers are free to conduct monetary and
Hscal policy in any way that they believe '>Viii help advance the objectives of stabilization
policy in any given situation. The idea is that policy mal<ers should use all the available
in formation, including advice from economic experts, and take account of any special
circumstances which might prevail.
Note that the distinction between m les and discretion is not equivalent to the distinc-
tion between passive and active macroeconomic policy. For example. while the Friedman
rule requires monetary policy to be passive by sticking to a constant monetary growth rate
8 1Sorensen-Whitta- Jacobsen:
Why and how?
© The McGraw-Hill
Companies. 2005
regardless of the state of the economy, the Taylor rule implies th at the central bank
actively reacts to changes in inllation and the output gap.
It might seem that discretion should be preferred to rules, since reliance on a simple
fixed policy rule reduces the ability of policy makers to react to all relevant information.
Many economists nevertheless believe that stabilization policy should be based on simple
rules. They argue that rules can help policy makers to establish credibility which will help
them realize the goals of stabilization policy. For example. if the natural unemployment
rate is rather high, the public may suspect that policy makers have an incentive to drive
down the actual unemployment rate by creating surprise inflation. In these circum-
stances it may be diflicult for policy makers to convince the private sector that they are
committed to securing a low and stable rate ofinllation. The fear ofinflation will then keep
the expected and actual inflation rate at an uncomfortably high level which can only be
reduced by generating a serious recession. However, suppose policy makers put their rep-
utation at stake by publicly announcing that they are bound by a policy rule requiring
them to respond strongly to any increase in inllation. It may then be easier to convince the
markets th at inflation will indeed be kept low. In this way a policy rule may serve as an
anchor lor (low) inllation expectations (we shall elaborate on th L~ argument in
Chapter 22). Similarly, if policy makers announce a rule wh ich implies that policy is auto-
matically tightened when output rises and automatically eased if output falls, this may
help to stabilize the growth expectations of the private sector which in turn '>\Till help to
stabilize aggregate demand.
Of course. policy makers can only 'buy' credibility by announcing a llxed policy rule
if they can convince the public that they are really bound by the rule. This will be easier if
the rule is written into the Jaw or into the mandate or statutes of the central bank, and if
policy makers face some kind of sanction if they break the rule. For example, the law reg-
ulating the central bank could specify that the central bank governor will be llred if he
consistently misses the bank's inflation target. For a rule to be credible, it must also be rea-
sonably simple so the public can easily understand the rule and check that policy makers
actually stick to it. 7
The advocates of discretion argue that fixed policy rules can only establish credibility
if they are overly simple and rigidly adhered to. Hence credibility can only be bought at the
price of.flexibility: by sticking to a simple policy rule no matter what the situation is, policy
makers lose the ability to account for whatever special circumstances might prevail. For
instance. suppose th at because of some unexpected event the stock market suddenly takes
an exceptional plunge which provides good reason to believe that the economy is headed
lor a deep recession. In that situation, should the central bank really keep the interest rate
unch anged just because stock prices do not enter into the policy reaction function it has
announced?
Most adherents of rules acknowledge that complete loss of llexibility would be a
problem. For example. John Taylor has not argued that central banks should slavishly
follow his rule. Rather, they should use it as a guideli11 e to be followed under normal
circumstances, but deviations from the rule would be acceptable when exceptional
circumstances prevail.
7. In Chapter 22 we will analyse the issue of rules versus discretion and the case for central bank independence in
greater detail.
Why and how?
© The McGraw-Hill
Companies. 2005
lCD
In practice central banks do not announce a fJXed quantified interest rate reaction
function, presumably because they wish to preserve some amount of policy flexibility.
Nevertheless, as we saw in Chapter 17, the interest rate policies of the most important
central banks seem to be fairly well described by the Taylor rule. In the following we will
therefore assume that stabilization policy is in fact based on rules.
20.2 M.?.I.l.~~~.r.:Y. . S.~.?:~.i.l~. <:l~.i.C>.~ ..P.?.l.i~y .......................................................................................... .

Having di~cussed the goals of stabilization policy. we will now study how such policy
should be conducted, given th at policy makers wish to minimize a social loss function of
the form (1), and given that policy follows a rule. In this section we will focus on monetary
stabilization policy, postponing the investigation oftlscal policy until Section 3.
The theoretical framework
Our analysis will be based on the stochastic AS-AD model set up in Chapter 19. restated
here in the usual notation, where variables without a subscript refer to the current time
period:
Goods market equilibrium: (4)

Real interest rate: r= i- .n:~l' ( 5)
Price formation : .n: = n ' + y(y - y) + s. (6)
Monetary policy rule: i = f + .n:~ 1 + h(n - .n:*) +b(y - [J) . (7)
Expectations: (8)
Inserting (5) into (7), substituting the resulting expression into (4), and assuming lor the
time being that public spending stays on trend (g = g) , we get the AD curve:
(9)
where we recall that the variable v captures shocks to private demand (shifts in expected
future income growth). Note from Eq. (33) in Chapter 17 that when g = g, our previous
demand shock variable z is related to v by z = v/ (1 + a 2 b).
The short-run aggregate supply (SRAS) curve is found by inserting (8) into ( 6 ):
n = JL 1 + y(y - y) + s. ( 10)
In the lollm>Ving, we will use the model (9) and (101 to investigate the optimal choice
of the monetary policy parameters h and b, given th at the central bank has committed itself
to the Taylor rule (7) lor monetary policy. In the next chapter we shall argue that it is
actually (close to) optimal for the central bank to follow some form of Taylor rule.
We should emphasize that this framework for analysing stabilization policy rests on
two strong simplifications. The first one is that expectations are entirely backward-looki11g, in
the sense that expected inflation depends solely on the actual inflation rate observed in the
Why and how?
© The McGraw-Hill
Companies. 2005
past (n ' = JL1 ) . As we shall see in the next chapter. stabilization policy becomes more
challenging if private agents rationally anticipate the effects of policy changes on current
inflation. Our second simplification is the assumption embodied in (7) that the central
bank can immediately observe and react to changes in current output and inflation
without any delay. In Chapter 22 we will discuss the dil\lculties arising when policy
makers can only react with a lag to changes in the state of the economy. Nevertheless.
despite its simplicity the AS-AD model above serves as a useful starting point for a
discussion of the problems of stabilization policy.
We assume that policy makers wish to minimize the social loss function (1) which
includes the variance of output arow1d trend, a~, and the variance of the inflation rate
around the target rate. a.!. We will also assume that policy makers take the long view, so
a; o;
we will interpret and as the long-run variances of output and inflation, that is, the
asymptotic variances emerging when we consider a very long (in principle, an infinitely
long) sequence of short-nm macroeconomic equilibria. In the appendix to this chapter we
show that if the shock variables sand z = u/(1 +a 2 IJ) are 'white noise' variables, 8 then the
asymptotic variances of output and inflation will be given by:
(11)
(12)
a2 h
a=---, (13)
1 +a 2 b
where the magnitudes a;

and a;
are the variances of the shock variables and res- z s.
pectively, a ~ is the variance of the private demand shock variable v, and a is the inverse of
the slope of the AD-curve, as you can see from (9). Not surprisingly, we see that the greater
the variances of the shocks to demand and supply, the greater are the long-run variances
of output and inflation. ceteris paribus.
We will now study how the central bank's choice of the monetary policy parameters
h and fJ aiTect o~ and e1,!. How should h and b be chosen if monetary policy makers wish to
minimize fluctuations in output and inflation? Is it possible to minimize both types of
fluctuations at the same time. or does society face a trade-off between output variability
and inflation variability ? In analysing these questions. it is useful to start by considering
the special case where the shocks to the economy originate solely from the demand side.
The monetary policy reaction to demand shocks
When there are no supply shocks, we have o~ = 0. Using (13). we may then '-vrite (11)
and (12) as:
(14)
8. Recall from C hapter 19 that a 'white noise' variable is a stochastic variable which is identic ally and independently
distributed over time, w ith zero mean and a constant variance.
20 STABILIZA TION POLICY 613
, ya;
(15)
a; =)'0:2 + 2a
From these expressions we see that the variances of output and inflation are smaller, the
higher the values of the monetary policy parameters hand 11. In other words, the volatility
of output and inflation will be smaller the stronger monetary policy reacts to the inflation
gap and to the output gap.
Figure 20.4 should help you to understand the intuition for this result. The figure
illustrates the case where the economy is hit by a negative demand shock. In the initial
period 0 the economy is in long-run equilibrium at pointE where output and inflation are
both at their target levels. However. in period 1 a negative demand shock occurs due to
more pessimistic private sector expectations. According to (9) such a drop in v will shift
the AD curve downwards. causing output as well as inflation to drop below their targets.
Now consider how the economy's reaction to the negative demand shock depends on
the way monetary policy reacts to a change in the output gap, captured by the parameter
b in the Taylor rule (7). According to (9) the numerical slope of the AD curve is equal to
(1 +a 2 b)/a 2 h. Since a 2 and hare positive, this expression shows that the AD curve will be
steeper the greater the value of IJ. As illustrated in Fig. 20.4, a steeper AD curve implies
that the fall in output as well as inflation will be smaller for any given downward shut of
the AD curve. In other words, by pursuing a coulltercyclical monetary policy (b > 0) involv-
ing a cut in the interest rate when output falls below trend (and vice versa), the central
bank can cushion the economy against demand shocks and reduce the variability of
output and inflation. Of course the reason is that a lower interest rate tends to stimulate
private spending. thereby ollsetting the negative shock to aggregate demand.
SRAS
.1To = n *
.nf
AD 1 (b > 0)
Y1 Yi Y y
Figure 20.4: The short-run effects of a negative demand shock: the importance of cou ntercyclical
monetary policy
Why and how?
© The McGraw-Hill
Companies. 2005
It is easy to see from Fig. 20.4 that the steeper the AD curve. that is, the higher the
value of b, the smaller will be the change in output and inflation when the economy is hit
by a demand shock. From this observation it might be tempting to conclude that IJ should
be set at a very high level (in principle, that it should be in finitely high) to minimize the
impact of demand shocks. However, such a prescription ignores the constraint that the
nominal interest rate crumot fall below 0. If b were very high, a shock which reduces the
rate of inflation might require that the nominal interest rate be driven down below 0. But
since this is impossible. there is a limit to how high b can be. If the shock variables v and s
have a h igh variance, causing substantial fluctuations in output and inflation. there may
be a fairly narrow limit to how high b can be. given that the non-negativity constraint
for the nominal interest rate must be respected in each period. Thus the implication of
the graphical analysis in Fig. 20.4 is that b should be as high as allowed by the non-
negativity constraint. i, ~ 0 .
Consider next how interest rate policy should react to inflation when the economy is
hit by a demand shock. The central bank's response to a change in the rate of inflation is
captured by the policy parameter h in the Taylor rule (7). From Eq. (9) we see that a higher
value of h will have two effects on the aggregate demand curve. On the one hand, it will
reduce the numerical slope (1 + r1 2 b)/a 2 h of the AD curve. As we saw in Fig. 20.4. a flatter
AD curve will tend to ampl~fy the change in inflation and output lor any given vertical shift
of the curve. On the other h and, it follows from (9) that whenever v falls by some amount
~ v. a h igher value of h will reduce the distance (~ v/a 2 h) by which the AD curve shifts
down. Clearly this effect of a higher h will tend to dampen the fall in output and inflation
induced by a negative demand shock. Equations (14) and (15) reveal that the latter eflect
is the dominant one, since we see that a higher value of h will reduce the magnitudes of cJ;
and c1.;. In other words, the greater the interest rate cut in response to a drop in inflation,
the smaller is the fall in output and in flation generated by a negative demand shock.
A similar reasoning implies that higher values of h and b will help to dampen the
increase of y and n above their target values whenever the demand shock is positive
(~ v > 0). Thus we may conclude that, when the economy is hit by demand shocks, a
strong central bank reaction to changes in inflation and output will help to stabilize the
economy. Indeed. when demand shocks are the only disturbances to the economy. our
simple model implies that the magnitudes of hand b should be as high as possible, subject
only to the constraint that the nominal interest rate cannot become negative in response
to a negative demand sh ock.
A basic insight fi-om this analysis is that there is no trade-ojf between output instability
and inj)ation in~tabili ty when business .fluctuations are driven by dernand shocks: a strongly
countercyclical monetary policy (a large value of h) and a strong interest rate response to
changes in inflation (a high value of h) will simultaneously reduce the variance of output
as well as inflation. In this way an active monetary policy will unambiguously serve to
reduce the social welfare loss from shocks to aggregate demand.
Unfortunately things are not that simple when the economy is hit by supply shocks,
as we shall see below.
The monetary policy reaction to supply shocks
To highlight the dilemma posed by supply shocks. let us locus on the case where there are
20 S TA B ILIZ ATIO N PO LIC Y 615
no demand shocks (a; = 0). From (11 )-(13) we then get:

}
} (}';
(16)
a;; = y 2 + 2y(1/a) '
(17)
Recall that 1/a is the numerical slope of the AD curve. Thus we see from (16) that a steep
AD curve will help to keep the variance of output small when business cycles are driven by
supply shocks. According to (13) a steep AD curve (a small value of a) can be achieved by
choosing a high value of band a low value of h. However, from (17) we see that a policy
which keeps a small will result in a high variance of in flation. If monetary policy makers
are primarily concerned about stabilizing the rate of inllation. it follows from (13) and
(17) that the central bank should choose a low value of b (indeed a negative b) and a high
value of h in order to make the AD curve as flat as possible.
Figure 20.5 illustrates the dilemma. When the economy's initial long-run equilib-
rium E is disturbed by an unfavourable supply sh ock. the short-run aggregate supply
cmve will shift upwards. U' policy makers have chosen a low (or even a negative) value of
band a high value of h. the economy will end up in the new short-run equilibrium E 1
where most of the shock is absorbed by a fall in output whereas the rise in inllation is
modest. By contrast. if the central bank focuses only on stabilizing the level of output. it
will choose a high value of band a low value of h to make the AD curve as steep as possible.
The economy then moves from E to E; where the shock is almost fully absorbed by a rise
in inflation.
LRAS
AD (b high and h low)
Yi Y y
Figure 20.5: The short-run effects of a negative supply shock: the importance of the monetary
policy response
Why and how?
© The McGraw-Hill
Companies. 2005
616 PART 6: THE SH O RT-RUN MO DEL FO R THE C LO SED ECON OM Y
The trade-off between output volatility and inflation volatility
In our study or demand shocks we found that the central bank will never face a trade-off
between output stability and inllation stability: a monetary policy which reduces output
stability will also imply a more stable inllation rate. However. when the economy is hit by
supply shocks. it follows from (16) and (17) that monetary policy makers face an
inescapable trade-oil' between output stability and in flation stability. If they choose the
policy parameters h and b so as to reduce the fluctuations in output, they will have to
accept a greater volatility of inflation, and vice versa. We shall now show that rational
policy makers will also face such a trade-offin the more general case where the economy
is hit by demand sh ocks as well as supply shocks.
As our graphical analysis has indicated, the central bank can inlluence the variance
of output (a; ) and the variance of inflation (a; ) via its choice of the policy parameters h
and b. A rational central bank acting in the interest of society will therefore choose h and
b so as to minimize the social loss function (1). Assuming that the non-negativity con-
straint ir;;. 0 is not binding. the llrst-order conditions for minimization of (1) with respect
to h and b are:
oSL = 0 oa~ oa.; = 0, (18)
a11
- ==* - + K-
oh oh
oSL = 0 oa~ oc1; (19)
a& a& a&
- ==* - · + K- = 0
Since the parameter K is positive, Eqs (18) and (19) imply that any ch ange in h orb which
reduces the variance of inllation will increase the variance of output, and vice versa.
Therefore, if the policy parameters hand bare chosen appropriately so as to minimize the
social loss from economic instability, policy mal<ers will inevitably face a trade-oil' between
output instability and inllation instability. The specific choice between the two types of
instability will depend on the magnitude of the parameter K which measures the policy
makers' aversion to fl uctuations in inllation relative to output volatility.
Summing up: optimal monetary policy under the Taylor rule
Table 20.2 summarizes our findings regarding the proper monetary stabilization policy
when the central bank sets the interest rate in accordance with a Taylor rule of the form
shown by (7). The table distinguishes between the case where demand shocks are the
dominant source of lluctuations. and the situation where business cycles are mainly
driven by supply shocks.
One important conclusion from Table 20.2 is that the real interest rate should gener-
ally be raised when the rate of inflation rises, that is, h should generally be positive. The
only modificat!on to this famous ;Taylor principle' is the case where supply shocks are all-
dominant and policy makers care only about minimizing the variance of output. In that
special scenario the optimal value of h approaches 0, since policy mal<ers will then want
the AD curve to be almost vertical, as we have seen above.
A second conclusion from Table 20.2 is that interest rate policy sh ould be counter-
cyclical (b > 0) except when supply shocks are dominant and policy makers care mainly
about stabilizing inllation.
Why and how?
© The McGraw-Hill
Companies. 2005
ICID
20 STABILIZ ATION POLICY 617
Table 20.2: Optimal monetary policy under the Taylor rule

Dominant shocks
Policy preference
for Demand shocks Supply shocks
Output stability h> O h -+ 0

b >O b> O
Inflation stability h> O
b< O
While concern about output stability calls lor a low value of h when shocks originate
from the supply side, the value of h should nevertheless stay (slightly) positive. In other
words, the nominal interest rate should always rise by more than one percentage point
when the in flation rate goes up by one percentage point, and vice versa, no matter what
type ofshock drives the change in inflation. To see why, recall from Eq. (19) in Chapter 19
th at our AS-AD model with static inflation expectations can be reduced to the difference
equation:
1 a21!
fJ=-. a=---. (20)
1 + ccy 1 + (l 2b
If h turns negative. we see from (20) that,B becomes greater than 1 so that output will tend
to move further and further away from its trend value once the long-run equilibrium has
been disturbed by a shock. In mathematical terms, the dill'erence Eq. (20) will be unstable
when (J > 1. To ensure long-run stability (I,8 < 1 1), we need I! > 0.
Empirical illustration: the US monetary policy experience
The danger of choosing a negative I! is illustrated by recent American business cycle

history. Figure 20.6 shows the estimated reaction of the nominal short-term interest rate
to the rate of inflation in the US lor the two periods 1960- 1979 and 1987- 1997. The
slopes of the two solid straight lines reflect the estimates of the coefllcient 1 + h in the
Taylor rule i 1 = f + n 1 + h(n1 - n *) + b(y1 - y) in the two periods. We see that 1 + I!< 1 in
19 60- 19 79 . implying a negative value of h in this period.
From the dillerence equation (20) we would then expect the variance of output
around trend to have been greater during 1960- 79 than during 198 7-97 where h L~
seen to be positive. In Fig. 20.7 the dotted horizontal lines indicate the average output
fluctuations implied by the estimated standard deviations of the cyclical components of
real US GDP in the two periods. We see that output instability was indeed greater during
19 60- 79 when higher inflation was allowed to reduce the real rate of interest.
The failure of the US Federal Reserve Bank to raise the nominal interest rate sutll-
ciently when faced with rising inflation in the late 1960s and the 19 70s is nowadays seen
as a serious policy mistake by many observers. In Chapter 2 2 we shall see that this failure
of US monetary policy makers to raise the real interest rate in response to inflation was
apparently due to the fact that they overestimated the economy's potential output during
the 1960s and 1970s. Thus they may not have been exceptionally 'soft' on in 11ation;
Why and how?
© The McGraw-Hill
Companies. 2005
618 PART 6: TH E SHORT- RUN MODEL FOR TH E CLOSED ECONOMY
Inflation rate
Figure 20.6: Estimated monetary policy rules in the United States: 1960-79 versus 1987-97
Source: Figure 7.3, p. 33 1 in John B. Taylor, 'A Historical Analysis of Monetary Policy Rules', Chapt er 7 in John B.
Taylor (ed.), Monetary Policy Rules, NBER Business Cycles Series, Volume 31, The University of Chicago Press ©
1999 by the National Bureau of Economic Research.
0.05
0.04
rt A
0.03
h 11 · ~
c 0.02 ·-j
J
"'c0
0.
0.01 M 14~r, I -K- I~ r~\ -A---~ ,.Ju
E
0
0.00 "
/\NV ~ \ I V"
Af"-Jv'
v
0
- 0.01 - IJ____i ______
~ --- -L''----
~
- 0.02
0
0' v vI J
- 0.03
- 0.04 v \
- 0.05
v
- 0.06
co
Ll)
0
<!l
co 0co
,... co
co 0
(j)
"'
(j)
'<t
(j)
<!l
(j)
co
(j)
0\ (j) (j) (j) (j) (j) (j) (j) (j) (j)
Year
Figure 20.7: The volatility of real GOP in the United States, 1958-98
Source: Bureau of Economic Analysis. The time series for real GOP has been detrended by HP filtering.
instead they seem to have been victims of serious macroeconomic measurement errors.
Chapter 22 will analyse the implications of measurement errors for optimal stabilization
policy.
~.i.S..~.':l~ . S.~.':l~.i~~~.':l~.i?.J?:..I?.?.l.i~y ...................................................................................................... .

Since we wanted to focus on monetary policy, the previous analysis assumed that public
expenditure was kept at its trend level (g 1 - g). reflecting a passive llscal policy. Let us now
consider whether active llscal policy can help to stabilize the economy. As a starting point,
recall that the equilibrium condition for the goods market may be written in the form:
(21)
Suppose that llscal policy can react to the same information on the inflation gap and the
output gap as monetary policy and that public expenditure is adjusted in accordance with
the followingjlsca/ policy rule:
(22)
We may then ask what the sign and the magnitude of the fiscal policy parameters c, and
cy should be if llscal policy makers wish to assist monetary policy makers in minimizing
the social loss function {1)? For example, should llscal policy be coul!tercyclica/, reacting
with an increase in public spending when output falls below trend (cy > 0)? To investigate
th is, we insert (22) along with (7) and (8) into (21) to get the AD curve with active jiscal
policy:
(2 3)
Equation 123) has exactly the same qualitative form as the AD curve (9) in our AS-AD
model with passive fiscal policy. The only modillcation is that the slope of the AD curve is
now affected by the llscal policy parameters c" and cY' and that the impact of the demand
shift variable v now depends on fiscal policy (via c, ). However, these parameters enter in a
way which is quite symmetric with th e impact of th e correspon ding monetary policy para-
meters h and b. For example. if llscal policy makers decide to cut public spending more
sharply in response to rising in flation, the resulting increase in c" will ailect the slope of
the AD curve and thevarianceofthedemand shifttenn v/(a 1 en + a 2 h) in the same way as
when monetary policy makers decide to raise the interest rate more aggressively in reac-
tion to higher inflation. Furthermore, we see from (2 3) that a more active countercyclical
fiscal policy will have the same qualitative impact as a more activist countercyclical mon-
etary policy, since cy and b affect the slope of the AD curve in the same manner.
We may draw two important conclusions from these observations. First. the factors
determining the optimal sign and magnitude of the fiscal policy parameters c, and cy are
the same as those determining the optimal monetary policy. Hence our previous analysis
of optimal monetary stabilization policy also fully ch aracterizes the optimal llscal stabi-
lization policy. In particular. the results summarized in Table 20.1 carry over to llscal
policy if his replaced by c" and cy is substituted for b. Second, if there are 110 constrai11ts on
Why and how?
© The McGraw-Hill
Companies. 2005
acroeconomics Economy
620 PART 6 : TH E SHORT-R UN MODEL FOR TH E CLOSED ECONOMY
the policy parameters h. b, en and cY' 11scal stabilization policy can achieve nothing in
terms of reducing the variability of output and inllation which could not have been
achieved through monetary policy, and vice versa.
The explanation for these results is that 11scal as well as monetary policy work
through the economy's demand side, by allecting the slope of the AD curve and the vari-
ance of the shifts in this curve. 9 From this one might be tempted to conclude that either
11scal stabilization policy or monetary stabilization policy is redundant: if one of these
instruments is available, we do not need the other one. Yet such a conclusion would be
unwarranted, since the policy parameters h, b. c!f and cu are in fact constrained. As you
remember. the choice of the monetary policy parameters hand b is constrained by the fact
that the nominal interest rate cannot fall below 0 . Hence there is a limit to the feasible cut
in the interest rate when a large negative demand shock drives output and inllation
far below their target levels. In such a gloomy scenario where a large negative output
gap persists despite a short-term nominal interest rate close to 0, as many countries
experienced during the Great Depression of the 19 30s, and as the world's second largest
economy (Japan) has experienced in recent years, monetary policy may become impotent.
and active 11scal policy may be the only means by which the economy can be dragged
out of recession. Indeed, the view that 11scal policy is an important potential tool of
stabilization policy was developed as a result of the experience of the Great Depression.
On the other hand, there are serious constraints on the ability of 11scal policy makers
to implement immediate and large changes in taxation and public spending in reaction to
changes in the business cycle. The level and structure of taxation and public expenditure
have important etlects on resource allocation and income distribution, so large and
abrupt changes in these variables may have considerable negative ellects on the welfare of
(some) citizens, even if they help to stabilize the macro economy. Moreover. in practice it
may take time before a change in taxes or public spending can be implemented. as we
shall discuss in Chapter 22 . For these reasons there are limits on the values of the 11scal
stabilization policy parameters c, and cy that are feasible in a short-term business cycle
perspective. Hence there is a clear potential for monetary policy to help to stabilize the
economy.
.~.~r.I.l.r.I.l.~~Y.......................................................................................... """""""""""""""""""""""""""".
1. Stabilization policy is the active use of monetary and fiscal policy to influence the aggregate
demand for goods and services. The goal of stabilization policy is to minimize the social
welfare loss from the volatility of output and inflation.
2. The instability of output and the associated fluctuations in real income impose a welfare loss
on risk averse consumers who prefer a smooth to an uneven time path of consumption.
Problems of moral hazard and adverse selection prevent consumers from insuring themselves
fully against the unexpected temporary income losses caused by business cycles, and credit
9. In a more complete and realistic model one should allow for the fact that changes in fiscal policy instruments (e.g .,
tax rat es) may also affect the economy's supply side by affecting the econonic incentives of households and firms.
For s implicity we leave out such incentive effects which are t he subject of ptblic finance theory.
20 STABILIZATION POLICY 621
constraints often prevent unemployed consumers from borrowing against their expected
future labour income. Hence the unanticipated income losses generated by recessions force
some consumers into cutting their consumption, resulting in welfare losses which are not
fully offset by the welfare gains from higher consumption during economic booms, due to
diminishing marginal utility of consumption.
3. The employment fluctuations resu lting from business cycles also create welfare losses for
(some) workers. Market imperfections drive a welfare-reducing wedge between the marginal
product of labour and the marginal rate of substitution between income and leisure. This
wedge goes down when employment rises during a boom, but typically the wedge increases
a lot more when employment falls during the subsequent recession. For this reason the
average level of welfare could be increased if employment could be stabilized around its
average (natural) rate.
4. Empirical evidence indicates that the potential gains from stabilization of employment are
considerable. Empirical studies have also documented that the losses of income and con-
sumption caused by recessions are borne d isproportionately by low-paid unskilled workers.
Hence business cycles tend to exacerbate the unequal distribution of income. This is another
reason why society may w ish to stabilize output and employment.
5. Stabilization policy should aim at stabilizing output around its tren d (natural) level. Due to
market imperfections, natural output is lower than the Pareto efficient output level, but if policy
makers were to target the efficient output level - seeking to keep unemployment permanently
below its ratural rate - the result wou ld be ever-accelerating inflation. Policy makers may use
structural policies to move natural output closer to the efficient output level, while stabilization
policy should be used to keep actual output close to natural output.
6. Apart from stabilizing output and employment, another goal for stabilization policy is to mini-
mize the volatility of the rate of inflation around its target rate. Inflation fluctuations generate
welfare losses because they are typically unanticipated, causing relative prices (including real
wages and real interest rates) to deviate from the expected levels on which household and
business plans are based.
7. Even anticipated inflation creates welfare costs arising from an inefficiently low demand
for cash balances, ' menu costs' due to frequent price changes, relative price d istortions
stemming from non-synchronized nominal price adjustments, and a d istorted measurement
of taxable income due to an unindexed tax system. Yet it is w idely agreed that the target
inflation rate should be kept above 0 to enable the central bank to cut the nominal interest rate
signifi cant~ in a serious recession without hitting the zero bound for nominal interest, and to
facilitate the downward adjustment of real wages to a negative shock in labour markets with
downward nominal wage rigidity.
8. A fiscal or monetary policy rule prescribes how the instruments of stabilization policy should
be set, given the observed state of the economy. Following a fixed policy ru le may help policy
makers to establish credibility and to increase the predictability of economic policy so as to
minimize welfare-reducing expectational errors in the private sector. The alternative to fixed
policy ru les is discretionary policy. Under d iscretion, policy makers may conduct monetary
and fiscal policy in any way they believe w ill help advance the goals of stabilization policy,
taking account of any special circumstances which might prevail. Discretion allows more
Why and how?
© The McGraw-Hill
Companies. 2005
622 PART 6 : THE SHORT-RUN MODEL FOR THE CLOSED ECONOMY
flexibility than ru les, but typically at the cost of reduced credibility and predictability. The evi·
dence suggests that the monetary policy of the most important central banks is fairly well
described by some form of Taylor rule, although policy makers never mechanically follow a
fixed rule.
9. When business fluctuations are caused solely by demand shocks, the central bank interest
rate should decrease sharply in response to a fall in inflation and output to minimize the social
welfare loss from the volatility of output and inflation. The interest rate reactions should be as
strong as possible, subject to the constraint that the nominal interest rate cannot fall below 0.
There is no dilemma between reducing the variance of output and that of inflation when
business cycles are driven by demand shocks.
1 0. If business cycles are generated exclusively by supply shocks, and monetary policy makers
are mainly concerned about minimizing the fluctuations in output, they should cut the real
interest rate aggressively in response to a fall in output and increase it very little in response
to a rise in inflation. By contrast, if a central bank faced with supply shocks focuses mainly on
keeping the inflation rate low and stable, it should raise the real interest rate sharply in reac·
tion to a fall in output and increase it aggressively in response to a rise in inflation. Whatever
the main policy objective, a lower variance ot output can only be achieved at the cost ot a
higher variance of inflation, and vice versa, when the business cycle is driven by supply
shocks.
11. In the general case where the economy is exposed to demand shocks as well as supply
shocks, there is always a trade-off between reducing the variability of output and reducing the
variability of inflation when stabilization policy is optimally designed.
12. A general feature of the optimal monetary stabilization policy based on the Taylor rule is the
so-called Taylor principle, according to which the real interest rate should always be raised
in response to a rise in the rate of inflation, regardless of whether shocks to the economy
originate from the demand side or from the supply side.
13. Fiscal stabilization policy may take the form of changes in public spending in response to
changes in the output gap and in the inflation gap. Whenever it is optimal to tighten (relax)
fiscal policy, it is also optimal to tighten (relax) monetary policy, so as long as the zero interest
rate bound on the nominal interest rate is not binding, fiscal stabilization policy cannot achieve
anything which could not be achieved through monetary policy. However, in a deep recession
where the short-term nominal interest rate has been driven down close to 0, expansionary
fiscal policy may help monetary policy makers to pull the economy out of the recession.
Chapter Appendix: calculating the asymptotic varian ce of

20.5 output and inflation
................................................................................................................................................................................
This appendix explains the derivation of the asymptotic variances of output and inflation pre·
sented in Eqs (1 1) and (12) in Section 2. For simplicity, we will assume that our shock
variables z, and s , are 'white noise' processes, each with a zero mean and a constant variance.
We start by considering the variance of output around its trend level. Specifying the
output gap as j 1 "' y,- y, and using Eq. (20) in the main text, we get the following expressions
for the deviations of output from trend from period 1 through period t:
.91 = /390 + f3(z1 - z0 ) - af3s 1

.92 = (3y, +f3(z2-z,)-af3s2
.9,_1 = f3 ji,_2+f3(z1_ 1 -z1_ 2)-af3s,_1

.9, = f3ji,_1 + f3(z 1 -z1_ 1) - af3s,
By successive substitutions of the expressions for ji1_ 11 ji1_ 2 , etc., into the expression for y, we
get:
1- 1
1
• {J''y 0 + 'L.-
y,= \ ~-' z,_,- z,_, _1 -
Rll + l( as1_ , ) , /3"'- . (24)
n- o 1 +ay
The asymptotic variance of .9, is the variance of the expression on the right-hand side of (24)
as t tends to infinity. Hence this measure of the variance of output accounts for all of the
shocks to which the economy has been exposed in the past.
We immediately see that .9, will tend to infinity for t~ oo if f3 > 1, as w ill be the case if c1. < 0
(as long as 1 + ay> 0). Hence it cannot be optimal to choose policy parameters such that
c1.< 0, as we explained in the text. In the following we w ill therefore assume that a> 0 so that
0 </3 < 1. For t -> oo the term f3'.9o in (24) will then ten d to zero so that (24) may be written:
(25a)
j = 0, 1 '2, 3, ... (25b)
We will now calculate the variance of the infinite sum on the right-hand side of (25a). For this
purpose, recall that if X is a stochastic variable and a is a constant, we have:
Var(aX) = a 2 Var(X), (26)
where Var(X) denotes the variance of X. One implication of (26) is that:
Var(-X) = Var(X). (27)
Furthermore, if the stochastic variables XII x2.x3.... ,X, are distributed independently of each
other (having zero covariances), we know that:
(28)
Note that all the stochastic terms on the right-hand side of (25a) are in fact uncorrelated,
since z and s are uncorrelated white noise processes. Hence the variance of the sum on the
right-hand side of (25a) is simply the sum of the variances of the individual terms in the sum.
Moreover, because z and s are white noise, all the z-terms have the same variance a;, and all
the s-terms have identical variance a~. From (25b) - (27) it then follows that:
(29)
Using (26), (28) and (29) and denoting the asymptotic variance of y by a;, we may write the
variance of the sum on the right-hand side of (25a) as:
(30)
Why and how?
© The McGraw-Hill
Companies. 2005
624 PART 6 : THE SHOR T-RU N MODEL FOR THE CLOSED ECONOMY
Since 0 <fJ < 1. the infinite sum in the last bracket on the ri~ht-hand side of (30) is:
p2
p 2 + jJ4 +{is + ps + ... = 1 - (3 2 '
so from (30) we get:
2
(3 2(2a; +a 2 a~)
(31)
ay= 1- (3 2
Using the definition jJ"' 1 / (1 + ay), the expression in (3 1) may be rewritten as Eq. (1 1) in the
main text.
Consider next the asymptotic variance of inflation around its target rate. In eq. (19) of
Chapter 19 we found that the solution for the inflation gap fr r"' n 1 - n* in our AS· AD model
w ith static expectations is:
itt= f3nr_, + yfizr+ fisr. (32)
Using backward substitution, (32) implies that
t- 1
•
1lr = fJRl1r• o + 'L,.
\ ' an+ 1(
fJ
)
YZt- n + Sr- n • (33)
n- o
The asymptotic variance of the inflation gap (a;) is the variance of the expression in (33) as t
tends to infinity. Note how (33) resemb les (2 4). Following the same procedure as the one
used to derive (30), and exploiting the wh ite noise properties of z and s, we find from (33) that
a~= ( y2a; +a~)(fi2+ (34+(3 6 +(3 s + .. ·)

2 (3 2( y2 a;+ a~)
a ·' = 1 _ p2 I
(3 4)
which in turn may be transformed into Eq. (1 2) in the main text, using the definition of (3.
20.6 Exercises
Exercise 1. Optimal monetary policy in a stochastic world
Consider an economy w ith stochastic demand and supply shocks where the social loss from
macroeconomic instability is given by the following loss function which assumes that the
target inflation rate is zero:
SL = a;+lw~, K > 0, (35)
The economy's supply side is captured by the simple Phillips curve:
n = y(y-J)+s, E[s] = 0, (36)
where s is 'wh ite noise' reflecting stochastic supply shocks wh ich are identically and indepen·
dently distributed over time. Ignoring fiscal shocks, the goods market equilibrium condition is
E[v ] = 0, (37)
where the wh ite noise variable v captures stochastic shocks to private demand.
We assume that although the central bank knows the probability d istribution of the
shocks, it cannot immediately observe the current realizations of the shocks s and v. Instead
it can only observe the current output gap and the current inflation rate, so monetary policy
follows the Taylor rule w ith a zero inflation target:
i = r + n + h:n + b(y- y). (38)
The problem for monetary policy makers is to choose the values of hand b which w ill minimize
the social loss function (1 ).
1. Find the solutions for the output gap and the inflation rate in the model (41 ) - (44) and show
that the variances of output and inflation are g iven by
(39)
(40)
2. Suppose now that all shocks originate on the economy's demand side so that a~ = 0. Explain
the optimal monetary policy in this special case. Is there a trade-off between stabilizing output
and stabilizing inflation? Is the optimal policy procyclical or countercyclical? What are the
constraints on the values of h and b? Is it really necessary for the central bank to react to the
output gap as well as the inflation rate, or is it sufficient to react to just one of these variables?
Explain.
3. Suppose alternatively that all shocks emanate from the supply side. Set a~ = 0 in (45) and
(46) and show that in this case:
~ _ 2a~h( 1 + a 2 b)a~
(41)
(Jh - fj, 3
oa~ 2yc.l ~h( 1 + a 2b)a~

at; = fj,3
(42)
Is there a trade-off between stabilizing output and stabilizing inflation in th is scenario? Usc a
graphical analysis to illustrate your point.
4. Use the results (4 7) and (48) to show that the ratios ¥,/ %~ and ¥/;!~i are exactly the same.
Explain on this basis why it is unnecessary to react to the output gap (i.e., why the central
bank can choose b = 0) provided h is chosen to achieve the optimal trade-off between output
volatility and inflation volatility?
5. Using your insight from Question 4 that one can set b = 0 when a~ = 0, show that in this case
the optimal value of h is:
(43)
Explain intuitively why the parameters y, 1< and a 2 enter the expression for the optimal value o f
h in the way they do. Try to explain why the variance of the supply shock does not enter the
solution for the optimal h? (Hint: does a~ affect the ratio between oa;/oh and (Ja~/oh?)
Why and how?
© The McGraw-Hill
Companies. 2005
Exercise 2. Optimal stabilization policy in a world of productivity shocks
In the main text we assumed that policy makers want to stabilize actual output around trend
output, )i. This seems a natural goal if shocks to aggregate supply mainly take the form of flue·
tuations in wage and price mark-ups wh ich do not change the economy's productive capac·
ity. But suppose the dominant source of output fluctuations is productivity shocks. One can
then argue that policy makers should allow output to expand when productivity is unusually
high, to take advantage of the fact that more output can be obtained w ith the same input.
Similarly, it seems reasonable to accept a fall in output if productivity is unusually low so that
it takes more than the normal effort to produce a given amount of output.
Against this background, let us introduce a distinction between trend output, )i, defined
in the usual way as the {log of) the level of output forthcoming when unemployment is at its
natural rate and productivity is at its normal (trend) level, and current potential output, y, which
now represents the output which can be produced at the natural unemployment rate when
productivity is at its actual current level. Thus the two concepts of output are linked by:
(44)
where § measures the percentage deviation of labour productivity from trend, reflecting
exogenous technological shifts in society's production possibility frontier. Assuming that
these technological shocks are the only source of supply shocks, it follows that we can write
the SRAS curve w ith static expectations as:
:rr = :rr_, + :(y-9) =:rr_, + y(y- .Y)- ys. (45)
You are now asked to analyse the optimal monetary stabilization policy in the case where
monetary policy makers are targeting potential output. In other words, we now assume that
policy makers wish to minimize the social loss function:
(46)
We w ill maintain the assumption that interest rate policy follows the Taylor rule (7) in the
main text, so the aggregate demand curve is still g iven by Eq. (9). Thus, even though the
central bank seeks to minimize deviations from potential output, the interest rate still responds
to the deviation of output from its trend level (y- )i). The reason might be that the central bank
c.;annot immediately observe the supply shocks oc.;c.;urring in the c.;urrent period, and henc.;e it
cannot observe the current potential output level (whereas it can estimate trend output )ion
the basis of historical data).
1. Suppose that the economy is mainly exposed to the technological supply shocks mentioned
above. Use a graphical analys is as a basis for d iscussing how the central bank interest rate
should respond to changes in the inflation rate, assuming that the social loss function is g iven
by (52). Should the value of the policy parameter h in the Taylor ru le be positive or negative?
Should the value of h be large or small? Try to provide an intuitive explanation for your finding.
(Hint: assume that the economy starts out in long -run equilibrium and then assume that a pos·
itive technology shock occurs, raising §from 0 to some positive value. Illustrate the short-run
effects on output and inflation and explain how these effects depend on the value of h.)
2. Suppose again that technology shocks (shifts in s) are the dominant source of disturbances
and that policy makers are targeting the unobserved potential output y. Use a graphical
analysis as a basis for d iscussing how the interest rate should respond to changes in the
estimated output gap y- Ji. Should monetary policy be countercyclical or procyclical? Should
the value of b be large or small? G ive an intuitive explanation for your result.
3. Is there a trade-off between output variability and inflation variability when policy makers are
targeting potential output rather than trend output and technology shocks are much more
important than demand shocks?
Exercise 3. Principles and problems of stabilization policy
In this exercise you are invited to restate and d iscuss some of the main results and arguments
presented in this chapter.
The goals of stabilization policy
The case for macroeconomic stabilization policy rests on the assumption that it is desirable to
stabilize the rates of output and inflation around some target values. This raises a number of
questions.
1. Discuss why it is socially desirable to stabilize the rate of inflation around some constant
target value. What are the costs of a fluctuating rate of inflation?
2. What are the arguments for avoiding fluctuations in output and employment? What factors
determine the magnitude of the welfare costs of output and employment instability?
3. Explain why even a constant rate of inflation generates welfare costs. Do you consider these
costs to be large or small? Discuss the factors which policy makers should take into account
when they choose the target inflation rate. Is zero inflation an optimal inflation target?
4. Discuss the factors which should be considered when policy makers choose the target level
of output. Should they choose the 'efficient' level of output, the trend level of output (Ji), or the
potential output y defined in Exercise 2? Discuss the arguments fo r and against the d ifferent
output targets.
5. Should macroeconomic stabilization policy follow fixed rules, or should policy makers be left
with discretion to make whatever macroeconomic policy choice they consider appropriate
in any given situation? What are the arguments in favour of rules, and what is the case for
discretion?
Optimal stabilization policy
6. Try to restate the arguments underlying the results on optimal monetary stabilization policy
summarized in Table 20.1 (which assumes that target output is equal to trend output). In
doing so, you may want to use an AS-AD diagram to illustrate how demand and supply
shocks affect the inflation gap and the deviation of actual output from trend output. Explain the
constraints on the choice of the policy parameters h and b in the Taylor rule.
7. Explain why Table 20.1 also characterizes the optimal fiscal stabilization policy, given that
fiscal policy reacts to inflation and the output gap. Discuss whether it is useful for policy
makers to be able to use fiscal as well as monetary stabilization policy.
Cl l Sorensen-Whitta- Jacobsen:
21. Stabilizati on policy
with rational expectations
© The McGraw-Hill
Companies. 2005
Chapter
Stabilization policy
with rational
expectations
conomic activity today depends crucially on expect.ed economic conditions tomor-
E row. A drop in the economy's expected future growth rate will tend to reduce the
propensities to consume and invest by reducing the expected future earnings of
households and firms. Hence the aggregate demand curve will shift do'lllrn, causing an
immediate fall in current output. As another example, a change in the expected rate
of inflation will shift the aggregate supply curve by feeding into the nominal wages
negotiated by workers and firms . It may also move the aggregate demand curve through
its impact on the expected real rate of interest. The expected inflation rate is thus an
important determinant of current economic activity.
Conventional macroeconomic models often assume that the expected future values of
economic variables depend only on the past history of those variables. Indeed, in the pre-
vious chapters we typically postulated that the expected inflation rate for the current
period is simply equal to the actual inflation rate experienced in the previous period. ThL~
assumption of backward-looking e).:pectations may be plausible in 'quiet' times when the
macroeconomy is not subject to signiHcant shocks. When people have no particular
reason to believe that the tightness of labour and product markets next year will be much
dill'erent from what it is today, it seems reasonable for them to assume th at next year's
inflation rate will be more or less the same as this year's. However. if the economy is hit by
an obvious and visible shock such as a dramatic change in the price of imported oil. or if
there is a clear change in the economic policy regime, say, due to a change of government.
it does not seem rational for people to assume that next year's economic environment
will be the same as this year's. Instead of just mechanically extrapolating the past into
the future, rational households and firms will seek to utilize all the relevant information
available to them when they form expectations about the future state of the economy.
In the early 1970s, some macroeconomists took this idea of'forward-looking expecta-
tions to its logical limit by advancing the rational expectations hypothesis (REH) . 1 According
1. The rational exp~c tations hypothesis was originally introduced in a microeconomic setting by John Muth, 'Rational
Expectations and the Theory of Price Movements', Econometrica, 29, 1961, pp. 315-335. Later on, the REH was
introduced into 11acroeconomic theory by Robert E. Lucas, 'Expectations and the Neutrality of Money', Journal of
Economic Theory, 4, 1972, pp. 103- 124, and by Thomas J. Sargent, 'Rational Expectations, the Real Rate of
Interest, and the Natural Rate of Unemployment', Brookings Papers on Economic Activity, 2, 1973, pp. 429-472;
followed by many others.
628
Sorensen-Whitta-Jacobsen: I Part 6- The Short-run 21. Stabilization policy © The McGraw-Hill
Introducing Advanced Model for the Closed with rational expectations Companies. 2005
21 STABILIZATION POLI CY WI TH RATIONAL EXPECTATIONS 629
to the REH, people use all the available information to make the best possible forecasts of
the economic variables which are relevant to them. Moreover, tire available irrjimnatior1
includes ilifimnntion about the structure of tile economy. The idea is that, even though in
practice the layman may not know much about the way the economy works. the eco-
nomic forecasts produced by professional economists are available to the public through
the media. so in this way people have access to the most competent forecasts of. say. next
year's rate of inllation. Economists should therefore model the fonnation of expectations
as if people use the relevant economic model to predict inl1ation and other economic vari-
ables which are important lor their economic decisions. In other words. rational expec-
tations are model-consistent expectations: they are identical to the forecasts one would
make by using the available knowledge of the structure of the economy. as embodied in
the relevant economic models. Another way of putting it is to say that economic analysts
should not assume that they are smarter than the economic agents whose behaviour they
are trying to predict. Instead, they should assume that agents form their expectations in
accordance with the analysts' own description of the economy. If they did not. and if the
analysts' model is correct, then agents would be making systematic expectational errors,
and presumably this would induce them to change the rules of thumb by which they form
their expectations until there is no discernible systematic pattern in their forecast errors.
This idea of rational expectations essentially revolutionized macroeconomic theory.
The REH is obviously a very strong assumption, and as 'Ne shall see. it can be criticized on
theoretical as well as empirical grounds. But before addressing these criticisms. this
chapter will explain the case lor the REH in more detail and derive some of its striking
implications. Our main purpose is to illustrate the importance of the way expectations are
formed. In particular. we will show how the effects of macroeconomic stabilization policy
may differ signillcantly depending on whether expectations are rational or backward-
looking. The 11nal sections of the chapter will discuss the validity of the REH. drawing on
theoretical arguments as well as empirical evidence.
The case against backward-looking expectations
One way of justifying the assumption of rational expectations is to examine more carefully
the implications of a macro model with backward-looking expectations. As we will
illustrate in this section, in some circumstances the assumption of backward-looking
expectations implies that economic agents are implausibly na'ive.
Consider the model of aggregate demand and aggregate supply set up in Chapter 19,
and suppose for simplicity that public spending always stays on trend so that g, =g. In the
usual notation, our AS-AD model may then be restated as follows:
Real interest rate: (2)
Price formation: n,= n;+ y(y, - y) + s,, (3)

21. Stabilization policy
© The McGraw-Hill
Companies. 2005
630 PA RT 6: TH E SH ORT-RUN M ODEL FO R T HE C LOSED ECONOMY
Monetary policy rule: (4)

Expectations: (5)
Equation (5) assumes a particularly simple form of backward-looking expectations by

postulating that the expected inflation rate for the current period equals the actual
inflation rate observed during the previous period. We will now illustrate the implications
of this assumption of 'static' expectations by solving the model (1 )- ( 5).
For the moment. we will simplify by setting the exogenous aggregate demand and
supply shock variables equal to their zero mean values, v , = s, = 0 . On th is assumption we
find by substituting (2), (4) and (5) into (1) that the aggregate demand curve may be
written as:
a ,h
y,- [I = a(n*- :nt), (l = - --- . (6)
1 + (12/J
while substitution of (5) into (3) yields the aggregate supply curve:
n, = :n,_J + y(y,- [1). (7)
Inserting (6) into (7) and rearranging. we obtain the first-order linear dillerence equation:
1
(3= - , (8)
1 + ya
which has the solution:

:n, = n * + f3'(:n 0 - :n*), t = 0,1,2, ... (9)
where n 0 is the predetermined initial value of the inflation rate in period 0. Since we see
from (8) that 0 < f3 < 1, it follows from (9) that the inflation rate will converge monotoni-
cally towards its target rate n * as t tends to infinity .
From (9) we may calculate the i11jlation forecast error, defined as the dillerence
between the actual and the expected inflation rate. Given the assumption of static expec-
tations (n~ = n,_1) , the inflation forecast error during the phase of adjustment to long-run
equilibrium is:
:n, - n~ = n, - n, _1 = fJ '(n 0 - n *) - {3 1- 1(n 0 - n*) *'*

n, - n~ = (3 1 1
- (n 0 - :n*)(/3 - 1). (10)
Suppose now that at the end of period 0 the government appoints a 'tough' new central
bank governor who announces a signiilcant reduction in the target inflation rate from the
start of period 1. For concreteness, suppose the inilation target :n* is reduced from 3 per
cent to 0 per cent per year. Usin g empirically plausible parameter values likey = 0.18 and
(1 = 0.878 (in accordance with our estimates in Chapters 18 and 19, respectively), we
may then simulate the evolution of the inflation forecast error implied by Eq. (10),
assuming that the initial inflation rate :n0 was equal to the previous in flation target of 3
per cent. The result of the simulation is sh own in Fig. 21.1.
We see that throughout the period of adjustment to the new inflation target of 0, the
public systematically overestimates the actual rate o.fit~flatioll. The reason for these systematic
21 STABILIZATION POLI CY WI TH RATI ONAL EXPECTATIONS 631
- 0.05 ~
/
/
~ - 0.1
I
.53' - 0.15
...... ~
0 0
~ ~ - 0.2
I
0~
~ ~ - 0.25
I
oe
-.,
§S - 0.3
I
·~
~ - 0.35 I
- 0.4 I
- 0.45
1 3 57 91 1 13 15 17 19212325272931 3335373941 43454749
Year t
Figure 21 .1: The inflation forecast error during a disinflation with static expectations (simulation)
mistakes in forecasting inflation is that the public mechanically extrapolates last period's
observed inflation rate into the filture. Thus, even though the central bank is determined
to bring inflation down by setting a high interest rate as long as n , > n *, people never-
theless continue to believe period after period that next year's inflation rate '>Viii be the
same as this year's.
Clear~, this behaviour is not very intelligent, especially not if the new central bank
governor has publicly mmounced his determination to kill inflation. Presumably. informed
citizens '>Viii observe or at least gradually learn that the monetary policy regime has
changed, and this should aflect the way they form their expectations ofintlation.
This is a statement of the case against the assumption of backward-looking expecta-
tions: if important aspects of the economic environment (such as the economic policy
regime) change, rational agents are likely to realize that the future path of the economy
cannot be projected by simply observing how the economy behaved in the past. To put it
another way, rational economic agents will utilize all relevant information when they
form their expectations about the future. including information about changes in
economic policy and other new developments which are likely to in fluence the course of
the economy.
Defining rational expectations

The rational expectations hypothesis takes the above line of reasoning one step further by
suggesting that ecmwrnic agents do not make systernaticforecast errors of the kind illustrated
in Fig. 21.1. To be sure. since the economy is often hit by stochastic shocks (which
were ignored for simplicity in Fig. 21.1). agents usually do make mistakes when they
try to predict the future state of the economy. But according to the REH there will be
no systematic bias in these forecast errors. For example. sometimes the rate of in flation will
© The McGraw-Hill
Companies. 2005
be overestimated, and sometimes it will be underestimated, but on average people's

inflation forecasts will be correct. The justillcation for this assumption is that economic
actions based on erroneous expectations cause losses of profits and utility, and hence
agents have an incentive to minimize their forecast errors. Ifthe forecast errors revealed a
systematic pattern such as persistent overestimation or underestimation, rational agents
should be able to detect this pattern and would have an incentive to revise their methods
of expectations formation to weed out systematic biases in their guesses about the future.
In an uncertain environment. the economic variables about which agents form their
expectations may be seen as stochastic variables. In such a setting we may forma lize the
hypothesis of rational expectations by saying that the sufJjective expectation of some
economic variable X for timet, X~. is equal to the objective mathematical expectation of X.
conditional on all available information at the time the expectation is formed. Thus, if the
expectation lor period t is formed at the end of period t - 1. the expected value of X for
period tis:
-
objeclive
subjeclive conditional
expectalion expectation
1 r,_ 1]
x~ =
.---..
E[X, (11)
where E[·] is the mathematical expectations operator, and I, _1 is the information set avail-
able to the agent at the end of period t - 1. Hence E[X,I I, _ 1] is the mean value of the sto-
chastic economic variable X lor period t, calculated at the end of period t - 1 using all the
information available at that time.
The definition of rational expectations can also be stated in slightly more general
terms. If,('( X;) is an agent's perceived probability that the stochastic economic variable X
will assume the value Xi' andf(X;) is the true probability that one can calculate. using all
the available information, the hypothesis of rational expectations says thatf'(X;) =.f(X;)
for all possible values of X;. In other words, the REH postulates that the agents' subjective
probability distributions for the relevant economic variables are equal to the true objec-
tive probability distributions. The mean value of the subjective probability distribution
of X and the mathematical expectation of the (conditional) objective distribution are.
respectively:
and (12)
Forf''(X;) =.f(X;). (12) obviously implies (11).

The specification in (11) captures the idea that expectations are on average correct.
even though they are hardly ever precisely correct. given the unpredictable stochastic
shocks which are hitting the economy all the time. Formally, Eq. (11) states that economic
agents know enough about the stochastic process determining X, to be able to calculate
the correct conditional mean value of X,. In this sense rational expecations are model-
consistent: the rationally expected value of X is equal to the mean value of this variable
which one would calculate from the correct stochastic economic model describing the
determination of X. Thus agents form their expectations tzs !l they knew the 'correct'
model of the economy (or the model of the subsector of the economy in which they are
interested). even though in practice they may have arrived at their forecasting rules
throug h a more intuitive trial-and-error learning process.
21 STA BILIZATION POLI C Y WI TH RATI ON AL EXPECTATI O NS 633
Later on we will discuss the realism of this assumption, but first we \vill explore some
of the striking implications of the rational expectations hypothesis.
We may use our model of aggregate demand and aggregate supply to study how rational
expectations affect the scope for macroeconomic stabilization policy. For convenience. let
us introduce the notation
(13)
to denote the rational expectation of variable X for period t, based on the in formation
available up until the end of period t - 1.
The Policy Ineffectiveness Proposition
Now suppose that when the central bank sets the interest rate lor period t, it does not
know what the actual levels of inflation and output for that period will be. Instead,
because macroeconomic data are only available with a time lag, the central bank has to
base its interest rate decision on the expected inflation rates n~.t- 1 and n~+ J.t- 1 and on the
expected general activity level !1~. 1_ 1 • Since r1 = i 1 - n ~+ u -1' we may then respecify the
monetary policy rule (4) as:
r 1 = f + h(n~. 1 _ 1 - n * ) + b(y~.t- 1 - jj). (14)
lvfaintaining the assumption that g1 = g, we still have the goods market equilibrium
condition:
(15)
and using the notation in (13) the aggregate supply curve may be written as:
(16)
To complete the model, we must specify the stoch astic properties of the exogenous
demand and supply shock variables. For simplicity we assume that these variables are
'white noise', being identically and independently distributed over time, with zero means
and constant variances (1 ~ and a;.
respectively:
(17)
E[s,] = 0 . (18)
To solve a model like (14)- (18) where expectations are rational, we must derive the
expectations :n:~.t- L and !J~.t- L which are consistent with the model, given the information
© The McGraw-Hill
Companies. 2005
available at time t - 1 . We w ill solve the model by going through th e following three
steps: 2
Step 1 Solve the model for the endogenous variables y 1 and n 1 in terms of the exogenous
variables and the expectations variables Y~. 1_ 1 and n~.~- 1 •
Step 2 : Find the solutions for Y~.t- 1 and n~.t- 1 by calculating the expected value of the
expressions found in Step 1, given the information available at timet - 1.
Step 3: Insert the solutions for y~ 1_ 1 and n~.t- l into the expressions found in Step 1 to
obtain the flnal solutions for y 1 and n 1 in terms of the exogenous variables.
Let us illustrate the mechanics of this procedure.
Step 1. Inserting (14) into (15). we flnd:
y, = y + v, - a 2 [h(n~1 _ 1 - .n*) + b(y~1 _ 1 - y)]. (19)
If we substitute (19) into (16). we then get:
.n, = n~.t- 1 - ya 2 [h(n~.t- 1 - .n*) + b(!J~.t- 1 - y)] + yv, + s1• (20)
We have now expressed the actual values of y, and n , as functions of the exogenous vari-
ables and of the expected values of y 1 and n 1, as required in Step 1.
Step 2. The next step is to use (19) and (20) to calculate the rational expectations of y ,
and n 1• The rationally expected value of output in period t is the mean value of the ex-
pression on the right-hand side of (19). calculated on the basis of information available
at time t - 1. In calculating this conditional expectation, we may use the facts that
E[n~.t- 1I r,_ 1] = .n~.t- J and E[y~.t- 1I r,_1] = Y~.t- 1. This simply says that agents do. of course.
know at time t - 1 what their own expectations are at that time. We also recall from (17)
that v 1 has a zero mean value, so at timet - 1, before 'nature' has drawn the value ofv for
period t, a rational agent's best guess is that this variable will assume its mean value of
zero during the next period. From (19) we then find that the rationally expected value of
output in period t is:
Y~t- 1 = 9 - a l [h(n~t- 1 - .n*) + b(y~.t- 1 - 9)], (21)

given that agents are assumed to know the structure of the economy. including the values
of the parameters appearing in Eq. (19). In a similar way, since E[s 1] = E[v 1] = 0, we may
use Eq. (20) to calculate the rationally expected inflation rate for period t:
n~t- 1 = nft-t - ya2[h(n ft-t - n *) + b(y~t- 1 - .Y)] ~

ll(n~t- 1 - .n*) + b(y~.t- 1 - 9) = 0. (22)
Inserting (22) into (21) assuming that h * 0 . we get:
Y~t- 1 = Y· (23)
which may then be substituted into (22) to give the solution for the expected inflation rate:
.n~.t- 1 = .n*. (24)
2. The solution p ro.:edure described here worl<s in simple models like the present one which only includes expectations
relating to the current period. More general models typically also include e;pect ed values of variables for one or
several future time periods. The solution of such models requires more advanced techniques w hich w ill be left for a
future macro course.
Step 3. Now we need to insert(23) and (24) into (19) and (20) to get the final solutions for
inflation and output in terms of the exogenous variables and parameters of the model.
Taking this final step, we find:
(25)
(26)
Notice that the policy parameters n*, band h do not appear in the solution for y, given in
(26). Hence systematic monetary stabilization policy does not affect the evolution of real
output! This is the famous Policy InejJectiveness Propositioll (PIP) which says that system-
atic demand management policies ca11not il\fluence real output and employment whe11 expecta-
tions are rational. To understand the intuition for this striking result, note that the
aggregate supply curve (16) may be rearranged as:
y, = [J + (1/y)(n,- n~. 1 _ 1 ) + (1/y)s,. (27)
Since [I and s, are exogenous. the supply curve (2 7) implies that monetary policy can only
affect real output by influencing the inflation forecast error n, - n~.t- l' that is. by creating
unanticipated inilation. But since wage and price setters have the same information on
macroeconomic data as the central bank, and since they know the monetary policy rule
(14) as well as the way the interest rate aflects the economy through its impact on aggre-
gate demand. private agents call perfectly anticipate the ejfect of systematic monetary policy on
the current inj)ation rate. Thus systematic monetary policy (whether it takes the form of a
Taylor rule or some other fixed policy rule) cannot generate surprise inflation. and hence
it cannot cause output and employment to deviate from their natural rates.
Of course, purely erratic changes in the interest rate set by the central bank could
create unanticipated inflation and thereby allect real output (you are asked to demon-
strate this formally in Exercise 1). However, to be unpredictable. such policy changes
would have to be purely random and completely unrelated to the state of the economy.
Such a random policy would have negative welfare ell'ects by creating expectational
errors. inducing agents to make economic decisions they would probably regret ex post.
Clearly, such central bank behaviour would not qualify as stabilization policy.
The PIP was a frontal attack on the conventional wisdom. Hence it created a strong
controversy among macroeconomists when it was initially presented. 3 Because of its
important implications for public policy, we will now discuss the robustness of the PIP.
Policy effectiveness under rational expectations
The model of the previous section assumes that the interest rate as well as the wages and
prices for period t have to be set at the end of period t - 1, based on the information avail-
able at that time. In other words, it is assumed that the central bank cannot act on the
basis of more updated information than wage and price setters in the private sector. Th is
is hardly realistic. Wage contracts often fb' the nominal wage (or specify the evolution of
3. The PIP was originally put forward by Thomas J. Sargent and Neil Wallace, 'Rational Expectations, the Optimal
Monet ary Instrument, and the Optimal Money Supply Rule', Journal of Political Economy, 83, 1975, pp. 241-254;
and in a paper by the same authors entitled: 'Rational Expectations and the Theory of Economic Policy', Journal of
Monetary Economics, 2, 1976, pp. 169- 183.
© The McGraw-Hill
Companies. 2005
636 PART 6: TH E SHORT-R UN MODEL FOR T HE C LOSED ECONOMY
the nominal wage) for a considerable period ahead, and many firms only change their
prices at infrequent intervals. By contrast. the central bank can change its interest rate
quite quickly if it feels that new information on the state of the economy warrants such a
change. This means that the central bank can act after many wages and prices have
been set, so even if private agents fully understand the ellects of the interest rate change
on the economy, they may not have the opportunity to adjust their wages and prices
immediately to offset the ellect of the policy change on the real economy, because they are
temporarily locked into the existing nominal contracts.
A realistic macro model should therefore allow for the possibility that the central bank
can react to economic developments occurring after (some of) the nominal wages and prices
in the private sector have been set. To capture this, we will assume that the central bank
interest rate can react to the actuaL levels of current output y 1 and inflation TC 1, whereas
wages and prices are set on the basis of expectations formed at the end of the previous period
(so that n~.t- 1 is still the relevant expected inflation rate to include in the aggregate supply
curve). The monetary policy rule (14) is then replaced by the more familiar equation: 4
r 1 = i' + h(n1 - n*) + b(y,- 'fj). (28)
In practice. the central bank may not have perfect information on the current levels of
output and inflation, but Eq. (28) is just a convenient way of modelling the fact that the
central bank can react to observed economic developments occurring after (some of) the
wages and prices in the private sector have been set. The qualitative results derived below
will go through as long as the central bank can react to new information arriving after
some of the wage and price setting decisons for the current period were made.
Let us now study the implications of the policy rule (28). assuming that the AD curve
and the AS curve are still given by (15) and (16), respectively, and that the stochastic shocks
have the properties stated in (17) and (18). Expectations are still rational. so we must solve
this revised model by going through the three steps described in the previous section.
Step 1: Solve the mode/for y 1 and n 1 in terms of the exogenous variables and the expectations
variables. If we insert (28) into (15) and rearrange, we get:
(29)
From (16) we have:
n 1 - n* = n~.t- J - n* + y(y, - [J) + s,, (30)
which may be inserted into (29) to give:
(31)
4. Equation (28) assumes that the central bank sets the nominal interest rate as:
i, = 7 + " ~· •·•-• +h(;r, - n *) + b(y1 - y),
w here n~. ..,_ , is the private sector's expected inflation rate between the current and the next period, assuming that
private spending decisions in period t are based on information available up until the end of period t- 1. In other
words, the cent·al bank knows that the private sector's expected inflation rate is "~• 11_ 1 , and hence it is able to
control the ex ante real interest rater, = i,- "~• • .•- • through its control of the nominal int~rest rate i,.
Furthermore, we may substitute (29) in to (30) to get:
(32)
Step 2: Find n;,,_ 1 by taki11g expected values in (32), col!ditional on information availaiJ/e at
timet - 1. Performing this operation and remembering that E[s,] = E[z,] = 0 . we get:
.il~ r- 1 = n* (33)
Step 3: Find the.fillal solutiol!sfor n, a11d y 1 by inserting the solutiollfor the mtionally expected
inflation rate (33) into (32) a11d (31) . Taking this last step, we obtain :
(34)
(35)
The important message from (35) is that real output is no\V influel!ced by systematic monetnry
policy, since the policy parameters band h appear in the solution for y ,. For example, we see
that a more activist countercyclical policy (a higher value of /J) will reduce the impact of
demand and supply shocks on output. Hence the Policy [nellectiveness Proposition breaks
down. even though expectations are rational. The reason has already been suggested
above: since the expectations governing the formation of wages and prices are formed
before the central bank sets the interest rate, monetary policy can react to new information
which was not avaliable when the private sector formed its expectations lor the current
period. The inllation forecast error. n 1 - n~.•- 1' is therefore affected by the central bank's
systematic reactions to new events, and consequently systematic monetary policy will
influence real output. according to the aggregate supply curve (2 7).
Since it seems realistic to assume that the central bank can indeed act after (some
agents in) the private sector has temporarily locked itself into nominal contracts. most
economists today consider the Policy Inell'ectiveness Proposition as theoretically interesting,
but not very relevant in practice.
Optimal stabilization policy under rational expectations
Our conclusion is that monetary policy can in fact influence real output and employ-
ment even if expectations are rational. 5 But will the kind of monetary stabilization policy
which we found to be optimal under backward-looking expectations also be optimal
5. In Exerc ise 2 we ask you to demonstrate that systematic fiscal policy can also be used to stabilize output under
rational expectations, provided fiscal policy makers can act on information obtained after nominal wages have been
set.
© The McGraw-Hill
Companies. 2005
under rational expectations? To investigate this important issue, notice from (34) and
(35) that the variances of the output and inflation gaps are:6
J , , 2 2
o;,
-'
=E ((
y.- y-. ) -7 )
·
=
<J ~+G:i l <J,
[1 +az(b +yh)]-
, , (36)
y a~ + (1 + a 2 b) o·;
2
= . ( _
2
2 * 2] _
a" - E[ n, n ) - [1 +a (b+yh)] 2 (3 7)
2
When business fluctuations are generated solely by demand shocks, i.e., when 0, we a;=
see that the variances of output and inflation will be smaller the higher the values of the
policy paramewrs I! and b. This was also the case in the model with backward-looking
expectations analysed in the previous chapter. Regardless of the way expectations are
formed. the central bank should thus react with a sharp decrease (increase) in the real
interest rate when the inflation and output gaps fall (rise) in response to a negative (posi-
tive) demand shock, subject to the constraint th at the nominal interest rate cannot
become negative.
However, by comparing Eqs (36) and (37) above to Eqs (11) and (12) in the previous
chapter, we see th at the functional relationships between the policy parameters and the
variances of output and inflation are not quite the same under rational and static expect-
ations. This means that although the optimal qualitative policy response to demand shocks
is similar under forward-looking and bachvard-looking expectations, the qua11titative
solutions lor the optimal values of hand b will diller depending on the way expectations
are formed.
Consider next the opposite polar case where fluctuations are driven solely by supply
shocks ((J~ = 0). In that case it is convenient to take square roots in (3 6) and (3 7) to obtain
the following expressions lor the standard deviations of the output and inflation gaps (<1u
and a Jr ' respectively):
(38)
(39)
A policy that reduces the standard deviations o Y and a" will obviously also reduce the
variances entering the social loss function SL = (1~ + Ka!. We see from (38) and (39) that:
da,, -a~hos
~= <0. (40)
(Jb [1 + az(b + yi!W
0<1, = - a 2 y(1 + a 2 b)as < O.

(41)
(JJr [1 +az(b +yh)] 2
6. Since " 1 and s 1 are identically and independently distributed over time, and there are no persistence-generating
mechanisms in our simplified rational expectations model, the variances for period I stated in (36) and (3 7) are also
the asymptotic (long-run) variances. They are thus comparable to the asymptotic variances in the model w ith back·
ward -looking expectations analysed in the previous chapter.
Sorensen-Whitta- Jacobsen: I Part 6- The Short-run 21. Stabilization policy © The McGraw-Hill
In the limiting case where th e preference parameter K in the social loss function tends to
infinity, reflecting that policy makers are only concerned about securing a stable rate of
inflation, it follows from (41) that they should choose the highest possible positive value of
h, subject to the non-negativity constraint lor the nominal interest rate. In other words. to
stabilize inflation the central bank should raise the interest rate sharply in response to a
rise in the inflation gap. Moreover, when K ~ oo monetary policy should be procyclical
(b < 0) to make the numerator in (39) as small as possible. However, note that the central
bank cannot reduce the variance of the inflation rate to 0 simply by setting b = - 1/o 2 ,
since th e solution to our model is not valid in this case (forb = - 1/a 2 th e denominator in
(29) would become zero).
On the other hand, if policy makers focus only on minimizing fluctuations in output
(K = 0), we see from ( 40) that they should follow a strongly countercyclical policy, setting
the value of bas high as possible. It might seem from ( 40) th at the central bank can ensure
a zero variance of output by setting I! equal to 0, but this is not the case since the solutions
to our model assume that h * 0 (recall the derivation of(24)). Hence (38) and (39) would
not be valid if h = 0 , so if it wishes to minimize output instability when laced with supply
shocks. the central bank must choose a value of h which is very low. but bounded away
from 0 . 7
In the more general case where policy makers care about output stability as well as
inflation stability (0 < K < oo), there is a trade-oil' between min imizing the variance of
output and minimizing the variance of inflation when fluctuations are rooted in supply
shocks. To demonstrate this. we take derivatives in (3 8) and (39) and lind that
o<J11 = <xi1 + <x2 b)a, (42)

7 > 0,
oil [1 + aib + yil)]-
o<Jtr = a;yha5
(4 3)
- 7 > 0.
i)b [1 + aib + yh)]-
Equations (40) and (43) imply that a more strongly countercyclical monetary policy
involving a higher value of b will reduce the variance of output. but only at the expense of
a higher variance of inflation. Similarly, we see from (41) and (42) that while a more
aggressive interest rate response to a ch ange inn will help to stabilize the inflation rate, it
will also increase the instability of output. Following a procedure similar to the one
described in Chapter 20. one can show that this dilemma between stabilizing output
and stabilizing inflation remains in the general case where the economy is exposed to
demand shocks as well as supply shocks, provided the policy parameters h and fJ are
chosen optimally to min imize the social loss function SL = a~+ Ka; .
All of these qualitative conclusions regarding the optimal monetary stabilization
policy under rational expectations are similar to those derived in the previous chapter
when we analysed optimal stabilization policy in an economy with backward-looking
expectations. Hence the principles of optimal monetary stabilization policy may still be
summarized in a table like Table 20.1 in the previous chapter. Moreover. the economic
intuition lor these principles is still the same, so we will not repeat it here.
7. Technicallf, the rational expectations solut ion to our AS-AD model would be indeterminat e if h were zero. Hence any
solution fer the expected inflation rate would be consistent w ith the model. In such a c ase w ithout any anchor for
expectat ions, the economy could become very unstable, being strongly affected by changes in market psychology.
© The McGraw-Hill
Companies. 2005
Although it is reassuring that our qualitative analysis of optimal stabilization

policy remains valid whether expectations are forward-looking or backward-looking, we
should stress once again that the optimal quantitative values of fl and b depend on the way
expectations are formed, due to the different quantitative relationships between the policy
parameters and the variances of output and inflation under rational as opposed to static
expectations.
Expectations formation and macroeconomic dynamics
The dillerences in the variances of macroeconomic key variables under static versus ratio-
nal expectations reflect that the gradual adjustment of static expectations to observed
changes in the actual inflation rate generates persistence in output and in flation, as we
have seen earlier. By contrast, in our model with rational expectations the expected infla-
tion rate is pinned down by the central bank's target inflation rate. so there is no adjust-
ment over time in expected inflation following a shock. and hence no persistence
mechanism rooted in expectations formation. Indeed. we see from Eqs (3 5) and (34) that
output '>\Till fluctuate randomly around the natural rate under rational expectations.
and that inflation will fluctuate stochastically around the inflation target. As soon as a
temporary shock disappears, and assuming that no new shocks occur, the economy
immediately returns to its long-run equilibrium because there are no gradual shifts in the
SRAS curve stemming from the updating of expectations.
In Chapters 14 and 19 we saw that many real-world macroeconomic time series
display considerable persistence. The absence of such persistence in our simple model with
rational expectations might seem to speak in favour of the assumption of backward-
looking expectations. However, it is possible to generate persistence in macro models with
rational expectations. lor example by introducing autocorrelated supply shocks (see
Exercise 4), or by allowing lor the dynamics of capital accumulation. In such extended
models the basic point remains that the way expectations are formed will influence the
dynamics of the macroeconomy.
The optimality of a modified Taylor rule under forward-looking expectations
We have so far assumed that monetary policy follows a Taylor rule of the form (28).
In reality. the evidence on central bank beh aviour reviewed in Chapter 17 suggests that
the monetary policy of the major OECD economies is described somewhat better by a
slightly modi.lled Taylor rule where the lagged interest rate is included on the right-hand
side:
i, = r + .n: 1 + h(.n:, - .n:*) + b(y, - y) + ci,_ 1 • c > 0. (44)
This equation reflects that central banks tend to engage in interest rate smoothing. that is.
if the interest rate was high during the last period, it '>Viii also tend to be high during the
current period, and vice versa.
The rational expectations hypothesis may help us to understand why it may actually
be optimal for the central bank to follow a modified Taylor rule like (44) rather than some
other monetary policy rule. This issue has recently been addressed by economists Julio
Rotemberg and Michael Woodford. 8 They have developed an empirical AS-AD type model
lor the United States in which the representative consumer's lifetime utility loss from eco-
nomic instability may be approximated by a function of the same form as our social loss
function SL = o~ + KCJ;. Their model implies that the central bank can minimize the con-
sumer's welfare loss by setting the interest rate as a limction of all observed past values of
the output gap, all past values of the inflation gap, and all past values of the interest rate,
with declining coelllcients on past variables observed further back in history. However,
they also find that if the central bank follows a simple interest rate rule of the form (44)
with appropriately chosen positive values of the parameters h. IJ and c. it can ensure
almost the same level of consumer welfare as the welfare level attainable under the
optimal policy rule. In other words, a modified Taylor rule which includes the lagged
interest rate on the right-band side comes close to being optimal, according to Rotemberg
and Woodford.
The rationale for the modified Taylor rule (44) is that including the lagged short-term
interest rate i ,_ 1 in the interest rate reaction function increases the ability of the central
bank to inlluence aggregate demand. To understand this. recall that the variable i,
appearing in (44) is the sh ort-term interest rate controlled by the central bank Th e
greater the change in the longer-term market interest rates induced by a change in i ,, the
greater is the change in aggregate demand caused by a change in the short-term interest
rate. since investment in long-lived assets depends mainly on the long-term interest rate. 9
In Chapter 17 we showed that the long-term interest rate is an average of the current and
expected .future short rates. If financial market participants have static expectations, they
will expect that future short-term interest rates are simply equal to the current short rate.
In that special case where the yield curve is always flat, the short-term interest rate set by
the central bank becomes a powerful instrument for inlluencing the long-term interest
rate. But as we saw in Chapter 1 7, the yield curve is not always flat. reflecting that market
participants try to look ahead to anticipate future ch anges in the short-term interest rate.
When llnancial market agents do not mechanically extrapolate the current short-term
interest rate into the future, it becomes more dilllcult for the central ban k to allect the
long-term interest rate through its control of the short rate. However, if market par-
ticipants understand that monetary policy follows the rule (44) with c > 0 , they will
realize th at an increase in the curren t sh ort rate i 1 will also imply higher sh ort-term rates
in thejilture, ceteris paribus. Compared to the standard Taylor rule where c = 0, a change in
the short-term interest rate will therefore have a stronger impact on long-term interest
rates, and hence a stronger impact on aggregate demand when c > 0. In th L~ way the
central bank's interest rate policy becomes a more ellective instrument lor managing
aggregate demand. In particular, when interest rate policy has a stronger impact on
demand, it is less likely that the non-negativity constraint on the short-term nominal
interest rate becomes binding for monetary policy: if even a small cut in the short-term
8 . Julio J. Rotemberg and Michael Woodford, ' Interest Rate Rules in an Estimated Sticky Price Model' , Chapter 2 in
John B. Taylor (ed.), Monetary Policy Rules, NBER B usiness Cycle Seties 31 , University of Chicago Press, 19gg,
9 . This is true even if investment is financed by a sequence of short-t erm loans. W ith such a financ ing pattern, the
expected profitability of investment w ill depend on the current and exJBcted future short·term interest rat es prevail-
ing over t ~e lifetime of the asset. But since the long·term int erest rat e is an average of the current and the expected
future short·term rates, this means that the long·t erm interest rat e is still the relevant indicator of the firm's cost of
capital.
© The McGraw-Hill
Companies. 2005
interest rate has a substantial positive impact on aggregate demand, the central bank
would rarely end up in a situation where it would want the nominal interest rate to be
negative.
The issue of the optimal monetary policy rule is still subject to much research and
debate among monetary economists, but the findings ofRotemberg and Woodford suggest
that following a relatively simple modified Taylor rule may be a good strategy for monetary
stabilization policy when the markets are forward-looking. u:
The Lucas Critique
The Policy Ineflectiveness Proposition played a prominent role in the so-called rational
expectations revolution which swept through the tield of macroeconomics in the 19 70s.
But as we have seen, the PIP is not robust to plausible changes in assumptions, so although
it is regarded as an interesting theoretical benchmark, it is not really taken literally by
today's macroeconomists.
A more lasting in fluence of the rational expectations revolution was the so-called
Lucas Critique of macroeconometric policy evaluation. advanced by Nobel Laureate
Robert Lucas. 11 The Lucas Critique says that an econometric macro model with backward-
looking expectations which was estimated u11der a previous economic policy regime cannot he
used to predict economic behaviour under a new policy regime. The reason is th at a change
in the policy regime will allect private sector behaviour, including the way in which
expectations are formed, and this '<\Till change (some of) the parameters of the relevant
economic model.
Our AS-AD model with rational expectations provides a simple illustration of the
Lucas Critique. Consider Eq. (33) which shows that the rationally expected inflation rate
equals the central bank's target inllation rate n *. Suppose now that the government
appoints an 'inflation hawk' as a new central bank governor to implement a more anti-
in flationary monetary policy, implying a fall inn*. According to (3 3) rational agents '<\Till
then immediately reduce their expected inflation rate. If the economic analyst does not
allow for this effect of the change in policy regime on expectations, he '<Viii miscalculate the
etl'ect of the policy change on inflation and output. For example, if the analyst assumes
static expectations. we saw in Fig. 21.1 that he will predict a long period where agents
will overestimate the inflation rate. lollm<Ving the downward revision of the inflation
target. As a consequence. the analyst assuming static expectations will forecast a long and
protracted recession. But if expectations are actually rational. and if the reduction in
the offlcial inflation target is considered to be credible, (3 5) shows that the cut inn* can
be implemented '<Vithout any loss of output, because it is immediately translated into a
corresponding fall in the expected inflation rate.
The Lucas Critique is relevant for structural policies as well as lor stabilization policy.
For example, the labour market models presented in Chapters 12. 13 and 18 imply that
10. Empirical studies of central bank interest rate reaction func tions also i n d icat~ that a modified Taylor rule of the form
(44) w ith a positive value of c provides a better description of interest rate policy than the simple Taylor Rule w ith
c = 0. So central banks seem to act in accordance with the prescriptions of monetary theory!
11 . See Robert E. Lucas, Jr., 'Econometric Policy Evaluation: A Critique', Carnegie Rochester Conference Series on
Public Policy, 1, 1976, pp. 19-46.
the natural rate of unemployment depends on the level of unemployment benefits, among
other th ings. According to these models, a labour market reform involving a cut in
unemployment benefits is likely to reduce stmctural unemployment. A macroeconomic
model incorporating a Phillips curve estimated on historical data for periods when
beneflts were higher \>\Till then tend to overestimate the natural unemployment rate after
the reform, leading to inaccurate economic forecasts .
The Lucas Critique is a warning that one cannot mech anically extrapolate past
economic behaviour into the future. To avoid this problem. Lucas argued that the analyst
must build economic models with rational expectation~ and explicit micro foundations.
In such a setting it is possible to describe economic behaviour as functions of the
government's policy instmments and of the 'deep' parameters representing tastes and
technology which are not influenced by policy changes. Armed with such a micro-based
rational expectations model. one can in principle predict how a change in the policy
regime will affect the economy.
This methodological approach has had a profound influence on the way macro-
economists are nowadays trying to evaluate economic policies.
Announcement effects
Another interesting implication of rational expectations is the phenomenon of al'!nounce-
mellt ejfects of economic policies. In economies \>\lith forward -looking expectations,
markets will react to new illformatiol'l on factors which are likely to affect the future course
of the economy. Announcements of future changes in economic policy- or even just
olllcial statements which are interpreted as signals of future policy changes-will therefore
influence the state of the economy even before the policy changes are implemented. To
illustrate, financial market participants typically pay close attention to the public state-
ments of central bankers and immediately react when they believe that these statements
indicate future changes in monetary policy. A famous example of this was given on
5 December 199 6. when the US Federal Reserve chairman Alan Greenspan gave a speech
in which he argued that the booming stock prices at that time reflected the ;irrational ex-
uberance' of stock market investors. This was interpreted as a sign that the Fed was ready
to tighten monetary policy in order to bring stock prices down from an unsustainable
level. As a result, stock prices immediately tell quite significantly, not only in the US. but
around the world, as illustrated in Fig. 21.2 . 12
To show how forward-looking expectations give rise to announcement effects. we
will consider a simple model of the stock market. As you recall from Chapter 15, the (fun-
damental) market value 11, of shares outstanding at the start of period t is given by the dis-
counted value of expected future dividends. Assuming that dividends are paid out at the
end of each period and that the representative shareholder's real discount rate is expected
to stay constant over time at the valuer. we thus have:
D'' D' D'
V =- '·-' +~+~+ ·· · (45)
' 1 + r (1 + r) 2 (1 + r) 3 '
12. In the end, as you probably know, the Fed was not w illing t o tighten monetary policy sufficiently to halt the
stock market rally, so in the years after 1996 the stock market rose to even more exuberant height s before finally
c rashinQ in the year 2000.
© The McGraw-Hill
Companies. 2005
644 PAR T 6: THE SHORT-RUN MODEL FO R T HE C LOSED ECONOMY
USA
1.5
<»
C)
c
<1S 0.5
..c
0
<»
C)
0
c<»
<1S
~ -0. 5
<»
Q.
-1
- 1.5
29- 11 02-1 2 03- 12 04-1 2 05- 12 06-1 2 09- 12 10-12
1996 1996 1996 1996 1996 1996 1996 1996
Date
United Kingdom
1.5
<»
C) 0.5
c
<1S
..c
0
0
<»
g> - 0.5
c<» -1
0
Q;
Q. - 1.5
-2
- 2.5
29- 11 02-1 2 03- 12 04-1 2 05- 12 06- 12 09- 12 10-12
1996 1996 1996 1996 1996 1996 1996 1996
Date
Germany
3
G reenspan "
---- ""'
2
<»
/\__.-speech L~
C)
c
<1S
..c 0
/ \ I
0
<»
C)
-1
"'-./ \ I
c<» I
<1S
-2 \
0
\ I
v
Q;
Q. -3
-4
-5
29- 11 02- 12 03- 12 04- 12 05- 12 06- 12 09-1 2 10- 12
1996 1996 1996 1996 1996 1996 1996 1996
Date
Figure 21.2: Stock market reactions to Alan Greenspan's speech on 'Irrational exuberance' on
5 December 1996
© The McGraw-Hill
Companies. 2005
lCD
Japan
2
Q)
0:
~ 0 +--.~----------~~~~------~-------~----------
~
u
& - 1+------""-
c
0)
~ -2 +-------------------------------~--f------------
cf.
-3
29- 11 02-12 03-12 04-12 05-12 06-12 09-12 10-12

1996 1996 1996 1996 1996 1996 1996 1996
Date
Figure 21.2: (continued)

Note: The figures measure the percentage change from the previous trading day in closing day values of stock
indices. The stock indices used for Germany and Japan are, respectively, the OAX 100 and t he Nikkei 225 indices.
Sources: Japan and t he UK: Hanson & Partners AB. Germany: EcoWin AB. US: Dow Jones & Company.
where D~.,. 1 is the after-ta"~ real dividend which the shareholder expects to receive at
the end of period t + rr. given the information available at the beginning of period t. For
simplicity, suppose that the pre-tax real dividend d fluctuates stochastically around a
constant mean value so that: a
(46)
where t is a stochastic 'white noise' variable with zero mean. Suppose further that
dividends are initially taxed at the proportional rate rO' According to (46), the rational
expectation of the after-tax dividend received at the end of period t + 11 is then given by:
(47)
as long as the dividend tax rate is expected to stay constant. Inserting (47) into (45) and
collecting terms, we get: 13
(48)
Now suppose that at time t = t 0 the government suddenly announces that it will per-
manently reduce the dividend tax rate to a lower level r 1 < r 0 from some future time
13. In deriving (48), we have used the general formula:
for - 1 < a < 1.

1 -a
In our particular case we have a= 1/ ( 1 + !).
© The McGraw-Hill
Companies. 2005
t 1 > t 0 . Thus the actual net dividends accruing after timet 1 will be:
(49)
For the time periods from t = t 0 and up until t = t 1 - 1 the expected net dividend will still be
given by (4 7) (with t + 11 ~ t 1 - 1), since the dividend tax cut does not take etfect until
period t 1• However, from (49) it follows that the rational expectation of net dividends paid
out from period t 1 and onwards will be:
and (SO)
Inserting (47) lor t 0 ~ t + 11 < t 1 and (50) into ( 45), we obtain the value of the stock market
in the time interval between t 0 and t 1: L4
V = -
t (~{[
r 1-
(1 +1r)'·-t ] (
1- r +[
ol (1 +1r)'·-t ] (1 - r I)} (51)
We now have a complete picture of the evolution of the stock market. Before time t0
the value of the market is given by (48 ). Between the time of the policy announcement
and the time when the tax cut takes efl'ect, the market value is given by (5 1), and from the
time the tax cut is implemented we have
(52)
by analogy to (48) . Figure 21.3 illustrates the evolution of the stock market implied by
these three equations.
Note that the market value of shares immediately jumps at time t0 when the new in for-
mation about future tax policy becomes available to the market, but before the new policy
has actually been implemented. As the government announces its intention to change the
future tax rules, investors start to anticipate higher future after-tax dhridends, and this
expected rise in future earnings is inunediately capitalized in stock prices as a result of the
forward-looking behaviour of the market. The magnitude oft he initial jump in stock prices
can be derived by setting t = t 0 in (51) and subtracting (48) from the resulting expression to
find the rise in the value of the stock market between time t 0 - 1 and time t 0 :
(53)
This result is quite intuitive: the larger the future tax cut (r 0 - r L), the greater is
the expected rise in future net dividends. so the greater is the in itial rise in stock prices.
14. To derive (5 1), we use the formula in the previous footnote plus the generalformula:
1 -a"
1 +a+a 2 +a 3 +· · + a"- 1 = - -
1 -a
In the present case we have a= 1/ (1 + r) and n = t,- I.
- ,-?:,)d
(1 -
to t,
Figure 21.3: Effects of a d ividend tax c ut announced at time 10 and implemented at time 11
On the other hand, the longer it takes before the dividend tax is cut (the greater the
diilerence between t 1 and t 0 ), the more heavily the market discounts the rise in expected
future earnings, so the smaller is the initial upward jump in stock prices.
The magnitude 1 - r is sometimes called 'the retention ratio' because it measures the
fraction or the pre-tax dividend which the shareholder is allowed to retain for himself.
From (51) we see that bea·veen the time of announcement and the time when the tax cut
takes etl'ect, shares are valued as if the current retention ratio were a weighted average of
the lower initial retention ratio 1 - r 0 and the new h igher retention ratio 1 - r 1 which will
prevail after timet 1• According to (51) the weight 1/(1 + r)''-1 given to the new retention
ratio will be heavier, the shorter the time interval t 1 - t between the current period t and
the time when the dividend tax is cut. Again this is intuitive: the sooner the dividend tax
will be cut the more heavily the tax cut is capitalized in stock prices. After the initial jump
in stock prices at time t 0 , the value of the stock market will therefore gradually rise as the
date t 1 tor the tax cut moves closer. When that day arrives. the tax cut has already been
fully capitalized by the market, so from timet 1 and onwards there is no further increase in
stock prices. as illustrated in Fig. 21.3 . 15
For concreteness we have focused on the ellects of anticipated changes in future
economic policy. However, the method of analysis described in this subsection can also be
used to study the ellects of anticipated changes in other economic variables which are
relevant for the valuation of assets with fully flexible prices. In all such cases the general
rule is that asset prices 'jump' immediately at the time when new information becomes
available, and after that time the asset price moves gradually and continuously towards
its new long-run equilibrium value. In Exercise 5 we invite you to explore announcement
effects in a simple model of the housing market.
15. Of course, in practice the change in s tock prices after t ime / 0 w ill influence aggregat e private investment and con-
sumption and this in tum may have feedback effects on corporate earnings and dividends which w ill also affect the
evolution of stock prices. These complications are ignored here.
© The McGraw-Hill
Companies. 2005
648 PA RT 6: TH E SH ORT-R UN M ODEL FO R T HE C LOSED ECONOMY
21.3 Do people have rational expectations?

················································································································································································
Defending rational expectations
The REH asswnes that people's expectations accord with the predictions of the relevant
economic model. The hypothesis does not postulate that ordinary people literally apply
complicated economic models to form expectations about future economic conditions.
This would obviously be a highly wrrealistic assumption. But as we mentioned earlier, the
economic forecasts produced by professional economists are widely publicized by the
media and are thereby available to the general public. By using this information, people
should be able to form unbiased forecasts of, say. the rate of inflation.
Critics have pointed out that economic experts have different views on how the
economy works. Hence they often differ in their forecasts of future economic develop-
ments, so it is not obvious on which forecast ordinary persons should base their expect-
ations. This is probably one of the most compelling criticisms of the REH. We know that
the available economic models are wrong to some extent. since even the brightest
economic experts do not fully understand how the economy works.
Nevertheless. since economic agents have an incentive to avoid making forecasting
errors, it is not attractive to assume that their forecasts diller systematically from the fore-
casts of the preferred model of the analyst. Such an assumption would imply that the
analyst could permanently hide his supposedly superior information from other agents.
even though they would benefit from getting access to it. In the context of economic policy
making, if policy makers applied an economic model which assumes that private sector
expectations differ systematically from the model's predictions, they would efi'ectively be
assuming that they could permanently fool the public. It is hardly desirable to base
economic policy on the premise that policy makers are systematically wiser than the
private sector. As Abraham Lincoln said long before the rational expectations hypothesis
in economics was suggested: 'You can fool all of the people some of the time; you can even
fool some of the people all of the time, but you can't fool all of the people all of the time.'
Thus, although nobody really knows the 'true' stmcture of the economy in all of its
detail, and although for this reason agents cannot literally calculate the tme objective
mean values of economic variables, one can argue that it is not sale for the macro-
economic analyst to assume that the private sector is ignorant of the in formation em-
bodied in his preferred economic model. Instead it seems preferable to assume
mode/-corzsistent expectations, and this is exactly what the REH implies.
The importance of expectations for consumer behaviour
In the end the purpose of a working hypothesis like the REH is to generate theoretical
predictions which can be compared '>\lith the empirical facts. If its implications are not
rejected by the data, the REH may be a useful assumption even if it does not a priori seem
very realistic. [n the following we '>Viii investigate whether the implications of the REH
for the most important component of aggregate demand - private consumption - are
consistent '>\lith the data for aggregate consumption. This will give us an opportunity to
revisit the theory of consumption developed in Chapter 16 and to illustrate the surprising
consequences of the REH lor consumer behaviour.
To simplify the exposition, we will foUow the procedure in Ch apter 16 and split the
consumer's time horizon into two periods, period 1 ('the present') and period 2 ('the
future'), but it should be stressed that all our results carry over to a setting with many
periods. Applying the notation from Chapter 16. the representative consumer enjoys
instantaneous utility u(C 1 ) from consumption in period 1 and instantaneous utility u(C2 )
from consumption in period 2. But since the consumer is impatient. the present value of
future consumption is only u(C2 )/(1 + qy) in utility terms, where¢ is the exogenous rate of
time preference, so from the perspective of the present the consumer's actual lifetime utility
is u( C1) + u( C2)/( 1 + qy). As an extension of the analysis in Chapter 16, we will now explic-
itly allow lor the fact that the consumer does not know his future income with certainty.
Thus he does not know in period 1 exactly how much he can aflord to consume in period 2,
so the best thing he can do is to choose his present consumption C1 so as to maximize his
expected lifetime utility, taking his intertemporal budget constraint into account.
We will denote the expected instantaneous utility of future consumption by E1[u(C1 )],
where the subscript '1 ' below the expectations operator E indicates that the expectation is
formed in period 1. Thus the consumer's expected lifetime utility is u( ~) + E 1 [u(~)J/(1 + ¢).
If he decides to save an extra euro in period 1. he incurs a current utility loss equal to the
marginal utility of current consumption u'(C1) . But ifthe real interest rate is r. he will be able
to increase his consumption in period 2 by 1 + r. and this will increase the expected lifetime
utility from future consumption by the amount (1 + r)E[u'(C1)]/(1 + ¢),where E[u'(C2 )] is
the expected marginal utility of consumption in period 2. 16 For the consumer's choice of C1
to be optimal, he must be indiflerent between consuming an extra euro and saving an extra
euro in period 1. implying:
(54)
Equation (54) is a straightforward extension of the Keynes-Ramsey rule derived in Eq. (9)
of Chapter 16. According to (54) the marginal utility of an extra unit of current con-
sumption (the left-hand side) must equal the expected marginal utility from an extra unit
of saving (the righ t-hand side) for the consumer to be in optimum.
For concreteness. let us assume that the instantaneous utility function takes the
quadratic form:
u( C,) = C1 - '2a C,2, t = 1,2. (55)
which implies that (expected) marginal utility decreases linearly with the (expected) level
of consumption:
and (56)
In the benchmark case where the real interest rate equals the rate of time preference
(r = rp), it then follows from (54) and (56) that:
( 57)
16. Al though we allow for uncertainty about the consumer' s future labour and transfer income, we assume for simplic·
ity t hat tt e real retum on saving r is known w ith certainty. This w ill be the case if t he consumer invest s his saving in
a risk·free indexed bond.
© The McGraw-Hill
Companies. 2005
In other words, the consumer's expected future consumption level will correspond t o his
current consumption. since the assumption r = ¢> implies that he prefers to smooth his
consumption perfectly over time.
So far we have made no specific assumption about the formation of expectations. The
result E 1 [u'(C2)] = 1 - aE 1[C2] holds whether or not expectations of future consumption
are rational. To explore the implications of expectations formation lor consumption
behaviour, we need to introduce the consumer's budget constraints. As you recall from
Eqs (3) and (4) in Chapter 16, these are given by
Budget constraint for period 1: (58)
Budget constraint lor period 2: (59)
where V, is real financial wealth at the start of period t. Y, is after-tax labour and transfer
income in period t. and where we assume that all payments are made at the begim1ing
of each period. Using (58) to eliminate V2 from (59) . we obtain the consumer's inter-
temporal budget constraint:
C, Y,
C1 + --- = v 1 + Y 1 + - -- (60)
1+r 1+r
stating that the present value of lifetime consumption (the left-hand side) must equal the
present value of the consumer's lifetime resources (the right-hand side). Even if he does
not have rational expectations. the consumer knows that over the life cycle he cannot
spend beyond his means, so in period 1 he expects that his lifetime consumption is subject
to the constraint:
(61)
which is obtained by taking expectations of ( 60), assuming that the consumer knows hL~
current income and wealth V1 + Y 1 but not his future income Y2 • Inserting (57) into (61)
and using (59). we get:
= v,
cl = (1 + r)(Vl + yl - Cl) + EI[Y1]
= Vz + EI[Yz], (62)
which may be subtracted from (59) to give:
(63)
Thus the rise (fall) in consumption over time equals the amount by which the consumer
underestimated (overestimated) his future income when he formed the expectations
governing his first-period consumption.
21 STA BILIZATION PO LI C Y WI TH RATI ON AL EX PEC TATI O NS 651
From ( 63) we may derive the implications of alternative hypotheses regarding

expectations formation . Suppose first that expectations are static, implying that expected
future income equals actual current income:
(64)
According to (63) and ( 64) the change in consumption over time will then be equal to the
change in income:
(65)
Suppose alternatively that expectations of future income are rational, being an unbiased
prediction of the actual future income so that:
E[x] = 0 . (66)
where x is a stochastic variable with zero mean. rellecting that agents do not make
systematic forecast errors. Substituting ( 66) into ( 63 ), we get:
(67)
Following a similar procedure. one can show that the results in (6 5) and (6 7) also hold lor
a consumer who lives for many periods. Thus we have:
Static expectations: (68)
Rational expectations: C, - C, _ 1 = x ,. E[xJ = 0. (69)
Notice the striking result in (69) : under rational expect.atwns the change ill consumption over
time is entirely unpredictable. According to the REH, the best forecast of consumption
tomorrow is the level of consumption observed today. Indeed. current consumption is the
only unbiased forecast of future consumption. Because (69) implies that consumption
may change by any amount in any direction. it is said that consumption follows a randorn
walk. This implication of the REH was first derived in a seminal paper by American
economist Robert Hall. 17
Testing the random-walk hypothesis
In summary, under rational expectations the change in consumption over time equals the
unpredictable part of the change in income. whereas under static expectations the change
in consumption equals the total change in income, including that part of the income
change which could have been predicted by a forward -looking consumer. In another
inlluential paper, economists John Campbell and Greg Mankiw proposed a simple way of
testing the empirical relevance of these two competing hypotheses on consumption
behaviour. 18 Campbell and Mankiw assumed that a fraction -1 of aggregate income
1 7. See Robert E. Hall, 'Stochastic Implications of the Life Cycle·Permanent Income Hypothesis: Theory and Evidence',
Journal of Political Economy, 86, December 1978, pp. 971-987. At the time the random·walk hypothesis was so
provocative that when Hall first presented his paper, one prominent economist told him that he must have been on
d rugs w hen he w rote the paper.
18. See John Y. C ampbell and N. G regory Mankiw, 'The Response of Consumption to Income - A C ross-Country
lnvesti(lation', European Economic Review, 35, 199 1, pp. 723-767.
© The McGraw-Hill
Companies. 2005
652 PA RT 6: TH E SH ORT-RUN MODEL FO R T HE C LOSED ECONOMY
accrues to consumers who behave in accordance with ( 68), as if they have static expec-
tations. The remaining fraction 1 - A oftotal income was assumed to accrue to consumers
with rational expectations whose consumption follow a random walk. From (68) and
(69). the change in total consumption '<Viii then be given by:
0.;; A.;; 1. (70)
This equation spans the two alternative hypotheses on expectations: if-1 = 1 all consumers
have static expectations, and if A= 0 they all hold rational expectations. Using data on
consumption and income, Campbell and Mankiw provided an econometric estimate of the
magnitude of the parameter A in a number of countries. Their results are summarized in
Table 21 .1. where the ligures in brackets indicate standard errors. 19
In all of the countries the magnitude of A- was found to be signillcantly greater than 0 ,
ranging from about 0 .2 to 0.4. In other words, a large part of consumption seems to be
undertaken by consumers who are not forward-looking. This is clearly at odds with a
strict interpretation of the REH. On the other hand. forward-looking consumers do seem
to account for the bulk of consumption in most countries. so the REH certainly cannot be
dismissed as being irrelevant. Rather. it appears that consumers are divided into two
groups; one which is forward-looking. and another which is backward-looking. Note that
consumers obeying ( 68) do not necessarily all have static expectations. It is also possible
that they are simply myopic, living from 'hand to mouth' . immediately consuming all of
their current income '<Vithout caring about the future. Indeed, this is the interpretation
adopted by Campbell and Mankiw. However, in both cases the fact remains that these
consumers do not behave as predicted by the REH.
Table 21.1: The proportion of consumers with static expectations (.A.)

Estimate of tl
Country Sample period (standard errors in brackets)
Canada 1972(1 )-1988(1 ) 0.225

(0.1 07)
France 1972(1 )-1988(1 ) 0.401
(0.208)
Sweden 1972(2)-1988(1 ) 0.203
(0.092)
United Kingdom 1975(2)-1988(2) 0.351
(0.11 7)
United States 1953(1 )-1985(4) 0.357
(0.1 73)
Source: J.Y. Campell and N.G . Mankiw, 'The Response of Consumption to Income', European Economic Review, 35, 1991, p. 736.
19. Campbell and Mankiw actually measured the changes in income and consumption in logarithms and included a
constant tenn on the right -hand side of Eq. (70) to capture long·tenn growth. S ince periods with surprise increases
in income probably often coincide w ith periods with rapid g rowth in total income, the explanatory variable Y,- )'; _1
is likely to be positively correlated w ith the error tenn x r Readers trained in econometrics will know that the esti·
mation method of O rdinary least Squares w ill then generate an upward bias in the estimate for .t To deal w ith this
problem, Campbell and Mankiw used the method of Two·Stage least Squares. The first st age of this estimation
method involved the use of lagged changes in consumption (which are necessarily uncorrelated w ith x) in an OlS
regression to predic t Y.- )';_1• In the second stage, the estimated values of Y.- )';_1 were used as the explanatory
variable in (70) to estimate the value of A in another Ol S regression.
Of course, the results of this analysis should be interpreted with some caution. First of
all. to derive the random walk hypothesis, we assumed a quadratic utility function (Eq.
(55)) combined with rational expectations. Thus the econometric study by Campbell and
Mankiw is really a joint test of a hypothesis on expectations formation and a hypothesis on
the form of the utility function. In principle. if the estimated value of A deviates sig-
nillcantly from 0 . it could be because the utility function is not quadratic rather than
because expectations are not rational. However, one can actually derive the random walk
hypothesis using approximations to more general utility functions. so the assumption of
quadratic utility is not as critical for the findings of Campbell and Mankiw as it may seem.
As another potential objection, a positive value of A may indicate that some con-
sumers are credit-constrained rather than myopic. If a forward-looking consumer expects
his future income to be higher than his current income. he may want to borrow against
his anticipated future income in order to smooth his consumption over time. But in the
absence or collateral, his bank may be unwilling to accommodate his credit demand lor
fear that he may default on the loan if his future income turns out to be lower than
expected. Then the best thing such a credit-constrained consumer can do is to consume all
of his current income ( C, = Y,). implying C, - C, 1 = Y, - Y, 1 • as if expectations were
static.
Although the hypothesis of credit constraints may sound plausible, there are some
arguments against this interpretation of the empricial findings of Campbell and Mankiw.
Up until the 1980s, governments in the OECD countries typically intervened in Hnancial
markets in an ellort to regulate the volume of credit. During the 19 80s these capital
market regulations were lifted in most countries. If liquidity constraints were widespread.
one would expect such capital market liberalization to increase the volume of credit,
thereby relaxing the credit constraints of many consumers. Presumably this would mean
that our parameter A should decrease. if this parameter mainly reflects credit constraints
rather than myopic or backward-looking behaviour. However, Campbell and Mankiw
found no cross-country evidence of a statistically signillcant drop in A in the latter part of
their sample period where capital markets had been liberalized. In a careful analysis of
macroeconomic data from Sweden. economists Jonas Agell and Lennart Berg likewise
failed to llnd any signillcant ellect of financial deregulation on the value of A. 20 This
indicates that consumption is not signiHcantly constrained by credit rationing.
One can also approach the issue of liquidity constraints from another angle. If such
constraints are important. one would expect that current consumption will tend to be
governed by current income for households without liquid assets. whereas households
with significant liquid assets will be able to smooth their consumption over time. because
they can simply sell their assets rather than asking the bank lor a loan. However, using
microeconomic data on income and consumption in the United States, economist John
Shea found no evidence of a significant diflerence between the consumption behaviour of
households '<\lith and without liquid assets. 21
Since all of this evidence suggests th at liquidity constraints are not a very important
determinant of consumption. our preferred interpretation of the study by Campbell and
20. Jonas Agell and Lennart Berg, 'Does Financial Deregulation C ause a Consumption B oom?' S candinavian Journal
of Econcmics, 98, 1996, pp. 579-601.
21. See John Shea, ' Union Contracts and the Life·Cycle/ PermanenHncon e Hypothesis' , American Economic Review,
as, 1995, pp. 186-200.
© The McGraw-Hill
Companies. 2005
Mankiw is that many consumers behave as if they have static expectations or as if they are
simply myopic. At the same time a large part of total conswnption seems to be governed
by fonvard-looking rational expectations.
Towards a more general theory of expectations
Thus both of the competing hypotheses on expectations formation which you have
encountered in this book - backward-looking expectations and rational expectations -
appear to have some empirical relevance. One might therefore specify the expected
inflation rate as:
n~ = Ant- 1+ (1 - A)n~t- L • (71)
Here .n~ is the average expected inflation rate calculated across the entire population, and
is the rationally expected inflation rate. based on all information available up until
.n~.t- J
the end of the previous period. Equation (71) assumes that a fraction A of the population
has static expectations. while the remaining fraction has rational expectations. If we go
back to our AS-AD model (1)-(5) and replace (5) by (71) . we obtain a model with partly
rational expectations. To solve such a model. one still has to go through the three steps
used to solve a model with purely rational expectations. Making the first two steps, 12 we
lind the expected inflation rate for those agents who have rational expectations:
a l l!
e l
n u- =.n* + (-A-)<n .t-1 - .n*) · (1 = --- (72)
A + ycc 1 + a1b
The rational agents know that the expectations of the backward-looking agents have
some influence on the actual intlation rate via the expectations-augmented Phillips curve.
Hence the rational forecast of intlation stated in (72) accounts for the tact that the expec-
tations held by the backward-looking agents are not anchored by the central bank's infla-
tion target. This is why the term (n ,_ 1- .n*) enters the right-hand side of ( 72) with a
positive coefficient wh ich is larger the greater the fraction of the population with back-
ward-looking expectations.
Using (71) and (72), one can show that our AS-AD model with partly rational ex-
pectations can be reduced to th e following two di!ference equations in the output gap.
[J , = y,- [1. and the inflation gap, n,=.n,- n *:
(73)
.n, = -A-) .n,_l+---+---

A ( yz,
A s, . (74)
A + ya 1 + ycc 1 + ya
The model analysed in Chapters 19 and 20 is the special caseof(73) and (74) where A = 1
(if you compare (73) and (74) to Eqs ( 41) and (42) in Chapter 19 and setA = 1 and</> = 0 ,
implying that all of the population has static expectations, you will see that the two sets of
equations are exactly identical). The model set up in Section 2 of this chapter is the other
22 . Here we maintain our previous assumption that the expected inflation rat e "~·• entering the monetary policy rule (4)
is the private sect or's expected inflation rate determining the real interest rate in (2).
special case where A. - 0 (in this case (73) and (74) collapse to (3 5) and (34), given that
a=a 2 h/(l + a 2 b) and z 1 = (1 + a 2 b)v 1) . In the more general case where O<A< 1, (73)
and (74) still have the same general form as the dill'erence equations in the output and
inflation gaps implied by our AS-AD model with purely static expectations. From an
empirical perspective the model (73) and (74) with A> 0 has the attractive property that
it generates persistence in the output and inflation gaps, that is, the current gaps depend
positively on their own lagged values. As you recall from Chapter 14. real-world macro-
economic time series do display such persistence.
Note that substitution of (72) into (71) yields:
ya(1 - A.)
n~ = qm* + (1 - cp)n 1_p O~cp= ~1. (75)
A+ ya
This shows that, in an economy \>\lith partly rational and partly static expectations. the
average expected inflation rate is a weighted average of the central bank's inflation target
and last period's actual inflation rate. When all agents are rational (A. = 0). we have cp = 1
and n~ = :n*, and when they all have static expectations (A = 1), we get cp = 0 and n~ = n 1_ 1•
Equation (75) was derived on the assumption that a part of the population forms
rational expectations in accordance with (72). As we have discussed, this assumption that
(some) people behave as if they know the entire structure of the economy is indeed a
strong one. But there is an alternative interpretation of(75) which does not require that
agents are quite as sophisticated as postulated by the REH: even if the forward-looking
part of the population may not always have su1Ilcient information to be able to form the
strictly rational expectation given by (72). these people may at least be informed about the
central bank's inflation target n*. Therefore, if the central bank has credibility (an issue to
which we return in the next chapter), it may make good sense for agents to assume that,
on average, the inflation rate will correspond to the target rate of inflation. Thus we may
interpret the parameter cp in {7 5) as the fraction of the population which is informed about
(and has confidence in) the central bank's inflation target. Since this is also the long-run
equilibrium rate of in11ation, this forecasting behaviour may be called 'long-term rational
expectations' or 'weakly rational expectations'. We will return to this hypothesis in
Chapters 24 and 2 5 when we consider the open economy.
21.4 .~.~.~~.<:J:~Y....................................................................................................................................................
1. The assumption of backward-looking (static or adaptive) expectations is hard to reconcile
with rational behaviour because it implies that economic agents may make systematic fore-
cast errors.
2. As an alternative to backward-looking expectations, economists have developed the rational

expectations hypothesis {REH) according to which an agent's subjective expectation of an
economic variable equals the objective mathematical expectation of the variable, calculated
on the basis of all relevant information available at the time the expectation is formed.
3. The REH assumes that the information available to agents includes knowledge about the
structure of the economy. Hence rational expectations are model-consistent: they correspond
© The McGraw-Hill
Companies. 2005
656 PAR T 6: THE SHORT-RUN MODEL FOR T HE C LOSED ECONOMY
to the predictions of the relevant economic model. This does not require that ordinary people
are able to solve economic models, since the average person may rely on the public ly avail·
able forecasts of professional economists.
4. Some macroeconomic models with rational expectations have led to the Policy Ineffective·
ness Proposition (PIP) which claims that systematic demand management policy cannot
affect real output and employment because the private sector w ill fully anticipate the effects
of systematic policy on the rate of inflation.
5. The PIP is nowadays considered unrealistic, since the central bank can typically change its
policy in reaction to new economic developments occurring after nominal wages have been
temporarily locked into existing contracts. Because nominal wages and prices only respond
to economic shocks w ith a lag, systematic monetary policy can affect output and employment,
even if the policy is fully antic ipated by rational agents. In these c ircu mstances the optimal
monetary stabilization policy is qualitatively similar to the optimal monetary policy response to
shocks under backward-looking expectations, although it is quantitatively d ifferent. Thus ratio·
nal expectations do affect the optimal policy, but do not make stabilization policy ineffective.
6. The REH has led to the Lucas Critique wh ich says that an econometric macro model wh ich
was estimated under a previous economic policy regime cannot be used to predict economic
behaviour under a new policy regime. The reason is that a c~ange in the policy regime w ill
affect private sector behaviour, including the way in which expectations are formed.
7. Under rational expectations the announcement of future changes in economic policy will influ·
ence the economy already at the time of announcement, even before the new policy is imple·
mented. In particu lar, the flexible prices of financial assets such as stocks w ill 'jump'
instantaneously at the time of announcement and w ill then gradually adjust towards its new
long-run equilibrium value as the date of implementation of the policy change comes c loser.
8. The REH has been criticized for being unrealistic because economists d iffer in their views of
the workings of the economy, making it d ifficu lt for the average person to base their expecta·
tions on expert forecasts . Defenders of the REH argue that it is not safe to base economic
policy evaluation on the assumption that policy makers are systematically better informed than
the private sector. Hence policy makers should assume that the knowledge embodied in their
economic models is also available to the private sector, as implied by the REH.
9. When consumers with rational expectations seek to smooth their consumption over time,
private consumption w ill follow a random walk, changing only as new information about future
incomes becomes available. The random walk hypothesis implies that the current consump·
tion level is the best forecast of futu re consumption (adjusted for underlying trend growth) and
that consumption changes only in response to unpredictable changes in income. By contrast,
under static expectations the change in consumption corresponds to the change in the con·
sumer's total income. Empirical evidence suggests that a large part of aggregate consump·
tion follows a random walk, consistent with the REH, but at the same time many consumers
behave as if they have static expectations.
10. In an economy where some consumers have rational expectations and others have static
expectations, the average expected inflation rate may be written as a weighted average of the
central bank's inflation target and last period' s inflation rate. The weight g iven to the inflation
tarQet may be interpreted as the fraction of aQents informed about the monetary policy tarQet,
21 STABILIZATION PO LI CY WI TH RA TIONAL EX PECTATI ONS 657
even if these people do not have all the information needed to form strictly rational expecta-
tions regarding the short run.
21.5 Exercises
Exercise 1. Issues in rational expectations
1. Define and explain the concept of rational expectations and d iscuss the arguments for and
against this hypothesis.
2. Explain the Policy Ineffectiveness Proposition and d iscuss its relevance.

3. Explain the content of the Lucas Critique and its implications for the evaluation of the effects
of economic policy.
4. Explain what is meant by 'interest rate smoothing' ? Explain why interest rate smoothing may
be a desirable ingredient in an optimal monetary stabilization policy when expectations are
rational.
5. Explain the content of the random walk hypothesis for private consumption. Explain the dif-
ference between the dynamics of private consumption under static and under rational expec-
tations.
Exercise 2. Monetary and fiscal policy under rational expectations
Suppose that the central bank does not always react systematically to changes in macro -
economic conditions so that monetary policy may be described by the interest rate rule:
r, = r + h(:n:,,_, - :n*) +a,, h > 0, (76)
where a, is a 'white noise' stochastic variable reflecting the non-systematic part of monetary
policy. Equation (76) states that the central bank bases its policy decisions on the expected
inflation gap, since it does not have full information on the current inflation rate at the time
when it sets the interest rate. For simplic ity, we assume that the bank does not react to the
expeGted output gap.
As usual, the economy's demand and supply sides are described by:
SRAS: :n,= :JC.~I- 1 + y(y, - ji) + s,. (78)
where z, and s 1 are wh ite noise reflecting demand and supply shocks.
1. Assu me that expectations are rational and find the model solution for real output by going
through the three steps in the solution procedure described in the main text. On this basis,
show that the variance of output is given by:
(79)
where a; and a~ are the variances of z and a, respectively. ls monetary policy 'effective' in this
model? What would be the effect of greater predictability of monetary policy? Would greater
predictability be desirab le? Discuss.
© The McGraw-Hill
Companies. 2005
658 PAR T 6: THE SHORT-RUN MODEL FOR T HE C LOSED ECONOMY
The variable z, - a 1(g,- g)+ v1 includes deviations of public spending from trend as well
as the private demand shock variable v,. Suppose now that fiscal policy reacts to the
expected inflation and output gaps in the following systemat c way, reflecting an intended
countercyclical and anti-inflationary policy:
g,- g = c,(n~,- t - n *) + c,(9 - Y~t- 1) ~
z, =a t c,(117,1_t - n*) + Cl. t c,(9 - Y~t- t ) + v,. (80)
Here we assume that u 1 is white noise. To focus on fiscal policy, suppose further that mone·
tary policy is passive, keeping the real interest rate at its natural level:
r,= r. (81)
2. Assume rational expectations and demonstrate that the Policy Ineffectiveness Proposition
holds in the model consisting of Eqs (77), (78), (80) and (81). Explain why fiscal policy is
ineffective.
Suppose instead that the fiscal authorities can react on current information on the actual
output and inflation gaps whereas nominal wages are pre-set at the start of each period when
inflation expectations can only be based on the information available up until the end of the
previous period. Thus (78) is still valid, but (80) is replaced by:
(82)
3. Show that the Policy Ineffectiveness Proposition no longer holds in the model consisting of
(77), (78), (80) and (81) even if expectations are rational. Explain why fiscal policy is now
effective.
Exercise 3. Nominal GOP targeting with rational expectations
You are now asked to study the properties of an economy where the fiscal and monetary
authorities pursue a target for the growth rate of nominal GOP. Specifically, we assume that
actual g row1h rate target growth rate
of nominal GOP of nominal GOP
,......-"-, ,-------"-,
Y7- 17-1=.u + I'](Jt- y,_1) + v,, 'I> 0, (83)
where y;' is the natural log of nominal GOP, and where the stochastic white noise variable u1
reflects that the authorities cannot perfectly control the growth in aggregate nominal demand.
The first expression on the right-hand side of (83) shows that policy makers try to speed up
the growth in nominal GOP if the previous period's real output y 1_ 1 has been below the trend
level of real output, and vice versa. Thus the authorities follow a countercyclical demand man·
agement policy, but since they can only observe output with a lag, they react to the previous
period's activity level rather than to current activity.
In our previous notation, y7 "' p , + y,, and n 1 "' p 1 - p 1_ 1 , where p andy are the logs of the
price level and of real output, respectively. By definition we thus have:
Y7- 17-1=n,+ y,- Y r- t· (84)
Equations (83) and (84) describe the economy's demand side. The supply side is g iven by the
AS-curve:
n, = 7l~r- 1 + y(y,- .9) + s,, (85)
where the stochastic supply shock variable s 1 is wh ite noise.

21 STABILIZATION POLICY WI TH RA TIONAL EX PEC TATI ONS 659
1. Find the solutions for real output and inflation on the assumption that expectations are
rational. Does the Policy Ineffectiveness Proposition hold in this model? Explain your results.
Suppose next that the authorities can react to current output so that the policy rule (53)
is replaced by
(86)
2. Find the solution for real output and check whether the Policy Ineffectiveness Proposition
holds. Which of the scenarios (Question 1 versus Question 2) do you consider to be most
realistic? Give reasons for your answer.
Exercise 4. Output persistence under rational expectations
In the main text we noted that our basic AS-AD model with rational expectations does not
generate any persistence (autocorrelation) in the deviations of output from trend, in contrast
to what we observe empirically. This exercise asks you to show that persistence in output w ill
emerge if we allow for autocorrelation in our supply shock variable. Thus we now describe the
economy by the following equations, where we assume for simplicity that the central bank
only reacts to the inflation gap:
SRAS: n, = n~t- 1 + y(y, - y) + s,' (87)
Monetary policy ru le: r1 = 7+ h(n 1 - n *), h > 0, (89)
Supply shock: s, = ws 1_ 1 + c,, 0 < (t) < 1, (90)
Demand shock: z 1 = pz1_ 1 + x, , O~p < 1. (9 1)
The stochastic variables c 1 and x, are assumed to be wh ite noise. When they form their
inflation expectations for the current period, nf.1_ 1 , the information set available to private
agents includes knowledge of the model (8 7)-{91) plus information on the shocks observed
during the previous period, s 1 _ 1 and z 1_ 1 • However, the private sector's information set does
not include information on the 'innovations' to the shocks, c 1 and x,.
1. Show tho.t the ro.tiono.l cxpccto.tions solution for the output go.p in the model {87) - {91) is
g iven by:
Yt• "' Y1 -y- =pR(X1 -a 2 hC 1) - WS 1-

--,
1
(92)
y
Try to explain why the output gap is affected by s 1_ 1 , but not by z 1_ 1•
2. Show that output d isplays persistence by using (90) and (92) to write the solution for the
output gap in the form:
ji, =aj,_1+ r., , (93)
where 0 <a< 1, and where ~: 1 is a (composite) white noise variable. Write the explicit expres-
sions for a and E 1• {Hint: start by lagging (92) by one period and then use the fact from (90)
that ws 1_ 2 = s 1_ 1 - c 1_ 1 to write s 1_ 1fy as a function of ji1_ 1 , x 1_ 1 and c 1_ 1 • Then insert the
resu lting expression for s 1_ 1/y into (92) and collect terms.) Discuss whether output
persistence can be said to be endogenous or exogenous in th is model.
© The McGraw-Hill
Companies. 2005
660 PAR T 6: THE SHORT-RUN MODEL FO R T HE C LOSED ECONOMY
Exercise 5. Policy announcement effects in the housing market
In the main text we saw that when expectations are forward -looking, announcements of future
policy changes will have an effect on the stock market right from the time of announcement.
Here we ask you to analyse announcement effects on the housing market. We consider a rep·
resentative consumer who owns a fixed stock of housing yielding a flow of housing services
whose real rental value h 1 fluctuates stochastically around the constant mean value li. We
therefore assume that, during any period, the home-owner expects that the real value of his
housing service w ill be li. The home-owner pays a property tax which is levied in the real
amount r 1 per square metre. For simplic ity, the property tax is thus assumed to be unrelated
to the market value of the house, but the tax may vary over time. Suppose that the consumer
owns one unit of housing w ith a market price 0 1 at the beginning of period t. Suppose further
that, at the start of period t, the market price at the start of the next period is expected to be
0 ~. 1 • 1 • If the real interest rate is r (assumed for convenience to be constant), the cu rrent
market price of housing must then satisfy the following arbitrage condition for consumers to
be willing to own the existing stock of owner-occupied housing:
after· tax value

opportunity cost of of housing expected cap~al
home-~ersh ip servtee gain
~
r0 1 li- r 1 + 0~+ 1,1- 01. (9 4)
The left-hand side of (9 4) measures the consumer's opportunity cost of owning his home
rather than selling it at the going market price and investing the proceeds in the capital
market, in which case he would earn an interest on the wealth invested in his house. The right·
hand side of (94) is the return to home-ownership, consisting of the after-tax value of the
housing service yielded by the consumer's housing wealth plus the expected capital gain on
that wealth over the period considered. Rearranging (94), we get:
0 = fi-r,+ 0 ~+ 1,1
(95)
1
1+ r ·
Rational home-owners know that an arbitrage condition similar to (95) must also hold in future
periods; they just do not know w ith certainty what the future property tax w ill be. At the start
of period t, the expectations of future housing prices will thus be given by:
• = li- r ~. 1.1 + o ~. 2.1 0 = li- 1.' ~+ 2, 1 + 0 ~+ 3, 1

0 1+1,1 etc. (96)
1+r ,. 2.1 1+r
where r~+n, t is tne property tax rate expected in period t to prevail in period t + n. We assume
that agents do not expect the real price of housing to rise systematically at a rate in excess of
the real interest rate, so the expected housing price satisfies the boundary condition:
lim 0 ~+n,l = 0 . (97)

n-~ 1+r
1. Explain more carefully why the arbitrage condition (94) must hold for the housing market to be
in equilibrium. Explain the economic mechanism which establishes this equilib rium.
2. Show by using (95)- (97) that the current market price of housing is:
(98)
21 STABILIZATION PO LI CY WI TH RA TIONAL EX PEC TATI ONS 661
where the last equality in (98) exploits the fact that r ~ 1 = -r1, since the home-owner is assumed
to know the current property tax rate from the start of the period. (Hint: you may use the same
procedure as the one we used to derive the fundamental stock price (6) in Chapter 15.)
Give a verbal interpretation of the result in (98) and compare with the expression for the
fundamental share price g iven in Eq. (6) in Chapter 15.
Assume that the real property tax is kept constant at the rate r 0 in the periods between
0 and t0 :
for 0 < t < t0 • (99)
At the start of period t0 , the government suddenly announces that it w ill cut the property tax
to the lower constant level r 1, taking effect from the start of the futu re period t 1• In other
words,
for t ;;. t1 > t0 • (100)
3. Use (98) - (1 00) to show that:
0Ti- r
0, = - - for 0 < t< t0 , ( 1 0 1)
r
0 --
Ti - -1 {[ 1- 1 ]r 0 + [ 1 ] r 1} for t0 EO; tEO; t1 , (102)
,- r r ( 1 + r)'·-• ( 1 + r)'·- •
Ti- r
0 ,=--
1 for t ;;. t1 . (103)
r
(Hint: follow the same procedure as the one we used to derive announcement effects in the
stock market, including the formulae in Footnotes 13 and 14). Draw a d iagram to illustrate the
evolution of the housing price from time 0 onwards. Give an interpretation of (1 02) and
explain why the housing price reacts already at the time the future change in tax policy is
announced. Would the same resu lts emerge if expectations were static?
4. Use (10 1) - (102) to derive an expression for the size of the price jump (010 - 0 10 _ 1) between
period t0 -1 and period t 0 when the futu re p roperty tax cut is announced. Explain the factors
determining the size of the initial price jump.
22. The limits to
stabilization policy:
© The McGraw-Hill
Companies. 2005
Macroeconomics Economy Credibility, uncertainty and
time lags
Chapter
The limits to
stabilization policy
Credibility, uncertainty and time
:• lags
•
•
•
•
: D o monetary and fiscal policy makers have the ability to stabilize the macro
: economy. thereby reducing the social costs ofbusiness cycles? In the two previous
• chapters our answer to this basic question in macroeconomics has been: yes.
However, our analysL~ was based on some important simplifying assumptions. First of all.
we assumed that policy makers have perfect informatioll about the current state of the
economy. Second. we assumed that they can react immediately to this information and
that their policy actions have predictable and well-known quantitative ejfects. Third, in the
case with forward-looking agents we postulated that any policy rule announced by policy
makers is always considered fully credible by the public implying, for example. that policy
makers never have any problem convincing the public that they will stick to an anti-
infl ationary policy. Taken together, these assumptions are very optimistic and not very
realistic. In this chapter we shall study the problems of stabilization policy when these
strong assumptions are replaced by more realistic ones.
We will start by studying the problems of establishing the credibility of an anti-
infl ationary monetary policy. This part of our analysis will show how our AS-AD model
combined with the hypothesis of rational expectations may provide a theoretical case
for delegatio11 of monetary policy to an independent cep·rtr-a! bank, as a lot of countries
have actually done in recent years. Other parts of the chapter investigate how uncertainty
about the current state of the economy and time lags in the implementation of policy
limit the scope lor stabilization policy. In the llnal section we focus particularly on Hscal
policy, investigating whether llscal policy makers have historically been able to dampen
the business cycle.
22.1 Policy r ules versus discretion : the credibility problem

................................................................................................................................................................................
The time· inconsistency of optimal monetary policy
In the previous chapter we saw how changes in the rational expectations of the private
sector may sometimes oflset the intended eflects of stabilization policy. Another basic dis-
covery made by the rational expectations school in macroeconomics was the insight that
662
Sorensen-Whitta-Jacobsen: I Part 6- The Short-run 22. The limits to © The McGraw-Hill
Introducing Advanced Model for the Closed stabilization policy: Companies. 2005
time lags
22 TH E LIMI TS TO STABI LIZ ATION POLICY 663
it may not be possible to implement an optimal economic policy because it lacks credibility.
For example, when the previous chapter derived the rational expectations solution for
expected inllation. n ' = n *, it was assumed that private agents are confident that the
central bank will always stick to the announced monetary policy rule with a target infla-
tion rate. n*. But such credibility of economic policy may be difficult to achieve when
policy makers can undertake discretionary policy ch anges after private agents have formed
their expectations. 1 If monetary policy makers announce that they will keep the inflation
rate down to a certain target level. the private sector may not consider such a statement to
be credible if the central bank can boost output and employment by generating unantici-
pated inflation.
We will now use our AS-AD model to illustrate the credibility problem arising under
discretionary policy with rational expectations. To simplify (without invalidating our
qualitative conclusions) . we will set the aggregate supply curve parameter y = 1. For the
moment, we will also abstract from demand and supply shocks (v 1 = z1 = s 1 = 0). The
expectations-augmented Phillips curve may then be written as:
ltt = 11~.1- I+ Y t - [J, (1)
and the goods market equilibrium condition simplilles to:
Yr - [J = -air 1- f). (2)
As before, we assume that the central bank can observe the expected rate of inflation (say,
through consumer surveys or by observing the dillerence between the interest rates on
indexed and non-indexed bonds) when it sets the nominal interest rate. Thus the central
bank can set the real interest rate r1 after the private sector has formed its expectations of
inflation. It then follows from (2) that the central bank can control the current output gap
y 1 - [J. According to (1) it can therefore determine the actual inflation rate, through its n,.
choice ofy 1 - [J, for any given expected inflation raten~.t- J ·
We assume that monetary policy mal<ers would like to minimize the social loss function:
}( > 0. (3)
According to (3) society loses welfare when output deviates from its target level y* and
w h en inflation deviates from its target rate which we now take to be zero for simplicity.
The quadratic form implies that large deviations of output and inllation from their respec-
tive targets cause disproportionately larger losses than small deviations. The parameter K
indicates the strength of the social preference for price stability relative to output stability.
Note that if y 1 - y * and n 1 were stochastic variables (because of the presence of stochastic
demand and supply shocks). and if the target level of output were equal to trend output
(y * = [J) . the expected value of the social loss in (3) would be equal to the social loss fi.mc-
tion (1) introduced in Chapter 20 (with n * = 0). In the present setting without stochastic
shocks. there is no need to distinguish between the actual and the expected values of
y 1 - y* and n - n*. so Eq. (3) is just the deterministic analogue of our previous social loss
function (l) from Chapter 20.
1. Remember from Chapter 20 that w hen policy is discretionary, policy m >kers do not follow a fixed policy rule like the
Taylor rule. Instead, they can adjust their instrument s in any way they believe will serve the goals of stabilization policy
in a particular situation.
CD I Sorensen-Whitta-Jacobsen: I Part 6- The Short-run 22. The limits to © The McGraw-Hill
time lags
664 PART 6: THE SHORT-RUN MODEL FOR THE C LOSED ECONOMY
Deviating from Chapter 20, we now make the important assumption that the
trend level of output is lower than the 'etllcient' level of output which policy makers are
targeting, because of imperfections in labour and product markets:
y* = Jj + W, (J) > 0. (4)
The parameter w reflects the magnitude of the distortions in labour and product markets.
The less competitive these markets are. the lower is the normal rate of utilization of eco-
nomic resources, and the greater is the difl'erence between trend output y and the efflcient
(desired) output level y*.
From (1) it follows that !it = y + n ,- n~.t-1' so according to (4) we have !it- y* =
nt - n~.1 _ 1 - w . Substituting this into ( 3), we get:
SL, = (:n:, - n~.t- 1 - w) 2 + Kn?. (5)
Now suppose for a moment that the central bank follows a Taylor rule with a zero inflation
target:
(6)
The public is assumed to have rational expectations and to know the policy rule (6).
Recalling that v , = s, = 0. it then follows from Eqs (34) and (3 5) in the previous chapter
that the economy will end up in the following equilibrium:
Equilibrium lll'!der the Taylor rule: n t = n~. 1_ 1 = n * = 0, y, = Jj . (7)
But would the central bank actually want to stick to the Taylor rule? To investigate
this, suppose the central bank were to deviate from the rule by generating surprise in-
flation after the private sector has formed its expectations n;·,~_ 1 = 0. How would social
welfare be allected by such a policy of 'cheating'? Calculating the derivative of the social
loss function (5) \.vith respect to the inflation rate at the initial point where n, = n~.1 _ 1 = 0 ,
we find:
dSL/dn, = - 2w < 0. (8)
Starting from the Taylor rule equilibrium (or any other equilibrium where n , = n~.r- 1 = 0).
the social loss can thus be reduced if the central bank decides to drive output closer to its
desired level y * by generating surprise inflation. The reason is that if n 1 = n~.r- 1 = 0 so that
y , = y. Eqs (3) and (4) imply that the marginal social cost of a slight rise in inflation is
zero, whereas the marginal social benefit from a slight rise in output is positive. Hence a
centrnl bank which can engage in discretionary policy will not want to stick to a policy ntle that
ger1erates price stability. Indeed, for any given expected intlation rate, the central bank \.vill
want to set the actual inflation rate such that the social loss function (5) is minimized. The
first-order condition for the optimal choice of the rate of inflation is dSLJdn, = 0 which
implies:
marginal
• !J1 - !t • marginal incre.ase ln SL
reduclion in .\1. due to due to higher
higher output inllation
2(n, - n~.l-l - w) +~= 0. (9)

Sorensen- Whitta- Jacobsen: I Part 6- The Short-run 22. The limits to © The McGraw-Hill
time lags
If the central bank has led the public to believe that it Vlrill ensure price stability (.n:~. 1 _ 1 - 0),
it follows from (9) and (1) that the economy will actually end up in the
w w
'Cheaiing' outcome with surprise in}Jation: n 1 = --, lJ 1 = i j + - - . (10)
l+l< ' ' l+K
If SL R is the social loss incurred in the Taylor rule equilibrium (7), and SL c is the social loss
in the 'cheating' outcome (10), we can use (3), (4), (7) and (10) to calculate the social
welfare gain from cheating:
=SLR - SLc = -
wl
Temptation to cheat -. (11)
1+K
Equation (11) is intuitively appealing: the greater the difference between natural output
and desired output. w , the greater is the temptation to create surprise inflation in order to
raise output above the natural rate. On the other hand, the stronger the social aversion to
inflation, K., the smaller is the gain from surprise inflation.
The important point is that the policy maker bas no incentive to actually implement
the policy .n: 1= :n:* = 0 if he can mal<e the private sector believe that he will do so. In other
words. over time the policy maker will not want to act in a manner consistent with the
rule he previously announced. Economists therefore say th at a rule-based equilibrium like
(7) with zero inflation is dynamically inconsistent or time-inconsistent when policy makers
have discretion. 1
Time-consistent monetary policy
Of course, rational agents who know the preferences of policy mal<ers will realize that the
central bank will not really want to implement the policy :n: , = 0. This is the credibility
problem: if the central bank cannot make a binding commitment to stick to the Taylor rule
or some other rule ensuring price stability, the announcement that the bank intends to
follow a policy of price stability will not be credible and will hence fail to eliminate expec-
tations of inflation. This is because private agents know that the central bank will h ave an
incentive to deviate from price stability in an ell'ort to stimulate output and employment.
More precisely, rational agents will recognize that the central bank will set the interest
rate so as to achieve the inflation rate implied by the llrst-order condition (9). Thus neither
the rule-based equilibrium (7) nor the 'cheating' outcome (10) will be realized, since
these outcomes are not true rational-expectations equilibria when agents know that
policy makers have discretion and seek to minimize the social loss function (5). Instead
rational agents will form their expectations on the basis of (9), implying that
:n:~.t-l- n~.,_ 1 - w + K:rt~. 1_ 1 = 0 , or .n~. 1 _ 1 = w/K. Inserting this solution lor the expected
inflation rate into (9) and using (1), we obtain the
Time-consistent rational expectntions equilibrium:

w
n l = n~·.t-1 = - , (12)
K
2. The problem of time inconsistency was first analysed by the Nobel prize w inners Finn E. Kydland and Edward C.
Prescott, · ~ul es Rat her than Discretion: The Inconsistency of Optimal Plans', Journal of Political Economy, 88, 1977,
pp. 86 7-896. The problem was later elaborated in another famous paper by Robert J. S arro and David B. Gordon,
'A Positive Theory of Monetary Policy in a Nat ural Rat e Model', Journal of Political Economy, 91 , 1983, pp. 589-6 10.
CD I Sorensen-Whitta-Jacobsen: I Part 6- The Short-run 22. The limits to © The McGraw-Hill
time lags
666 PART 6: THE SHORT-RUN MO DEL FO R THE C LOSED ECONOMY
The equilibrium in (12) is said to be time consistent because policy makers have no in-
centive to deviate ex post from the inllation rate n 1= w/K. given that this rate of price
increase is derived from the frrst-order condition (9). In the time-consistent equilibrium
expectations ofinllation have driven the actual inflation rate up to a level which is so high
that policy makers do not wish to generate further (surprise) inflation, even though they
have the ability to do so through discretionary policy.
The results in (12) illustrate the unfortunate implications of the credibility problem
under discretionary monetary policy. Though the realized inllation rate permanently
exceeds the target inllation rate by the amount w/K. inllation is fully anticipated. so no
output gains are obtained in return for the excess inllation . Clearly this outcome is
worse than the outcome (7) which would emerge if the central bank could make a binding
commitment to stick to a policy of price stability. Let SL 0 denote the social loss in the time-
consistent equilibrium with discretionary policy. Inserting (4) and (12) into (3). we Hnd
that:
-
social less due soc i al lo~
tc: due to
ineflkk-:ntly intlation
SL 0 = -.,
low output
w- +
w·
/C
7
(13)
The second term on the right-hand side of (13) could be eliminated if the central bank
could somehow commit itseli'to a rule of price stability (to check this, you should use (3).
(4) and (7) to show that SL R = w 2 ) . This term therefore represents the welfare loss from the
inability to commit. We see that the presence of market distortions (w) and the resulting
temptation to generate surprise inflation in order to boost output creates an injlation bias
under discretionary monetary policy.
The points made above are illustrated graphically in Fig. 22.1 . The concentric ovals
are social indillerence curves showing alternative combinations of output and inllation
which generate a constant social loss. The equation for the indillerence curve correspond-
ing to the social loss Cis found by setting the expression on the right-hand side of (3) equal
to the constant C. 3 The minimum social loss ( = 0) is achieved at the 'bliss' pointE* where
y = y* and n = 0 . Larger ovals further away from E* correspond to higher levels of social
loss. The flrsl-besL optimum E* catmol be:: allaiuc::U. iu au equilibrium wuerc:: expeclalious
are fullllled, since it follows from (1) that y = y when n ' = n (to put it another way, mone-
tary policy cannot eliminate the market imperfections rellected in w = y*- y). The second-
best optimum ER is the equilibrium which emerges when the central bank can mal<e a
binding commitment to a policy of price stability. In that case the private sector will ratio-
nally expect stable prices, and the economy's short-run aggregate supply curve will be
given by the curve SRAS(n' = 0) corresponding to a zero expected rate of inllation.
However, if the central bank can engage in discretionary policy so that the policy rule
3. Along a social indifference curve we thus have dSL = 0. According to (3) this implies
tbr y*- y
(y -y*) · dy+ K.• · dn=O <=>- = - -
dy ""
Thus the slope of the indifference curves becomes 0 w hen y = y", whereas the slope t ends to infinity w hen inflation
tends to 0, as illustrated in Fig. 22. 1.
time lags
LRAS
(l)
SRAS (n• = /()
(l)
(J) /(
1 + /(
Figure 22.1: Monetary policy in the Barre-Gordon model: the rule·based equilibrium (En}. the
'cheating' outcome (Eel. and the time-consistent equilibrium (E0 ).
:rc = 0 is not truly binding, it will have an incentive to create surprise inflation to move the
economy from pointER to the 'cheating' outcome Ec where the SRAS curve is tangent to
a social indillerence curve. Point Ec represents the lowest possible level of social loss. given
a zero expected inflation rate. The trouble is that rational private agents anticipate the
central bank's incentive to cheat, so according to (12) they will expect an inllation rate
equal to w/K under discretionary policy. The actual short-run aggregate supply curve will
then be given by the curve SRAS(.n' = w/K) in Fig. 22.1. and the economy will end up in
the time-consistent rational expectations equilibrium E1y This is a third-best optimum
where the social loss is minimized given the private sector's positive expected inflation
rate.
The model of inflation summarized in Fig. 2 2.1 is often referred to as the BarTo-
Gordon model, named after its inventors (see the reference in Footnote 2). It has had a
strong influence on the way economists think about monetary policy. and it helps to
explain why many economists have come to favour binding policy rules over discre-
tionary policy. In particular, the Barro-Gordon model has motivated economists and
policy makers to think about ways of securing commitment to policy rules in order to
overcome the problem of inllation bias in monetary policy.
22 •2 Dealing with inflation bias: reputation-building and

~~~-~g.':l~~~-~--~.f..~~-~-~~~.r.Y.P~.~-~~Y.........................................................................................
Building a reputation
One situation where a rule-based policy of price stability may be sustainable is w hen
polic:y m:JkP-rs m ly on mputntinn. ThP- :Jn:Jiysis in th P. prl\vions sP.c:tion impli r.itly :JssmnNl
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
668 PART 6: THE SHORT-R UN MODEL FOR THE C LOSED ECONOMY
that policy makers arc short-sighted, caring only about the economic outcome in the
current period. In that case it is always rational for them to 'cheat' by creating surprise
inflation if the initial inflation rate is (close to) zero. Hence rational agents will never con-
sider an announced policy of price stability to be credible if they believe policy makers to be
myopic. However, policy makers who interact with the private sector period after period
will have an incentive to consider the implications of their current actions for the future
behaviour of private agents. For example, if the creation of surprise inflation implies that
the policy maker \>\Jill face higher expected inflation rates in the future, he may prefer to
stick to an announced policy of price stability. In this way he will earn a reputation for
being a reliable protector of monetary stability. and this will keep future expected and
actual inflation rates down. 4
To illustrate how such a mechanism of reputation-building might work, suppose that
the public believes the announcements of the central bank as long as the bank does not
generate any unanticipated inflation. Thus, if there were no inflation surprises last period
(n,_ 1 = n:~-u- 2), the central bank has credibility in the current period. Monetary policy
makers can use such credibility to eliminate expectations of inflation by announcing that
thP.y will follow il polic:y m iP. P.ns11ring iln infliltion riltP..nn P.C}ll:ll to 7.P.rn. WP. thP.n hilvP.:
n:~.r- 1 = nR = 0 (14)
If the central bank 'cheats' in some period t - 1 by deviating from price stability, it loses its
credibility lor the subsequent period t. The public will then form its expectations for period
ton the assumption that the central bank will pursue the optimal discretionary policy
yielding the inflation rate w/K derived in (12) above. Hence we have:
(J)
n~.t- 1 = nv = - (15)
K
where :n: 0 is the optimal inflation rate under discretion. Knm>Ving that this is the public's
expected inflation rate, the best thing the central bank can do in period t is to announce
and implement the policy n t> = w/K. In this way it regains credibility in the next period by
carrying out its atmounced plan in the current period. We may say that the public is
playing 'tit-for-tat' against the policy maker. If the policy maker behaves well by sticking
lo his promises, he is 'rew;u·deu' by :a::ro expected iullalion iu Lhe nexl periou, enabling
him to avoid the social loss associated with inflation. If the policy maker 'misbehaves' by
creating surprise inflation, he is 'punished' by high expectations of inflation in the next
period.
Suppose we start out in a period in which the policy maker has inherited credibility
from the past. The policy maker must then decide whether to stick to the policy n R = 0
generating the current-period social loss SLR associated with the rule-based equilibrium
(7). or whether to cheat in order to reduce the current-period social loss to the lower level
SL c associated with the 'cheating' outcome (10) . If the poliqr maker decides to cheat, the
net social gain in the current period will be SLR - SLc At the same time the cheating will
4. This idea was developed in another influential paper by Robert J. Sarro and David B. Gordon, 'Rules, Discretion and
Reputation in a Model of Monetary Policy', Journal of Monetary Economics, 12, 1983, pp. 101- 121. The simplified
version of the dynamic Barro·Gordon model presented below is heavily inspired by Ben J. Heijdra and Frederick van
der Ploeq, Foundations of Modern Macroeconomics, Oxford University Press, 2002, Section 10.1.3.
time lags
22 TH E LIMITS TO STABI LIZ ATION POLICY 669
geuerale a uel soda! loss of SL a - SL R in lue next period where the economy will eud up in
the third-best equilibrium (12). whereas it could have ended up in the second-best equi-
librium (7) if the policy maker had not cheated in the current period. If the policy maker
has a positive rate of time preference p, he will discount next period's social loss when
comparing it to this period's social gain . Instead of Eq. (11) which is relevant only for a
myopic policy maker. we then get the following modiHed expression for the
next-period
current-period loss from
gain from cheating
cheating ~
~ (SLv - SLR)
Temptation to cheat: SLR - SLc - . (16)
1+p
Our previous expression (11) lor the temptation to cheat is just the special case of
(16) occurring when the policy maker is very short-sighted. In that case his rate of time
preference p appoaches inllnity so that the second term on the right-hand side of (16)
vanishes.
If the expression in (16) is positive, the policy maker will always want to cheat. The
mle-based policy of price stability will then be unsustainable, and the economy w ill end up
in the third-best time-consistent equilibrium in every period. just like before. But if (1 6) is
negative, the policy maker has no incentive to deviate from price stability. and the rule-
based second-best equilibrium \>Viii then be sustainable and implemented every period.
Inserting (3), ( 4), (7), (10) and (12) into (16), we lind that the temptation to cheat may be
\>Vritten as:
w 2 (pK. - 1)
(17)
x(1 + x)(1+p) '
Equation (17) shows that the policy maker will not want to cheat if his discount rate p is
sufficiently low. that is, if he cares sufficiently about the future. With a low discount rate
the short-term gain from cheating will be outweighed by the future social loss from the
inllation that follows from the Joss of credibility. We see from (17) that a low value of the
inllation aversion parameter K also helps to increase the likelihood that the policy maker
will not want to ch eat. Th e reason is t hat a low value of K generates a high rate ofin!lation
in the third-best equilibrium emerging when the policy maker has lost credibility. Hence a
low value of K implies a high value of SL v (see (13)) which makes the policy maker more
eager to avoid a loss of credibility. Finally. it follows from (17) that the magnitude of the
market distortions (w) does not inlluence the sig11 of the expression lor the temptation to
cheat. since a higher value of w increases next period's loss as well as this period's gain
from cheating.
Delegation of monetary policy
The insight from the preceding analysis is that the in flation bias in monetary policy may
be eliminated by the policy maker's incentive to build a reputation, provided he places
sulllcient weight on the future. But casual observation suggests that governments are
often very preoccupied \>\lith the short tenn. perhaps because they are mainly concerned
22. The limits to
© The McGraw-Hill
Companies. 2005
Macroeconomics Economy Credibility, uncertai nty and
time lags
670 PA RT 6: THE SH ORT-RUN MODEL FO R THE C LOSED ECONOMY
about '<\ltnnlng the next election. 5 In other words, governments often seem to act as If they
put little weight on the more distant future, indicating a high discount rate. For this
reason many economists doubt that the incentive for reputation-building '-vill be suf-
ficiently strong to eliminate the inflation bias if monetary policy is controlled directly by
the government. As an alternative way of otlsetting the inflation bias, it has therefore been
suggested that politicians should delegate monetary policy to an independent central bank
with a strong mandate to resist inflation.
In practice. central banks can be more or less independent of the government. In
Table 22 .1 the degree of central bank independence (CBI) is measured in three main
dimensions. Personnel independence reflects the degree to which government of11cials are
represented on the governing board of the central bank: the extent to which board
members are appointed by the government; the length of the term of of11ce of the gover-
nor(s), etc. The jinnncial independence indicator focuses on the stringency of limitations on
the ability of the government to borrow from the central bank, and the policy independence
indicator measures the degree to which the central bank can set its policy instruments
(instrument independence) and its policy goals (goal independence) without h aving to take
instructions from the government. The upper part of Table 22.1 shows the values of the
various indicators assigned to a central bank which is deemed to be fully independent. The
weights h ave been normalized such that the maximum total independence score which
can be assigned to a central bank is unity. The column headings refers to the authors of
four ditlerent studies of CBI (there are several other such studies, but the ones shown here
are quite representative), and the lower part of the table reports the estimated total degree
ofCBI in various countries. All authors seem to agree that the European Central Bank and
the German Bundesbank are very independent, and that the US Federal Reserve Bank also
has quite a high degree of independence. The central banks of Canada, Japan and the UK
seem to occupy a middle ground, although there is somewhat less agreement on the
ranking of these banks (which is not surprising. since the interpretation of central bank
statutes and legislation leaves room for judgement).
We will now show how the delegation of monetary policy to a (more or less) indepen-
dent and 'conservative' central bank may help to reduce the inflation bias. 6 Suppose the
governor of the central bank considers the loss from instability of output and prices to be
given by the loss function:
e > 0, (18)
where e measures the degree to which the inflation aversion of the central banker exceeds
the inflation aversion of the government. We may say that the parameter e is a measure of
the central banker's degree of 'conservativeness' (h is 'excess' aversion to inflation. taking
the preferences of the government as the benchmark).
5. O f course, if voters actually care about the (more distant) future, politic ians seeking re·election should also support
policies which place a reasonable weight on the future. If actual polic ies nevertheless seem to be biased towards the
short run, there must be some kind of imperfections in t he political process w hereby voter preferences get translat ed
into political outcomes. Such imperfecti ons are one of the subjects of the theory of Political Economy which we shall
have to leave for another course.
6. This idea was oiginally put forward by Kenneth Rogoff, 'The Optimal Degree of Commitment to an Intermediate
Monetary Target', Quarterly Journal of Economics, 100. 1985. pp. 11 69- 1 189.
22. The limits to
© The McGraw-Hill
Companies. 2005
lCD
time lags
Table 22.1; Indices of central bank independence

Grilli, Masciandaro Eijtfinger and
Measure Alesina and Tabellini Schaling Cukierman
M aximum total sco re 1
Personnel independence 0.5 0.375 0.4 0.2
Financial independence 0.25 0.3125 0.5
Policy independence 0.25 0.3125 0.4 0.3
of which
Instrument independence 0.1875 0.15
Goal independence 0.125 0.15
Germany 0.8125 1 0.66
Canada 0.5 0.6875 0.2 0.46
Japan 0.75 0.375 0.6 0.16
United Kingdom 0.5 0.375 0.4 0.31
United States 0.75 0.75 0.6 0.51
European Central Bank 0.875 0.94
Source: A dapted f rom Table 2.2 in Sylvester C.W. Eijlfinger and Jakob de Haan, European Monetary and Fiscal Policy, O xford University Press,
2000. The various indices have been normalized so that the maximum total seo<e equals 1 (lull central bank independence).
Let {3 be an index of CBI which can assume a value between 0 (no independence) and
1 (full independence). The parameter {3 measures the degree to which the government has
delegated monetary policy to the central bank. The greater the value of{3, the greater is the
weight of the central banker's preferences in the determination of monetary policy. Since
the government's loss !unction is still given by (3), we may then assume that monetary
policy is determined by minimization of the modil1ed loss function:
0 ~ {3 ~ 1, (19)
reflecting a compromise between the government and the central bank. The loss function
(19) is similar to (3) except that the in flation aversion parameter K has been replaced by
K + {3c. All the results from the previous section therefore carry over if we just substitute
K + {h forK. From (12) we then obtain the time-consistent rational expectations equilibrium
with delegation of monetary policy:
w
J(t = Jl~.t-1 = - - - ' (20)
K +{3£
A comparison of (1 2) and (20) reveals that by delegating monetary policy authority to a
central/Jallker who is nwre conservative than itselj; the govemment call reduce the inflation /Jias
in discretionary mo11etary policy. Indeed. by appointing a very conservative central banker
who cares only about price stability (t ~ oo) , the government can move from the third-
best equilibrium (12) with positive inflation to the second-best equilibrium (7) with zero
inflation. The reason is that a policy maker who is extremely averse to in flation has no
incentive to cheat, since cheating involves the creation of inflation. The paradox is that,
by voluntarily tying its own hands and delegating monetary policy to a policy maker
whose preferences deviate from its own, the government can achieve an outcome which
is mon~ nl'$i rFlhl~ from its own vi ~wpoi n t. To s~~ this. Wi\ mFly in s~rt th ~ r~su l ts from (20)
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
672 PA RT 6: TH E SH ORT-RUN MODEL FO R THE C LOSED ECONO MY
into the government's loss function (3) and usc (4) to get the social Joss with delegation of
monet.ary policy:
2
KW
a ), .
1
SL = w- + (21)
(K + p£. -
Clearly this loss is smaller when policy is (partly) delegated to a conservative central banker
({3t > 0) than when monetary policy only reflects the government's own preferences
(/Jt.; = 0).
Note that two conditions are required to obtain this beneficial outcome:
1. The central bank must have some independence ({3 > 0).
2. The central banker must be more conservative than the government (t > 0) .
If the central bank has no independence at all, the preferences of its governor are irrele-
vant, since they will be completely overruled by the government. And if the central bank
has the same preferences as the government. it will obviously make no difference whether
policy is made by the bank or by the government itself.
This analysis suggests that, other things equal, we should expect to observe a lower
average rate of inflation in countries with a higher degree of central bank independence. 7
A number of empirical studies have indeed found a clear negative correlation between CBI
14
~ 12
(')
['-
E
0>
Q)
10
u
c
·c;;
Q) 8
~ 8
c
0 6
·~
~ 4
OLS: y = 11.23 - 10.03x, Ffl = 0.38
Q)
~ _{se.....3.45)._
Q)
2
~
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7
Index of central bank independence
Figure 22.2: Central bank independence and inflation
Source: The inde>. of central bank independence was constructed by Alex Cukierman, Central Bank Strategy,
Credibility and Independence: Theory and Evidence, Cambridge, MA and London, MIT Press, 1995. Inflation is
measured by the official consumer price index.
7. The usual caveat 'other things equal' is important here, since the government of a country w ith a more independent
central bank might want to appoint a less conservative central bank governor when there is a trade·off between fight·
ing inflation and minimizing the variability of output. Below we w ill explain the circumstances in which such a trade·
off w ill arise.
time lags
and inflation, using alternative indices of CBI. Figure 22 .2 summarizes the findings from
one such study.
Influenced by this experience, many countries have moved towards a much higher
degree of central bank independence over the last decade. For example, the constitutional
framework for the European Monetary Union aims to secure a maximal degree of inde-
pendence of the European Central Bank. In specifying the mandates for their central
banks. governments in recent years have also been keen to stress that central bankers
should act 'conservatively' by pursuing price stability (defined as a low rate of inflation) as
their main goal.
Credibility versus flexibility
So far our analysis might seem to indicate that the greater the degree of central bank in-
dependence and/or the stronger the conservatism of the central bank, the better is the
macroeconomic outcome. Unfortunately things are not that simple when we allow for
the possibility of aggregate supply shocks. To demonstrate this. let us replace the simple
ilP.tP.rministic: AS c.nrvP. (1) hy thP. mnrP. rP.:cJ iist ic. supply r.nrvP.:
(22)
where s, is a stochastic supply shock variable \>\lith zero mean and constant variance 0';.
We continue to assume that monetary policy is (partly) delegated to an independent and
conservative central bank. so monetary policy is still determined by minimization of the
modified loss function (19). Using (22) to eliminate y, from (19) and recalling that
y* = y + w, we may rewrite (19) as:
SL = (n,- n;·,_1 + s,- w) 2 + (K + fh)n~ . (2 3)
Assuming that policy makers cannot credibly commit to a rule of price stability, they \>Viii
set the inflation rate so as to minimize the loss function (23), given the private sector's
expected rate of inflation. and given the current-period supply shocks, which we assume
to be observable at the time monetary policy has to be decided. Thus the Hrst-order condi-
tion for the solution to the policy makers' problem under discretion is:
d.SL/dn, =O = 2(.n, - n~.1_ 1 + s, - w) + 2(K + ,8t).n1 = 0

+ s, + W
Jt~.t- I
~ n,= . (24)
1 + K +,Be
Rational private agents know that policy makers will set the inflation rate in accordance
\>\lith (24). so the expected inflation rate is found by taking expected values on both sides of
(24), remembering that E[s,] = 0:
Jl~.t- 1 +w w
.n:~.t- I ~ Jl~.t- 1 = - (25)
1 + K +,Be K+ ,BE:
Substituting (25) into (24), we get:
w s,
.n:, = - - + . (26)
K+,Bt 1 + K +,Bt
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
and inserting (25) and (26) into (22) we find:
y,= y - ( K+{Je
' 1 + K + {Je
)s,. (27)
We may now substitute (26) and (2 7) into the government's loss function SL = (y 1 - y*) 2
+ Kni to obtain (using y * = y + w):
2
SL = " + j3e ) s +w ) + x: (- -w+ - - - -s,' - -)2
(( 1 + K+ {Je 1 K + {Je 1 + x: + j3e
2
=
K + {Je ) s,2 + w 2 + 2w ( x: + {Je )S
( 1 + K+ {Je 1 + x: + {Je
1
( 2-Kw-)( s, )
2
+K- w- + ) x:si + (28)
( K - {Je (1 + x: + j3e) 2 K + {Je 1 + K+ {Je ·
To lind the average loss experienced by the government. we calculate the mean value of the
expression in (28), using E[s,] = 0 and E[sn We then get: =a;.
1
E[SL]=w·, [ 1 + K ] + a,2[ ( K + {Jt ) + K , ]. (29)
(~< +(h:) 2
· 1 + x: +fJe (1 + x: + fte)-
The value of (J;; will be higher the greater the degree of CBI {/3) and the stronger the con-
servativeness of the central banker (e). We may say that(Je measures the 'eflective degree
of central bank conservativeness'. If there are no supply shocks. that is. if 0, we see a; =
from (29) that a higher ellective degree of central bank conservativeness will a lways
reduce the government's loss. This is in line with our previous analysis. However, in the
presence of supply shocks ((1; > 0) a higher value of j3e will not necessarily reduce the
expected social loss. Taking the partial derivative of the expression in (29) with respect to
{Je, we lind after some manipulations that:
oE[SL]
---= -
w2K
+ .
f3w;
(30)
o({Je) (K + j3e) 3 (1 + K+ {Je) 3 •
If the variance of the aggregate supply shocks is high (if a;
is large), this expression may
well be positive. implying that more CDI and/or a more conservative central bank will
actually increase the government's social loss from instability of output and prices. To
understand why, note from (26) that changes in s, will only be allowed to allect prices to
a limited degree when j3e is high. Hence the supply shocks must be absorbed mainly by
changes in output (you can verify from (2 7) that the coellicient on s1 in the solution for
output will indeed be larger the larger the value ofj3t). If the government is not too con-
cerned about price stability so that K is relatively small, it will incur a welfare loss if the
central bank pursues price stability at the expense of output stability.
Thus governments are faced with a trade-otr. They may reduce the inflation bias by
delegating monetary policy to an independent central bank which is given a mandate to
pursue price stability. But if the central bank adheres rigidly to its goal of price stability.
the result may be greater instability of output in periods when supply shocks are im-
portant. Recent experience in the Euro area suggests that this dilenuna is not just a theo-
retical possibility. In Fig. 22 .3 we show the evolution of oil prices, consumer prices and
Sorensen-Whitta- Jacobsen: I Part 6- The Short-run 22. The limits to © The McGraw-Hill
time lags
22 THE LIM ITS TO STABI LI ZA TION POLI C Y 675
3.5 120
,,,"-------...........
3 ,"' ............ 100
~ '
Q)
0>
2.5
'' Q)
0>
c ' 80
c
"'
.s::::
{) "'
.s::::
{)
Q) 2 '' Q)
0>
60 0>
c"' 1.5
' ', c"'
Q) ',', Q)
{)
Q;
Crude oil prices
40 e
',,~right axis)
Q)
a. a.
'iii '' 'iii
::J
c '' 20 ::J
c
c
<(
0.5 ' c
<(
0 0
- 0.5 - 20
2001 2002
Year
Figure 22.3: Demand, inflation and oil prices in the EMU countries
Note: The price of crude oil is shown with a one·year lag. The oil price change is a four·quarter moving average of
the percentage changes in the price of Brent Blend.
Source: ECB Monthly Bulletins.
real domestic demand in the EMU cow1tries at the start of the present decade. The rate of
oil price inflation has been smoothed (by taking four-quarter moving averages) and
plotted with a one-year time lag. because it takes time for oil price changes to feed into
consumer prices and nominal wage rates. Thus the oil price ch ange plotted for 2001 and
2002 actually retlects the development of the oil price in 2000 and 2001, respectively.
Figure 22.3 indicates th at the Euro area (which is a net importer of oil) was h it by a nega-
tive supply shock in 2001 in the form of a sharp increase in the price of oil. In part because
of this. the European Central Bank was unable to keep the inflation rate in the Euro area
below the level of 2 per cent deemed consistent with 'price stability'. despite the fact that
domestic aggregate demand started to falL Many observers thought that the ECB should
have lowered its interest rate more significantly to stimulate aggregate demand in the face
of recession. but this would probably have driven the inflation rate further above the
target level. As a newly established central bank, the ECB was keen to build up a reputa-
tion for sticking to its prime goal of price stability and hence felt unable to pursue a more
expansionary interest rate policy.
We may say that policy makers face an unpleasant trade-oil between credibility and
flexibility. To reduce the variability of output, policy makers might like to react to sh ocks
in a flexible manner by allowing a greater rise in inflation when the economy is hit by an
inflationary supply sh ock than when it is hit by an inflationary demand shock. But if the
central bank does not take a tough anti-inflationary stance regardless of the cause of a rise
in inflation. it is in danger of losing its credibility as a protector of price stability.
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
676 PART 6: TH E SHORT-RUN MODEL FOR THE C LOSED ECONOMY
22•3 Groping in the <lark: the implications of macroeconomic

measurement errors
So far our discussion of stabilization policy has implicitly assumed that policy makers have
Iilii information on the current state of the economy, that they can instantaneously react
to changes in macroeconomic conditions. and that they can fully predict how the
economy will react to any change in policy. Neither of these assumptions is very realistic.
and the present and the next section will discuss the implications of relaxing them.
Specifically. we will describe the dill1culties of stabilizing the economy when there is
uncertainty about the current state of the economy and when economic policy changes
only take full ell'ect with a time lag. In this section we focus on the former problem.
Measurement errors
In the real world macroeconomic policy makers have imperfect information about the
current state of the economy when they have to make their policy decisions. This is
because economic data typically only become available with a certain delay, and because
the data are frequently revised - sometimes substantially so - at a later stage when statis-
ticians have had time to check the numbers more carefully. Fig. 22.4 illustrates this trivial
but nevertheless important point. The figure shows the initial preliminary estimates of
real GDP growth published by Statistics Denmark compared with the final official growth
ligures in the national income accounts when all revisions to the data have been com-
pleted. Ideally, the latter time series rellects the 'true' historical rate of growth, but the
problem for policy makers is that they typically have to base their judgements and policy
4
0. 3.5
0
0
.... 3
~ 2.5
0 2
~ 1.5
~
3 1
e 0.5
....::>
Cl
c: 0 - Preliminary data
~ -0.5 - Final data 1-- - - - - - - - - - - - - - - -\-
1
1998:3 1999:2 2000:1 2000:4 2001 :3 2002:2 2003:1
Quarter
Figure 22.4: Preliminary versus final data on real GDP growth in De~mark
Note: Preliminary data are defined as the initially published estimat es of GDP growth. Rnal data are the latest esti·
mates of GDP gro.vth.
Source: Konjunktl.l'statistik, S tatistics Denmark. Authors' own calculations.

time lags
decisions on the preliminary estimates even though they know that these estimates may
be subject to substantial revision.
To estimate the current output gap policy makers not only need reliable information
on actual current output; they also face the difl'icult problem of estimating current potential
output which we have also referred to as trend output or natural output. As you recall from
Chapter 14. there are several ways of estimating potential output and they typically yield
diilerent results. leaving uncertainty regarding the true level of natural output.
The implications of measurement errors for stabilization policy
A dramatic example of the problems raised by macroeconomic measurement errors is

given in Fig. 22.5. The dotted graphs show the preliminary data which were available to
policy makers in 'real time' when they had to make their policy decisions. while the solid
graphs represent the revised data available in 2002. The upper part of Fig. 22.5 illustrates
the sharp increase in the rate of inflation in the United States in the late 1960s and in the
19 70s. The lower part of the figure gives a hint why policy makers failed to prevent this
P.pisnih~ whic:h is nnw:1clays rp,ff\rmcl to :1s thP. GrP.at Tnl1:1tinn. Assuming that th P. mnrP.
recent revised numbers are the more correct ones, we see that during the 1960s and
19 70s policy makers signillcantly overestimated the amount of slack in the economy. For
example. in 1975 monetary and fiscal policy makers believed that US output was more
than 16 per cent below its potential level, whereas the data available today indicate that
output was in fact less than 5 per cent below potential in 1975. In particular. policy
makers at the time were slow to recognize the increase in the natural unemployment rate
and the slowdown in the trend rate of productivity growth which took place in the 1970s.
For that reason they expanded monetary and llscal policy to llght the massive perceived
slack in the economy, but because potential output was considerably lower than esti-
mated at the time. the rate of intlation ended up at a much higher level than intended.
In Chapter 20 we presented evidence to suggest that the Great Inflation in the US was
due to a failure of monetary policy to follow John Taylor's prescription of raising the real
interest rate in response to a rise in inflation. In that analysis we had the benellt of hind-
sight by being able to use the revised macroeconomic data which are available today.
However. recent research has shown that monetary policy in the 19 70s actually followed
a standard Taylor rule fairly closely. given the real time data available when monetary
policy was made. 8 Thus the Great Inflation did not arise because policy makers were
unusually 'soft' or irresponsible; the problem was that the data available at the time indi-
cated a very large negative output gap which called for an expansion ary monetary policy.
What are the implications of measurement errors for the optimal monetary stabili-
zation policy? We will now use our AS-AD model to analyse this question. We start by
speciljring the actual output gap and the actual inflation gap as:
(31)
(32)
8. This is document ed by Athanasios Orphanides, 'Historical Monetary Policy Analysis and the Taylor Rule', Journal of
Monetary Economics, 50, 2003, pp. 983-1022.
0 1Sorensen- Whitta- Jacobsen:
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
nflation
14 I! q I!
I i I; I i
13 I i lj I j
12 - - Inflation: 2002
11 - - Inflation: Real-Time
10
9
8
'$ 7
6
5
4
3
2
1
0
1951 1955 1959 1963 1967 1971 1975 1979 1983 1987 1991 1995 1999 2003
Output Gap
8
7
6
5
4
3
2
1
0
-1
-2
-3
"" -4
<$' - 5
-6
-7
-8
-9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 1 7+->T+rrr+rrrrrrrrnrrlrn,;oh,.~~T. . .,.,r+rrrrrrrrrrrr+
1951 1955 1959 1963 1967 1971 1975 1979 1983 1gB7 991 1995 1999 ~003
Figure 22.5: Real-time versus retrospective views of the US economy

Reprinted from Journal of Monetary Economics, Vol 50, Athanasios Orphanides, ' Historical Monetary Policy
Analysis', pp. 983- 1022, 2003 with permission from Elsevier.
Suppose now that the estimated output gap y~ and the estimated inflation gap it~ deviate
from the respective true gaps by some random measurement errors,u, and c 1:
E[uJ = 0, E[Jtn =a;. (33)
E[c,] = 0, E[ci) = (1~ (3 4)

Sorensen- Whitta- Jacobsen: I Part 6- The Short-run 22. The limits to © The McGraw-Hill
time lags
The variableft 1 captures errors in the measurement of actual as well as potential output,
while the variable c, represents errors in the measurement of the actual inflation rate.
The constant variances rJ1~ and oi reflect the degree of uncertainty in measurement.
Operating in 'real time', the central bank must base its interest rate policy on the estimated
rather than the actual output and inflation gaps. Hence we assume that the central bank
follows a Taylor rule of the form: 9
(3 5)
In Chapter 14 we saw that demand shocks seem to be the main driver of business cycles in
most countries. To keep the exposition simple. we will therefore abstract from aggregate
supply shocks. As we have seen earlier, if the private sector has rational expectations and
the central bank has full credibility, the expected inflation rate will be equal to the central
bank's inflation target n *. We may then speci(y the aggregate supply cunre as:
(36)
Allowing lor demand shocks z,. the goods market equilibrium condition may be written
as:
g, = z, - air,- r), E[z,] = 0 . (37)
Equations (3 5)-(3 7) constitute a complete AS-AD model which may be solved to give the
following expression for the output gap (you may want to derive this result as an exercise):
(38)
This expression shows that a negative bias in the measurement of the output gap (ft 1 < 0)
or an underestimation of the inflation rate ( c 1 < 0) will tend to generate a positive actual
output gap (and hence a positive inflation gap) by inducing the central bank to set a lower
interest rate. thereby stimulating aggregate demand. As we have seen above, there was in
tact a negative bias in the estimated output gap in the United States during the Great
Inflation of the 19 70s.
Assuming that the stochastic variables z" c, and fl 1 are all uncorrelated, and recalling
lhaL lhey all uave zero meau values. il follows from (38) Lh al lhe varian ce of Lhe oulpul
gap is:
(39)
Equation ( 39) shows that errors in the measurement of macroeconomic data contribute
to macroeconomic instability because they make the central bank set the ·,.vrong' level of
interest rates which is not adequately tuned to the true state of the economy.
From (36) we see that a monetary policy minimizing the variance of the output gap
will also minimize the variance of the inflation gap in the present model without aggregate
9. If Jl, reflects errors in the measurement of potential out put, these errors w ill generally cause the central bank to mis·
measure the equilibrium real interest rate r. For simplicity we abstract from this complication here. In Exercise 4 you
are invited to demonstrate that even if r is mismeasured, we still obtain the expressions for the optimal values of b
and h given in Eqs (43) and (44) below.
Cl l Sorensen- Whitta- Jacobsen:
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
supply shocks. To minimize the social loss from variability in output and inflation. the
central bank should therefore choose those values of the policy parameters hand b which
will minimize the expression in (39). The llrst-order conditions for the solution to this
optimization problem are:
(40)
(41)
from which one can show that:
(42)
Thus, the greater the uncertainty c1~ in the measurement of the output gap relative to the
uncertainty a~ in the measurement of inllation. the larger is the central bank's optimal
response to the estimated in11ation gap relative to its optimal response to the estimated
output gap. Using (41) and (42), one can also show that: w
(43)
(44)
These results are quite intuitive since they show that monetary policy should react more
cautiously to the measured output and inflation gaps the greater the variances o~ and oi.
that is. the greater the average errors in the measurement of the gaps.
As we have noted, policy makers in the United States (and elsewhere) tended to
overestimate the size of the negative output gap for several years during the 19 70s.
This suggests that measurement errors are likely to display some degree ofpersL~tence. We
may formalize this by assuming that the measurement errors f't and " 1 are given by the
autoregressive processes:
0 < p < 1, E[dn = a~. (45)
O<fJ<l. E[k n = a~. (46)
where the parameters p and fJ quantify the degree of persistence in the measurement
errors, and where d1 and k 1 are stochastic white noise variables. By successive sub-
stitutions. we find from (45) and (46) that
2 3
fit = d t + pdt- 1 + P dt-2 +P dt -3 + .. . (4 7)
E't = kt + (Jkt- 1 + () 2 k t-2 + () 3 k t-3 + .. . (48)
10. We are assuming that the values of band h implied by (43) and (44) do not lead to a violation of the non-negativity
constraint on the nominal interest rate.
time lags
22 THE LIMI TS TO STABI LI ZA TION POLICY 681
Since cl 1 a11d k 1 are while noise variables, illullows from (47) ami (48) Lhal E(,u 1] = B[ c 1] = 0,
as we assumed in (33) and (34). Recalling that 0 <p < 1 and 0 < 8 < 1, we also see from
(47) and (48) that:
(49)
ai =E[c 2
I
] =a-k1 (1 + 8 + 8 2+ 8 3+ ... ) = OJ: .
1- 8
(50)
Inserting these expressions into (43) and (44), we obtain :
a ; (l - p)
b= J ' (51)
az o;;
(52)
Quite intuitively, we see that the greater the degree of persistence in the measurement
errors (the higher the values of p and 8), the smaller are the optimal values of b and fz,
that is, the more cautiously monetary policy should respond to the measured output and
inflation gaps.
To sum up the insights from this section, measurement errors contribute to economic
instability and reduce the scope for activist stabilization policy.
22.4 Time lags

················································································································································································
So far we have assumed that policy makers can instantaneously change the monetary
and llscal policy instruments as soon as they perceive a movement of output or inflation
away from their targets. For purposes of theoretical analysis, this is an interesting bench-
mark case, but it is not very realistic. In practice, there may be considerable Jags from the
time the economy is hit by a shock to the time when the economic policy response attains
its maximum impact. This section discusses the nature and implications of such lags.
The inside lag versus the outside lag
The time lags in economic policy are often divided into the 'inside lag' and the 'outside
lag'. The inside lag is the period from the time an economic disturbance arises until the
time when a change in the economic policy instrument has been implemented. The inside
lag thus arises from delays inside the system of policy making. The outside lag is the period
from the time the policy instrument is changed until it achieves its maximum impact on
the economic target variable.
The inside lag may be further divided into a recognition lag, a decision lag and an
implementation lag. The recognition Jag arises from the fact that a change in the state ofthe
economy cannot be observed immediately, because it takes time to collect, process and
publish economic data and statistics. Thus the recognition lag is related to the problem of
measurement errors discussed above. In particular, it often takes a while before analysts
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
682 PAR T 6: THE SHORT-R UN MODEL FOR THE C LOSED ECONOMY
and policy makers recognize that the business cycle has turned around. For example, it
was not until the end of November 2001 that the Business Cycle Dating Committee of the
US National Bureau of Economic Research concluded that the American economy bad
already entered a recession in March 2001. Such a long recognition lag reflects that it
takes a while before one can be sure that an observed movement in economic activity is
not just an erratic and very short-lived disturbance. A further problem already alluded to
is that statistical time series on economic variables are often revised, as more reliable data
become available, and as statisticians flnd more time to process the data.
The decision Jag stems from the fact that it may take time to decide a change in eco-
nomic policy even after policy makers have recognized that a signiflcant change in the
state of the economy has occurred. This may be because the formal procedures of decision-
making require some time. or because a decision may involve difllcult political trade-offs.
for example. in the case of a negative supply shock which drives up inflation at the same
time as it reduces output.
Finally. even after a policy decision has been made. it may take a while to implement
the policy change if the new policy requires new administrative procedures and routines.
'l'his ml'ly r~snlt in l'ln i111I?IP.rmmtntinn1ng.
On top of these inside lags comes the outside lag which is due to the fact that the
economy does not react instantaneously to a policy change. For example. the experience
from many Western countries is that it takes about two years for a change in the interest
rate to attain its maximum impact on the rate of inflation because of delayed behavioural
responses in the private sector (to be elaborated below).
Normally the inside lag is shorter in monetary policy than in flscal policy. The
recognition lag can be expected to be roughly the same for the two types of policy, but
once the need tor a policy change is recognized, the central bank can typically decide and
implement a change in its interest rate quite quickly. By contrast. changes in taxation
and public spending are subject to parliamentary procedures and may require time-
consuming negotiations, because they may involve controversial effects on resource
allocation and income distribution. Hence the decision lag in Hscal policy may be long.
Moreover, some flscal policy changes may take time to implement if they require the
setting up of new administrative routines and institutions.
On the other hand, since changes in public consumption and investment directly
allect aggregate demand whereas monetary policy works indirectly through changes in
the rate of interest. the outside lag may be shorter in llscal policy. Nevertheless. because of
the long inside lag in the flscal policy process. many economists believe that monetary
policy is a more effective tool of stabilization policy.
Coping with the outside lag through inflation forecast targeting
While it may be possible to reduce the inside lag by investing more resources in the collec-
tion and improvement of economic data and by reforming the procedures for policy deci-
sion-making. it is much harder for policy makers to do anything about the outside lag.
since this lag is rooted in the frictions and delays in the private sector's reactions to
changes in the economic environment. However, if policy makers know the approximate
length of the outside lag, they can account for it when designing the rules for stabilization
policy. In particular, we will demonstrate how a speciflc version of the Taylor rule for
time lags
monetary policy may be an optimal way of coping with the outside lag. For convenience,
we \!\Till now assume static expectations. but the qualitative conclusions can also be sh own
to hold under forward-looking expectations. 11
Consider the following AS-AD model \1\Tith static expectations, deliberately simplified
to highlight the outside lag in the most transparent way:
Goods market equilibrium: (5 3)
SRAS curve: (54)
The time period of the model may be thought of as one year. Thus Eq. (53) assumes that
it takes one year before a change in the real interest rate attains its (full) impact on aggre-
gate demand. This is not altogether unrealistic, since it takes time for firms to make
new investment plans in response to a change in the cost of capital, and since it is time-
consuming lor firms in the investment goods sector to produce new equipment and (in
particular) new business structures.
Equation (54) further assumes that it takes a year for a ch ange in economic activity
to affect the rate ofinllation. Again, a certain delay in the impact of a change in activity on
inflation is realistic, since nominal wages and prices tend to be sticky in the short run
because of the menu costs of wage and price changes {see Chapter 1), or because workers
and firms are temporarily locked into nominal contracts.
The model (53) and (54) captures the stylized empirical fact mentioned above that it
typically takes about two years for monetary policy to achieve its full efiect on the rate of
inflation. According to the model it takes one year for a change in the interest rate to aflect
aggregate demand. and then it takes another year before the resulting change in output
and employment forces an adjustment of the pace of infl ation.
Suppose now that the government has delegated monetary policy to an independent
central bank wh ich has been given the prime mandate of ensuring a low and stable rate of
inflation at the target level :rc*. Given this one-dimensional goal, we may speci(y the
central bank's loss function for period r as:
{55)
The central bank must account for the fact that a change in the interest rate in year t \!\Till
not allect the inflation rate until year t + 2. In year t the bank must therefore choose the
nominal interest rate, i 1• so as to minimize the expected value of SL two years in the ft1ture.
Of course, the bank is also concerned about the losses from the inflation gaps emerging
after year i + 2, but since it can reset the interest rate at thestartofeach future period in the
light of any new shocks which may have occurred since the previous period. the bank only
needs to worry about the effect of i 1 on :n: 1+ 2 : the impact of monetary policy on inflation from
year t + 3 and onwards can be optimized through the choice of future interest rates. 12
1 1. The mocel and the analysis in this subsection is a simplified version of the one found in Lars E.O. Svensson,
'Inflation Forecast Targeting: Implementing and Monitoring Inflation Targets', European Economic Review, 41 ,
1997, pp. 11 11 -1 146.
12. To put it another way, by an appropriate future choice of i,. ,, i,. 2 , i,. 3 , etc., the central bank can always neutralize
any potentially undesired effect s of i, on n,. 3 , n ,. 4 , etc. Hence the bank only needs to account for the effect of i, on
11 h-2·
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
To sec hmv the current nominal interest rate affects the rate of inflation two years
later, we note from (54) that.n:,+ 1 = n , + y(y, - y) + s,+ 1 . Inserting this into (54) along with
(53). we get:
Jlf + 1 Yt + 1- fi
(56)
When the central bank sets the interest rate for the current period, it can observe the
current inflation rate and the current output gap. 1 3 The bank is also assumed to know the
structure of the economy, summarized in Eqs (53) and (54), but obviously it does not
know the magnitude of the future supply and demand shocks, s, + 1• s 1+2 and z, + 1 .
Assuming that the stochastic shocks s and z have zero means and covariances and are
independently distributed over time. it follows from (56) that a rational central bank's
forecast of the inflation rate in year t + 2, based on the information set I, available in year
t. will be:
(57)
Not surprisingly, the inflation forecast depends on the choice of the current interest rate.
To minimize the expected social loss in year t + 2, we see from (55) that the central bank
should choose i 1 so as to ensure:
(58)
Thus the central bank should follow a policy ofirl}1ation.forecast targeting: if the forecast for
the inflation rate two years ahead is higher than the inflation target, the nominal interest
rate should be raised untiln~+ Z. t = n * . If the opposite is the case, the interest rate should be
lowered until the inflation forecast is on target. To find the interest rate which will ensure
that the inflation forecast corresponds to the target inflation rate. we insert {58) into (57)
and solve for the nominal interest rate to get:
i, = r + n 1 + h(n1 - n:*) + b(y, - y), (59)
Remarkably, we see that interest rate policy should follow a version of the Taylor rule!
Indeed, our AS-AD model with an outside lag implies that tile Taylor rule is all optimal mon-
etary policy provided the coell1cients on the inflation and output gaps reflect the way these
variables affect the future inflation rate (through the parameters a 1 andy). It is interest-
ing to note that monetary policy should react to the output gap even though this variable
does not directly enter the social loss function (55). The reason is that the current output
gap is a predictor of future inflation. For example, if y,- fi rises by one unit. we see from
(56) that this will. ceteris paribus, raise n t+ 1 and n:,+ 2 by y units. But if the central bank
reacts to this by raising i 1 by l/a 2 units, it follows from (56) that this would reduce aggre-
gate demand and output in the next period byy units which in tum would reducen: 1+2 by
a similar amount. Thus, if the two-year-ahead inflation forecast is initially on target, the
central bank can keep it there by raising the interest rate by an appropriate amow1t in
response to a rise in the current output gap.
13. We now abstract from measurem ent errors to focus attention on the role of the out side Ia~.
I Part 6 -The Short-run
22. The limits to
© The McGraw-Hill
Companies. 2005
lCD
time lags
Similarly, if the current inflation rate were to rise by one percentage point. the
(forecast for the) inflation rate two years ahead would. ceteris paribus, rise by 1 + ya 2
percentage points, partly because of the direct impact of the rise inn, on inllation ex-
pectations for years t + 1 and t + 2, and partly because of the indirect impact (captured by
the term ya 2 ) through the fall in the real interest rate and the resulting increase in output
in year t + 1. If it raises the current interest rate by 1 + h = (1 + ya 2 )/ya 2 units in response
to the unit increase inn,. we see from (56) that the central bank can neutralize the
increase inn,+ 2 •
In swnmary, the outside lag implies that policy makers should target forecasts lor the
.future values of the macroeconomic variable(s) entering the social loss function, and
should react to current macroeconomic conditions with the purpose of steering the fore-
casts towards their target values. This insight is reflected in the contemporary practice of
many central banks who often justify their interest rate decisions by reference to inflation
forecasts for the next couple of years.
The difficulties of fine tuning aggregate demand
The analysis above showed that in principle there are ways of dealing with the outside lag
in stabilization policy. However, the analysis abstracted from the inside lag. and it was
based on the optimistic assumptions that policy makers know the structure of the
economy, including the exact length of the outside lag. and that there are no errors in the
measurement of current macroeconomic conditions. fn practice, there is uncertainty
about the true structure of the economy, about current business cycle conditions. and
about the length of the outside lag which may vary from one situation to another. For
these reasons lew economists nowadays believe in the possibility of'llne tuning' aggregate
demand through macroeconomic stabilization policy. We do not know exactly when and
by how much a change in monetary (or llscal) policy allects the economy. The inevitable
time lags and our imperfect understanding of the economy imply a real danger that an
overambitious stabilization policy actually ends up destabilizing the economy, because it
only takes full effect long alter the economic shock it was meant to ollset. or because its
effect turns out to be stronger than intended. Many economists therefore believe that
monetary policy, and in particular fiscal policy, should only be used actively when the
economy is hit by significant shocks which are expected to last for some time.
22.5 Has fiscal policy been stabilizing?

················································································································································································
We have so far locused on problems of monetary stabilization policy. In this llnal part of
the chapter we will consider the experience with fiscal stabilization policy.
As we have mentioned, demand shocks seem to be the dominant driver of business
cycles in most countries. In such a setting we know from our AS-AD model that the
optimal llscal policy is countercyclical. We may therefore ask whether fiscal policy has, in
fact, been countercyclical historically? If the answer is no, it may be because the problems
of time lags. measurement errors and perhaps also 'imperfections' in the political process
are so serious that it is very dilllcult to implement optimal fiscal stabilization policies.
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
686 PA RT 6: TH E SH ORT-R UN MODEL FO R THE C LOSED ECONO MY
In recent decades many economists have expressed scepticism regarding the possibil-
ity of using Hscal policy as a successful tool for countercyclical stabilization policy, given
the way parliamentary political processes tend to work. This scepticism was nurtured by
the experience of mounting public debts from the late 19 70s to the mid 1990s, as illus-
trated in Fig. 22 .6. Such an increase in public sector indebtedness was without historical
precedent in peacetime. The growing public debts indicated that Hscal discipline was very
hard to achieve. and that governments found it dilllcult to tighten Hscal policy in times of
economic expansion. In Europe, this experience was a crucial motivation for the rules on
fiscal discipline written into the Maastricht Treaty on Economic and Monetary Union in
19 9 2 and into the EU' s so-called Stability and Growth Pact in 19 9 7.
However, the graphs in Fig. 22.6 do not in themselves tell us whether fiscal policies
have in fact been procyclical or countercyclical. A countercyclical fiscal policy implies
that policy is tightened when economic activity rises. and that it is relaxed when the busi-
ness cycle turns down. A tighter Hscal policy will appear as an improvement in the public
budget balance, but we cannot measure the stance of Hscal policy by simply looking at the
actual budget balance, since the budget deficit will automatically decrease when output
:mn P-mplnyment risP., nnP. to thP. sn- r:~IIP-1'1 twfmnntir. .~tnhiliur.~. ThP. ~n tnm~ tir: st~hili7.P.rs
include the marginal rates of direct and indirect tax. and the rates of unemployment ben-
efits and social assistance benefits. These policy instruments imply that public revenue
automatically rises and that public expenditures on transfer payments automatically fall
when an economic expansion increases the tax base and reduces the number of un-
employed workers. In this way the public budget improves even if fiscal policy mal<ers
90
80
70
60
a..
0 50
0
0
,g 40
0
- - Denmark
30
---- EU1 5
20 - USA
10
0
0 0 C'< <0 0 C'< <0 0
..... C'<
.....
~
..... .....
<0
.....
00
00 00
~
00 00
00
00 (1) (1)
~
(1) (1)
00
(1) 0
(1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) 0
C'<
Figure 22.6: The evolution of public debt, 1970- 2001 1
Note: 1. General government consolidated gross debt.
Source: Eurostat.
time lags
remain passive, undertaking no active changes in any policy instrument. For fiscal policy
to be actively countercyclical, policy makers must implement an increase in some tax rate
or a cut in public spending that goes beyond the automatic fall in expenditure.
To measure such deliberate and active changes in fiscal policy, economists have
developed measures of the so-called structural primary budget deficit, defined as the deficit
which would arise in the primary public budget (the budget excluding net interest pay-
ments on the public debt) when economic activity is at its normal trend level. Thus a rise
(fall) in the structural dellcit indicates that fiscal policy has been relaxed (tightened)
through active policy decisions. If 'f, and G, are the cyclically adjusted (or 'structural')
levels of tax revenue and primary public spending. respectively, andY, is the trend level of
output (potential output). the structural primary budget dellcit as a share of potential
output (a,) is given by:
'T _ ~ - 'f,
u, - - (60)
Y,
The structural levels of revenue and spending may be estimated by correcting the actually
observed revenue and spending figures T1 and G1 in the follmving manner:
0. > 0, (61)
f3 < 0. (62)
These specifications account lor the fact that tax revenues and public expenditures will,
ceteris paribus, vary with the output gap. due to the automatic stabilizers. The revenue and
spending elasticities o. and f3 have been estimated by the OECD, using information on the
tax-transfer systems of OECD member countries.
Based on these estimates plus its estimates of the output gaps, y 1 - y,. the OECD has
produced time series for the structural budget detlcits of its member countries. Recently
economists Jordi Gali and Roberto Perotti have used these data to study whether fiscal
policies in the OECD have followed a procyclical or a countercyclical pattern. 14 For all of
thP. 'oln' ORr.n mP-mhP-r st il tP.s for w hi r.h long timP. sP.riP.s ilrP- il Vilililhh Gilli ilnn PP-mtti
estimated a regression equation of the form
(63)
where c 1 i~ a white noise random variable. b1 _ 1 is last year's stock of public debt, and
E1_ 1 [y, - jj 1] is the expected output gap for the current year, estimated on the basis of the
previous year's output gap in the domestic economy and in the trading partners of that
economy. By using the expected rather than the actual output gap as an explanatory vari-
able. one accounts for the fact that fiscal policy decisions have to be made before the
current output gap is known. Moreover. by including b1 _ 1 and a;_1 on the right-hand side,
one allows for the possibility that past debt and deficits impose a constraint on current
llscal policy (lor example, a higher inherited stock of debt may induce policy mal<ers to
14. See Jordi Gali and Roberto Perotti, ' Fiscal Policy and Monetary Integration in Europe', National Bureau of Economic
Research Working Paper 9773. June 2003.
22. The limits to
Companies. 2005
time lags
688 PA RT 6: THE SH ORT-RUN M ODEL FO R THE C LOSED ECONO MY
pursue a tighter policy in order to bring down the debt burden) . In the present context, we
are particularly interested in the sign and magnitude of the coefficient n 1 in ( 63 ). If n 1 is
negative. the country has tended to follow a countercyclical fiscal policy, as prescribed by
economic theory, whereas a positive sign of a 1 indicates thatthe country's policy has been
procyclical. involving more fiscal laxity in times of economic booms.
Table 22 .2 shows the estimated values of a 1 for a number of countries for the period
1980-91 (before the Maastricht Treaty) and for the period 1992-2002 (after the
Maastricht Treaty). Separate estimates were made lor these two sub periods because some
observers have argued that the fiscal discipline required by the Maastrich t Treaty (a
general government budget deficit no higher than 3 per cent of GDP and a ratio of public
debt to GDP no higher than 60 per cent) forced several EU countries to follow a procycli-
cal liscal policy in the 19 90s, tightening their public budgets even in periods with very low
growth . Table 22 .2 does not provide any support for this hypothesis. On the contrary. we
see that before the Maastrich t Treaty a lot more EU countries were actually following pro-
cyclical fiscal policies, as indicated by the fact that 11 out of the 14 EU countries included
in the table had positive estimated values of a 1 in the period 1980- 91, whereas only four
RTJ mP.m hP.r statP.s h:1cl positivP. v:1h1P.S of n 1 :1ftP.r thP. M:1:1strkht TrP.:1ty. Th n s. Hlthongh
Table 22.2: The response of the structural primary budget deficit to the expected output gap
Estimates of a 1 (t values in brackets)
Before Maastricht Treaty After Maastricht Treaty
(1980-1991) (1992-2002)
EU countries
Austria -0.05 (-0 .24) -0.59 (-1.17)
Belgium 0 .38 (-1 .52) -0.84 (-1.06)
Denmark - 1.40 (-2.11) -0.24 (- 0.47)
Finland 0 .23 (-0 .64) -0.35 (-1.81 )
France 0 .14 (-1.55) 0.10 (0.56)
Germany 0.41 (-3.4) 0 .32 (0.84)
Greece 0 .12 (-0 .55) 0 .35 (0.68)
Ireland 0 .26 (-1.1 ) -0.07 (- 0.23)
Italy 0 .35 (-1 .53) -0.86 (-1.24)
Netherlands 0 .29 (0 .98) -0.72 (- 1.1 1)
Portugal 0.49 (5 .24) 0.16 (0.90)
Spain 0 .10 (0 .92) -0.06 (-0.37)
Sweden -0.52 (-1.18) -1.61 (-2.39)
United Kingdom 0 .11 (1.81 ) -0.90 (-2.92)
Other OECD countries

Australia -0.19 (-0 .66) -0.13 (- 0.62)
Canada -0.15 (-0 . 75) -0.39 (- 0.76)
Japan 0 .16 (-1.57) -0.33 (- 0.85)
Norway -0.39 (-1.12) - 1.22 (- 2.66)
United States -0.04 (-0 .23) - 1.07 (- 3.53)
All countries 0 .02 -0.44

Source: Table 2 in Jordi Gali and Roberto Perotti, 'Fiscal Policy and Monetary Integration in Europe' , NBER Working Paper 9773, June 2003.
time lags
important EU countries like Germany and France have continued to pursue destabilizing
procyclical policies in recent years, it does seem that a growing number of Western
European countries have recently moved towards the kind of countercyclical fiscal policy
which would help to even out business fluctuations . 15
22.6 Summary
................................................................................................................................................................................
1. W ith rational expectations monetary policy may suffer from a credibility problem when the
socially desired output level exceeds the natural level and the central bank can act in a
discretionary manner after the private sector has formed its expectations. If monetary
policy makers announce that they w ill keep the inflation rate down to a certain target level, the
private sector may not consider such a statement to be credible when the central bank can
boost output and employment by generating unanticipated inflation. The problem is that a
ru le-based policy of price stability is not time-consistent.
2. In a time-consistent rational expectations equilibrium with discretionary monetary policy, ratio-

nal private agents anticipate the central bank's incentive to stimulate output by generating
surprise inflation. The expected and actual inflation rate then becomes so high that the central
bank does not w ish to generate additional (unanticipated) inflation. The end result is an equi-
librium w ith inflation above the target rate and an output level equal to the (suboptimally low)
natural rate. Hence monetary policy suffers from an inflation bias under discretionary policy. A
socially superior outcome could be achieved if the central bank could credibly commit to pur-
suing a ru le-based policy of price stability.
3. If the policy maker cares sufficiently about the future, he may not have an incentive to gener-
ate surprise inflation in the short run, because he realizes that this will cause an unfavourable
upward shift in the expected inflation rate in the longer run. In such a case the inflation bias in
monetary policy may be held in check by the policy maker's desire to build up a reputation as
a defender of price stability. The inflation b ias may also be reduced by delegating monetary
policy to an independent and 'conservative' central bank which puts more emphasis on price
stability than the government itself.
4. Errors in the measurement of macroeconomic data such as the output and inflation gaps
imply that monetary policy makers sometimes set the 'wrong' level of interest rates which is
not adequately tuned to the true state of the economy. In this way measurement errors con-
tribute to macroeconomic instability. The greater the average magnitude of the measurement
errors, and the higher the degree of persistence in these errors, the smaller is the optimal
interest rate response of the central bank to changes in the estimated output and inflation
gaps.
5. In practice, it is also a serious problem for macroeconomic stabilization policy that there may
be considerable lags from the time the economy is hit by a shock to the time when the
15. One shoJid not exaggerate the importance of the econometric results in Table 22.2, since the magnitudes of the t·
statistics in the brackets indicate that many of the estimated a 1 coefficients are not significantly different from 0 (the
absolute 1-value has to be almost 2 for the estimated coeffici ent to be statistically significant at the usual 5 per cent
significance level).
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
economic policy response attains its maximum impact. In the meantime the state of the
economy may have changed so that the policy change turns out to be inappropriate. The
inside lag is the period from the time an economic disturbance arises until the time when a
change in the economic policy instrument has been implemented. The outside lag is the
period from the time the policy instrument is changed until it achieves its maximum impact on
the economy.
6. Typically it takes about two years before a change in the interest rate attains its full impact on
the rate of inflation. To cope w ith th is outside lag, a central bank wh ich has been given an
inflation target should aim at a forecast for the inflation rate two years ahead. If the inflation
forecast is above the target inflation rate, the interest rate sho·Jid be raised until the inflation
forecast is on target, and vice versa. One can show that such inflation forecast targeting leads
to a version of the Taylor rule for interest rate policy where the coeffic ients on the inflation gap
and on the output gap reflect the way these variables affect the future inflation rate.
7. Many economists believe that it is easier to stabilize the economy through monetary rather
than fiscal policy because monetary policy is generally subject to much shorter inside lags.
The empirical evidence shows that in many Western European countries fiscal policy has
in fact tended to be procyclical in recent decades, thereby amplifying the business cycle. At
the same time there are indications that several countries have shifted from destabilizing
procyclical to stabilizing countercyclical fiscal polic ies during the last decade.
22.7 Exercises
Exercise 1. Time-consistent monetary policy
When we analysed the credibility problem under d iscretionary monetary policy, we assumed
for simp licity that the policy maker's target rate of inflation (n*) was zero. Now we assume
that, instead of being given by Eq. (3) in the main text, the policy maker's loss function is:
n* > 0. (64)
1. Discuss reasons why it may be reasonable to have a positive target rate of inflation (recall the
arguments made in Chapter 21 ).
Suppose that there are no stochastic demand and supply shocks so that the economy may
be described by the following equations:
Aggregate supply: nt=n~t- 1 + Yt - Ji, (65)
Goods market equilibrium: Yt - Ji = -a2(rt- i'), (66)
Announced monetary policy rule: ft = r + h(nt- n *) + b(y t - Ji), (67)
Target output: y* = ji + w, w > 0, (68)
2. Explain briefly why it may be reasonable to assume that target output exceeds trend output,
as we have done in Eq. (68).
I Part 6 -The Short-run
22. The limits to
© The McGraw-Hill
Companies. 2005
lCD
time lags
3. What levels of output and inflation will emerge if private agents have rational expectations and
if they believe that the central bank will stick to the announced monetary policy rule (67)? W ill
the monetary policy maker actually want to stick to this ru le if he can act in a discretionary
manner? Explain.
4. Derive the 'cheating' solutions for output and inflation if the public believes that the central
bank follows the rule (67) whereas the bank actually follows the optimal d iscretionary policy.
Compare your results w ith Eq. (10) in the main text and eY.plain the d ifference.
5. Explain the concept of time consistency. Derive the time-consistent rational expectations
equilibriurr values for output and inflation. Compare your resu lts w ith Eq. (12) in the main text
and explain the d ifference. Draw a diagram analogous to Fig. 22. 1 and illustrate the Taylor ru le
equilibriurr (where the private sector believes that the policy maker w ill follow the rule (67)),
the 'cheating' outcome, and the time-consistent equilibrium in the present situation, where
:rc* > 0.
Exercise 2. Inflation bias and delegation of monetary policy
When we discussed the effects of delegating monetary policy to an independent central

bank, we assumed that the central banker had the same output target y* as the government.
However, central bankers often claim that they are not targeting an output level above the
natural rate, since they know that this w ill only lead to inflation.
Against this background you are now asked to analyse the effects of delegating monetary
policy in a setting where the central bank targets trend output y, whereas elected politicians
have a more ambitious output target y* > .Ji. Thus we assume that the government's loss
function is:
}( > 0, (6g)
where
y* = y+w, (V > 0, (70)
wh ile the central bank's loss function is given by:
SLa =(y , - y) 2 + K:rc~. (7 1)
Note that the government and the central bank are assumed to have the same degree of aver-
sion to inflation (K).
The central bank has been granted a degree of independence f3 from the government.
Hence monetary policy is determined by minimization of the following loss function, repre-
senting a compromise between the preferences of politicians and the preferences of the
central bank:
SL = (1 - f3)SL + f3SL 8 , 0 ~{3~ 1. (72)
Finally, the rate of inflation is g iven by the SRAS curve:
(73)
By controlling the (real) interest rate, monetary policy can control the output gap y ,- y and
hence indirectly the rate of inflation.
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
1. Use the procedure described in Section 1 to derive the time consistent rational expectations
equilibrium emerging when monetary policy is determined by minimizing (72). Compare the
solutions for output and inflation w ith Eq. (12) in Section 1 and comment on similarities and
differences.
2. What is the socially optimal degree of independence for the central bank (the optimal value
of p)? G ive reasons for your answer. How may central bank independence be achieved in
practice?
Exercise 3. The implications of macroeconomic measurement errors
In (43) and (44) in the main text we derived the following formulae for the optimal values of the
monetary policy parameters b and h when there are errors in the measurement of the output
and inflation gaps:
(74)
(75)
Give an intuitive interpretation of these results in which you explain the role of all the para-
meters a;, a;, ai, y and a2 .
Exercise 4. Optimal stabilization policy with measurement errors
In Section 4 we noted that when policy makers err in measuring the output gap, they are also
likely to mismeasure the equilibrium ('natural') real interest rate. Suppose therefore that the
economy is described by the following system, where r• is the estimated equilibrium real inter-
est rate, y~ and :Ji~ are the estimated output and inflation gaps, and where the measurement
errors a,, It, and E" 1 are stochastic wh ite noise variables:
Aggregate supply: n,=yy,, .

:rr,"'n, *
- n' .9,"' y,- y, (76)
Goods market equilibrium: E[z,) = 0, (77)
Monetary policy: r,= r• +h ie~+ by~, (78)
Estimate of r: r•=r+ a, E[a,] = 0, E[a?l = a;, (79)
Estimate of y,: .9~ =.9, + p, E[.u,] = 0, E{,tt~] =a;, (80)
Estimate of ii 1: ic~=ic, + E", E[€,] = 0, E[€~] = a~. (81)
Equation (76) assumes that the central bank's inflation target is credible so that rational
agents have the expected inflation rate n• = n*. For s implicity, the model abstracts from
supply shocks and from persistence in the measurement errors. Note that since (76) implies
that the variance of the inflation gap (a~) is proportional to the variance of the output gap (a;),
1m; must minimize the
a; "' E[yn
an optimal policy wh ich minimizes the social loss function SL =a; +
variance of the output gap
time lags
1. Solve the model (76) - (8 1) for the output gap and show that the variability of output is given
by:
Comment on this expression and explain the effects of the measurement errors, in particular
the effect of the error in measuring the natural rate of interest.
2. Use the procedure described in Section 4 to show that the optimal values of b and h are still
g iven by Eqs (43) and (44) in the text. Try to explain why a~ does not affect the optimal values
of band h.
Exercise 5. The outside lag and inflation foreca st targeting
Consider the following model which assumes static expectations and a one-year 'outside lag'
due to a delayed effect of changes in economic activity on the rate of inflation:
Goods market equilibrium: y,- y= z,- air,- f), (82)
Real interest rate: r,= i,-n, , (83)
Aggregate supply: n,. , = n 1 + y(y,- J?) + s,. (84)
The shock variables z, and s, are assumed to be stochast c white noise. Monetary policy has
been delegated to an independent central bank which has been given the sole mandate of
maintaining low and stable inflation at the target rate n*. In each period t, the central bank
therefore sets the nominal interest rate i, with the purpose of minimizing next period's
expected social loss from inflation variability, g iven by:
(85)
To minimize (85), the central bank must engage in inflation forecast targeting, that is, it must
choose i, so as to ensure that
(86)
Here n~+ l , r is the inflation forecast which can be calculated from the model (82)- (84), using
all the information available to the central bank in period I, including knowledge of y,, n, , z,
and s,.
1. Is the time lag embodied in the SRAS curve (84) p lausible? G ive reasons for your answer.
2. Use Eqs (82)- (86) to show that the optimal monetary policy is described by the interest rate
rule:
* z, s,
i1 =7+n1 +h(n1 -n ) + - + - , (87)
(1.2 ya. 2
Compare this rule to a conventional Taylor rule and explain the various factors influencing the
optimal interest rate policy.
22. The limits to
© The McGraw-Hill
Companies. 2005
time lags
Suppose now that there are also lags in private sector behaviour on the economy's aggre-
gate demand sde so that (82) is replaced by:
(88)
This specification reflects that it takes time for changes in real income and the real interest
rate to influence aggregate demand.
3. Discuss whether it is reasonable to assume the lags embodied in (88)? Suppose that
the length of each period in the model is one year. How long does it take for a change in the
interest rate to affect the rate of inflation in the model consisting of (83), (84) and (88)? Is an
outside lag of such length plausible?
4. Show that when the demand side is described by (88), the optimal monetary policy rule is
given by a version of the Taylor rule. Explain your results and comment on the factors deter-
mining the coefficients on the output and inflation gaps in your Taylor rule, including the role
of the parameter a 1 •
Exercise 6. Explaining time lags and their implications
Explain why there may be time lags in macroeconomic stabilization policy and why such lags
reduce the ability of policy makers to stabilize the economy. Discuss whether time lags are a
more serious problem for fiscal than for monetary policy.
I Sorensen-Whitta- Jacobsen: I Part 7- The Short-run I 23. Aggregate demand and I © The McGraw-Hill
Introducing Advanced Model for the Open aggregate supply in the Companies. 2005
Macroeconomics Ee:nnnmv nnP.n P.r.nnnmv
••
•
•
Part
he Short-run
;Model for the
•
•
~ Open Economy
•
•
•
••
•
•
23 Aggregate demand and aggregate supply in the open

economy
······································································································································································-
24 .T~~. ?.P..~.~.~~.?.~?..~Y..~~~~..~~~~..~~.~~~.~J~~~..~.~~~.~....................................
25 .T~~. ?.P..~.~.~~.?.~?..~.Y...~~~~..~~~~~.~~..~~.~.~.~~.~~..~~~~.~............................
26 The choice of exchange rate regime and the theory of
optimum currency areas
........................................................................................................................................................................
.!.\.P.P..~~~.~~:. ~~~.~~. ~~w.:~~~.~.?.~..~~~~.Y..~~.~................................................................
695
Cl l Sorensen- Whitta- Jacobsen:
Model for the Open
23. Aggregate demand and
aggregate supply in the
© The McGraw-Hill
Companies. 2005
Macroeconomics Economy open economy
Sorensen-Whitta- Jacobsen: I Part 7- The Short-run 23. Aggregate demand and © The McGraw-Hill
Chapter •
•
•
•
~ Aggregate demand
•
·and aggregate supply
in the open economy
n Part 6 we focused on the closed economy. Understanding the workings of the closed
I economy before moving on to study the open economy is useful because most of
the key mechanisms in the closed economy are also present in the open economy.
However, in some important respects openness to international trade in goods and capital
does change the way the macroeconomy works. For this reason openness may signill-
cantly all'ect the scope lor and the ellects of macroeconomic policy. It is therefore time now
to put the spotlight on the open economy.
One important insight which will emerge from our study of the open economy is that
the short-run macroeconomic dynamics and the effects of monetary and fiscal policy will
depend on the exchange rate regime. In Chapter 24 we will consider a fixed exchange rate
regime where the central bank intervenes in the foreign exchange market to keep the
nominal exchange rate fixed. Chapter 25 will study a llexible exchange rate regime where
the exchange rate is determined by the forces of supply and demand in the foreign
exchange market.
In the present chapter we derive some important economic relationships prevailing
under both exchange rate regimes. The analysis will allow for the possibility that the
exchange rate may change over time, as it typically does under llexible exchange rates
(outside long-run equilibrium). Ifthe exchange rate is actually fixed, one can simply set
the nominal exchange rate equal to a constant in all the relationships presented in this
chapter \<\lith out invalidating any of our conclusions.
We will start by documenting the trend towards increased international economic inte-
gration and by laying out our key assumptions regarding the open economy. Section 2 then
considers the implications of openness for the formation of nominal and real interest rates.
Follm<Ving this, Section 3 explains how international trade and capital mobility allects the
economy's aggregate demand curve. and Section 4 discusses the modelling oft he aggregate
supply side in the open economy. Section 5 confronts the aggregate supply curve with the
aggregate demand curve to characterize the long-nm equilibrium in the open economy.
The analysis in this chapter will set the stage for the next two chapters where we \<Viii
use our AS-AD model lor the open economy to highlight the characteristics of alternative
exchange rate regimes, as a prelude to the discussion ofthe choice of exchange rate regime
in Chapter 26.
697
Model for the Open
© The McGraw-Hill
Companies. 2005
698 PAR T 7: THE S H ORT-R UN MO DEL FOR TH E O PEN ECO NOMY
23 .1 T~.~.~. ~.!!~.~~~. ~P..~.~.~.~~~.~.~..~~~.~.~.~~~x. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..

The internationalization of the economy
During the last hall~century there has been a tremendous increase in cross-border
economic transactions. In the llrst decades after the Second World War this process of
international economic integration mainly took the form of an increase in the volume of
international rrade. Thus. while the Western countries liberalized their international
trade regimes by reducing tariffs and eliminating quantitative restrictions on imports,
they maintained substantial restrictions on the private export and import of capital. One
motivation for this policy was that capital controls made it easier lor governments to
defend the llxed exchange rate parities under the so-called Bretton Woods system of llxed
exchange rates established after the Second World War. Another motivation was th at
capital controls were necessary to implement the regulations of borrowing and lending
which were seen by most governments as an essential part of their monetary policies.
However, during the 1970s the Bretton Woods system of fixed exchange rates broke
down. and quantitative restrictions on international capital flows and on domestic credit
came to be seen as increasingly inellective and harmful to economic efllciency. In the
1980s and 1990s a large number of countries therefore abolished their capital controls.
At the same time the rapid improvements in communication and information tech-
nologies significantly reduced the transactions costs associated with international invest-
ment. As a result, the last two decades of the twentieth century witnessed a truly dramatic
increase in international capital mobility.
Table 23.1 gives an impression of the strong increase in foreign trade relative to total
output since 1960. Figure 23.1 shows a time series for the swn of total foreign assets and
liabilities relative to total foreign trade for an aggregate of industrial countries for which
data are available. The fact that this ratio has increased markedly indicates that inter-
national capital flows h ave grown even laster than foreign trade in recent years.
Some key assumptions
In this chapter we w ill show h ow our AS AD framework can be extended to a llow for
international trade in goods and capital. Specillcally, we will explain how one can model
aggregate demand and aggregate supply in an open economy with free capital mobility.
Our analysis will be based on three key assumptions. The first one is th at the domestic
economy is so small that it cannot significantly affect macroeconomic conditions in the
rest of the world. For example. if a purely domestic recession strikes our small open
economy. this will not allect its export market since foreign economic activity will remain
the same. The reason is that if the domestic economy is small. its imports are only a very
small fraction of total foreign output. so even if' imports fall due to the domestic recession,
this fall in demand will hardly be felt by the rest of the world. For most countries except the
very large ones like the United States. the assumption that the domestic economy is small
relative to the world economy is a reasonable ilrst approximation.
Our second key assumption is that the small domestic economy is specialized in the
sense that the goods produced domestically are imperfect substitutes lor the goods produced
abroad. This means that the price of domestic goods can vary relative to the price of
23 AGGREGAT E DEMAND AND AGGREGA TE SU PPLY IN T HE O PEN ECONOMY 699
Table 23.1; Ratio of foreign trade to GOP, 1960-2000

1960 1970 1980 1990 2000
Australia 0 .14 0.14 0 .17 0.17 0.22
Austria 0 .24 0.30 0.37 0.39 0.51
Belgium 0.39 0.50 0 .58 0.70 0.84
Canada 0.21 0 .27 0 .26 0.43
Denmark 0.34 0.29 0.33 0.33 0.41
Finland 0 .22 0.25 0.32 0.24 0.38
France 0.15 0 .22 0 .22 0.28
Germany 0.21 0 .27 0 .30 0.34
Greece 0 .14 0.13 0 .26 0.23 0.30
Ireland 0.33 0.39 0 .53 0.55 0.91
Italy 0 .13 0.16 0 .23 0 .20 0.28
Japan 0 .10 0.10 0 .14 0 .10 0.10
Luxembourg 0 .85 0.87 0 .94 1.02 1.41
Netherlands 0.46 0.45 0 .52 0.53 0 .65
Norway 0.37 0.37 0.40 0 .37 0.38
Portugal 0 .18 0.24 0.30 0.36 0.37
Spain 0 .08 0.13 0 .16 0 .18 0.31
Sweden 0 .22 0.24 0 .29 0 .29 0.43
Switzerland 0 .26 0.30 0.34 0.34 0 .41
United Kingdom 0 .21 0.22 0 .26 0.25 0 .29
United States 0 .05 0.05 0 .10 0.10 0 .13
Note : Foreign trade is measured as the a·1erage of exports and imports.
Source : OECD Economic Out look database.
6.0 __/
5.5
5.0
4.5
4.0
...
.. /
_/
"'
---
/
----
.&
3.5
/"
3.0 •
2.5
2.0
1.5
1.0
0.5
0.0
1985 1989 1993 1997 2001
Year
Figure 23.1: Foreign assets and liabilities relative to foreign trade

Note: The C·)Untries included in the sample are: US , UK, Austria, Bel gium, Germany, Italy, Net herlands, Norway,
Sweden, Switzerland, Canada, Japan, Finland and Spain.
Source: Figure 2.4 of Philip R. lane and Gian Maria M ilesi-Ferretti, ' International Financ ial Integrat ion', /MF Staff
Papers, 50, Special Issue, 2003, pp. 82-113.
Model for the Open
© The McGraw-Hill
Companies. 2005
foreign goods. Indeed. as we shall sec in the next two chapters, the endogenous adjust-
ment of the relative price of domestic goods is a basic mechanism through wh ich the small
specialized economy adjusts to a long-run macroeconomic equilibrium. 1 The manufac-
tured goods which make up the bull< of exports in most industrialized countries are typi-
cally somewhat dillerentiated from the manufactured goods produced in other countries.
Hence our assumption that countries specialize in the production of dillerent goods also
seems plausible.
Rellecting the deep international integration of capital markets prevailing today. our
third key assumption is that international capital mobility is perfect. You may recall from
Chapter 4 thac under perfect capital mobility domestic and foreign llnancial assets are
perfect substitutes. and llnancial investors can instantaneously and costlessly switch
between domestic and foreign assets. For simplicity. we \>\Jill also assume that investors are
risk neutral. arranging their portfolios so as to maximize their expected rates of return
without worrying about the variance of the rates of return. 2
The next section derives some important implications from our key assumptions for
the formation of nominal and real interest rates in the open economy.
Capital mobility, interest rate parity and purchasing power

23•2 parity
................................................................................................................................................................................
Capital mobility and nominal interest rate parity
When capital mobility is perfect and investors are risk neutral, it follows that foreign and
domestic assets must yield the same expected rate of return. For example. if domestic
bonds had a lower expected return than foreign bonds. investors would immediately sell
domestic bonds in order to buy foreign bonds, thus driving down domestic bond prices and
pushing up foreign bond prices until the expected rates of return are equalized. Given that
this arbitrage can occur instantaneously and costlessly, the expected returns must be
equal at any point in time.
Let us be more precise. Following common European practice. we dellne the nominal
e:rchange rate. E, as the number of domestic currency units needed to buy one unit of the
foreign currency. If i is the domestic nominal interest rate and jl is the foreign nominal
interest rate, perfect capital mobility implies the arbitrage condition:
1 + i = (1 + / ) E' )
(
f . (1)
where E is the nominal exchange rate at the start of the current period, and 1 is the E:
nominal exchange rate expected to prevail at the beginning of the next period. The
1. The assumption that foreign and domestic goods are imperfect substitutes distinguishes our open economy AS·AD
model from the open economy growth model set up in Chapter 4. In that chapter we made the simplifying assump·
tion that d omestic goods were perfect substitutes for foreign goods.
2. Alternatively, we might assume that the domestic interest rate includes an exogenous risk premium (positive or
negative), reflecting risk aversion coupled w ith country-specific investment risks. Working with an exogenous risk
premium. as we did in Section 4 of Chapter 4. would not c hange the properties of our model in any important ways.
Sorensen-Whitta-Jacobsen: I Part 7- The Short-run 23. Aggregate demand and © The McGraw-Hill
23 AG G REGATE DEMAN D AN D AG GREG ATE SU PPLY IN T HE O PEN ECONOM Y 701
left-hand side of (1) measures the amount of wealth accruing to an investor at the end of
the current period if he invests one unit of the domestic currency in the domestic capital
market at the beginning of the period. As an alternative to such a domestic investment,
the investor could have bought 1/E units of the foreign currency at the start of the period
lor the purpose of investment in the foreign capital market. At the end of the period he
would then have ended up with an amount of wealth (1/E)(1 +if) in foreign currency. At
the start of the period when the investment is made, the investor expects that this end-of-
period wealth will be worth (E ~ 1 /E)(1 +if) units of the domestic currency. Thus equation
(1) says that domestic and foreign investment must generate the same expected end-of-
period wealth and hence must yield the same expected rate of return. 3
If we take natural logarithms on both sides of (1) and use the approximation
ln(1 + x)"' x, we get:
e =lnE. (2)
The magnitude (e~ 1 - e) x 100 is the expected percentage rate of depreciation of the
domestic currency against the foreign currency. This is the expected capital gain on foreign
bonds relative to domestic bonds over the period considered. Thus Eq. (2) says that if
the domestic currency is expected to depreciate by x per cent, the domestic nominal inter-
est rate must exceed the foreign nominal interest rate by x percentage points to make
domestic and foreign assets equally attractive.
Equation (2) (and its approximate equivalent (1)) is known as the condition for wzcov-
ered interest rate parity. The term 'uncovered' refers to the fact that the investor h as not
covered his risk: when he invests in the foreign capital market, he expects a capital gain
< 1 - e, but this gain is uncertain, so he is exposed to risk. If he wants to cover his risk at the
time of investment. he can use the forward market for foreign exchange. In this market he
can sell an amount of foreign currency (1/E)(1 +if) for delivery one period from now at
the forward exchange rate E+1 currently prevailing in the market. The one-period forward
exchange rateE+1 is the domestic-currency price of one unit of foreign currency delivered
one period from now. Since£+ 1 is known at the start ofthe current period when the foreign
investment is made. the investor knows lor sure that he \.viii end up with (E+1/E)(1 +if)
units of the domestic currency at the end of the period if he covers his foreign investment
via the forward market. For domestic and foreign investment to be equally attractive, we
thus have the arbitrage condition:
1 + i = (1 + /)(£~ 1 ). (3)
Tal<ing logs on both sides of (3 ), we get the approximate relationship:
i = / + e+ l -e, (4)
Equations (3) and (4) are known as the condition for covered interest rate parity. Is it possi-
ble for covered and uncovered interest rate parity to hold at one and the same time?
According to (2) and (4) the answer is 'yes', provided
(5)
3. As already mentioned, ( 1) also covers the case where exchange rates are completely fixed, as indeed they are if the
countries considered belong to a monetary union w ith a common currency. In t hat case we simply have E~ ,/E = 1.
Model for the Open
© The McGraw-Hill
Companies. 2005
702 PART 7: THE SHORT-RUN MODEL FOR THE OPEN ECONOM Y
The term on the left-hand side of the first equation in ( S) is the fonvard fort:ign e:~o:chmzgc
premium, defined as the (percentage) dillerence between the price offorward exchange (the
price of foreign exchange lor delivery one period from now) and the current spot exchange
rate E (the price of foreign exchange for immediate delivery). Equation ( 5) says that if the
spot exchange rate is expected to increase by (e~ 1 - e) x 100 per cent over the next period.
then the forward foreign-exchange premium should also equal (e~ 1 - e) x 100 per cent.
When investors are risk neutral, this condition must hold. To see this. suppose that
<
e+1 < 1 . In that case it would be possible to score an expected profit by buying foreign
exchange in the forward market today and then selling the foreign currency in the spot
market when it is delivered one period from now. Similarly, ife+1 > investors will want e:,
to sell foreign exchange in the forward market today and buy foreign exchange in the spot
market one period ahead when the foreign currency is to be delivered, thus scoring an
expected profit of e+ 1 - e:
1 at the time of delivery. To rule out such arbitrage opportunities
for pure prollt making, Eq. (5) v.rill have to be satisfied, so covered and uncovered interest
rate parity will hold simultaneously when investors are risk neLitral.
Equations (2) and ( 4) suggest that when exchange rates can vary a lot, the interest
riltP.s of cliflP.rP.nt c:mmtriP.s c::c~n :c~ l so clP.vi:c~tP. snhst:c~nti:c~lly from P.:c~c:h othP.r. HowP.vP.r, whP.n
countries move towards greater llxity of exchange rates, the expected exchange rate
changes should tend towards zero. forcing national interest rates into equality.
Figure 23 .2 conllrms this hypothesis. Around the mid 1990s the exchange rates within
the European Monetary System could in principle vary within an exchange rate band of
±15 per cent around the central parity. and countries like Italy and Spain with a history
of devaluations against the German mark had relatively high nominal interest rates.
reflecting the perceived probability of devaluation of the Italian and Spanish currencies.
At the same time countries like Germany. Netherlands and France. whose currencies
were perceived to be strong, had nominal interest rates below the average EU level.
- - France
- Germany
- Italy
- - - Netherla nds
- Spain
- 2 +-----~~~~~~------------------------~
- 3+-ro-..-,-,.-,-,,-.,,.-,,-,-,.-,,-,-,.-1
vvvv~~~~~~~~~~~~oooooooom~mmo
mmmmmmmmmmmmmmmmmmmmm~mmo
~~~o~~~b~~~o~~~o~~~o~~~o~
ooo~ooo~ooo~ooo~ooo~ooo~ o
Figure 23.2: Nominal interest rate differentials in Europe, 1994-2001

Note: Difference between domestic ten·year government bond yield and the corresponding EU average.
Source: IMF International Financial Statistics.
23 AG G R EGATE D EMAN D AN D AG GREG A TE SU PP LY IN T H E O PEN EC O NOM Y 703
However. when financial markets became convinced that political leaders in Europe were
determined to establish a monetary union with completely fJXed exchange rates (and
ultimately a common currency), the nominal interest rates of the different EU countries
quickly converged, as fears of substantial exchange rate movements vanished.
Figure 23.3 tells a similar story about Denmark. From the early 19 70s to the early
1980s, the Danish krone was devalued against the German mark on numerous occasions,
and expectations of future devaluation kept the Danish nominal interest rate far above the
German interest rate. But as the policy of systematic Danish devaluations was abandoned
after 19 82 and the krone became a much more stable currency, the Danish interest rate
gradually converged towards the German level.
We have so far focused on the implications of capital mobility for the cross-country
relationship between nomina/ interest rates. but the aggregate demand for goods and ser-
vices depends on the real interest rate. defined as the nominal interest rate minus the
(expected) rate ofinflation. To see what international economic integration implies for the
cross-cow1try link between real interest rates, we must therefore explore the link between
national inflation rates created by international trade. This is the agenda for the next
sn hsP.r.ti on.
The real exchange rate, relative purchasing power parity and real interest
rate parity
A country's international competitiveness is often measured by the price of foreign goods

relative to the price of domestic goods.4 This important variable is called the real exchange
-
- Danish-German nominal interest rate differential (average bond yield)
- Percentage depreciation of the Danish krone against the German D·mark
14
12
IJ1v ~\ fl
) /, I
I-- I I A
IJW Y$\
"\\II '\(\$7'1
J
.f I~ I ?\ "\ /\~
I If ~
-2
v 1\f
I
'V
V~t--1
-4
-------- ---- -
"'.... ....«t ....<D ....co co co
0
co "'co «t
co
<D
co co
0
Ol "'
Ol
«t
Ol
<D
Ol Ol
0
0
Ol Ol Ol Ol Ol Ol Ol Ol Ol Ol Ol Ol Ol
-
Ol 0
"'
Year
-
Figure 23.3: Interest rate differential and currency devaluation in Denmark, 1972- 2000
Source: M ONA database, Danmarks Nationalbank.
4. St rictly speaking, internat ional competitiveness should be measured by t he relative prices of t hose goods which can
be traded int ernationally. In our AS·AD model we w ill make the si mplifyin~ assumption that all ~oo d s can be traded.
8 I Sorensen-Whitta- Jacobsen: I Part 7- The Short-run
Introducing Advanced Model for the Open
© The McGraw-Hill
Companies. 2005
704 PAR T 7: THE SHORT-R UN MODEL FOR THE OPEN ECONOMY
rate. E•, dellned as

EPf
E'=-
- p . (6)
where pf is the price of foreign goods denominated in foreign currency, EPf is the price of
foreign goods measured in domestic currency, and Pis the price of domestic goods in units
of the domestic currency. The real exchange rate indicates the number of units of the
domestic good which must be given up to acquire one unit of the foreign good. The higher
the real exchange rate. the cheaper are domestic goods relative to goods produced abroad.
The inverse of the real exchange rate (1/E'J is referred to as the international terms of
trade, since an increase in the real exchange rate implies a deterioration in the terms on
which domestic goods can be traded for foreign goods.
For later purposes it will be convenient to measure the real exchange rate in natural
logarithms. We therefore dellne:
e' =In E' = e + pf- p. pf =In pf, p =ln P. (7)
The log oflhe real exdtauge rale iulhe previou~ period is lheu given by:
(8)
which may be subtracted from (7) to give:
e' - e~ 1 = tJ.e +nf - n, (9)

tJ.e=e- e_ 1• n =p - p- 1·
where tJ. e is the percentage depreciation of the domestic currency. n f is the foreign rate of
inflation, and.n is the domestic inflation rate. According to (9) the percentage depreciation
of the real exchange rate is thus equal to the nominal exchange rate depreciation plus the
inflation difl'erential between the foreign and the domestic economy.
The dellnitional relationship (9) has an interesting implication when we combine it
with the concept of a long-run macroeconomic equilibrium. As we shall see below. the
real exchange rate is an important determinant of the trade balance. For the economy to
be in long-run equilibrium. the real exchange rate therefore has to be constant (otherwise
the trade balance would keep on changing over time). If we insert the long-run
equilibrium condition e' = e~ 1 into (9) . we get the long-ru11 equilibrium co11dition:
(10)
This condition is known as relativepurchasing power parity. It says that. over the long run.
a country's rate of nominal exchange rate depreciation must correspond to the excess of
the domestic over the foreign inflation rate. In this way the country's international com-
petitiveness (its real exchange rate) is kept constant over time. Relative purchasing power
parity (RPPP) will play an important part in our characterization oflong-run equilibrium
in the open economy.
Figure 23.4 shows that the hypothesis ofRPPP provides a reasonably good description
of the long-run relationship between price levels and the nominal exchange rate between
Denmark and her largest trading partner, Germany. For many years the Danish price
level increased relative to the German price level, but this loss of Danish competitiveness
23 AGGREGATE DEMAND AND AGGREGA TE SU PP LY IN T HE O PEN ECONOMY 705
1 50 ~--------------------------------------------,
Consumer prices in Denmark relative to

consumer prices in Germany
50+------.-------.------.------.-------.------1
1970 1975 1980 1985 1990 1995 2000
Figure 23.4: Bilateral exchange rate and relative prices between Denmark and Germany
Source: Chart 4. 1 of Monetary Policy in Denmark, Danmarks Nationalbank, 2003.
was more or less ollset by a corresponding depreciation of the exchange rate until some
time in the 1980s when Demnark switched to a fixed exchange rate policy vis-a-vis
Germany.
Combined with our assumption of perfect capital mobility, relative purchasing power
parity implies that in the long run the domestic real interest rate is tied to the real interest
rate abroad. This may be seen as follows: in a long-run equilibrium the rate of exchange
depreciation must be correctly anticipated, that is, the expected rate of depreciation must
equal the actual rate, e: 1 - e = t>e+1• Hence we may write the condition for uncovered
nominal interest rate parity as i = / + fi e+1 , and the condition for relative purchasing
power parity may be rewritten as n +1 = .n:~ 1 + t>e+ 1• Subtracting the latter from the former
equation gives:
(11)
The left-hand side of (11) is the ex post domestic real interest rate, and the right-hand side
is the foreign real interest rate measured ex post. 5 In a long-run equilibrium where infla-
tion is correctly anticipated. the ex ante real interest rates (based on expected inflation
rates) are identical to the ex post rates, so in the long run the real interest rate in the small
domestic economy will be equal to the exogenous foreign real interest rate. In other
words. while capital mobility establishes a link between the nominal interest rates at home
and abroad, the combination of capital mobility and foreign trade also implies a long-run
link between the domestic and the foreign real rates of interest. The relationship (11) is
5. The definition of the real int erest rate given here is the so·called real producer rate of interest, defined as the nominal
interest rate minus the rate of increase of the price of dom estic goods. This is the rate of int erest which det ermines
the profitability of investm ent from the viewpoint of domestic finms. Fro IT the viewpoint of consumers, the relevant real
interest rate is the so·called real consumer rate of interest, defined as the nominal interest rate minus the rate of
inc rease of consumer prices (which include the prices of imported consumer goods). By focusing on the real pro·
ducer rate of interest, we are implicitly assuming that the interest rat e affects aggregate demand mainly through its
impact on business investm ent decisions, whereas household savin~s are relatively insensitive to int erest rat es.
© The McGraw-Hill
Companies. 2005
706 PART 7 : THE SHORT-RUN MODEL FOR THE OPEN ECONOMY
UK-USA
20 .---------------------------------------------~
1 5 ~--------------~--------------------------~
1 0~~------------~-~-~--------------------------~
5 -lH.- - - - - -l
o+-¥-~~,P~~~~~~--~~~~~~r-~~--~~
-5~---4+--1 1---~-~~------V--~~~
- 1 0 ~---~---v----------------~
- 1 5 +------r----~.-----~-----T------~----~----~
1920 1930 1940 1950 1960 1970 1980
Denmark-UK
12,-----------------1-----~
10~----------------------------------n-----------1
8~----------------~
6 +-----------~~-------------------~1--4-1~----~
4 +--.-----~-~------.~~~1-~tA-~
2 +---~~----74-~4--~-~~--~----~~~+-~~--~~~
0 +-~~~~~~~~~~~~~~~-~--~~
- 2~-~---~~---~--------~~~~)~--------------~
-4~/--~~--------~L--------~
- 6 ~~------------.-------.-------.-------.-----~
1930 1940 1950 1960 1970 1980
Sweden-UK
8 ,---------------------~
6 +--------~---------lr----~
4~-----------------~~-------------------H--•------~
2 -~-----.
0
-2
- 4 +-----~-----J----1~~----~----~------~--~--~
-6 +---,----.~~+.---r---~---.-~
1922 1932 1942 1952 1962 1972 1982
Netherlands-UK
1 0 +=====~==========================~
5+-----~Y-~----~~r----r~~-----------~~----~
0~~~~-H~~-A~~H- . .~~~~~~~~~~~~
.g. - 5
- 1 0 ~-----------l------------l ~--------------------~
- 1 5 +---~----~--~-----r~--r---~----~--~----~
1900 1910 1920 1930 1940 1950 1960 1970 1980
Norway-UK
10 , - - - - - - - - - - - - - - - - - - - - - - ,
a +-------------1.-------------------------------~
6~------------f-~----------------------------~H
4~-------H~---------.-
2 ~--1\-/11-------~-1
o~~T?~-T~~~~~~--~~~~~~r-~.r~
- 2 +-~~-4.-1~~~~\~~~~~~~~-~--~-1-+~v-~
-4
-6 +---~---~------~------~
-8 +----.---.-----.------.------.-----~
1930 1940 1950 1960 1970 1980
Figure 23.5: Long -term real interest rate differentials

Source: Interest rates from S. Homer and R. Sylla, A History of Interest Rates, Rutgers University Press, 199 1.
known as rt'lll interest rate parity and it will play an important role in our analysis of the
open economy. 6
If real interest rate parity holds, the difference between the ex post real interest rates
across any two countries in the world economy should tend to fluctuate around a zero
mean value. Figure 23.5 suggests that this is indeed the case. The figure shows that long-
term real interest rate diflerentials have fluctuated quite a lot over time - indicating that
the economy can tluctuate substantially around its long-run growth trend - but the real
interest rate differential displays a clear tendency to revert towards a mean value of zero,
as implied by the hypothesis of real interest rate parity.
Note mat our derivation of (11) did not rely on any speciflc assumption regarding the
exchange rate regime. Thus real interest rate parity should hold in the long run whether
nominal exchange rates are fixed (t. e = 0) or flexible (t. e * 0). This is consistent with
Fig. 23 .5 which covers a long period during which international exchange rate regimes
have varied considerably.
23.3 Aggregate demand in the open economy

................................................................................................................................................................................
Drawing on the analysis above, we will now study the determinants of aggregate demand
in an economy which is open to trade and capital mobility.
The trade balance and the real exchange rate
In the open economy, the condition for equilibrium in the goods market is:
Y = D+ G+NX, (12)
where Y is real GDP, D is total private demand lor goods and services, G is public demand
for goods and services. and NX is the trade balance. defined as exports minus imports (net
exports). Thus net exports add to the total demand for domestically produced goods. All of
the components of demand on the right-hand side of (l2)- including the trade balance -
are measured in units of the domestic good. In other words. we are using the domestic
good as our numeraire good. If the quantity of imported foreign goods is M, the nominal
value ofin1ports measured in domestic currency units will be EPf M . Given that one unit of
the domestic good sells at the price P. the volume of imports measured in units of the
=
domestic good will then be EPf MjP E' M. If the quantity of domestic goods which is
exported abroad is denoted by X, we thus have:
NX =X- E'M. (13)
It is reasonable to assume that the volume of exports. X, depends positively on the inter-
national competitiveness of domestic producers. For example, Fig. 23.6 shows that the
market share ofDanish exporters tends to vary positively with the price charged by foreign
competitors relative to the price charged by Danish exporters. In our stylized model this
6. In Chapter 4 we directly assumed real interest rate parity. We have now provided a deeper understanding of this 'law'
for the long run.
Model for the Open
© The McGraw-Hill
Companies. 2005
708 PART 7: THE SHORT-R UN MODEL FOR THE OPEN ECONOMY
120 . ----------------------------------------------.
0
~ 90 ~~~~~--------------~~
Figure 23.6: Danish price competitiveness and export market share

Note: The market share is calculated as the total volume of Danish exports less exports to Gennany divided by a
weight ed average of the imports of the remaining 21 recipient markets. Price competitiveness is the market defla-
tor in DKK divided by the Danish export price.
Source: Chart 2 of Heino Bohn Nielsen, 'Market Shares of Manufactured Exports and Competitiveness', Danmarks
Nationalbank, Monetary Review- 2nd Quarter 1999, pp. 59-78.
price ratio is captured by the real exchange rate E' . It is also natural to assume that the
export volume depends positively on total output yf in the rest of the world, since higher
economic activity abroad leads to a larger export market for domestic producers. Hence
we may write X= X(E' , y f), where X varies positively with both E' and y r_
In Chapter 17 we saw that total private demand for goods in the closed economy
depends on domestic output Y. the total tax bill T. the real interest rate r, and the state of
confidence (the expected future growth rate) e. In the open economy all of these variables
are also likely to affect imports, since some of the goods demanded by the private sector are
imported from abroad. In addition, the quantity of imported foreign goods will depend
negatively on the real exchange rate, since a higher relative price of foreign goods will
reduce domestic consumer demand for foreign goods. partly because a rise in the price of
imported goods reduces the purchasing power of domestic nominal incomes (the income
eflect), and partly because a higher price of foreign goods induces consumers to substitute
towards domestic goods (the substitution ellect). We may therefore specify the import
function as M = M(E' . Y. T, r. e), whereM varies negatively withE". Tand r, and positively
with Vande.
With these specifications of exports and imports. it follows from (13) that net exports
are given by:
NX = X(E', y f ) - E' M(E' , Y, T, r. e), (14)

++ - +-- +
where the signs below the variables indicate the signs of the partial derivatives. We see
that a rise in Y or e will raise the level of imports. because an increase in these variables
will stimulate total private demand for goods and services. On the other hand. a rise in Tor
r will reduce imports by dampening total demand for goods.
i\ crucial question is how a change in the real exchange rate 1'1-'i ll a ffect the trade
balance. To investigate this, we calculate the partial derivative of the net export function
(14) with respect toE':
oNX ()X oM
- - = - - E' · - - M . (15)
()E' ()E' ()E'
Let us denote the initial magnitudes of X, M and E' by X0 , M 0 and E~l· respectively. As a
bench mark case we will assume that the trade balance is initially in equilibrium so that
X0 = E:;MO' We may then rewrite Eq. (15) as:
()NX (()X E~) oM E~) )

oE' = Mo ()E' Xo - ()E' Mo - 1
()NX ax w
()E' = Mo(TJx + TJ.w- 1 ), IJx =()E' X > 0, (16)
where TJ x is the elasticity of exports \'l'ith respect to the real exchange rate, and TJ ,~.~ is the
numerical elasticity of imports with respect to the real exchange rate. Equation (16)
shows that a real depreciation of the domestic currency - that is, a rise in the real exchange
rate E'- will improve the trade balance provided the sum oft he relative price elasticities of
export and import demand is greater than 1 (TJx + 1'/,,1 > 1). This important result is called
the Marshall-Lerner condition. named after its discoverers. Numerous empirical studies lor
a large number of countries suggest that the Marshall-Lerner condition is almost always
satislled. at least when the time horizon is one year or longer, so that economic agents
have h ad time to adjust to the price change. In the following. we will assume that this
condition is in fact met. 7
The aggregate demand curve in the open economy
We are now ready to derive the aggregate demand curve lor the open economy. Recalling
the factors influencing the trade balance and the private demand lor goods. and assuming
that total tax payments correspond to total public spending (T = G). we may restate the
goods market equilibrium condition (12) as:
Y = D(Y, G, r , e, E') + NX(E' , yf, Y, G. r, t:) + G

= D(Y, G, r, t: , E', yf) + G, D =D + NX. (1 7)
=
The magnitude D D + NX measures the total private demand for domestic goods ema-
nating from the domestic as well as the foreign private sector. Notice from ( 17) that the
real exchange rate E' influences D due to the income effect mentioned earlier: when the
price of imports increases relative to domestic prices, the purch asing power of domestic
money incomes is eroded, and this \'\Till. ceteris paribus. reduce total private demand lor
goods because domestic residents become poorer.
7. In the very short run 'lx and IJMmay actually be so small that the Marshaii·Lerner condition is violat ed , reflecting that
firms and households do not instantaneously adjust their patterns of demand and supply to a relative price change.
Thus the trade balance tends to follow a 'J·curve' pattern over t ime, following a real depreciation : in the very short run,
the trade balance tends to det eriorate due the worsening of the terms of trade, but as the quantities of exports and
imports start to adjust. the trade balance gradually improves.
© The McGraw-Hill
Companies. 2005
While an increase in domestic economic activity Y will stimulate total private

demand lor goods. D, it will also increase imports and hence reduce the trade balance NX.
=
Thus it would seem that the sign of the partial derivative D,, oD/oY is ambiguous.
However, since only a fraction of total demand for goods will be directed towards imports,
the rise in imports will be less than the rise in total demand lor goods. Hence we
=
haveDy iJD/iJY + iJNX/iJY = iJD/iJY - E'(oM/iJY) > 0. Maintaining our assumption from
Chapter 17 that oD/oY < 1, it then follows that:
• ()D ()NX
0 < D 1, =()Y + iJY < 1. (18)
The fact that only a fraction of total demand is directed towards imports also implies that:
- ()D ()NX
D =-+--<0. (19)
r or or
_ ()D i)NX
D, =-:;-+-=>
aE:
-
u£
> 0. (20)
Given our ass umpliou of a balauceu goverumeul b uugel, a rise iu goverumeul speu uiug G
will lead to a corresponding rise in taxation. Consumers react to this fall in disposable
income by reducing consumption, but the fall in consumption is not as great as the rise in
taxation, since private saving will also fall when the tax bill goes up. Hence we h ave
- 1 < oD/oG. Furthermore. only a part of the fall in consumption will manifest itself as a fall
in imports; the remaining part will appear as a fall in the demand lor domestic goods. From
this it follows that:
- ()D ()NX
- 1<Dc = - + - - < 0. (21 )
' oG oG
Finally, we assume that:
_ oD i)NX
DIJ =;-;:
uE
+ :;-r > 0.
uE
(22)
Recall that ()NX/ iJE r > 0 when the Marshall-Lerner condition is met. However, because of
the negative income effect of a higher real exchange rate, we have oD/ oEr < 0 . The
assumption in (22) that De > 0 thus requires that the Marshall-Lerner condition is satis-
fied '.vith a certain margin so that IJ x + IJ M is sulllciently greater than 1. Empirically this
has turned out to be a reasonable assumption (except for the very short run where the
price elasticities in export and import demand are small, due to inertia in consumer
reactions to relative price changes) .
Using the above dellnitions of and assumptions on partial derivatives. and following
the procedure lor log-linearization explained in Section 2 of Chapter 1 7, the appendix to
this chapter derives the following log-linear approximation to the goods market equilib-
rium condition (17):
(23)
The notation x indicates the initiallong-nm equilibrium value of variable x. The variables
y, yi and g are measured in natural logarithms. and all the fJ coefficients are positive.
We will now show that {23) implies a negative relationship behvccn domestic output
and the domestic inllation rate. just as we found in the closed economy. First we recall
from (9) that
(24)
Second, we note from the condition for uncovered interest parity (2) that the domestic (ex
ante) real interest rate is given by
(25)
We may choose our units of measurement such that the real exchange rate E' = 1 in the
=
initial long-run equilibrium, implying e" In Er = 0. Further. we know from the condi-
tion for real interest rate parity (11) that the domestic long-run equilibrium interest rate,
f, equals the foreign equilibrium real interest rate which we denote by rf . Inserting (24)
and (2 5) into (23) along with e' = 0 and f = pf , we then obtain the following preliminary
expression for the aggregate demand curve for the open economy:
y - [J = {J 1 (e~ 1 + 6 e + n:f- .n) - {J 2 ( / - .n~ 1 + 6 e' - rf) + i. (26)

z ={J 3 (g - g) + {J 4(!/- Tf) + {J 5 (ln £ - ln c). (27)
Equation (26) is preliminary since it does not constitute a complete theory of aggregate
demand until we h ave described the formation of inllation expectations (.n~ 1 ) and
exchange rate expectations (6e''). In the next two chapters we shall argue that these
expectations are likely to depend on the exchange rate regime, and we shall therefore
postpone the speciflcation of expectations until then. However, remembering that fJ 1 > 0,
we can already see from (26) that an increase in domestic inllation will. ceteris paribus,
reduce aggregate demand for domestic output. The reason is that higher domestic inj1ation
erodes the international competitiveness of dom estic producers by reducing the real exchange
rate. Thus, a rise in .n raises the relative price of domestic goods, thereby reducing net
exports. To see that this is indeed the mechanism, note that the term ( e~ 1 + 6 e + .nf- .n) on
the right-hand side of (26) is simply the (log of the) current real exchange rate, e'. In the
appendix we show that the response of aggregate demand to the real exchange rate
depends on the price elasticities of export and import demand (rJ x and rJ M) which are
incorporated in the parameter (3 1• The higher the value of these elasticities, the stronger is
the reaction of net exports to a change in the real exchange rate. and the llatter is the
aggregate demand curve in ( y . .n) space.
A second important observation is that the preliminary AD curve deflned by (26) will
change position from one period to the next whenever relative purchasing power parity
fails to hold. that is. whenever the economy is out oflong-run equilibrium. This follows
from (24) which shows th at the real exchange rate will change over time whenever
*
6 e .n - nf . Hence the value of the lagged real exchange rate. e~ 1 • appearing on the right-
hand side of (26) will also change from one period to the next when the rate of exchange
depreciation deviates from the inllation differential between the domestic and the foreign
economy, and this will cause the AD curve to shift even ifthe other variables on the right-
hand side of (2 6) are constant. 8 In the open economy. shifts in the AD curve are thus an
8. The fact that the AD curve w ill be shifting over time as long as 6 e * :rr- :rr 1 explains our earlier c laim that a long-run
equilibrium requires relative purchasin ~ power to hold.
Model for the Open
© The McGraw-Hill
Companies. 2005
inherent part of the adjustment to long-run equilibrium. This contrasts w ith our AS-AD
model of the closed economy where the adjustment took place solely through shifts in the
AS curve.
23.4 ~.~&.r.~~. <:l~.~.. ~. ~P.P~Y...~~ ..~.~.~ ..?.P..~P. ~~.?.~?..~.Y.............................................................. .

To complete our AS-AD model lor the open economy. we need to confront the aggregate
demand curve with the aggregate supply curve. Modelling aggregate supply in the open
economy requires some care because of the possibility that wage formation may be
influenced by changes in import prices, and hence by changes in the real exchange rate.
To illustrate this possibility. suppose that consumer preferences are given by a Cobb-
Douglas utility function defined over domestic goods and imported goods. One can then
show that the consumer price index P" will take the form :
0 < 1/J < 1, (28)

WhP.rP. t hP. nxNl prp,fp,rp,nc:P. p::!r::JmP.tP.r 1/1 is th P. sh::~rP. of importP.n goons in C:Ol1SllmP.r
budgets. and EP' is the price of these goods measured in domestic currency units. We see
from (28) th at the consumer price index will move whenever there is a change in the real
=
exchange rate E' EP1/P. If wage setters seek to achieve a certain level of the real consumer
wage, dellned as the nominal wage rate W dellated by the consumer price index P", a
change in the real exchange rate will then obviously induce wage setters to change their
nominal wage claim.
However, if wage setters focus instead on relative wages, trying to maintain some
target ratio between their own wage rate and the average wage level in the rest of the
economy, a change in the real exchange rate caused by a change in import prices will not
influence wage setting, since a movement in import prices does not disturb the existing
pattern of relative wages.
The theories of wage formation developed in Chapters 12. 13 and 18 all imply that
wages are set as a mark-up over the representative worker's 'outside option ' which
includes the rate of unemployment benellt. Whether wage setters locus mainly on real or
on relative wages th erefore depends on the rules lor indexation of nominal unemployment
benellts. If nominal benellts are indexed to consumer prices so as to keep the real unem-
ployment benent constant, it follows from (28) that a change in the real exchange rate will
allect the nominal unemployment benellt. Hence it will also affect domestic nominal
wages and prices because a higher nominal unemployment benefit will drive up nominal
wage claims and lead to higher domestic prices th rough the mark-up price setting offirms.
This means that the position of the economy's aggregate supply curve will depend on the
level of the real exch ange rate.
However, if nominal unemployment benefits are instead indexed to nominal wages.
nominal benefits will not automatically change when the real exchange rate changes. In
such a setting we shall see that it is rational lor wage setters to focus on their relative wage
position even u· they only derive utility from their own real income. When benefits are
indexed to wages, the real exchange rate will not directly inlluence the formation of
domestic wages and prices, and the aggregate supply curve lor the open economy will
have the same form as the SRAS curve for the closed economy.
We shall now substantiate this claim by deriving the aggregate supply curve for an
open economy where unemployment benefits are indexed to wages, as is the case in most
countries where benefits are either linked to the unemployed worker's own previous wage
or to the general wage level (which amounts to the same thing in our representative agent
model). We will base our analysis on the theory of elllciency wages developed in
Chapter 12. but we emphasize that the same end result can be derived from the model of
trade union wage setting presented in Chapter 13. ~
Efficiency wage setting and aggregate supply
We will consider a simplified version of the so-called 'shirking' model of efllciency wages
introduced in Chapter 12. 10 The basic idea underlying this model is that work ell'ort
generates disutility, so workers have an incentive to 'shirk' on the job (taking coffee
breaks, chatting with colleagues, using the olllce computer to surf on the internet. etc.)
instead of working hard all the time. To keep the incentive lor shirking in check, firm
managers therefore have to monitor the work effort of the employees and impose some
s:mc:tinn snc:h i!S tiring il wnrl<f~r if c:Bngh t shirking. Hmwwe.r, sinc:e. pe.rff\c:t Bnil r.onstBnt
monitoring is practically impossible or prohibitively expensive. there is always some
chance that a worker can get away with shirking \>Vithout being caught.
To simplify matters. we assume that a worker can either work all the time or shirk all
the time. If be works all the time, the expected utility level u;" of a worker employed by firm
i \>\Till be:
u;" = (;.:)(1 - 1,). 0 < fL < 1, (29)
where W; is the nominal wage rate paid by tlrm i. and pc• is the expected consumer price
level so that W;/P"'' is the expected purchasing power of the money wage ollered by firm i.
Equation (29) assumes that the worker's utility varies in proportion to his real wage. It
also assumes that the disutility of work is given by 11 · (W;/Pr'), where fL is a parameter
reflecting the worker's preference for 'leisure on the job'. The higher the worker's income,
the more highly he thus values leisure on the job. so the more income be is willing to
sacrifice in return lor the benefits from coll'ee breaks, internet surllng, etc. 11
If the worker chooses instead to shirk all the time. he faces a probability 8 of being
caught by the manager and llred lor poor work performance. In that case he must look for
a job elsewhere in the labour market. but since there is unemployment. he cannot be sure
to find one. We assume that the worker expects to be able to obtain the utility level v if he
is flred due to shirking. The variable v is referred to as the worker's 'outside option', i.e., his
fall-back position if sacked (we will discuss the determinants of v in a moment). The
probability that the worker is not caught sh irking is 1 - 8. In that situation the worker's
utility \>Viii simply be his real consumer wage W/P'. since he obviously does not incur any
9. We now choose to work with an efficiency wage model rather than a trade union model in order to illustrate that
efficiencl wage theory offers yet another possible microeconomic foundation for the familiar expectations·
augmented Phillips curve.
10. Even if you have not studied Chapter 12, you should still be able to follow the exposition below.
11. If we did not allow the valuation of leisure to rise with the worker's income level, the incentive to shirk would tend
towards zero as the real wage grows over time due to increasing productivity. Assuming that the demand for leisure
varies positively with income is equivalent to assuming that leisure is a 'normal' good.
0 I Sorensen- Whitta- Jacobsen: I Part 7- The Short-run
© The McGraw-Hill
Companies. 2005
disutility from work cfl'ort w h en he is shirking. Overall, a shirking worker's CJ.'Pcctcd

utility Uf will thus be given by:
u• =1 (w.) ev
(1 - 8) -pte' + · ' o< 8 < 1. 8 > 1t . (30)
A utility-maximizing worker will choose to shirk if Uf > U~ and will choose to work when
u: < u;•. In the borderline case where u: = U~. the worker is in principle indillerent
between working and shirking. but we assume that in this case social norms induce him
to work.
When workers choose to work, their output per person is determined exogenously by
the firm's level of technology. It is now easy to derive the cost-minimizing wage rate which
the employer will want to oller. If the wage is so low that if; > U;". workers will produce
nothing and the employer will make a loss as long as the wage rate is positive. Firing the
shirking workers will not solve the problem since newly hired workers will have an
incentive to shirk as well. assuming that they have the same preferences as those who
were flred. On the other hand. if the wage is so high that u: < u;•. the employer is paying
more th an necessary to induce work effort without obtaining a higher labour productivity
in return (since the productivity of a non-shirking worker is exogenously determined by
technology). Hence the cost-minimizing wage rate is found where the non-shirking condi-
tion U~ ~ u:
is met with equality. Setting U;" = u:
and using (29) and (30), we thus
obtain the cost-minimizing wage rate offered by Hrm i:
-----------
w.)
u;• u;
.......--.....
(w)
(pr; (1 - !t) = (1 - 8) P"; + ev
w. ( e )
P'~ = 8 - t v. (31)
1
We see that the expected consumer real wage is set as a mark-up over the representative
worker's outside option v . The mark-up factor 8/(8 - p) is positive and greater than 1.
given our assumption in (30) that 8 > p. A higher value of 8 reduces the firm's optimal
wage rate, because a higher risk of being fired for shirking strengthens the incentive to
work. thereby lowering the wage rate which is necessary to induce efl'ort. On the other
hand, the stronger the preference p. for 'leisure on the job', the higher is the wage rate
which must be paid to secure efl'ort.
Let us now specif)r the representative worker's outside option, i.e., the utility he expects
to obtain if Hred for shirking. For an average worker who is sacked, the probability that he
will be able to find a job elsewhere in the labour market is given by the employment rate
1 - u, where u is the rate of unemployment. If he succeeds in finding a job, it follows from the
analysis above that he may expectto obtain a utility level us =uw= (W'/Pr')(1 - fl), where
W' is the expected average money wage level in the labour market. On the other hand, the
worker faces a probability 11 of remaining unemployed, in which case he expects to be able
to collect a nominal unemployment benefit B' with an expected real purchasing power
B'/¥'. We may therefore specify the representative worker's outside option as follows:
v =
(W') (B').
(l - u)(1 - ft) pre + u p•·• (32)
The rate of unemployment benefit is assumed to be indexed to the general wage level, with
a constant replacement ratio c which is known to all agents. The expected benefit rate is
then given by B" = cW' . We impose the restriction that 0 < c < 1 - lt· since othenvise the
expected real wage net of the disutility of work, (W'/P~'')( 1 - lt), would be lower than the
expected real rate of unemployment benefit, cW'/P'''', so that unemployed workers would
have no incentive to look for a job.
Inserting (32) and B' = cW'' into (31). and multiplying through by P ..... we lind the
optimal nominal wage rate oll'ered by firm i:
8
wi = (- - )[(1 - u)(1 - ,.t) + cu]W' . (33)
8 - p.
We see that the (expected) consumer price level p re does not affect wage formation. 12 As
(33) makes clear, the individual employer seeks to maintain a certain relationship
between his 0\>\111 nominal wage rate and the expected general nominal wage level. A
change in consumer prices does not disturb this relationship. since it aflects workers inside
and outside each firm in the same manner. Thus our elllciency wage model implies
'relative wage resistance' in the sense th at individual firms resist a ch ange in the ratio of
their own wage to the general wage level. However, the model does not imply 'real wage
resistance', since a rise in consumer prices does not automatically lead to a rise in nominal
wages.
If we make the synun etnJ assumptio11 that expectations as well as the parameters 8 and
p are the san1e across all firms, ( 3 3) implies that all employers will set the same wage rate.
Hence we have W; = W, where Wis the general level of nominal wages. From (3 3) we then
get:
W = M '"(1 - yu)W' . (34)
To establish the link between wages and domestic prices, we now consider the repre-
sentative firm's price setting behaviour. For simplicity, suppose the production function is
linear and that one unit of labour produces a units of output, where a is the exogenous
level of average and marginal labour productivity. The representative firm's average and
marginal unit costs '"'ill
then be Tl'l /a. According to the theory of price formation developed
in Chapter 18, a monopolistically competitive firm will set its price as a mark-up. MP, over
its marginal labour cost. Since all firms pay the same wages. using the same technology
and charging the same mark-up. the price of domestically produced goods charged by the
representative firm will then be:
w
p = MP. - . (3 5)
a
Let us assume that MP is constant over time whereas labour productivity fluctuates
around some trend level a. Equation (35) then implies that, on average. the relationship
12. This result relies on our assumption that the nominal rate of unemployment benefit is indexed to the nominal wage
level. If benefits were instead indexed to the consumer price level, (expected) c hanges in P would affect the
workers' out side option and thereby influence wage setting. In reality, most countries index unemployment benefits
to waQes rather than to consumer prices.
0 I Sorensen- Whitta- Jacobsen:
Model for the Open
© The McGraw-Hill
Companies. 2005
716 PAR T 7: TH E S H ORT-R UN MO DEL FOR TH E O PEN ECO NOMY
between the general wage level and the domest ic price level is W = lfPjMP. Suppose that.
by experience. wage setters are aware of this average relationship between wages and
prices. Suppose fi1rther that, at the time when the individual linn sets its own wage and
price lor the next period, it expects the general domestic price level to be p •· during that
period. It is then reasonable to assume that the representative individual 11rm will expect
the general wage level to be: 13
lf
W' = - · P' . (36)
MP
Using (35) and (36) to eliminate Wand W '' h·om (34). taking natural logarith ms on both
sides of the resulting equation, and exploiting the approximation ln(1 - yu)"" - yu, we get:
a·
-
M 1'
w
p w = M "'(1 - vu ) -lf · P'
'
------
MP
wt
p "" p' + m ••- yu + In lf - In a, (3 7)
p = In P, p' =In P' , m•• =In Mw.

Uwe subtract p_ 1 on both sides and use the concept of the natural unemployment rate IT.
we may rewTite the price setting equation (3 7) in terms of the actual and expected domes-
tic inflation rate:
:n: = :n:'' + y(IT - u) + ln lf - In a, (38)
:n: =p - p-1' :n:' =p''- P- J· i7 =m"'/y.
We see that ( 38) is an expectations-augmented Phillips curve. From this we may derive an
aggregate supply curve. Denoting the total labour Ioree by Nand the trend level of output
by Y, we introduce the following definitions and approximations from Chapter 18:
Y =a(1 - u)N ==* y =In Y "" In a + In N- u, (39)

Y =lf(1 - fl)N ==* [I =In Y "" h1lf + In N- fl. (40)
Inserting (39) and (40) into (38), we end up with a short-run aggregate supply curve of
the form:
SRAS: n = :n:' + y(y - [J) + s. s =(1 + y)(ln a- ill a), (41)
where s is a supply shock variable capturing productivity shocks. 14 Equation ( 41) shows
that our theory of efficiency wage setting leads to an aggregate supply curve of exactly the
same form as the SRAS curve for the closed economy.
As we mentioned. our aggregate supply curve (41) is based on a theory of 'relative
wage resistance', because wage setters tend to resist changes in their relative wage
13. Equation (36) may seem a roundabout way of forecasting W (by first estimating W/ P and then forecasting P).
However, recall from Chapt er 18 that the mark·up price setting behaviour of the representative firm i was derived
from a demand curve of the form Y,= (Pj P)- (Y/ n). To forecast its own demand, the firm thus has an incentive to
form an estimate of the general price level, P, and then it may as well use this estimate to forecast the general wage
level, exploiting its knowledge of the average relationship between Wand P. Moreover, even if the firm forms a
direct estimate of W, (36) will still hold provided the firm's (separat e) expectations about Wand Pare consistent
w ith its knowledge of the average ratio of wages to prices.
14. For simplicity, we have assumed that the wage and price mark-ups m" and mPare constant. If they are fluctuating
over time, the supply shock variables would also include the cyclical component s of m" and mP.
position as long as the level of economic activity is unchanged. As you recall. this result is
based on the assumption that unemployment benefits are indexed to wages. If nominal
benefits are instead indexed to consumer prices. wage setting will be characterized by 'real
wage resistance' in the sense that wage setters \>Viii resist changes in their real consumer
wage at any given level of activity. The position of the SRAS curve \>Viii then be influenced
by the real exchange rate. as we have already explained. Here we \>Viii stick to the assump-
tion that benefits are indexed to wages, partly because this is the most common practice in
the real world, and partly because it leads to a simpler specification of aggregate supply
which is directly comparable to the SRAS curve for the closed economy. In Exercise 3 we
ask you to explore the alternative case of real wage resistance.
23.5 ~?..~.~ ~~~~..~~l~.~~!.~E~~~..~~..~?.:~. ?.P.~~..~.~?..~.?.~Y...................................................

Conditions for long-run equilibrium
As noted earlier, a complete AS-AD model which is able to describe the slwrt-nm dynamics
of the open economy must include a theory ofthe fonnation ofinflation and exchange rate
expectations, and these expectations are likely to depend on the exchange rate regime. In
the next two chapters we will study expectations formation and short-run macro-
dynamicsunder alternative exchange rate regimes. However, at this stage we can already
characterize the Jong-nm equilibrium on which the small open economy will converge if it
is stable.
We have seen earlier that the real exchange rate must be constant in a long-run
equilibrium. Otherwise the AD curve (26) would keep on shifting over time. We h ave also
seen that constancy of the real exchange rate implies relative purchasing power parity,
which leads to real interest rate parity when capital mobility is perfect. In the long run
the domestic real interest rate is thus tied to the exogenous foreign real interest rate,
regardless of the exchange rate regime:
(42)
= n::
Moreover. we know that r i- n:~ 1 = jl + t.e'- 1 . Inserting these relationships plus
(42) into (26). we obtain the open economy's long-run aggregate demand curve (LRAD) :
r y- y - z
e = /31 . (43)
z =- (J (rf- F)+ (J (g - g)+ /3/Y'- [/) + fJ 5(ln f - In f) .

2 3
The LRAD curve specifies the relationship between real output and the real exchange rate
which is consistent with long-run equilibrium in the market for domestic goods.
Unconventionally, the curve is upward-sloping in (y , er) space since a real exchange rate
depreciation (a rise in e") increases total demand lor domestic goods by improving the
international competitiveness of domestic producers. To maintain equilibrium between
the demand for and the supply of domestic goods, domestic output must therefore
increase. The exogenous variable z captures demand shocks stemming from the foreign
and the domestic economy. By construction, z is zero in the initial long-run equilibrium.
© The McGraw-Hill
Companies. 2005
718 PAR T 7: THE SH ORT-R UN MO DEL FOR TH E O PEN ECO NOMY
To determine the long-run equilibrium values of output and the real exchange rate.
we must combine the LRAD curve with the economy's long-run aggregate supply curve
(LRAS). In a long-run equilibrium inflation expectations must be confirmed (n' = n), and
productivity must be at its normal trend level (In a = In lf). Inserting these conditions into
the SRAS curve (41), we get the LRAS curve:
y = [J . (44)
This shows that the long-run aggregate supply curve is vertical in ( y, e' ) space, given the
assumption of relative wage resistance underlying the SRAS curve (41). In the long run
output will be at the natural level given by (40), corresponding to the natural rate of
unemployment IT= mwfy.
Figure 23.7 illustrates the long-run equilibrium in the open economy. Output is
determined exclusively from the supply side. and the equilibrium real exchange rate is
found where the upward-sloping LRAD curve crosses the vertical LRAS curve. A perma-
nent sh ock to aggregate demand - that is. a shift in the LRAD curve generated by a
permanent change in z - will be fully absorbed by a change in the equilibrium real
P.xr.hflngP. rfltP., IP.flving th P.Iong-rnn ontpnt lP-vP.I nnCJffP.r.tP.cl
The long-run neutrality of the exchange rate regime
Our analysis of long-run equilibrium in the open economy has one striking implication
which is worth emphasizing: the long-run equilibrium values of' real' variaiJles such as the real
interest rate, tlte real exchange rate a11d real output are all independent of the exchange rate
regime. This follows from the simple fact that we did not have to make any assumption
e'
LRAS
LRAD
Figure 23.7: Long·ru n equilibrium in the open economy

regarding the exchange rate regime when we derived the long-run equilibrium values of
r. e' andy. The analysis in Fig. 23.7 is valid regardless of whether exchange rates are fixed
or flexible.
We may therefore say that the exchange rate regime is neutral in the long run, since
it has no impact on the long-run equilibrium values of real variables. This is a parallel to
the long-run neutrality of money in the closed economy. In a closed economy setting. the
authorities can control the nominal money supply M (or at least the monetary base), but
in the long run they cannot control the real money supply M/P, because the price level is
endogenous. Similarly. in the open economy policy makers can choose to fix the nominal
exchange rate E rather than allowing it to float, but they cannot lix the long-run value of
the real exchange rate. EPf/P.
This might seem to suggest that the choice of exchange rate regime is unimportant,
but such an interpretation would be wrong. As we shall see in the next two chapters, the
mechanisms through which the economy adjusts towards its long-run equilibrium are
different under fixed and flexible exchange rates. The choice of exchange rate regime may
therefore make an important dillerence for the short-run and mediwn-rw1 dynamics of the
mHr.rn P-r.onomy. TnrlP-P-rl, WP. shi!ll SP.P- th i!t th is r.hoir.P. will clP.tP-rminP. w hP.thP.r or not mon -
etary policy can be used as a tool of short-run stabilization policy.
Moreover. although most economists agree that the exchange rate regime is at least
approximately neutral in the long run, they do not agree on how good this approximation
is. Some even argue that the choice of exchange rate regime may have important struc-
tural effects in the long run. For example, some researchers have claimed that the lower
volatility of exchange rates under a llxed exchange rate regime may significantly promote
international economic integration by reducing the degree of risk and uncertainty relat-
ing to cross-border economic transactions. In Chapter 26 we shall return to this theme
when we discuss the theory of optimum currency areas which has played a prominent
role in the economic debate on monetary union in Europe.
Some economists also question the usefulness of the sharp distinction between short-
run dynamics and long-run equilibrium on which the long-run neutrality of the
exchange rate regime is based. These critics point out that the economic system may
display 'path-dependence' or 'history-dependence' in the sense that the economy's long-
run equilibrium may depend on events occurring along th e dynamic path of adjustment
to equilibrium. For example, if a serious recession pushes a lot of workers into long periods
of unemployment. some of these workers may lose (part of) their skills during the unem-
ployment spell, and some may reduce their job search if they become discouraged by or
accustomed to unemployment. Such negative effects of unemployment on labour skills
and job search will increase the equilibrium unemployment rate. In this way the natural
rate of unemployment may depend on the history of actual unemployment, and a reces-
sion may leave a permanent scar on the economy even if it is temporary. If the economy is
characterized by such 'hysteresis', 15 the exchange rate regime may have long-term real
15. This term is borrowed from physics w here hysteresis refers to the fai ure of an object to return to its original value
aft er being changed by an external force, even after the force is removed. The theory of hyst eresis in the labour
market is controversial. It was developed in an effort to explain the persistence of high unemployment in continen·
tal Europe in recent decades. However, many economists still believe that the economy tends to return to a unique
natural unemployment rate in the very long run. This view is supporte:l by Fig. 12 .1 in Chapt er 12 which suggests
th::~t in thP. lo ng run unAmrlnymP.nt g r::~v it;:ltA!=: tow::~rrl l=\ ::1 r.on!=:t:=~n t unP.mr lnymAnt r::~t A ::~rnunrl !i-7 pAr r.P.nt.
© The McGraw-Hill
Companies. 2005
720 PAR T 7: THE SH ORT-R UN MO DEL FOR TH E O PEN ECO NOMY
ell'ccts because it affects the way in which the economy adjusts to a shock over time. For
instance. if a switch from fixed to flexible exchange rates allowed a country to pursue more
expansionary policies during recessions, such a regime shift might lower the average
long-term unemployment rate ifthe labour market exhibits hysteresis.
For these reasons the theory of the long-run neutrality of the exchange rate regime
should only be seen as a rough approximation. In Chapter 26 we will return to a discus-
sion of the long-run eilects of the choice of exchange rate regime.
Wealth effects and the balance of payments in the long run
Whenever an open economy is running a surplus (deficit) on the current account of the
balance of payments - that is. whenever it is exporting (importing) capital - it will accu-
mulate (decumulate) foreign assets. When the current account imbalance stems from
private sector transactions. it will therefore generate a change in the private sector's
aggregate net wealth. In Chapter 16 we saw that changes in private wealth are likely to
all'ect private consumption. In an economy with no long-run growth, a 'true' long-run
P.qnili hrinm won lcl th P.n rP.qnirP. t hi!t thP. rP.HI (inlli!tion-HcljnstP.cl) hH lHnc:P. on thP. r.nrrP.nt
account be zero. Otherwise real private wealth would be changing over time, inducing
changes in aggregate demand which would be incompatible with long-run equilibrium.
However, in an economy with secular growth the real balance of payments must
deviate from zero in the long run to allow real private wealth to grow in line with output
and income (see Exercise 5). Moreover, real-world experience suggests that it tal<es a very
long time before the economy fully adjusts to wealth effects stemming from current
account imbalances. The Danish experience is a good example of this. As shown in
Fig. 2 3.8, Denmark ran a persistent current accountdellcit from the late 19 50s to around
1990. The dellcit was so h igh that the ratio of foreign debt to GDP was rising throughout
most of this period. On the other hand. since 1990 Denmark has run a persistent current
3 ~----------------------------------
"'
Ul
2 +---~~-----------------------------~l--+-----~
"'
§o.
co
g0 0
co-
c ~ - 1 +-~----~-1+--------~~---------------+--------~--~
eOl
~ ~ - 2 ~-...L-----1H
§
~ ~
e
- 3 -l----------1--------'\
~ - 4+------------------------1+-4+--~4.~---------------1
"iii
m -5 -l------------------~---------4
-6+---.----.---.---.----.---.----.---.---.--~
1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 2002
Figure 23.8: The c urrent account on the Danish balance of paymen: (o/o of GOP)
Source: Statistics Denmark, Statistical Yearbook.
23 AGGREGAT E DEMAND AND AGG REGATE SU PPLY IN T HE O PEN ECONOMY 721
account surplus. The fact that the economy seems to respond very slowly to current
account imbalances in the short and medium run indicates that it may be a reasonable
simplification to ignore current account dynamics unless one wants to focus on the very
long run . To keep matters simple, our model of the open economy therefore abstracts from
wealth etlects stenuning from current account imbalances. just as we have previously
ignored changes in private wealth arising from possible imbalances on the government
budget.
23.6 Summary
................................................................................................................................................................................
1. This chapter develops a model of aggregate demand and aggregate supply for an open
economy. The economy is assumed to be so small that it cannot significantly affect macro-
economic conditions in the rest of the world. It is also assumed to be specialized in the sense
that the goods produced domestically are imperfect substitutes for the goods produced
abroad. Tt;is means that the price of domestic goods can vary relative to the price of foreign
goods. A third important assumption is that capital is perfectly mobile across borders.
2. Under perfect capital mobility the arbitrage behaviour of risk neutral investors will enforce
uncovered nominal interest rate parity. This implies that the domestic nominal interest rate will
be equal to the foreign nominal interest rate plus the expected percentage rate of deprecia-
tion of the domestic currency against the foreign currency. If a group of countries moves
towards credibly fixed exchange rates, the interest rate differentials between them will
therefore tend to vanish.
3. In the forward market for foreign exchange investors can buy or sell foreign currency for future
delivery. The arbitrage behaviour in the forward market enforces covered nominal interest rate
parity. This means that the domestic nominal interest rate equals the foreign nominal interest
rate plus the forward foreign exchange premium. The latter is defined as the percentage
difference between the forward exchange rate and the current spot market exchange rate.
Covered and uncovered interest rate parity will hold simultaneously if the forward exchange
premium equals the expected rate of change in the exchange rate over the period considered.
This equal ty must hold if investors are risk neutral.
4. The real exchange rate is the price of foreign goods relative to the price of domestic goods.
The real eY.change rate is inversely related to the international terms of trade. The percentage
change in the real exchange rate equals the percentage rate of depreciation of the nominal
exchange rate plus the difference between the foreign and the domestic rate of producer
price inflation. In long-run equilibrium the real exchange rate must be constant, implying that
relative purchasing power parity (RPPP) must hold. Under RPPP the rate of depreciation of
the nominal exchange rate equals the difference between the domestic and the foreign
inflation rate.
5. When capital is perfectly mobile, relative purchasing power parity implies that in the long run
the domestic real interest rate is tied to the foreign real interest rate. This long-run relationship
is referred to as real interest rate parity and tends to hold empirically.
© The McGraw-Hill
Companies. 2005
722 PART 7: THE SHORT-RUN MODEL FOR THE OPEN ECONOMY
6. The trade balance is the difference between exports and imports of goods and services, also
denoted net e)ports. When the Marshaii-Lerner condition is met, a depreciation of the real
exchange rate will improve the trade balance. The Marshaii-Lerner condition requ ires that the
sum of the numerical price elasticities of export and import demand exceeds unity.
7. The short-run aggregate demand curve in the open economy implies a negative relationship
between the rate of domestic producer price inflation and the aggregate demand for
domestic goods. The reason is that higher domestic inflation will, ceteris paribus, erode the
international competitiveness of domestic producers. The higher the price elasticities of
export and import demand, the stronger is the reaction of net exports to a change in the real
exchange rate, and the flatter is the aggregate demand curve in the output-inflation space.
When the economy is out of long-run equilibrium, the short-run aggregate demand curve will
gradually shift as the real exchange rate changes over time.
8. The properties of the short-run aggregate supply curve in the open economy will depend on
whether there is relative wage resistance or real wage resistance. Under real wage resistance
wage setters have a target for the real consumer wage, defined as the nominal wage rate
rlP.fl~tP.rl hy thP. ~onsumP.r rri ~P. inrlP.X. RP.~I W~QP. rP.sist~n~P. imr liP.s th ~t thP. short-run ilQQrP.-
gate supply curve will shift when the real exchange rate changes, as workers respond to
changing import prices by adjusting their nominal wage claims. Real wage resistance will exist
when nominal unemployment benefits are indexed to consumer prices.
9. Under relative wage resistance individual wage setters seek to maintain a certain relation
between their own wage rate and the wages set in the rest of the economy. Such behaviour
emerges when nominal unemployment benefits are indexed to nominal wages, and it leads to
an aggregate supply curve of the same form as the SRAS curve for the closed economy. The
specification of aggregate supply adopted here assumes relative wage resistance since
indexation of unemployment benefi ts to wages is the most common international practice.
10. The open economy's long-run aggregate demand curve (LRAD) shows the relationship
between real output and the real exchange rate which is consistent with long-run equilibrium
in the market for domestic goods. The LRAD curve is upward-sloping in the output-real
exchange rate-space since a real exchange rate depreciation increases total demand for
domestic goods by improving the international competitiveness of domestic producers.
11. Under relative wage resistance the open economy's long-run aggregate supply curve (LRAS)
is vertical at the natural rate of output. The long-run equilibrium level of output is then uniquely
determined by the position of the LRAS curve, and the long-run equilibrium real exchange rate
is found where the LRAD curve intersects the LRAS curve.
12. Our AS-AD model of the open economy implies that the long-run equilibrium values of the real
interest rate, the real exchange rate and real output and employment will be independent of
the exchange rate reg ime. In the long run the exchange rate regime is thus neutral with
respect to real variables. The proposition that the exchange rate regime is neutral in the long
run should be seen only as an approximation, since the exchange rate regime may have some
influence on the degree of international economic integration.
13. When there are wealth effects on aggregate demand, the accumulation of net foreign assets
via the current account on the balance of payments will influence the evolution of the
economy. In a long run equilibrium w ithout secular growth, the current account balance
measured in units of domestic output must then be zero to ensure constancy of the real stock
of net foreign assets. However, the evidence suggests that in practice it takes a very long time
for current account imbalances to adjust via wealth effects of domestic demand, suggesting
that these effects are quite weak. To simplify the analysis, our AS-AD model of the open
economy therefore ignores wealth effects stemming from the current account.
Ch apter Appendix: deriving the aggregate demand curve for

23.7 the open econ omy
................................................................................................................................................................................
This appendix shows how to derive the log-linear approximation to the goods market equilib-
rium condition given in Eq. (23) of the main text. Our procedure will be similar to the one which
was used to derive the aggregate demand curve for the closed economy in Section 4 of
Chapter 17.
We start by restating Eq. (17):
Y = 6(Y, G,r, e, E', Y' ) + G (45)
Assuming that we start out in a long-run equilibrium, and using the notation for partial deriva-
tives introduced in Eqs (18)- (22), we calculate the total differential of (45) to get the linear
approximation
Y- Y'= Oy(Y- f) +D,(r-1) +DEf.E' - E')

+D.(e - c)+Dr·(Y'- '? 1) +(1+OG)(G- G) =
0 0
Y- Y'= ( £_ )cE' -P)+( '- )er-r)
1 -Or 1 -Or
6
r_ )cY'-f') +( 1-0y
1
6
+( 1-0y
·_ )ce - e) +(1-0y +~13 )(G-G), (46)
where long-run equilibrium values are indicated by a bar above the variables. It is natural to
assume that, other things equal, an increase in world economic activity will increase the
domestic economy's export market in proportion to the domestic economy's initial weight in
the world economy. In that case we have:
(47)
When a relationship like (47) holds for all countries, it simply means that an increase in world
output Y 1 does not in itself change the individual country's share of the world market.
Our rext step is to rewrite (46) in terms of relative changes of the various variables
(except the real interest rate which is already expressed in percentage terms). Doing this, and
using (47). we get:
Y- Y' ( E'DE
T = '?(1-Dr)
)(E'-P) +(f(16,
--y,- -Dr) (r-1)
)
i:O,. (e-f) ( 1 )(Y'-f') (G(1+0G))(G-G)

+( f( 1- Dr) 7 + 1-D r ---yt + f(1- Dr) ~ . (48)
Model for the Open
© The McGraw-Hill
Companies. 2005
Using the definitions
y "' In Y, y'"' In Y1, e'"' In£', g"' lnG,

and remembering that the relative change in some variable x is approximately equal to the
change in the ratural logarithm of x, we may write (48) as:
{4g)
where
-6
{32 "' '?( 1 - t)y) I
t:6,.
We see that (49) is identical to (23) in the main text.

As we shall see in the next two chapters, the magnitude of the parameter fJ 1 is crucial for
the speed with which the economy adjusts to long-run equilibrium. It is therefore useful to
express fJ 1 in a form which will enable us to estimate its likely magnitude. According to (16)
and (22) we have:
= (JO _ (!NX
6 E-(!£' ()£''
(!NX
() £' = M,li'Jx + I'JM-1) .
Furthermore, let us specify the numerical elasticity of total private demand with respect to the
real exchange rate as:
Using these relationships, and choosing units such that the initial real exchange rate 'E' = 1,
you may verify that the expression for {3 1 may be rewritten as:
{J , = (MJY'J(I'Jx + I'JM-_1)- (T5J"?'JI'Jo . (50)

1 - Oy
This specification of {3 1 will be used in the next two chapters.
23.8 Exercises
Exercise 1. Important concepts in the theory of the open economy
1. Explain what is meant by a 'small specialized economy'?
2. Define the concept of perfect capital mobility and explain the conditions for uncovered and
covered interest rate parity. Which assumption is necessary for both parity conditions to
hold at the same time? Are these parity conditions short·run relationships or long-run
relationships?
3. Define the concepts of the real exchange rate and the international terms of trade. Explain the
conditions for relative purchasing power parity (RPPP) and real interest rate parity. Are these
conditions short-run relationships or long-run relationships?
4. Explain the Marshaii-Lerner condition. Is fu lfilment of this condition sufficient to ensure that a
real exchange rate depreciation will raise the demand for domestically produced goods?
5. Suppose that individual domestic exporters are each selling a differentiated product and that
each of them therefore has a monopoly position in the international market (and thus faces a
downward-sloping demand curve). G ive a theoretical reason why, in these circumstances, the
Marshaii-Lerner must necessarily hold. (Hint: a monopoly firm will produce and sell up to the
point where marginal revenue equals marginal cost. What does this imp ly for the size of the
price elasticity of demand in the firm's optimum?)
6. Explain the concepts of relative wage resistance and real wage resistance and how these
phenomena are related to the rules for indexation of unerrployment benefits.
Exercise 2. Permanent shocks under relative wage resistance

In this exercise you are asked to undertake a graphical analysis of the long-run effects of
permanent shocks in a small open economy with relative wage resistance. The conditions for
long-run equilibrium in such an economy were summarized in Eqs (43) and (44) in the text,
and the long-run equilibrium was illustrated graphically in Fig. 23.7. You may use this diagram
along with (43) and (44) as a basis for answering the following questions.
1. Suppose the foreign real interest rate permanently increases. G ive a graphical illustration of
the long-run effects of this demand shock on the real exchange rate. Explain the effect.
2. Assume that the domestic government permanently increases public consumption. Illustrate
and explain the long-run effect of this policy change on the real exchange rate.
3. Suppose that domestic productivity permanently increases so that the supply shock variable
s "' (1 + y)(ln a - In a) permanently declines from zero to some negative amount. Illustrate and
explain the long-run effects of this productivity shock on domestic output and on the real
exchange rate.
Exercise 3. Permanent shocks under real wage resistance
This exercise invites you to study the long-run effects of permanent shocks in an open
economy with real wage resistance and to compare these effects with those emerging under
relative wage resistance. For this purpose you must first derive the aggregate supply curve
under real wage resistance and characterize the economy's long-run equilibrium under this
wage setting reg ime.
Consider a representative monopoly trade union which sets the nominal wage rate W
with the purpose of achieving a target real consumer wage. The nominal wage rate is set at a
time when the union does not have perfect information on the average consumer price level.
According to (28) in the main text the actual consumer price level is pc = (E') v· P, where E' is
the real exchange rate and P is the price of domestically produced goods. Hence the
expected consumer price level is p ee= (E'") "'p •, where E'" and p • are the expected levels of
E' and P. We assume that the nominal rate of unemployment benefit is indexed to the
Model for the Open
© The McGraw-Hill
Companies. 2005
consumer price level, so the real rate of unemployment benefit, b, is unaffected by changes in
Pc. The union aims to obtain an expected real consumer wage wh ich is a mark-up, mw, over
the real unemployment benefit. Thus the nominal union wage claim is given by:
(5 1)
It is reasonable to assume that the union will moderate its wage c laim when the rate of
unemployment (u) rises. We may model this by specifying the union mark-up factor as:
w > 0, y > 0, (5 2)
where w and y are constants.

The representative domestic firm uses a linear production technology by which one unit
of labour produces a units of output. Hence the marginal cost of production is Wfa, and the
firm sets its price as a mark-up, mP, over marginal cost:
P = mP · - ,
w mP > 1. (53)
a
Finally, we assume that the real unemployment benefit is tied to the 'normal' level of produc-
tivity, that is, the trend value of a, denoted a:
b = c ·a. (54)
This specification assumes that, over the long run, the unemployed are allowed to share in the
real income gains generated by productivity growth.
1. Show by taking logarithms th at Eqs (51) - (54) lead to an expectations-augmented Phillips

curve of the form:
ln(mPwc)
:;r = :;r 8 + y(!J - u) + lf.le'• +In a - In a, U!l:ll I (55)
y
n"" In P-In P_1, e'• "" In £ '•,
where n and n • are, of course, the actual and the expected domestic inflation rate, respec-
tively. (Hint: use the approximation ln(1 - u) "' -u.) Explain in economic terms why the (log of
thP.) P.XpP.r.tP.OrP.::l l P.Xr.h::ln!)P. r::ltP., P.' 0 , ::lppP.::lrS in (55).
2. Use (55) along w ith Eqs (39) and (40) in the main text to show that the short-run aggregate
supply curve under real wage resistance takes the form:
n = n • + y(y- y) + lf.le'• + s, s "" (1 +y)(ln a- In a), (56)
where s is a supply shock variable reflecting productivity shocks.
3. Derive the long·run aggregate supply curve (LRAS) in an open economy with real wage resis-
tance by inserting the long-run equilibrium conditions n • = n and e'• = e'"" In E' into (56).
Draw the resulting LRAS curve in a d iagram withy along the horizontal axis and e ' along the
vertical axis, and g ive an economic explanation for the slope of the LRAS curve. Draw the
long-run aggregate demand curve (the LRAD curve g iven by Eq. (43) in the text) in your (y, e1
diagram. Explain the slope of the LRAD curve and identify the economy's long-run equilibrium.
W illy necessarily be equal toy in long-run equilibrium? Make a graphical comparison of the
long-run equilibrium under real wage resistance and under relative wage resistance.
4. Assume that the domestic government permanently increases public consumption. Illustrate
the long-run effects of this policy change on domestic output and on the real exchange rate
under the two alternative scenarios of real wage res istance and relative wage resistance.
Compare and explain the effects.
5. Suppose that domestic productivity permanently increases so that the supply shock variable
s "' ( 1 + y)(ln a -In a) perm anently declines from zero to some negative amount. Illustrate the
long-run effects of this productivity shock on domestic output and on the real exchange rate
under the two alternative scenarios of real wage resistance and relative wage resistance.
Compare and explain the effects.
Exercise 4. The open economy with wealth effects on aggregate demand
In this exercise you are asked to study an open economy with relative wage resistance and
wealth effects on private consumption demand. We noted in the text that imbalances on the
current account of the balance of payments may affect aggregate demand by changing the
private sector's stock of wealth and the associated amount of capital income accruing to
rlomP.stic: rP.siciP.nts.
To concentrate on the role of the current account in the process of wealth accumulation,
we will ignore other sources of changes in private wealth. Thus we will identify private wealth
with the private sector's stock of net fo reign assets, denoted by F and measured in units of
domestic output (just as we have measured the trade balance in units of domestic output).
Ceteris paribus, a rise in Fwill generate an increase in private consumption. Hence we must
respecify our previous goods market equilibrium condition (17) in the following way, where
we do not explicitly include the exogenous confidence variable c and the exogenous foreign
activity level Y':
Y = D( Y, G, r,E', F) + NX(Y, G,r,E', F) + G, (57)

oD oNX oNX
oF"' oF > o, oF < o, Dp+ oF > o.
By stimulating private consumption, a rise in net foreign assets will cause a deterioration of
the trade balance, because part of the rise in consumer demand will be directed towards
imports. However, since the marginal propensity to import is less than 1, the net effect of a
rise in F on the demand for domestic goods is positive, as indicated by the last inequality
above.
By definition, we have:
EF'
F " 'p- (58)
where F 1 is the nominal stock of net foreign assets denominated in foreign currency. As an
accounting identity, we also have:
E+1 F'.. 1 -EF' = P· ( E -E)

NX + i 1 +~ EF
1
(59)
This equation says that the increase in the nominal stock of net foreign assets (the left-hand
side) arises through a nominal surplus on the trade balance, P · NX, through interest earnings
on the existin~=J forei~=Jn assets, i 1 • EF 1, or throuQh exchan!=Je rate Qains on the current asset
Model for the Open
© The McGraw-Hill
Companies. 2005
728 PAR T 7: THE SHORT-RUN MODEL FO R THE OPEN ECONOMY
stock, (~)EF'. Dividing by P. , "" (1 + n . ,)P in (59) and using (58) p lus the approximation
~e .,= (E. , - El/E, we find that the increase in foreign assets measured in units of domestic
output is given by:
(60)
We can now characterize the long-run equilibrium in the open economy with wealth effects.
Since we are abstracting from long-term growth, a long-run equilibrium requ ires that the stock
of net foreign assets be constant, since aggregate demand would otherwise keep on shifting
over time. We also know that relative purchasing power parity and real interest rate parity
must hold in the long run. Moreover, if we assume relative wage resistance, we know that the
long-run equilibrium level of output is an exogenous constant, f . Hence we have the long-run
equilibrium conditions:
(6 1)
which may be inserted into (57) and {60) to give:
f = O(f, G, r1, E', F) + NX( f , G, r1, E', F)+ G, (62)

NX(f, G,f, E', F) + r' F = 0 . (63)
Equations (62) and (63) determine the long-run equilibrium values of the two endogenous
variables E' and F. Notice from (63) that in the long run the current account has to balance in
real terms to ensure constancy of the real stock of net foreign assets.
1. Use Eqs (62) and (63) to illustrate the economy's long-run equilibrium in a diagram in which
F is measured along the horizontal axis and E' is measured along the vertical axis. The equi-
librium combinations of F and E' implied by (62) may be termed the 'GME curve' (GME for
Goods Market Equilibrium). Show that the slope of the GME curve is:
dE' OF+ ()NX)

()F
(dF) CME
=- aNx < 0 '
( Q + --
(64)
E d£'
which is seen to be negative since we assumed in the main text that DE+ (()NX/aE') > 0. Give
an economic explanation for the negative slope.
The equilibrium combinations of F and E' implied by (63) may be denoted the 'CAB
curve' (CAB for Current Account Balance). Show that the slope of this curve is:
(65)
Empirically, a value of r1 in the reg ion of 5 per cent per annum is plausible. In many empirical
studies based on annual data the magnitu de of the derivative OF in the numerator of (64) is
also estimated to be close to 0.05. Assuming OF"" r1, you can now determine the sign of the
slope of the CAB curve and say whether it is steeper or flatter than the GME curve. {Hint:
what is the sign of the derivative DE e CJO/oE'?)
ldentif.; the long-run equilibrium values of F and E' in your diagram.
2. Suppose the domestic government permanently raises public consumption. Give a graphical
illustration of the long-run effects of this policy change on net foreign assets and the real
exchange rate. Explain the effects.
3. Assume that the foreign real interest rate permanently increases. Illustrate and explain the
long-run effects of this demand shock on net foreign assets and on the real exchange rate.
4. Assume alternatively that the domestic economy is hit by a positive supply shock which per-
manently increases domestic potential output f . Illustrate and explain the long-run effects of
this supply shock on net foreign assets and on the real exchange rate.
Exercise 5. The sustainable balance of payments
In Exercise 4 we saw that in an economy with wealth effects on aggregate demand and with
no rP.;il growth in thP. long run, thP. c:urrP.nt ;iC:C:ount m u~t h;il;inC:P. in rP.;il tP.rm~ in thP. long run
to keep the stock of private wealth constant. However, if there is secular real growth, a long-
run equilibrium no longer requires constancy of real private wealth; it only requires that real
private wealth grows at the same rate as real GDP. This means that the real balance on the
current account no longer has to be zero in the long run, although there is a limit to how
much it can deviate from zero. You are now asked to illustrate these points by some simple
numerical examples.
Consider an open economy where real GDP is expected to grow at an average annual
rate of 2 per cent in the long run. Suppose that this economy starts out with a stock of net
foreign debt equal to 25 per cent of GDP. Suppose further that the average inflation rate at
home and abroad is expected to be 2 per cent per annum. Assume finally that the international
real interest rate is 4 per cent per annum.
1. What is the magnitude of the nominal current account deficit relative to nominal GDP which
the countr; can afford to sustain without increasing the ratio of foreign debt to GDP?
2. What is the magnitude of the real current account deficit {the deficit measured in units of
domestic goods) which the country can afford to sustain without increasing the debt- GOP
ratio?
3. What is the magnitude of the nominal trade balance relative to nominal GDP which the
country must sustain to maintain a constant ratio of foreign debt to GDP?
Model for the Open
24. The open economy w~h
fixed exchange rates
© The McGraw-Hill
Companies. 2005
Chapter
The open economy

.• with fixed exchange
T
he AS-AD model of the open economy developed in the previous chapter implies
that the long-run equilibrium values of real variables such as real output. the real
interest rate and the real exchange rate are unaffected by the exchange rate
regime. In this and the next chapter we shall see that a country's choice of exchange rate
regime is nevertheless a fundamental economic policy choice because of its influence on
the economy's short-run dynamics. First of all, the choice of exchange rate regime will
determine whether monetary policy can be used as a tool of macroeconomic stabilization
policy. Second, the exchange rate regime will affect the way in which the economy adjusts
to its long run equilibrium and h ow it responds to exogenous shocks in the short and
medium term.
The rest of this book will elaborate on these points by analysing the workings of the
macroeconomy under alternative exchange rate regimes. The present chapter focuses on
a small specialized eco11omy with .fixed exchange rates and perfect capita/mobility. We start by
studying the characteristics of fixed exchange rates as an economic policy regime and by
discussing why a country might want to adopt a fixed nominal exchange rate. We then
adapt the AS-AD model of the open economy set up in Chapter 23 to analyse how the
economy adjusts to long-run equilibrium under Hxed exchange rates. The flnal parts of
the ch apter use our AS-AD model to study the eflects of macroeconomic policy under fixed
exchange rates.
24.1 Fixed exchan ge rates as an economic policy regime

................................................................................................................................................................................
Under a flxed exchange rate regime the central bank fixes the nominal price of (some)
foreign currency in terms of domestic currency. This ofllcial price of foreign currency is
referred to as the exchange rate parity. In the purest version of a iixed exchange rate
regime, the central bank stands ready to buy and sell unlin1ited amounts of foreign cur-
rency at the exchange rate parity so that the nominal exchange rate is completely flxed . In
practice. most countries \>\lith 'fixed' exchange rates have allowed the market price of
foreign currency to fluctuate within some band around the exchange rate parity. Of
730
Sorensen-Whitta-Jacobsen: I Part 7- The Short-run 24. The open economy w~h © The McGraw-Hill
Introducing Advanced Model for the Open fixed exchange rates Companies. 2005
24 THE O PEN EC ONO MY WI TH FI XED EXC HA NG E RATES 731
course, the wider the band within which the exchange rate is allowed to vary, the closer
the 'Hxed' exchange rate regime comes to a regime with freely floating exchange rates.
Moreover, many countries with 'Hxed' exchange rates have occasionally changed the
ofllcial exchange rate parity, in effect following a policy referred to as fixed but adjustable
exchange rates. The dill'erence between Hxed and flexible exchange rates is therefore a
matter of degree. and the precise dividing line betv.reen the two types of regime is not
clear-cut.
In Chapter 2 6 we shall describe the different types of exchange rate regimes found in
the world and study some important historical examples of Hxed exchange rate arrange-
ments with varying degrees of 'llxity' of the exchange rate. For analytical purposes the
present chapter will start out by focusing on the theoretical benchmark case where the
nominal exchange rate is completely IL"<ed. Later in the ch apter we will consider a regime
with Hxed but adjustable exchange rates.
The impotence of monetary policy under fixed exchange rates and free
capital mobility
Under a regime '<\lith fully fixed exchange rates and free capital mobility it becomes impos-
sible to use monetary policy for the purpose of stabilizing the domestic economy. To
understand this crucial point. recall from the previous chapter that when international
capital mobility is perfect and investors are risk neutral, the following condition lor
uncovered interest parity must hold:
e = In E. (1)
Here i is the domestic nominal interest rate, if is the foreign nominal interest rate, E is the
nominal exchange rate (the price of foreign currency in terms of domestic currency), and
E~ 1 is the nominal exchange rate expected to prevail in the next period. Hence e~ 1 - e is the
expected percentage rate of depreciation of the domestic currency against the foreign
currency, that is, the expected capital gain on foreign bonds relative to domestic bonds
over the period considered. Thus if + e~ 1 - e is the total expected return on foreign assets
which must equal the return i on domestic assets when the tv.ro asset types are perfect
substitutes. Under a so-called bard peg where the exchange rate is credibly fixed, the
expected exchange rate change '<Viii be zero, th at is. e! 1 - e = 0 . It then follows from (1)
th at:
i = jl, (2)
In other words, under jixed exchange rates and perfect capital mobility the domestic nominal
interest rate is tied to the .foreign nominal interest rate.
As far as the market-determined interest rates are concerned, the equality in (2) is
enforced by the arbitrage between domestic and foreign assets described in the previous
chapter.
When it comes to the sh ort-term domestic interest rate set by the central bank, Eq. (2)
is enforced by the need to protect the country's foreign exchange reserves. If the central
bank tries to keep the domestic short-term interest rate below the foreign short-term
interest rate, capital will flow out of the domestic economy. As investors sell domestic
currency and buy foreign currency to invest in foreign assets. the central bank will have
Model for the Open
© The McGraw-Hill
Companies. 2005
732 PART 7 : TH E SHO RT-RUN MO DEL FO R TH E O PEN ECONO MY
to sell foreign currency and buy domestic currency to maintain the 11xed exchange rate. In
this way the country's foreign exchange reserves will be exhausted if the central bank
insists on keeping the domestic short-term interest rate below the foreign level. Indeed. the
market will most likely interpret such an interest rate policy as a signal that the central
bank is not committed to defending the exchange rate. In that case expectations of a
future devaluation will arise. and a speculative attack on the domestic currency will
occur, leading to an even faster depletion of foreign exchange reserves. Once the reserves
are gone, the central bank can no longer intervene in the foreign exch ange market to keep
the exchange rate fixed, and the domestic currency will have to fall.
On the other h and, if for some reason the central bank keeps the domestic short-term
interest rate above the foreign short-term interest rate. a massive inflow of capital from
abroad will occur. This will swell the foreign exchange reserves and create a constant
upward pressure on the exchange rate. The huge capital inflow will also lead to a large
expansion of domestic bani< credit and a rise in domestic asset prices which will generate
strong inflationary pressures.
The upshot is that if the central bank wishes to maintain a 11xed exchange rate, it
cannot set a short-term interest rate which deviates from the short-term interest rate
abroad. The situation in Denmark illustrates this point very well. Th e Danish krone is
pegged to the euro. and whenever the European Central Bank decides to change its
leading interest rate, the Danish central bank interest rate is almost always changed by
exactly the same amount within the same hour.
Monetary policy thus becomes impot.e11t under tixed exchange rates and perfect
capital mobility. When the exchange rate is 11xed. the central bank cannot pursue an
independent interest rate policy aimed at managing aggregate demand. Instead.
monetary policy is fully bound by the commitment to defend the exchange rate. In
contrast, if a country is willing to accept exchange rate variability, it can set its interest
rate independently of the foreign interest rate. as indicated in Eq. (1).
In principle. a country with a fixed exch ange rate can also pursue an independent
monetary policy if it is able to maintain effective control over private cross-border capital
flows. thereby preventing the arbitrage which would otherwise drive the domestic interest
rate into equality with the foreign interest rate. However. nowadays practically all the
developed cow1tries in the world h ave abandoned capital controls. partly because they are
difllcult to enforce, and partly because they are considered undesirable, since they prevent
investors from seeking out the most prolltable investment opportunities and from diversi-
fying their portfolios to hedge against country-specillc invesnnent risks. 1
The case for fixed exchange rates
If capital controls are inellective or are seen as undesirable, the price of maintaining fixed
exchange rates L~ the loss of monetary policy autonomy. Given this fact, why would a
country nevertheless choose a fixed exchange rate regime? Advocates of fixed exchange
rates usually point to the disadvantages of exchange rate uncertainty and to the benefits
of using a fixed exchange rate as a nominal anchor in the fight against inflation. Let us
briefly summarize these arguments.
1. In Chapter 4 we offered a more extensive discussion of the long·run benefits from free capital mobility.
1. The first argument is that the large exchange rate movements often observed under
floating exchange rates may hamper international trade and investment by creating
exchange rate uncertainty. The impediment to trade may not be large, since short-
term foreign exchange risk can be covered through the forward exchange market at a
modest transaction cost. But the opportunities for covering long-term foreign
exchange risks are much more limited, and this might discourage long-term trade
contracts and long-term international investments if the exchange rate is highly
unstable.
2. A related argument starts from the observation that exchange rates often seem to
'overshoot' their long-run equilibrium values under floating exchange rates. as we
shall see in the next chapter. The concern is that such excessive exchange rate move-
ments may cause an unintended redistribution of real income across different sectors
in the economy. For example, a sharp appreciation of the domestic currency will erode
the international competitiveness of the domestic export industry and of domestic
industries which are competing against imports. As a consequence, profits and
employment in those sectors will be squeezed, whereas sectors relying on imported
inputs and selling their output in the domestic market without any competition from
abroad will benefit from the appreciation. If the resulting redistribution of income is
significant, it may cause social tensions.
3. Another main argument is that a fixed exch ange rate can provide a nominal anchor
which may help to bring down inflation. In the longer term a country can on ly main -
tain a lixed exchange rate against a foreign trading partner if the domestic inflation
rate does not systematically exceed inflation in the foreign country. Policy makers
may therefore signal a commitment to keep the domestic inflation rate low by
announcing that they will peg the domestic currency to the currency of a foreign
country with a history of low inflation. If this commitment is credible, it will keep
domestic inflation expectations in check. and this in turn will make it easier to keep
the actual inflation rate low. Thus, by pegging to a stable foreign currency, domestic
policy makers may 'import' some of the credibility and discipline of foreign policy
makers who have been successful in fighting inflation. Such a strategy is most likely to
succeed if the political costs of giving up the peg to the foreign currency are perceived
to be considerable. Oth erwise the public may not consider the llxed exch ange regime
to be credible.
The belief that pegging to a stable foreign currency can help to bring down
domestic inflation motivated many previous high-inflation countries in Europe to tie
their currencies closer to the German D-mark after the mid 1980s. After a setback
caused by the speculative attacks on the European i'vfonetary System in 1992-93 (to
which we shall return later in this chapter). this policy of pegging to the D-mark grad-
ually led to lower exchange rate variability which did in fact drive national inflation
rates in the EU closer to the low German level. as shown in Figs 24.1 a and 24 .1 b.
Economists have a long-standing debate on the costs and benefits of fixed versus
flexible exchange rates. In this section we have briefly restated the main arguments in
favour of fJXed exchange rates. In the next two chapters we shall consider the case lor
flexible exchange rates. At this stage we can already see one basic trade-o!Tin the choice of
exchange rate regime: by fixing the exchange rate, a country may eliminate nominal
Model for the Open
© The McGraw-Hill
Companies. 2005
734 PART 7 : THE SHORT-RUN MODEL FOR TH E OPEN ECONOMY
- France
§ - ~
'(ii -t 15 +-- - -1+ - - - - - - - - - - - - - . , . - - - -+-1 ....... Portugal
i~ 10 +---:f+-\H L- - - - - - - - - - - --i-- --l-1-===S=p=a=in=J

"'c
~~
c
~ CJ
~
5 ;~----r~~'1;1-cR-
i$~
~= 0
~ ~
~ ·~
~"' - 5
Year
Figure 24.1 a: Exchange rate variability in Europe, 1980-2000

Note: Percentage rat e of currency depreciation against the German D·mark.
Source: IMF Financial S tatistics.
30,- - - - - - - - - - - - - - - - - - - - - - -
- France
25 -1-----...,~'-\--------------1 - - Germany
/ \ ........ Italy
,/ \ --- Portugal
20 -1-:..-....-,-,"'"<----~\------------1- Spain
,,"·······.... \
~ 1 5~~-~---~------------------
0
0
00
a>
- "'
co
CJ) CJ) CJ)
(')
co co co co
CJ)
'<t
CJ)
lO <D
00
CJ)
c-. co
00
CJ)
co co
CJ)
CJ)
CJ)
0
CJ)
CJ)
- "'
CJ)
CJ)
CJ)
CJ)
(')
CJ)
CJ) ""
CJ)
CJ)
lO <D
CJ)
CJ)
CJ)
CJ)
c-.
CJ)
CJ)
co
CJ)
CJ)
CJ)
CJ)
CJ)
0
0
0
0
0
Year "' "'

Figure 24.1 b: nflation rates in Europe, 1980-2000
Source: OECD Main Economic Indicators.
exchange rate uncertainty and provide a nominal anchor for domestic prices, but these
benellts can only be reaped by giving up monetary policy autonomy. In contrast. flexible
exchange rates offer monetary policy autonomy, but only if policy makers are willing to
accept the possibility oflarge fluctuations in exchange rates. Given this ditllcult trade-off.
it is hardly surprising that dill'erent countries with diflerent economic histories and
dill'erent economic structures have chosen dill'erent exchange rate regimes.
Inflation and real exchange rate dynamics
Before analysing the ell'ects of economic policy in an open economy with fJXed exch ange
rates. it is useful to study the dynamic process through which such an economy adjusts to
long-run equilibrium. As we shall see in this section, the evolution of the real exchange
rate plays a key role in the macroeconomic adjustment process. From the previous
=
chapter you will recall that the real exchange rate is dellned as E' EPfjP. where E is the
nominal exchange rate. pf is the foreign price level measured in foreign currency, and P is
the price of domestic goods. When the nominal exchange rate L~ fixed, it follows that the
percentage change in the real exchange rate (the change in the log of the real exch ange
rate) is given by:
(3)
e' =In E' , nf =In pf - In P~., n= LnP - lnP_1 ,
where n f and n are the foreign and the domestic in flation rate, respectively. Thus the real
exchange rate can only remain stable over time if the domestic inflation rate corresponds
to the foreign inflation rate. Since a constant real exchange rate is one condition for long-
run equilibrium, it follows that the domestic inflation rate will converge on the foreign
inflation rate if the long-run equilibrium is stable {as indeed it will be under the
assumptions made below).
We will now illustrate the role of the identity (3) in our AS-AD model of the open
economy with fixed exch ange rates.
Aggregate demand and aggregate supply
In Eq. (26) of' the previous chapter we presented the general expression for the aggregate
demand curve in the open economy. Allowing for the possibility of exchange rate flexibil-
ity , that equation included both the actual rate of nominal exchange rate depreciation
which affects the country's international competitiveness. and the expected rate of
nominal exchange rate depreciation w hich influences the domestic rate of interest. Under
a policy regime where nominal exchange rates are credibly fixed - that is, where the
public believes in the government's declared commitment not to devalue the currency -
the expected as well as the actual rate of nominal exchange rate depreciation will be zero.
Inserting this condition (~ e = ~ e' = 0) into Eq. (2 6) of the previous chapter, we get the
following expression for the goods market equilibrium condition in an economy with
credibly fixed exchange rates:
(4)
where pf is the initial foreign real interest rate, n: is the expected rate of domestic inllation
1
© The McGraw-Hill
Companies. 2005
between the current and the next period, and z is a demand shock variable. From ( 3) we
see that the magnitude e~ 1 + n f- n in (4) is simply the current real exchange rate, e' . Thus
aggregate demand varies positively with the competitiveness of domestic producers.
measured by the current real exchange rate. We also see that total demand varies
negatively with the real interest rate. ji - n:
1•
In modelling aggregate supply. we will rely on the hypothesis of relative wage
resistance explained in the previous chapter. As you recall, this hypothesis leads to a
standard aggregate supply curve of the form:
n = n '' + y(y - y) + s. (5)
To complete our AS-AD model with fixed exchange rates, we now only need to specify the
formation of inflation expectations.
Expectations formation under credibly fixed exchange rates
As a starting point for doing so. let us return to the idea presented at the end of Ch apter 21
thrJt snmP. pP.nplP. r~rP. hrJdn>~mrrl -l nnking whP.rP.rJs nthP.rs rJrP. IorwrJrrl-lnnking. SpP.dfi r.r~ll y.
suppose the population is divided into two groups. The first group (Group 1) has purely
backward-looking expectations and simply believes that this period's intlation rate will
equal the rate of inflation observed in the previous period. This is, of course, the familiar
hypothesis of static expectations.
The second group of people (Group 2) is more sophisticated. This group realizes that
under fixed exchange rates domestic inflation cannot systematically exceed foreign infla-
tion for a long period of time, since this would undermine the international competitive-
ness of domestic producers and lead to an ever-increasing trade deficit. An ever-growing
trade deficit would ultimately force the domestic authorities to bring down domestic infla-
tion in order to protect the country's foreign exchange reserves. Similarly, Group 2 people
also understand that if domestic inflation were systematically lower than foreign inflation.
the persistent fall in the relative price of domestic goods would lead to ever-increasing
demand for domestic output which would ultimately pull up the domestic inflation rate.
In short, Group 2 agents realize that on average the domestic inflation rate will have to
equal the foreign in flation rate, but they do not h ave enough information to predict th e
short-run fluctuations of the inllation rate. Instead, they assume that the domestic infla-
tion rate for the current and next period will be at its average level given by the foreign
inllation rate. We may say that these individuals have 'weakly rational' expectations.
since they hold correct belief's about the long-run equilibrium inflation rate but cannot
perfectly predict the short-run inflation rate.
If Group 2 constitutes a fraction <p of the total population. these assumptions imply
that the average expected in llation rate will be:
O<;; <p <;;l. (6)

In Exercise 6 we ask you to study the general case where 0 < <p < 1. However, in the
following we will consider the special case where all agents have weakly rational
expectations, <p = 1, so that n ' = n~ 1 = nf , where foreign intlation is assumed to be con-
stant at the level nl . We will concentrate on this case because it greatly simplilles the
graphical exposition of our AS-AD model and allows us to focus on the role of real
exchange rate dynamics in the adjustment to long-run macroeconomic equilibrium.
The assumption that :rr:' = :rr:~ 1 = :ref may be seen as a compromise between the case of
purely backward-looking expectations and the strong postulate of strictly rational,
forward-looking expectations. Our hypothesis on expectations formation is a simple way
of formalizing the idea that, under credibly fixed exchange rates, the authorities have
implicitly adopted an inflation target equal to the foreign inflation rate, and this inflation
target serves as an anchor for domestic inflation expectations.
The complete AS-AD model with fixed exchange rates
Inserting :rt' = :rr:~ 1 = :ref into (4) and (5), isolating:rr: on the left-hand side of( 4), and remem-
bering that ri = if- :rei is the foreign real interest rate, we may now state our complete
model of the open economy with fixed exchange rates as follows:
AD: :rr: = e~l + :ref- (;Jy- y - z), (7)
z =-f3iri- F) + f3ig - g)+ f3iyf - yi) + {3 5 (ln e - In l').

SRAS: :rt = :rei + y(y - y) + S, (8)
Real exchange rate: (9)
Equation (7) is a restatement of the aggregate demand curve. Since P1 > 0, we see that a
higher domestic inflation rate will, ceter is paribus. be associated with lower aggregate
demand for domestic output. In the closed economy the negative impact of inflation on
demand stems from the fact that higher inflation induces the central bank to raise the
real interest rate. In the open economy with fixed exchange rates this mechanism is
suppressed, because the domestic interest rate is tied to the foreign interest rate via perfect
capital mobility. Instead, the reason for the negative slope of the AD curve (7) is that
higher domestic inllation erodes the international competitiveness of domestic producers
by reducing the real exchange rate. Thus, a rise in :rc raises the relative price of domestic
goods. thereby reducing net exports.
The variable z in (7) captures so-called real shocks to aggregate demand. In the open
economy. these shocks include changes in foreign real output, yf, and in the foreign real
interest rate, I . as well as changes in domestic real government spending, g. and in
domestic private sector confidence, f. . A change in any ofthese variables will cause a shift in
the aggregate demand curve. In addition, ( 7) shows that the domestic economy may be
exposed to a so-called nominal shock in the form of a change in the foreign intlation rate, :rei .
Such a shock will also shift the AD curve. In addition, a foreign inflation shock will shift the
short-run aggregate supply curve (8) by changing the expected rate of domestic inflation.
From short-run to long-run equilibrium
As the economy enters the current period. last period's real exchange rate e~ 1 is a pre-
determined constant. The current rates of output and inllation. y and :rc, are then determined
simultaneously by the AD curve (7) and the SRA.<:> curve (8 ), given z and :ref and the supply
Model for the Open
© The McGraw-Hill
Companies. 2005
738 PART 7 : TH E SHO RT-RUN MO DEL FO R TH E O PEN EC ONO MY
shock variables. The current level of inflation in tum determines the current real exchange
rate via (9). Ifni * n it follows that e' * e~ 1, and the economy will then enter the next period
with a new predetermined real exchange rate. According to (7) this means that the position
of the AD curve will shift from the current period to the next one. This will generate new
short-run values of output and inflation during the next period which in turn will change
the level of the real exchange rate, and so on.
Figure 24.2 illustrates this dynamic adjustment process. The bracket attached to
each AD curve in the llgure indicates the previous period's real exchange rate w hich
determines the position of the AD curve lor the current period. according to Eq. (7). In
period 1 the AD curve has the low position AD 1 determined by the real exch ange rate in
period 0, e0. The weakness of aggregate demand in period 1 means that the economy is in
recession, with low levels of output and in flation (y 1 and n 1) . But this is only a short-run
equilibrium: since domestic in flation is lower than foreign inflation, the real exchange rate
increases during period 1 so that e; = e[) + nf- n 1 > e[>· Thus the international competi-
tiveness of domestic producers improves by the amount Jtf- n 1 during period 1. and
according to (7) and (9) the AD curve shifts upwards by a corresponding distance as the
economy moves from period 1 to period 2. As a result, domestic output and inflation
increase to the levels y 2 and n 2 in period 2.
However. since domestic inflation is still lower than foreign inflation in period 2 ,
there is a further gain in domestic competitiveness, so the AD curve shifts upwards by the
distance n l - n 2 between periods 2 and 3, causing a further rise in output and inflation in
period 3, and so on. In this way the gradual improvements in domestic competitiveness
will continue to pull the economy up along the short-run aggregate supply curve until
domestic output reaches its trend level. [J. At th at point domestic inflation catches up with
JC
LRAS
Y1 Y2 Y3 ji y
Figure 24.2: The adjustment to long·ru n eq uilibrium under fixed exchange rates
foreign inflation so that no further changes in the real exchange rate will take place. that
is, the economy will be in long-run equilibrium. Provided the supply shock variables is
zero. output will be at its trend level yin this long-run equilibrium. since it follows from (8)
th at y = y when rr = rrf and s = 0 .
Notice the difference between the dynamic adjustment mechanism in the closed and
in the open economy. In the closed economy a recession will lead to falling inflation which
induces the central bank to cut the real interest rate so that aggregate demand recovers.
Thus economic policy is crucial lor the dynamic adjustment of the closed economy. In the
open economy with fixed exchange rates a recession drives domestic inflation below
foreign inJlation, causing a gradual recovery of demand by lowering the relative price of
domestic goods. This endogenous adjustment of the real exchange rate will pull the open
economy towards long-run equilibrium even if economic policy remains passive.
Stability and speed of adjustment under fixed exchange rates
What determines the speed of adjustment to long-run equilibrium in the open economy
with 11xed exchange rates? To investigate this, we will now solve the model analytically.
We start by defining the output gap, fj, and the inflation gap, ir, in the usual manner:
fJ, =u,- g, ( 10)
We may then rewrite the model consisting of (7), (8) and (9) as:
(ll)
it,= 1Yt + Sr. (12)
e,r =e,_L- n ,.
1 •
(13)
Now insert (12) into (11) and solve lor e~- 1 to get:
r
e,_ 1 + Yf31
L= ( _{3_1_ ) Yr+ 1 ) z,
• sr - ( {31 (14)
Substituting (14) and the analogous expression for < into (13) along with the expression
for ir, given in (12 ), we get the following first -order linear dillerence equation in the output
gap:
1
{3=-- . (1 5)
1 + Yf3t
Equation (1 5) has the same structure as the dillerence equation lor the output gap in
the closed economy given in Eq. (19) in Chapter 19. Indeed, the only dillerence between the
two equations is that the parameter a in the closed economy model is replaced by the para-
meter {3 1 in the open economy model. This reflects the dillerence in the macroeconomic
Model for the Open
© The McGraw-Hill
Companies. 2005
adjustment mechanisms: in the closed economy, the parameter a incorporates the central
bank's interest rate response to changes in the inflation gap and in the output gap as well as
the interest ela~ticity of aggregate demand lor goods. In the open economy. the parameter
f3 1 includes the price elasticities of export and import demand which determine the eflect of
a change in domestic inflation on the demand lor domestic goods.
When Z 1+ 1 = Z 1 = s, = 0 . the solution to (15) is:
t = 0, 1. 2, ... (16)
=
We see from (15) that the parameter f3 1/(1 + y{J 1) is positive but less than 1. According
to (16) this ensures that the economy will indeed converge towards a long-run equili-
brium, provided the shock variables z and s stay constant over time. In other words. the
open economy with flXed exchange rates is stable. The speed of convergence to long-run
equilibrium will be higher, the smaller the value of/3. Hence the adjustment to equilibrium
will be faster the greater the value of f3 1• that is, the larger the price elasticities in foreign
trade. This is intuitive: if exports and imports are very sensitive to relative prices. a
negative demand sh ock which drives domestic inflation belmv foreign inflation will lead to
a large increase in net exports which will quickly pull the domestic economy out of reces-
sion. The magnitude of the price elasticities in foreign trade ·will depend on the structure of
product markets. If competition in international markets is tough, the price elasticities
will tend to be large, and the economy's adjustment to equilibrium will then be relatively
fast.
Let us consider a numerical example to get a feel lor the likely order of magnitude of
the open economy's speed of adjustment. In the appendix to Chapter 23 we showed that
fJ _ (MofY)(TJx + TIM - 1) - (D/Y)IJv oDE:

~ - 1 - by ' TJv = - aw o ·
where TJ vis the numerical elasticity of total private demand lor goods with respect to the
real exchange rate. M1JY' is the ratio of imports to GDP. and D/Y'is the ratio of total private
demand to GDP (which in turn equals one minus the ratio of government consumption to
GDP when foreign trade is balanced). Empirical studies indicate that if the length of the
time period is one year or longer, the sum of the relative price elasticities in export and
import demand arc a lmost always greater than 1 . For concreteness. suppose that
TJx+ TJM= 3, TJv = 0.3, Mof¥' = 0 .3. D/Y= 0.8,

These parameter values are not implausible for a relatively open economy. The justillca-
tion lor the choice of the value of IJv is that if about 30 per cent of demand is directed
towards imports, then a 1 per cent rise in the relative price of imported goods should erode
the real purchasing power of domestic consumers by about 0.3 percentage points and lead
to a similar percentage fall in demand. In Chapter 19 we assumed that the Phillips curve
parameter y = 0.05 when the length of the time period is one quarter. We therefore
assume that y = 0.2 on an annual basis. Using these parameter values, we Hnd that:
1
fJ =- - .. 0.8 7.
1 + yf31
From the analysis in Chapter 19 we know that the number of time periods t 1' which must
elapse before half of the economy's adjustment to long-run equilibrium is completed is

given by:
In 2 0 .693
t'1 =- - - =- - - .
In .8 In ,8
For .8 = 0.87 this implies that t" "' 4.98. In other words, given the parameter values
assumed above, it \>Viii take about live years before the economy has moved halfway back
towards the long-run equilibrium after it has been disturbed by a shock.
24.3 ~.~.~.~.~~.P?..l.i~Y ........................................................................................................................................... .

We will now use the AS-AD model developed above to study the effects of macroeconomic
policy under llxed exchange rates. As we h ave already seen, when fixed exchange rates
are coupled with high capital mobility, there is no role for an independent monetary
policy. This leaves llscal policy and discretionary changes in the exchange rate parity as
the two major remaining instruments of macroeconomic stabilization policy. In Section 4
we shall consider the ellects of exchange rate policy. The present section focuses on fiscal
policy. We start by analysing the repercussions of a llscal policy shock before moving on to
consider the ellects of a systematic countercyclical llscal policy.
Unsystematic fiscal policy: a temporary fiscal shock
Figure 24.3 illustrates how an open economy with fixed exchange rates will react to a
temporary fiscal expansion. In period 0 the economy is in long-run equilibrium at A0
LRAS
Y2 Y1 y
Figure 24.3: Effects of a temporary fiscal expansion

Model for the Open
© The McGraw-Hill
Companies. 2005
where output is at its trend level and domestic inllation equals foreign inflation. Suppose
then that the government temporarily increases its spending in period 1, but cuts spend-
ing back to its original level from period 2 onwards. In period 1 when the government
raises its spending so that z goes up, the AD curve shifts up frmn AD 0 to AD I' and the
economy moves to a new short-run equilibrium A 1 where domestic output and inllation is
higher. In period 2 when the government cuts back its spending to its original level. one
might think that the AD curve would simply fall back to its original position AD 0 so that
long-run equilibrium would be restored frmn period 2 onwards. In fact, however, the AD
curve for period 2 will fall to the position AD 2 below the original aggregate demand curve.
The reason is that the higher rate of inflation in period 1 reduces the real exchange rate in
that period (since e~ = e[1 + :nf - :n 1) so the economy enters period 2 with a weaker inter-
national competitiveness. In formal terms, the lagged real exchange rate, e~ l' in Eq. (7)
=
falls by the amount l:l:n 1 :n 1 - :nf between period 1 and period 2, thereby pulling down
the aggregate demand curve by a similar vertical distance. As a result, the economy falls
into recession in period 2 when the temporary tlscal stimulus has vanished and has left the
economy with an inflated cost and price level.
From period 2 onwards, we see from Fig. 24.3 that domestic inllation is below the
foreign inflation rate. This generates a gradual improvement of domestic competitiveness
which pulls domestic output and inflation back towards their long-run equilibrium levels
through successive upward shifts in the AD curve.
When the economy has moved all the way back to point A0 , all real variables (includ-
ing the real exchange rate) are back at their original levels. The interesting point is that
the temporary tlscal expansion takes the economy through a boom-bust cycle where a
short-lived economic expansion is followed by a protracted period with below-normal
activity. Of course, a similar adjustment pattern would occur if our demand shock variable
z temporarily increased due to some event other than a change in fiscal policy.
Systematic fiscal policy
The analysis above focused on a fiscal policy shock rooted in exogenous political events. It
is also of interest to study the eflects of a systematic fiscal policy which follows a fixed policy
rule. For example, suppose that fiscal policy is countercyclical so that public spending is
raised above trend when output is below the natural rate, and vice versa. In that case we
have:
g- g=a([l - !J). a> o. (17)
where the policy parameter a indicates how strongly fiscal policy reacts to changes in the
output gap. Inserting (17) into the definition of z given in (7), we obtain a new equation
for the AD curve:
1+{3 3
:n = :nf +e~ 1 - - {3- -
a)([1 - y)+ {3z , (18)
(
1 1
z =- {Jz(! - F)+ fJil - y/) + {3 5 (ln t - In E).

We see that a countercyclical fiscal policy involving a positive value of a makes the
AD curve steeper compared to a passive policy where a = 0. The implications of this are
Sorensen- Whitta- Jacobsen: I Part 7- The Short-run 24. The open economy w~h © The McGraw-Hill
illustrated in Figs 24.4 and 24.5 where we assume that the economy starts out in long-
run equilibrium in period 0.
Figure 24.4 shows the short-run ellects of a negative demand shock which lowers
our variable z from zero to some negative amount in period 1. thereby shifting the AD
curve (18 l downwards by the vertical distance 7./f3 1• In the case of a passive Hscal policy
(a = 0) the economy will then end up in the short-run equilibrium A 1 in period 1, but with
an active countercyclical policy (a> 0) the new short-run equilibrium will be given by
point A '1• In this case we see that the steeper slope of the AD curve ensures a smaller drop
in output as well as in flation. Of course, this is because the countercyclical policy implies
an offsetting rise in public sector spending when the exogenous fall in aggregate demand
reduces the level of output.
Figure 24.5 illustrates the case where the SRAS curve shifts upwards due to an
unfavourable supply shock. Again we see that the steeper AD curve implied by a counter-
cyclical Hscal policy will ensure a smaller drop in output but at the same time it will imply
a larger increase in domestic inflation. Thus policy makers face a trade-oil between
reducing the volatility of output and reducing the variance of inflation when the economy
is exposed to supply shocks. This dilemma is well known from our analysis of the closed
economy.
The countercyclical Hscal policy also involves another dilemma: although it reduces
the short-run output fluctuations caused by exogenous shocks, it also reduces the
economy's speed of adjustment back to long-run equilibrium. If we replace our previous
AD curve {7) by (18) and follow the same procedure as the one used to derive Eq. {15), we
find that the dynamics of the output gap under the countercyclical fiscal policy are given
AD1 (a > 0)
y y
Figure 24.4: Short-run effects of a negative demand shock under a passive versus a countercyclical
fiscal policy
© The McGraw-Hill
Companies. 2005
lf
LRAS
AD(a >0)
y, y, ji y
Figure 24.5: Short-run effects of a negative supply shock under a passive versus a countercyclical
fi scal policy
by:
(19)
As you can easily verify. the coefllcient,B on the lagged output gap is larger the greater the
value of n, that is, the stronger fiscal policy reacts to the output gap. A more vigorous
countercyclical policy will therefore slow down the convergence towards long-run equi-
librium. The reason is that once output starts to move back towards the natural rate after
having been pushed into recession by a negative shock, fiscal policy is automatically
tightened under a countercyclical policy. Obviously this will reduce the speed at w hich
aggregate demand and output are able to increase.
Yet a countercyclical fiscal policy will also reduce the initial displacement from long-
run equilibrium when the economy is bit by a shock. With a quadratic social loss function
of the form assumed in our analysis of stabilization policy in Part 4 . governments are
presumably willing to accept some reduction in the economy's speed of adjustment if they
can thereby amid large deviations of output from trend. This is because large deviations
from the natural rate cause a disproportionately greater social loss than small deviations.
But our analysis shows that governments cannot have it both ways: they cannot reduce
the short-run impact of shocks through countercyclical policy and at the same time
increase the economy's speed of adjustment.
In closing this discussion of fiscal policy, we should remind you of the warnings given
in Chapter 22 . The simple policy rule (17) assumes that fiscal policy makers can instanta-
neously react to the current output gap. This may not be very realistic, since the so-called
inside lag in the llscal policy process may sometimes be quite long due to the parlia-
mentary procedures needed for approval of llscal policy changes. Because of time lags,
there is a danger that llscal policy measures which were intended to be countercyclical
end up being destabilizing because of inappropriate timing.
24A ~?C~.~.~~~~.!.~.~~.P~.~.~~Y...................................................................................................................
So far we have analysed a so-called hard currency peg where the government does not use
changes in the nominal exchange rate as an instrument of economic policy. We will now
consider a ;softer' type of llxed exchange rate regime where the authorities may occasion-
ally adjust the exchange rate. This is sometimes referred to as llxed but adjustable
exchange rates. In practice, most countries with such a policy regime have tended to
devalue their currencies from time to time in order to compensate for the impact of high
domestic inflation on international competitiveness or to stimulate domestic economic
activity.
When analysing the efl'ects of exchange rate policy, it is crucial to distinguish
between anticipated and unanticipated devaluations. When a devaluation is foreseen by the
private sector. it will already have an anticipation eflect on economic activity before it
occurs, whereas an unanticipated devaluation will have no impact until it is imple-
mented.
We will start by considering an unanticipated devaluation since this provides a useful
benchmark when we move on to analyse the more complicated case of a devaluation
which is (partly) anticipated.
An unanticipated devaluation
Suppose the economy starts out in a long-nm equilibrium like the one illustrated by point
A 0 in Fig. 24.6. Suppose further that the government finds that the unemployment rate
corresponding to this equilibrium is too high and needs to be brought down quickly before
the next election. As a ;quick Hx' . policy makers may then engineer a short-run improve-
ment in international competitiveness through a devaluation of the domestic currency.
Let us assume that the devaluation is undertaken at the start of period 1. Given that the
devaluation is unanticipated. it will not alfect the expected exchange rate for period 1.
When the devaluation occurs. the government solemnly declares that this is a unique,
one-time event which will never be repeated; from now on the government is fully
committed to a solidly Hxed exchange rate. Suppose the public actually believes the
government. With these assumptions, the devaluation will never cause the expected rate
of exchange rate depreciation e~ 1 - e to deviate from 0, so according to the uncovered
interest parity condition (1) the domestic nominal interest rate will remain equal to the
foreign interest rate both before and after the devaluation. Moreover, since the public
expects no further devaluations in the future. the expected domestic inflation rate remains
equal to the foreign rate of inflation. so there is no ch ange in the domestic real interest rate
and no upward shift in the aggregate supply curve.
Given these very optimistic assumptions (which will be relaxed later on), the only
short-run etl'ect of the devaluation is to stimulate aggregate demand by raising the real
© The McGraw-Hill
Companies. 2005
LRAS
AD0
y Yt y
Figure 24.6: E'fects of an unantic ipated devaluation
exchange rate. In period 1 which starts right after the devaluation occurs. the real
exchange rate will bee; = e[l + M + n f- Jtp where t. e =In E1 - In E0 > 0 is the percentage
rate of devaluation. Setting z = 0, we may then use (7) to write the AD curve for period 1
as:
.n:l = e[) + !le + Jtf - ( ; ] )cy-[J) . (20)
which shows that the AD curve shifts upwards in period 1 when the devaluation occurs.
The economy therefore moves to a new short-run equilibrium like point A 1 in Fig. 24.6,
where domestic output has increased due to the gain in competitiveness generated by the
devaluation. However, the new short-run equilibrium A 1 also implies that domestic in fla-
tion rises above the foreign inflation rate, so from period 2 the real exchange rate starts to
fall over time as the excess domestic inilation gradually erodes the initial gain in competi-
tiveness. 2 Hence the AD curve starts to shift down from period 2 onwards. This process
\-viii continue until domestic inflation has been brought dovvn in line with foreign in fla-
tion, that is, until the economy has moved all the way down the SRAS curve to the initial
long-run equilibrium A 0 in Fig. 24.6.
We see from th is analysis that a devaluation will be entirely neutrnl in the long run.
In other words, a devaluation will have no impact on the long-nm equilibrium value of
any real variable. To veril)r that a devaluation of the nominnl exchange rate cannot influ-
ence the long-run equilibrium value of the real exchange rate. recall that in long-run
2. From period 2 onwards the evolution of t he real exchange rate is once again given by (13).
equilibrium e' = e~ 1 • :rc = :n:f and y = y - y- 1 s. where s is a permanent supply shock.

Inserting these conditions into (7) and rearranging. it follows that the long-run equilib-
rium value of the real exchange rate is given by:
(21)
which is seen to be independent of the nominal exchange rate. Thus. in the long run the
real exchange rate is affected only by the permanent supply and demand shocks. sand z.
The preceding analysis shows th at policy makers may temporarily drive output above
its natural rate through an unanticipated devaluation. However, since exchange rate
policy cannot permanently influence real output and employment. it may be more
natural to use it as an instrument for speeding up the economy's adjustment to long-run equi-
librium and then use structural policy to steer the natural rate of employment towards its
desired level.
Figure 24.7 illustrates how an unanticipated devaluation may be used to shorten the
length of a recession. In period 0 the economy is in deep recession in the short-run equi-
librium A 0 • In the absence of policy intervention. the economy would gradually pull itself
out of the recession and move up the SRAS curve towards the long-run equilibrium A,
because an output level below the natural rate keeps the domestic inflation rate below the
foreign inflation rate so that international competitiveness is gradually improving.
However, this sell'-regulating market mechanism may work very slowly. Indeed, for
plausible parameter values we have seen that it may take around five years for the
economy to move just halfway towards long-run equilibrium. If the economy starts out
LRAS
SRAS
AD~
Yo Yo y y
Figure 24.7: An unanticipated devaluation to speed up the adjustment to long·ru n equilibrium

Model for the Open
© The McGraw-Hill
Companies. 2005
very far below the natural rate of employment. such a long period of adjustment may be
politically unacceptable. To speed up the adjustment process. policy makers may therefore
decide to devalue the domestic currency in period 0. If the devaluation is unanticipated.
the economy will then jump immediately from the short-run equilibrium Au to the new
short-run equilibrium A;,. In this way the policy makers save the time it would otherwise
take to move from Au to A;, through a gradual gain in competitiveness generated by a
protracted period of high unemployment which keeps domestic inflation low. Still, the
conclusion remains that the devaluation only brings a temporary stimulus to economic
activity, since the economy will ultimately end up in the same long-run equilibrium A
whether or not the currency is devalued.
An anticipated devaluation
Our analysis of an unanticipated devaluation serves as a useful theoretical benchmark.

but in the real world devaluations rarely take the private sector by complete surprise. This
is especially true in countries which have devalued on several previous occasions. In such
countries the government's declared commitment to a 'ftxed' exchange rate will typically
lack credibility, and expectations of a future devaluation may easily arise whenever the
state of the macroeconomy provides (or is believed to provide) a temptation for policy
makers to devalue the currency. Thus, if the authorities devalue the domestic currency
from time to time. households and firms will start to incorporate the risk of future devalu-
ation into their forecasts of inllation and asset returns. A realistic analysis of exchange
rate policy must include such ellects on expectations.
To illustrate this point, we will now consider an economy which is in long-run equi-
librium in period 0 and which undertakes a devaluation in period 2. We may assume that
the decision to devalue is motivated by the policy makers' desire to bring about a quick
expansion in a situation where the natural unemployment rate is unacceptably high .
However, our qualitative results would still be valid if we assumed instead that the
economy starts out in a short-run equilibrium where output is below trend.
Because the country considered has devalued previously on several occasions, we
assume that fears of a future devaluation arise in the private sector in period 1 already.
The economy therefore goes through three stages of reaction to the devaluation. The first
stage may be termed 'the anticipation stage' where the emergence of devaluation expect-
ations starts to allect the real economy even before the devaluation occurs. The second
stage may be called 'the implementation stage' and includes the short-run ellects of the
devaluation occurring in the period when it is implemented, whlle the third and ftnal
'adjustment stage' includes all the subsequent macroeconomic adjustments which take
the economy back to the long-run equilibrium. We will consider each of these stages in
turn.
Stage 1: The a11ticipntion ejfects of a devaluation

The anticipation stage coincides with period 1 where expectations of a filture devaluation
have arisen but where the devaluation has not yet been implemented. The actual per-
centage rate of devaluation implemented in period 2 is t. e2 • We introduce a parameter¢
to indicate the ratio behveen the rate of devaluation which is expected to occur in period 2
and the rate of devaluation which actually occurs. In period 1, the percentage devaluation
which is expected to take place in period 2 (e~- e1) is therefore given by:
0 ~ ¢ ~ 1, (22)
where e~ is the log of the nominal exchange rate expected to prevail in period 2. In the
borderline case where the parameter ¢ is equal to 0 the devaluation is completely
unanticipated. whereas¢ = 1 reflects the opposite borderline case where the devaluation
is fu lly foreseen. In principle. it is of course possible that the private sector overestimates
the rate of devaluation so that¢> 1. but this case will not be considered here.
Since perfect capital mobility implies that the expected returns to domestic and
foreign assets must be the same, it follows from (22) and the arbitrage condition (1) that
the domestic nominal interest rate in period 1 will be:
(23)
Thus the domestic interest rate must equal the foreign interest rate / (assumed to be
constant throughout our analysis) plus the expected exch ange rate gain ¢ !:l e 2 on foreign
assets.
A devaluation raises the price of imported goods, and we will assume that households
and firms expect this to lead to a temporary increase in the domestic inflation rate. As we
shall see, our analysis of the implementation stage will in fact justify such an expectation.
Specifically. we assume that investors in period 1 expect the inflation rate for period 2 to
be:
(24)
The superscript eb in (24) indicates that the expectation is lormed before the magnitude of
the devaluation is known with certainty. The parameter 8 2 measures the extent to which
the anticipated devaluation in period 2 is expected to drive the domestic inflation rate lor
period 2 above the foreign inflation rate. In general. we assume that 8 2 < 1 so that the
devaluation is not expected to be fully reflected in the domestic inflation rate in the short
run. Since the domestic real interest rate in period 1 depends on the expected rate of price
increase between periods 1 and 2, it follows from (23) and (24) that:
(25)
We see that when the anticipated devaluation is not expected to be fully and immediately
passed through to the domestic inflation rate (that is. when e2 < 1). the domestic real
interest rnte will rise above the foreign real interest rate b4ore the devaluation, provided the
devaluation is at least partly anticipated so that ¢ > 0. To see how this will allect the
macroeconomy. we note from Eq. (2 3) in the previous chapter that when there are no
other shocks than those arising from exchange rate policy (i.e.. when g =g. yf = r/ and
t = l) . the equilibrium condition lor the goods market in period 1 can be written as:
(26)
where we have chosen units such that the initial real exchange rate e' = e[1 = 0, and where we
have assumed that in the initial period 0 (before the fear of devaluation arises) the economy
is in a long-run equilibrium with f = J.r. From (9) we have e; = e[1 + .n1 - n 1 = .nf- n 1•
given e~ = 0. Inserting this along with (2 5) into (2 6) and rearranging. we get the following
Model for the Open
© The McGraw-Hill
Companies. 2005
expression lor the aggregate demand curve for period 1:

AD curve for period 1: (2 7)
In period 0 when there is no expectation of a future devaluation and the economy is in a

long-run equilibrium with e~ 1 = z = 0, it follows from (7) that the equation for the AD
curve is simply n 0 = n f- (1/{3 1)(y 0 - [J) . Comparing this expression to (27). we see that
the aggregate demand curve will shift downwards in period 1 when expectations of a future
devaluation emerge, because the fear of devaluation pushes the domestic real interest rate
above the foreign real interest rate. The downward shift in the AD curve between periods
0 and 1 is illustrated in Fig. 24.8.
Since the devaluation does not occur until period 2, it does not allect the short-run
aggregate supply curve for period 1, because the position of this curve depends only on the
expected increase in prices from period 0 to period 1. The reason L~ that wages are reset
each period, so wage setters do not have to worry about the expected inflationary impact
of the devaluation already in period 1, since this ellect can be accounted for when wages
are reset at the start of period 2. 3 Hence we see from Fig. 24.8 that the downward shift in
the AD curve will cause output in period 1 to fall to the level y 1, while inflation in period 1
will fall to n 1• In other words, when the private sector starts to expect a future devalua-
tion, the .fear of devaluation will push the economy into recession because it feeds into the
domestic real interest rate .
.'ll
LRAS
SRAS 0 = SRAS 1
AD0
AD 1
Y1 y
Figure 24.8: The anticipation effect of a devaluation
3. We assume that wages and prices for period 1 are set at the start of the period right before the expectations of a
future devaluation emerge, so wage setters do not take into account that the anticipated devaluation w ill influence
domestic inflation already in period 1 by affecting the position of the AD curve for that period.
Stage 2: The implementation ejfects of an anticipated devaluatiOI'I

Suppose now that the domestic currency is actually devalued by the amount 6 e 2 at the
beginning of period 2, right before wages and prices for that period have to be set. and just
before the investment decisions lor period 2 are made. 4 Since workers and llrms are thus
able to form their expectations for period 2 alter the magnitude of the devaluation has
become known, the setting of wages and prices lor period 2 will be based on an expected
inflation rate equal to:
(28)
In the absence of supply shocks (s = 0) the relationship between domestic inllation and
output is given by the SRAS curve. n 2 = n; + y(!h - y). Inserting (28) into this expression,
we get the SRAS curve for period 2:
(29)
Compared to the SRAS curve lor period 1. Eq. (29) contains the additional term 8 2 6 e 2 .
This shows that the short-nm aggregate supply curve will slr!ft upwards in the period when the
devaluation occurs, as illustrated in Fig. 24.9.
But what happens to the aggregate demand curve? The AD curve for period 2 is given
by the general expression:
Y2 - !J = fJ 1 e~ - f3k2 - r1). (30)
Accounting for the eflect of the devaluation. the real exchange rate for period 2 is:
(31)
We assume that the implementation of the devaluation in period 2 eliminates any fears of
a further devaluation in the near future. When the exchange rate is expected to stay con-
stant between periods 2 and 3, perfect capital mobility ensures that i 2 = jf, If investors
expect that the devaluation which occurred in period 2 will continue to have some impact
on domestic inflation in period 3 so that .n~ = rrf + 8 3 Lle2 (where 8 3 could be zero), the
domestic real interest rate for period 2 will then be:
(32)
Substituting (31) and (32) into (30), we obtain the AD curve for period 2:
.n2 = n1 + e~ + Lle2 - (1/{J 1)(y2 - y) + ((3283/f3t)Lle2. (33)
Comparing (33) to the AD curve lor period 1 given in (27), recalling that the AD curve for
period 0 is .n0 = .nf- (1/(3 1)(y0 - y), and noting that e~ = e[l + .nf- rr 1 = .nf - n 1 > 0 (since
domestic inflation fell below .nf in period 1), we see that tire devaluation will cause the AD
curve in period 2 to shijt upwards to a position above the AD curves for periods 0 and 1, as shown
in Fig. 24.9. There are two reasons for this. First of alL the temporary drop in domestic
inflation in period 1 as well as the devaluation in period 2 improve the international
competitiveness of domestic producers, thereby inducing domestic and foreign consumers
4. A lternatively, we might assume t hat the devaluation occurs right after the decisions on wage and price setting
and investment for period 2 have been made. This would slight ly complicate our analysis but would not c hange our
qualitative conclusions.
Model for the Open
© The McGraw-Hill
Companies. 2005
LRAS
SRAS0 = SRAS1
f
Jl
Y2 y
Figure 24.9: The implementation effects of a devaluation w hic h is partly antic ipated
to substitute domestic for foreign goods. This expansionary effect is captured by the term
e~ + ~e 2 in (33). Second, both the elimination of the fear offuture devaluation and the
expected inflationary ellect of the devaluation cause the domestic real interest rate to fall.
The vanishing fear ofli.1ture devaluation is reflected in the fact th at the AD curve for period
2 does not include the negative term - (/J 2 r!J/fJ 1)(1 - 8 2 )!l e 2 which appeared in the
AD curve for period 1. and the impact of higher expected domestic inflation on the real
interest rate in period 2 is captured by the term (/3 2 8 3/(J 1 )~ e2 in (3 3).
Although the SRAS curve also shifts upwards in period 2, the effect of the devaluation
on the AD curve is sufllciently strong to ensure that output increases above the trend level
in period 2, given the assumptions we have made. To prove this, you may solve (29) for
n f- n 2 and insert the resulting expression into ( 3 3) to get:
(34)
e;
Since and M 2 are both positive. and since 0.;; 8 2 , 8 3 .;; 1 by assumption, it follows from
(34) thaty 2 - !J > 0 . as illustrated in Fig. 24.9 .
The basic insight from this complex analysL~ is that anticipated devaluations may
generate a 'bust-boom' movement in the domestic economy. Before the adjustment of the
exch ange rate, when fears of a future devaluation build up, the impact on the economy is
likely to be contractionary. as the perceived risk of devaluation and the associated capital
outflow drives up the domestic real interest rate. But when the devaluation has occurred.
the economy will expand. as domestic competitiveness improves and the domestic real
interest rate falls (assuming that the exchange rate adjustment does not foster
expectations of further imminent devaluations).
Stage 3: The longer-term adjustme11t to all anticipated devaluation

While the devaluation will stimulate the economy in the period in which it occurs, its
ellects in the subsequent periods depend on the exact way in wh ich the AD curve and the
SRAS curve will shift over time. After the devaluation (in period 2), we see from Fig. 24.9
that the domestic inflation rate is higher than the foreign inflation rate. This will gradually
erode the domestic economy's initial competitive gain from the devaluation. Hence the
AD curve will gradually shift downward due to a falling real exchange rate e' . Ceteris
paribus. this will tend to push domestic output and inflation back towards their original
levels. 'fj and n f. As domestic inflation falls due to falling aggregate demand, and as the
devaluation becomes an event of the past. firms and households will most likely reduce
their estimate of the effect of the devaluation on next period's domestic inflation rate. This
fall in expected inflation will cause the SRAS curve to shift dm.vnwards, back towards its
original position. At the same time the lower expected rate of inflation and the associated
rise in the domestic real interest rate will push the AD curve further down. As a result of
these shifts in the two curves, the domestic inflation rate will continue to fall back towards
the foreign inflation rate. and the downward-shifting AD curve will help to pull output
back towards the natural rate.
Thus, although the devaluation does indeed raise the real exchange rate e' in the
short run. it will be followed by periods in which the domestic inflation rate exceeds the
foreign inflation rate, and this situation will continue until the real exchange rate is back
at its original level. Depending on the specific dynamics of expected domestic inflation, one
can even imagine that output and inflation may fall below their original levels lor a while
before the economy ends up in the original long-run equilibrium.
A .final word o.f'caution

Although plausible, the devaluation scenario described above is not the only possible one.
The exact short-run effects of a devaluation will depend on the speciHc way in which
private sector expectations are all'ected, and this in turn may depend very much on the
specific h istorical and political context. In Exercise 4 you are invited to consider another
possible devaluation scenario involving dillerent ellects on our expectations parameters¢>
and B.
Empirical illustration: the EMS crisis of 1992-93
The dramatic crisis in the European Monetary System !EMS) in 1992- 93 provides some
empirical support for our stylized analysis of an anticipated devaluation. Under the EMS
system introduced in 19 79 , the participating EU countries had obliged themselves to keep
their bilateral exchange rates within a band of ±2.5 per cent around the fixed exchange
rate parities. Germany also participated in this arrangement, but the independent German
central bank (Bundesbank) had indicated from the beginning that it would not com-
promise on its strong historical commitment to maintain a low German inflation rate. 5 As
a consequence, German monetary policy continued to be directed mainly towards the
5. In the aftermath of the two world wars of the twent iet h century, Germany had experienced two episodes of devas-
tating hyperinflation. Because of this traumatic experience, German voters and policy makers have given high prior·
ity to the goal of price stability in the post-war period.
Model for the Open
© The McGraw-Hill
Companies. 2005
goal of domestic price stability, so the EMS system eflectively implied that the other EU
countries pegged their currencies to the D-mark and hence had to subordinate their
interest rate policy to the policy followed by the Bundesbank. For a while during the
1980s this arrangement worked quite well in the sense that the other EU countries were
able to bring down their rates of inflation signiHcantly by voluntarily subjecting them-
selves to the German monetary discipline (see Fig. 24.1b above). Indeed. around 1990
Finland, Norway and Sweden all decided to unilaterally peg their currencies to the ECU (a
unit of account consisting of a basket ofEU currencies) in an eilort to import the low rate
of in flation prevailing in the EU area.
However. the reunification of East and West Germany in late 1990 initiated a strong
construction boom concentrated in the former DDR and led to massive debt-financed
income transfers from the German federal government to the new liinder in the East. Not
surprisingly, this fiscal expansion tended to create excess demand in the German economy
and induced the Bundesbank to raise its leading interest rate in order to stop German in fla-
tion from rising. This tightening of German monetary policy forced the other member
countries in the EMS system to raise their interest rates to defend their exchange rates
against the D-mark even thoug h their economies were not exposed to any expansionary
forces like those felt by the reunifying Germany. As a result of having to mimic the tight
German monetary policy, the rest of the EU therefore fell into a recession which deepened
as the high German interest rates persisted.
During the summer of 1992, a growing number of international financial market
participants began to doubt that this situation would remain politically acceptable to the
non-German members of the EMS. Speculations arose that several EU countries would be
tempted to devalue their currencies against the D-mark in order to escape from the reces-
sion. These speculations were strengthened when Danish voters rejected the Maastricht
Treaty on European Monetary Union in June 199 2, and when French voters only barely
accepted the Treaty with a very narrow majority in September of the same year. Given this
scepticism, would European politicians remain committed to the creation of a monetary
union with completely lixed exchange rates, or would they fall back on their earlier prac-
tice of using devaluations as a temporary relief during economic recessions?
As doubts about the policy mal<ers' commitment to fixed exchange rates were
growing, the currencies of most El'viS member countries came under violent speculative
attacks in September 199 2. Within a few days Finland. Italy and the United Kingdom had
to drop out of the EMS system and allow their currencies to float. and the Spanish peseta
had to be devalued. A further devaluation of the peseta and the Portuguese escudo fol-
lowed in November. and Sweden and Norway had to move to a floating exchange rate in
November and December 1992 , respectively. The spring ofl993 saw yet another round
of devaluations of pesetas and escudos as well as a devaluation of the Irish pound. In the
summer of 1993 the crisis culminated in a massive attack on practically all the remaining
currencies in the EMS. so on 2 August the EU llnance ministers felt forced to widen the
exchange rate band around the EMS parities to ±15 per cent, effectively giving up fDced
exchange rates lor the time being.
The diagrams in Fig. 24.10 show the evolution of short-term interest rates. nominal
exchange rates against the D-mark and industrial production from month to month in a
number of European countries in the period from the start of 19 92 until the end of
1994.
24 THE O P EN ECONOMY WI TH FIXED EXCHANGE RATES 755
Denmark
17 110
~
0
15 ,,.,, Exchange 105 ;;::
' ',,,_______.... --,___
<li ::.:: c
/ \ rate
ci. g~
__ , ,.
I
,g 13 I Q) ::J
100 ~-g
0 I
~ 11
Q)
C)~
c «!
~
0:::
95
CU·i::
0 ~1;;
~
9
(Q ::J
(3 Q)"'
(ij . ~
c0 90 c-
~
·- 0
7 E x
C"'
0 Q)
E
(') -o-
c
5 85 )(
Q)
"0
.E
0 80
'<t ['- 0 '<t ['- 0 '<t ['- 0
E E E ~
E E E E E E ~
'<t '<t '<t

"'a>a> "'a>a> "'ma>E E E
(') (') (')
"'a>a> a>
a>
a>
a>
a>
a>
(')
a>
a>
a>
a>
a>
a>
a>
'<t
a>
m m
France
16 104
Exchange
rate ~
1 -,.~,
0
14
I \ 102 -.
--
I
I \
\ 0::: c:
u. 0
<Q
a.
'$ 12 ,.,,_ __,
: I
\ ',_,...'"' ~--- .... ... _, 100
~·+='
Q)
-('0"0
::J
u
I
of ~ 0
Q) ~
~ a.
98 C)
co;
0::
0 10 ~·s
u (/)
96 Ol"'
(Q )( ::J
il: (ij.~
c0 8 c-
~
·- 0
94 E X
C"'
0 Q)
E
(')
-o-
c:
6 92 X
Q)
"0
.E
4 90
~ '<t ['- 0 ~ '<t ['- 0 '<t ['- 0
E E E E E E E E
..;-
E ~
E (') (') (') E '<t ..;- E

"' "' "' "'a>
a>
a>
a>
a>
a>
a>
a>
a>
a>
a>
a>
a>
(')
a>
a>
a>
a>
a>
a>
a>
..;-
a>
a> a> a>
Figure 24.10: The behaviour of interest rates, exchange rates and output before and after the EMS
crisis
We see that. before the devaluations/switches to floating exchange rates, all countries
had to raise their interest rates sharply in reaction to mounting expectations of a future
devaluation/depreciation of their currencies. The high interest rate policy prevented output
from growing and even forced a decline in industrial production in several countries. When
the previous exchange rate parities were abandoned and the currencies were allowed to fall.
interest rates could be lowered, and output started to increase in all countries. This is in
accordance with our graphical analysis of an anticipated devaluation in Fig. 2 4 .9.
© The McGraw-Hill
Companies. 2005
United Kingdom
12 125
Interest
11 120 ~
,,,--.... ___ Cl
(ilc
c<i 10
ci. ,-' ~o
,-" 11 5 Q) ·;;
- 0
'$ 9 "'
~'U::J
<1i Q) 0
11 0 C!)~
~ 8 ca.
0:
0 7 105 -5~ )( (/)
CD Q) ::J
::::; 6 (ij-g
C·-
100
~
'E
0 5
Eo
0 )(
E c Q)
95 -0 'U
(') c
4 x-
Q)
90 'U
3 .E
.... ....
-- .... .... .... ....
2 85
0 0 0
E
C'<
<j)
E
C'<
0>
E
C'<
<j)
E
C'<
E
(')
<j)
E
(')
0>
E
(')
<j)
E
(')
....
E
<j)
....E
<j)
E
....
0> ....
E
"' (])
"'
(])
0> "' (])
"' 0>
<j) "' 0> 0> 0>
<j)
Finland
18 145 ~
Cl
c<i
16 135 ~
~c
ci. ~ 0
'$ 14 ~·<3
125 ~ ::J
<1i Q)'U
-o
~
0:
, "'
~ ~
a.
12 I 115 ~~
0 I
CD I
I ~~
::::; 10 ~ ::J
UJ
I
I 105 ~]
J:
~
---------J <UO
'E 8 95 E .,
.£X
0
E o"
c£
0
(')
6 85 )(
-
Q)
'U
- -
4 75 .E
.... .... 0
- .... .... 0 .... .... 0
E
C'<
<j)
E
C'<
0>
E
C'<
<j)
E
C'<
E
(')
<j)
E
(')
<j)
E
(')
<j)
E
(')
....E
0>
....E
<j)
....E
0> ....E
<j) 0> <j) <j) <j) <j) <j) <j) 0> <j) <j) <j)
<j) <j) <j)
The vulnerability of a fixed exchange rate regime
The EMS crisis in 1992-93 illustrates that a fixed exchange rate regime coupled with free
capital mobility is vulnerable to speculative attacks. When financial investors believe that
some currency X may soon be devalued, they have an incentive to borrow a large amount
in this currency, exchange it into some other currency and convert the funds back into
currency X at a later time when a devaluation may have occurred. If currency X has
actually been devalued in the meantime. this transaction will generate an exchange rate
Sorensen-Whitta- Jacobsen: I Part 7- The Short-run 24. The open economy w~h © The McGraw-Hill
24 THE O PEN ECONOMY WI TH FIXED EXCHANGE RATES 757
Sweden
20 145
~
0
18 ,,..... __ '"'
eg
Exchange 1 -
135 0!
---
...
.; I \ rate I I ',
ci. 16 / "''...--1I
/
/ \
' ....,........ _.,
I
:.:::·..=
<f. 125 ~
Q) ::>
u
~e
~ 14
115 ~~
Q)Q.
0::
0 12 ('I) ~
.<:'li)
CD u ::>
F 105 ~-g
(/)
10 m:.:
c0
.<: c 0
95 'E X
E 8 0 Q)
c'O
c
(')
0-
6 85 X
Q)
'0
£
-
4 75
0 0 0
E
~
<:1'
~
E "'E
~ E
E
(')
<:1'
E
(')
"'E
(') E
E
<:1'
<:1'
E
<:1'
"'E
<:1' E
0> 0> 0> ~ 0> 0> 0> (') 0> 0> 0> <:1'
0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0>
OJ OJ OJ
Italy
20 150
~
18 140 0
Exchange
/_,
,, ,. ~ c
rate :.:J 0
a. 16
oi
<f. /
/
.... ---........_,....,,_/"' /
/
130
-·~
.,u
-<'G'O
::>
0
14
~
/
<lS Q) ~
O)Q.
~ 120
Q; 12 ~~
""'(ij
u(i;
X ::J
110 Q)'O
(ij .£
E
>- 10 .£ 0
Q)
c
0 100 E
0 Q)
X
::E 8 c'O
-0 -
c
6 90 X
Q)
'0
- --
£
-
80
<:1' 0 <:1' 0 <:1' 0
E
~
E
~
"'E
~ E
E
(')
E
(')
"'E
(') E
E
<:1'
E
<:1'
"'E
<:1' E
0> 0> 0> ~ 0> 0> 0> (') 0> 0> 0> <:1'
0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0>
0> 0> 0>

Sources: lncustrial production: IMF International Financial Statistics; exchange rates: Bank of England; IBOR·rates:
various central banks, for Italy the source is IFS.
gain, and if no devaluation occurs, it will only generate a small transaction cost. Thus
speculation against a fixed exchange rate is virtually a one-way bet: you have a chance of
scoring a large gain at the risk oflosing very little.
An expectation that currency X could be devalued may therefore generate an
extremely large additional supply of X. imposing a strong downward pressure on the
exchange rate (an upward pressure on E). To be sure, the central bank issuing currency X
Model for the Open
© The McGraw-Hill
Companies. 2005
may try to reduce the attractiveness of speculation by raising its interest rate and thereby
increase the cost of borrowing in currency X. However, if the market believes that a large
devaluation is just around the corner, it may take an extremely high interest rate to ollset
the expected gain from speculation against the currency. For example. on 16 September
1992 the Swedish central bank raised its overnight lending rate to 500 per cent in an
attempt to ward oil' a speculative attack on the Swedish krona. Very lew governments can
live with the economic and social consequences of such exorbitant interest rates for very
long. 6 This is why a speculative attack may become selrJi.tljillillg in the sense that it may
force a government to let the currency fall even if policy makers would not have devalued
if the attack had not occurred. The sustainability of a llxed exchange rate regime is there-
fore crucially dependent on the credibility of the government and the central bank. This
credibility in turn depends on the ability of policy makers to avoid large macroeconomic
imbalances which are seen by the market as providing a temptation to devalue the
currency. The historical experience shows that a speculative attack can occur very
suddenly once the markets start to doubt a government's political will to defend its
exchange rate.
The vulnerability of a fixed exchange rate regime to speculative attacks is an
important reason why many countries have moved towards flexible exchange rates in
recent years when growing capital mobility has increased the scope for currency specula-
tion. In the next chapter we shall study how the macroeconomy works under flexible
exchange rates.
20.5 .~.~~.~.~Y....................................................................................................................................................
1. In an open economy with credibly fixed exchange rates and perfect capital mobility monetary
policy is impotent in the sense that the domestic interest rate has to follow the foreign interest
rate as a consequence of the condition for uncovered interest parity. Despite the loss of
monetary autonomy a country may nevertheless choose a fixed exchange rate regime to
reduce the uncertainty associated with exchange rate fluctuations. By pegging the domestic
currency to the currency of a country with a history of low inflation, domestic policy makers
may also succeed in bringing down the expected and actual domestic rate of inflation.
2. If exchange rates are credibly fixed, private agents are likely to realize that the domestic infla·
tion rate cannot systematically deviate from the foreign inflation rate for an extended period of
time. Our AS·AD model for the open economy with fixed exchange rates therefore assumes
that the expected domestic inflation rate is tied to the foreign inflation rate. On this assump·
tion the economy's short-run aggregate supply curve will not shift during the adjustment to
long-run macroeconomic equilibrium. Instead adjustment will take place through shifts in the
aggregate demand curve as the international competitiveness of domestic producers
changes over time whenever the domestic inflation rate deviates from the foreign inflation
rate.
6. In September 1992 the Swedish central bank did, in fact, win the first battle against the speculators, but on 19
November of the same year a renewed speculative attack forced the bank to abandon its fixed exchange rate and
allow a sharp d~p rec iat ion of the krona
24 THE O PEN ECONOMY WITH FIXED EXCHANGE RATES 759
3. A temporary fiscal expansion will generate a boom-bust cycle: At first output expands, but
when the fiscal stimulus disappears the economy falls into a recession because of the loss of
international competitiveness incurred during the previous expansion. The economy then
gradually recovers from the recession as international competitiveness is gradually restored
through a low domestic rate of inflation.
4. A systematic countercyclical fiscal policy will reduce the short-run output fluctuations
generated by exogenous shocks to aggregate demand and supply, but it will also reduce the
economy's speed of adjustment towards long-run equilibrium, since a countercyclical fiscal
policy rule implies an automatic tightening of fiscal policy as the economy recovers from a
recession, and an automatic relaxation of fiscal policy when the economy starts to move back
towards a normal activity level following a boom.
5. An unanticipated devaluation will temporarily stimulate domestic output, but in the long run
the economy settles in an equilibrium where all real variables are unaffected by the devalu-
ation. Hence a devaluation is neutral in the long run because the initial gain in international
competitiveness is gradually lost again due to the higher domestic inflation rate following the
devaluation. However, in a domestic recession, an unanticipated devaluation may speed up
the adjustment to a normal activity level.
6. In practice, a devaluation is often (partly) anticipated.ln that case it will generate a bust-boom
cycle: before the devaluation the economy will be pushed into recession because the fear of
a future devaluation generates a capital outflow which drives up the domestic real interest
rate. When the devaluation occurs, output expands, even though domestic inflation also rises,
in part because of a higher expected rate of inflation. Over time, the higher inflation rate
gradually eliminates the gain in domestic competitiveness stemming from the devaluation, and
the economy returns to the original long-run equilibrium.
7. A fixed exchange rate reg ime is vulnerable to speculative attacks in a world of high capital
mobility. The vulnerability stems from the fact that currency speculation is virtually without any
risk under such a regime. The historical experience shows that a speculative attack can occur
very suddenly once financial markets start to doubt a government's commitment to defend its
fixed exchange rate.
24.6 Exercises
Exercise 1. Issues in the theory of the open economy with fixed exchange
rates
1. Explain why monetary policy is impotent in an open economy with perfect capital mobility and
(truly) fixed exchange rates. Discuss some reasons why a country might nevertheless want to
adopt a fixed nominal exchange rate.
2. Explain the basic macroeconomic adjustment mechanism under fixed exchange rates and
compare to the basic adjustment mechanism in the closed economy.
3. Explain why a devaluation is neutral in the long run.
4. Explain why a system of fixed but adjustable exchange rates may lead to anticipation effects
in advance of a devaluation. Explain the nature of these anticipation effects and how they are
likely to affect the economy.
Model for the Open
© The McGraw-Hill
Companies. 2005
5 . Explain why a system of fixed exchange rates is vulnerable to speculative attacks. Explain
briefly what is meant by a 'self-fu lfilling' speculative attack.
Exercise 2. Temporary and permanent shocks in the open economy with

This exercise asks you to undertake a graphical analysis of the effects of temporary and per·
manent shocks, using the AS·AD model for the open economy w ith fixed exchange rates
summari2ed in Eqs (7)- (9) in the text. In all scenarios you may assume that the economy
starts out in long-run equilibrium in period 0.
1. Do a graphical analysis of the effects on output and domestic inflation of a permanent fiscal
expansion. Illustrate and explain the short-run effects as well as the economy's adjustment
over time. Explain the macroeconomic adjustment mechanism under fixed exchange rates and
make sure that your diagram indicates precisely how much the AD curve shifts from one
period to the next (indicate the position of the AD curves for periods 0, 1, 2 and 3, and use
arrows to indicate the movement of the AD curve after period 3). What happens to the real
exchange rate in the long run?
2. Illustrate the effects on output and domestic inflation of a temporary negative demand shock
which lasts for one period. Explain the short· run effects as well as the economy's adjustment
over time. (Hint: it may be useful if you go through the analysis in Fig. 24.3 once again.)
3 . Illustrate the effects on output and domestic inflation of a temporary positive supply shock
which lasts for one period. Explain the effects in the short run and over time. Do the effects of
the shock go away as soon as the shock has died out?
4. Suppose now that the positive supply shock is permanent. Illustrate and explain the evolution
of the economy from the initial long-run equilibrium to the new long-run equilibrium. What
happens to the real exchange rate in the long run?
Exercise 3. The effects of a global recession
You are now asked to use the AS· AD model of the small open economy w ith fixed exchange
rates to analyse the domestic effects of an international recession. In Questions 1- 3 you may
assume for simplic ity that the g lobal recession only affects foreign output, y', and the foreign
inflation rate, n 1, but not the foreign real interest rate, r1• In all q uestions you may also assume
that the domestic economy starts out in long-run equilibrium in period 0.
1. Suppose that the foreign economy is hit by a negative demand shock wh ich generates an inter·
national recession in period 1. The recession lasts for one period only (which may be thought
of as a year), so all foreign macroeconomic variables return to their original values from period
2 onwards. Use the AS·AD diagram to illustrate how the domestic economy is affected over
time by the temporary global recession. Explain the effects in period 1 as well as the subse·
quent adjustments. (Hint: start by explaining how y 1and n 1 are affected by the global recession.)
2. Suppose alternatively that the international recession lasts for two periods so that y 1 and n 1
only return to their original values from period 3 onwards. Illustrate and explain the dynamic
effects on the domestic economy in this scenario and compare w ith the scenario in
Question 1 where the recession lasted only one period.
3. Now assume instead that the international recession is triggered by a negative supply shock
which occurs in period 1. Suppose further that y 1 and :rr: return to their normal values from
period 2 onwards. Illustrate and explain the effects of the global recession on the domestic
economy in the short run and over time. Compare to the effects found in Question 1. (Hint:
start by explaining how y' and :rr 1 are affected in period 1 when the international recession
originates from a negative supply shock.)
4. Discuss briefly how the answers to Questions 1-3 are likely to be modified if we allow for the
possibility that the international recession may affect the international real interest rate, r', and
the confidence variable e. (Hint: how are foreign monetary policy and domestic confidence
likely to react to the recession?)
Exercise 4. The effects of a fully anticipated devaluation
In this exercise you are invited to study the effects of a devaluation in the borderline scenario
where the devaluation which occurs in period 2 is fully and correctly anticipated already from
period 1, and the private sector expects that the rate of domestic inflation in period 2 will rise
by the same amount as the percentage rate of devaluation (and that domestic inflation will fall
back to the initial level :rr' from period 3 onwards).
1. Use the AS-AD diagram for the open economy to illustrate the effects of a fu lly anticipated
devaluation. (Hint: what are the values of the parameters</> and 8 in this scenario?) W ill the
devaluation have any effects on real economic variables? W ill there be any difference
between the short-run and the long-run effects of the devaluation? Discuss the realism of the
assumptions underlying this scenario.
Exercise 5. The open economy with fixed exchange rates and rational
expectations
The model in the main text of this chapter assumed a form of 'weakly' rational expectations
where the expected domestic inflation rate is anchored by the foreign inflation rate, :rr• = :rr 1,
provided the exchange rate is credibly fixed. In this exercise we assume instead that expect-
ations are strictly rational in the sense defined in Chapter 2 1. Thus the expected domestic
inflation rate, :rr~1_ 11 is the expected inflation rate for period t predicted by our AS-AD model,
given all the information available up until the end of period t- 1. The model of the small open
economy with fixed exchange rates consists of the following equations, where the stochastic
demand and supply shock variables are white noise, and where the foreign inflation rate, :rr 1, is
assumed to be constant:
Goods market equilibrium: E[z 1] = 0, E[zn = a;, (35)

SRAS: :rr, =:rr~1_ 1 + y(y,- y) + s,. E[s,] = 0, E[sn =a;, E[z,s,] = 0. (36)
Real exchange rate: e; = e;_1 + :rr 1
- :rr 1• (3 7)
1. Show that under rational expectations the solutions to the model (35)- (37) are given by:
z, - {J 1 s 1
(38)
y, - y = 1 + yfi 1 '
' ' s, + yz,

:rr,=:rr + e,_1+ - -{J- . (39)
1 +y 1
Model for the Open
© The McGraw-Hill
Companies. 2005
{Hint: solve the model by the procedure outlined in Section 2 of Chapter 21 ). Discuss whether
this model has plausible persistence properties? {Hint: does the model display persistence in
output?)
Now suppose instead that a fraction A of the population has 'weakly' rational expectations,
expecting the domestic inflation rate to equal the foreign inflation rate, while the remaining
fraction has strictly rational expectations. The average expected inflation rate is then g iven as:
0 <A< 1. {40)
and the SRAS curve becomes:
{41 )
where n~ is g iven by {40). The agents w ith rational expectations form their expectations for
period t knowing that the structure of the economy is g iven by the equations {35), {3 7), {40)
and {4 1), and using all information available up until the end of period t- 1.
2. Show that the rate of inflation expected by the 'strictly rational' part of the population is:
n 1•1
0 ' ( y/3 , ) ,
_ 1 = n + A+ y{J , e,_1• {42)
3. Show by using {42) that the output gap may be w ritten as:
{43)
Use this resu lt along w ith {3 7), {40), {41 ) and {42) to show that the output gap y, "' y,- y
evolves according to the d ifference equation:
{44)
{Hints: use {40), {41 ) and {42) to find an expression for n 1 -n1• Use this expression along with
{43) to rewrite {37) in terms of j ,. 1, y, and the shock variables z,. 1 , z,and s,. ,. Then collect
terms to get {44)).
4. Does the model with 0 <A,< 1 have more plausible persistence properties than the model w ith
strictly rational expectations where A = 0 ? How is the economy's speed of adjustment to long·
run equilibrium affected by a higher degree of rationality in expectations formation {a higher
value of A) ? Try to g ive an economic explanation for your answer.
Exercise 6. Simulating an AS-AD model for an open economy with fixed

exchange rates and a mixture of backward-looking and forward-looking
expectations
This exercise invites you to undertake computer simulations with the following generalized
version of our model of the small open economy w ith fixed exchange rates, where we apply
the usual notation:
Goods market equilibrium: {45)
SRAS: {46)
Expectations: .n:~=qm:'+( 1- cp)n,_ 1 , (47)

1
Real exchange rate: e;= e;_,+.n: - .n: 1. (48)
Equation (47) gives the average expected domestic inflaton rate, assuming that a fraction cp
of the population has 'weakly rational' expectations which are anchored by the foreign infla-
tion rate, .n:1, whereas the remaining fraction of the population has static expectations, expect-
ing that this year's inflation rate will correspond to the inflation rate observed last year, .n: 1_ 1 •
In the special case where cp = 1, we obtain the basic model analysed in the main text of this
chapter.
1. Discuss briefly whether Eq. (47) is a plausible specification of expectations. (Hint: for this
purpose you may want to go back to the last subsection of Section 3 in Chapter 21).
2. Show by means of (45) and (47) that the equation for the aggregate demand curve is:
_
.7r I - Jl
1
+ ({3i31) ,
I e1- 1 -
( i311 )(JI - .9- zI) ' (49)
Before implementing the model on the computer, it is useful to reduce it to two difference
equations in the output gap and the inflation gap. Defining y, "' y y and n 1"' n 1 -
1
- .n: 1, the
model consisting of (46) - (49) may be summarized as:
AD: (50)
SRAS: .ii1 = ( 1 - cp).ii1_1 + yy1 + s, , (5 1)

Real exchange rate: e; = e;_, - .ii,. (52)
From (50) and (52) it follows that:
(;Jr -y,_,-
= e~_, - e;_ 2
.ii, - .ii1_, = (~:)~ - 1 (z1-z1_,)]. (53)
while (51) implies that:
.iii- fr,_, = -r:p.iil- 1 + yy, + sl . (54)
It also follows from (51) that
(55)
3. Equate (53) and (5 4) to find an expression for.n, and substitute the resulting expression into
(5 1) to show that the model may be reduced to the following second-order difference
equation in the output gap:
91+2- a,yl+t + aoYt = f3(zt+ 2- zt+ ,) - /3(1 - r:p)(zl+' - zl)- fi,f3sl+2 + f3(iJ ' - /3,)st+ ,' (56)
1
/3, ___ ' a1 "'/3[2- cp+y(ft ,- /3,)], a0 "'/3(1 - r:p).
1 + y/3,
Furthermore, insert (55) into (53) and show that the model may alternatively be condensed to
the following difference equation in the inflation gap:
(57)
Model for the Open
© The McGraw-Hill
Companies. 2005
Verify that (56) collapses to Eq. (15) in the main text in the special case where cp = 1. (Hint:
remember to use the definition of p 1
.)
You are now asked to simulate the model (56) and (5 7) on the computer, say, using an
Excel spreadsheet. Recall that:
/31 ;a {31 - fJ 2( 1 - cp).
Using the definitions g iven in the appendix to Chapter 23, and assuming that trade is initially
balanced and that the marginal and the average propensities to import are identical, one can
show that:
m(IJx+ 17M -1) -17o(1 - r)

{3 1 = 1-Dy+m '
where m ,. Mof''Y is the initial ratio of imports to GOP, 17 x and 17M are the price elasticities of
export and import demand, r ,. (Y- D)/Y is the net tax burden on the private sector,
17 0 ,. -(aDj() E')(E'I D) is the elasticity of private domestic demand w ith respect to the real
exchange rate {reflecting the income effect of a change in the terms of trade), and Dr is the
private marginal propensity to spend.
Under the assumptions mentioned above, one can also show from the definition of {3 2 in
the appendix to Chapter 23 that
17( 1 - r)( 1 - m)
{3 2 "' 1 - Dy+ m '
where 17 is the change in private demand induced by a one percentage point change in the
real interest rate, measured relative to private d isposable income (we already encountered
this parameter in Chapter 19). G iven these specifications, your first sheet should allow you to
choose the parameters y, cp, m, 17 x• 17 M• r , 1], 17 0 and Dr to calcOJiate the auxiliary variables {3 1 ,
fJ 2 , /3 1 and {3 (from which your spreadsheet can calculate the coeffic ients a 1 an d a 0 in (56)).
You should construct a deterministic as well as a stochast c vers ion of the model. In the
deterministic version you just feed an exogenous time sequence of the shock variables z, and
s, into the model. In the stochastic version of the model, the shock variables are assumed to
be given by the autoregressive processes
0 .;;; o< 1, x, - N(o,a;), x, i.i.d. (58)
s,..., = ws 1 - cl+ 1 , 0 .;;; w < 1, c, - N(O, a~), c, i.i.d. (59)
so your first sheet should also allow you to choose the autocorrelation coefficients c) an d w
and the standard deviations ax an d a c· We suggest that you simulate the model over 1 00
periods, assuming that the economy starts out in long· run equilibrium in the initial period 0 (so
that all of the variables y,ir, e', x, c, z and s are equal to zero in period 0 and period -1 ). From
the internet address www.econ.ku.dk/pbs/courseslmodels&data.htm you can download an
Excel spreadsheet w ith two different 1 OO·period samples taken from the standardized normal
d istribution. Choose the first sample to represent the stochastic shock variable x,, and the
second sample to represent the shock variable c 1 . To calibrate the magnitude of the shocks
x, and c 1, you must multiply the samples from the standardized normal distribution by the
respective standard deviations ax and a c ·
Apart from listing the parameters of the model, your first sheet should also list the standard
deviations of output and inflation as well as the coefficient of correlation between output and
inflation and the coefficients of autocorrelation for these variables (going back four periods)
emerging from your simulations of the stochastic model version. It will also be useful to
include diagrams illustrating the simulated values of the output gap, the inflation gap, and
(using (52i) the real exchange rate.
For the simulation of the deterministic version of the model, we propose that you use the
parameter values:
q:>= 0.1' y= 0.2, m = 0.3, YJx = I'J~.~= 1.5,
r = 0.2, 1'J = 0.5, 1'Jo = 0.3, Dr= 0.8.
4. Use the deterministic model version to simulate the dynamic effects of a temporary negative
demand shock, where z 1 = -1 in period t = 1 and z 1= 0 for all t;;. 2. Illustrate the evolution of
the output and inflation gaps in diagrams and comment on your results. Does the model
embody a theory of business cycles? Discuss. (Hint: is the adjustment to long-run equilibrium
monotonic or cyclical?)
5. Now use the detem1inistic model to simulate the dynamic effects of a temporary negative
supply shock, where s 1 = +1 in period t = 1 and s 1 = 0 for all t;;. 2. Illustrate the evolution of the
output and inflation gaps in diagrams and comment on your resu lts. Is the adjustment to long-
run equilibrium monotonic or cyclical? Illustrate the importance of expectations formation for
the economy's dynamic properties by varying the value of the parameter q:>.
6. Simulate the stochastic version of the model, setting:
q:>=O=w = 0.5,
while maintaining the other parameter values given above. From this starting point, try to
adjust the values of q:>, o, w, ax and a c so as to achieve a better match between the model-
simulated standard deviations and coefficients of correlation and autocorrelation for output
and inflation and the corresponding data for the Danish economy presented in
Tables 14.2 - 14.4. For your information, the standard deviation of the cyclical component in
the Danish inflation rate was 0.48, according to the data underlying Tables 14.3 and 14.4.
(When trying to reproduce the observed correlation between output and inflation, you should
focus only on the contemporaneous coefficient of correlation. Moreover, note that our simpli-
fied model cannot be expected to reproduce the data with great accuracy.) Assuming that
your parameter values are plausible, what are the implications of your analysis for the relative
importance of demand shocks and supply shocks as drivers of the Danish business cycle?
8 I Sorensen-Whitta-Jacobsen: I Part 7 -
The Short-run
Model for the Open
flexible exchange rates
© The McGraw-Hill
Companies. 2005
Chapter
•
•
•
•
;The
•
open economy
~ with flexible
•
•
...._____ ~ exchange rates
A.
s we have seen. a llxed exchange rate regime is vulnerable to speculative attacks
in a world with high capital mobility. In recent years many governments have
learned this lesson the hard way. We have already reported how the fJXed
exchange rate regime of the European Union (the European Monetary System) came
under heavy attack in 1992-93. A few years later, a number of other countries w hich
pegged their currencies to the US dollar were attacked by currency speculators. This
happened to Mexico in 1994. to the East Asian countries inl997, to Russia inl998, to
Brazil in 1999, and to Argentina in 2000-2001, just to mention the most spectacular
examples.
Countries have reacted in different ways to the growing vulnerability of fJXed
exchange rate regimes arising from the growing mobility of capital. Contrary to the pre-
diction of many observers at the time. the EMS crisis in 1992-93 strenghtened the resolve
of European political leaders to move towards the ultimate llxed exchange rate regime by
forming a monetary union where the use of a common currency rules out the possibility
of currency speculation.
However, during the 1990s a minority of Western European countries as well as a
large number of emerging market economies and developing cow1tries have reacted to
the growing frequency of speculative attacks against llxed exchange rates by moving
towards flexible exchange rates where currency speculation is much more risky because
there is no llxed exchange rate parity against which to speculate. Thus these countries
have followed the lead of big economies like the United States and Japan which have
allowed their currencies to float right from the breakdown of the Bretton Woods system of
Hxed exchange rates in the early 19 70s.
Against this background the present chapter studies the workings of an open
economy with flexible exchange rates. We continue to focus on a small specialized
economy, and in line with the previous chapter we assume perfect capital mobility, since
significant capital controls are nowadays very rare among developed countries.
The Hrst section of the chapter dellnes the ch aracteristics of a flexible exch ange rate
regime and describes how a number of Western cow1tries have designed their monetary
policies under this regime. This part of the chapter also highlights the role of exchange
rate expectations under flexible exchange rates. In the second section we adapt our AS-AD
766
Introducing Advanced Model for the Open flexible exchange rates Companies. 2005
25 THE O PEN ECONO MY WITH FLEX IB LE EXCHA NG E RATES 767
model to a regime with flexible exchange rates and identi(y the similarities and dillerences
in the macroeconomic adjustment process under flexible and fixed exchange rates. The
third and final section of the chapter then studies how an open economy with flexible
exchange rates reacts to demand and supply shocks and compares these reactions to those
occurring under fixed exchange rates.
25.1 Flexible exchange rates as an economic policy regime

................................................................................................................................................................................
Inflation targeting as a nominal anchor
A crucial characteristic of a flexible exchange rate regime is that it allows the domestic
central bank to pursue its own monetary policy even if capital mobility is perfect. This
follows from the condition for uncovered interest parity with which you are already
familiar:
e = In E. (1)
According to (1). the domestic central bank may set the domestic nominal interest
rate, i. independently of the foreign nominal interest rate. if, provided the bank is willing to
accept whatever magnitude of the expected percentage exchange rate depreciation,
e: 1 - e, which is necessary to make the holding of domestic and foreign interest-bearing
assets equally attractive.
Since the expected change in the exchange rate depends on the current level of the
(log of the) exchange rate e, we see from (1) that the domestic central bank can pursue a
fully independent interest rate policy only if it is willing to leave the determination of the
exchange rate completely to the forces of the market. Obviously this means that the
exchange rate cannot serve as a nominal anchor. in contrast to a fixed exchange rate
regime where pegging to the currency of a low-inflation country can help to keep the
expected and actual domestic inflation rate low.
As an alternative way of providing a nominal anchor for inflation expectations. many
countries with freely floating exchange rates have therefore adopted a monetary policy
regime of in.flation targeting. Under inflation targeting policy makers specify a target for the
rate of inflation which is considered to be consistent with an acceptable degree of price
stability. Monetary policy is then given the task of ensuring that the actual inflation rate
stays close to the target. Typically the target inflation rate is quite low, say, 2 per cent per
annum. Realizing that monetary policy cannot perfectly control the rate of inflation, infla-
tion targeting countries also specify a 'tolerance band' . that is, an acceptable range of
fluctuation of the actual inflation rate around the target rate. Indeed, some countries only
specify a target range for the inflation rate without explicitly setting a particular target
rnte.
To make the policy goal of a low inflation rate as credible as possible. most inflation
targeting countries have delegated monetary policy to an independent central bank. In
Chapter 22 we saw that such delegation may be a way of overcoming the credibility
problem arising from the possible in flation bias in monetary policy. As you recall. an infla-
tion bias may exist when policy makers with a short time horizon are tempted to stimulate
output by creating surprise inflation.
Model for the Open
© The McGraw-Hill
Companies. 2005
768 PART 7 : TH E SHO RT-RUN MO DEL FO R THE O PEN ECONOMY
More generally, the creation of an independent central bank may be a means of

convincing the public that the goal of price stability will not be compromised by a govern-
ment trying to manipulate monetary policy to its own short-term political advantage.
However, to ensure that monetary policy makers can be held accountable for their actions
by democratically elected politicians. many inflation targeting countries require the
central bank to justify its policy decisions at regular intervals in reports to the government
and/or parliament and to the general public. Through such a reporting procedure to
promote transparency in monetary policy making, it becomes easier to check whether the
central bank adheres to the goal of price stability in a satisfactory manner. This may
strengthen the credibility of the inflation target and at the same time make the delegation
of monetary policy democratically acceptable.
Table 25. L lists a number of Western countries which h ave practised inflation
targeting over the last decade. We see that all of these countries have chosen inflation
targets of 2- 2.5 per cent. with a typical tolerance band of ±1 per cent. In recent years
many disinilating countries in Eastern Europe and the developing world have also adopted
inflation targeting. These countries are in the process of gradually reducing their inflation
targets down to the level chosen by the 'best-practice' countries listed in the table.
Although the European Central Bank (ECB) has not olllcially adopted inflation
targeting, it has announced th at it aims to keep inflation in the euro area below but close
to 2 per cent. The ECB bases its monetary policy on an analysis of a wide range of
economic and llnancial variables, but the purpose of this analysis is to evaluate whether a
change in the interest rate is needed to maintain a low and stable rate of infl ation. In
several important ways the ECB ·s monetary policy regime is therefore similar to a regime
of inflation targeting.
In the USA, the ofllcial goal of the Federal Reserve System is to promote maximum
employment, stable prices and moderate long-tenn interest rates. However. at least since
the early 1980s the actions of the Fed h ave revealed that it attaches great weight to the
maintenance of a low and stable rate of inflation, so although the US is not officially an
inflation targeting country. US monetary policy also bears some similarity to inflation
targeting.
Table 25.1: Inflation targets in different countries (for 2002)

Target inflation Tolerance band
rate (%) (% points)
Australia 2.5• 2- 3
Canada 2.o• 1- 3
Israel 2.5• 2- 3
New Zealand 2.o• 1- 3
Norway 2.5 1.5- 3.5
Sweden 2.0 1- 3
United Kingdom 2.5 1.5- 3.5
• Inflation target implied by midpoint of tolerance band. Only the t olerance band is publicly announced.
Source: New Zealand Parliamentary Library: Inflation Targeting, Table 1,

p. 10. Background Note of September 30, 2002.
Monetary policy under inflation targeting
In theory one may distinguish between 'strict' and 'flexible' inflation targeting. Under
strict inflation targeting monetary policy aims exclusively at fuliHling the inflation target,
whereas flexible inflation targeting implies that monetary policy also reacts to the evolu-
tion of real output and employment. In practice, inflation targeting central banks have
tended to adhere to a 11exible regime. In Exercise 3 we will ask you to analyse such a
regime of flexible inflation targeting, but the main text of this chapter will simplify by
considering a regime of strict inflation targeting. As we shall see later, even strict inflation
targeting may reduce the volatility of output compared to a regime with fixed exch ange
rates. provided business cycles are driven mainly by demand shocks.
Specillcally, we will assume that the inflation-targeting domestic central bank sets
the interest rate in accordance with the policy reaction function:
i= ri + <] + l!(:n: - :n:*), I! > 0. (2)
:n::
where :n:* is the target inflation rate, :n: is the actual current inflation rate. and 1 is the
expected rate of inflation between the current and the next period. As before, Eq. (2)
assumes that the central bank can observe the public's expected in flation rate, say,
through consumer and business surveys. or by observing the difference between the
interest rates on indexed and non-indexed bonds. Through its control over the nominal
interest rate. i, the central bank can therefore also control the ex ante real interest rate.
i - :n:: 1• in the short run. However, the bank also recognizes that in long-run equilibrium
the domestic ex post real interest rate must equal the foreign real interest rate. rf. as we
have shown in Chapter 23 . Thus the policy rule (2) says that the central bank raises the
domestic real interest rate above its equilibrium level when the inflation rate exceeds its
target, and vice versa. Note that this is just a special version of the Taylor rule where the
coelllcient on the output gap has been set at zero.
In practice, aggregate demand and inflation only react to a change in the interest rate
with a certain time lag. Infl ation targeting central banks must therefore base their interest
rate policy on a .forecast lor the inflation rate expected to prevail one or two years ahead, as
we saw in Section 4 of Chapter 22 . and as we will explain in more detail in Chapter 26.
However, since there is some persistence in the inflation rate. a rise in the current inflation
rate, Jc. '"'ill typically increase the central bank's forecast of future inflation, thereby
triggering a rise in the interest rate, as assumed in (2 ).
Specifying the inflation target
Table 2 S.1 shows that smaller Western inflation targeting economies have chosen roughly
the same inflation target of 2 .0-2 .5 per cent. In recent years the euro area and the United
States have also experienced a rate of inflation very close to this level. This suggests that
Western inflation targeting economies tend to choose an inflation target which is equal to
the average rate of inflation in the OECD area, either because they see the foreign inflation
rate as an appropriate nominal anchor. or simply because there is a strong international
consensus on what the appropriate level of inflation is. In other words, it seems realistic to
assume that the target inflation rate is roughly equal to the foreign inflation rate :ref:
:n:* = :;cf , (3)
Model for the Open
© The McGraw-Hill
Companies. 2005
770 PART 7 : TH E SH O RT-RUN MO DEL FO R THE O PEN ECONOMY
Apart from the empirical justification, there is also a theoretical rationale for this choice of
inflation target. As you recall from Chapter 24, a long-run equilibrium requires fulHiment
of the condition lor relative purchasing power pnrity to ensure constancy of the real
exch ange rate:
(4)
In the long run an infl ation targeting country which manages to keep the domestic
inflation rate equal to the foreign inflation rate will therefore also be able to maintain a
stable nominal exchange rate, since :rt = n:l implies !:le = 0 according to (4). The inter-
national experience suggests th at although inflation targeting countries do not pursue a
specific target for the nominal exch ange rate, and although over the medium term the
exchange rate can move quite far away from its historical average, countries usually
try to avoid policies which lead to a systematic depreciation or appreciation of their
currencies over very long periods of time. Equation (3) combined with ( 4) is our way of
modelling this apparent policy preference for long-run stability of the nominal exchange
rate.
Inflation expectations under inflation targeting
Let us now consider the formation of inllation expectations under inllation targeting.
From the monetary policy rule (2) it follows th at the economy can only be in long-run
equilibrium when the inllation rate equals its target rate, since the real interest rate.
i - .n::l' will only attain its long-m n equilibrium value. rf. when n: = .n:*. In parallel to the
previous chapter. we will assume th at agents are 'weakly' rational in the sense that they
can form a correct estimate of the long-run inllation rate even though they do not have
enough information to predict the short-run fluctuations in inflation. Under inflation
targeting, agents thus understand that on average the inllation rate has to equal the
central bank's oillcial inflation target. Using (3), we therefore assume that:
( 5)
where the foreign inflation rate, .n:f , is assumed to be constant. Equation (5) is the simplest
way of formalizing our assumption that the central bank's inflation target has credibility.
It is the natural analogue to the analysis in the previous chapter where our specification
n::, = .n:' = n:f reflected the assumption that the central bank's commitment to a fixed
exchange rate was credible. 1
Exchange rates and interest rates
Consider next the formation of exchange rate expectations. From (1) we see that these
expectations are crucial for the link between interest rates and exchange rates. Whereas
1. A more general assumption than (5) would be
reflecting the idea from Chapter 24 that a fraction 'I' of the population has static expectations while the remaining
fraction has weakly rational expectations. The main text of this chapter focuses on the limiting case where <p = 0
whereas Exercise 4 asks you to explore the more general case where 0 < 'P < 1.
Introducing Advanced Model for the Open flexibl e exchange rates Companies. 2005
the central bank's otlicial inflation target provides a natural anchor for inflation expect-
ations, there is no similar policy-determined focal point lor exchange rate expectations,
since there is no official exchange rate target under freely floating exchange rates. Our
theory of exchange rate expectations is therefore based on a different line of reasoning
than the one underlying our simple theory of inflation expectations. Specifically, we will
adopt a hypothesis oCregressive · exchange rate expectations which has often been used in
the literature on the open economy. This hypothesis postulates that the exchange rate is
expected to rise if it is currently below its perceived normal level, and vice versa. If the
expected exchange rate for the next period is e~+ l and the perceived 'normal' exchange
rate is et, the hypothesis of regressive expectations thus says that:
(J > 0. (6)
In other words, agents believe that there is some normal or average level towards which
the exchange rate tends to move over time. 2 The perceived normal exchange rate is likely
to depend on the actual exchange rates observed in the past. with greater weight being
attached to the more recent observations. As a rough approximation capturing this idea,
we will assume that the 'normal' exchange rate is simply identified with last period's
observed exchange rate:
(7)
We emphasize that none of the results derived in this chapter depend crucially on this
strong simplification. In Exercise 1 we ask you to derive a very similar AS-AD model from
an alternative theory where the perceived normal exchange rate e is simply treated as
exogenous. However, the link between interest rates and exchange rates implied by (7)
does find some empirical support, as we shall see in a moment.
Inserting (7) into (6) and dropping the time subscript for the current period lor
convenience. we get:
(8)
which says that if the exchange rate was rising (falling) between the previous and the
current period, it is expected to fall (rise) over the next period. From (8) and (1) it then
follows that:
(9)
which shows that when the domestic interest rate is raised above the foreign interest rate,
the domestic currency will appreciate. and vice versa. The reason is that when i is raised
above / , domestic assets tend to become more attractive than foreign assets, other things
being equal. To maintain a capital market equilibrium where all assets are equally
attractive. it is therefore necessary lor the domestic currency to appreciate, since this will
create an expectation th at the domestic currency will depreciate over the next period (see
(8)) so that the higher interest yield on domestic assets is on:~et by an expected exchange
rate loss.
2 . We do not restrict the value of the parameter 0 in (6) to be less than 1. 1f 0 > 1, agents expect that the exchange rat e
w ill tend to overshoot its normal l evel in the short run. As we w ill explain later in this chapter, exchange rat es do in
fact tend to overshoot their long·run equilibrium values in t he short term. The regression results reported below are
also consistent w ith the hypothesis that 0 > 1, since the estimated value of o-• in Eq. (10) is smaller than 1.
Model for the Open
© The McGraw-Hill
Companies. 2005
772 PART 7: THE SHORT-RUN MODEL FOR TH E OPEN ECONOMY
Equation (9) was derived from (1) which assumes that investors are risk neutral and
that foreign and domestic bonds are perfect substitutes. But suppose that financial
investors are risk averse and that foreign and domestic bonds are considered to have
different risk characteristics because they are issued in different political jurisdictions
using dillerent currencies. In equilibrium the expected returns to domestic and foreign
bonds must therefore diller to compensate for the dillerent risk characteristics of the two
assets. Let us call this difference the 'risk premium', and let us assume that it consists of a
'systematic' component. Ii. plus a stochastic component, t, which fluctuates around a zero
mean value, reflecting random shifts in the market's evaluation of the riskiness of the two
assets. or shifts in the 'appetite' for risk-taking. Using (8), the arbitrage condition (1) then
modiHes to:
E[t] = 0 . (10)
Note that the systematic component v in the rL~k premium can be either positive or nega-
tive. For example. if the risk characteristics of domestic bonds are seen as more attractive
than those offoreign bonds- say. because historically foreign bond prices have been more
volatile than the prices of domestic bonds - v will tend to be negative. Since the mean
value oft is assumed to be zero. (10) still predicts that on average the relationship between
the interest rate difJerential i - /and the percentage change in the exchange rate !:l e
should be negative. although in periods with large temporary increases (decreases) in the
risk premium one may observe positive (negative) values ofi - /as well as !:l e.
In Fig. 2 5.1 we have plotted recent monthly observations for the percentage depreci-
ation of the bilateral exchange rate (!:l e) against the corresponding bilateral short-term
interest rate diflerential i - / for a number of countries with floating exchange rates.
According to ( lO) the observations for each pair of countries should be scattered around a
downward-sloping straight line. As one would expect given the simplistic theory of
expectations formation (7) which is embodied in (10), the negative relationship between
i - /and !:l e observed in Fig. 25.1 is far from tight. A regression analysis nevertheless
shows that the line of best flt through the observations does indeed have a negative slope
which is statistically significant for all the countries considered.
Of course, it is possible that a more sophisticated theory of expectations formation
could have given us a better flt. but the theory specified in ( 6}and (7) has the advantage of
being very simple. Hence we shall stick to it. since its prediction of a negative relationship
between the interest rate difi'erential and the change in the exchange rate seems to be
empirically correct on average. To keep the exposition as simple as possible. most of our
analysis will be based on the deterministic equation (9). but later on we will also discuss
some implications of the more general stochastic relationship (10).
'Dirty' floating
The specifications above relate to a regime of 'clean' floating where policy makers do not
in any way try to influence the market-determined exchange rate. Yet historically many
countries have practised so-called 'dirty' floating by intervening in the foreign exchange
market in order to reduce the fluctuations in the exchange rate. The exchange rate can
usually be controlled over a few days by relying on 'sterilized' foreign exchange market
interventions where the central bank buys or sells foreign exchange without changing
25 THE OPEN ECONOMY WITH FLEXIBLE EXCHANGE RATES 773
1OO*(e- e-1)
2.5 UK-Euro area
2 • Regression line:
1OO*(e- e _1) = 0.39696 - 0.29726*(i-/)
t = - 2.53
1.5 Ff = 0.1007
••
---·•--··••
•
0.5 •• •• •
~ • • . .f
0 I-I
.1
•••• ·~
0.~.· 3
• •
- 0.5
• • •• •
•
• ••
•
-1
•
- 1.5
•
-2
100*(e- e- 1)
Sweden-Euro area
"'
Regression line:
• 100*(e- e _1) = 0.07519 - 0.24766*(i-/)
.v t = - 2.47
ff = 0.0965
•
... •
•
•
Er.5- ---.
•
• ••••••
~ • •
r- • •• • • . .f
•
I-I
• ·- 1.5 •
v
.5 -1 1.5
• • ~ • • l
();5-
••
~·
• •••• •
• •
1 •
·"
Fig. 25.1 : Interest rate differentials and rates of exchange rate depreciation
llexible exchange rates
© The McGraw-Hill
Companies. 2005
1OO*(e-e_1)
,,...,
~~
UK-USA
• L
Regression line:
1OO*(e-e_1) = 0.21977 - 0.22200*(;-/)
t = - 2.65
• 1.e f?? = 0.1099
•
----.-. ·~
-·--
•
• • • • ••
- 6.5i
v
•
•• .. •• . .f
I-I
- 0.5
~
1 0.5 1
-0~- •
• •• • •• •
•
•
- ·'-'
•
- .<. •
' L,V
100*(e-e_1)
.v
Canada-USA
Regression line:
1 OO*(e-e_1) = 0.00444 - 0.25202*(;-/)
!• • • I = - 2.96
f?? = 0.1332
•
.. \T.v
•• .
• •• • •• •
~v • ;
• . -1
I-I
~
1 - 0 .5 2 2. 5
-••
•
~.5
•
• • -
• • • •
•
- .v
•
- L
- £, \.)
•

Note: Monthly data, 1999: 1 through 2003:12. The interest rates are one-month money market rates.
Sources: Bank of England; US Federal Reserve System; B ank of Canada; Sveriges Riksbank.
the interest rate. The evidence shows. however, th at to have a persistent e!Tect on the
exchange rate. foreign exchange interventions must be 'unsterilized', th at is. they must
involve a change in the interest rate which supports the movement in the exchange rate
th at the central bank is trying to achieve.
To model a policy regime where inllation targeting is supplemented by unsterilized
foreign exchange intervention to smooth lluctuations in the exchange rate. we might
modify the monetary policy rule (2) in the following way:
h > 0, A.> 0. (11)
According to (11) an observed depreciation of the exchange rate (e > e_1) induces the
central bank to raise the domestic interest rate. In this way the bank may generate a
capital inflow which will increase the supply of foreign exchange and the demand for
domestic currency, thereby moderating the depreciation. By analogy. when the bank
observes a tendency for the domestic currency to appreciate (e < e_1). it reduces the
interest rate to induce a capital outllow which will increase the supply of domestic
currency and the demand for foreign currency. Such a policy is sometimes described as
'leaning against the wind'. because the central bank goes against the tendencies in the
foreign exchange market.
In Exercise 2 we will ask you to show that the AS-AD model emerging under the
policy of dirty lloating specifled in (11) is qualitatively similar to the model implied by the
clean floating regime (2). although the quantitative properties of the model will be
allected by a policy ofleaning against the wind. In the rest of the chapter we will maintain
the assumption of clean lloating, partly because it slightly simpliHes the exposition, and
partly because many countries with flexible exchange rates have in fact moved towards
'cleaner' floating by signilicantly reducing the frequency of their loreign exchange market
interventions in recent years.
We are now ready to set up our version of the AS-AD model for the open economy with
flexible exch ange rates. After having done so. we will study the economy's adjustment to
long-run equilibrium and compare the adjustment process to the one characterizing a
fJXed exchange rate regime. Through this comparison we \<Viii gain a deeper under-
standing of the special leatures of the two alternative exchange rate regimes.
The aggregate demand curve
As you recall from Eq. (26) of Chapter 23, the goods market equilibrium condition lor the
open economy may be \.vritten as:
Y - yr = P1
a (er- 1 + !:le + nf- n) - R
P2 (l - n '+1 + e'+1 - e- r~f\J + i ' (12)
i = fJ 3(g - g)+ {3 4 (yi- r/) + (Js(ln c - In 'if),
where we have used the usual notation. To derive the AD curve under llexible exch ange
rates. we start by noting from (1) that C: 1 - e = i - / . Inserting this along with (9) into
© The McGraw-Hill
Companies. 2005
776 PART 7 : TH E SHO RT-RUN MO DEL FO R TH E O PEN ECONOMY
(12). and using the fact that/- .n: 1 = / - nf = ,.I, we then get:
exchange rate interest rate
channel channel
(13)
Equation (13) illustrates the two channels through which monetary policy a!Tects aggre-
gate demand under flexible exchange rates. One channel is the familiar interest rate
channel which also operates in the closed economy: when the interest rate rises, there is a
direct negative effect on private investment and possibly also on private consumption. The
strength of this eflect depends on the parameter {3 2 which captures the interest sensitivity
of aggregate demand. The second channel in the monetary transmission mechanism is
the exchange rate channel: when the domestic interest rate increases relative to the inter-
est rate abroad. there is a tendency for the domestic currency to appreciate. This leads to a
loss of international competitiveness which reduces net exports. The greater the value of
the parameter ,8 1 , that is. the higher the price elasticities of e>.:port and import demand, the
stronger is this eflect of monetary policy through the exchange rate channel. By increas-
ing the response of the exchange rate to a rise in the domestic interest rate, a lower value
of the parameter () will also strengthen the impact of monetary policy on aggregate
demand through the exchange rate channel.
The second step in the derivation of the AD curve is to note from (2). (3) and (5) that:
= ij'
(14)
Substituting (14) into (13) and using the definition of z, we get an expression for the AD
curve under flexible exchm-1ge rates:
y - [J = f3t e~ 1 - /3 1(n - n 1) + Z, PI=/3 1+ h({32 + e-l /31). (15)

z =- f3irl - F)+ /3 3(g - g) + {3 4(!/- yf) + f3 5 (ln t - In 'if) .
For comparison, Eq. (7) in Chapter 24 gives the analogous expression for the flD-curve
under jh.:ed e;rcl!ange rates:
(16)
We see that the AD curve h as the same structure under the two exchange rate regimes.
Under both regimes the demand shock variable z includes the same disturbances, and a
rise in the domestic inflation rate lowers aggregate demand. However. since /J 1 > {3 1' we
see from (15) and (16) that a rise in domestic inflation causes a larger fall in aggregate
demand under flexible than under fixed exchange rates.
By taking a closer look at the expression lor /3 1 in (15), we can identifY the various
effects generated by a rise in domestic inflation under a floating exchange rate. First, there
is a direct eflect as the rise inn reduces net exports by eroding the economy's international
competitiveness. This direct effect - which is the only effect arising under fixed exchange
rates - is reflected in the term /3 1 in the expression for ~ 1 . However, under floating
exchange rates a rise inn generates two additional effects stemming from the fact that a
flexible exchange rate provides scope for an independent domestic monetary policy. As
domestic inflation rises. the central bank reacts by raising the domestic interest rate. This
creates additional downward pressure on aggregate demand, partly through the interest
rate channel (rellected in the term hf3 2 in the expression for fi 1). and partly through the
exchange rate channel (captured by the term htr 1f3 1 in the definition of/3 1) . This explains
why the AD curve is }latter under tlexible than under fixed exchange rates. As a conse-
quence, the economy will respond dillerently to exogenous shocks under the two
exchange rate regimes, as we shall see later on.
Nominal and real exchange rate dynamics
As was the case under fixed exchange rates, we see from (15) that the AD curve under
llexible exchange rates includes the lagged value of the real exch ange rate. e~ 1• As long as
the economy is out of long-run equilibrium. the dynamics of the real exchange rate will
therefore cause the short-run AD curve to sh ift from one period to the next. Under llexible
exchange rates, these shifts in the AD curve are driven by the identity:
(17)
Substituting (14) into (9). we find that the dynamics of the nominal exchange rate are
given by the relationship:
(18)
which may be inserted into (17) to give the dynamics of the real exchange rate:
direct effect on monetary policy
competitiveness effect
~ ...-----...
er = e~l + nl - :n + hfr\n'- n) (19)
Thus the real exchange rate '<\Till depreciate (that is. e will rise) whenever the foreign infla-
tion rate exceeds the domestic inflation rate. Part of this real depreciation is due to the
direct efTect on competitiveness. The other part is due to a monetary policy effect: when
domestic intlation falls below the inllation target n1~ the central bank cuts the domestic
interest rate, inducing a depreciation of the nominal exchange rate.
The complete AS·AD model with flexible exchange rates
We may now sununarize our complete AS-AD model of the open economy with llexible
exchange rates. Following the previous chapter. we continue to use a specification of the
aggregate supply side based on the hypothesis of relative wage resistance. Rearranging
(15) and (19), we then end up with the following model:
n = nl + ( ~
1
AD:
1 )e~1 - )(y -y -
( ;
1
z), (20)
SRAS: n = n f + y(y - y) + s, (21)
The structure of this model is similar to the structure of the model with fixed exchange rates
(you may want to compare with Eqs (7)-(9) in Chapter 24). At the start of each period. the
previous period's real exchange rate e~ 1 is predetermined, so the current rates of output and
intlation are determined by the intersection of the aggregate demand curve (20) and the
© The McGraw-Hill
Companies. 2005
778 PART 7 : TH E SHO RT-RUN MO DEL FO R TH E O PEN ECONOMY
short-run aggregate supply curve (21), where the position of the AD curve depends on the
lagged real exchange rate, the foreign inflation rate n/ and the real demand shock variable
z. while the position of the SRAS curve depends on :ref and on the supply shock variable s.
The current level of inflation then determines the current real exchange rate through (22).
If n f * nit follows that er * e~ 1 , and the economy will then enter the next period with a new
predetermined real exchange rate. causing a shift in the position of the AD curve. This will
generate new short-run values of output and inflation during the next period which in tum
will change the level of the real exchange rate, and so on.
Figure 2 5.2 illustrates this adjustment process in a case where the economy starts
out in recession at the output level y 1 and a corresponding domestic inflation rate n 1
below the foreign inflation rate. From (22) and (20) it then follows that the AD curve will
shift upwards by the vertical distance (j3 JP1)(nf - n 1) between period 1 and period 2.
The definition of P1 given in (15) implies fJ 1/ P1 < 1. so the upward shift in the AD curve
will be smaller than the distance n f - n. as illustrated in Fig. 2 5.2. The new AD curve
generates a new short-run equilibrium in period 2 where output is !hand inflation is n 2 •
Between period 2 and period 3. the AD curve shifts upwards again by the distance
({3 1/fi 1)(nf - n ~) < n1 - n 2 , causing a further rise in output and inflation toy 3 and n 3 • and
so on, until the economy reaches the long-run equilibrium at point A . Since er = e~ 1 in
long-run equilibrium, we see from (22) that the domestic inflation rate must equal the
foreign inflation rate in the long run. As noted earlier, the central bank's inflation target
will thus be realized, and the nominal as well as the real exchange rate will stay constant.
According to (21) the condition n f = :rr: also implies that output will equal its natural rate
y = [J in a long-run equilibrium without supply shocks (s = 0).
f
1l
LRAS
Y1 Y2 Y3 Yt
Figure 25.2: The adjustment to long -run eq uilibrium under flexible exchange rates
The speed of adjustment under alternative exchange rate regimes
From the previous chapter you may recall that under fixed exchange rates the vertical shijt
in the AD curve from one period to the next equals the jitll distance n f - n . Hence one
might be tempted to conclude that the adjustment towards long-run equilibrium L~ faster
under fixed than under flexible exchange rates. However. this would be '<Vrong since we
have also shown that the slope of the AD curve is }latter under flexible than under fixed
exchange rates. It is easy to visualize from Fig. 2 5.2 that with a flatter AD curve, a given
vertical shtl't in this curve will cause output and inflation to move closer towards their
long-run equilibrium values. Thus we catmot say a priori whether the economy's speed of
adjustment is faster under llxed than under flexible exchange rates.
Let us now identify the parameters which are cmcial for the relative speed of adjust-
ment under the two exchange rate regimes. For this purpose we introduce our familiar
'gap' variables:
(23)
and '<Vrite the model (20)- (22) as:
(24)
itt = YYr + st, (25)

< e;_
= 1- (1 + he- 1
)n,. (26)
Using (25) to eliminaten, from (24) and solving for e~- 1' we get:
, (1_f3_t_
e,_ L=
+y{31)·y,+!i: - "'!i;'tJ1 st Zt
(27)
By substituting (2 7) and the corresponding expression for e~ into (26) along with the
expression for n1 given in (2 5), you may verily that the evolution of the output gap under
jlexible exchange rates is given by the following diflerence equation:
(28)
(29)
For zt+ 1 = z, = st = 0 the solution to (28) is:
t = 0. 1. 2, .... (30)
From (29) we see that the coetncient a is positive but smaller than 1. According to (30)
this guarantees th at the economy with flexible exchange rates is stable and converges
monotonically on its long-run equilibrium. as illustrated in Fig. 25 .2 above.
For comparison with (30), we recall from the previous chapter that the solution for
the output gap under .fixed exchmrge rates is:
1
{3= - - <1. (31)
1 + y{31
© The McGraw-Hill
Companies. 2005
780 PART 7 : TH E SH O RT-RUN MO DEL FO R TH E O PEN ECONOMY
The speed of adjustment to long-run equilibrium will be lower (higher) under ftxed
than under flexible exchange rates if the coefllcient {3 in (31) is larger (smaller) than the
coetficient a in (30). From the deHnitions of 11 . {3 and/] 1 we ftnd that:
(32)
The presence of the monetary policy parameter h in the numerator of (32) shows that the
dillerence in the speed of adjustment under the two exchange rate regimes arises from the
fact that only a flexible exchange rate leaves scope for an independent monetary policy. In
general (32) does not allow us to say which exchange rate regime implies the fastest
convergence on long-run equilibrium, thus conilrming our earlier conclusion. The reason
is that the macroeconomic adjustment process under flexible exchange rates includes one
mechanism which makes for faster convergence. but another mechanism which mal<es
for slower convergence than under fixed exchange rates. On the one hand. when domes-
tic inflation falls below its steady-state leveln/ under flexible exchange rates, the central
bank reacts by cutting the domestic interest rate, thereby inducing a nominal exchange
rate depreciation which helps to pull domestic inflation back up again by stimulating
aggregate demand. This mechanism - which is reflected in the term e-1 in the numerator
of (32)- tends to speed up the adjustment to long-run equilibrium and is obviously absent
under fixed exchange rates. On the other hand, as inflation rises back towards its steady-
state level, the central bank also reacts by gradually raising the domestic interest rate
under flexible exchange rates. thereby dampening the increase in aggregate demand which
is pulling the economy back towards long-run equilibrium. This mechanism clearly tends
to slow down the adjustment process under flexible exchange rates and L~ captured by the
term - yf3 2 in the numerator of (32).
The relative magnitude of the parameters 8 and y{3 2 will determine whether the net
efi'ect of these two ollsetting mechanisms is to speed up or to delay the convergence to
long-run equilibrium under a flexible rate regime compared to a fixed exchange rate
regime. However, even though we cannot say for sure whether deviations of output and
employment from their natural rates will last longer under fixed than under flexible
exchange rates. theory does allow us to predict the relative magnitude of the short-run
fluctuations in output and inflation occurring under the two exchange rate regimes when
the economy is hit by shocks. This is the topic for the next section.
25•3 A floating exchange rate: shock absorber or an amplifier of

shocks?
An important issue is whether a flexible exchange rate can help to absorb shocks to the
economy so that short-run fluctuations in output and inflation are reduced, or whether
the response of a floating exchange rate will actually tend to ampli(y the effects of shocks
on output and inflation? We will now use our AS-AD model of the open economy to
analyse this question.
25 THE OPEN ECONOMY WITH FLEX IBLE EXCHANGE RATE S 781
The short-run effects of aggregate supply shocks
We start by considering the short-run eflects of aggregate supply shocks under alternative
exchange rate regimes. In the previous chapter we saw that the numerical slope of the
aggregate demand curve is 1/fJ Lunder fixed exchange rates, while Eq. (20) above shows
that the numerical slope of the AD curve is 1/ PLunder flexible exchange rates. Since
/J 1 > fJ 1 • it follows that the AD curve is )latter under flexible than under fixed exchange
rates. The reason is that under flexible exchange rates a fall in the rate of domestic
inflation Vtrill not only stimulate aggregate demand by the direct impact on competitive-
ness; it will also induce the central bank to lower the interest rate, thereby boosting
investment and generating a depreciation which further strengthens international com-
petitiveness.
In Fig. 25.3 the curves 'AD(fix)' and 'AD(flex)' represent the AD curves under fixed
and under flexible exchange rates, respectively. Suppose now that the economy is hit by a
negative supply shock which shills the SRAS curve upwards. As illustrated in the tlgure,
this will have a larger negative short-run effect on output and a smaller positive impact on
inflation under flexible than under fixed exchange rates.
The explanation for the dillerent effects of supply shocks under the two exchange rate
regimes is that whereas the interest rate stays constant under fixed exchange rates. under
flexible exchange rates the inflation targeting domestic central bank responds to higher
inflation by raising the interest rate. thereby depressing aggregate demand via the interest
rate cha1mel and the exchange rate channel.
LRAS
Figure 25.3: Short-run effects of a negative supply shock under fixed versus flexible exchange rates
© The McGraw-Hill
Companies. 2005
Thus we conclude that under a flexible exchange rate regime with strict inflation
targeting, adjustments in the interest rate and in the nominal exchange rate help to
dampen the fluctuations in inflation arising from supply shocks, but only at the cost of
increased fluctuations in the level of output compared with a fixed exchange rate regime.
The short-run effects of aggregate demand shocks
Let us next compare the short-run effects of aggregate demand shocks under the two
exchange rate regimes. This is not so easily done by a graphical analysis. since the vertical
sh!ft in the AD curve generated by a given shock to aggregate demand will be different
under the two regimes at the same time as the slope of the AD curve differs across regimes.
It is therefore more convenient to resort to a mathematical analysis. Solving (2 5) and (24)
for the short-run values of the output and inflation gaps, and using the dellnition of P1
stated in (15), we get:
• z,
Flexible excha11ge rates: (33)
Yr = 1 + yf31 + yh(f3z + e - 1/3 1) ,
- yz,
(34)
1 + y/31 +yh(fJ2+ 8 1/31).
nt =
<-
where we have set 1 = s, = 0 . since we are now focusing on demand shocks. and since
the value of e~_ 1 is immaterial for the present short-run analysis. In a similar way, we lind
from Eqs (7) and (8) in the previous chapter that the short-run solutions lor the output
and inflation gaps under llxed exchange rates are (lor e~_ 1 = s, = 0):
Fixed exchange rates: (3 5)
(36)
From these expressions we see that the demand shock variable z 1 has a smaller short run
effect on output as well as inflation under flexible than under llxed exchange rates. Under
the latter regime, the impact of a positive demand shock is dampened only by the loss of
competitiveness occurring as higher domestic activity drives up the domestic inflation
rate. This dampening ellect is reflected in the term y/3 1 in the denominators of (3 5) and
(3 6). But under flexible exchange rates the rise in domestic inflation generated by a
positive demand shock also induces the central bank to raise the interest rate. causing a
further dampening of the rise in output and inflation through the interest rate chatmel
and the exchange rate chatmel. This shock-absorbing effect of the change in the interest
rate and in the nominal exchange rate is captured by the term yh(/3 2 +8- 1/3 1) in the
denominators of ( 3 3) and (34).
This analysis suggests that if demand shocks are the dominant source of macro-
economic fluctuations. a country wL~hing to minimize the variability of output and infla-
tion may be better served by a flexible than by a llxed exchange rate, because a floating
exchange rate allows the interest rate and the nominal exchange rate to help absorb
demand shocks, thereby reducing the need for adjustments in output and inflation.
Table 25.2 summarizes our results regarding the dillerences in the short-run ellects
of demand and supply shocks under the two exchange rate regimes. Because of the
Table 25.2: Short-run effects of shocks under flexible exchange rates,

compared to the effects under fixed exchange rates
Type of shock
Short-run
fluctuations in Demand shock Supply shock
Output Smaller Larger

Inflation Smaller Smaller
qualitative similarities in the structure of the open economy under the two regimes, the
qualitative medium-term and long-term effects of demand and supply shocks are similar
under llxed and flexible exchange rates. In particular, the dynamic, qualitative effects of
temporary and permanent llscal policy shocks under flexible exchange rates correspond to
those found in the previous chapter. Hence we will not repeat the graphical analysis of
such shocks in the present chapter. However. notice that since our variable z1 includes
llscal policy shocks. the mathematical analysis above implies that llscal policy has a
smaller quantitative impact on economic activity under flexible than under llxed
exchange rates.
Is the exchange rate really a shock absorber?
Table 25 .2 suggests that a floating exchange rate can help to cushion the economy
against demand shocks and that it also helps to reduce fluctuations in the inflation rate
under an inflation targeting regime. However. based on empirical observations some
economists have questioned whether a flexible exchange rate is really an effective shock
absorber rather than an independent source of shocks. To illustrate how a floating
exchange rate could be a source of shocks, note from Eq. (10) that when llnancial
investors are risk averse and foreign and domestic bonds have dillerent risk character-
istics. the change in the nominal exchange rate may be '.vritten as:
t1e = e-1
(/ - i + v + ~.:) . (37)
Experience shows that risk premia in llnancial markets can vary quite a lot. as some
unexpected economic or political event motivates International Investors to revise their
evaluation ofthe relative riskiness of assets issued in different countries. For a given stance
of domestic and foreign monetary policy- that is, for given values of i and / - we see !rom
(3 7) that fluctuations in the (stochastic component of the) risk premium t must generate
fluctuations in the exchange rate which may in turn contribute to instability in output
and inflation. If the shifts in risk premia are lrequent and significant, a floating exchange
rate may thus become a source of macroeconomic instability rather th an an absorber of
shocks arising elsewhere in the economy.
If a floating exch ange rate tends to act as an absorber of shocks to real output and
employment. we should observe a negative correlation between exchange rate volatility
and output volatility. In Fig. 2 5.4 we measure exchange rate volatility. o <' as the standard
deviation of each country's nominal ellective (i.e.. trade-weighted) exchange rate, while
output volatility, o Y' is measured by the standard deviation of the rate of growth of indus-
trial production. The diagram is based on monthly data for 17 OECD countries lor the
llexible exchange rates
© The McGraw-Hill
Companies. 2005
784 PART 7 : TH E SHO RT-RUN MO DEL FO R TH E O PEN ECONO M Y
ae
0.016
•
Regression line:
0.01 4 •• a 0 = 0.00553 - 0.08 273*oy
t = - 3.31
••
,.. •
Ff = 0.0458
0.01 2 • •
•
0.0 1 0.02 0.03 0.04 0.05 0.06 0.07 0.08

ay
- 0.002
Figure 25.4: The relationship between exchange rate volatility and output volatility in 17 OECD countries, 1975-
2000/ 02
Note: Exchange rate volatility is measured by the s tandard deviation of the nominal effective exchange rate. Output volatility is measured
by the standard deviation of the change in the log of seasonally adjusted industrial production. The regression line w as estimated using
standard O LS.
Source: IMF Int ernational Financial Statistics.
period 1975- 2002 . This period was divided into 28 two-year intervals for which the
standard deviations were calculated, giving a total of 28 x 17 observed combinations of
exchange rate volatility and output volatility. Fitting a regression line through the data,
we find that higher exchange rate volatility does tend to be associated with lower output
volatility, but this relationship is driven very much by a lew outliers in the data represent-
ing periods when output volatility was unusually high. i'vloreover. the regression explains
very little of the total variation in the data. so the main impression is that higher exchange
rate flexibility has at best a very weak stabilizing influence on output. 3
Using a statistical technique called vector autoregression analysis (VAR), several
economists have tried to evaluate more carefully whether the exchange rate is mainly a
source of shocks rather than a shock absorber. A V AR analysis allows the researcher to
estimate that part of the current movement in variables such as output and inflation
which is due to new shocks occurring in the current period. These shocks may then be cat-
egorized into dillerent types using basic assumptions (so-called identi(ying restrictions)
from economic theory. For example. we know from theory that only supply shocks can
have a permanent ellect on output. so an estimated shock which has a permanent
3. The idea for the simple analysis in Fig. 25.4 came from a paper by Robert P. Aood and Andrew K. Rose, 'Fixing
Exchange Rates- A Virtual Ouest for Fundamentals', Journal of Monetary Economics, 36, 1995, pp. 3-37. B ased
on a more sophisticated analysis of the data, these authors also concluded that there is no clear trade·oft between
reduced exchan~e rate volatility and macroeconomic stability.
2 5 THE O PEN ECONO MY WITH FLEX IB LE EXCHA NG E RATES 785
statistical effect on output is categorized as a supply shock. As another example, a so-

called real demand shock (e.g., a permanent Hscal shock) is a disturbance which has a
lasting statistical ellect on some real variable other than output (such as the real exch ange
rate). whereas a so-called nominal shock has no permanent effect on any real variables. A
shock to the nominal exchange rate arising from a temporary shift in risk premia is an
example of a nominal shock.
The economists Michael Artis and Michael Ehrmann have carried out a comparative
VAR analysis of data for industrial output. inflation, domestic and foreign short-term
interest rates and nominal exchange rates for Canada. Denmark. Sweden and the UK lor
the period 1980- 1998. The purpose of their analysis was to evaluate the role of the
exchange rate as a shock absorber or source of shock. Their main results are summarized
in Table 25 .3 which shows the proportion of the short-run variance of output. inflation
and exchange rates which can be ascribed to the diilerent types of shock. A monetary
policy shock is an 'unsystematic' movement in the short-term nominal interest rate,
that is, a movement which cannot be predicted from a monetary policy rule such as
Eq. (2). These shocks as well as exchange rate shocks are categorized as nominal shocks.
Over the period considered, all of the four countries have allowed considerable
variation of their bilateral exchange rates against their major trading partner (Germany in
the case of Denmark, Sweden and the UK: the United States in the case of Canada). This
is also true for Denmark. even though that country held its exchange rate against
Table 25.3: Proportion of the variance of output, inflation and nominal exchange rate
explained by different types of shocks, 1980-1998
Percentage of variance explained by
Real Foreign and
Supply demand domestic monetary Exchange
Country Variance of shocks shocks policy shocks rate shocks
Output 15.5 83.4 1.0 0.1
Canada Inflation 53.9 1.8 24.6 19.7
Exchange rate 4.1 0.7 87.9 7.3
Output 90.6 5.9 3.0 0.5

Denmark Inflation 3.0 0.3 54.9 41.8
Exchange rate 3.1 1.3 51.7 43.9
Output 55.2 41.8 2.3 0.7

Sweden Inflation 50.7 15.9 32.6 0.8
Exchange rate 1.2 0.6 2.3 95.9
O utput 83.9 11.6 3.1 1.4

United Kingdom Inflation 7.0 17.1 43.1 32.8
Exchange rate 1.5 1.5 36.2 60.8
Note: Variance calculated over one-year horizon.
Source: Michael J. Artis and Michael Ehrmann, 'The Exchange Rate- A Shock-Absorber or Source of Shocks?
A Study of Four Open Economies'. Centre for Economic Policy Research. Discussion Paper No. 2550, September 2000.
© The McGraw-Hill
Companies. 2005
786 PART 7 : TH E SH O RT-RUN MO DEL FO R THE O PEN ECONOMY
Germany within a rather narrow band towards the end of the period. It therefore
makes sense to study the role of the exchange rate as a potential shock absorber in these
countries.
If the nominal exchange rate is mainly an absorber of shocks to the real economy
(output) rather than an independent source of shocks, then most of the movements in the
exchange rate will be driven by the same types of shock as those causing variations in
output. This hypothesis finds little support from the results reported in Table 25.3. The
bulk of the short-run variance in output is explained by supply shocks or by real demand
shocks. but these shock categories explain less than 5 per cent of the variance of the
exchange rate in all countries.
However. Table 25 .3 indicates that a substantial part of the variability in inflation
has been created by monetary policy shocks which also seem to explain quite a lot of the
variance in the exchange rate in Canada, Denmark and the UK. This suggests that
nominal demand shocks arising from unsystematic monetary policy have been absorbed
by changes in exchange rates as well as by adjustments in goods prices.
Still. the most striking result in Table 25.3 is the very high proportion of the variance
in the exchange rate which is explained by shocks to the exchange rate itself, particularly
in Sweden and the UK. This suggests that, rather than reacting to shocks originating else-
where in the economic system, the exchange rate moves mainly in response to shocks to
the foreign exchange market such as shills in risk premia. In Denmark and the UK. and to
a smaller extent in Canada. the exogenous exchange rate shocks seem to have generated
a large part of the variance in the inflation rate, so while the exchange rate may have
absorbed part of the monetary policy shocks to inllation, it may also have been a breeder
of inflation shocks itself. At the same time we see from the table that exchange rate shocks
appear to have had very little influence on the variance of outpu t.
Results like those reported in Table 2 5.3 should be interpreted with great care
because of the inherent diiliculties of separating the eflects of shocks from the underlying
systematic movements in economic time series. In particular, identifying shocks to the
nominal exchange rate is very hard because economists have had great difl'iculty in
modelling the determination of floating exchange rates. But with th is proviso. Table 2 5.3
suggests that exchange rates tend to move in response to shocks other than those that
drive output fluctuations and that a large part of exchange rate variability originates from
disturbances to the foreign exchange market itself. A number of other empirical studies
have found similar results. This does not support the idea that the exchange rate plays an
important role as a shock absorber.
Exchange rate overshooting
On top of the short-run fluctuations, floating exchange rates also tend to undergo rather
large swings in the medium term. For example, Fig. 2 5.5 illustrates the large medium-nm
movements in the exchange rate between the US dollar and the euro (a weighted average
of the euro area currencies before 1999). Such large swings in floating exchange rates
may seem hard to rationalize. Alter all, the hypothesis of relative purchasing power parity
formalized in Eq. (4) says that over the long run the movement in the exchange rate
should just oflset the difference between national inflation rates. Since these diilerences
tend to be small across developed countries (and are. indeed. nowadays close to zero), the
Introducing Advanced Model for the Open flexibl e exchange rates Companies. 2005
25 THE OPEN ECONOMY WITH FLEX IBLE EXCHANGE RATE S 787
1.3
e 1.2
::>
Q)
Q;1.1
a.
"'
~
0
0 0.9
0.8
0.7
0.6
0 C'l <t ~ co 0 C'l <t ~ co 0 C'l <t
co co co co co "' "' "' 0 "' 0 0 0
"' "' "' "' "' "' "' "' 0 "' 0
C'l 0
C'l 0
C'l
Figure 25.5: The dollar·euro exchange rate

Not e: Before 1999 the graph shows the exchange rate of the dollar against a weight ed average of the currencies
of the countries in the Euro area.
Source: Darish Economic Council.
large medium-run exchange rate swings might seem to reflect an irrational overreaction
in foreign exchange markets.
In a seminal contribution. the late German-American economist Rudiger Dornbusch
showed that rather than reflecting irrationality. the apparent 'overshooting' of flexible
exchange rates may be compatible with rational behaviour in foreign exchange markets. 4
Dornbusch used the example of a monetary policy shock to make his point. but here
we will illustrate that exchange rate overshooting may also occur in response to other
disturbances such as a fiscal policy shock.
Inserting (6) and (5) into (1), we may write the condition lor uncovered interest
parity as:
r1 = r{ + 8(e1 - e1) . (38)
Following Dornbusch, we now assume that the expected 'normal' exchange rate e1 in (3 8)
corresponds to the economy's long-term equilibrium exchange rate. In other words,
financial investors have correct (rational) expectations of the economy's long-term equi-
librium nominal exchange rate. Suppose now that the domestic government engages in a
permanent fiscal expansion. Figure 25 .6 shows how this will allect the economy's real
exchange rate in the long run, using the analytical apparatus from Chapter 23 .
As you recall, LRAD is the long-run aggregate demand curve which is upward-
sloping since a real exchange rate depreciation (a rise in er) stimulates the demand for
domestic output by improving the country's international competitiveness. The LRAS
4. See Rudiger Dornbusch, 'Expectations and Exchange Rate Dynamics', Journal of Political Economy, 84, 1976,
pp. 11 61-11 76.
Model for the Open
© The McGraw-Hill
Companies. 2005
788 PART 7: THE SHOR T-RUN MODEL FOR THE OPEN ECONOMY
e'
LRAS
LRAD0
LRAD 1
e,r
Figure 25.6: The long-run effect of a fiscal expansion
curve is the long-run aggregate supply curve. A permanent fiscal expansion shifts the
LRAD curve to the right. 5 This implies a new long-run equilibrium at point A 1 where the
real exchange rate has appreciated.
Under a floating exchange rate. part of this real appreciation will occur through an
appreciation of the long-run nomil'wl exchange rate e,. We can now see how exchange rate
overshooting occurs. In the short and medium term. the fiscal expansion drives up the
domestic levels of output and inflation. inducing the domestic central bank to raise the
domestic real interest rater,. At the same time rational investors correctly anticipate that
the long-run equilibrium exchange rate e, has fallen . It then follows from (38) that the
current exchange rate e, has to fall by even more than e, to maintain uncovered interest
parity. In other words. in the short run the exchange rate must 'overshoot' its new long-
run equilibrium value in order to generate expectations of a ji.tture depreciation of the
domestic currency so that domestic assets are no more attractive than foreign assets.
despite the fact that the domestic interest rate has risen.
This analysis suggests that overreactions in the exchange rate relative to its long-nm
equilibrium may be an unavoidable feature of a flexible exchange rate regime. Some
economists see such overreactions as an argument in favour of fixed exchange rates.
whereas others argue that the benefits from an independent national monetary policy
outweigh the costs of exchange rate overshooting under floating exchange rates. These
observations take us into the great debate on the choice of exchange rate regime which is
the subject of the next chapter.
5. In Chapter 23 we showed that the LRAD curve is given by the relationship e' =fii '(y - r- z). A permanent fiscal
expansion implies a permanent rise in z so that aggregate demand permane1tly increases for any given level of the
real exchange rate e', thereby shifting the LRAD curve to the right.
2 5 THE OPEN ECONOMY WITH FLEXIBLE EXCHANGE RATES 789
25A Summary
................................................................................................................................................................................
1. A crucial characteristic of a flexible exchange rate regime is that it enables the domestic
central bank to pursue an independent monetary policy even if capital mobility is perfect.
2. Under a floating exchange rate regime the nominal exchange rate cannot serve as a nominal
anchor for inflation expectations. To provide an alternative nominal anchor, many countries
with flexible exchange rates have adopted a monetary policy regime of inflation targeting
where the central bank seeks to keep inflation within a narrow band around some target rate
which is typically set to 2-2.5 per cent. Under strict inflation targeting the central bank's
interest rate reacts only to deviations of inflation from the target, whereas flexible inflation
targeting implies that the central bank also reacts to the evolution of output and employment.
The fact that inflation targeting countries in the OECD have chosen roughly the same inflation
target suggests that these countries try to avoid a systematic depreciation or appreciation of
their nominal exchange rates over the longer run.
3. Most inflation targeting countries have delegated monetary policy to an independent central
bank to strengthen the credibility of the inflation target. When the target is credible, it will
serve as an anchor for domestic inflation expectations. The model in this chapter assumes
that agents have weakly rational expectations in the sense that the expected inflation rate
equals the central bank's inflation target. G iven that countries choose (roughly) the same
inflation target, this means that the expected domestic inflation rate is (roughly) equal to the
foreign inflation rate.
4. Under flexible exchange rates and perfect capital mobility the difference between the
domestic and the foreign nominal interest rate is given by the expected rate of change in the
nominal exchange rate. According to the hypothesis of regressive exchange rate expect-
ations, the exchange rate is expected to rise if it is currently below its perceived normal level,
and vice versa. The perceived normal exchange rate is likely to depend on the actual
exchange rates observed in the past. The model in this chapter makes the simplifying assump-
tion that the perceived normal exchange rate equals last period's actual exchange rate. From
this assumption plus the assumption of uncovered interest rate parity it follows that a positive
differential between the domestic and the foreign interest rate will cause an appreciation of
the domestic currency, whereas a negative interest rate differential will generate a deprecia-
tion. The empirical evidence lends some support to this hypothesis.
5. Under 'clean' floating the central bank does not intervene in the foreign exchange market.
Under 'dirty' floating the central bank intervenes with the purpose of reducing fluctuations in
the exchange rate. In the case of 'sterilized' interventions the central bank buys or sells foreign
currency without changing the interest rate. To have a lasting impact on the exchange rate,
interventions have to be 'unsterilized', involving a change in the interest rate which supports
the movement in the exchange rate that the central bank is trying to achieve. When the central
bank systematically raises (lowers) the interest rate in response to a depreciation (appre-
ciation) of the domestic currency, it is said to follow a policy of 'leaning against the wind'.
6. In a flexible exchange rate reg ime monetary policy affects aggregate demand both through
the interest rate channel and through the exchange rate channel. The interest rate channel is
Model for the Open
© The McGraw-Hill
Companies. 2005
the d irect impact of a change in the interest rate on investment and consumption demand.
The exchange rate channel is the impact through the foreign exchange market: when the
interest rate is lowered, there is a tendency for the exchange rate to depreciate so that net
exports increase.
7. Under flexible exchange rates a fall in the domestic rate of inflation boosts aggregate demand
through three effects. First, there is a direct positive effect on net exports as domestic com-
petitiveness improves. Second, a lower rate of inflation induces an inflation-targeting central
bank to reduce the interest rate, thereby stimulating investment and consumption. Third, the
lower interest rate causes a depreciation which g ives a further stimulus to net exports. The
latter two effects are absent under fixed exchange rates. Hence the aggregate demand curve
is flatter under flexible than under fixed exchange rates.
8. The AS-AD model for the open economy has a similar structure under fixed and under flexible
exchange rates, but the quantitative properties of the economy w ill d iffer under the two
regimes. A priori one cannot say whether convergence on long-run equilibrium w ill be faster
under one or the other reg ime.
9. UnciP.r fiP.xihiP. P.Xc:h;mgP. r~tP.S w ith stric:t infl<'l tion t~rgP.ting, surrly shoc:ks w ill C:<'lUSP. l~rgP.r
short-run fluctuations in output but smaller fluctuations in inflation than under fixed exchange
rates. Our AS-AD model also predicts that demand shocks ( ncluding fiscal policy shocks)
w ill generate smaller fluctuations in output as well as inflation under flexible than under fixed
exchange rates, because a floating exchange rate allows the central bank to counteract
demand shocks through its interest rate policy.
10. When domestic and foreign assets have different risk characteristics and financial investors
are risk averse, fluctuations in requ ired risk premia can cause fluctuations in nominal
exchange rates under floating exchange rates. Empirical evidence suggests that exchange
rates tend to move in response to shocks other than those that drive output fluctuations and
that a large part of exchange rate variability originates from disturbances to the foreign
exchange market itself. This does not support the idea that the exchange rate plays an
important role as a shock absorber.
11. Floating exchange rates tend to undergo large swings in the medium term and often seem to
'overshoot' their long run equilibrium values. Such apparently irrational behaviour may be
quite compatible with rationality. For example, financial investors may correctly anticipate that
a permanent positive demand shock requires a long-term appreciation of the exchange rate.
If the central bank raises the domestic interest rate above the foreign interest rate in response
to the shock, the exchange rate w ill then have to appreciate even more in the short run than
in the long run to generate market expectations of a future depreciation so that domestic
assets are no more attractive than foreign assets.
25.5 Exercises
Exercise 1. The aggregate demand curve under flexible exchange rates
The derivation of the AD curve in the main text assumed that the expected ' normal' exchange
rate is equal to the actual exchange rate observed during the last period. This exercise asks
you to demonstrate that under certain assumptions, one can derive a similar AD curve under
flexible exchange rates by simply assuming that the expected normal exchange rate is
exogenous.
We start by assuming that the real interest rate affects aggregate demand mainly through
its influence on business investment. Assuming that it only takes one period for firms to adjust
their capital stock to its desired level Kd, and abstracting from depreciation on the existing
capital stock, total gross investment in the current period, I, w ill equal the desired increase in
the capital stock:
(39)
Reflecting the underlying growth in aggregate demand, the desired capital stock varies
positively with time, but at the same time it varies negatively with the real interest rate.
Adopting a linear specification for convenience, we thus ~ave:
K~=at -br, = l, =a-b(r,-r,_1 ), (40)
where a and bare constants. In other words, the level of investment depends on the change
in the real rate of interest. Dropping the time subscripts for the current period, we may there-
fo re approximate the goods market equilibrium condition by:
y- y=/3 1 e' - f3 2 (r- r_1) + 'Z, (41)
Z. ""/3 3 (g- g)+ f3ir' - Jl') + /J 5 (1n c -In f),
where we apply the usual notation (the constant a has dropped out because (4 1) considers
a deviation from trend, and the parameter b is incorporated in /3 2 ) .
Now suppose that the central bank targets the foreign inflation rate by raising the real
interest rate when domestic inflation is above foreign inflation, and vice versa:
r-r_1 = h(n - n 1) . (42)
In addition we have the condition for uncovered interest rate parity:
"
J. = t•/ +e+1 -e, (43)
where we assume that exchange rate expectations are regressive:
(J > 0. (44)
Here we treat the expected normal exchange rate, e, as an exogenous variable, although it
n:
may change from time to time.
By definition, the ex ante domestic real interest rate is r ""i- 1• The central bank has
credibility, so the expected domestic inflation rate equals the central bank's inflation target,
that is, n! 1 = :;r~ 1 = n 1• Hence we have:
r =i-n1. (45)
Finally, we have the familiar definition of the foreign real interest rate,
r1 "" i1- n 1, (46)
and the bookkeeping identity fo r the current real exchange rate:
e' = e~ 1 + e- e_1 + n 1 - :;r, (4 7)

Model for the Open
© The McGraw-Hill
Companies. 2005
792 PART 7 : THE SHORT-RUN MODEL FOR THE OPEN ECONOMY
1. Use Eqs {41 )- {4 7) to show that the aggregate demand curve takes the following form:
{48)
{49)
where ~e"' e- e_, and M 1 "' r 1 - r~ 1 • {Hint: as an intermediate step, use {43) - {46) to derive
an expression for r- r _1 in terms of e- e_1, ~e. and ~r1• Then insert {42) to obtain an expres-
sion fore- e _, in terms of n - n 1, ~e. and ~r 1.) Make a brief comparison between these results
and the AD curve {15) in the main text.
2. Suppose that expectations of a permanent weakening of the domestic currency arise. How
will this affect the aggregate demand curve? G ive some examples of economic events which
might generate expectations of a permanent depreciation of the domestic currency.
Exercise 2. 'Leaning against the wind' under flexible exchange rates
This exercise invites you to explore the implications of a monetary policy regime where the
central bank follows a policy of 'leaning against the wind' by setting the interest rate in accor-
dance with the following policy rule, explained in the section on 'dirty floating':
h > 0, l > 0. {50)
Because of perfect capital mobility and risk neutrality, the condition for uncovered interest
parity must hold:
e "' lnE, {51)
We also continue to assume regressive exchange rate expectations so that:
{5 2)
and we maintain the assumption that the central bank's inflation target is credible and equal
to the foreign inflation rate:
{53)
The goods market equilibrium condition corresponds to Eq. {12) in the main text, repeated
here for convenience:
___....._...._
e'
y- y= f3 1 (e~ 1 + ~e + n 1 - n) - f3 2(i1 - n! 1 + e!, - e- r') +i , {54)
Finally, we continue to work with an aggregate supply curve of the form:
:Jr. = :Jr. . + y(y - y) + s. {55)
1. Discuss briefly why the authorities might want to adopt a policy of 'leaning against the wind'.
2. Use Eqs {50) - (54) to show that the economy's aggregate demand curve takes the form:
{56)
(Hint: follow the procedure described in Section 2.) Explain the economic mechanisms under-
lying the negative slope of the AD curve. Explain how the policy of leaning against the wind
affects the slope of the AD curve.
3. Derive the equations determing the change over time in the nominal and in the real exchange
rate. Does the policy of leaning against the w ind amplify or dampen the impact of the inflation
gap n - n 1 on nominal and real exchange rates? Explain. (Hints: use (50) - (53) to express
e- e _, in terms of Tr - n 1 and the parameters h, 8, and A.) G ive a graphical illustration of the
economy's adjustment to long-run equilibrium and explain the adjustment mechanisms.
4. Does leaning against the w ind slow down or speed up the economy's speed of adjustment to
long-run equilibrium? (Hint: derive a d ifference equation of the same form as (28) in the main
text and irvestigate by d ifferentiation how the parameter A affects the coefficient a on the
lagged output gap.) Try to provide an economic explanation for your result.
5. Use the procedure described in Section 3 of the main text to investigate how the policy of
leaning against the w ind affects the economy's short-run reaction to supply and demand
shocks. Does leaning against the wind amplify or dampen the short-run effects of the shocks?
G ivP. iln P.r:onomic: P.xrlilniltion for your finrlings.
6. Demonstrate that if the policy of leaning against the wind is very aggressive (so the parameter
A tends to infinity), the economy w ill work approximately as if the exchange rate were fully
fixed. (Hint: compare the d ifference equation derived in Question 4 for the case of ,t --> oo with
the corresponding difference equation for the output gap under fixed exchange rates, g iven in
Eq. (1 5) of Chapter 24.) G ive an intuitive explanation for your finding.
Exercise 3. A flexible exchange rate regime with flexible inflation targeting
In the main text we assumed that the domestic central bank pursued so-called strict inflation
targeting, reacting only to changes in the inflation gap. This exercise studies a reg ime of so-
called flexible inflation targeting where interest rate policy follows a standard Taylor rule with
a positive coefficient on the output gap. Our open economy w ith flexible exchange rates is
thus described by the following equations (in our usual notation):
e'I
Goods market: y- JI =,B,(e: , + e-e_1 + n 1 -n) - ,82 (i-Tr! 1 - r') +i , (57)

1
Inflation expectations: n ! , = n• = n , (58)
Monetary policy: i = r' + n ! , + h(n - n) + b(y- Jl), b > 0, (59)
Uncovered interest rate parity: i = i1 + e!, - e, (60)
Exchange rate expectations: e! ,- e = -B(e- e_,), (6 1)
Foreign real interest rate: r' = i1 - n 1, (62)
SRAS: n=n• + y(y-JI) + s, (63)
1
Real exchange rate: e'= e: , + e- e_, +Tr - n. (64)
1. Use Eqs (57) - (62) to show that the economy's aggregate demand curve is g iven by:
(65)
Model for the Open
© The McGraw-Hill
Companies. 2005
(Hint: follow the procedure used in Section 2 to derive the AD curve.) Explain in economic
terms how the central bank's reaction to the output gap affects the slope of the AD curve.
2. Use a (y, .rr) d iagram to undertake a graphical analysis of the way the economy reacts to
demand and supply shocks in the short run (i.e., in the first period), assuming that the
economy is in long -run equilibrium in period 0 . Illustrate and e~plain what d ifference it makes
for your results that b is positive rather than zero. Can you think of a situation where policy
makers wou ld not want to choose a positive b?
3. Use (58)- (62) to show that
e' = e~ 1 - (1 + e-1 h)(.n- .rr 1) - e-1 b(y- y). (66)
Explain in economic terms why positive inflation and output gaps generate a real exchange
rate appreciation (a fall in e1.
Defining y1 "' y ,- y and ir. 1 "' .n 1 - .n1 and using (58), (63), (65) and (66), we may summarize
our model as follows:
(fJ 1) 1 + b(/32 + e-l fJ 1)) Z1 (67)

e,_1 -
f (
AD: Pt y, + P1 '
A A
n,= "'{;
SRAS: ic1 =yy1 + s 1, (68)

4. Use (67)- (69) to show that the model may be condensed to the following di fference equation
in the output gap:
9,~ 1 = ay, + {J(z 1 ~ 1 - z,)- /3P 1 (s 1~ 1 - s,) - fJ/3 1( 1 + fr 1 h)sp (70)
1
fJ -= 1 +r~ 1 + b(/32 + e-1flt ) , a "' {3[1 + f3 2 ( b + yh) ].
Is the economy's long-run equilibrium stable? Does an increase in the value of b have an
unambiguous effect on the economy's speed of adjustment? Try to g ive an economic expla-
nation for your conclusion on the latter question. (Hint: it may be helpful for you to reconsider
our explanation for the result found in equation (32) in the text.)
Exercise 4. Simulating an AS-AD model for an open economy with flexible

exchange rates and a mixture of backward-looking and forward-looking
expectations
Following up on Exercise 6 in the previous chapter, this exercise seeks to deepen your
understanding of the workings of an economy w ith flexible exchange rates by asking you to
implement the following model on the computer:
e'I
1
Goods market: y 1 - y = {3 1( e;_1 + e 1 - et-t + n 1 - n 1) - fJ 2 (i1 - .rr~~ ~ - 7 ) + 21, (71 )
Inflation expectations: n~ = qm + (1 - ({l)n 1_ p
1
0 ~({I~ 1. (72)
Monetary policy: i 1 = r + n~~ ~ + h(.n 1 - :n~,

1
(73)
Uncovered interest rate parity: i1 =i' + e~.., 1 - e11 (74)
Exchange rate expectations: e~~ 1 - e,= - B(e, - e,_1), (75)

Foreign real interest rate: r' = i' - :rc', (76)

SRAS: :rc 1 =:rc~+y(y, - ji)+ sp (77)
Real exchange rate: e; = e;_, + e 1 - e 1_ 1 + :rc 1 - :rc1• (78)
Equation (72) gives the average expected domestic inflation rate, assuming that a fraction cp
of the population has 'weakly rational' expectations which are anchored by the central bank's
inflation target, :rc* = :rc 1, whereas the remaining fraction of the population has static expect-
ations, expecting that this year's inflation rate w ill correspond to the inflation rate observed
last year, .n 1_ 1 • In the special case where cp= 1, we obtan the basic model analysed in the
main text of this chapter.
1. Show by means of (71 ) - (76) that the equation for the aggregate demand curve is:
n 1=:rc +
1
(~:)e;_,- (;Jr,- y -z,), (79)
f;, ""p, +fJ 2 h +/3, e-1( 1 - cp+ h), z,"" - /J 2(r 1 - r') +z,.
(Hint: use the procedure followed in Section 2 to derive the AD curve.)
To facilitate imp lementation of the model on the computer, we now want to reduce it to two
difference equations in the output gap and the inflation gap. Defining y, "" y,- y and
1
fr 1 ,:rc 1 - :rc and using (72) - (76) to derive
• e 1-e 1_ 1 = -e- 1(1 - qJ+ h), the model consisting of
(77) - (79) may be summarized as:
(80)
(81)
From (80) and (82) it follows that:
ir, - ft,_, = (~)1 + e- 1

(1- qJ+ h)]( -n,_1) - (;,)rr,-p,_ 1- (z, - z,_1) ] , (83)
wh ile (81 ) implies that:
(84)
It also follows from (81) that:
(85)
2. Equate (83) and (84) to find an expression for J!: 1 and substitute the resulting expression
into (81 ) to show that the model may be reduced to the following second-order difference
equation in the output gap:
9,. 2 - a 1.Y,. 1 + a0jl 1 =f3(z,. 2 - z,. 1 ) - /](1 - cp)(z,. 1 - z,) - (3 1 tJs,. 2 + /Jh/] 2 s,. 1 , (86)
1
f3"" 1 + yp1'
© The McGraw-Hill
Companies. 2005
Furthermore, insert (85) into (83) and use the definition of (J 1 to show that the model may
alternatively be condensed to the following difference equation in the inflation gap:
(87)
Verify that (86) collapses to Eq. (28) in the main text in the special case where q; = 1. (Hint:
remember to use the definition of (J 1 .)
You are now asked to simulate the model (82), (86) and (87) on the computer to study the
dynamics of output, inflation and the real exchange rate under flexible exchange rates. We
proceed in a manner identical to the procedure followed in Exercise 6 of the previous chapter,
but in case you haven't solved that exercise, we repeat all the steps here.
Using the definitions given in the appendix to Chapter 23, and assuming that trade is ini-
tially balanced and that the marginal and the average propensities to import are identical, one
can demonstrate that:
m(IJx+ IJM- 1) -1] 0 (1 - r)

P1= 1- Oy+m
,
where m = M0 j"l is the initial ratio of imports to GOP, YJx and '7 M are the price elasticities of
export and import demand, r "' ("? - 0 )/'Y is the net tax burden on the private sector,
1Jo"' - ('dO/ CJ£') (£'/ 0) is the elasticity o f private domestic demand with respect to the real
exchange rate {reflecting the income effect of a change in the terms of trade), and Or is the
Under the assumptions mentioned above, one can also show from the definition of P2 in the
appendix to Chapter 23 that:
R = 1J( 1 - r) (1 -m) 0,
fJ2 -
1- Or + m
I
IJ "' - (1 - r )'Y '
where IJ is the change in private demand induced by a one percentage point change in the
real interest rate, measured relative to private disposable income (we introduced this para-
meter in Chapter 19). Given these specifications, your first sheet should allow you to choose
the parameters y, q;, h, 8, m, 1J X• 1J M• r, 1J, 17 0 and Or to calculate the auxiliary variables P, P2 ,
{3 1 and P(from which your spreadsheet can calculate the coeff cients a 1 and a 0 in (86)).
You should construct a deterministic as well as a stochastc version of the model. In the
deterministic version you just feed an exogenous time sequence of the shock variables z 1 and
s 1 into the model. In the stochastic version of the model, the shock variables are assumed to
be given by the autoregressive processes:
0 E;; 0 < 1, X1 - N(O, a;), X1 i.i.d. (88)
s t+1 = wsr- c t + 1• 0 E;; w < 1, C1 - N(O, a~), C1 i.i.d. (89)
so your first sheet should also allow you to choose the autocorrelation coefficients 0 and o;
and the standard deviations ax and a c· We suggest that you simulate the model over 1 00
periods, assuming that the economy starts out in long-run equilibrium in the initial period 0 (so
that all of the variables y, ir, e', x, c, z and s are equal to zero in period 0 and period -1 ). From
the internet address www.econ.ku.dk/pbs/courseslmode/s&data.htm you can download an
Excel spreadsheet with two different 1 00-period samples taken from the standardized normal
distribution. Choose the first sample to represent the stochastic shock variable x~ and the
second sample to represent the shock variable Cr To calibrate the magnitude of the shocks
x, and c 1, you must multiply the samples from the standardized normal distribution by the
respective standard deviations a, and a c·
Apart from listing the parameters of the model, your first sheet should also list the standard
deviations of output and inflation as well as the coefficient of correlation between output and
inflation and the coefficients of autocorrelation for these variables (going back four periods)
emerging from your simulations of the stochastic model version. It will also be useful to
include diagrams illustrating the simulated values of the output gap, the inflation gap and the
real exchange rate.
For the simulation of the deterministic version of the model, we propose that you use the
parameter values
q:>= 0.5, y= 0.2, m = 0.3, f!x = I'J M= 1.5, fJ = 2,

h =0.5, T = 0.2, f! = 0.5, 'lo = 0.3, Oy = 0.8.
3. Use the deterministic model version to simulate the dynamic effects of a temporary negative
demand shock, where z 1 = -1 in period t = 1 and z 1 = 0 for all t;;. 2. Illustrate the evolution of
the output and inflation gaps in diagrams and comment on your resu lts. Investigate how the
economy's adjustment to the shock depends on the conduct of monetary policy by varying the
value of the parameter h. Try to explain your findings.
4. Vary (one by one) the parameters q:> and fJ to explore how the economy's dynamic reaction to
the temporary demand shock depends on the way expectations are formed. Try to explain
your resu lts.
5. Now use the deterministic model to simulate the dynamic effects of a temporary negative
supply shock, where s 1 = + 1 in period t = 1 and s 1 = 0 for all t;;. 2. Illustrate the evolution of the
output and inflation gaps in diagrams and comment on your results. How do the results
depend or the monetary policy parameter h? Does monetary policy face a dilemma? Explain.
6. Simulate the stochastic version of the model, setting o=w = 0.5, and a; = a~ = 1, while main-
taining the other parameter values given above. From this point of departure, try to adjust the
values of cp, o, w, a;
and a~ so as to achieve a better match between the model-simulated
standard deviations and coefficients of correlation and autocorrelation for output and inflation
and the corresponding data for the UK economy presented in Tables 14.2-14.4. For your
information, the standard deviation of the cyclical component in the UK inflation rate was 0.06,
according to the data underlying Tables 14.3 and 14.4. (When trying to reproduce the
observed correlation between output and inflation, you should only focus on the contempora-
neous coefficient of correlation. Moreover, note that our simplified model cannot be expected
to rep roduce the data with great accuracy.) Assuming that your parameter values are plausi-
ble, what are the implications of your analysis for the relative importance of demand shocks
and supply shocks as drivers of the UK business cycle?
Model for the Open
26. The choice of exchange
rate regime and the theory
© The McGraw-Hill
Companies. 2005
Macroeconomics Economy of optimum currency areas
Chapter
The choice of exchange

rate regime and the
theory of optimum
currency areas
ne of the great debates in economics is whether the exchange rate should be fixed
0 or hP. :Jll owP.cl to fl o:Jt. This mntrovP.rsy is still nnrP.~olvP.cl.
exchange rate regime involves some dillkult trade-o!Ts. and because different coun-
tries have had diflerent experiences with their exchange rate policies. A further reason why
hf\r.:JnsP. thf\ r.hoir.P. of
the debate continues is that economists and policy makers diller in their views on the work-
ings of the foreign exchange market. While some stress the ability of a floating exchange
rate to act as an absorber of shocks to the economy, others emphasize that a floating
exchange rate may be an independent source of shocks, as we explained in Chapter 25.
Building on the three previous chapters. this chapter reviews the most important
arguments which have been made in the debate on the choice of exchange rate regime. To
provide some background, Section 1 starts by categorizing the different exchange rate
regimes found in the world today and discusses some historical trends in the international
monetary system. Section 2 then explains how the choice of exchange rate regime may be
seen as a choice of framework lor monetary policy and stab~ization policy. In this part of
the analysis we maintain the view embodied in our AS-AD model that the exchange rate
regime does not allect any real economic variables in the long run, although it does affect
the way the economy reacts to shocks in the short and medium term. The tina! section goes
beyond our AS-AD model by considering some microeconomic factors suggesting that the
exchange rate regime may have some structural long-run effects on the real economy alter
all. This part of the analysis will tal<e us into the theory of optimum currency areas which
has played a prominent role in the debate on monetary union in Europe.
26.1 ~.~.~.(;l!l:~~.!.~.~~.E~.~~.~.~.: ..P..(;l~.~ ..(;l!l:~..P.E~.~~!l:.~..........................................................

Alternative exchange rate regimes
In Fig. 2 6.1 we have grouped the various exchange rate regimes found in the world into
three categories denoted 'hard peg', 'intermediate' regimes, and 'float'. 1 The exchange
1. Figure 26.1 is updated from Stanley Fischer, 'Exchange Rate Regimes - Is the Bipolar View Correct?' Journal of
Economic Perspectives, 15, Spring 2001, p. 4.
798
Sorensen-Whitta-Jacobsen: I Part 7- The Short-run 26. The choice of exchange © The McGraw-Hill
Introducing Advanced Model for the Open rate regime and the theory Companies. 2005
26 CHOICE OF EXCHANGE RATE REGIME AND THE THEO RY OF OPTIMUM CURRENCY AREAS 799
'iii 70%
§ O:lv/o
0
Q) 60%
(98) 1111 1991 0 2001
I
0)
c"'
Q) 50% 45%--
e
Q)
(80)
Q.
40% -
"'
Ul
31%
(59)
"'
Ul
Q)
·;::
30%
26%
(4m 2-3o/o- -
c::> 16%
(36)
0 20% -
u
0
~
:;; 10% 1- --- --- -
.D
E
::>
z 0%
Hard peg Intermediate Float
Figure 26.1: Exchange rate regimes in IMF member countries, 1991 and 2001
Note: Perce1tage of all countries; number of countries in brackets.
~ource: IM~ Annual Keport 2001.
rate arrangements described as hard pegs include situations where countries have no
national currency. either because they are in a currency union like the European
Monetary Union, or because they have 'dollarized' by formally adopting the currency of
some other country, typically the US dollar. A hard peg can also take the form of a currency
bonrd where the central bank is obliged by law to exchange domestic currency for a
speciHed foreign currency at a completely flxed exchange rate. and where domestic
currency can only be issued against foreign currency so that the domestic monetary base
is fully backed by foreign exchange reserves.
The 'tloating' group contains economies whose systems are described by the Inter-
national Monetary Fund as either 'independently floating' or as a 'managed float'. Under
independent floating the exchange rate is market-determined. To the extent that the
central bank intervenes in the foreign exchange market by buying or selling domestic
currency against foreign currency, such interventions only aim to moderate undue fluc-
tuations in the exchange rate. but the interventions do not seek to establish a particular
level for the exchange rate. Under a managed float the central bank influences the move-
ments of the exchange rate through active intervention in the foreign exchange market
without specifYing. or precommitting to, a predetermined path for the exchange rate. This
is what we referred to as 'dirty' floatin g in the previous chapter.
The 'intermediate' group consists of economies with a variety of exchange rate
arrangements fallin g bet\<veen the hard pegs and the floating group. For example. some
countries have set a central parity for the exchange rate against a particular foreign cur-
rency or against a basket (a weighted average) of currencies. but the actual exchange rate
is then allowed to fluctuate within a flxed band around the parity. Other countries operate
a 'crawling band' where the band for the exchange rate is allowed to move over time, or a
'crawling peg' where the exchange rate is temporarily flxed, but where it is allowed to shift
gradually over time. Table 26.1 summarizes the exchange rate regimes prevailing in the
world's developed market economies at the beginning of the twent:y-frrst century.
Model for the Open
© The McGraw-Hill
Companies. 2005
800 PAR T 7: THE S HORT-R UN MODEL FOR TH E O PEN ECO NOM Y
Table 26.1; Exchange rate regimes in developed market economies 2001

Euro area Other
Ex change arrangement Exchange arrange ment
Austria No separate legal tender Australia Independent float

Belgium No separate legal tender Canada Independent float
Finland No separate legal tender Denmark Pegged rate in horizontal band
France No separate legal tender Hong Kong SAR Currency board
Germany No separate legal tender Japan Independent float
Greece No separate legal tender New Zealand Independent float
Ireland No separate legal tender Norway Managed float
Italy No separate legal tender Singapore Managed float
Luxembourg No separate legal tender Sweden Independent float
Netherlands No separate legal tender Switzerland Independent float
Portugal No separate legal tender United Kingdom Independent float
Spain No separate legal tender United States Independent float
Source: IMF Annual Report 2001.
Going back to Fig. 26.1, we observe a remarkable polarization of the choice of

exchange rate regimes over the last decade. In 1991 almost two-thirds of all countries had
some form of intermediate exchange rate arrangement. A decade later a lot of countries
had moved either towards a hard peg or towards floating. This is not coincidental. since
almost all of the serious foreign exchange crises th at occurred in the 1990s involved some
form of intermediate exchange rate regime. This was tme of the crisis in the European
Monetary System in 1992- 93 and of the crises in Mexico 1994, Thailand, Indonesia and
South Korea in 199 7. Russia and Brazil in 1998, and Turkey in 2000. 2 The background
for these crises was the huge increase in international capital mobility during the 1990s
which greatly increased the scope for speculative attacks against 'soft' fJXed exchange rate
regimes. The weakness of a fixed exchange rate regime where the exchange rate can be
adjusted is that speculation is virtually risk-free: if an investor moves out of a currency
which is expected to be devalued, he will obviously gain from this move if devaluation
actually occurs. and if the exchange rate is maintained h e will lose nothing except a small
transaction cost. Alerted by the numerous foreign exchange crises during the 1990s.
many countries with a preference for stable exchange rates moved towards a hard peg.
whereas countries with a preference for monetary policy autonomy moved towards lloat-
ing. Indeed. it is now widely believed that intermediate exchange rate regimes tend to be
unsustainable in the long run in the modern world of high capital mobility. 1
2. Argentina was also hit by a serious c risis in 2001 even though the countr1 had a currency board. However, the
Argentinean currency board had a 'soft' element since it allowed foreign exchange reserves to fall below the level of
the domestic monetary base in exceptional c ircumstances.
3. Denmark is a rare example of a country w ith an 'intermediate' exchange rate regime w hich has nevertheless escaped
serious speculative attacks on its currency for many years. The Danish exchange rate parity against the German
D·mark - and later on against the euro - has been unchanged since 1987, and except for a short time during
the EMS crisis n 1993, the market exchange rate of the krone has been kept very close to the central parity. The
stability enjoyed under the Danish fixed exchange rate regime is often ascribed to the fact that fiscal policy has been
quite disciplinec in Denmark where successive governments have run public budget surpluses for many years.
26 CHOICE OF EXCH ANGE RAT E REGIME AND THE THEO RY OF O PTI MUM CURRENCY AREAS 801
The classical gold standard
Historically the developed countries in the world have made several attempts to establish
an in ternational monetary system based on fixed exchange rates. One important and
interesting historical example of a fixed exchange rate regime was the classical gold stan-
dard which had its heyday from around 1870 up until the outbreak of the First World
War in 1914. During this period almost all of the countries participating in the inter-
national division oflabour gradually made their currencies convertible into gold at a ftxed
price. All agents in the private sector could thus exchange domestic currency for gold with
their central bank at the quoted official price.
This convertibility of currency into gold implied that exchange rates could only
fluctuate within a fairly narrow band determined by the cost of shipping gold from one
country to another. To see this. consider Fig. 26.2 where the DD' curve and the SS' curve
indicate the demand for and the supply of foreign currency which would prevail in an
unregulated market lor foreign exchange. If the exchange rate (the price of foreign
currency) were allowed to float freely. the market for foreign exchange would equilibrate
at the exchange rate Ef in Fig. 2 6.2 . However, suppose the domestic as well as the foreign
country have made their currencies fully convertible into gold and that the oflicial price of
one ounce of gold is one unit of the foreign currency and E units of the domestic currency.
Suppose further that gold can be transported between the two countries at a domestic-
currency cost of c per ounce. A domestic resident will then only be willing to pay the price
E + c lor a unit of the foreign currency. If the market exchange rate were higher than this,
it would be cheaper for the domestic resident to buy an ounce of gold in the domestic
central bank at the otncial price E. ship the gold abroad at the cost c and sell it to the
foreign central bank in return lor a unit of the foreign currency. Clearly the total cost of
acquiring the foreign currency through this transaction in gold would be E +c. and no
optimizing firm or household would be willing to pay a higher market price for foreign
currency. Hence there would be no market demand for foreign currency at an exchange
rate in excess of E +c. In Fig. 26.2 we h ave indicated th is by 'dotting' the DD' curve above
the level E+ c.
At the same time as the market demand lor foreign exchange collapses to zero, the
supply of foreign exchange would tend to become intinitely large if the exchange rate were
to exceed E+c. For example. suppose the market exchange rate were Ef > E+c. A domes-
tic investor could then score a pure profit by buying gold domestically at the price E, ship-
ping it abroad at the cost c. selling the gold to the foreign central bank in return for a unit
of the foreign currency, and then selling the foreign currency in the exchange market,
thereby earning a net profit equal to Ef- (E +c). Of course. as all investors tried to take
advan tage of this arbitrage opportunity, their collective trading would make the market
supply of foreign currency infinitely elastic at the level E+ c, as illustrated by the dotting of
the SS' curve in Fig. 26.2. In the situation depicted in the flgure. the free-market exchange
rate Ef would thus never be realized. Instead. the exchange rate would hit the ceiling E + c
where gold would start to fl ow from the domestic to the foreign country.
By a similar line of reasoning, you should convince yourself that demand and supply
in the foreign exchange market under the gold standard would become infinitely elastic at
the exchange rate E- c where gold would start to flow from the foreign into the domestic
economy. Thus the convertibility of currencies in to gold eflectively established a system of
Model for the Open
© The McGraw-Hill
Companies. 2005
802 PAR T 7: TH E S H ORT-R UN MODEL FOR TH E O PEN ECO NOM Y
0' , 1 S'
\ I
\ I
\ I
\ I
E' --------------- ::X
I \
I \
I \
I \
~ - A 1
'B
J ::~---------I--------\-------- I \
I \
I \
I \
I \
/ \
s 'o
Q uantity of foreign currency
Figure 26.2: The mechanics of the classical gold standard
llxed exchange rates where market rates could only fluctuate within a band of width 2c
determined by the cost of transporting gold across borders. as illustrated in Fig. 2 6.2.
One intriguing feature of the classical gold standard was that it tended to be self-
regulating. In the situation depicted in Fig. 26.2, where market forces generate an excess
demand lor foreign exchange equal to the distance AB. there is a corresponding outflow of
gold from the domestic economy. As domestic agents exchange domestic currency for gold
in the central bank, the domestic supply of base money shrinks. and the price of domestic
liquidity (the domestic interest rate) tends to rise. This makes it more prolltable lor inter-
national portfolio investors to invest in domestic interest-bearing assets, so the supply
curve, SS' in Fig. 26.2, shifts to the right as foreign investors ofl"er foreign exch ange in
return for domestic currency in order to acquire domestic assets. At the same time the
demand curve. DD', shifts to the left. since the higher domestic interest rate makes it less
prolltable for domestic residents to invest abroad rather than at home (so that domestic
portfolio investors need less foreign currency), and since there is lower demand for foreign
currency lor the purpose of importing foreign goods, as the rise in the domestic interest rate
reduces import demand via lower domestic economic activity. In this way the outflow of
gold tends to eliminate the excess demand lor foreign exchange by shifting the point ofinter-
section between theDD' curve and the SS' curve down towards the level E + cin Fig. 26.2.
Indeed, under the classical gold standard central banks typically raised their interest rates
as soon as there was a tendency lor gold to flow out of the country. Thus monetary policy
helped to keep market exchange rates '.vi thin a narrow band around the gold parity E.
Despite these self-regulating forces, the classical gold standard broke dmoV!l under the
pressure of the First World War. Many governments found it impossible to llnance the war
ell"ort without resorting to the age-old practice of printing money, so they abolished the
convertibility of th eir currencies into gold in order to be able to expand the money supply.
After the war most countries tried to restore the gold standard. but the system broke down
again during the Great Depression of the 19 30s when many countries tried to stimulate
output and employment through a devaluation of their currencies.
The international monetary system after the Second World War
After the Second World War a new international system of 11xed exchange rates was
established as part of the so-called Bretton ·woods agreement. Under this system countries
pegged their currencies to the US dollar. allowing their exchange rates to tluctuate within
a narrow band of ±1 per cent against the dollar. The dollar itself was made convertible into
gold at a tlxed price and became the main international reserve currency used by central
banks. reflecting the dominant role of the United States in the world economy. Countries
were supposed to devalue their currencies only in the event of a so-called 'fundamental
disequilibrium' on their balance of payments. and the International Monetary Fund was
established to provide international credit to countries which ran into temporary balance-
of-payments crises.
The fixed exchange rates of the Bretton Woods system came under speculative attack
in the late 1960s as international capital flows expanded. and the system broke dm.vn in
a massive wave of speculation between 19 71 and 19 73 . Since that time the major cur-
rencies in the world have been tloating against each other. but at the same time the
member states of the European Union h ave made repeated efforts to create exchange rate
stability within Europe. During the 19 70s most of the EU cow1tries made a half-hearted
attempt to keep their bilateral exchange rates within a fairly narrow band called the
'currency snake' (because the band could tl uctuate vis-a-vis the currencies of third coun-
tries) . and from 19 79 they established a more ambitious system of 11xed but adjustable
exchange rates in the form of the European l'vlonetary System. In 1999 these etlorts
culminated in the formation of the European Monetary Union; the ultimate fixed
exchange rate arrangement where member countries h ave irrevocably 11xed their
bilateral exchange rates by giving up their national currencies in favour of a conunon
currency, the euro. Yet. even for EMU countries the ejfective exchange rate - the weighted
average exch ange rate of the euro vis a vis the trading partners of the EMU is not fiXed
because the euro is floating against the most important currencies such as the US dollar,
the British pound and the Japanese yen.
The macroeconomic trilemma
In Chapter 24 we saw that a system of 11xed exchange rates coupled with free capital
mobility makes monetary policy impotent. This is an illustration of a fundamental macro-
economic 'trilemma', also known as the 'Impossible Trinity'. The trilemma arises because
a macroeconomic policy regime can include at most two out of the following three policy
goals:
1. free cross-border capital !lows,
2. a 11xed exchange rate. and
3. an independent monetary policy.
Model for the Open
© The McGraw-Hill
Companies. 2005
804 PAR T 7: THE SHO RT-R UN MODEL FO R THE OPEN ECONOMY
The Impossible Trinity is easily understood once we recall the familiar condition for
uncovered interest parity which assumes perfect capital mobility. In our usual notation
this condition reads:
(1)
If a country fixes its exchange rate while at the same time allowing free international
capital flows, it follows from (1) that the domestic nominal interest rate i becomes tied to
the foreign nominal interest rate i~ since credibly fixed exchange rates imply that the
expected rate of exchange rate depreciation, e~ 1 - e, becomes zero. Alternatively, if a
country wants to be able to set its own interest rate independently of the foreign interest
rate while allowing capital mobility, it follows from ( 1) that it must also allow its exchange
rate to vary. As a final alternative. since (1) is enforced by capital mobility, a country must
impose capital controls if it wants to pursue an independent interest rate policy while at
the same time llxing its exchange rate.
Over the years countries have resolved the fundamental macroeconomic trilemma in
dillerentways. Table 26.2 provides a simpliHed summary of the historical experience with
the trilemma. drawing on a comprehensive study by US economists Maurice Obstleld and
Alan Taylor. During the era of the classical gold standard up until the First World War.
there was a broad consensus among Western governments about the desirability of fixed
exchange rates and free international mobility of capital. Thus governments were willing
to subordinate their monetary policies to the goal of protecting their gold reserves rather
than using monetary policy to stabilize the domestic economy. Indeed, under the domi-
nant laissez-faire philosophy oft hat time, policy makers did not perceive a need for activist
stabilization policies since they tended to believe strongly in the self~regulating forces of
the free market.
When the economic crisis of the interwar years struck the Western world, this laissez-
faire attitude came under heavy attack, and during the Great Depression most countries
started to use monetary policy to stimulate the domestic economy. To be able to do so,
they either had to impose capital controls. or they had to abandon their fixed exchange
rates and adopt a policy of devaluation or of flexible exchange rates. The latter route was
Table 26.2: The trilemma and major phases of capital mobility

Resolution of trilemma -
Countries choose to sacrifice:
Activist Capital Fixed

Era policies mobility exchange rate Notes
Gold standard Most Few Few Broad consensus

Interwar (when off gold) Few Several Most Capital controls especially in Central
Europe, Latin America
Bretton Woods Few Most Few Broad consensus
Float Few Few Many Some consensus; except currency
boards, dollarization, etc.
Source : Reprinted from Table 1, Chapter 3 : 'Globalization and Capital Markets', by Obstfeld and Taylor
in Michael D. Bordo, Alan M. Taylor and Jeffrey G. Williamson (eds),
Globalization in Historical Perspective. The University of Chicago Press © 2003 by the National Bureau of Economic Research.
26 CHOICE OF EXCH ANGE RAT E REGIME AND THE THEORY OF OPTIMUM CURRENCY AREAS 805
chosen by a large number of countries which were thereby able to recover more quickly
from the depression th an the countries that stuck to the gold standard.
After the Second World War very few cow1tries were willing to give up the possibility
of pursuing an activist monetary stabilization policy, but at the same time they feared a
return to the beggar-thy-neighbour policies of the 1930s where countries h ad tried to
gain competitiveness at each other's expense through large aggressive devaluations. As a
consequence. the new Bretton Woods system of ibced exchange rates explicitly allowed
countries to maintain capital controls to gain some monetary policy independence.
However, over time capital controls became h arder to maintain, as communication
and information technologies developed and the volume of international trade trans-
actions increased. Moreover, restrictions on the ability of investors to seek out the most
prolltable international investment opportunities came to be seen as a source of economic
inelllciency. Still, the most important countries remained unwilling to give up an inde-
pendent monetary policy. As a consequence. several countries moved towards floating
exchange rates combined with free capital mobility during the 1970s and 19R0s, but at
the same time the formation of the European Monetary Union at the turn of the century
created an important European region with completely llxed exchange rates. free capital
mobility and a common monetary policy. involving a sacrifice of national monetary
policies.
26.2 ~~.~~.~.Y.~~~.~.~..~.~.~~~.~~..~~g~~.~~..~.?:~~'~'''''"'''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''
For many decades economists have debated the pros and cons of llxed versus flexible
exchange rates. This controversy will surely continue, but going through the main argu-
ments will help us understand some of the factors determining whether a particular
country is better served by one or the other exchange rate regime. Indeed. our analysis
will suggest that dill'erent exchange rate regimes may be suitable for dill'erent countries,
depending on their individual circumstances.
As we saw in Chapter 23, our simple AS-AD framework implies that the exchange
rate regime is 11eutral in the long run. In the present section we will maintain this asswnp-
tion. If the exchange rate regime cannot influence the long run equilibrium values of any
real variables. the main issue in choosing between llxed and flexible exchange rates is
whether the short-run fluctuations in output and inflation will be systematically smaller
under one regime or the other? Related to this is the question whether the economy's
speed of adjustment to long-nm equilibrium will be faster under one regime or the other?
The analysis in the previous chapter suggests that the answers to these questions depend
on the nature of the shocks hitting the economy and on the parameters of the aggregate
demand curve. In particular, we found that when demand shocks are dominant, the
short-run instability of output will be smaller under flexible than under llxed exchange
rates, but when supply shocks are the dominant drivers of business cycles, output will be
less unstable under a fbced exchange rate. Since the nature of shocks and the economy's
capacity to absorb them may vary over time and across cow1tries, there is no reason to
believe that all countries will prefer the same exchange rate regime at all times.
In the following we will elaborate on the factors determining which exchange rate
regime offers the best potential cush ion against short-run macroeconomic instability.
© The McGraw-Hill
Companies. 2005
The exchange rate regime as a framework for monetary policy
Under free capital mobility we have seen that the choice of exchange rate regime is inti-
mately linked to the choice of monetary policy regime: if a country faced with high capital
mobility wants to pursue an independent monetary policy, it must adopt a flexible
exchange rate. It is therefore useful to consider the goals and targets for monetary policy,
for once these have been decided, they will dictate the choice of exchange rate regime in a
world with mobile capital.
In recent years a widespread consensus has emerged that the prime goal of monetary
policy should be to secure a low and stable rate of inflation. This is based on the view also
embodied in our AS-AD model that in the long run monetary policy can only influence
nominal variables, but not real variables. Of course monetary policy can affect real vari-
ables like output and employment in the short and medium term. However, if the public
believes that price stability (a low and stable inllation rate) is the prime goal of monetary
policy, an expansionary macroeconomic policy can actually become a more powerful tool
in a recession, since it is easier to stimulate aggregate demand without igniting inflation
expectations when the central bank's commitment to long-run price stability is credible.
Making low and stable inflation the main o!Iicial goal of monetary policy is therefore seen
by most economists and policy makers as the best contribution monetary policy can mal<e
to economic stability.4
If this is indeed the main goal of monetary policy, it might seem most logical to adopt
an inflation targeting regime and let the exchange rate Itoat to enable monetary policy to
concentrate on the domestic goal of price stability. But monetary policy cannot directly
control the intlation rate, since it only works indirectly and with a time lag through its
influence on aggregate demand (and possibly its impact on expectations). As an alter-
native to inflation targeting with flexible exchange rates, a country may therefore choose
to peg its exchange rate to the currency of a country with a record of low and stable
inflation. Let us consider this option more closely before returning to inflation targeting.
A fixed exchange rate as an intermediate target for monetary policy
A fixed exchange rate may be seen as an intermediate target for monetary policy. A fixed
exchange rate regime means that the instruments of monetary policy - the short-term
interest rate and foreign exchange market interventions - become fully and exclusively
dedicated to achieving the intermediate target of a fixed exchange rate. However. this
intermediate target is adopted with the purpose of realizing the ultimate policy goal of price
stability. By credibly pegging to a hard foreign currency, a country can import a low and
stable inflation rate from abroad. since the domestic inflation rate must equal the inflation
rate of the hard-currency country over the long run for the fixed exchange rate to be
sustainable.
In general, the rationale tor adopting an intermediate target for monetary policy is
that the central bank cannot directly control the ultimate goal variable such as the rate of
inflation. It may then be useful to choose an intermediate target variable which is easier to
4. We should not exaggerat e the degree of consensus on this point. Economists do disagree on the extent to w hich
monet ary policy can affect the economy in the short and medium run, and therefore they also disagree on the extent
to which st abilization of the real economy should be included in the official goals of monetary policy.
control by means of the monetary policy instntments and w hich is systematically related
to the u ltimate goal variable. Ideally, an intermediate target variable should have a high
correlatio11 with the goal. so that achieving the intermediate target is an ellective way of
fullllling the ultimate policy goal. The intermediate target should also be easier to control
than the goal - otherwise there is no point in adopting the target rather than focusing
directly on the goal. Finally, it should be easy for outsiders to observe whether the inter-
mediate target is ach ieved, since this makes it easier for the political authorities and the
general public to hold the central bank accountable for its actions.
A fixed exchange rate regime has obvious advantages by the last two criteria. In
normal times when the country is not exposed to a speculative attack on its currency, it is
easy for the central bank to control the exchange rate via foreign exchange market inter-
ventions combined with an interest rate policy that follows the policy of the foreign
anchor country. 5 When the exchange rate is only allowed to move within a narrow band,
it is also easy for outsiders to observe whether the intermediate target is achieved or not:
either the exchange rate is (roughly) tixed. or it is not.
However. a fixed exchange rate is not necessarily closely correlated with the ultimate
goal of low and stable inflation. As you recall. the evolution of the (log of the) real
exchange rate e' is given by:
(2)
where t. e = e- e_1 is the rate of exch ange rate depreciation, n f is the foreign inflation rate,
and n is the domestic inflation rate. From (2) we see that under fixed exchange rates
(t. e = 0) the domestic inflation rate may be written as:
(3)
A low and stable domestic inflation rate thus requires a low and stable inflation rate in the
foreign anchor country plus stability of the real exchange rate, i.e., e' = e~ 1 • The Hrst con-
dition will be met if the anchor country is indeed committed to low and stable inflation,
but this does not in itself guarantee stability of the real exchange rate. Such stability
depends on the similarity between economic conditions at home and in the foreign anchor
country. To illustrate this point, Fig. 26.3 shows the case of a negative supply shock
wh ich permanently reduces th e natural level of output. As a result of the sh ock, the long
run aggregate supply curve of the domestic economy shifts to the left. If the supply shock
is 'asymmetric' - that is, if it only a fleets the domestic but not the foreign economy - there
will be no shock to net exports since foreign economic activity will be unchanged. The
LRAD curve will then be unallected, and the domestic economy will end up in a new long-
run equilibrium at point A 1 where the real exchange rate has appreciated relative to the
original equilibrium point A0 .
However, suppose instead that the supply shock is 'symmetric', allecting the foreign
economy to the same extent as the domestic economy. Then foreign economic activity will
decrease, and this will shift the LRAD curve to the left by reducing net exports trom the
domestic economy. If the long-run aggregate demand curve shills to the new position
LRAD', the domestic economy will end up in a new long-run equilibrium at point A'1
where the real exchange rate is unchanged. In practice, the shift in the LRAD curve may
5. We use tte term 'anchor country' for convenience: of course this 'country' could be a region such as the euro zone.
Model for the Open
© The McGraw-Hill
Companies. 2005
808 PAR T 7: THE SHORT-R UN MODEL FOR TH E OPEN ECO NOMY
e'
LRAS1 LRAS0
LRAD'
LRAD
y, Yo y
Figure 26.3: The long· run effects of an asymmetric versus a symmetric negative supply shock
not be exactly sufficient to ensure an unchanged real exchange rate, but a symmetric
shock will always imply a smaller change in the equilibrium real exchange rate than an
asymmetric shock.
If the foreign and the domestic economies are closely integrated and have very similar
structures, it is more likely that a shock afl'ecting one country will have a roughly similar
impact on the other country. For example. if the negative supply shock is a permanent
increase in the relative price of imported oil. the avo countries will be affected equally if
they are equally dependent on imported oil. In that case the effect of the shock would be
described by the movement from point A0 to point A; in Fig. 26.3 . But if the domestic
economy is frequently hit by asymmetric shocks. then it may need frequent adjustments
of its real exchange rate. and under a fixed nominal exchange rate this can happen only
through fluctuations in the domestic rate of inflation. For instance, in the case of a nega-
tive asymmetric supply shock, the real appreciation implied by the movement from the
original equilibrium 110 to the new equilibrium A 1 in Fig. 26.3 can occur only through an
adjustment period in which the domestic infl ation rate exceeds the foreign inflation rate.
To sum up the argument so far, a close correlation between the intermediate target of
a fixed nominal exchange rate and the ultimate goal of a low and stable inflation rate
seems to require that the domestic economy is hit by roughly the same type of shocks as
the foreign anchor economy - and that the two economies react in roughly the same way
to the shocks - so that large and frequent adjustments in the real exchange rate are not
needed.
But th is argument needs to be modified. Consider again our example of a negative
asymmetric supply shock which shifts the LRAS curve to the left. If the domestic
Sorensen- Whitta- Jacobsen: I Part 7- The Short-run 26. The choice of exchange © The McGraw-Hill
government wants to neutralize the impact of this sh ock on the equilibrium real exchange
rate. it can permanently tighten fiscal policy since such a permanent negative demand
shock will shill the LRAD curve to the left. By an appropriate dose offiscal contraction, the
domestic govermnent can thereby move the economy fi·om the equilibrium A0 to the new
equilibrium A't in Fig. 26.3. In this way it can avoid the temporary rise in domestic
inflation which would otherwise be needed to take the economy to the alternative equili-
brium A 1• In general, by assigning to fiscal policy the task of minimizing fluctuations in
the real exch ange rate (keeping the domestic inflation rate close to the foreign in flation
rate). a country with a fixed nominal exchange rate should be able to avoid substantial
fluctuations in the rate of inflation even if it is hit by asymmetric shocks.
Sceptics argue that it is unlikely that fiscal policy will be able to live up to such a task.
First of all, the required frequent changes in taxes or public spending may have undesir-
able side effects on resource allocation and income distribution and may therefore be polit-
ically dill1cult to implement. In particular, in cases when a fiscal contraction is needed,
this may be very hard to carry out because of its political unpopularity. Second. usually
there are long lags in the fiscal policy process. as we explained in Chapter 22, and this may
make it very dillicult to achieve the appropriate timing of temporary fiscal measures
intended to ollset the eflects of temporary shocks.
For these reasons many economists believe that fiscal policy L~ not a very ellective
instrument of stabilization policy, at least when it comes to dealing with shocks of a
shorter duration. To the extent that fiscal policy does not succeed in minimizing fluctua-
tions in domestic inflation. a flxed exchange rate regime may then imply that monetary
policy becomes procyc/ical, that is. destabilizing. To see how this may happen, let us return
to the idea presented in Chapter 24 that only a fraction qy of the population has 'weakly
rational' expectations (n" = n f) while the remaining fraction may have static inflation
expectations (n' = n _1) . The average expected inflation rate will then be given by:
n ' = rpnf + (1 - qy)n _1 • (4)

In Chapter 24 we focused on the benchmark case where qy = 1, but suppose instead that
qy < 1. An asymmetric negative demand shock which forces the economy into recession
will drive down the domestic rate of inflation. According to (4) this will reduce the
expected inflation rate in the subsequent period. But since the central bank has to keep the
domestic no1ninal interest rate equal to the (unchanged) foreign nominal interest rate to
deli'md the fixed exchange rate. it follows th at a fall in the expected inflation rate will drive
up the domestic real interest rate in a time of recession. Obviously this will further depress
aggregate demand. which is exactly the opposite of what is needed during a recession. By
analogy, under a boom which raises the actual inflation rate. the real interest rate for the
following period will fall when the nominal interest rate has to be kept constant to fix the
exchange rate and part of the population has backward-looking expectations. &
The vulnerability of a fixed exchange rate regime to speculative currency attacks also
tends to increase the risk that monetary policy becomes procyclical. For example, if the
6. This argun ent against fixed exchange rat es is sometimes referred to as the 'Walters Critique' because it was formu-
lated by the B ritish economist Alan Walters who served for a time as an economic policy adviser to former Prime
M inister Margaret Thatcher. See Alan Walt ers, Britain's Economic Renaissance, O xford, Oxford University Press,
1986.
Model for the Open
© The McGraw-Hill
Companies. 2005
810 PAR T 7: THE S H O RT-R UN MODEL FO R TH E O PEN ECO NOM Y
economy is in a deep recession with mass unemployment, international investors may

begin to speculate that the authorities will leel tempted to devalue the currency in order to
escape from the recession. If such expectations become widespread, a speculative attack
may occur, forcing the domestic central bank to raise the domestic interest rate to defend
the currency. Clearly this procyclical policy will only tend to deepen the recession.
This line of reasoning combined with scepticism regarding the scope for llscal stabi-
lization policy explains why quite a few economists believe that business lluctuations will
tend to be larger under llxed than under flexible exchange rates, other things equal.
The alternative: flexible exchange rates with inflation forecast targeting
Given these di!Ilculties with a flxed exchange rate as a framework lor monetary policy,
many countries have recently adopted llexible exchange rates combined with inllation
targeting to secure an anchor for inllation expectations. Cnder inflation targeting. the
central bank's inflation forecast effectively becomes an intermediate target for monetary
policy. 7 As we explained in Chapter 22, the reason is that monetary policy only allects the
inflation rate with a certain time lag which is typically between one and two years. The lag
arises from the fact that it takes time before aggregate demand reacts to a change in the
interest rate, and it also takes time before a change in demand achieves its maximum
impact on inflation. The central bank must therefore set the interest rate today knowing
that this will not significantly allect the inflation rate until l !-2 years from now. Hence
monetary policy must be based on a forecast of future inllation over a time horizon corre-
sponding to the time it takes for a change in the interest rate to inlluence the inllation rate.
If the inflation forecast exceeds (falls short ol) the target inflation rate when the interest
rate is held constant, then the interest rate must be raised (lowered) until the inllation
forecast corresponds to the target. In this sense the inllation forecast is used as an inter-
mediate target lor monetary policy, that is, infl ation targeting implies inflation forecast
targeting.
Of course. the realized future inllation rate will typically deviate from the inllation
forecast since the central bank does not have perfect knowledge of the way its policy
allects the economy. and new unpredictable shocks may hit the economy between the
time the interest rate is changed and the time it attains its impact on the inflation rate.
However, provided the central bank optimally uses all relevant information in producing
its forecast. inllation forecast targeting will make the deviations of actual inllation from
the target rate of inflation as small as existing economic knowledge permits. This is why
advocates of infl ation targeting believe that this monetary policy regime makes it easier to
achieve the goal oflow and stable inflation than a llxed exchange rate regime.
Moreover. the inflation forecast is obviously easier lor the central ban k to control
than the in llation rate itself. and it is also easy for the public to observe. provided the
central bank publishes its forecasts. From these perspectives the inflation forecast seems
to be an attractive alternative to a llxed exchange rate as an intermediate target for
monetary policy.
7. This interpretat ion of inflation target ing regimes as well as several other poin1s made above are well explained in the
highly readable article by Lars E.O. Svensson, 'Exchange Rate Target or Inflation Target for Norway?', in A nne Berit
Christiansen and Jan Fredrik Ovigstad (eds), Choosing a Monetary Policy Target, Oslo, Scandinavian Universi1y
Press (Universitetsforlaget A S). pp. 120-138.
Sorensen- Whitta- Jacobsen: I Part 7- The Short-run 26. The choice of exchange © The McGraw-Hill
However, those w h o are sceptical ofllexible exchange rates h ave pointed to a number
of potential problems with this regime. First of all, in practice cen tral banks do not target
the rate of producer price inllation (our variable n), as we have so far assumed for sim-
plicity. Instead, they target the rate of consumer price inflation which is a weighted
average of the inllation rate for domestic and imported goods. Thus, if the share of imports
in private consumption is !fJ, the rate of consumer price inflation (.n r) targeted by the
central bank is:
inllation rate for inflation rate for
imported goods domestic goods
--.. ~-
n;C = !fJ · (toe+ .nf) + (1 - 1/J) · n , 0 .;;; !fJ .;;; 1. (5)

As we saw in the previous chapter. under the floating regime the nominal exchange rate
(and hence to e) may be subject to considerable exogenous shocks which cannot be ollset
easily by monetary policy in the short run. Such nominal exchange rate shocks will
contribute to volatility in the targeted rate of inflation through their impact on the prices
of imported consumer goods, according to (5). 8
Second, we found in Chapter 2 5 that aggregate demand shocks have a smaller
impact on the economy under llexible than under tixed exchange rates. On the one hand
this means that the economy may be less vulnerable to shocks under flexible rates, but it
also implies that fiscal policy (which is included in our demand shock variable z) becomes
a less ellective tool of stabilization policy. The greater scope lor monetary policy under flex-
ible exchange rates is therefore achieved at the expense of the ellectiveness offlscal policy.
Third, apart from being volatile in the short run, floating exchange rates may
undergo very large fluctuations in the medium run because they tend to overshoot their
long-run equilibrium values. Both types of exchange rate instability may hamper inter-
nationa l economic transactions by increasing their riskiness. It may also cause an unin-
tended redistribution of real income across dillerent sectors in the economy, as we
explained in Chapter 2 4 .
This third point amounts to saying that nominal exchange rate stability should not
only be considered an intermediate target; it may also be a legitimate ultimate policy goal
itself. We shall now elaborate on this argument which takes us into the theory of optimum
currency areas.
26.3 The theory of optimum currency areas

················································································································································································
We have so far maintained the assumption that the exchange rate regime only affects the
economy's short-nm and medium-term reaction to shocks without changing the long-
run equilibrium value of any 'real' economic variables. In practice. there are some reasons
to believe that the exchange rate regime can have a permanent impact on the structure of
the economy. For example. greater exchange rate stability might stimulate international
8. The exten: to w hich exchange rate fluctuations actually cont ribute to nflation instability depends on the degree t o
which changes in the exchange rate are passed through to domest c consumer prices. The empirical evidence
suggests that some foreign producers of imported goods follow a 'pricing ·to·market' strategy, pref erring to keep t heir
prices rel<.tively stable measured in the local currency of their export mar1<et. In that case exchange rate fluct uations
tend to be absorbed by fluc tuations in profit margins rather than in domestic consumer prices.
Model for the Open
© The McGraw-Hill
Companies. 2005
trade and investment by reducing the riskiness of international economic transactions.

According to the theory of optimum currency areas the choice of exchange rate regime
then becomes a matter of trading oil' such microeconomic benents against the potential
macroeconomic costs of fixing the exchange rate.9
The theory of optimal currency areas may be summarized using Fig. 26.4. The
diagram illustrates a country's total costs and beneilts (relative to GDP) of moving from a
flexible to a fully ilxed exchange rate, possibly by adopting the same currency as the major
trading partners. The horizontal axis measures the country's degree of economic integra-
tion with its trading partners measured, say, by the ratio of foreign trade and investment
to GDP. The BB curve indicates the microeconomic benefits from exchange rate stability.
One such benefit is that a credibly fixed exchange rate reduces the riskiness of foreign
trade and investment. This is welfare-improving when economic agents are risk averse.
Further benellts are gained if exchange rate stability is achieved through a common cur-
rency. In that case society saves the real resources (mostly labour time and computer soft-
ware) which are needed to exchange one currency into another and to keep books in
dillerent currencies. It has also been argued that the adoption of a common currency will
generate a welfare gain from stronger international price competition because a single
currency facilitates the comparison of prices across borders. Moreover, by increasing the
volume of transactions carried out in the same currency, a monetary union tends to
increase the liquidity of the markets for financial assets and paves the way lor more
competition and greater economies of scale in the financial sector. 10
The greater the volume of international transaction~ relative to the size of the
economy. the larger are the microeconomic benefits from a stable exchange rate and a
common currency relative to GDP. This is why the BB curve slopes upwards in Fig. 26.4 .
Since several of the benefits mentioned above can only be reaped by adopting a common
currency. the position and positive slope of the BB curve will be higher when the fixed
exchange rate is achieved by entering a monetary union.
The CC curve in Fig. 26.4 indicates the macroeconornic costs of switching from a
flexible to a fully fixed exchange rate, measured as a percentage of GDP. The fact that this
curve lies above the horizontal axis reflects the assumption of optimum currency area
theory that the short-run volatility of output and inllation will tend to be lower under
flexible than under fixed exchange rates. As you recall from the previous chapter, this is
consistent with our AS-AD model if the shocks to the macroeconomy mainly originate
from the demand side. More generally, the assumption of positive macroeconomic costs of
fixing the exchange rate is based on the idea that the ability to pursue an independent
monetary policy under llexible exchange rates makes it easier to stabilize the economy.
But as indicated in the diagram, these macroeconomic costs will be smaller relative to GDP
the more the domestic economy is integrated with the economies of its trading partners.
There are four reasons for the negative slope of the CC curve. First, the value of
nominal exchange rate flexibility is greater the more often the country is exposed to
9. The theory of optimum currency areas was pioneered by the later Nobel Prize winner Robert A. Mundell in 'A Theory
of Optimum Currency Areas', American Economic Review, 51 , 1961, pp. 657-665. This theory has played an
important role n the economic debate on monetary union in Europe.
10. A 'liquid' market is a market with a large and fairly steady volume of day·to·day transactions. In a liquid market, a
seller in need of cash can normally expect to be able to sell his asset without having to accept a significant cut in
its price.
26 CHOICE OF EXCHANGE RAT E REGIME AN D THE THEO RY OF OPTIMUM CURRENCY AREAS 813
c 8
8 c
z X
Degree of economic integra:ion
Figure 26.4: The costs and benefits of a common c urrency
asymmetric shocks requiring a signillcant adjustment ofits real exchange rate. Remember
from the analysis in Fig. 2 6.3 that when a shock hits the domestic country and its trading
partners symmetrically, there is no need lor a real exchange rate adjustment. Moreover, in
the case or symmetric shocks a common interest rate policy will be equally appropriate lor
the domestic and for the foreign economy (assuming that the two countries or regions
have roughly the same social preferences for output stability relative to intlation stability) ,
so there is no need for nominal exchange rate flexibility to allow for dillerent national
monetary policies. Now comes the point: as the domestic and the foreign economies
become more integrated, it is more likely that they will be exposed to the same type of
shocks. 11 Hence they have less need for exchange rate flexibility as the degree of inter-
national economic integration increases. For this reason the macroeconomic cost of
moving to a llxed exchange rate will fall as we move to the right on the horizontal axis in
Fig. 26 .4 .
The second reason for the negative slope of the CC curve is that if the domestic
country allows its nominal exchange rate to depreciate in order to absorb, say, a negative
asymmetric shock to its export demand, the increase in import prices stemming from the
depreciation will be transmitted more quickly to the domestic wage and price level the
11. Actually this relationship is not self-evident, since inc reased international trade may lead different countries to spe-
cialize in different types of production subject to different industry·s~ecific shocks (so-called inter-industry trade).
However, most of the trade among the O ECD countries takes the form of intra-industry trade, that is, exchange of
different variants of the same type of products. For example, France may export C itroen cars to Gennany while
Germany exports BMWs to France. Hence both France and Gennany are vulnerable to shocks hitting the car
industry.
© The McGraw-Hill
Companies. 2005
larger the ratio of imports to GDP. The gain in competitiveness obtained through a flexible
exchange rate will therefore be more short-lived- and consequently the cost of giving up
exchange rate flexibility will be smaller - the more the domestic economy depends on
international trade.
Third, as the domestic and foreign economies become more integrated in terms of
trade and investment. the international mobility oflabour is also likely to increase, since
cross-border economic transactions tend to reduce the information barriers and cultural
barriers to migration. Higher international labour mobility means that if the economy is
hit by a negative asymmetric shock which creates unemployment, some domestic
workers will emigrate to look lor jobs abroad. Hence it becomes less urgent to deal with the
unemployment problem through a depreciation of the domestic currency, so the cost of
giving up exchange rate flexibility falls to the extent that increased economic integration
implies increased labour mobility.
The fourth potential mechanism generating a negative slope of the CC curve is more
speculative and much less certain to occur: if deeper economic integration also leads to
more political integration (think of the EU context. for example), and if political inte-
gration motivates the trading partners to engage in joint supranational activities financed
through a common budget (again, think of the EU budget), then such a budget may be
used as a means of transferring resources from countries benellting from positive asym-
metric shocks to countries bit by negative shocks. Such transfers would reduce the need
for nominal and real exchange rate adjustments to counter asymmetric shocks. However,
there is of course no guarantee that increased economic integration will automatically
induce the integrating countries to implement such an international transfer mechanism.
At a level of economic integration corresponding to point A in Fig. 26.4, the micro-
economic benefits from a llxed exchange rate/common currency are just ollset by the
macroeconomic costs. At a higher degree of economic integration, the domestic country
would benefit in economic terms from fixing its exchange rate or from entering a currency
union with its trading partners (remember that the position of the BB curve will be higher
and its slope steeper in the latter case). At lower degrees of integration, the macro-
economic costs of exchange rate fixity exceed the microeconomic benelits.
Optimum currency area theory does not ofler a quantitative method lor estimating
whether a particular country is to the right or to the lett of the critical point A in Fig. 2 6 .1 .
But the theory does help us to focus on the factors which are important for evaluating
whether fiXing the exchange rate or joining a currency union is a good or a bad idea. The
theory also has one important qualitative long-run implication: if international economic
integration continues to deepen in Europe and elsewhere. then a growing number of
countries should lind it in their economic interest to form/join a currency union with their
most important trading partners. Thus the theory suggests that more European countries
will want to adopt the euro as time goes by.
In the reasoning above we have taken the degree of international economic inte-
gration as given, but of course the adoption of a common currency may in itself tend to
increase the degree of integration by stimulating cross-border trade and investment.
Hence one may conceive that a country which starts out from a point like Z in Fig. 26.4
may end up at a point like X after having joined a currency union with its trading
partners. In other words, even though union membership seems to imply an economic
loss ex ante. it turns out to generate a net economic gain ex post. Much of the economic
Sorensen-Whitta- Jacobsen: I Part 7- The Short-run 26. The choice of exchange © The McGraw-Hill
debate on the Economic and Monetary Union in Europe can be understood in this way.
Those economists who were/are sceptical about EMU membership have tended to argue
that economic integration among the (potential) EMU countries has not yet proceeded far
enough to justify the adoption of a common currency, at least as far as the more peripheral
EU member states are concerned. By contrast. those in favour of monetary union have
tended to argue that. in so far as the preconditions for successful monetary unillcation are
not already present, the adoption of a common currency will in itself promote economic
integration to the point where the economic gains outweigh the costs.
Against this background, let us end this chapter by considering some empirical
evidence bearing on optimum currency area theory.
Some evidence on the relevance of optimum currency area theory
As we have noted. the creation of a common currency area may be expected to stimulate
trade among the members of the currency union because the adoption of a common
currency reduces the transactions costs and riskiness of international trade. Thus we
would expect that. after the introduction of the euro. trade v.rithin the euro area bas
increased by more than trade between EMU and non-El\ill countries. The data in Fig. 26.5
are consistent v.rith this hypothesis. The curves in the diagram indicate the average value
of indices for bilateral trade between the 23 OECD countries included in the data set. The
tendency lor trade to fall towards the end of the sample period reflects the global recession
at thestartofthe present century. However, the clear impression from Fig. 26.5 is that the
euro (and the greater exchange rate stability in the few years before the adoption of the
euro from the start of 1999) has tended to boost trade among EMU countries compared to
the evolution of trade among non-EMU countries. Careful statistical analyses controlling
lor other factors affecting trade have conllrmed this impression . suggesting that the euro
may so far have increased bilateral trade within the EMU by somewhere between 9 per
cent and 3 7 per cent. depending on the estimation method used. 12 Hence the micro-
economic benefits from a common currency may be significant.
It is interesting to see from Fig. 26.5 that the euro also seems to h ave stimulated trade
between EMU and non-EMU countries relative to trade outside the euro area. This may be
because the euro enables firms outside the euro area to save some transaction costs, since
they can use euros earned in transactions with one EMU country to settle transactions
with business partners in another EMU country.
The theory of optimum currency areas claims that the microeconomic benefits from
a common currency will increase with the volume of trade within the currency area, and
that the macroeconomic costs of abandoning national currencies v.rill be smaller the lower
the frequency of asymmetric shocks to individual countries. An absence of large and fre-
quent asymmetric shocks ,>Vi[) be reflected in a high degree of co-movement of national
output levels across the countries in the currency union. From the perspective of optimum
currency area theory. we would thus expect that those EU countries which decided to stay
12. These results can be found in the following two articles evaluating the early experience with the euro: Alejand ro
Micco, Ernesto S tein and Guillenno Ordonez, 'The C urrency Union Effect on Trade: Early Evidence from EMU',
Economic Policy, October 2003, pp. 3 15 -356; and David Barr, Francis B reeden and David Miles, 'Life on the
Outside: Economic Conditions and Prospects Outside Euroland' , Economic Policy, October 2003, pp. 573-613.
© The McGraw-Hill
Companies. 2005
115
110
~ ~ ......
105
100
~
- ~__ ... ...... _.,' ,
""-
II'
' ',,
--- ''
''
95
90
I! ''
''
85
,'J/ ''
''
/ ;f'
80
--- Non EMU -Non EMU
EMU-EMU
I
75 EMU-Non EMU
70
1993 1994 1995 1996 1997 1988 1999 2000 2001 2002
Figure 26.5: The evolution of international trade before and after monetary unions in Europe
Source: Reprinted from Figure 1, p. 3 19 in Alejandro Micco, Ernesto Stein and Guillermo Ordonez, 'The Currency
Union Effect on Trade: Early Evidence from EMU'. Economic Policy, Blackwell Publishing, October 2003,
pp. 3 15-356.
out of the EMU should be those countries which have the lowest degrees of trade inte-
gration and the lowest degrees of co-movement of output with the rest of the EU, since
such countries will have smaller net economic benefits (or larger net costs) of EMU
membership.
The data presented in Table 26.3 suggest that economic considerations may indeed
have played a role in the decisions ofDenmark. Sweden and the United Kingdom to opt out
of the EMU lor the time being. The first column in the table shows the average volume of
bilateral trade as a percentage of GDP among individual countries in the country groups
considered. For example. the fifth figure in the first column shows that, on average.
Denmark's trade (measured as the average of exports and imports) with each EMU
country amounts to 0 .29 per cent of Danish GDP. This trade volume is significantly less
than the typical volume of bilateral trade between any tv.ro EMU countries (0.49 per cent
ofGDP). Sweden and the UK are also seen to have less trade integration with the EMU than
the average EMU country. Hence the potential microeconomic benefits from EMU mem-
bership lor the three outsider countries appear to be less than the benefits for those EU
countries which have already adopted the euro.
The second column in Table 26.3 shows the average absolute deviation between
output gaps (relative to their time series mean) across countries in the different country
groupings. A higher value of this variable means that on average the dillerences in output
gaps have been larger. indicating a lower degree of output co-movement, that is, a lower
degree of synchronization of national business cycles. We see that for all of the three out-
sider countries, the output co-movement with the EMU area has been lower than output
co-movement within the EMU, suggesting that Denmark. Sweden and the UK have tended
to be more exposed to asymmetric shocks than the countries in the EMU. In that case the
Sorensen-Whitta- Jacobsen: I Part 7- The Short-run 26. The choice of exchange © The McGraw-Hill
26 CHOICE OF EXCH ANGE RATE REGIME AND THE THEO RY OF OPTI MUM CURRENCY AREAS 817
Table 26.3; Indicators of the economic attractiveness of a currency union in the EU

Average for the period 1978-91
Bilateral trade as a Output co-movement

percentage of GOP (lower figure = more co-movement)
All EU countries (EU-15) 0.40 2.5

EMU countries 0.49* 2.1 **
UK with EMU 0.31 3.1 **
Sweden w ith EMU 0.34 2.5*
Denmark w ith EMU 0.29 3.0**
Period averages, all EU countres
1978- 88 0.39 2.2
1988- 98 0.41 1.7
Not es: B ilat eral trade is the sum of exports and imports divided by 2. Output co· movement is an indicator of the average absolute deviation
between output gaps.
* S ignificant difference between EMU and non· EMU countries at the 10 per cent level.
** Significant difference between EMU and non-EMU countries at the 1 per cent level.
Source: Tab lco 1 and 2 in David Bo.rr, Francie B rccdon and D avid Milco, 'Life on t he Outoidc:
Economic Conditions and Prospects outside Euroland', Economic Policy, October 2003, pp. 573-613.
three outsider countries would tend to face higher macroeconomic costs of giving up their
national currencies. according to optimum currency area theory. 13
While these findings may help to explain why certain EU countries decided to stay out
of the EMU, they do not explain the timing of the decision to form a monetary union. The
two bottom rows in Table 26.3 are suggestive in this respect. They show that between the
periods 1978-88 and 1988-98. trade integration as well as output co-movement within
all of the EU increased. According to optimum currency area theory, these developments
made monetary union more attractive from an economic perspective. This goes some way
towards explaining why the euro was not adopted earlier.
The evidence summarized in Table 26.4 is also in line with optimum currency area
theory. The table groups a total of 116 countries into four categories with different
exchange rate arrangements in the year 2000. As we move from left to right in the table
Table 26.4: Exchange rate arrangements and openness

Exchange rate arrangements in 2000
Peg Limited flexibility Managed floating Free floating
Number of countries 52 27 29 8
Average openness 1 52 41 38 24
1. Average of exports and imports as a percentage of GOP.
Source: Table 1 in Pietro Lovaand Jens S0ndergaard, 'When Should Monetary Policy Target the Exchange Rate?' . Mimeo. Georgetown
University, November 2003.
13. While the UK and Sweden have flexible exchange rat es and are therefore able t o pursue independent monetary
policies, it is less obvious what Denmark gains by staying outside the EMU, since the Danish krone is pegged to the
euro. Presumably the benefit for Denmark is that maintaining a national currency has an 'option value', since it
enables Denmark to adjust the exchange rat e in case a major asymmetric shock should occur, or in case Danish
policy preferences should start to deviate significantly from those of the euro area.
© The McGraw-Hill
Companies. 2005
818 PAR T 7: THE SHORT-RUN MODEL FOR THE OPEN ECONOMY
the degree of exchange rate llexibility increases. We see that countries with a higher ratio
of foreign trade to GDP appear to prefer less exchange rate flexibility. Ceteris pnribus, this is
what we would expect, since the net benents from exchange rate stability tend to be larger
the more a country depends on international trade.
26A §.l:l.J?.I?..(:lJY....................................................................................................................................................
1. This chapter discussed the economic factors which are relevant for a country's choice of
exchange rate regime. The various exchange rate regimes found in the world can be cate·
gorized into hard pegs, intermediate regimes, and floating. Under a hard peg the exchange
rate is fully fixed, and the national currency may even have been abandoned as in the case of
membership of a monetary union. Under floating the exchange rate is market-determined,
although the central bank may sometimes intervene in the foreign exchange market to
moderate exchange rate fluctuations. The intermediate exchange rate regimes fall between
the hard pegs and the floating regimes by allowing some exchange rate flexibility within a
(possibly shifting) band around the parity. In recent years there has been a worldwide
tendency for countries to move away from intermediate regimes towards a hard peg or
towards floating exchange rates.
2. Macroeconomic policy choices are subject to the macroeconomic trilemma. The trilemma
arises because a macroeconomic policy regime can include at most two out of the following
three policy goals: (i) free cross-border capital flows, (ii) a fixed exchange rate, and (iii) an
independent monetary policy. Under the classical gold standard before the First World War
most countries chose to sacrifice monetary autonomy. Under the Bretton Woods system of
fixed exchange rates between 1945 and 1971 most countries maintained restrictions on
international capital flows, but since the early 1970s the largest economies have abandoned
fixed exchange rates while allowing capital mobility and pursuing independent monetary
policies. More recently the majority of EU member states have adopted a common monetary
policy and a common currency as a means of ensuring irrevocably fixed exchange rates in a
common market with free capital mobility.
3. There is currently a widespread consensus that the ultimate goal of monetary policy should be
to maintain a low and stable rate of inflation. A country may adopt a fixed exchange rate as an
intermediate target for monetary policy with the purpose of achieving the ultimate goal of low
and stable inflation. However, under fixed exchange rates stable inflation requires not only a
stable inflation rate in the foreign anchor currency country; it also requ ires stability of the real
exchange rate. This may be hard to achieve if the domestic economy is often exposed to
asymmetric demand and supply shocks.
4. As an alternative to a fixed exchange rate, a country may adopt a flexible exchange rate with
an inflation target to serve as a nominal anchor. Under inflation targeting the central bank's
inflation forecast effectively becomes an intermediate target for monetary policy, since the
central bank can only affect inflation with a time lag of 1 years. If the inflation forecast !- 2
exceeds (falls short of) the target inflation rate when the interest rate is held constant, the
interest rate must be raised (lowered) until the inflation forecast corresponds to the target.
26 CHOICE OF EXCHANGE RATE REGIME AND THE THEORY OF OPTI MUM CURRENCY AREAS 819
5. Proponents of flexible exehange rates with inflation targeting argue that this policy regime will
make the deviations of actual inflation from the target rate of inflation as small as existing
economic knowledge permits. They also point out that a fi:<ed exchange rate regime is vulner-
able to speculative currency attacks. Proponents of fixed exchange rates argue that a floating
rate may be an independent source of shocks and that a stable exchange rate should be
seen as a goal in itself, since exchange rate uncertainty may hamper international trade and
investment.
6. The theory of optimum currency areas sees the choice of exchange rate regime as a trade-off
between the microeconomic benefits and the macroeconomic costs of a fixed exchange rate.
One microeconomic benefit is that a credibly fixed exchange rate reduces the riskiness of
foreign trade and investment. Further benefits are gained if exchange rate stability is achieved
by entering a currency union where the adoption of a common currency reduces international
transactions costs, improves market transparency and increases the liquidity of financial
markets. The macroeconomic costs arise from the fact that a fixed exchange rate/ common
currency excludes the possibility of an independent national monetary policy to stabilize the
domestic economy.
7. The microeconomic benefits of a fixed exchange rate/ common currency increase with the
degree of international economic integration whereas the macroeconomic costs decrease
with economic integration. When integration proceeds beyond a certain point, it therefore
becomes optimal to switch from a flexible exchange rate to a fixed rate/ common currency. It
has also been argued that even if joining a currency union is not optimal ex ante, it may
become optimal ex post because the adoption of a common currency will in itself promote
economic integration.
8. Optimum currency area (OCA) theory suggests that the macroeconomic costs of giving up
exchange rate flexibility within a group of trading partners will be relatively small if there is a
low frequency of asymmetric shocks, a high degree of labour mobility across countries, and
an international transfer mechanism securing a transfer of resources from countries hit by
positive shocks to those hit by negative shocks. OCA theory also implies that the micro-
economic benefits of a fixed exchange rate/common currency will be greater the greater the
volume of trade and investment across borders. The evidence indicates that those EU coun -
tries which have so far chosen to opt out of the Economic and Monetary Union do indeed ten d
to be more exposed to asymmetric shocks and to trade less with their EU partners than those
countries which have already joined the EMU.
26.5 Exercises
Exercise 1. Alternative monetary and exchange rate regimes
1. Explain the classification of alternative exchange rate regimes used by the International
Monetary Fund. Discuss briefly why there has been a polarization of the choice of exchange
rate regimes in recent years.
2. Explain the nature of the macroeconomic 'Trilemma'. Why is the trilemma inescapable, and
how have countries tended to deal with it over time?
© The McGraw-Hill
Companies. 2005
3. Explain what is meant by an intermediate target for monetary policy. What arc the desirable
characteristics of an intermediate target? Explain how the nominal exchange rate becomes an
intermediate target for monetary policy in a fixed exchange rate regime, and how (why) the
inflation forecast becomes an intermediate target for monetary policy under flexible exchange
rates with inflation targeting. Discuss the arguments for and against each of two inter·
mediate targets, assuming that the ultimate goal of monetary policy is a low and stable rate of
inflation.
Exercise 2. The classical gold standard
1. Explain the mechanics of the classical gold standard. In particular, explain the arbitrage
operations which prevent the exchange rate from falling below the level 'E- c in Fig. 26.2.
2. Explain the sense in which the gold standard could be said to be self-regulating.
3. Explain briefly some of the factors which led to the temporary breakdown of the gold standard
during the First World War and the more permanent breakdown during and after the Great
Depression.
Exercise 3. Symmetric versus asymmetric shocks and the need for (real)
exchange rate flexibility
1. Explain why a low and stable domestic inflation rate under fixed exchange rates requ ires a
stable real exchange rate as well as a low and stable inflation rate in the foreign anchor
country.
2. Explain the difference between symmetric and asymmetric shocks in open economies.
Illustrate the effect on the long·run equilibrium exchange rate of a permanent positive supply
shock when the shock is asymmetric and compare with the effect of a symmetric shock.
Explain why a fixed nominal exchange rate may be problematic in the scenario with an
asymmetric shock.
3. Discuss the potential role of fiscal policy in a country with fixed exchange rates exposed to
asymmetric shocks.
4. Assume that a group of countries with fixed but adjustable exchange rates and strong inter·
national trade links are exposed to a symmetric negative demand shock. What would happen
if all (or most) countries tried to escape from the resulting recession by devaluing their
currencies? What would be the most rational economic policy response from an international
perspective?
Exercise 4. Illustrating the Walters Critique
As we explained in Section 2, the British economist Allan Walters has argued that a fixed
exchange rate regime tends to generate macroeconomic instability by causing the domestic
real interest rate to rise during a recession, and vice versa. The purpose of this exercise is to
explore this so-called Walters Critique of fixed exchange rates. (You will gain more from this
exercise if you have already solved Exercise 6 in Chapter 24.)
The point made by Walters is most easily seen if we assume that all agents have static
expectations (setting q; = 0 in Eq. (4) in the text) so that ne = n _1 • Under fixed exchange rates
26 CHOICE OF EXCHANGE RATE REGIME AND THE THEO RY OF O PTI MUM CURRENCY AREAS 821
and perfect capital mobility the domestic real interest rate then becomes i-.n: ~ 1 = i1 -.n:. The
condition for goods market equilibrium may therefore be written as:
...----....
e'
y-y = f3 1(e: 1 + n
1
- n)-f3 2 (i 1 - :rr -r 1)+ z ~
AD: n =n +
1
(f3,~1,8Je~ 1 - ({3 1~ fJJy- ji- z), z "" -,Bir'- r') + z. (6)
where we have assumed that the foreign inflation rate is constant, n~ 1 = :rr 1, so that the ex post
1 1
foreign real interest rate can be written in the usual manner as r = i - n 1•
Using the definitions g iven in the appendix to Chapter 23, and assuming that trade is
initially balanced and that the marginal and the average propensities to import are identical,
one can show that:
q(1 - r)(1 - m)
,82 "" 1 -0 y + m '
where 17 is the change in private demand induced by a one percentage point change in the
real interest rate, measured relative to private d isposable income, m "" M 0 /"? is the initial ratio
of imports to GOP, r "" (? - 0)/ ? is the net tax burden on the private sector, and Oy is the
When we simulated our model with fixed exchange rates in Exercise 6 in Chapter 24, we
assumed a rather low value of 1J which implied ,8 2 < fJ 1 , to make sure that the AD curve had the
usual negative slope (note from (6) that the s lope of the AD curve is -1 /(fJ 1 -,8 2 )) . However,
empirical research g ives no reason to believe that ,8 2 should necessarily be smaller than ,8 1. In
this exercise we therefore assume that ,8 2 > fJ 1 to highlight the potential stability problem
pointed out by Walters. For concreteness, we w ill actually assume that fJ 2 = 2{3 1. This
particu lar ratio between the two parameters is by no means crucial for the Walters Critique,
but it w ill facilitate the graphical analysis.
For fJ z = 2{3 1, the AD curve (6) changes to:
AD(,8 2 = 2,8 1): :rr=:rr' -e: 1 +(;Jy-y-z). (7)
G iven the assumption of static expectations (n"=n _1), the economy's supply side may
be specified as:
SRAS: :rr=:rr-1+y(y- ji) +S, (8)
and the dynamics of the real exchange rate under fixed nominal exchange rates is given by the
usual identity:
(9)
Note from (7) that the AD curve now has a positive slope. In the analysis below you
should assume that 1/fJ 1 > y so that the AD curve is steeper than the SRAS curve. Notice also
that a real exchange rate depreciation (a rise in e~ now causes a downward shift in next
period's AD curve, and vice versa, in contrast to our stan dard model in Chapter 24.
1. Suppose the economy is in long -run equilibrium in period 0 and is then hit by a temporary
supply shock in period 1 wh ich lasts fo r only one period. Thus we have s 0 = 0, s 1 > 0, and
© The McGraw-Hill
Companies. 2005
s 1 = 0 fort;;> 2. Use a (y, n:) diagram to illustrate the effects of the supply shock on output and
inflation in period 1. Explain in economic terms why output rises in reaction to the cost shock
and why inflation increases by more than the increase in s.
2. Illustrate graphically how the temporary supply shock in period 1 will affect output and infla·
tion in periods 2 and 3. (Hint: be careful to use Eqs {7)-{g) to locate the exact positions of
the AD curve and the SRAS curve in each period. You should find that y 2 = y and y 3 < Ji).
Does it look as if the economy's long·run equilibrium is stable? Comment.
3. Suppose again that the economy is in long·run equilibrium in period 0 but is then hit by a
temporary positive demand shock which lasts only during period 1 (z 0 = 0, z 1 > 0, and z 1 = 0
for t;;. 2). Use a (y, .n) diagram to illustrate the effects of the temporary demand shock on
output and inflation in periods 1, 2 and 3. {Hint: again you should find that y 1 > y, y 2 = y and
y 3 < y). Do you see any reason why the oscillations in output and inflation should stop after
period 3?
4. If you have already solved Exercise 6 in Chapter 24, go back to the deterministic version of
your simulation model with fixed exchange rates and set cp = 0 and 1J = 1 while maintaining all
the other parameter values given in that exercise. Then simulate the effects of a temporary unit
demand shock (z 1 = 1) occurring only in period 1, using the deterministic version of the
model. You may also simulate the effects of a temporary unit supply shock. Do your simula·
lions confirm the Walters Critique? (Hint: is the economy stable?) How far must 17 be lowered
to ensure convergence to long·run equilibrium after a shock?
Exercise 5. Optimum currency area theory
1. Explain and discuss the nature of the microeconomic benefits from a fully fixed exchange
rate/ common currency and how these benefits depend on the degree of international
economic integration. What difference does it make whether the exchange is fixed or whether
national currencies are abandoned in favour of a common currency?
2. Explain and discuss the nature of the macroeconomic costs of a fully fixed exchange rate and
how these costs depend on the degree of economic integration.
3. Many economists have argued that even though a fixed exchange rate/common currency
among a group of countries may not seem optimal ex ante, it may nevertheless turn out to be
optimal ex post. Explain and discuss this argument.
4. Discuss the relevance of optimum currency area theory for (the debate on) monetary
integration in Europe.
Sorensen-Whitta-Jacobsen: I Back matter Appendix- Basic © The McGraw-Hill
Introducing Advanced regression analysis Companies. 2005
Macroeconomics
Appendix: Basic
·regression analysis
:T
:
his appendix gives a brief introduction to regression analysis as used in this book. It
r.iln in nn WilY snhstitntP. for il rP.ill (intrnclnr.tnry) r.m1rsP. in P.r.onnmP.trks, 1 hu t it
• should enable the reader to understand the estimations undertaken in this book.
Our appendix presupposes some basic knowledge of probability theory and statistics such
as empirical and theoretical distributions and their means and variances.
1 Model and data

Assume that according to some economic theory, ceteris paribus one economic variable x
affects (or explains or causes) another economic variable y according to the following
linear relationship:
(1)
where fJ 0 and fJ 1 are parameters. We call x the independent or explanatory variable andy
the dependent or explained variable.
For example, x could be household income andy could be household consumption, or
:r could be the length of an individual's education in years andy could be the individual's
hourly earnings.
The meaning ofthe ceteris paribus assumption above is that other variables believed to
affect y are held constant. In economics, where data rarely come from laboratory experi-
ments, it i~ most often 11ot possible to hold other variables constant. The economist may
then wish to include as many important explanatory variables as possible. This often leads
to an equation with several explanatory variables such as:
(2)
In our earnings and education example. an important additional explanatory variable

would be experience (years in the labour market).
1. For a real guide to econometrics we can refer the reader to Jeffrey tv. Wooldridge, Introductory Econometrics: A
Modern Approach, Thomson, South·Western, 2003, or to James H. Stock and Mark W . Watson, Introduction to
Economerrics. Addison Wesley. 2003.
823
I Back matter Appendix - Basic
regression analysis
© The McGraw-Hill
Companies. 2005
Macroeconomics
824 PAR T 7: THE S H ORT-R UN MODEL FOR TH E O PEN ECO NOM Y
The linear form of(l) or (2) may seem special, buty and/or x and z could, for instance,
be the logs of some underlying economic variables. For instance. in estimations of earn-
ings and education relations in labour economics, they in (2) is typically the log of h ourly
earnings, so {3 1 and {3 1 are semi-elasticities (giving the percentage increase in earnings
resulting from an additional year of education or experience) . If both the dependent and
the independent variables are measured in logarithms, the parameters are elasticities. The
linear form may therefore not be as restrictive as it ftrst appears.
We assume that the economist collects data in the form of coupled observations of x
(and z) andy, or of variables that come as close as practically possible to the theoretical
variables x. y and z. The data set is thus of the form (x ;· Y;) or (x ;• z;. Y;), where i is an index
for observation number running from 1 ton.
Such a data set can be a 'cross-section' where the index i runs across entities such as
persons or tlrms or countries. For instance. for each person i in a group of persons, the
economist observes average hourly earnings, education level and experience in some
specific year. Alternatively the data set can be a 'time series', where the index i goes across
time. so, for instance. the data set contains for each year i between 19 60 and 2000 aggre-
gate income and aggregate consumption of a specific country. The index i can go across
both time and entity. In that case the data set can be a 'pooled cross-section' reporting.
for instance, for two different years the average hourly earnings, education levels and expe-
riences of a number of people. but the people need not be the same in the two years, or it
may be a set of'panel' or 'longitudinal' data where, in the earnings and education example.
the people for whom earnings, etc.. are observed would be the same in the two years.
We assume throughout that for any collected data set not all observations of an
explanatory variable are identical. and in the case of several explanatory variables the
observations of x; and z; are not perfectly correlated, that is, they are not situated exactly
on a straight line.
A data set for the case of one explanatory variable is illustrated in Fig. A.l where
there are 2 7 observations (x ;· y;). Note that the points are not situated exactly on a
straight line. but rather as a cloud seemingly clustered around a line. In practice, data
always involve some 'disturbance' because in practice it is not possible to include literally
all factors aflecting the dependent variable.
If our model is strictly speaking (1) and our data set is as illustrated in Pig. A. l, the
model is immediately rejected: according to our model the points should be exactly on a
straight line. but they are not. Thus we have to moditjr our model for it to be consistent
with the data. We therefore assume that y is a random variable and that it is only the
conditional mean value of y that is given from x as {3 0 + {3 1 x. This amounts to saying that:
(3)
where 11 is a random 'disturbance' or 'error term' that can also be thought of as the
unobserved influence of omitted explanatory variables. For our exposition (which does
not aim at maximal generality) we will from the begilming impose the following strong
but classical conditions on the distribution of u: the distribution of the error term u is
independent of x and follows for any given x a normal distribution with mean zero and
variance <1 2 > 0. In standard notation:
11 is i.i.d., u - N(O. <J 1), (4)
where i.i.d. means 'independently and identically distributed'.
Macroeconomics
APPENDIX : BAS IC RE G RESS ION ANALYSIS 825
y
8
7 •
--=.
__:-----.
6 •
• • •
5
• ~ • ••
...
4
• •
•
3 .- ~.
•
2 •
•
0 X
0 3 4 5 6 7 8 g 10 11
Figure A.1: An example of a data set or sample (x;. y,) with 2 7 observations
In the case of multiple explanatory variables our model is:
y = f3o+f3lx+f32z + u, (5)
where u is now independent of :Y and z, and we combine again with (4) for the full model.
The name lor our model is the 'classical linear regression model'.
We assume that the data set (x 1, Yrl or (x;. z1• y;) is the outcome of a random sample
from the true 'population model' described by (3) or (5) and (4): somehow realizations X;
of x (and •r of z) occur. This could be deterministically in a laboratory experiment where
one can control x and z, but in most applications in economics the explanatory variables
are not controllable in which case observations such as x; and Z; are best viewed as real-
izations of random variables. 2 Each time an x1 (and z;) has been realized, a value Yr of y is
generated according to our model: it is picked randomly according to one of the distri-
butions N({J 0 + {3 1x r• 0' 2) or N({J 0 + {3 1 X;+ {3 2 z1• o 2) . Tllis is how the data set or sample,
(y r· x 1) or (Y;· x 1, z ;). i = 1, .... n. has been generated. The data set is thus to be considered a
stochastic variable that takes different contents with dill'e rent probabilities. and repeated
sampling would yield different data sets.
2 OLS estimators
The economist (econometrician) wants to use the data set to estimate the parameters {3 0 ,
{3 1 (and{J J of the model. Because of the randomness of the data set. one can only arrive at
2. These random variables are assumed to have finite (well-defined) means and variances. For the present purposes we
do not need to be more specific about their distributions than that
I Back matter Appendix- Basic
regression analysis
© The McGraw-Hill
Companies. 2005
Macroeconomics
the correct values by accident: dill'erent samples will, for any given estimation procedure,
yield different estimates. To distinguish estimates from true parameter values we use the
notation fJ 0 , /3 1, fJ 2 for our estimates. Given a specific sample, how should one arrive at a
reasonable value of. say, /3 1?
Consider the simple linear regression model with one explanatory variable. Given /3 0
and fJ 1 , the estimated value of y for a given observation x; of x would bey; = /3 0 + /3 1 x ;·The
observation ofy coupled with X; is !1;· so the residual (the difference between the observed
and the estimated value of y. given x ;) would be ii; !J; - {:J;. The so-called ordinary least=
squares (OLS) criterion says that fJ 0 and /3 1 should be set such as to minimize the sum of
squared residuals over the data sample:
=Iuf =I<Y; - /3 - P x;)

u u
where Q(/3 0 ,jJ 1) 0 1
2
. (6)
i =l i=l
By minimizing the sum of the squared residuals. rather than just the sum of the absolute
values of the residuals. one puts particular emphasis on avoiding large (as opposed to
small) residual~ . This makes intuitive sense. but there are also some more basic reasons for
adopting the OLS criterion. as we shall explain below. To understand this explanation. we
must lirst derive the implications of the or._c:; criterion. Note that the data sample is con-
sidered lixed here and the variables to be chosen are /3 0 and jJ 1• The first-order conditions.
()Q/o/J 0 = 0 and ()Q/o/J 1 = 0. lor this minimization are:
I" <Y; - f>o- /31 :x:;) = o.

i= l
(7)
L" X;(Y;- ,B(J -PIX;) = 0, (8)

i= l
and because of the quadratic form of Q, these conditions are also sullicient lor a (global)
minimum. Equation (7) can be 'vritten as:
LY; n/3 + P LX;.

,, ll
= 0 1
i= l i= l
Dividing here on both sides by nand letting averages be denoted by bars (x = ~ I;'= X; and
=
1
Y ~ Li'=1 Y;) givesy = f> 0 + /3 1xor:

Po = Y- Ptx. (9)
Inserting this expression for /3 0 in the second first-order condition, (8), gives:
L" x;(y; - ij + f3/S':. - {3 1 X;) = 0.

i= l
or
,, ,,
L x;(y; - y) = /3 L x;(x;- x) .
1
(10)
~I ~I
From this we can find /J 1 directly, but we prefer to 'vrite it slightly differently. Since
obviously I ;':/Y; - y) = 0 (show thL~). we have:
I =I I =L (x; - x)(.rr - .fJ),

II Jl Jl II
x;(y; - ij) x;(.LJ; - Ji) - x(.tJ; - Ji)

i= l i= l i= l i =l
Macroeconomics
and th erefore also in the special case where x; = y ;:
Ii=l x;(X; - x) =Ii=l (x;- :\Y

u ,
Inserting these into (10) gives:
/3 = I;'=l (x; - x)(y; - y) .

1 {11)
I;~ 1 (x; - x) 2
The formula in (11) is our estirnator (our general formula for estimation) of the slope {3 1 in
the simple linear regression model. An estimate is a particular value for the slope computed
from a specific data set. Clearly, an estimate of {3 0 follows immediately by inserting the
estimate 1~ 1 into (9) . Most often our main interest is in the slope estimate (a constant
measurement error on x will simply be included in the estimate of the intercept, but will
not affect the slope estimate).
Once one has fed the data set into a spreadsheet on a computer, it is easy to compute
the estimates jJ 1 and /30' Standard programs give the estimates and other relevant
information on the estimation at the press of a key. For the example in Fig. A. l, the OLS
estimates are jJ 0 = 2. 73 6 for the intercept and jJ 1 = 0. 342 for the slope.
If there are several explanatory variables as in {5), the first-order conditions for mini-
~~~
= 2
mizing the sum of squared residuals, Q(fJ 0 , {3 1' fJ 2) I;~ 1(y; - {3 0 - fJ 1 X; - fJ 2 z;) , are:
~ ~ ~
Ii=l (Y;- Po- /J 1

II
:t:; - /J2z;) = 0. (12)
Po - /3 1 X;- /3 2 z;) = 0,
II
I X;(Y; - (13)
i=l
- /3 - /3
II
I z;(y; 0 1 X; - /J 2 z;) = 0, etc. (14)

i=l
for even more explanatory variables. These systems of equations become more tedious to
solve the more explanatory variables there are, but simple computer routines can solve
such systems. So again, having typed in the data, the press of a key will give the estimates.
As i1 li1rth ~r st~p h~for~ ~xplilining thP. riltinnil iP. fnr th P. OT.S ~sti miltors, it is us~li1l to
show how one can get an indication of the goodness of llt of an estimation.
3 !..~.~. ~g?..4~.~.S..S...?.f..~~. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
Now that we have our estimates for /3 0 , jJ 1 and /3 2 • we can write down the OLS regression
line (or function):
y=/3 + /3 x+/3 z.
0 1 2
For the example in Fig. A.1 this 'line of best lit' is the line indicated in the figure:
y = 2.736 + 0.342x.
One can compute the observed values of residuals as ii; = Y; - Y;· where Y; = /3 0 + /J 1 X;+ /J 2 Z;,
lor each observation i. A property of our estimation is that I ;'=1 fl; = 0, as you may see
regression analysis
© The McGraw-Hill
Companies. 2005
Macroeconomics
directly from the llrst of the first order conditions above, (7) or (12). We can view each obser
vation Y; of the dependent variable as the sum of the predicted or fitted value, Y;· and the
residual. fi;, thatis, Y; = Y; + Ct;. Averaging on both sides gives: [I = y.so the sample average of
the dependent variable equals the average of the fitted values.
As a measure of the total variation of the dependent variable in our data set (the
variation we seek to explain) we may use the 'sum of squares, total':
II
SST= I (y;- D) 2 •
i= l
As a measure of the variation explained by our line of best f1t we may use the 'sum of
squares explained', defined as:
I =I
tt u
SSE= (fJ;- y)2 (fJ; - D)2,
i= l i =l
while letting the sum of squares of residuals measure the variation not explained:
II
SSR = Iul.
i= l
Each of these sums of squares is positive, and one can show that SST = SSE + SSR. As a
single numerical measure of how good the fit of the 'line of best fit' actually is, one often
uses the so-called R-squared or coellicient of determination. which measures the fraction
of the total variation of the dependent variable in the sample that is explained by the
explanatory variables according to the estimation:
, SSE SSR
R-=- = 1 - - .
SST SST
This R 1 is a number between 0 and 1. If all the observations Y; are exactly on the line of
best fit. y; = y; for all i, then R1 = 1. indicating the best possible fit. If, on the other hand, R2
is close to 0, very little of the variation in y is explained by x (while the residuals 'explain '
all the variation). Generally R 2 is considered an index of the goodness of fit. 3 The
R-squared is typically included in the output of standard computer programs. For the
example in Fig. A.1 we have R L = 0. 53 .
Many packages report an 'adjusted R-squared' as well. The R2 just defined necessarily
increases as more explanatory variables are added even ifthese do not help to increase the
explanatory power of the regression very much. To compensate for this tendency the
adjusted R-squared is defined as:
-, 11 - 1 J
R- = 1 - (1 - R-),
II - k- 1
where k is the number of explanatory variables. Increasing k makes R2 go up which tends

to increase IF, but the adjusted R-squared includes a 'cost' of adding new explanatory
variables. Evaluating the goodness of f1t by the adjusted R-squared rather than the
3. One can show that R 2 as defined is the square of the coefficient of correlation between Y; and y, which explains the
name R·squared and aQain indicat es that R 2 measures the Qoodness of fit.
Macroeconomics
Rsquared tends to favour parsimonious theories that include only few, important
explanatory variables. A value of 1 for JP still corresponds to a perfect Ht while a negative
value (which is now possible, for instance if R 1 = 0.2, rr = 21 and k = 5) corresponds to a
poor fit.
It is important to emphasize that an estimation that yields a low R 2 (or JP) may still
be very useful in the sense that it may give an accurate estimate. /3 1 say. If important
explanatory variables have been left out (perhaps because they are difficult to observe or
measure) , but the assumptions of the linear regression model nevertheless hold, then an
estimation based on a good (large and well dispersed) sample may give a sharp estimate of
the model's parameters and yet explain only a tiny fraction of the variation in the depen-
dent variable. However. our basic model assumptions, in particular that the mean of the
disturbance u is zero (and its variance constant) independentl!J of x, are less likely to be
fuliHled if important explanatory variables have been left out (try to explain why). This is
why we said above that the econometrician may wish to include as many important
explanatory variables as possible.
In conclusion, the econometrician would usually like to see a high R 1 , but should not
necessarily dispose of an estimation if the R 2 is low.
4 OLS estimators are BLUE

The OL.<:; estimators are themselves random variables since different data samples occur
with different probabilities and yield different estimates. Note that there are two sources
of randomness: the draw of the sample values of the independent variable(s), e.g.
=
(x 1. ... , x ,) x, and, given this draw, the randomness of the dependent variable due to the
randomness of the error terms.4 We now establish some properties of the OLS estimators
as random variables, with the purpose of explaining the rationale for the method of
Ordinary Least Squares.
Consider the slope estimator /3 1 of the simple regression model stated in (11). Since
I.;~ 1 (x; - x) = 0, we can write this as:
" (:~.:,- x)
I --2- vi. (15)
i =1 s,,.
=
where we have used the del1nition s~ I.;~ 1 (x;- x) 1 . [t appears from (15) th at /3 1 is a
linear combination of the observations !J; of the dependent variable given x. We say that
/3 1 is a linear estimator. From (9), /3 0 is also a linear estimator. It can be shown that the OLS
estimators for the model with multiple explanatory variables are linear as well. There
could be (and are) other linear estimators of f3 1 where the coefl'icient IV; on each!/; would
not be exactly (x;- x)/s~ as in (15). but a W ; depending in some other way on the sample
values X;.
4. If the sample values x 1 can be controlled, the randomness of samples comes solely from the error terms. As we
mentioned, most often in economics the sample values of the independent variables must be considered random
variables.
regression analysis
© The McGraw-Hill
Companies. 2005
Macroeconomics
Since /J 1 is a random variable we can ask: what is its mean value? According to our
model each observed y; can be expressed in terms of the observed :Y; as:
(16)
which is just a repetition of our model restricted to observed values of x. Note that the {3 0
and fj 1 occurring here are not estimates. but true values, of the parameters, and likewise
the II; are not the residuals z1; = Y;- [J; dellned above (that can be computed using the
estimates /J 0 and /J 1), but the true realizations of u. Inserting (16) into (15) gives:
In the numerator we may use that I.;~ 1(x;- x){3 0 = 0 and I.;'=1(A:; - x)x; =I.;'= 1(A:; - x) 2 as
well ass;= I.:=1(A:; - x) 2 to get:
(17)
Given the observations x ;• the estimator /J 1 is equal to the true {3 1 plus a linear combi-
nation of the true error terms u;· It is only because the error terms are typically not zero
that the estimate /J 1 does not deliver fj 1 exactly. However. since the eA:pected value of each
error term u; is (assumed to be) zero. taking expected values on both sides of (17) con-
ditional on x (that is. taking the sample values X; as given so they are not random) gives:
E(/J 1 I x) = {3 1• Since this holds for any given x, we also have without conditioning on x:
(18)
Likewise one can show that E(/J 0 ) = {3 0 . On average across a large number of samples the
estimators [J 0 and /J 1 will thus correspond to the true values of the parameters fj 0 and fj 1.
We therefore say that /3 0 and /3 1 are urr/Jiased estimators. To arrive at this result we did not
use all the distributional assumptions stated in (4), only that the mean of u given any xis
zero: E(u l x) = 0 . It can be shown that also with multiple explanatory variables the OLS
estimators are unbiased.
Next turn to the variance var( /3 1) of the slope estimator in the simple regression
model. We should emphasize that by var(/J 1) we mean the variance of /3 1 conditional on x
(we could have called it var(/J 1 Ix) to stress this fact). The term fj 1 in (17) is a non-random
constant, and given x so is each coefllcient (x;- x)/s~ on u ;·Since the u; are independent
and all have variance a 2 , it follows from the usual computational rules for variances that:
var(p
11) '
1 = L
" ((·
··; -·x~))z
-'·
-- - (1
2
=
"'" (•·
... i= l ··; - "")2
-'· ~2v
a 2 = 2, (19)
2 4
i= l sx s, s,
and the standard deviation of/J 1 is:
(20)
The two latter equations have an intuitive content: the larger a 2 is, the more spread out
around the true straight line the observations Y; will tend to be given the :Y;, and therefore
the less precise an estimate derived from the observations will be. On the other hand. the
Macroeconomics
APPENDIX : BAS IC REGRESSION ANALYSIS 831
greater the dispersion of the x 1 (the larger s ,. is). the more precise is the estimate Oust draw
a figure where the x 1 are very close to each other and see how easily a lltting line can be
twisted) . This warns us, in so far as we have any inlluence on the sample values:\: ;• to try
to have these 'well spread out' .
With more explanatory variables one can again find the variances of the OLS slope esti-
mators given the sample values of the independent variables. These look a bit more compli-
cated because they have to take account of any correlation between the independent
variables. However, independently of the number of explanatory variables the following
important result about OLS estimators known as the Gauss-Markov theorem holds:
Among all linear and unbiased estimators, tile OLS estilnator has the s1nal/est (uncondi-
tional) variance. ln other words, tile OLS estirnators are Best Linear Unbiased Estimators, or
BLUE. This is an important motivation for the criterion of 'least squares'. We give here a
nice and simple proof of the Gauss-Markov theorem for the slope estimator of the simple
linear regression model.
Consider an arbitrary linear estimator of {3 1: /3 1 =I ;'= 1 w1y 1• where the weights w 1
depend on x. As above, we may rewrite the expression for fJ 1 using Y; = {3 0 + {3 1 x 1 + u1 to
gP.t: ft 1 =fi 0 I ;'=• w 1 +/{ 1 I;~ . w 1x 1 +I;'= • 1.111 u1. 'T'Hking P.XpP.c:tP.n vHinP.s on hoth sin P.s c:on-
ditional on x, remembering that E(u ;) = 0, gives: E(P 1 I x) = {3 0 I ;'=1 w 1 + {3 1 I;~ 1 W;X 1.
p
Insisting that E( 1 I x) = {3 1 for any values of the true parameters {3 0 and f3 l' that is,
insisting on (conditional) unbiasedness, requires that for any x:
ll ll
I
i =l
w; = o and I
i=l
W ; X; = 1. (21)
Conditional on x, the variance of fJ 1 is: var(/3 1) = a 2 I ;'= 1 wf. Hence, the linear and
unbiased estimator that gives the smallest variance (given x) is one that minimizes
I ;'= 1 wf subject to the two restrictions in (21). The Lagrangian for th is problem is:
La= I-;~ 1 wf - . 1. 1 I ;'=1 w1 - A.z{I;'.. 1 w 1x1 - 1), where A 1 and A2 are the Lagrange multi-
pliers. The first-order conditions resulting from di!Jerentiating '>\lith respect to .1 1 and .1 2
and the w1 are the two equations in (21) repeated and: 2w; - A1 - A2 X; = 0, i = 1, .... n,
which can be rewritten as:
.1 1 .1 2
\\1· = - + - X·. (22)
l 2 2 l
Inserting this expression lor w 1 into each of the equations in (21) gives:
and (23)
From the first of these, A.J2 = - (A 2/2)I;'=1 x;/n = - (A 2/2)x. Inserting this expression for
. 1. 1/2 into the second equation in (23) gives:
A ("
-..3.
2 i= l
I xf - xI" x )
i= L
1 = 1.
A2 = 1
2 I:~ 1('-:,- x) 2 ·
regression analysis
© The McGraw-Hill
Companies. 2005
Macroeconomics
Inserting this expression for ?.. 2/2 into the expression for ?.. 1/2 we found above (?.. 1/2 -
- (?..z/2)x) gives:
Finally, inserting the expressions for}.. 1/2 as well as ?c 2/2 into (22) gives:
and this is exactly the value of W; in the 04<:) estimator (just compare to (15)) . Hence.
conditional on x, the OLS estimator S1 has smallest variance among all linear, unbiased
estimators. Since this holds for any given x, the variance of the OLS estimator is also
smallest unconditionally. This finishes our proof.
Note that we have still not used all of our assumptions on the distribution of u. The
additional assumption used above is that the variance of u is a 1 independently of x.
whereas normality of the distribution of 11 has still not been used. It will be now.
5 Inference
We know that the mean of the slope estimator, /3 1 , of the simple regression model is P1,
and the variance (given x) is a1/s; (the standard deviation is o/s,) . Furthermore, the
expression in (17) shows that S1 is the sum of a fixed term (.8 1) and a linear combination
of the error terms 11 ;· where the weights are llxed given x. This means that given x, the
estimator S1 is normally distributed (applying for the first time that the 11; are normally
distributed): /3 1 - N((3 1 , a 2/s;). Consequently, the transformation defined by:
/3, - P~ P1- !31

v = --- = --.- (24)
(1/s, std(p 1)
has the standard normal distribution: v- N(O. 1). Thls is potentially useful lor statistical
inference. For instance, let v 0 .975 be the 97.5th percentile for the standardized normal
distribution, that is. the particular number such that according to N(O, 1) the probability,
Pr(v < v 0 . 975 ), of drawing a value of v below v 0 .975 is 9 7. 5 per cent. This percentile is
known to be 1.960. Since N(O, 1) is symmetric around zero, the probability of a random
draw ending between -v0 . 975 and v 0 . 975 is 95 per cent, that is,
Pr(-v0 9 75 < v < v0 .975) = 0.95 .

Since v- N(O, 1), we have:
This says that there is a 9 5 per cent probability that the true value of {3 1 lies between
S1 - v 0 . 975 (f/s, and /3 1 + v 0 .975 o/s,, and the interval reaching from the lower to the upper
Macroeconomics
of these values is called the 9 5 per cent co11.fidence interval for f3 1 . Likewise. substituting the
99.5th percentile v 0 .99 5 (known to be 2.576) for v 0 . 9 75 gives the 99 per cent conlldence
interval for j3 1• Hence. given the estimate /3 1 and the standard deviation a/s,, we can
identify a confidence interval so that with probability 95 per cent (or 99 per cent) the true
value ofj3 1 lies in the interval, given the assumptions of our model. Such a statement is, of
course. very useful, in particular if the confidence interval is narrow. However, in practice
we almost never know the tme variance a 2 , so we cannot compute the above conlldence
intervals. We must estimate c1 2 from the data to do any statistical inference analysis.
The error terms u; = Y;- f3 0 - j3 1 X; cannot be observed because the true parameter
values j3 0 and .B 1 are unknown, but the residuals u; = y; - (J 0 - /3 1 X; can be computed and
are estimates of the u;. Since cr 2 = var(u) = E[u 2), one might think that the estimator lor o 2
should be z:;~ 1 uJ/n. but because of the intricate matter of 'degrees of freedom', which is
beyond the scope of this appendix. the estimatora 2 that ensures unbiased ness (E[a 2) = a 1 )
is:
{25)
For a given sample the estimated variance a2 , and hence the estimated standard deviation
a, can be computed. The latter is reported by standard computer programs. for instance
under the name of 'standard error of regression '. For our example, a=0.989. Since
std(/3 1) = cJ/s,.. the natural estimator for std((J 1) is:
This can be computed, which is always done by standard packages, and it is reported as
the 'standard error of the estimate /3 1' . (Sometimes it is less precisely referred to as the
'standard deviation of /3 1 ' although strictly speaking it is an estimate of the standard
deviation). In our example se( /3 1) = 0.064. With more than one explanatory variable
there are similar formulae for the standard errors of the OLS estimates. In the regressions
of this book the computed standard errors are widely reported and used lor the purposes
described in the following.
In view of what we did above it is now natural to dellne the transformed variable:
{26)
which is almost the same as v. but with the estimate ;; substituted for cJ (or the standard
error se(P1) substituted lor the standard deviation std(P1)) . The variable t is not normally
distributed (note from (2 5) that theY; enter into rr. as they do into {3 1, so even given x. the
variable t is not just a normally distributed variable times a constant plus a constant).
However, t is clearly a random variable since it depends on the random realizations of y.
Given the properties of the random variable y ;implied by our modelling assumptions, and
given x, the variable t defined by (2 6) bas been shm<\111 to follow the so-called t-distribution
with 11 - 2 degrees of freedom . The t-distribution is 'spread out' compared to N{O, 1), and
more so the fewer degrees of freedom there are. but it approaches N(O. 1) as the number of
observations n goes to infinity. For any given number of degrees of freedom, the t-distrib-
ution is fixed and knmvn and symmetric around zero. In particular its different percentiles
41 1
regression analysis
© The McGraw-Hill
Companies. 2005
Macroeconomics
are known and can be found in statistical tables. If the 9 7 . 5th percentile of the t distribu
tion with rt - 2 degrees of freedom is t 0 . 97 5 , then analogously to the above:
Hence. the 9 5 per cent confidence interval for fJ 1 goes from /3 1 - t 0 975 se( /3 1) to /3 1 +
t 0 9 75 se(/J 1) . and likewise the 99 per cent confidence interval goes from /3 1 - t 0 .995 se(/3 d to
/3 1 + t 0 99 5 se( /3 Jl. These intervals can be computed since they depend on {; (or se( /3 1) )
rather than on o (or std(/3 Jl) . One or both of these intervals are usually included in the
output of standard regression packages along with the estimated coefficients and their
standard errors, etc. In our example the 95 per cent confidence interval for (the estimate
/3 1 = 0.342 of) {J 1 is [0.209: 0.474) . while the 99 per cent confidence interval is
[0.163 : 0.521].
If you look into an old-fashioned statistical table listing percentiles for the t-distri-
bution as depending on degrees of freedom, you will see that already lor 25 degrees of
frec::J om (correspouuiug lo 2 7 observaliou s iu l!Jc:: simple:: regre~;~;iou moue!). £0 .975 is dose lo
a.vo (2.060), lor 50 degrees of freedom it is very close to two (2.009) and lor 60 degrees of
freedom it goes below two, from there falling further towards its boundary value of 1.9 60.
As long as we have a reasonable number of observations, we can use the mle of thumb
that the 9 5 per cent confidence interval is the estimate plus or minus two of its standard
errors. In our example. where /3 1 = 0. 342 and se(/3 1) = 0.064. you will see that the 9 5 per
cent confidence interval stated above is pretty close to being given this way.
If there is more than one explanatory variable, the 9 5 per cent confidence interval for
each estimate ,B; still goes from /3 ;- t 0 975 se(/3;) to /3; + t 0 . 975 se(,8 ;). just with the modifica-
tion that the number of degrees of freedom of the relevant t-distribution is the number of
observations rt minus the number of explanatory variables k minus one. that is. rt - k - 1.
specializing to n - 2 for k = 1. Hence. as long as the number of observations exceeds the
number of explanatory variables to a reasonable degree. t 0 .q75 will still be close to two, so
the rule of thumb just described still holds.
The standard errors se( /3 ;) and the associated conlldence intervals are useful since
they provide us with interval estimates of the parameters. Standard deviations and conll-
dence intervals are particularly useful lor hypothesis testing.
Assume that according to our theory, explanatory variable number j affects the
dependent variable and that we have arrived at an estimate /3; > 0, say. We may be in
doubt w hether {3 i is really zero (explanatory variable j does not aflect y after all) and we
*
just arrived at a /3; 0 by accident. We would therefore like to test the hypothesis th at
variable j has no influence on the dependent variable, fJ; = 0, against the 'two-sided' alter-
*
native that variable j h as some (positive or negative) influence. /3 i 0. Now, if zero is not in
the 9 5 per cent confidence interval, then there is a less than 5 per cent probability that
fJ; = 0 at the same time as we drew the sample leading to /3 i• se(/3 ;) and the associated con-
*
fidence interval. We therefore reject the hypothesis {3 i = 0 (against /3; 0) at the 5 per cent
significance level. Likewise. if zero is not in the 99 per cent confidence interval. we reject
that {3 i = 0 (in favour of{3; * 0) at the 1 per cent significance level. Of course, the 1 per cent
level gives a stronger rejection than the 5 per cent level. In our example. we must reject
fJ 1 = 0 both at the 5 and at the 1 per cent level.
Macroeconomics
We may have theoretical reason to believe in a particular value bi for j3 i' and would
*
like to test the hypothesis f3 i = bi against f3 i IJ i' where this time we may want not to have
the hypothesis rejected. If bi is not in the 9 5 per cent interval, then we must reject fJ i = bi (in
*
favour of {J i bi) at the 5 per cent level, while if bi is in the 9 5 per cent interval then we
*
cannot reject f3 i = hi (against fJ i bi) at the 5 per cent level.
We are often interested in testing the idea that .B i > 0 (or f3 i < 0) since our theory most
often gives an indication of the sign of an influence, so the 'null hypothesis' f3 i = 0 should
be tested against the 'one-sided' alternative f3 i > 0 (or perhaps .B i < 0). If.B i = 0 is true. then
(from the analogue of (26)) the random variable
t= fii~ (27)
se(f3i)
should follow a t-distribution with n - k - 1 degrees of freedom. We can compute the
actual t-value following from (27) and our estimates fi i and se(fii). Note that this is just the
number of standard errors that the estimate fJ i lies away from zero, which is exactly the
't-value' reported by standard packages along with the parameter estimates etc. In our
example, the t-value is 5.323. Let the 95th percentile of the t-distribution with n - k - 1
degrees of freedom be t 0. ~ 5 . If the hypothesis f3 i = 0 is true. then the probability that the
t-value exceeds t 1 ).~ 5 is only 5 per cent, so if t > t 0 . 95 we reject f3 i = 0 against f3i > 0 at the
5 per cent significance level. From a table one can see that even with lew degrees of
freedom . t 0.~ 5 is below two and t 0 .99 is below three. Therefore, as rules of thumb, when the
number of observations exceeds the number of explanatory variables just moderately, if
the t-value is above two (below minus two) we can reject f3 i = 0 against f3 i > 0 (f3 i < 0) at
least at the 5 per cent level. while if the t-value L~ above three (below minus three) we can
reject .Bi = 0 against{Ji > 0 (j3 i < 0) at least at the 1 per cent level. In our example therefore,
f3 i = 0 is rejected against f3 i > 0 at a significance level even stronger than the 1 per cent
level.
Along with the t-value of an estimate fJ i' standard software packages for regression
analysis usually report a P-value which is the probability mass in the tails of the relevant
t-distribution: if {Ji and (hence) the t-value are positive. the P-value is the probability
according to the t-distribution with 11 - k - 1 degrees of freedom th at a random draw of t
ends above the t-value or below minus the t-value, that is, it is the significance level at
which one will just (or just not) rejectfi i = 0 against/Ji * 0 (and similarly for fi i < 0). In our
example, the P-value lor .8i is 1. 62 · lo-S. so not to reject fJ i = 0 would require an extreme
signillcance level of at most 0 .00162 per cent.
You have probably realized that the estimated standard errors se(/J i) (along with the
parameter estimates themselves) gives us the information needed lor inference analysis.
This is why we locus so much on standard errors in the estimations in this book.
6 A final warning concerning causality

................................................................................................................................................................................
We were very explicit above saying that the economist's theory implies that x influences
y. not the other way round. Very often, however, the economist will be in doubt. For
instance, if x is national income and y is national consumption (and our data set is a time
41 1
regression analysis
© The McGraw-Hill
Companies. 2005
Macroeconomics
series). there may be reason to believe that:>.~ allects y because when people earn more they
can spend more, but also that y allects x because national consumption is a large part of
aggregate demand which affects aggregate income according to short-run macro-
economic theory. From the econometrics presented in this appendix one cannot draw
conclusions about the direction of causality. Finding a well-11tting and significant regres-
sion line. y = iJ 0 + S1 x, by the techniques we h ave explained does not exclude that it is
really y that allects x or that there are causal relationships in both directions. If. on the
other hand, x is household income andy is household consumption (and our data set is a
cross-section, say). there may not be a similar problem of 'reversed causality', since we
may have no reason to believe that high family spending should cause high family
income, but may have reason to believe that high family income results in high family
spending.
Even ifwedo have reason to believe that x affects y (according to a linear relationship)
and have no reason to believe that y affects x. we can still not be sure that a nice (signi-
flcant. etc.) result from an OLS regression of yon x indicates that x really causes y. This is
because of the possibility of 'spurious correlation': it may be that a third and omitted vari-
able really explains both "~ and y . say. allecting both positively. so a close relationship
between "~ and y may not indicate any causality. The earning and education example is
illustrative here. There may be little reason to believe that earnings allects education level
and good a priori reason to believe that education level aflects earnings. Nevertheless,
finding a tight relationship between education and earnings (perhaps including other
explanatory variables such as experience) could simply be an indication that 'innate capa-
bility' (a variable th at it is dilllcult to measure and include) explains both education level
and earnings, affecting both of these positively, without education itself having any
impact on earnings.
There are econometric techniques to test lor causality (or endogeneity), but they are
too advanced for this appendix.

MACROECON 1n

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

MACROECON 1n

Încărcat de

Drepturi de autor:

Formate disponibile

I Sorensen-Whitta-Jacobsen:

The Short Run

The economy in the

Wh y understan ding business cycles is in1portant

398 INTRODUC ING ADVANCE D MACROECONOM ICS

log of GOP United Kingdom log of GOP USA

log of GOP Denmark log of GOP Finland

14 TH E EC ONOMY IN TH E SHORT RUN 399

On several occasions in history, recessions have developed into severe economic

Figure 14.2: Unemployment and political extremism in Germany, 1924-1 933

400 IN TRODUCING ADVANCED MACROECONOMICS

14.2 Defining business cycles

This definition of business cycles emphasizes several points.

• Aggregate economic activity: Business cycles are characterized by a co-movement of a

14 TH E EC ONOMY IN TH E SHORT RUN 401

• Expansions and contractions: Business cycles are characterized by periods ofexpansion of

j 4.3 .I:>..CI.~i.~~--~-l:l.S..i~~S.S..9'<::~~S. ............................................................................................................. .

402 IN TRODUC ING ADVANCED MACROECONOMICS

Table 14.1: US business cycle expansions and contractions

Trough Peak Trough 1. Expansion 2. Contraction Cycle 1 1./2.

December 1854 June 1857 December 1858 30 18 48 1.67

Source : 1\ational Bureau of Economic Research.

14 TH E EC ONOMY IN TH E SHORT RUN 403

14.4 .~.~.~~~~.~.~.~. ~~.~~~~~~. ~Y..~~·~·~································ ..·································································

404 INTRODUC ING A DVANCE D M ACROECONOM ICS

14 THE ECONOMY IN THE SHORT RUN 405

1. A trough must be followed by a peak, and a peak must be followed by a trough.

406 IN TRODUCING ADVANCED MACROECONOMICS

log of GOP United Kingdom log of GOP USA

log of GOP Denmark log of GOP Finland

log of GOP Belgium log of GOP Netherlands

Figure 14.3: Real GDP and its underlying trend

14 THE ECONOMY IN THE SHORT RUN 407

408 INTRODUCING ADVANCED MACROECONOMICS

1983 1986 1989 1992 1995 1998

14 TH E EC ONOMY IN TH E SHORT RUN 409

14.5 yy~~.~. ~.~P..P.~.~.~. ~~.~~~&..~.~.~.~~.~~.~y~~~~................................................................

410 IN TRODUCING ADVANCED MACROECONOMICS

GDP 100 1.51 1.00

Sources: UK Office for Naticnal Statistics, Ecowin and Oatastrearn.

Average share Absolute standard Relative standard

GDP 100 1.66 1.00

14 THE ECONOMY IN THE SHORT RUN 411

Table 14.2b: Macroeconomic volatility in Denmark and Finland

GOP 100 1.6 7 1.00

Sources: Danmarks Nationalbank and Ecowin.

Average share Absolute standard Relative standard

GOP 100 2.29 1.00

Sources: S tatistics Finland and Bank of Finland.

Correlation, leads and lags

412 IN TRODUCING ADVANCED MACROECONOMICS

Table 14.2c: Macroeconomic volatility in Belgium and the Netherlands

GDP 100 1.06 1.00

Sources: Belgostat, Banque Nationale de Belg ique and Ecowin.

Average share Absolute standard Relative standard

GDP 100 1.22 1.00

Sources: Statistics Netherlands, De Nederfandse Bank and Ecowin.

14 TH E EC ONOMY IN TH E SHORT RUN 413

414 INTRODUCING ADVANCED MACROECONOMICS

Sources: See Table 14.2.

Coefficient of correlation between GOP and Xr

Sources: See Table 14.2.

14 THE ECONOMY IN THE SHORT RUN 415

j 4.3 .I:>..CI.~i.~~--~-l:l.S..iS.S..9'<::S. ............................................................................................................. .

14.4 .~.~.~~~~.~.~.~. .~~~~. ~Y..~~·~·~································ ..·································································

14.5 yy~~.~. ~.~P..P.~.~.~. .~&..~.~.~...~y~~~~................................................................

14.8 §. .Y.. ..................................................................................................................................................

15.2 T..~.~. P..r.~. ?..f..S..t.?.~ ............................................................................................................................ .