Sunteți pe pagina 1din 125

Economics of Underdeveloped Countries

Contents
Contents..................................................................................................................... 1
(A) Poverty Traps, Incentives and Development & (B) Quantitative Development
Economics and Kanbur Lectures................................................................................. 2
Lecture 1A - Overview: The Macro History of Development....................................2
Lecture 1B Institutions and Economic Growth......................................................7
Lecture 2A Poverty and Poverty measurement...................................................12
Lecture 2B Estimating an Absolute Poverty Line................................................22
Lecture 3A Poverty Traps.................................................................................... 25
Lecture 3B Testing for Efficiency of Rural Markets..............................................33
Lecture 4A Risk and Insurance............................................................................38
Lecture 4B Testing for Risk Sharing in Poor Communities..................................43
Lecture 5A Investment in Human Capital...........................................................47
Lecture 5B Evaluating Externalities in Education................................................51
Lecture 6A Physical Capital: Property Rights, Incentives and Growth.................56
Lecture 6B Property rights and Investment Decisions........................................61
Lecture 7A Political Economy..............................................................................65
Lecture 7B Testing the Neo-Con Agenda: Democracy in Resource-Rich
Economies............................................................................................................. 68
Lecture 8A Governance...................................................................................... 73
Lecture 8B A Free Press is Bad News for Corruption...........................................76
Econometrics Summary Guide.............................................................................. 79
Demography and Development................................................................................81
Lecture 1 Why do economists want to explain demographic behaviour?
Transition theory and the Malthusian model.........................................................81
Lecture 2 The Household Demand Approach and its Applications......................88
Lecture 3 Externalities to Childbearing and Social Norms: Demand for Children
vs. Supply of Family Planning................................................................................93
Lecture 4 Social Interactions and Fertility Transactions......................................98
Lecture 5 Mortality and Missing Women...........................................................103
Lecture 6 Current Policy Debates about Demography in Poor Countries..........109
Financial Liberalisation........................................................................................... 113
Part IIB Revision, Stephen Ward 1 Development
(A) Poverty Traps, Incentives and Development & (B)
Quantitative Development Economics and Kanbur
Lectures
Lecture 1A - Overview: The Macro History of Development
Large variation in wealth of nations and living standards
o Why some countries rich and others poor?
o What can be done about it?
o How is world dist. of income expected to change in future?
Countries at top of income dist. more than 30x richer than those at bottom
o 2000 US GDP/cap = $33,000
o Mexico < $9,000, China < $4000, India < $2500, Nigeria < $700
Penn World Tabled (PWT) all numbers at 1996 US$, PPP adjusted to allow for differences in
relative prices of different goods across countries
o At RER, divide all numbers by approx. 4 absolute difference far greater in RER than
PPP
Economic Development: understand micro-processes keeping poor indiv. and nations in poverty
Economic Growth: understand macro process of societies and regions growing and structurally
transforming their economies
Essentially ask the same question
1
st
conceptual revolution
Solow Model: growth driven by
o Capital accumulation
o Exogenous technological progress only factor that actually changes growth rate
o Human capital (in augmented Solow model) increases per-cap income but not growth
rate
Technology not exogenous
o Arrow (1960s): tech. and knowledge fundamentally different from other goods
o Romer, Aghion-Howitt, Grossman-Helpman: models of endogenous, sustained growth in
the LR because of innovation and invention
o Incentives to innovate: technologies endogenous because of diffusion and adoption of
technologies
Technological differences help understand wealth and poverty of nations
o Poor nations poor as
Dont use best available technologies
Dont use technologies and factors efficiently
o Traditional perspective societies poor as
Inefficient contractual arrangements
Dysfunctional credit mkts.
Lack of I in new tech.
Poor health and human K
2
nd
conceptual revolution
Look for fundamental causes of differences in
o Human & physical K
o Tech.
Part IIB Revision, Stephen Ward 2 Development
Mkts. dont work
o Mkt. structure
Geography view Sachs: importance of ecology, climate, disease environment, geography i.e.
factors outside human control. E.g. landlocked countries cannot trade easily, high transaction
costs outside control
Institutions view importance of man-made factors, especially organisation of society that
provides incentives to individuals and firms. Dont necessarily work in these countries e.g.
state of accounting ind., legal structures, incentive structures etc.
Historys accidents some countries unlucky and trapped in underdevelopment
Most plausible fundamental cause is institutions
Change in econ. success in many instances (reversal among former European colonies, growth
takeovers in South Asia India in 80s) evidence against both geography and historical accidents
views
Social organisations and institutions likely to be important for process of development
o Need more micro rather than macro catch all2 emphasis on institutions and need
theory of institutions. Need by context or group of countries
Sources of incentives
o Determined by institutions rules of game of society
o For businesses, securing property rights is most important
No I if businesses expect expropriation, high taxes or inefficient regulation
Institutions must make credible commitment to reward I
o Threats
Theft by others (often govt.)
Inefficient intervention or taxation by govt.
Solow
o Assume no mkt. failures (no externalities, constant r.t.s., no transaction costs,
complete info
o Suppose all output in economy can be aggregated. Aggregate prod. function represents
abil. To prod. aggregate good (GNP) as function of aggregate amounts of available
inputs
( ) ( ) [ ] ( ) ( ) [ ]

1
t L t A t K t Y
A(t) = labour augmenting tech. progress. A(t)L(t) = no. effective workers in LF
Let k = K/AL, y = Y/AL
o Assume L grows at exogenous rate n, A grows at exogenous rate g
( ) ( )
nt
e L t L 0 ( ) ( )
gt
e A t A 0
n
L
L

g
A
A

o Fixed proportion s of GNP put into K accumulation. Some K depreciates so (1 d)K of


previous K stock remains
dK sY K '
o Exogenous growth rate of effective labour is:
( ) ( ) [ ]
( ) ( ) [ ]
g n
t L t A dt
t L t A d
+
1
o As K accumulated through saving, K:L ratio changes through time. Differentiating k =
K/AL w.r.t time
L
L
A
A
K
K
k
k


Part IIB Revision, Stephen Ward 3 Development
o Combining equations and rearranging gives Solows fundamental equation of economic
growth, giving an indication of how process of K accumulation changes factor
proportions over time
( ) ( ) ( ) ( ) t k d g n t k sf k + + ' where f(k(t)) = k(t)

o If k>0, economy becoming K richer when k<0, becoming K scarcer. Steady state k
*
occurs where k = 0
o Richer countries have relatively more K-intensive production
o All countries reach the same steady state if they have the same n, g, d & s
o Assumption of concavity gives result in all countries K and income per effective worker
converge to unique steady state
For prod. functions not globally concave in accumulating factors, possibility of
multiple steady states. Could result in a high steady state and a zero steady
state. Extreme but shows how much increasing returns can change results
o Similar effects from supposing pop. Growth a function of income per cap. E.g. let
growth = n(k)
Standard, though controversial demographic views suggest there is rapid pop.
Growth at low income levels followed by demographic transition after which
income fertil. Declines
Can lead to possibility of poverty trap induced by high fertil. Growth
Again find threshold K:L ratio k
*
above which income per cap grows to income
level of khigh. Below threshold level, economy remains at low income level of
klow, even though has same tech. level as economy at khigh.
Suggests shocks that place economy on wrong side of threshold can have
persistent effects. Contrary to Solow, shocks can have LR effects
Wealth of Nations
Huge growth in income per capita in many parts of world over past 200yrs
o US 1990 dollars 1820 = $1,200, today = $30,000
o Av. income in W. Europe from $2,000 to $18,000
Changes not uniform large disparities across countries
o Africa av. income around $1,300 (stayed pretty flat) and much less in some countries.
Around 550 in Tanzania, Sierra Leone, Niger and less in Zaire
o Smaller but still substantial gap between US and smaller middle income countries like
Portugal
Due to
o Physical K
o Human K
o Efficiency Solow Residual
o Growth accompanied with more physical & human K, and greater efficiency
o Poor countries have less physical K, human K and lower efficiency
o With physical K, quantity explains only small part. Does matter but major factor is
efficiency (TFP)
Human K increases prod. of workers and possible creates benefit for other workers and firms
(exchange of ideas, innovations, complementarities etc.)
Part IIB Revision, Stephen Ward 4 Development
k
k
*
khigh
I per unit
effective
lab.
klow
Actual I: sf(k)
Break even I: (n+g+d)k
o Approx. 6-10% per 1 yr schooling for on effects
o Approx. 0-2% direct externalities on society
o Substantial but only sufficient to be part of major proximate channel for cross-country
differences
Why some countries have better tech. and more human K?
o Provide better incentives to those who invest in new tech. and human capabilities
o Select right type of indiv. And firms to become key investors and players
o Institutional variation big differences in institutions across countries
Democracy vs. dictatorship
Enforcement of property rights
Legal systems and contracting institutions (how easy to enforce contracts in
Turkey, about 10x as expensive as in US, but relatively fast)
Entry barriers
Causation:
o Correlation doesnt mean causation
Institutions endogenous, partially caused by income and determined by other
factors simultaneously influencing income
o Institutions important for LR econ. development?
Yes but must use IV estimation
Exploit sources of variation in institutions not directly related to
income e.g. features of history e.g. natural experiment of colonisation
(Acemoglu, Johnson, Robinson)
Strategies typically show v. important effects of institutional features
on LR growth, I and efficiency
Conclusion
Proximate causes of econ. growth: human K, physical K and tech.
o Human K special role in enabling tech. progress
Fundamental causes:
o Incentives (via institutions)
o Selection (via incentives)
Mankiw, Romer and Weil (1992) share of physical K about 1/3 empirically
Strong assumption assume countries in steady state
o Assumes + g constant across countries can drop parameters from equations
o Assumes prod. function CRS
Matters if human K introduced into regression
o If believe human K has effect on dependent variable. If believe has effect on n or
saving rate, get omitted variable bias
o Preferred saving rate of human K than levels of human K lead to less problems
o When include human K what happens to adjusted R
2
and when is added variable
significant?
Including human K how affect ? i.e. share of K
Calibration plug in = 1/3 and increase income by 1% which increases savings by 20%
o Implies sK should be 20x more in rich than poor countries, but this not the case
Solow rate of convergence - ( ) ( ) + + g n 1
o = 1/3, n = 1%, g = 2%, = 3% (from other studies)
o gives = 4% but know from studies this is 2%
Including human K, problems solved or at least improved coefficient C (effect of savings rate)
becomes 1, not i.e. doubles
Omitting Human K gives > 1/3, not good estimate of real world
o When include Human K, & = 1/3 and R
2
improves
Part IIB Revision, Stephen Ward 5 Development
Also effect on from different school measures
o Which give more realistic ?
Adjusted R
2
and significance of coefficients
Ambiguous results not clear which measure better
o Measure of human K important different measures give different results
o Difficulties in measuring human K
Education? Primary or secondary? Both? All subjects?
Health?
Combo?
o Klenov and Rudriguez-Clare (1997)
MRW results very sensitive to human K measure primary enrolment not vary
greatly between countries but secondary does. When include primary, effect
far smaller than when consider just secondary (which is what MRW do
exaggerates countries differences in human K)
Part IIB Revision, Stephen Ward 6 Development
Lecture 1B Institutions and Economic Growth
Fundamental causes of underdevelopment
Geography and Climate (Montesquieu 1748, Myrdal 1968, Diamond 1997, Sachs 1998)
Culture (Weber 1930, Landes 1998)
Colonial History and Institutions (Engerman and Sokoloff 1999, Acemoglu, Johnson and Robinson
1999)
Montesquieu
o Heat of climate so excessive body can be without strength
Prostration pass to spirit so no curiosity, enterprise, generous sentiment etc.
o People more vigorous in cold climates (braver)
o Argues lazy people tend to be governed by despots while vigorous people could be
governed by democracies
Hot climates conductive to authoritarianism and despotism (bad institutions)
Diamond
o Importance of geographical and ecological differences in agric. technology and
availability of crops and animals
Sachs
o Economies in tropical ecozones nearly everywhere poor while in temperate ecozones
generally rich
Certain parts of world geographically favoured with geographical advantages
Access to natural resources
Access to coastline and sea
Advantageous conditions for agriculture (Tropical agric. faces several
problems leading to reduced productivity of crops (particularly
staples))
Advantageous conditions for human health (e.g. lack of infectious
diseases)
Culture = relatively fixed characteristic of group/nation, affecting and preferences (e.g. religion)
(fixed effects that vanish over time)
Smith & Churchill suggest British cultural and political influence beneficial more so than
Spanish and French (Anglo-Saxon work ethic vs. Spanish predatory attitudes)
Why dont all countries develop growth promoting culture?
o Culture not useful in understanding Korean divergence (in growth rates of N and S)
o Inconsistent with several failures of British colonies in West Indies and Africa
Reversal of Fortune countries more wealthy in 1500 poorer now and vice versa
Rules out geography and culture explanations for good outcomes
Recent contributions state importance of different colonialisation strategies
European powers established good institutions with relatively equal dist. of wealth and
human K in some places these places flourished
Some places established exploitative institutions with very unequal dist. of wealth and human
K in other places these places stagmented
Acemoglu, Johnson and Robinson (2001) The Colonial Origins of Comparative Development: An
Empirical Investigation
Very influential paper
o Does pattern of colonisation determine institutions today?
Part IIB Revision, Stephen Ward 7 Development
o Do institutions cause growth?
Obviously institutions matter but how much? Cannot be inferred from OLS estimates due to
reverse causality
Growth literature assumes institution-free, or zero transaction costs: complete mkts., perfect
info, no enforcement costs
What gets in way of poor countries adopting better technologies, increasing savings and
accumulating K?
o Poor institutions e.g. property rights, public good provision, govt. regulation, other
transaction costs
Europeans more likely to benefit from good institutions when they are significant fraction of
population i.e. settlers
o Lower strata (lower social class) of Europeans place constraints on elites when there
are significant settlements
European settlements foster better institutions
But European settlements endogenous
o May be more likely to settle if society has greater resources or more potential for
growth, or less settlements when greater resources that can be extracted without
settlement
o Look for exogenous variation in European settlements disease environment
Some colonies, Europeans faced very high death rates because of diseases for
which had no immunity (particularly Malaria and Yellow Fever)
o Lower mortality = lower disease = better to invest in
o Areas Europeans didnt settle had higher pop. density and some immunity to diseases
Econometric argument
o Potential mortality of European settlers affected settlements, which affected past
institutions, which affected current institutions
o Instrument is early settler mortality
Econometric empirical difficulties
o Endogeneity
Reverse causality
Omitted variable bias variables omitted from analysis that determine incomes
are correlated with institutions
o Measurement error coefficient always biased towards zero
Measures of institutions constructed ex-post perhaps natural bias in
nominating better institutions in richer countries
Errors in measuring institutions
Why might institutions affect economic outcomes?
o Micro studies emphasise links between secure property rights and incentives to invest
(Besley 1995)
More secure property rights
More investment in physical and human K
Higher levels of income
2 difficulties in testing this
Reverse Causality - Economies that are rich can afford better institutions
Omitted Variables - Other fundamental differences between countries that
determine both differences in institutions and incomes per capita
o Solution find variable with 2 crucial properties = Instrument
Highly correlated with institutions
Does not directly influence todays economic outcomes
Here use differences in mortality rates faced by European settlers
Justification of settler mortality as instrument
Part IIB Revision, Stephen Ward 8 Development
o While culture shapes institutions, if look only at former colonies, colonisation strategy
driven in part by how feasible European settlement was
o Different sets of colonisation policies created different institutions
Extractive states (Leopolds Congo) no protection for priv. property, no
checks and balances against state expropriation
Neo-Europes replication of European states with strong protection of priv.
property rights (Australia, NZ, Can, US)
Choice of strategy influenced by feasibility of settlement If climate/disease
conditions favourable, migration and permanent settlement
o Colonial state and institutions persisted after independence
Setting up functioning institutions (law courts etc.) have sunk costs
Costs of switching to extractive institutions could be high (and vice versa)
o Small elites in past foster small elites at independence and persistence of extractive
institutions even in democratic climate (Lat. Am., Sierra Leone)
In Lat. Am., monopolies and regulations created by Spanish persisted after
independence because size of elite was small
Institutions may be complementary to irreversible investments in human and
physical K, leading to force for persistence by owners of such investment
o (Potential) settler mortality affected early settlements, which affected early
institutions which affected current institutions, which affected current performance
o Settler mortality explains 25% of variation in one measure of efficiency of todays
institution: Risk of expropriation obtained from political risk services but settler
mortality should not directly affect todays incomes
Data mortality of settlers (soldiers, bishops and sailors stationed in colonies between 17
th
and
19
th
Century only available for 64 of 163 countries)
o Current institutions
Index of protection against expropriation political risk services index from 0
to 10 with 0 highest risk
Constraints on executive in 1990 from polity III
Early institutions
o Measure of constraints on executive in 1900 (not available for countries still colonies in
1900)
o Index of democracy in1900 (not available for countries still colonies in 1900)
o Measure of constraints on executive at independence
o Fraction of population of European descent in 1900 (measure of European settlement in
colonies)
Regression
+ + + Y X R y
i i i
'
log
log yi is log of per cap income in 1995
Ri is measure of current institutions ( is critical parameter)
refers to all other variables that may be expected to directly affect per cap incomes in 1995
o Difficulty is that Ri may be influenced by log yi, so mortality rates used as instrument to
obtain consistent estimates of
yi is function or Ri, Ri is function of Ci (current institutions), Ci is function of Si (measure of
European settlements in colony) and finally Si is function of early settler mortality rates
Results only persuasive if mortality has no direct effect on econ. performance. Reasonable?
o Colonial authorities differed in kind of cultural norms/legal codes they brought with
them, so its not settler mortal. That matters but these other characteristics
o Settler mortal. correlated with current disease environment and climate (if true,
settled where things better, so bad instrument)
Part IIB Revision, Stephen Ward 9 Development
o Other possible controls ethno-linguistic fractionalisation (hard if country diversified to
set up efficient institutions), measures of soil quality and natural resources, whether
country land locked
Omitted variables
o Institutions could be picking up effect of climate/cultures and traditions, which also
affect current productivity
Counter: European mortal. caused by diseases against which local pop.
developed some immunity, so previous disease environment cannot be picking
anything up today
Putting in direct effect of climate does not affect coefficient on current
institutions
o Countries equidistant from equator have very different cultures
Checking reverse causality
o Table 3 panel A various measures of early institutions seem to explain current
institutions
o Table 3 panel B early institutions on settlement in 1900 and settler mortality are
negatively related
o Table 4 reports 2
nd
stage estimates
o Latitude has uncertain sign and is insignificant could mean studies that find large
effect of latitude actually picking up effect of institutions
Further worries
o Validity of approach depends on institutions being only (or main) channel through
which settler mortality affects todays outcomes
o Could be alternative channels
British institutions being better than others and British got the better colonies.
These had low settler mortal. and with better institutions grew faster. To
address need a dummy for British colony
o Robustness checks in table 5
Alternative takes
Sachs (2000) comment on AJR
o Both institutions and geography matter
o AJR find Geography doesnt matter as small sample of ex-colonies (with settler mortal.
rates) increasing sample size changes results
What determines expropriation risk and institutions? Culture? British institutions superior as
better economic incentives (La Porta et al. 1998)
Political factors France feudal lords more powerful relative to Britain (juries)
Sokoloff and Engermann in 18
th
Century, richest countries were in Caribbean (good soil, climat
and slave labour)
Initial dist. of human and physical K very unequal explains later collapse rel. to US and Can
AJR follow up paper countries rel. rich in 1500 rel. poor today rel. poor sparsely settled and
so induced Europeans to settle and develop good institutions
o In densely populated areas, extractive policies preferred
Institutions and Growth
Empirical analysis suggests if effect estimates causal, it would imply fairly large effect of
institutions on performance, accounting for approx. of income per cap. differences across
countries
Relationship could not be causal
o Reverse causality institution measure refers to current institutions plausible rich
economies able to afford or prefer better institutions
o Omitted determinants of income differences naturally correlated with institutions
Part IIB Revision, Stephen Ward 10 Development
o Measure of institutions constructed ex-post analysts may have had natural bias in
seeing better institutions in richer places
o These factors may all intro + bias into OLS but considerable measurement error in
institutions variable may bias OLS estimate downwards
Can solve all problems by instrumenting for institutions IV must bee important in accounting
for institutional variation observed but have no direct effect on performance
Ethnolinguistic fragmentation likely to be endogenous w.r.t. development (disappears during
growth and formation of centralised mkts.) and is correlated with settler mortality, the
estimate of protection from expropriation is likely understated the effect of institution
Part IIB Revision, Stephen Ward 11 Development
Lecture 2A Poverty and Poverty measurement
Choosing welfare indicator what constitutes a good life?
Must indicate standard of living growth, capabilities, ability to participate in society
Simple if could map into single measure but cant
Deprivation has multiple dimensions
Take one dimension income
o Aggregate different dimensions (HDI weighted sum of income, life expect, educ.)
Objection: way weight them, standard problems of creating P indices, income
may grow while LE may not
Number of different goals not aggregated Millenium development goals UN Millennium
Declaration Sept 2000 wish list with 2050 universal aim
o Eradicate extreme poverty & hunger, improve maternal health, universal primary educ.
o Combat HIV/AIDS, malaria and other diseases, promote gender equality & empower
women, ensure environmental sustainability
o Reduce child mortality, develop global partnership for development
Main goal of econ. development is reducing poverty and degradation and deprivation that
accompanies it
To target properly must know characteristics of poor and understand constraints that face
them in getting better off
Functional reasons linking poverty to mechanisms that actually create it and make it persist
over time. Think of regression: LHS = poverty, RHS = mkt. failures
Poverty Measurement
Household-survey based, consumption-based measures of standard of living. In most poor
countries, cons. mainly food plus essential non-food
Superior to income-based as
o Can be more readily measured (but needs care bought or from own prod.?) income
hard to pin down
o More stable than income (smoothed income faces shocks)
o From duality expenditure needed to achieve particular level of U can be seen as
money measure of that U level = money metric value of U welfare comparable across
people
Cons. better to income
o Perm income hypothesis this is what is important and is not observable
Consumers follow this gives better money measure of U so can use as an
approx. of welfare
Cons. is better functional mapping into well-being
Problems
o As measure of indiv. level of welfare non measurement of leisure, public goods and
non-mkt. goods
o How aggregate across indiv.? (Arrows impossibility thm. cant aggregate U across
household as with whole of society)
o Measurement error
o Commodities vs. capabilities space of commodities really appropriate?
o As comparable measure of welfare (anonymity, property etc.)
o Tastes people have different preferences
o Needs corrections for household size, economies of scale in prod. of goods & cons.
Part IIB Revision, Stephen Ward 12 Development
o P across time and space
Poverty = inability of indiv. or family to command sufficient resources to satisfy basic needs
Poverty Line = basket of basic needs defined by its value in terms of local prices (or $)
obtained. If indiv. (or households) income/cons. falls below this value, indiv. or household
deemed poor
Poverty and Welfare
Possible measures of welfare: per capita household real income (or cons.) Begin with c.d.f. (%
people below certain level) of x
o Social welfare as aggregate function: ( )
N
x x x x V W , , ,
3 2 1

where N = pop. size and x = indiv. welfare
V is:
o Increasing and monotonic in each of its arguments social welfare is therefore non-
decreasing in x with individual welfare the weak Pareto condition
o Focus axiom: concentrate only on poor, at poverty line z:
( ) ( ) 0 , 0 , 0 , , , , , , , ,
3 2 1 3 2 1
z x x x x V P x x x x V
N N

z is the last before the poor end. Dont care about others. P denotes measure of
poverty
o Symmetric (or anonymous) in its treatment of individuals it doesnt matter whose
welfare we talk about, provided we can compare across individuals all interested in is
total cons.
o More equal dist. preferred to less equal V quasi-concave (social indifference curves
convex to origin)
Weighted average of 2 equally desirable allocations superior to either
Transfer from poor to rich person will lower social welfare = Dalton principle of
transfers
Basis for z: absolute or relative?
o Absolute with minimum calorie needs?
o Relative? As in Europe (UK less than 60% mean income dubbed poor) not make sense
to do this 1
st
step all have enough to eat, then look at whether have as much to eat
as everyone else
Crucially, should be consistent
o Same across all x wish to compare
o Also when compare across countries or over time
Poverty line will shift with differences in tastes, activity levels, relative P and publicly
provided goods etc.
E.g. Difference between urban and rural poverty based on min. calorie intake per cap of 2300
cal per person + essential non food. Can violate consistency of cons.
o Engels law when becoming richer, cons. changes from food to non-food and also
within food groups
o Diet to get 2300 cal in urban areas more expensive compensate for this?
o P of some non-food goods often lower in urban areas, but food P often higher. D for
food less so food intake lower at any given cons. level. Work requires less energy in
urban areas meaning maintenance of body lower and so is food intake
o Urban households tend to prefer more expensive foods and pay more per calorie
o Someone above poverty line may move to another region e.g. rural to urban, and
become better off, but may be below poverty line in urban area
Leads poverty in urban areas to register increase even if indiv. better off from
moving
Constructing a poverty measure
Part IIB Revision, Stephen Ward 13 Development
Method 1
Start from explicit calorie norm 2300 cal per person per day in both rural and urban areas
(note in urban areas calories cost more must ship in etc. Therefore need more money to
survive? But urban typically expend less energy so may not be true)
Survey households on diets and calculate calorie cons. per capita as function of household
income per capita
Draw calorie Engel curve
= Food-energy Intake method of setting poverty lines
o Food-energy intake plotted against cons. expenditure
o Work out level of cons. necessary to achieve min. calorie intake = poverty line P
Problems with poverty line
Shifts with differences in tastes, activity levels, relative P, publicly provided goods etc.
o E.g.1: P of some non-food goods lower in urban areas (Food P higher though?). D for
food also less, so food energy intake less at any given cons. level. Does this mean
households poorer in urban areas? No.
work requires less energy in urban areas required maintenance of body lower
and so also food intake
Tastes different urban households prefer more expensive foods and pay more
per calorie
o E.g.2: Someone above poverty line migrates from rural to urban sector
May be better off but may be below poverty line in urban areas so poverty line
in urban areas registers increase even if absolutely better off by moving
No guarantee differences in poverty lines actually equal to difference in costs of acquiring
same basket of basic foods
Poverty measures - finding aggregate function
Represent x using c.d.f. of x
Poverty measures are different ways of representing area between poverty line and c.d.f. of x
2 steps
o Judgement of indiv. poverty: mapping xi into P(xi, z)
o Aggregating P: P=S(p(x1, z), p(x2, z)p(xN, z))
Examples of additive poverty measure:
( ) ( ) [ ]

N
i
i
z x p
N
P
1
,
1
( ) 0 , > z x p
i
if z x
i
and 0 otherwise
Head Count Index
Counting poor (p takes value 0 or 1)
Fraction of people below poverty line
( ) ( ) [ ]


N
i
i
z x I
N
P
1
0
1
where I = 1 if xi = z and 0 otherwise
Satisfies weak Pareto, symmetry but not Dalton
Poverty Gap
Accounts for degree of poverty
o Weight headcount by degree to which each poor person falls short of the poverty line
( ) ( )
1
]
1

,
_

N
i
i
i
z x I
z
x
N
P
1
1
1
1
let P1 = 0.25 total amount by which poor fall below poverty line = 0.25z population
Part IIB Revision, Stephen Ward 14 Development
Cost of removing poverty only if lump-sum taxes and subsidies are possible (no disincentive
effects) Cant do in real life how much in real money? In many cases very large
Continuous and concave in x
Sens measures incorporate measure of inequality among poor
Weighted average variant on squared poverty gap
[ ]
p sen
I I P P ) 1 (
0
+
Dont like as not differentiable
Not additive
measures average income shortfall amongst poor
z
x z
I
p

, and = Gini coefficient of


inequality amongst poor
P2
Variation on poverty gap also sensitive to dist. among poor
Squares shortfalls as a ratio of poverty line
Weights extreme poverty more than shortfalls just below poverty line
Gini Coefficient and Lorenz Curve
Lorenz Curve = way of examining dist. of income
o Plot of cumulative fraction of pop. On x-axis against cumulative fraction of income
received on y-axis
Gini = area between curve and 45
o
line
If inequality among poor = 0, Sens measure = poverty gap
Poverty measures must satisfy 3 things
Pareto add money to one person, social welfare should increase
Symmetry Everyone the same
Transfer take account of moving income between people
Sen conditions:
Dist. between rich doesnt matter for rich people
Monotonicity take money from poor and throw away
In poverty measures, use P instead to allow how much care about the very poor
P0 doesnt satisfy transfer or monotonicity
P1 doesnt satisfy transfer takes account of shortfall satisfies monotonicity
P2 satisfies all > 1 allows to satisfy transfer gives greater weight to poorer (further from
poverty line)
Poverty line = some way to cut off poverty
Part IIB Revision, Stephen Ward 15 Development
Individual Welfare
Contribution to Social
Welfare x z
-1
0
P1
P1
P0
Dont like Headcount ratio dont like. Impossible to
increase welfare by taking from rich and giving to poor

Z
Poverty Ratio = % people below z
If increase spread, poverty must increase
Mean increase and inequality decrease poverty falls double speed (East Asian Miracle of 60s,
70s, 80s)
Mean decrease poverty increases
Inequality increase poverty increases
India, Ghana, Brazil
Stylised fact
Where no growth, no poverty reduction need increase to get any poverty reduction
Where fast growth (sustained or for short period), rising inequality has accompanied growth
o Generally poverty fell as growth effect dominated in equality growth
Happened in China, Vietnam and Bangladesh too
Policy makers scared of inequality growth despite general growth what is going on in villages
etc. and common lands
May do well and reduce poverty nut election won by those focused on dist. concerns
Are official poverty numbers missing something?
Concerned with inequality only to point it affects lower tail?
Something should worry about over and above poverty concern?
Wellbeing of i
th
person = yi
o Ranked lowest to highest
o Need poverty line
Where z = poverty line. below the poverty line
What is poverty here?
o Headcount Ratio Measure = % people below poverty line
Sen (1976)
o Thought Experiment 1: Take 1 off poorest person and poverty must increase. Any
measure that does not include this must be problematic i.e.
Depth of poverty argument
Let = average income of those in poverty
= average shortfall of an in poverty person from the poverty line
= income gap ratio = response to above thought experiment
o Thought Experiment 2: Suppose replicate every poor person. Measure does not change
but surely increase poverty?
Should say when av. income of poor rel. to poverty line falls, poverty increases
and when increase people in poverty, poverty increases

Part IIB Revision, Stephen Ward 16 Development
Income
i.e. add all poverty gaps individually and normalise by population n
o Thought Experiment 3: Take off poorest and give to 2
nd
poorest. Doesnt affect but
what if value this as having made poverty worst? i.e. poorest person needs it more
To pick this up, can square the gap
o Can even use general power
= FGT (Foster, Greer, Thorbecke 1984) measures
Collect data stratify sample 15-20 samples in country and within those do random sampling (overall
could be all urban etc.)
Household survey on basis of responses construct measures
Ask questions on cons. much choose respondent within household
Use 30 or 7 days for recall period problem?
o Longer period gives lower flow rate really affected by recall period
Must be same when compare 2 surveys
Can give trends simply because change recall period
o India 94/95 survey, 30 day recall, as had always been
99/00 survey messed with this big problem sharp decline in poverty shown
so suddenly huge political significance, but cannot get clear answer here if
govt. claims poverty fallen
Official numbers missing something?
Population normalisation
% poor vs. number of poor
Sen (1976) axiom of population replication
Chakravarty, Mukherjee & Kanbur (2006)
o Ghana - % people in poverty fall (1% per yr i.e. very fast for Africa)
o If population increased 2% per yr
o Absolute poor must be an increase in 1% p.a. more street kids etc.
Pop. normalisation not so simple. On ground, reality = total number of poor
Even in countries where fall in incidence of poverty so great that despite population increase,
absolute number in poverty fallen (as in China)
Sens axiom
Consider 2 identical islands 40 poor out of 100 on each
Combine 2 islands into 1 200 people, 80 poor.
Poverty must be unchanged (seems reasonable)
But explains disconnect between on the ground and official numbers and responses
World Bank now also includes absolute numbers of poor
What ask households when collect info?
Lots on marketised cons. (good at)
Some on non-marketised cons. (e.g. crops in back yard)
o Consume crops that didnt buy?
o How much cost if had bought it? (become quite good at)
Not good at: public services how much would have cost for school/health etc.?
Part IIB Revision, Stephen Ward 17 Development
Shift from mkt. to non-mkt. can have big effect
Possible stats show improvement in poverty but may not actually be one
Intra-household inequality measure at level of household (not individually)
o E.g. 10,000 rupees household of 5 members, 10,000 rupees household of 4 members
Take per cap cons. of each household
Standard procedure allocate per household back to each indiv.
o Suppose 10,000 household of 3 adults, 2 kids and 10,000 household of 2 adults, 3 kids
Child has adult equivalent i.e. 3 x 0.6 for example
o For official stats, generally ignore composition and just do per cap.
Haddad & Kanbur (1990) estimate inequality understated by up to 30%
True inequality > official inequality
o Had info on indiv. level food cons. for Philippines. Convert to ca;orie intakes and had
info on indiv. level calorie intake
o If didnt have indiv., what would you get?
Aggregate up and do as official and would see results are different
o Apply same 1% point growth rate to both dist. fall in poverty is less for the more
unequal dist.
Shows why fall in true poverty < fall in official poverty
Forced into assumption that everything within household equal
o Huge poverty implications strong ev. that high inequality within households i.e.
allocation within households not equal
Apply some growth rate to 2 dist.
Fall in true poverty < fall in official poverty
Income Pooling Hypothesis
2 people bring in 2 incomes, take whole household pooled income = unitary model of household
If persons cons. screwed towards income, ev. against unitary model
If woman brings in money and her and childrens consumption increases, evidence against
unitary household model
Irrefutable evidence against imposes absolute equality in household
Official poverty numbers missing something cont.?
Something not right about one persons agony balanced by anothers gain i.e. 2 people exactly
swap incomes
To what extent right ethically to aggregate in this way?
o Even amongst poor some churning occurs
o In times of major structural shifts (which led to major growth rates and falls in
poverty) expect gainers and losers
But even though gainers gains dominate losers losses, no case to ignore losses
Panel data show considerable churning e.g. Grootaert & Kanbur (1995, 1997) Cote dIvoire
panel example
Part IIB Revision, Stephen Ward 18 Development

True dist.
Official dist.
o Overall fall in poverty composed of people falling in and out of poverty
Regional variations in poverty reduction (Kanbur, Venables 2005) e.g. Ghana different N and S
o Fall in poverty primarily concentrated in S where coco export growth increased. N far
less (Sahara desert)
o National shows fall whereas in N, measures that show depth of poverty stagnated or
even got worse
o Between group inequality as N Islamic and S Christian
Why normative concern on inequality?
Groups ethnic, religious, racial, regional
ANOVA or Theil decompositions into between-group or within-group
Salience of rising between group component
o E.g. spatial inequality Ghana (Kanbur-Venables Global Project)
o Means of racial groups inter-racial been coming closer but within groups, inequality
increasing especially blacks with rise of black middle class
o South Africa strange Between group narrowed, within group widened and total
inequality fell
Between group component that really worries people (pol. makers), even if total inequality
remains
Stylised fact: Indonesia/Asia overall inequality fell but change in nature of inequality fall.
From late 80s, between groups began to widen = another reason why not to celebrate poverty
In South Africa, even though between group inequality falling, due to history this is still the
only relevant issue
Summary
What axioms should functions have?
Focus axiom since interested in poverty, what happens above poverty line should not matter
at all have no effect on poverty
Doubling all incomes and poverty line should leave all incomes unchanged
Fall in poor persons income should always increase poverty
Sens axiom of population replication
All axioms put mathematical restriction on function what function satisfies axioms?
Intrahousehold inequality?
o Poverty over time follow indiv. panel data and churning
Measures emphasise depth of poverty not falling may in fact be rising
Equality of opportunity
If all take risk and all equal to start with
If not all equal at first, concerns
Circumstances cannot be altered by indiv.
Effort can be altered by indiv. (how distinguish in actual data?)
Can we really differentiate when a childs circumstances is a result of parents efforts?
All parents equal all take risks some go well, some badly kids very different
o Redist. across kids? Not fair taking away parents opportunity
Much of it luck?
o If so shouldnt be concerned should redist.
What can/should be done?
Are growth enhancing pol. bad for dist? What are they?
Is integration into world economy good or bad for growth?
Part IIB Revision, Stephen Ward 19 Development
o Success stories of countries with trade restrictions did develop generally dont focus
on those that dont develop
o Requires details of how trade pol. implemented
Chinas openness helped
Africa closing itself bad
Korea closing pol. helped enhance growth
o 50% and Q restrictions too high, need lower
o 15%-20% & no Q restrictions keep pushing doing OK
No matter what measures take, will have dist. consequences
o Trade pol. = blunt instrument
= macro instrument
Change rel. P
Net sellers / net buyers affected differently
Poor very heterogeneous
Economy: Traded and non-traded sectors
Rel P non tradables > tradables
o Nom. ER overvalued so over-pricing of X. Many reasons for:
o Bad for economy? Suppose large X sector good for economy
If nom. ER gives rel. P, distributional effects result. Poor and rich in both
sectors
Tradable sector better off and fall in poverty here
Non-tradable sector worse off and increase in poverty here
Total poverty change = weighted av. of both
FGT measures of poverty = Additively Subgroup Decomposable
A
X
A
B
X
B
1 +
B A
X X
Go to group A and calculate poverty in this group, then same for B
A
P
, B
P
,
What is relationship between national pol. and Sectoral pol.?
B B A A
P X P X P
, ,
+
o Suppose
A
P
,
falls and
B
P
,
increases, total effect is a weighted
average of both
Even when overall effect good, composed of big swings in
opposite directions
Generalised Compensation Mechanisms (GCM)
Suppose society where no one allowed to fall below min. How?
o Tax all above minimum level
o Does not matter who gainer/loser is. Every time someone hurt, pick them up again
Many types of compensation mech. being/have been tried by developing countries
Conditional Cash Transfers (CGTs)
E.g. Mexican Progressa/oportunidades (Lemy 2006) good example with good results
o Children not completing school. Progression = financial incentive for child to stay in
school. Monitoring and cash payment made to mother
o Targets poor, comes from general govt. tax therefore paid for by rich
Many similar schemes for all countries need to think of totality of these schemes as GCMs
Part IIB Revision, Stephen Ward 20 Development
Kuznets (1955) Kuznets Curve
Data for early 20
th
century and late 19
th
century only picked up 2
nd
half of the curve
Evidence for turndown?
o Now used cross section to see if relationship still holds (80s and 90s), but dont find
above relationship in cross section view
o Must address issue on specific case basis
Equity enhancing pol. good for growth?
Equity good for growth somewhat threshold effect. Need level of K to get going need certain
level of investment
Society with very unequal dist. of income, resources below initial threshold, failure of credit
mkts. leads to no effect at all for them
Increase returns at lower end and decrease returns at higher end of spectrum credit mkt.
failures
o Take from rich / give to poor pushes income dist. together
o More equitable dist. leads to more investment in economy / faster returns etc.
Political system delivers pol. favoured by median voter
o What really need is pol. preferred by mean voter
Difference between mean and median = measure of inequality
Need more info to illuminate disconnect between policymakers views and people on the street
Part IIB Revision, Stephen Ward 21 Development
Per cap.
GDP
Inequality
Lecture 2B Estimating an Absolute Poverty Line
Does growth reduce poverty? Despite high rates of growth, there is no consistent evidence about the
effects of growth on poverty
There was a net decrease in poverty in Asia, North Africa and Middle East
Did not change in Latin America and Sub Saharan Africa
Rose in Eastern Europe and Central Asia
Economic growth is typically pro-poor in that absolute poverty (measured against a poverty line
with fixed real value) falls with growth.
Does not follow that policies good for growth are good for poverty reduction.
Point is that growth is about average the average might move up but the lowest end might
not be shifiting very much at all.
Deeper point - Inequalities in income, education and endowments all change the growth elasticity of
poverty.
Poverty = inability of indiv. or family to command sufficient resources to satisfy basic needs
Exist pre-determined and well-defined levels of standard of living problem is that lines exist but
views on where located differ
Poverty line = basket of basic needs defined, value in terms of local P (or $) obtained. If indiv. (or
households income/cons. falls below this value, indiv. or household deemed poor
Difficult to distinguish between marginally better off and poor
2 methods of setting
Food Energy Intake Method (FEI)
Fix food energy cut-off in calories
Find cons. expenditure/income at which person typically attains that food energy intake then
run regression of calorie intake against cons. expenditures/income
Poverty line = level of total expenditure at which one can expect person to be adequately
nourished
o Unique only if average level of FE intake strictly increasing in cons.
o Method also includes allowance for non food cons.
Application Indonesia 1990 FEU used by Indonesian Central Bureau of Statistics
o Urban poverty line Rp 20,614, rural poverty line Rp 13,295
Does this reflect true differences of acquiring same bundle?
Part IIB Revision, Stephen Ward 22 Development
o Urban bundle has higher cons. of superior staple (rice over cassava or corn) and more
fashionable vegetables like tomatoes (rather than cassava leaves) as well as meat,
chicken and drink
o % of people below poverty line in rural areas is about 12%, whereas in urban areas it is
almost twice that at 21%
At any given poverty line this will always be true
o The difference in costs for acquiring the basic bundle must be 45% for poverty in rural
areas to be higher, but the difference in poverty lines here is about 55%
Cost of Basic Needs Method (CBN)
Specify reference household deemed to be poor
o Household with mean values of endowments of poorest 15% of Indonesia, ranked by
cons. per cap (same group deemed poor by Statistics Bureau)
Set line for each region
o Take reference cons. bundle multiply by (recommended food intake / energy intake
of reference household) the amount of typical food needed to reach caloric intake
Value this at local P
Person deemed poor if live in household that cannot afford this food bundle, chosen to yield
2100 calories and consistent with typical diet
Difficult problem = non-food
o What is bundle of basic needs? i.e. other goods, not just food
o Amount of basic food given up to have non-food item what level of non-food spending
one can allow to displace basic food spending, at poverty line
o Look at household that do not reach nutritional level but could do so and see how much
they spend
Begin with D function for food
Part IIB Revision, Stephen Ward 23 Development
ratio of total expenditure to cost of basic needs in food. If = 1, then household can just
afford basic food bundle
average food share of households that can just afford basic food bundle
o Method gives urban poverty line of Rp. 18,519 and rural poverty line of Rp. 15,693
o Now difference between poverty lines is 18% (as opposed to 55% by previous method)
Why so different results?
FEI method has much higher elasticity of poverty w.r.t. mean cons.
How get $ per day poverty line used to compare across countries?
Ravallion, Datt and Van De Walle (2001) and Chan and Ravallion (2001)
Use primary survey data household income or cons. normalised by household size
Use PPP estimates
Regression:
where zi = poverty line for country i with cons. per capita ci (with ci > c
min
the lowest cons. per
capita in the sample of countries)
o Intercept provides lower bound to log poverty line for poorest country in sample
Estimate was 3.46, representing $1.08 per day with 95% confidence interval
Hence the $ per day with 95% CI
o Results: % of population living below $ per day fell from 28% in 1987 to 23% in 1998
o Too little growth of average hhd incomes + persistence of high initial inequality in both
incomes and other things prevented poor from growing too
Part IIB Revision, Stephen Ward 24 Development
Lecture 3A Poverty Traps
Dollar and Kray (2002)
On av. in world:
1
ln
ln
% 20

meanincome
poorest
Y d
Y d
i.e. incomes of poorest quintile moved almost 1:1 with av. incomes
State Growth good for poor but ignores idea of poverty traps
Persistent poverty and poverty traps
Poverty has tendency to persist (people who poor tend to stay poor for while) suggests idea of
poverty as low-level equilibrium even if disturbances away from it, forces bring economy back
to it
Poverty Traps or Vicious cycle of poverty mechanisms that perpetuate poverty
o Critical threshold under which can easily slip but hard to escape
Evolution of wellbeing
o Governed by history (path dependence): given initial conditions, otherwise identical
indiv. locked in poverty and others not = equilibrium
o Multiple equilibria (empirical ev. v.hard to find) and coordination failures
More difficult to translate to micro dimension: some notion of group behaviour or group
endowment that drives outcomes (e.g. Geography, infrastructure, cultural norms i.e. parents
poor so must work kids, who therefore receive no education, hence kids are poor and become
parents etc.)
History dependent growth processes (initial inequality or bad luck decides steady state)
o Inequality in assets means some people locked out of credit mkts. so some trapped
forever as wage workers
o Assed inequality combined with missing mkts. in credit and insurance means people
with low assets compelled into low return activities with low risk. Difficult to climb out
of low asset-low income trap
Important for pol.
o Overall growth accompanied by increase D for lab, so growth will benefit poor (at leadt
to some extent)
o Traps are different: equilibrium positions with exclusion from econ. processes. People
may drift into poverty but large shocks needed to get them out
So growth may not benefit this group
o Some cases interventions pareto-improving i.e. could pay for themselves
Perfect mkts.
First Welfare theorem every comp. outcome pareto efficient suggests equality of returns to
FOP
o Route of poverty is poor have smaller share of initial endowments
Second Welfare theorem each possible Pareto efficient outcome is competitive equilibrium
for given dist. of endowments
o Simple view of poverty and its solution (poor should earn same as rich for same hectare
of land or same hours work)
Poor health and nutrition consequences
Muscle wastage, retardation of growth, increased illness
Inability to work for long or with concentration
o Weaker, less able to work hard for long time
Part IIB Revision, Stephen Ward 25 Development
o Mentally incapacitated
Lower mental capacity of unborn children and their growth
= vicious cycle: Poverty to under nutrition to lower abil. to work and earn income to under nutrition
Suppose poor household spends about 70% of income on food (norm in poor country)
o At low levels of nutrition (income), most nutrition for maintaining body or resting
metabolism
o So little energy left for work. Past this, energy available rapidly but soon diminishing
returns sets in, past turning point
Perfect mkts. case
o Firm demands more lab. Until marginal value product equal to factor prices (wage)
o Q = F(n): profit maximising D for lab. Given by: pF(n) = w
o Consider Q = F(n.(w)) with = efficiency unit or output piece, (w) > 0
Profit mac: find max n and w s.t. max obtained:
( ) ( ) wn w n F p . .
F.O.C.
n: ( ) ( ) ( ) 0 . ' w w w n F
w: ( ) ( ) ( ) 0 1 ' . ' w w n F
So ( )
( )
w
w
w

'
Plot (w), (w) > 0, (w) > 0 over region and then (w) < 0
o Capacity curve relates income and work capacity (productivity)
Found by linking different nutrition (or income) points to corresponding levels
of work capacity generated by indiv.
Higher income means better nutrition
Better nutrition first used by body for resting metabolism, then into higher capacity for work
o First, v. little energy left over for work so work capacity close to 0 and does not
increase quickly as levels change once resting metabolism taken care of, marked
increase in work capacity with nutrition
o Finally diminishing returns a natural bodily limits restrict conversion of increasing
nutrition to ever-increasing work capacity
Work capacity curve convex and intersects 45
o
line from bottom
Piece Wage Schedule
Part IIB Revision, Stephen Ward 26 Development
Income
(nutrition)
Work
Capacity
Capacity
Curve
Total income
Work Output
Piece rate
Piece rate is w/ shows how much income
get from each task perform
Wage v
*
at which people can just work creates a discontinuity
If D curve for lab. cuts above gap, mkt. determines equilibrium piece rate
If S > D, D curve may cut S curve in gap
o If piece rate > V
*
, excess S and wages go down
o If piece rate < V
*
, excess D and wages go up
V
*
is equilibrium allowing for U (which is involuntary) some workers get work (luck, 1
st
come
etc.) while others with same work capacity will not
Unemployed cannot bid down piece rate cannot supply desired work at lower wage (will not
have energy)
Involuntary U
o Person involuntary U if cannot find E in mkt. that does employ someone exactly like
him, and if that person through being employed is better off
o Imagine lottery that determines who gets employed = Poverty Trap
Low wages lead to low work capacity, which closes access to employment
Suppose workers have some non-lab. income from assets
o Rural workers likely to have income from various sources such as land own
Which direction will capacity curve shift? Workers have total income = rent +
piece rate wages
2 workers, one with land one without, which curve lies above other
Who more likely to be employed: rich or poor?
Who earns larger wage income if both employed?
o Functioning of lab. mkt. magnifies asset inequality!!
o Richer worker earns more wage income
Applies only to pool of the poor rich will be at top end of capacity curve
anyway
o People without assets doubly cursed dont enjoy non-lab. income and at disadvantage
in lab. mkt. relative tot those who do possess assets
Part IIB Revision, Stephen Ward 27 Development
Income (nutrition)
Work Capacity
Capacity Curve
V1 V2
V3
V. low piece rate (V3) means very low
capacity. Need to increase rate to get a
reasonable amount of work
Individual Labour Supply
Piece Rates
Aggregate Labour Supply
Piece Rates
V1
V2 = V
*
V3
V2
V1
Rank indiv. by amount land they own
o Income dist.: n
*
of them have no land
o Technology: Food = G(Land, Labour). G is constant returns to scale
Every household had D for labour unless without land
Landless are net suppliers of labour
G is CRS, lab. demanded including own is independent of dist. of land
o lab. demanded = aggregate lab. D
Labour S curves
Rank by land holding
o More land and more rents, the more the farmer can take care of some of his nutritional
needs
o Minimum wage necessary to work falls with wealth
Wealthy will also need higher compensation in order to work value of their time is higher
o Combine abil. to work with the willingness
o u-shaped part is minimum at which people able to work at low levels of non-lab.
income, abil. is the binding factor
3 possible regimes
o High lab. D full E
o Low lab. D involuntary U of landless (some) and voluntary U of the landed (some)
o Very low lab. D involuntary U of all landless and some landed
Falling part of curve shows when ability is the operative constraint whereas the upward sloping
portion represents zone where willingness is the operative constraint
For each piece rate in mkt., supply of lab. given by amounts worked by all whose min. piece
rates lie between going mkt. wage = indiv. who willing and able to work at going piece rate
o By varying piece rate, trace out supply curve
At given piece rate, horizontal distance between red line individuals employed
o To right, people able to work but dont wish to = voluntarily unemployed (landed
gentry)
o To left, unemployed as unable to get work at going piece rate = involuntarily
unemployed (going W rate does not allow them to carry out sustained work over time
without seriously impairing health and physical strength)
Evidence: India cost of one days calories is Rs 2.60 = $0.04
Average daily wage of agric. Labourer is Rs16.00 = $0.23
Evidence on employment probabilities is landless no less likely to be hired than landed
Part IIB Revision, Stephen Ward 28 Development
Individuals
Min piece rate
Non-labour
Income
Combined minimum
Min. piece rate for
abil.
Min. piece rate for willingness
More land you have means less willing
you are to work so require higher
wage
Labour Income
Dasgupta & Ray (1986,1987) - Children inherit income from parents. Eat all this income, work
and then pass this income on to children
o Suppose capacity curve always above 45
o
line
o Does final income of dynasty depend on where began?
o Suppose always below. What now?
o Suppose capacity curve cuts 45
o
line from below. What now?
How much does households income increase investments in human K and why important?
o Equity efficiency trade-off always at centre of tax and dist. pol. Dont want poor to
starve but in order to guarantee them an income, have to distort effort choices of the
rich
o Know asset redist. will raise both output and incomes if a poverty trap so asset
redistribution is efficient
o Idea that income effects large are basis of notion of poverty trap i.e. small change
income has enormous effect on working capacity
If income effects small, may be better to lower P of investment in human K (build school,
subsidise textbooks etc) rather than redist. assets
Effect of average per. Cap. income and other variables on the income of the poor
Necessary to instrument average per cap income by lagged growth and income because
o Simultaneity i.e. reverse causality some reverse causation from average incomes of
poor to average incomes or from log quintile share to average incomes
o Measurement error in average incomes leads to biases instrumenting does not remove
problem however does reduce it
o Omitted determinants that correlated with either X or average incomes can also bias
results
If growth is food for poor, log income should have a positive coefficient greater than 1 when
poor income is regressed on it
o Reject hypothesis coefficient = 1
Growth just average may not be good for all poor people
As regional dummies, may not be good for poor in all countries
Suggests may be good for poor, but cannot be sure
o Should focus on IV due to problems of OLS
o If > 1, poor gets more than rich redistributive
o If = 1, inequality doesnt change good for poor in that absolute position increases, but
inequality remains
o With regional dummies, some areas have growth with inequality and others without
must be careful
Inconsistent with Kuznets hypothesis (inequality varies with growth)
Although some studies state 1:1 relationship, many others dont find this. Differences in
institutions, samples (countries) used etc.
Basic conclusion that income changes and changes in inequality unrelated
Dollar and Kraay (2002) find very similar results in estimation of same model
Deininger and Squire (1996), Chen and Ravallion (1997) and Easterly (1999) come to same
conclusion though using smaller sample of countries
Forbes (2000) applies similar techniques in similar panel set to study effect of inequality on
growth
Main differences - lack of evidence of significant negative impact of openness to international
trade on incomes if poor. Consistent with Edwards (1997) but Barro (2000) finds all trade
volumes significantly positively associated with Gini coefficient.
Part IIB Revision, Stephen Ward 29 Development
Spilimbergo et al. (1999) find several measures of trade openness associated with higher
inequality and effect is lower in land or K abundant countries and higher in skill abundant
countries
o Factors such as difference in measure of inequality could contribute to differences, as
could differences in sample of countries used and time periods of data. Different
measures of openness will affect results as will differences in econometric specification
and technique
General relationship between income per cap. and income dist: does income per cap. have significant
influence on income dist. and are there significant regional effects in relationship?
Regress log of index of income dist. on log of per cap. income for both top decile and bottom
40%
Share of richest people first increases, then decreases (inverted u shape)
Share of poorest first decreases, then increases (u shape relationship)
Suggests inequality first increases, then falls i.e. supports Kuznets hypothesis
Country dummies
o Developed countries show less inequality supports Kuznets
Could be reverse causality income increase not necessarily cause inequality inequality could
increase income supports Kuznets but cannot say definitely right
Omitted variable bias
Kuznets country develops over time data here = cross-section of countries cannot use as
support
Cross-country analysis leads to potential heteroskedasticity problem (larger errors at higher
levels of income, or larger errors in LDCs than developed countries measurement difficulties),
meaning all standard errors are biased i.e. all tests unreliable
o Use Whites heteroskedasticity adjusted standard errors to solve
Regional effects could be significant
o Different institutions
o Different govt. systems e.g. long colonial history, extractive system, education
system etc.
Poor health can affect physical growth
o Muscle wastage leads to retardation of growth and increased illness
o If workers poorly nourished, may have inability to work for long or with much
concentration
o Under nutrition can have effects on future generations with lower mental capacity of
unborn children and effects on their growth
o Cycle of poverty leading to under-nutrition, lowering ability to work and earn income
leading to under nutrition etc.
o Poor unable to escape cycle and get stuck in poverty trap
When piece rate low, poor person may not take job (see capacity curve above)
o Even with higher piece rate, W of richer person > poorer person condition of the poor
is worse
Credit mkts. effectively allow capacity curve to shift left, which can effectively give a huge
change in labour income for a given piece rate (see capacity curve above) for V1, previously
would not work, however opening of credit mkts., and shifting of capacity curve left, allows to
work and earn labour income where could not before
Lack of collateral makes it hard to borrow
Statistics say rate of default higher among the poor less willing to give loans
Moral hazard
o Poor countries 70% of income spent on food
Part IIB Revision, Stephen Ward 30 Development
If poor workers could only find casual or ST employment:
o Firms not so worried about maintaining health of workforce lower wage
Metabolism effect
Capacity effect
o For certain work levels, capacity effect dominates as can work better. Means firms may
sacrifice some SR output or pay higher W so indiv. can take care of nutrition
o But in casual mkts., wont happen
No guarantee employee still there tomorrow
If invest in employee, employee in better health for other jobs bids up own W
rate employees have incentive to leave
Same for all employers non invest = prisoners dilemma
If Employment Guarantee Scheme (EGS) introduced by govt. in state of Maharashtra (India)
theoretically should benefit but in practice:
o Mis-targeted
o Still involuntary U
o Piece rates self select well nourished (see diagrams above)
Main aim = provide insurance scheme for poor
o Gaiha (1993) helps poor to stabilise income suggests well targeted at rural poor and
widely regarded as successful pol. intervention
o Always option which guarantees job when dont have job, can go get one
Scheme guarantees every adult who wants job in rural areas will be given one
provided person willing to do unskilled manual lab. on piece-rate basis
Official instruction work provided within 8km radius from workers residence
in practice guarantee operates at district level, which may require person t
travel long distance for only few days of work
EGS will make difference if encourages poor to work and earn reasonable wage, allowing them
to gain nourishment that would not be able to otherwise
Helped negotiate with agriculture jobs increased mkt. power allowed better W in other jobs
W rate usually low: piece rate < min. agric. W as govt. not want to disturb other mkts.
Lab. intensive job offered very poor cannot take advantage as poor nutrition does not allow
them to do so
o Majority of participants not the very poor
Gaiha (1993) participants followed U-shaped dist. in 1979 with majority
earning incomes well above poverty threshold. Similar relationship in 1989 but
share of non-poor higher
Both years, share of poor in EGS corresponded to share of poor in LF
Vast disparity in sex ratio with far more males in scheme and men also earned
far more than women did through the scheme
More affluent appeared to benefit from larger gains from EGS in 1989
o Poverty traps may persist
o Reasons = low piece rate and mis-targeting
Problems
o Not enough jobs for all who want to work
o Although increase in welfare of many, and these left jobs to better ones, programme
not able to generate significant income levels for other participants
Conclusions raise concerns about design and implementation of EGS and its potential to target
poverty
But, despite mistargeting, large subset of chronically poor depended heavily on EGS, which
enabled them to raise their household incomes more than moderately with significant welfare
gains
Part IIB Revision, Stephen Ward 31 Development
Participation in EGS highly sensitive to overall econ. conditions large withdrawals from EGS
prompted by improvement in employment prospects in rural areas which often carried higher
W
Datt and Ravallion (1994) ideal embodied in EGS of providing work on demand not being met
o Projects do appear to generate sizeable net income gains to participants and OC of
employment is low
o Welfare gains appear far greater than implied by using mkt. W rates for similar work to
value the forgone income
Part IIB Revision, Stephen Ward 32 Development
Lecture 3B Testing for Efficiency of Rural Markets
Udry (1997) Efficiency and Mkt. Structure: Testing for Profit Maximisation in African Agriculture
Do rural areas of poor countries have well-functioning (competitive) mkts.?
Main features of rural lab. mkts.
o Self-employment
o Auto consummation cons. own prod. from work on own farm/business
Possible specialisation not happening constrained on some way?
o Household labour often used as an input
Households make decisions about production, time allocation of household members and
consumption simultaneously
Claim: With complete mkts., production decisions of household separable from cons. choice =
Separation Hypothesis decisions on production not affected by preferences for consumption.
Production looking at prices only
Stage 1 household maximises profit
Stage 2 household maximises utility subject to income constraint that includes value of
maximised profits
If mkts. work well, household behaves as if could hire all inputs (or value the time of members
according to mkt.)
Can allocate resources to prod. as much as any independent firm would do
o Note incomes from prod. affect cons.
o If decisions separable, prod. decisions unaffected by cons preferences i.e. cons.
preferences dont dictate how going to produce
Household members get U from cons. of a good and leisure
o P = price of cons. good, W = wage rate
Prod. technology on farm: F(L,A), L = labour used, A = area of land
E = endowment of leisure
Problem
( ) l C MaxU ,
s.t.

[ ]

+ 0 wl pC wE
(1) = full income constraint
( ) 0 , rA wL A L qF (2)
0 , l C
(3)
Note (1) binding at solution and U increasing in . Max (2) then can plug into (1). What do to U depends
on what do to
Max profit w.r.t. L and A, then max U:
( ) l C MaxU ,
s.t.
( ) [ ]

+

0 , , 1 wl pC r w wE
Where ( ) 0 ,

rA wL A L MaxqF where 0 , A L
o Prod. decisions on any piece of land depend only on P and characteristic of that piece
of land
o
*
is only function of prices (P of outputs and inputs)
Part IIB Revision, Stephen Ward 33 Development
o Main issue test whether problem can be written this way are prod. and cons.
decisions separable?
Method 1
Examine dist. of yields (output per unit area) and inputs on similar plots and check whether
dispersion very large. If little deviation from perfect mkts., dispersion should be measurement
error and not significantly different within a village
Method 2
Check whether amongst any group of plots of land, facing the same P, the choices of inputs and
outputs are the same (controlling for differences across plots)
Test:
Method 1
vhtci vtc vhtci vhtci
X Q + +
h
vhtc vhtci vhtci
vhtci
X Q + +
where v = village, h = house, t = time, c = crops, i = plot
X summarises plot-specific characteristics no q as like a constant term exactly same characteristics
Assume constant r.t.s. prod. functions all yields (output per hectare) also input per hectare
2
nd
regression identical to 1
st
but in second fixed effect is household specific rather than village
specific
o Household specific effect about plots controlled by specific household
o Distribution of error terms here reflects variation across apparently identical plots
controlled by same household
Regression 1 about plots within village
o If distribution of errors here is same as in 2
nd
equation, the hypothesis of separation is
valid
o If separation hypothesis valid, identical plots have identical characteristics
In real world, not identical so account for differences fixed effect term takes
account of P
Method 2
vhtci vtc vhtci vhtci vhtci
E X Q + + +
where E is exclusion restriction. If separation is valid, coefficient should be zero and the only things
that should matter are plot characteristics and prices
This is the classic way of testing the complete mkt. hypothesis
3 possible candidates for E
o Farm size
o Household demographics
o Wealth or cash flow should not matter could borrow in world of perfect credit mkts.
Farm size restriction
o Consistent with large literature that finds inverse relationship between yield and plot
area, or farm size (area) and labour demand
Explanations
o Mkt. failure small farms face different opportunity costs than large farms
o Plot qual. not measured properly if small farms have higher qual. land, they are more
fertile = omitted variable problem
o Decreasing returns to scale how control for this?
Part IIB Revision, Stephen Ward 34 Development
Looking within
household across plots
Random error Fixed effect
Compare within same household control for area of plot and use total area on
other plots cultivated by same household as exclusion restriction
Household demographics
o Relationship between household size and yield do larger families farm more
intensively?
o Suppose lab. mkts. dont work well but land mkts. competitive (separation still valid),
however if households can choose qual. of land they farm, then household size will be
positively correlated with unobserved land qual.
Wealth or cash flow
o Good qual. land could also be correlated with wealth
May wrongly reject null hypothesis of separation because of correlation between omitted qual. of land
and exclusion restriction being tested
Each of the three tests, however, is subject to a similar important econometric caveat
Under separation, cropping decisions depend only on prices and plot characteristics
o Can account for prices through village fixed effects provided that the model is linear
o Plot characteristics, however, are problematic
Undoubtedly unobserved variation in land quality - omitted variables problem
and the bias induced could lead us to reject the null hypothesis of separation
when separation is in fact true.
Solution: Fixed effects regressions
Natural solution is to use IV estimates
o Unfortunately, while there are plenty of variables that are correlated with farm size,
plot size, household demographics and non-farm income, it is very difficult to make the
case that any of these are uncorrelated with unobserved land quality
o Therefore not able to identify likely instruments in any of the available datasets
Take 1
st
differences and eliminate fixed effects (plot qual.) and see what happens
Results
Tests for Burkina Faso:
Good data, collected over 4 years across 6 villages and 150 households
All farmers poor with annual per capita income < $60
o To test whether errors dispersed same within households within villages:

Part IIB Revision, Stephen Ward 35 Development

Village level variation higher (and controls for quality are very good in this data)
Also, indiv. have plots all over village
Under the null hypothesis that separation holds within a village, the distribution of the two
error terms should be the same. The estimates indicate that
v
has a much more diffuse
distribution than
h
(i.e. Village level variation is higher)
The much larger dispersion of
v
than
h
raises the proposition that the separation hypothesis is
violated
Suppose no measurement error or unobserved variation in plot qual., distribution of error for
household level less spread than would expect otherwise. Includes shocks still though
Suppose also complete mkts. in each village
o All villagers have access to mkts. and P same for all maximisation problem same for
all households
If replace household fixed effect with village fixed effect exactly the same
Table 2, Column 1:
Exclusion restriction is household size.
o In apparent violation of profit maximization, plot yield is increasing in household size,
conditional on prices and all observed characteristics of the plot
o Elasticity is about .2 and is significantly different from zero at any conventional
significance level
Table 2, Column 2:
Examines the relationship between plot yield and the area cultivated by the household on
other plots
o Again in apparent violation of the separation hypothesis, there is a highly significant
correlation between plot yield and other total plot area cultivated by the same
household (i.e. the area cultivated by the household on other plots)
o Striking result, because the implication of conventional land and labour market
imperfections would have been a negative correlation between farm size and plot
yield.
Result provides further evidence of violations of the separation hypothesis in Burkina Faso
Table 2, Column 4:
Conditional on short-term resource inflows and household size (and plot characteristics), plot
yield does not vary with the area cultivated on other plots. On the other hand, plot yield is
higher for larger households with large short-term inflows of non farm resources, conditional on
other household characteristics (i.e. Positive correlation between plot yield and short-term
cash inflows to the household)
Conclusions
Part IIB Revision, Stephen Ward 36 Development
What are the market failures?
Pattern of results
o Negative correlation between yield and the area cultivated on other plots
o Positive correlation between yield and short-term inflows of non-farm resources
o Positive correlation (possibly) between yield and household size
Have strong implications for the structure of rural mkts.
If augment log of household size in equation (to test if larger households farm more intensively), and
coefficient significant must be careful
If unobserved characteristics, under some assumptions (CRS), can be correlation between
unobserved land qual. and household size leading to omitted variable bias
o But here no unobserved variation in plot quality (characteristics) so size does increase
yield
Shows output not just function of P, but also plot characteristics separation
hypothesis rejected
Conclude failure in lab. mkt.
If failure in credit mkt., may get same result not easy to find which mkt. has
a failure
Udry paper has provided evidence against the hypothesis that farmers maximize profits in two African
settings.
In each case, shown
o Plot level production decisions affected by factors other than prices and the
characteristics of the plot itself
Thus the hypothesis that production decisions are separable from the rest of
the households allocation decisions is not correct
In both Kenya and Burkina Faso, economically large and statistically significant divergences
between actual input allocation decisions and those which would be predicted in a simple
model of a profit-maximising farmer
In neither case can cause of the divergence be determined with certainty using the broad-brush
methods of this paper - rejection of the separation hypothesis can be caused by any number of
possible market failures.
However, pattern of results is suggestive:
o In Burkina Faso, fact that plot output correlated with non-farm income and gifts as well
as farm size suggests further investigation include a close look at capital and insurance
markets
Part IIB Revision, Stephen Ward 37 Development
o
Lecture 4A Risk and Insurance
Farm household maximisation problem with lab. mkt. failures
( )
2 1
, , max l l C U
s.t. ( ) ( ) [ ]

+ + + 0
2 1 2 1
l l w pC E E w
( ) 0 , rA wL A L qF
0 , ,
2 1
l l C
2 1 2 1
l l E E L + +
Why does coping with risk in poor countries matter?
Low incomes also enormous volatility in incomes
Missing mkts. or imperfect mkts.
Informal institutions that might substitute for mkts.
Ask questions
What is risk sharing and is 1
st
best risk sharing feasible?
What is 2
nd
best risk sharing?
How effective are risk sharing schemes?
Full Insurance
Households risk-averse and if fair insurance available, households will buy it
If most risk idiosyncratic, households should come together to share risk
If do share risk, indiv. incomes should not affect cons.: only group level incomes should affect
it = 1
st
best outcome
Assume village community come together and offer mutual insurance
o Village can divide up amount of income Yc (sum of indiv. incomes) between members in
each period
o Any such division is a Pareto optimal outcome
Before community level of income known, imagine set of income sharing agreements that can
be written, binding into future
o Do this as where uncertainty and risk, plans will not be fulfilled
o Even identical households with identical initial inheritances and identical prospects will
end up with different cons. profiles over time
Some luckier than others
Some get good income draws or see investment do well
o If people averse to risk, this can be made better trade fluctuation in incomes for
smoother cons.
Households in same initial situation should be able to agree to share good and bad fortune
o Means increase in expected util. for all
o Will work if all mkts. complete and each state of nature can be insured against
How can optimal risk sharing be done?
o Each household has intertemporal U function written over states of nature and time
o Probability of each state of nature in every time period known
o To cooperate, households must decide who gets what share in the total Pareto
weights
o Assume social planner (village elders) decide. How?
Part IIB Revision, Stephen Ward 38 Development
Maximisation problem:
where i = pareto weight attached to each household i
= Lagrange multiplier on constraint that total household cons. no greater than aggregate
village cons.
Village elders control cons. of each household
Optimal (efficient) risk sharing how can it be done and what conditions must be satisfied?
Each household has intertemporal U function written over states of nature and time
probability of each state of nature in every time period known
To cooperate, households must decide who gets what share in total, which they call the Pareto
weights
Pareto Efficiency and Risk
i = 1,N indexes individuals in community
T periods, s = 1,S indexes states that can occur within each period with objective and
commonly known probability of occurrence s
Individual income yis and cons. cist
Suppose ( )

S
s
ist i s
T
t
t
i
c u U
1 1

u(.) twice continuously differentiable, u > 0, u < 0
Any efficient allocation can be characterised by

N
i
i i
c
U
ist
1
max s.t.

N
i
is
N
i
ist
y c
1 1
and 0
ist
c
i.e. social planner maximising inter-temporal U s.t. constraint that total household cons. not
greater than agg. village income
o Implies
( )
( )
t s j i
c u
c u
i
j
jst j
ist i
, , ,
'
'

i.e. efficient risk sharing characterised by condition that ratio of MUs of each
household equal to ratio of Pareto weights
Let ( )
x
i
e x u

,
_


1
i.e. cons. risk averse
Thus ( )
x
i
e x u

'
and taking logs ( )
j i jst ist
c c

ln ln
1
+
Which holds over all N households in st. Sum over these to get

,
_

N
j
j i
st
ist
N
c c
1
ln
1
ln
1

i.e. all idiosyncratic risk is pooled but aggregate risk cannot be pooled!
Suppose full set of financial mkts. and complete set of assets
o For each state and time period, there exists an asset that pays 1 unit if state occurs
and 0 otherwise
o Securities freely traded amongst households
= decentralised version of previous scheme
Part IIB Revision, Stephen Ward 39 Development
Insurance is natural device to deal with risky incomes
o A form of trading in risk and can be done because
Indiv. have different risk preferences rich may insure poor if less risk averse
Risk idiosyncratic when I have bad yr, you have good yr and vice versa
Limits of risk sharing arrangements
o Problems of moral hazard
Group members may not be sure that a particular member has suffered an
income loss or a consumption spike i.e. may be difficult/impossible to verify
the outcome, on which insurance transfers are predicated
Hard to say if this sort of information problem could occur within the village. A
Village that is segmented into subgroups may be unable to develop insurance
schemes that cut across these lines of segmentation, simply because the flow
of information is restricted
Suggests the need to extend empirical work to define more precisely what we
mean by the community, and to the extent that our notion of community fails
to caputre these points, we might fail to capture the existence of insurance
sub-networks simply because we are looking for one giant network, which
doesnt exist
o Problems of adverse selection
possibility that the occurrence of some insurable event can be influenced by
the unobservable actions of the individual e.g. in the presence of full
insurance, incentive to deliberately under-apply inputs is higher
o Difficulties of enforcing contract
A farmer with a good harvest may be required to make a transfer to the
common insurance fund but why would he do so?
Evidence on full insurance or 1
st
best risk sharing under this, cons. should track group income more
closely than it tracks individual income
Townsend (1993) - India
o Cons. smoother than income but indiv. income still affects cons. after controlling for
village income
o Neither U or illness affects cons.
o Poor and landless have more fluctuating incomes than any other group
Townsend (1995) - Thailand
o Indiv. incomes matter enormous in determining cons. full insurance fails
o Works better if farmers in north and NE alone considered
o Works worse for entrepreneurs and those near Bangkok
o Uses similar regressions, carried out on Thai data, and finds villages do relatively well
in exhibiting great deal of cons. smoothing
Deaton (1994) - Africa
o Full insurance fails but cons. smoother than income and households cons. moves
together
o Study of Cote dIvoire suggests although co-movement of household cons., full
insurance can be convincingly rejected household income does significantly influence
household cons.
Grimard (1997) Indonesia
o Defined community as ethnic group and found more evidence of insurance could be
related to info failures within village between people of same ethnic group
Other Reliance on savings to self insure
Other mechanisms may include
o Grain inventories (Paxson and Chaudhuri 1994)
o Holdings of cash (Lim and Townsend 1994)
Part IIB Revision, Stephen Ward 40 Development
o Sale and purchase of livestock (Rosenzweig and Wolpin 1993)
For efficient risk sharing, require idiosyncratic risk. Conditions for efficient risk sharing:
Complete mkts. and perfect info
Enforcement possible
Inability to pretend
No moral hazard
Need asset for each state so can insure self against each state
Second best risk sharing
2 households who want to share risk
Incomes observable to each only at cost k
Each household can have high or low income, with probability 0.5
If have same income, no transfer
If one has high and other low, share
2
l h
Y Y +
But since cannot actually know what other has, need to check. Households may say have a low
income when they have a high income. Assume household caught lying pays penalty f
Outcomes
o High-High either or both lie if both lie, no transfer
o Low-Low both tell truth i.e. no transfer
o High-Low person with high may lie - outcome is no transfer
Households always tell truth about low incomes if:
( ) f Y U
Y Y
U
h
l h
>

,
_

+
2
(To stop the high type lying when the other household has low income)

,
_

+
+ > f
Y Y
Y Y U Y U
l h
l h h
2
) ( (if both get high incomes and one person lies)
These both hold if the penalty is high enough, i.e. if:
2
l h
Y Y
f

> this reflects ability to enforce contract
Contract worthwhile if expected payoff to household bigger than expected U without contract i.e. if
) ( 5 . 0 ) ( 5 . 0
2 2
5 . 0 ) ( 25 . 0 ) ( 25 . 0
h l
l h
h l
Y U Y U
k Y Y
U Y U Y U + >

,
_

+
+ +
(if tell truth with arrangement, get high income with prob. 0.25, low income with prob. 0.25 and shared
income with prob. 0.5)
Expect risk sharing to work better where k low (easier to observe incomes of households) and f
high (higher penalty for lying) and where risks high (greater benefits from risk-sharing)
o Tend to be in small communities/networks
Holds when risk is high (Yh Yl high)
Holds when people risk averse
Possibility of contracting also depends on abil. to enforce contracts ex post
Also need penalty for those who refuse to honour agreements (renege or make smaller transfers)
Non-mkt. arrangements to share risk
Credit cooperatives
Interlinkages in agricultural contracts (sharecropping)
Credit arrangements that forgive those with bad shocks
Advantages
Part IIB Revision, Stephen Ward 41 Development
People in small communities know each other well
Peers can monitor behaviour better than formal institutions like banks
Can overcome moral hazard and adverse selection problems
Can enforce (even unwritten) contracts through social sanctions
Part IIB Revision, Stephen Ward 42 Development
Lecture 4B Testing for Risk Sharing in Poor Communities
Saving and cons. smoothing testing risk-sharing within villages in rural India
Interested in saving to understand links between saving and growth
Strong positive correlation between saving and growth
o Saving generate growth or a by-product of it?
Why higher growth accompanied by higher saving?
Saving helps examine how people deal with fluctuations in incomes
People in poor countries suffer large variation in incomes, even over very short periods like a
season
o Because engaged in agric. (returns only after harvest)
o Incomes fluctuate with variations in weather, from natural disasters (common shocks)
o Fall ill (idiosyncratic shocks)
o Particular field invaded by pests (idiosyncratic shocks)
So must insure themselves in come way against variation and keep cons. smooth, and not be
driven to extremities simply because income temporarily low
Can insure and cons. smooth by:
o Saving or borrowing in bad times = self insurance
o Pooling risks with other people note can only do so if risks not common
2 aspects
Keeping cons. smooth over time (like life-cycle hypothesis)
Keeping cons. smooth over different groups of people
o No man is an island in crisis obtain help from other people
Downside shock always worse than upside (bad crop compared to good one)
o Important only works so long as everyone doesnt suffer at same time
o Also called social insurance or group insurance
In any village some people richer than others have better endowments (land) and more assets
Any farmer just as likely to suffer from bad shock and need help from neighbours
Suppose farmers risk averse and willing to sacrifice some of average income for greater
stability
o Each has diminishing MU of cons. so loss of U from bad harvest greater than gain from a
good one
Pareto improving trades to make all better off?
o Agree pool surpluses (and shortfalls) of harvest incomes from indiv. means and make
transfers from pot so every farmer brought up to their mean income
Problems
o Common shocks all suffer shock at once e.g. forest fire or a flood, or disease on crops
o Moral hazard once in arrangement, incentive to not work as hard as income will be
made up to mean whatever
o Limited commitment people may hide certain resources/not be willing to full
participate when have good year
Examples of schemes
o Tithes to churches
o Poor relief funds set up within villages
Part IIB Revision, Stephen Ward 43 Development
Do such insurance schemes exist today and work?
India
o Data on 10 Indian villages by International Crop Research Institute for the Semi-Arid
Tropics
o Data for 10 yrs on incomes, cons., lab. mkts., fin. assets (bank deposits, cash) and
prod. = very good, complete panel
o Morduch uses 3 villages for which data more complete. Uses data on
Total cons. 76/77-81/82
Food cons. alone 76/77-83/84
o First need to check whether most shocks common or whether there is substantial
idiosyncratic variation
Decompose observed income of household i in each period t: Yitv
it itv t tv itv
Y Y
Take logs and estimate as regression treating household and common village
factors as fixed effects
Decomposes incomes into base incomes, common shocks, idiosyncratic shocks,
measurement error and year dummies
Village dummies will capture common shocks
Year dummies will capture seasonal shocks
Remainder is in residual (measurement error and idiosyncratic shocks)
V(C) = variability of cons.
1-R
2
is very high (0.75, 0.96, 0.84)
o Even if this is measurement error, still a lot will be idiosyncratic
Second column: compares two cases: all risk is pooled compared to no risk is pooled.
Third column, based around consumption. Idiosyncratic element is lower in the first village
Case 1: Incomes shared within village
Case 2: Each household depends on its own incomes alone i.e. self insurance
If cons. smoothing impossible
it it
Y C
it
Variance of household cons. ( )
2
1
1


T t
t
i it
Y Y
T
,
i
Y = household sample average
Case 1:
Suppose complete pooling of idiosyncratic shocks, subject to condition that there was no
income redistribution on average (household average still Yi)
Then
it i it
Y C , where Yt is total village income (sum of all indiv. income) in period t and i
reflects share of total income allocated to household
Under these assumptions, share is

T t
t
N i
i
it
T t
t
it
i
Y
T
Y
T
1 1
1
1
1

where N households in village, this is the share i has = av. village income over all times
Part IIB Revision, Stephen Ward 44 Development
Av. village income over
time
Sum of all incomes over all times
So potential for risk sharing here in eliminating total idiosyncratic risk to the household is
( )
( )

T t
t
i it
T t
t
i t
Y Y
T
Y Y
T
1
1
1
1

This is the fraction of consumption over time under risk-sharing to variance under autarky
Results
Lots of idiosyncratic risk
o Social insurance can help reduce this
Test:
Social planner sets out to maximise socially weighted sum of lifetime utilities across households
( )

t T k
k
k t
k
N i
i
t i
C U E Max
0 1

Subject to village resource constraint (standard budget constraint)
( )( )
+

+
+ +
1
1
1
1
1
1
it
N i
i
t it it
N i
i
it
Y r C A A
F.O.C.
( ) ( )
t jt j it i
C U C U ' '
where t is the lagrange multiplier on the village resource constraint the marginal utility of
village income
Test whether rate at which 2 MUs change over time is the same
( )
( )
( )
( )
t
t
jt
jt j
it
it i
C U
C U
C U
C U

1
1
1
'
'
'
'
+
+
+

Assume preferences are constant relative risk aversion (CRRA)
( )
( )
( )

1
1
it
it i
C
C U , ( )

it it i
C C U ' '
and household cons. is adjusted for household size and needs
Note F.O.C. for social planners problem mean
t
t
jt
jt
it
it
C
C
C
C

1
1
1 +

,
_

,
_

Testable result:

,
_

,
_

+ +
t
t
it
it
C
C

1 1
ln
1
ln
for all households and all periods
o Additive so can look at groups of households must be exactly the same for groups
Empirical test:
it
it
it
y t t
it
it
Y
Y
D
C
C
+

,
_

,
_

+ +

1 1
ln ln
o If full risk pooling, the time specific fixed effects (dummy variables for each year)
should explain systematic movements in cons. growth and y = 0
o If the model is completely wrong, y = 1
o The test shows
The majority of coefficients are significantly different from zero
Part IIB Revision, Stephen Ward 45 Development
The R-squared is higher for food consumption regressions, suggesting that food
consumption is better insured
Conclusions
Little evidence for risk-sharing of total consumption within villages
If there is any risk sharing at all, seems confined to food i.e. subsistence
Calls into question existence of a social welfare function that incorporates individual utilities
as arguments (social planner problem)
Social arrangements are limited to guaranteeing subsistence
Part IIB Revision, Stephen Ward 46 Development
Lecture 5A Investment in Human Capital
Why important?
Intrinsic value being educated and healthy valuable
Instrumental value
o Affects personal opportunities abil. to work, get a job, expand income, Sens terms
achieve larger set of functionings
o Social roles greater literacy and basic educ. can facilitate public discussion of social
needs, expand public services in directions people want
o Process of investment draw kids away from lab., expand range of people kids meet,
broaden horizons, discipline (punctuality)
o Re-dist. and empowerment roles
Help disadvantaged organise politically and get fair share
Affects dist. within family girls empowered by educ. to resist inequitable
allocations of resources within household for themselves and kids
Evidence on educ.
Returns to primary educ. highest concentrate investment here
Returns marginally higher for women
Example from Green revolution (Foster and Rosenzweig 1996) returns to primary educ.
increased in areas with new tech. nut not elsewhere
Rapid increases in educ. not produced growth several instances of stagnation in face of large
public investment
Taiwan increase supply of educated (compulsory educ. raised from 6 to 9 yrs) coincided with
demand
Conclusion necessary but not sufficient input for growth?
Barro (1991) schooling and growth correlated
Extra yr schooling in 1960 associated with 0.6% faster growth in GDP between 1960-80
o With linear production tech., increases in schooling raise growth. In Solow traditional
differences in human K explain temporary differences in growth rates
o Higher growth increases investment in schooling
Expected future growth lowers effective discount rates and increases
investment in schooling (schooling sacrifices income today for income
tomorrow and rising growth increases potential gains)
Endogenous Growth
Suppose learning today depends on past stock. If schooling not subject to diminishing returns
then qual. of educ. can grow relative to past
Growth in TFP not due to increases in quantity of schooling
o If schooling causes growth, must come largely from rising qual.
Empirical analysis predicts if rising school qual. raises growth, young cohorts have better
schooling than older cohorts and wage disadvantage is falling
Bils-Klenow (2000) find schooling to growth can explain only less than half Barros coefficient. Suggests
causality runs from growth to schooling
Enormous disparities in educ. outcomes across world, across regions in same country, by sex
within countries, by income levels etc.
MODEL: INVESTMENT IN HUMAN K
Part IIB Revision, Stephen Ward 47 Development
Parents make decisions for their human K (educ., allocation of food)
U function as function of schooling and earnings of child when he/she grows up is:
( ) ( ) ( ) S h y m S y U ln ,
S = schooling, h = cost of schooling, m < 1
Earnings of child when grown up are
( ) bS a y + ln
b = economic returns to education, and is linear and constant
MC of educ. can be written as
( ) S r s h + '
Familys problem:
Integrating above, ( )
2
2
S
rS s h

+
Maximising U gives
( )

r mb
S
So optimal choice of S depends on
o Share of childrens earnings parents can get, m (links to idea children as investment
good)
o Marginal return to educ., b
o MC of each yr schooling,
See The Probe Report for what determines these
Myths about education in India
Parents not interested. D for educ. low
o But 80% would like compulsory primary educ.
Child lab. is main obstacle
o But only small proportion of kids in full time work (3-10% of kids age 5-14)
W lab smallest part of this farm lab. or lab. at home far more important
More likely children work because they do not go to school
Elementary education free
o But direct costs, opportunity costs
Schools available but
o Access many schools deny admission to those from lower castes (teacher may be
upper caste)
o Returns to investment parents may not wish to invest in educ. if no schools available
for higher levels of educ. i.e. upper primary which offers educ. until grade 8
o Qual. and returns a poorly equipped school which offers little teaching is a waste of
time
Returns to educ. low this is not true
Externalities to education
Illiterate workers lab. earnings affected by educ. of others in same household (Basu, Narayan,
Ravallion 1998)
o Suppose household chooses cons. levels collectively, by scheme of efficient bargaining.
Would literate members of household help illiterate members?
Positive income effects of sharing literacy through more earnings for illiterate
Negative effects on composition of consumption illiterate members earn
more, perhaps affecting their bargaining power and changing allocation within
household
Household has 2 members, one literate and other illiterate. U functions UL and UO
Let x be vector of goods (assume 2 with prices p1 and p2) consumed by the household
Part IIB Revision, Stephen Ward 48 Development
Assume household has hhd welfare function where pareto weight is function of incomes
earned by each member
Assume yO and yL are income and yL constant so that ( ) ( )
O L O
y y y ,
( ) ( ) 1 , 0
O
y and ( ) 0 '
O
y
Household chooses x to max
( ) ( ) ( ) ( ) ( ) x U y x U y
L O O O
+ 1
s.t.
O L
y y px +
Solution is ) , ( p y x x
O
note: x also depends on yL and p which are constant
In period 1 (before problem above opens), literate member has option of sharing literacy with
others i.e. illiterate gain, and will if
o Have same preferences over cons. (Members have similar preferences)
o Bargaining weight independent of yo (Bargaining power not greatly affected by rising
incomes for illiterate)
In period 1, literate member chooses between setting yo = yP or yI where first is income if you
are proximate to literate person and other if youre isolated, where yP yI
o Set yo = yP iff ( ) ( ) ( ) ( )
I L P L
y x U y x U where ( )
P
y x and ( )
I
y x are solutions to
the maximisation problem in each case
o Claim 1: If ( ) ( ) x U x U
L O
then literate member will not exclude illiterate
member from benefits i.e. if have similar preferences, literature shared
in household. In this case the maximisation problem is
( ) ( ) ( ) ( ) ( ) ( ) x U x U y x U y
L L O O O
+ 1 max
s.t.
O L
y y x p x p px + +
2 2 1 1
o So ( ) ( ) x U x U
L O
for all x such that ( )
P
y px px
Since utility monotonic in goods, ( )
P L P
y y y px + and ( )
I L I
y y y px +
Hence ( ) ( ) ( ) ( )
I L P L
y x U y x U
o Claim 2: If ( )
O O
y y then F.O.C.
( )
( )
( )
( )
( )
( )
2
1
2 2
1 1
1
1
p
p
x
x U
x
x U
x
x U
x
x U
L O
L O



and
2 2 1 1
x p x p y y
O L
+ +
Conclusions
Literacy has positive externalities for household members
Illiterate women more efficient recipients of benefits from literate members
o Better able to absorb literacy and more likely to gain in income
Unitary model each funds own cons. (no incentive to share literacy) or family share income and
jointly decide cons. (incentive to share skills in order to increase wages)
Bargaining effect and income effect
o If preferences of illiterate and literate similar
Part IIB Revision, Stephen Ward 49 Development
o If increased income does not increase the bargaining power significantly
Only one of these need hold for effect of literate household member to be positive on the
illiterate member i.e. expect positive effect of literacy of another member of household on the
wage
Bigger effect for women
o Due to lower bargaining power effect for them?
o Women dont have access to further external sources of literacy?
Cultural and religious sanctions on womens social activities may mean they
have harder time than men accessing sources of help outside family e.g. paying
for services of scribe
o Like to devote time and income to kids?
Father prefers to share knowledge with mother
o Problem of pride in males too proud to learn
Empirical concerns
Reverse causation high earnings lead to high education
o High education more likely to attract educated spouse
o Works mostly through marriage remove endogeneity problem by looking at the
unmarried
Unmarried women likely to continue living with parents who help
o Male more likely to move out
Maybe from having money leads to literacy as opposed to vice versa endogeneity problem
maybe 1
st
results for male not unbiased were wrong
o Maybe no significant effect at all for men
Effect for women becomes greater
o Employers prefer someone not in family? Free and unmarried?
Can only be estimated for those in wage work
o Generic problem not just this question holds for any W regression in any country
o Doesnt look at participation decision
o Self selected sample? Those who see benefits of literacy work whereas others dont?
Sample Selection Bias when do regression, estimate for those in job mkt. and
receive wage. Ignore those out of mkt. if characteristics of in and out of job
mkt. similar, no problem as random sample of pop.
If factors determine entry into job mkt. and differs between group if this also
helps determine W leads to omitted variable bias
If variable observable, can deal with and include in regression to control for
specifically e.g. abil. to work or motivation to work (unobservable)
Heckman method probit model to determine p(participate in mkt.) correct
estimate for W becomes unbiased
Part IIB Revision, Stephen Ward 50 Development
Lecture 5B Evaluating Externalities in Education
Teacher Shocks and Student Learning Evidence from Zambia
Investigate impact of shocks to teacher inputs on learning achievement
o Use info on teacher absenteeism and child test scores from panel data
o In context of high absenteeism linked to health in HIV/AIDS affected Zambia
Where spending on teachers is 80%+ of total education spending
Boosting completion and quality of primary education is essential for pro-poor growth
Within Zambia, enrolment and completion success rates (e.g. 80% + NER) under threat
o World bank wanted to increase spending
o Zambia had large fall in income
Copper main produce P collapsed and so did the govt. budget and health
and education spending
o Where is it appropriate to ask where is all the education gone?
Empirical Evidence
Hanushek (US), Park (China), Rockoff (US)
o Significant teacher effects while much less clear evidence what else matters
Unpacking teacher effects role of absenteeism
o Ehrenberg School pass rates no clear effect
o LDC evidence
By Probe in India teachers often absent/teaching poorly
Kremer a lot of it (20%+ in Africa, 25%+ India), but impact?
45% down to 20% fall in absenteeism from camera experiment (photo
with kids, date and time, at start and end of day
Also found 43% extra chance primary kids go on to secondary with this
o Impact likely to be higher if
Substitute teaching not available
Poor health is cause
Approach
Matched panel of test score and teacher inputs from 2002 in 182 schools, 2000+ kids
Set of papers on impact of different inputs
o Impact of public expenditure on learning achievement
o Impact of teachers on learning
Explicit household optimisation framework where parents care about educ., and spend
optimally in response to teacher and school inputs
Limited basic objective quantify impact of teacher input shocks
Findings
High teacher absenteeism 42% of teachers 1 day or more absent in last 30 days
Limited learning gains (e.g. from 38% of maths questions right to 45% in 1 yr)
More than 60% absences linked to illness (own + family) or funerals
Child with 5% more absent teacher has 10% lower learning
Literature
Most tries to estimate school prod. function
o But prod. also involves parental inputs that respond to changes in school inputs
Todd-Wolpin (2003) identify prod. function beyond policy effect
Approach cont.
Part IIB Revision, Stephen Ward 51 Development
Here use demand approach, explicitly accounting for responses to pol. variables and other
anticipated changes
Households respond to info., shifting resources into/out of education, as one option for
spending (given deep poverty), up to optimal path substitute for school input is smooth path
of achievement
If some (technical) substitution between household and school inputs, overall impact of
anticipated increase in school inputs lower than unanticipated increases
o E.g. increases in funding in Das, Dercon et al. (2004)
Responses to increased uncertainty
o E.g. if households face more uncertainty regarding teacher quality, relatively more
spending by parents likely (given precautionary motive)
Theoretical Model
Households maximise
( ) ( ) [ ]
t t
T
t
t
t
X v TS u E +

where TS = test scores, X = other goods


Subject to
( ) ( )
t t t t t t
z X P y A r A + +
+
1
1
where A = initial assets, y = income, z = private school inputs, r = interest rate (price school
inputs numeraire)
and the production function
( ) , , , ,
1
s z m TS F TS
t t t t

where m = teacher inputs, s = fixed school/teacher, = fixed child characteristics and (Fm > 0,
Fmm < 0)
Let teacher inputs consist of 2 parts
o Quality + shocks
o
o Both unknown to parent
Suppose can buy cognitive achievement at price v, in period 1 to but 1 unit. In next period,
could sell (1 ) units (depreciation) and get current value =
( )
r
v
+

1
1
User Cost = cost of holding 1 unit of test scores for 1 period =
( )
r
v
v
+

1
1
With prod. function F, households buy test scores in period t by increasing Zt
Interested in cost of boosting achievement in one period only assume next period reduce zt+1
to ensure overall stock of test scores at t + 1 remains unchanged
Similar to
( )
r
v
v
+

1
1
above, the first term measures cost of taking resources at t and
transforming it into one extra unit of cognitive achievement
( )
( )
( ) ( ) . 1
.
.
1
1 +
+

t
t
t
z
TS
z
t
F r
F
F

When implemented through prod. function, P no longer constant


o If prod. function concave, the higher the initial levels of cognitive achievement, the
greater the cost of buying an extra unit as reflected in marginal value
( ) .
t
z
F
Part IIB Revision, Stephen Ward 52 Development
P of good
MP of spending money on child
Of additional unit bought in period t, amount left to sell in period t + 1 is
( ) .
t
TS
F
Second term thus measures the present value of how much of this one unit
will be left in next period expressed in monetary terms
Define user cost or rental P of boosting TS for one period only
( )
( )
( ) ( ) . 1
.
.
1
1
+

t
t
t
z
TS
z
t
F r
F
F

o Yielding optimal conditions, equating marginal utilities of boosting stock over time,
appropriately priced

,
_

+
+
1
1
t
t
t
t
t
TS
U
E
TS
U

o Leads to standard prediction


Any unanticipated change in user costs, e.g. due to shock to teacher inputs,
leads to shock in optimal path (contrary to anticipated changes)
Empirical Model
Path of test-scores determined by path of user costs and shocks, including teacher
absences/illnesses or shocks to quality = core of test
ijkt jkt jkt jkt
ijkt
ijkt
X t w
TS
TS
+ + + +

,
_

3 2 1 0
1
ln
o Test score of child i with teacher j in school k at t
o w = measure of shocks to teacher inputs
o t measures observable teacher qual. changes
o x = other variables affecting user costs (controls)
More negative 1 = larger impact of shocks
Measure shocks to teacher inputs i.e. absenteeism, using matched data on test scores and
teachers (grade 5 kids)
Econometric issues
o Assume ( ) 0 ,
ijkt jkt
w corr , but this means all changes of teachers qual. measured or
at least uncorrelated with shocks
o If ( ) 0 ,
ijkt
G
t
w m corr , 1 would be biased
o Can use rule in Zambian educational system that allows us to identify 1 at least from
sub-sample
For those staying with same teachers, 1 can be identified. For others possibly
promlematic
wjkt interacted with whether mover or stayer
Data
182 schools in Zambia, 2002
o Education and economy in crisis, teacher attrition, large class sizes, double shifts
Learning achievement
o Tests in Maths and English 5
th
grade at t
o Same sample of children, 1yr apart
o Small learning gains observed (38% to 45% maths questions right in 1 yr
Detailed data on schools, teachers sub-sample matched kids and parental characteristics
Controls for change in teacher qual.
Part IIB Revision, Stephen Ward 53 Development
o Teacher characteristics
o Sub-sample teachers staying with kids so no change in qual. for these kids used to
identify consistency impact of shocks
Absentees based on head teachers reports 42% of teachers in last month absent for at least
one day
Results
Focus on English results Maths qualitatively the same
Controls teacher characteristics (sex, experience, certificate etc.) and school characteristics
(funding, location, head teacher characteristics, PTA fees, priv. status)
Shocks matter (for maths, even without controls)
Significant and negative for non-movers (for both Maths and English)
Fact that qual. changes controlled for seems relevant
Size of effect: 5% more absent teacher results in 10% lower learning gains in period
NB. For sub-sample, have child illness no change in coefficient on teachers
Coefficients on non-movers*days absent last month v. significant. Non interacted not significant
o Impact on non-movers strongly negative. Coefficient on movers is larger. How
interpret?
Econometric reasons
Selective matching
o Unobserved qual. or qual. changes correlated with absenteeism
but then absenteeism positively correlated with higher quality
teacher (makes no sense better teachers more absent?)
o No evidence on such positive correlation on observables
Sample differences
o Movers and stayers quite different. Stayers = more urban,
wealthier
Schools better able to respond but effect of shocks
expected smaller for them
Schools better able to anticipate shocks again
expected effect smaller
o Formal test regression based on propensity matched sample
of movers (based on child and school characteristics); no effect
on coefficients found
Behavioural reasons (theoretical)
Households respond to info on schools/teachers in terms of spending
(compensating)
E.g. Evidence on household responses to anticipated increases in
funding in Das, Dercon et al. (2004)
Responses to increased uncertainty e.g. if households face more
uncertainty regarding teacher qual., relatively more spending by
parents likely
Suppose differences in uncertainty about some part of teacher inputs, teacher qual. . Will
this affect spending ex-ante?
o See optimal rule spending on educ. vs. other goods

,
_

t
t
t
t
t
TS
U
E
P
X
U

1
Impact depends on whether RHS increases if risk increases i.e. then more spending on
educ. and less on X
Part IIB Revision, Stephen Ward 54 Development
Education spending will increase if risk increases if
t
t
TS
U

is decreasing and convex in teacher


inputs
Sufficient conditions
o Household and teacher inputs technical substitutes
o Risk aversion
o Convex MU
o User costs increase in m at decreasing rate
More risk taken on if risk increases? Similar to precautionary savings
If less risk, less spending but then impact of other shocks higher (Fmm < 0)
Note stayers had no or less uncertainty about teacher (=rule), so predicted impact of shocks
more negative, or observed impact consistent with precautionary spending
Conclusions
Shocks matter a lot 5% more absenteeism gives 10% lower learning gains
Absenteeism linked with illness another piece of high costs of HIV/AIDS epidemic in Africa
Policy: health as well as cushion (sub. For teachers)
Effect for stayers seems well identified
Differences in results between movers and non-movers not easily explained using econometric
reasons and consistent with behavioural response related to precautionary educational
spending
Part IIB Revision, Stephen Ward 55 Development
Lecture 6A Physical Capital: Property Rights, Incentives and
Growth
Stable property rights important part of well functioning economy
o Many schemes in developing countries to try and improve such rights
Difficulty of establishing and enforcing property rights is the weakness of political and legal
systems
That property rights make a difference to investment often regarded as self evident
o Many channels where can work much beyond expropriation argument frequently put
forward
De Soto Agenda
Hernando De Soto one of main advocates of view that improvements in property rights needed
to improve economic performance
o Lots of weight on view that this intimately linked to collaterisability of assets and
improvements in credit mkts.
MODEL: PROPERTY RIGHTS (BESLEY)
kt = investment at time t
Rt+1 = property rights at period (t+1) when returns from investment realised
Returns/payoff to investment at time (t+1) ( )
1
,
+

t t
R k V
o V concave and increasing in k and R
Costs of investment ( )
1
,
+

t t
R k C
o C increasing in k and non-increasing in R
Max profits
( ) { } ( ) ( )
1 1 1
, , , max
+ + +

t t t t t t
k
R k C R k V R k W
t
At optimum k
*
( ) 0 ,
1

+

t k
R k W
When g(x,y) = 0, provided continuity and non-zero commodity. Use implicit function theorem:
y g
x g
x
g
g
y
dx
dy

If Wkk 0 then there exists a function k


*
such that it solves the above and
kk
kR
k
k
t
W
W
k W
R W
R
k



1
By concavity of profit function, 0 <
kk
W at a maximum
o Implies that investment increases as rights are improved if 0 >
kR
W
3 cases and test if 0 >
kR
W in each, and under what conditions
o Freedom from Expropriation
Security of rights: model probability of expropriation as decreasing function of
R
In each period, some chance indiv. will have land expropriated and prob. this
happen is decreasing function of rights he enjoys
Models property right insecurity like random tax on land
Part IIB Revision, Stephen Ward 56 Development
Seems most relevant for situations where rule of law broken down and
indiv. able to appropriate each others assets or that state seizes
indiv.s property after a revolution
May also be relevant for Lat Am. Where squatters who gained some
rights to land through prolonged residence face threat of eviction
Risk of Expropriation insecurity like random tax on returns of investment
Level effect overall reduction in investment
Composition effect invest in assets less subject to expropriation
invest in more portable assets (e.g. livestock)
o Access to Credit (land as collateral)
If land easier to use as collateral, banks offer lower interest rates (marginal
return to investment), so investment rises
If credit rationed, better abil. to use land increases amount credit available
(Feder et al 1988)
Link from land rights to collateral to lending to investment
Assume
Improvements in land rights encourage lenders to recognise land as
collateral
Borrowers dependent on access to credit for investment in land
Then 2 ways land rights can be linked to investment
Competitive credit mkt. without problems, improved access to
collateral reduces risk premium on lending, hence reduces interest rate
faced by borrower
With credit mkt. imperfections, collateral can reduce agency problems
and can improve access to credit (e.g. reduce credit rationing)
Key assumption that better land rights lower foreclosure cost
Improved rights reduce equilib. Interest rate and since this set equal to
MPK invested in land, investment stimulated
View implies land rights at household level rather than field level matter to
investment decisions
No reason why farmer cannot collateralise any of his fields to invest in
another = important observation used in empirical analysis
No reason to find link between land rights on particular field and
investment on that field when farmer has more than one field some
measure of rights at household measure should matter
o Gains from Trade
Improved rights to transfer land (rent etc.) encourages investment
Better rights, leading to expanded trading opportunities and abil. to exploit
gains from trade, enhance investment incentives
Gains from trade may be realised by institutions where land is disposed of to
someone with better use for it (standard comparative advantage argument)
Incomplete property rights impair trade in land, particularly for rights to sell or
rent land, improving possibilities for trade in land can improve investment
incentives
So when gains from trade raise marginal returns to land
Superior transfer rights modelled as lowering costs of exchange if land either
rented or sold
Desire to trade generated by stochastic outside opportunities such as
offer to buy/rent land from current owner
Gains from trade may only be realised institutions where land disposed of to
someone who has better use for it (standard comp. ad. Argument)
Expropriation
Part IIB Revision, Stephen Ward 57 Development
denotes physical return from investment
Probability of expropriation is
( )
1 + t
R with ( ) 0 '
1
<
+ t
R
Then ( ) ( ) [ ] ( )
t t t t
k F R R k V
1 1
1 ,
+ +

So ( ) ( ) 0 ' '
1
>
+ t t kR
k F R V and assume costs independent of
1 + t
R

kR kR kR
C V W
Which implies 0 >
kR
W if costs of investment are linear
1 + t
R for instance (MC independent of
R)
Collateral
Stochastic return to land
o with probability q
o 0 with probability 1 q
Loans backed by collateral and lenders can access collateral at cost ( )
1 + t
R where
( ) 0 '
1
<
+ t
R
Assume by social convention, cannot ruin borrower his utility cannot fall below u
Bt = amount borrowed at interest rate rt
Expected utility is
( ) ( ) ( ) ( ) u q b r k qu k b y u
t t t t t t
+ + + 1
o Farmer chooses (bt,kt) to maximise expected utility
o Implies ( )
t
r k

' i.e. MPk = interest rate


o And ( )
t t
r g b and ( ) 0 ' <
t
r g
F.O.C. give:
For kt:
( )
( ) ( )
( )
t
t t t
t t t
k q
b r k u
k b y u
'
'
'

+
For bt:
( )
( ) ( )
t
t t t
t t t
qr
b r k u
k b y u

+
'
'
Loan function: ( ) y r g
t
, which describes amount farmer would wish to borrow as function of
interest rate and income
= lenders OC of funds (mkt. P of land), P
*
= value of land after default, x is level of cons that
gives min. U, u
Lenders Profits
( ) ( ) ( ) ( ) ( ) ( ) [ ] y r pg x P R q y r g r q R r
t t t t t
, 1 , ,
*
1 1
+
+ +

Equilibrium interest rate is given by
( ) C R r r
t

+1
, :
where C is fixed costs of entering credit mkt. ( ) C R
t
<
+1
, i.e. lending at OC will be
unprofitable
o must be increasing in but also increases with since it affects the amount that
the lender can recover in the event of default
o Increasing reduces and hence increases optimal investment (since optimal
investment decreasing in )
Part IIB Revision, Stephen Ward 58 Development
Gains from Trade
Investment in period 1
o At beginning of period there is a shock that affects productivity of owner if he
continues to operate the land
Productivity of K is where [ ] ,
Outside option has distortion of returns, g()
Cost of trading is with
Nash bargaining determines price P at which land traded

( ) [ ]( ) { } p k k p
p
+ max
o Then ( ) ( ) ( )
1
]
1

'

+
+ + 1 1
2
1
, max ,
t t
R E R k V
o ( ) ( ) ( ) ( ) 0 '
1 1
>
+

t t kR
R d g R F V

So investment is increasing in property rights


Trade takes place when buyers valuation exceeds sellers by sufficient amount more trade
happens when property rights better
Endogenous rights: ( )
t t
k R
+1
then even greater incentive to invest
Why investment low in Africa?
Main distortions in product and credit mkts.
High risk appears to be biggest deterrent to investment
Where contracts can be designed to ensure repeat contracting, risk sharing and investment are
higher
Idiosyncratic risk very high relative to community risk e.g. aggregate P volatility understates P
volatility to indiv. since mkts. fragmented
Financial depth poor informal credit mkt. limited
o Low investment coexists with high returns (true if high risk deters investment)
o Unstable macro pol. environment
Evidence from firm level surveys
Investment in manufacturing (CSAE, Oxford University)
o Median investment rate is zero in Cameroon, Ghana, Kenya and Zimbabwe
o Low investment, high profit rates and low growth of value-added co-exist
o Only Kenya saw sustained per cap. income growth over 20 yrs
o All countries had high and variable inflation rates
o All countries had large changes in nominal interest and exchange rates
o Small firms financially constrained but high capital costs biggest problem across firms
High K costs associated with high uncertainty and high risk constrains investment
Availability of finance is not main constraint
Agricultural organisation
Key place in developing countries
UNDP, 1996 agriculture employed
o 60% of LF while contributing 20% GDP of LDCs
Implies agric. In LDCs rel. less productive w.r.t. non-agric.
o 2% of LF while contributing 2% GDP of DCs
Implies agric. In LDCs also rel. less productive w.r.t. agric. in DCs
Part IIB Revision, Stephen Ward 59 Development
o Yet provides livelihood of majority of people in LDCs
Backward tech., infrastructure, bad policies = reasons (but they affect other sectors too)
Missing mkts., transaction costs, inefficiencies of agric. organisation = other reasons
Stylised facts
o Small farms more prod. than large farms inverse farm-size productivity relationship
(Berry & Cline 1979)
o Sharecropping important form of agric. org., even though less productive than owner-
cultivation or fixed rate tenancy
Suggest land reform, other than promoting equity, can also raise productivity
o some evidence of this (tenancy reform policies)
An explanation for first fact is diminishing returns to land
o But land mkts. should take care of it
Explanation for 2
nd
fact is agency costs
o Again why does land mkt. not get rid of these inefficiencies?
Key Stylised Facts
Small farms more productive than large farms inverse farm-size productivity relationship
(Berry and Cline 1979)
o Explanation is diminishing returns to land but land mkt. should take care of
Sharecropping important form of agric. org., even though less productive than owner-
cultivation or fixed rate tenancy
o Explanation is agency costs but again land mkt. should get rid of inefficiencies
Suggests land reform, other than promoting equity, can also raise productivity
o Some evidence that land reform and tenancy reform policies improved productivity
Endowments matter only with mkt. imperfections
Consider producer
o Values cons. and leisure u(c,l)
o Has some land and labour endowments T L, , and prod. technology ( ) T L f y ,
o Wage rate w and rental rate r
o Profits are
o
( ) l c u
l c
, max
,
s.t. ( ) T r l L w c + +
or ( ) T r L w rT wL T L f wl c + + + ,
o Shows in choice of L and T, preferences or endowments do not matter
o Could hire in labour or hire out. Same for land
Separation of profit max. behaviour as prod. and U max. behaviour as cons.
If breaks down, farm with lower will have different productivity than farms with large
With frictionless mkts., factors efficiently allocated and farm sizes adjust endogenously
Why frictions in land or lab. mkts.?
o Agency costs (from informational problems) and transaction costs (from problem of
commitment and enforcement)
o Suggests viewing agric. organisation as response to deal with these problems
Key questions:
o What drives choice of agric. org./contracts?
o Does it affect productivity?
o If it does, why doesnt everyone choose the most efficient organisation?
Part IIB Revision, Stephen Ward 60 Development
Lecture 6B Property rights and Investment Decisions
Why might property rights matter for investment?
Freedom from expropriation
Access to credit (land as collateral)
Gains from Trade improved rights to transfer land (rent etc.) encourages investment
Model of investment (profits re-invested) regressed on property rights index at the level of the firm and
a vector of other variables that may affect re-investment
One regression for all firms, one for spin-offs from existing firms
Since investment doesnt come from banks expropriation not really working here?
Spin-off firms known by others PR kind of pre-established
o If assume spin-off related to parents, PR just look like parents PR
Index coefficient much lower for spin-offs. This could be the source of the
insignificance
If expected significant variable and in fact not significant in regression, one of 1
st
things to do
is check variation in the variable
Parents control decisions of spin-off firms - firm doesnt decide level of investment pay out
excess profit as dividend rather than re-invest?
o Relationship between variables broken here
Potentially endogeneity
o Reverse causality share of profits re-invested could have causal effect on property
rights
o Omitted variable bias some unobserved variables?
Investment ability, shareholder attitude
Deal with using fixed effects dummy variable for each household to capture
the unobserved factors
PR = subjective index highly subject to measurement error
Use IV past rights? Removes reverse causality to higher investment but still measurement
error
Use 4 different variables length of time of firm? Number times gone to court about firm?
o Use instruments and regress to see if results different or not
Lack info on failed firms not reported
Sample of all firms consistent with theories whereas sample of spin-offs does not show PR
enhances investment
Hard to judge which one significant out of 3 models of expropriation (Freedom from
expropriation, access to credit or gains from trade)
o To test for collateral something in level of household
Must create PR at this level and test for it
Hard to say much about collateral here not related to credit from banks
Model:
Decision to invest kt on a field
o Returns to investment at time (t + 1) = ( )
1
,
+ t t
R k V where R denotes property rights
V concave and increasing in k and R
o Costs of investment = ( )
1
,
+ t t
R k C
Part IIB Revision, Stephen Ward 61 Development
C increasing in k and non-increasing in R
o
( ) { } ( ) ( )
1 1 1
, , ,
+ + +

t t t t t t
k
R k C R k V R k W Max
t
3 cases
Security of rights model prob. of expropriation as decreasing function of R
o Applies to each plot individually if household has many plots, there will be positive
correlation between plot land rights and investment
Credit and Collateral if land easier to use as collateral, banks offer lower interest rates =
marginal return to investment so investment rises or if credit rationed, better abil. to use land
increases credit available
o Must apply at level of household no reason why plot pledged as collateral is same plot
invested in
Gains from trade better transfer rights make it less costly to organise trades
o Or lower transaction costs in selling and renting improve investment
o Must hold at plot level should suggest at greater level
o Implies connections between rights and investment are at plot level but is rank order
to rights
Data
With fully codified land rights, cant investigate sub-Saharan Africa useful to investigate
o Urban slums too good to measure the link
2 regions in Ghana
Wassa
In west-central Ghana
o Data on 10 villages, 217 households and 1074 plots of land
Grows cocoa and only significant investment is in tree crops (takes time to grow - every 2 yrs
for example)
o 66% have increased investment in tree crops
Land traditionally controlled by tribal authority but now more individualistic
Anloga
Data on 5 villages, 117 households and 494 plots of land
Land improvements many and diverse grow shallots on small fields
Continuous manuring, draining, excavating land, irrigating etc.
Most land rights do not require lineage approval de facto individualistic
Data on sex and age composition of household, some measures of durable wealth
Use and transfer rights on every plot operated:
o Right to sell
o Right to rent out
o Right to bequeath
o Right to pledge as collateral
o Right to mortgage
o Right to gift
Each right takes 2 forms
o Right to do without lineage approval
o Right to do with lineage approval
Gives 12 dummy variables measuring rights
In general, investment in plot j by household i at time t
k = investment, z = household characteristics, x = plot characteristics, R = future property
rights
Part IIB Revision, Stephen Ward 62 Development
( )
ijt it t ij ijt
x z R f k , ,
1 , +

Problem 1: endogeneity if property rights can be created through investment


( )
ijt it t ij ijt t ij
x z R k g R , , ,
, 1 ,

+
o Leads to overestimate of rights on investment (if higher investment raises rights in
future i.e. positive correlation between errors and future rights)
Problem 2: R measured with error
o Correlation from past rights to present rights
Not measured particularly well - measured with error
o Classic measurement error biases coefficient towards zero
Both need IV methods
Empirical Model
Dependent variable = Investment on field j owned by farmer i at time t
o 1
ijk
y if investment type k by household i on field j - = dummy variable
Explanatory variables - land rights in future periods , measured characteristics of the plot
and household / farmer characteristics
ijk ij k ij k i k ijk
e x R z y + + +
(1)
household fixed effect
ijk ij k ij k ki ijk
e x R y + + +
(2)
This equation estimated for each region
Problem: household fixed effects cannot test hypothesis 2 collateral based story
Problem 1 future land rights may be endogenous to current investment need instrument
o Use past / current land rights as instrument
Problem 2 measurement error: formal transfer rights declared do not capture rights that
farmers think matter for investment
o Use instruments use if land gifted, bequeathed etc. i.e. past land rights but special
types of
Problem 3 measured rights proxy for omitted ability of farmers leading to spurious link
between rights and investment
o Solution is to use household fixed effects
o Has its own problems difficult to distinguish effect of average household-land-rights
as in collateral arguments
Results
Equations (1) and (2) estimated with and without instrumentation for land rights
Instruments
o Whether there is a transfer deed for plot; whether household has ever sued / been
sued over the plot; how field was acquired; how long its been owned
o Good instruments?
Tell story well need a narrative
Formal test F test for instruments must be significant
Wassa
o Land rights matter for investment in trees; effects vary by how strong rights are (with
approval effect is 2.5%-11%, without approval 2%-12%)
o Both OLS and IV using sum of rights (with and without lineage approval) suggest link
from rights to investment but effect is weak
o IV estimates higher than OLS suggests measurement error in rights matters more then
endogeneity (biases downwards)
Analoga
Part IIB Revision, Stephen Ward 63 Development
o 6 different types of investment considered: drainage; manuring; land excavation;
irrigation; mulching; making shallot beds
o Positive and significant effect of rights on investment without instrumenting
o Collapses after introducing household level fixed effects or instrumenting
Suggests strong endogeneity
Results shaky and little can be learned from them
How test between models?
Collateral based model suggests household rights rather than field specific rights ought to
matter
If rights variable significant even after controlling for household fixed effects, no real support
o In Wassa, this so plot specific rights seem to matter
o Plays with Anloga support for field/plot level rights
Is there hierarchy of rights?
o Gains from trade argument suggests rights to sell & rent matter most
o Land rights decomposed into 6 categories
In Anloga, rights to sell & bequeath highly correlated (cannot tell right to sell different to right
to pledge) In Wassa, not so
o If rights to sell and rent jointly insignificant, ev. against this hypothesis
o However, dont seem to matter for investment directly, which is odd
Probably argue against credit link but cannot really favour one of the others (cannot test H2
collateral based link due to problems with data measurement at household level)
Part IIB Revision, Stephen Ward 64 Development
Lecture 7A Political Economy
Do natural resources reduce growth? Yes, they could
Poor governance
Dutch Disease
o 1960s, Netherlands experienced vast increase in wealth after discovering large natural
gas deposits in North Sea unexpectedly was disaster for traditional Dutch X
o Dutch guilder became stronger, making Dutch non-oil X less competitive
o U rose sharply
o Corden & Neary (1982) consider economy with 3 sectors
2 traded goods sectors booming X sector and lagging X sector
Non-traded goods sector supplies domestic residents and includes retail trade,
services and construction
When country catches Dutch disease, traditional X sector gets crowded out by
other 2 sectors
Suppose country discovers oil
o Increase in oil X initially increases incomes as more foreign
exchange flows in
o If spent entirely on imports, no direct effect on money supply,
or demand for domestically produced goods
o Suppose foreign currency converted into local currency and
spent on domestic non-traded goods
Fixed ER conversion of foreign into local currency increases money
supply. Pressure from domestic demand increases domestic prices. An
appreciation of the real ER
Flexible ER increased S of foreign currency increases value of
domestic currency, nominal ER rises and an appreciation in the real ER
In both cases, real ER appreciation weakens competitiveness of countrys X
Traditional X sector shrinks = spending effect
Resources (K & lab.) shift into prod. of domestic non-traded goods to
meet the increase in domestic D and into the booming oil sector
Both these transfers lower prod. in now lagging traditional X sector
= resource movement effect
Pol. responses depend on whether rise in inflows due to oil discovery temp. or
perm.
Democracies better for growth when natural resources discovered? (Collier and Hoeffler 2006)
Democracies handle resource rents better?
Or democracies with poor checks and balances do worse with resource rents?
Better
Autocrats may be particularly predatory in presence of resource rents
o Autocrat as stationary bandit must limit predation to sustainable rate
o Stationary bandit invests in order to increase returns by capture
Resource wealth different no need to invest atypically immobile and highly taxable, natural
resource rents reduce need for autocrat to invest in growth of non-resource priv. economy
Well functioning democracy
o High accountability of public spending to the priorities of citizens
o Pay-off to democracy proportional to share of public spending in GDP
Part IIB Revision, Stephen Ward 65 Development
So increases resource rents
Worse
What politicians do with public money depends on restraints in place resources can be used
either for provision of public goods or for priv. patronage
o If democracy functions well, politicians who corrupt will be defeated and perhaps put
on trial for fraud etc.
What do politicians do in democracy?
Persuade voters to vote for them by
o Offering programme of public goods
o Directly buy votes usually infeasible as
Contracts with voters non-enforceable due to secret ballots (functioning
electoral system)
Bribe transactions liable to criminal prosecution (functioning judiciary)
Antagonise un-bribable voters (either social norm or disgruntlement that not
bribed)
Too expensive to be affordable by all political parties (consequence of above)
If vote-buying cost-effective it will predominate unless restrained by lack of finance
o Can a lot of public revenue be embezzled?
If difficult, even corrupt politicians will deliver public goods despite themselves
o In this world, public goods provided not to win votes but because checks and balances
present in system prevent them from diverting all revenues to patronage
What determines checks? This a public good, so endogenous politicians would liket to tax
heavily to generate revenue for patronage, but constrained from doing so as high taxation
provokes intense scrutiny
o Tax rate a check? If too high, citizens get cross. If too low, not enough funds available
for politicians to embezzle
MODEL: POLITICIAN PATRONAGE
i.e. offer rewards to supporters
t = tax rate
Y = taxable income
tY = tax revenue
e = fraction of tax revenue embezzled
P = etY = total amount available for patronage
Politicians wish to max P s.t. e(t), e(t) < 0
Let ( ) t e 1 where = rate of embezzlement
Politicians problem:
( ) tY t P
t
1 max
( ) 0 2 1 2

t Y Y t Y
t
P

2
1

t
4
max
Y
P

Remainder of taxes can be spent on public goods


( )
4
2
4 2
Y Y Y
G


Politicians would like to tax heavily in order to generate revenue for patronage, but
constrained in doing so as high taxation provokes intense annoyance from citizens
Part IIB Revision, Stephen Ward 66 Development
Natural resources:
o Suppose country can also generate revenue from natural resources, rY, but this
revenue is not taxed directly but can be embezzled separately. Potential govt.
spending on public good is:
( ) rY Y r t G + 1
But politicians could embezzle part of the rY
o So wish to max w.r.t. t
( ) ( ) [ ]Y r r t t P + 1 1
( )( ) 0 2 1 1

Y r t r Y
t
P

or
( )
( ) r
r
t
2 2
2 1


Means tax rate is lower the higher is revenue from resource rents. The level of scrutiny is lower
and so rate of embezzlement is higher
Share of govt. revenue from income is now fixed
( )
( )
( )
( )
( )
2 2
2 2 1
1
2 2
2 1
1
Y rY Y r
rY Y r
r
r
rY Y r t G
+
+

+
o So given total income, revenue for patronage rises as result of resource rent
o Can now raise same money while not annoying citizen less needs to be diverted to
provision of public goods
Corollary: comparing 2 societies with same level of income but different shares of natural resource
rents, one with the higher share will have worse provision of public goods
Will country that discovers natural resources worsen provision of public goods?
Depends on scale of resource discovery, r, and rate of embezzlement,
e.g. suppose r = (resource discovery doubles Y)
o
( )
( )
0
2 2
2 1


r
r
t
r
o ( ) Y G
r
2 1
Before discovery, t = and ( )
4
2
Y
G
G G
r
<
if ( ) ( )
4
2 2 1
Y
Y < i.e. if > 6/7 = 0.86
If r > 0.5, politicians can raise G
r
(effective taxation is negative)
e.g.
o Norway high r, very low ,
G G
r
>
Relatively small resource discovery in society with strong prior tradition of
scrutiny, so very small
o Saudi Arabia very very high r, high ,
G G
r
>
Resource discovery so large that even though high, provision of public goods
improved
o Nigeria medium r, high ,
G G
r
<
Moderate sized discovery and high initial value of , where discovery probably
worsened public goods provisions: across range of social indicators Nigeria
ranked below other African economies with fewer resources
Part IIB Revision, Stephen Ward 67 Development
Lecture 7B Testing the Neo-Con Agenda: Democracy in
Resource-Rich Economies
Empirical Literature
Growth, inequality and democracy growth itself affects the quality and quantity of
institutions
o Democracy v. difficult to measure
o National resources and democracy hand in hand?
Alesina-Rodrik (1994) find negative correlation between inequality and growth but no
significant coefficient on democracy-inequality interaction term
( ) + + + + +
i i i i i i
Democracy Inequality Democracy Inequality GDP Growth * ln
4 3 2 1
o Inequality = gini coefficient never look at over time as doesnt change much
Persson-Tabellini (1994) find significant effects only in democracy, inequality reduces growth
Difference largely driven by different data sets on inequality (and to minor degree by definition
of democracy)
o Effects mitigated in presence of democracy? Expect 4 significantly positive?
Tavares and Wacziarg (2001)
o Results suggest democracy fosters growth by
Improving accumulation of human K
Less robustly by lowering income inequality (not so strong)
o Democracy hinders growth by
Reducing rate of physical K accumulation
Less robustly by raising ratio of govt. cons. to GDP
o Once all of these indirect effects are accounted for, the overall effect of democracy on
econ. growth is moderately negative
Blow for neo-con agenda not clear from econ. point of view that introducing
democracy leads to growth
o Find democratic institutions responsive to demands of poor by expanding access to
educ. and lowering income inequality but do so at expense of physical K accumulation
Jones and Olken (2005)
o Does leadership change matter?
Leadership change does not affect growth in democracy while it does in
autocracy
In democracy, really good leader cannot have big effect
o Show different kind of effect on growth democracy mutes leadership effect on growth
Collier and Hoeffler (2006) Testing the Neo-Con Agenda
Data set need 2 measures
o Resource rents
Rents = natural resource P the extraction costs
E.g. for oil, World Bank database provides average of 4 spot crude oil P
P are global, thus vary over time but are same across countries
Extraction costs are country specific and vary over time
Step 1 obtain spot P, Step 2 multiply natural resource rents per unit of
output by total volume extracted (affected by extraction costs)
Add these total rents for variety of natural resources: oil, gas, coal,
lignite, bauxite, copper, iron, lead, nickel, phosphate, tin, zinc, silver
and gold
For each yr, divide sum of resource rents by GDP
Part IIB Revision, Stephen Ward 68 Development
Rent variable constructed for 969 panel data observations
o Measure of democracy
Polity IV = 11 point ordinal scale from zero to ten higher values indicate
greater openness of democratic process
Regression analysis uses four yr averages so data averaged the data over 8 sub
periods (to get growth rates per period) (1970-73, 1974-77, , 1998-2001)
Natural resource rents as percentage of GDP shows heavily skewed frequency
o 158 countries no resource rents
o 363 countries less than 1%
o 180 countries 1%-5%
o 79 countries 5%-10%
o 187 countries 10% or higher (high resource rent countries)
o = v. asymmetrical dist. with long tail of zeros
Is growth affected by resource rents?
Medium term growth rate of economy as measure of econ. performance. LHS variable
t t t t t t
t t t t t t
ummies Continentd Checks Natres Checks Natres Natres
Democracy Natres Democracy Natres GDP Growth


+ + + + + +
+ + + +
8 7 7 1 6
2
5
4 3 2 1
ln
Checks = checks and balances
Column 1 = basic spec.: countries with initially higher democracy score have on average
higher growth rates. Coefficient on natural resource rents is insignificant
Column 2 = interaction term (rents x democracy): interaction term negative and
significant at the 10% level. Democracy appears to enhance growth except in presence of
substantial natural resources
3 possible variations
Squared resource rents: allows for possibility of diminishing returns to rents (column 3) but find
no evidence of such non-linear effects
Allow for lagged effects: large case-study literature on nat. resource rents has many examples
of public expenditure being increased to unsustainable levels (see column 4). Term is
Part IIB Revision, Stephen Ward 69 Development
Given level of checks, does more nat.
resources make growth worse or not? Expect
(-) but not sure
Main variable when more vs. less
democratic country, what is effect on
natural resources
Changes measure
of democracy but
results unaffected
Significant and positive i.e.
if many checks, pol. leaders
cannot loot natural
resources and leads to
growth
Should always leave in when
include interaction term
involving this
Baseline specifications
(ones to look at)
Democracy with checks and
balances is something to
aim for
significant, negative and substantial resource rents generate unsustainable increases in level
of output
Since natural resource rents have no significant direct effect, drop this term in favour of this
lagged effect (column 5). Now interaction of democracy and resource rents negative but not
significant at conventional levels (p = 0.166)
Other variants:
Focus on abil. of other agents to restrain govt. Democracy score and the checks variable are
correlated (=0.72) includes both in same regression
While direct effect of checks insignificant, interaction of resource rents with checks and
balances is positive and significant
Adverse interaction effect of democracy and natural resources now becomes highly significant
o Whereas democracy per say distinctively detrimental for resource rich countries,
checks are distinctively beneficial
Robustness checks
Replace democracy as continuous variable with dummy which takes value of unity if democracy
score greater or equal to five (column 8)
o Results unaffected in particular, interaction term between democracy and natural
resource rents is still negative and significant at the 1% level
Control for fixed effects (important not able to do it). Since model contains time invariant
variables (continent dummies) and variables which are in general only changing slowly over
time (political economy variables), cannot estimate model by conventional fixed effects
o Add country dummies a third at a time and only retain ones that significant.
o Results dont change (very odd procedure though - fundamental of what not to do)
Allow for potential endogeneity of democracy should note since only concerned with
interaction effect of democracy, rather than its direct effect, problem less serious than were
we attempting to infer direct causal connection from democracy onto come outcome
o Interaction effect analogous to difference-in-differences approach
Does difference in democracy have different effect depending upon resource
rents? However instrument for democracy. Instrument = historical data on
settler mortality
Conclusions
Resource rich countries tended to be autocratic and also tended to use resource wealth badly
o Neo-con agenda of promoting democratisation in resource-rich countries thus offers
hopeful prospect of better use of their econ. opportunities
In developing countries, combo of resource rents and democracy been significantly growth-
reducing
In absence of resource rents democracies outperform autocracies
In presence of large resource rents, autocracies outperform democracies
o In interests to keep country going so can take money from resources
Result robust to controlling for potential endogeneity of democracy and to fixed effects
Antidote to adverse effects of democracy is intensified checks and balances
Good idea to invade Iraq?
Countries with large resource rents need checks and balances this not what they get
Resource rents tend gradually to undermine checks and balances (resource rents = of primary
interest politically, both from US and Iraqi view)
In developing societies where state has most command over resources, the democratic process has
been least effective at controlling them for the public good
Part IIB Revision, Stephen Ward 70 Development
Implication for neo-con agenda is that it either needs to be scaled up or scaled down (no
middle ground cannot have a little bit of democracy)
On criterion of econ. performance, targeting electoral competition on resource-rich societies
appears to be particularly inappropriate unless complemented by checks and balances
Unfortunately, whereas electoral comp. easy to establish since strong incentives for
participation, checks and balances are public goods liable to be undersupplied
o Of necessity different in quality to electoral process
o Different types matter for different cons.
Govt. determine tax rate s.t. generates max revenue for patronage
Patronage expenditure determined by product of tax rate, t, taxable income, Y, and the
proportion of revenue which can be embezzled for patronage, E:
EtY P where
kt
e E

and 0 > , 1 > k


Optimum tax rate given by
tY e P
kt
t

max
which gives
k
t
1

i.e.
ke
Y
P

This gives provision of public goods (tax revenue resources used for political patronage):

,
_



e k
Y
P Y t G

1
Suppose after discovering oil reserves, countrys income increases to . Share of resource
income in is equal to r and resource rent accrues direct to govt. i.e. ( ) [ ]Y r r t G
~
1 +
instead of tY above
Optimum tax rate now given by
( ) [ ]Y r r t e P
kt
t
~
1 max +

which gives
r
r
k
t

1
1

( )
2
1
1
r sr
dt

which implies the higher is resource revenue from resource rents, the lower is
the tax rate
o The level of scrutiny is lower, so the level of public embezzlement can be higher
In developing countries, govt. does not like to develop scrutiny of people
In first case, govt. revenue G:
k
Y
Y t G

i.e.
k Y
G 1
= constant
In 2
nd
case
( )
k
r
r r
r
r
k Y
G

1
]
1

,
_


1
1
1
1
~
so as r increases, G/Y falls
o Indirectly, increase income from other than tax resources through tax when r
increases, t falls dominates direct effect leading to negative total effect
o Directly increase income from resources effect is positive
( ) [ ] ( ) [ ] ( ) ( ) [ ]Y r r t e Y r r t e Y r r t P G G
kt kt
~
1 1
~
1
~
1
*
+ + +


o Substituting in for t*
( )
1
]
1

,
_

r
k
e
k
Y
r
G
r
k r
1
1 1
1
1 1 *

Part IIB Revision, Stephen Ward 71 Development


>0 as both >0
>1 (given)
>1
<1
>0
>0
>0
Hence 0
*
<

r
G
and the country with a higher share of resource rents will have
the worse provision of public goods
Empirically
Main result that countries with initially higher democracy score have on average higher growth
o Effect of democracy on growth is positive unless introduce natural resources then
effect decreases but cannot explain until later
Add lagged resources as growth rate can change depending on discoveries e.g. double
discoveries lead to 2x growth
Govt. spending and GDP can grow sharply but not sustain in the next period
Growth can cause democracy and other way round = endogeneity problem
o As richer, more money to spend on building new institutions as required for democracy
More education, more money and more desire to participate in democracy
o Mortality a good instrument as doesnt affect growth today directly not correlated
with error term
o Could be omitted variables affecting growth
o Natural resource rents measured as %GDP possible to combat problem using total
rather than rel. resource rents or using instrument for natural resource rents such as
product of a commodity P index and Q of sub-soil assets (amount of reserves in earth)
(as used by Collier and Goderis 2007)
If dont have checks and balances, democracy can make situation worse overall effect of
democracy depends on whether have natural resources or not
o G/D in paper gives threshold of NR = 7.5%
Above this, effect of democracy on growth becomes negative
Below, democracy still has a positive effect
Part IIB Revision, Stephen Ward 72 Development
Lecture 8A Governance
Effective allocation of resources requires that decision makers have good incentives
Private Governance firms
o Good ev. qual. of priv. governance important in allocating resources in fin. mkts.
Public Governance government
o Increasing concern with problem of good governance
Good governance
Reducing predation
o Temptation to expropriate
o Protection of political and civil liberties
Efficiency in government
o Reducing waste in policy process
o Selecting Pareto efficient policy
Pursuing broad based social objective
o Limiting corruption
o Reducing poverty
Democracy
Allowing citizens a direct say in government that they have but many different institutions of
democracy
o Type of electoral system
Role of direct democracy
o Creation of independent agencies central banks and judiciaries
o Decentralisation putting more decisions at local level
o Enhancing political representation of disadvantaged groups
MODEL: POLITICAL CAPTURE (BESLEY-PRAT)
N newspaper
2 periods govt. can be good or bad
Election at end of 1
st
period
Rational voters, informed by newspapers
Govt. can try to bribe newspapers - how bribe Media?
o Envelope full of cash (McMillan and Zoido 2004)
o Fire/promote employees (if state owned)
o Public decision benefiting media company (TV license)
o Public decision benefiting firm (Italys Fiat, Russia etc.)
o Selective access to scoops US, UK
McMillan and Zoido (2004) Peru in 1990s
o Detailed records of monthly payments to judges, elected politicians, media etc. and
explicit contracts
Amounts paid to media about 100x larger than those paid to judges and
politicians
How much cost to buy a democracy?
Preconditions for successful governance
Human K
Social K
Part IIB Revision, Stephen Ward 73 Development
Other complimentary institutions watchdogs such as the media
An example of governance difficulties = corruption
Corruption and Transparency
2 period retrospective voting model (adverse selection model)
In 1
st
period, incumbent exogenously in power
o 2 possible types ( ) g b, and ( ) g Pr where g = good and b = bad
Good incumbent delivers benefit of 1 to voters while bad incumbent provides zero
At beginning of time an incumbent selected who good with probability
(politicians dont make choices come as types)
Accountability voters cannot observe payoffs (what politicians do for them)
o Can choose to get rid of incumbent in favour of another randomly drawn politician
(good or bad)
o Voters get to observe whether incumbent is bad with probability q (explained below)
o Voters can kick out bad incumbents (only works when informed as to whos bad)
o Turnover here is a proxy form real accountability as only bad incumbents kicked out
Turnover rate q(1 )
In this model, voter welfare increasing in turnover
Endogenous Information Provision
n active providers (media)
If incumbent good, observe no verifiable info. If bad, with probability [ ] 1 , 0 q , receive a
verifiable signal that incumbent is bad
Only verifiable info can be printed
Allow that incumbents can manipulate news and look at bargaining between politician and
media
Info providers can capture rent a (advertising?) by printing news
o Rent per newspaper = a/m where m is number of outlets that print news (not all do
and only those that print news get revenue)
Govt. can offer ti to each media outlet to buy silence
Only ti/ gets to outlet measures difficulty of bribing
Politicians get rent r out of which they must pay these bribes
Incumbent gets

I i
i
t r
if re-elected and

i
i
t
if not. I = number of outlets that are
bribed
Media gets signal with prob. q if incumbent is bad
Politician gets to see signal media has received then makes offer of bribe

a
r

= level of rent incumbent enjoys relative to silencing a newspaper (outlet)


If media is to stay silent, not enough to offer each of them a/n. Each have to be paid not lost
revenues a/m
o But amount outlet would get if were the only one to broadcast news = na
Equilibrium in media mkt. can be one of 2 kinds
o If
a
r
n

> , media industry is free media outlets report info to voters


o If
a
r
n

< , media industry is captured each media outlet suppresses its info in
exchange for a bribe a t
i

Media capture is a determinant of real accountability
Part IIB Revision, Stephen Ward 74 Development
With
a
r
n

< , there is no turnover. With


a
r
n

> turnover is ( ) 1 q
Suggests the following are good for real accountability
Pluralistic media (high n)
Independent media (high )
Low political rents (low r) incumbent less willing to offer bribes to media when small rent
associated with political survival
Motivated information providers (high a)
Part IIB Revision, Stephen Ward 75 Development
Lecture 8B A Free Press is Bad News for Corruption
Brunetti, Aymo and Weder (2003)
Is media freeform control of corruption in govt.?
3 routes through which corruption may be controlled
Internal mechanism and incentives within bureaucracy
o Negative if: lack of explicit standards; nepotism rather than meritocratic means of
obtaining reward; low level of public vs. priv. wages
External mechanisms (checks and balances): independent judiciary, press (rank higher than
judiciary? Bribes 100x higher than judges)
o Positive if: independent judiciary; citizens committees; generally high educ. /
development
Levels of rents that can be appropriated are low
o Positive if: low govt. intervention in mkts.; regulation and distortive P
Claim: press freedom is external control that works both against extortive and collusive corruption
Extortive = civil servants can delay or refuse projects but firms could report them (question
why do they have such powers?)
Collusive = firms may collude with indiv. agents and may have no incentive to report, so need
press
Data
Measure of press freedom assembled by Freedom House (75% US govt. funding, 25% priv.
funding) criticisms - CIA linkages, but also praised
o Since 1996, expanded indices of press freedom for 145 countries based on
Experts opinions
Findings of international human rights groups / press organisations
Analysis of publications and news services
Reports of govts. on related subjects
8 dimensions of press freedom in 4 categories
Laws and regulations that influence media content reflects judgement of the degree of actual
impact on press freedom (0-15, 15 worst) what can talk about
Political influence over media content captures political pressure on the content pol. govt.
links
Econ. influence over media content: reflects competitive pressures in priv. sect. that distort
reportage etc. report favourable to specific ind. groups
Repressive actions measures actual acts which constitute violations of press freedom e.g.
Zimbabwe throw reporters in jail
Corruption
Part IIB Revision, Stephen Ward 76 Development
Main measure or corruption = indicator collected by International Country Risk Guide (ICRG)
and varies from 0-6 (0 worst)
Use average for 1994-98 which available for 128 countries
In addition, use corruption measures from 3 alternative sources: World Bank, Institute for
Management Development and Transparency International
Specification:
i i i i i
RULE BUREAU PRESS CORR + + + +
3 2 1 0
where 1 is the key coefficient of interest should be negative
Controls for other ways in which controls on corruption could operate
BUREAU = measure of qual. of bureaucracy
o Provided annually for large number of countries by ICRG based on evaluations from
country experts
RULE = measure of external control and is also provided by ICRG
o Marks presence of sound political institutions, strong court system and provisions for
orderly succession of power
Expect 2 and 3 positive and significant
Broader Specification:
i i i
i i i i i i i
ETHNIC BLACK
TRADE HUMPCAP GDP RULE BUREAU PRESS CORR


+ + +
+ + + + + +
8 7
6 5 4 3 2 1 0
GDP = level of per cap GDP in 1995
HUMCAP = educational attainment
Proxy for external controls: the abil. of civil society to judge government performance and act as
external control on corruption. Expect 4 and 5 positive
TRADE = exposure of economy to foreign trade
o Openness implies competitiveness hence less rents
o 6 > 0
BLACK = black mkt. premium on FX
o Broad indicator of degree of govt. created distortions in economy
o 7 > 0
ETHNIC = degree of ethnolinguistic diversity
o 8 > 0
Many econometric difficulties:
Endogeneity of press freedom - reverse causality corrupt regimes may seek to repress
freedom
Relationship hold up where country not oppressive?
3 checks
o Exclude countries with highly repressive regimes from sample
o Use several instruments for press freedom pol. rights being one
o Exploit tome series dimension of alternative data set on press freedom
Cannot fully resolve the issue of causality since only have imperfect instruments and little time series
variation to work with
Results
Part IIB Revision, Stephen Ward 77 Development
Overall, the results presented in Table 1 indicate
There is a close relationship between press freedom and corruption which is robust to
specification and sample variation
Effect of press freedom on corruption is sizable
o Absolute value of the coefficient varies between 0.017 and 0.028 for the full sample of
countries - means that an improvement of 46 points in the press freedom indicator
(that is a move of the average country to full press freedom), could reduce corruption
by about 1point
When testing robustness, more mixed evidence - not so robust or so clear
Conclusions
The empirical evidence presented in this paper shows a strong association between the level of
press freedom and the level of corruption across countries
The results suggest that an independent press may represent an important check against
corruption
The result is not sensitive to the specifications estimated and for alternative measures of
corruption and press freedom
Theoretical considerations, estimations with various instruments as well as panel data evidence
suggest that the causation runs from more press freedom to less corruption
All in all the results suggest that press freedom might be an important check on corruption
Part IIB Revision, Stephen Ward 78 Development
Econometrics Summary Guide
Basic econometrics deployed in papers so far:
Following table summarises how estimates of parameters , in regression
+ + X Y
stand up to
violations of Gauss-Markov assumptions (real world hardly any met), where X is a vector of k variables
i
X , k i .. 1
Violations Impact on i
Multicollinearity in X (Xis highly correlated) Unbiased, (estimated variance, s
2
, higher, biased)
Heteroskedasticity (variance not constant) Unbiased, (estimated variance, s
2
, higher, biased)
Omitted variables Biased bias depends on corr(X,)
Measurement error in Xi Coefficient on i biased towards zero
Measurement error in Y Unbiased, estimated variance biased
Endogeneity of Xi Biased, could reverse sign altogether (opposite to
true effect expected)
Summary of bias in i when an omitted variable (say Z)
Corr(Xi,Z) > 0 Corr(Xi,Z) < 0
Corr(Z,Y) > 0 Positive bias Negative Bias
Corr(Z,Y) < 0 Negative Bias Positive Bias
Corr(Z,Y) = 0 No Bias No Bias
Endogeneity of explanatory variable is a serious problem
o Possible fix = find a good instrument
o Good instrument:
Variable that has no reason to be in regression (i.e. no direct effect on Y)
Highly correlated with endogenous variable Xi
Plausible i.e. reason for its omission from regression can be explained in
economic narrative
Measurement error also calls for use of instruments
Omitted variables effectively cause same problem as endogeneity in a structural model
Apart from instruments, panel data (data over time on the same units of observation) can be
useful, since first differencing an equation may remove correlation with an omitted variable
o Only works if omitted variable fixed over time
o E.g.
it it it
X Y + +
Suppose
it i it
u Z + where Z is the omitted variable (like entrepreneurial spirit or
motivation, some attribute fixed over time but specific to unit i) and uit is white noise
1 1 1
+ +
it it it
X Y
( )
it it it
X Y + +
1
where ( ) ( )
it it it i i it
u u u Z Z +
1

Estimating equation in 1
st
differences as above has eliminated difficulty of correlation
of X with omitted variable Z
Have also employed exclusion restrictions to test hypotheses
o Have structural model (such as complete mkts. hypothesis or that of perfect risk
sharing), where only the independent variables in the regression (if this is the
maintained hypothesis) should be prices (say)
o By adding other variables (that in principle should not affect Y), were able to test
whether maintained hypothesis of perfect mkts. holds
o More importantly, if pattern of significant coefficients on supposedly irrelevant
variables can tell us anything about which mkts. fail
Part IIB Revision, Stephen Ward 79 Development
Part IIB Revision, Stephen Ward 80 Development
Demography and Development
Lecture 1 Why do economists want to explain demographic
behaviour? Transition theory and the Malthusian model
Why do economists want to explain demographic behaviour?
Historical trends in pop. growth
o History matters
Population bomb or population explosion
o Population harmful or beneficial for econ. development?
o Birth of pig increases national income, birth of child decreases it
Contemporary population debates echo what Malthus (1798) set out in An Essay on the
Principle of Population
o Obvious truth pop. must always be kept down to level of means of subsistence and
its a view of these means which forms the strongest obstacle in way to any very great
future improvement of society
o Argued pop. growth harmful for econ. development
Population trends
Year World Pop Year World Pop Year World Pop Year World Pop
1 AD 300m 1800 950m 1933 2,057m 1995 5,716m
1500 600m 1850 1,200m 1950 2,515m 2000 6,130m
1750 800m 1900 1,650m 1985 4,800m 2030 8,200m*
*Projected estimate, based on current trends
Natural pop. Increase when birth rate > death rate
Worldwide pop. Growth rate is 1.5% per yr and expected to continue (about 230,000 per day)
o 90% of increase forecast in LDCs (SAsia, SubSahAf and MidEast)
Mortality 20
th
century LDC mortality lower than in 19
th
century LDCs
o Due to spread in medicine after WWII
o DR decline preceeded decline in BR by 20 yrs
Very high rates of pop. Growth of 3-4% a yr. 20
th
century LDC fertility is 150%
greater than max achieved in 19
th
century
Post 1960 and demographic momentum: birth rates also declining rapidly in most LDCs
Strong regional differences between LDCs, SubSahAf and MidEast (both fertile. Rates of about 6
kids per woman but very different incomes, so what causes?
Higher av. income gives lower fertility relationship. SubSahAf (4-6 kids) vs. EAsia and LatAm (2-
3 kids)
o Outliers = ME (sig. exceptions), China (high income, lower fertil), India (high income,
higher fertile.)
Faster growth during 20
th
century than during 19
th
century, particularly in LDCs
o Related to balance between trends in mortality and fertility
o Direction of causation between pop. and development unclear
May think hinders growth but may be other factors
Birdsall (1988) evidence indicates strong negative relationship between TFR and income per
cap.
Pop. Growth good or bad for econ. development?
Part IIB Revision, Stephen Ward 81 Development
Improvement in development indicators: world food prod., infant survival, LE at birth, literacy
all improved despite rapid pop. growth
Positive externalities of human genius (Boserup, Simon) may increase supply of human genius
leading to positive externalities for everyone else. But, if cannot get primary education, no
point being a genius!
o Boserup In 19
th
century high pop. growth and density good for agric. development,
increasing cons. D (enabling E of S), in turn lowers cost of dist. and implementation of
tech advances as well as providing large, low cost lab. supply
Limitation does not control for other differences between 19
th
and 20
th
centuries benefits may be dependent upon other institutional factors. LDCs
not currently developing as result how long need to wait? Historical ev. mixed
much of Europe had rising pop. without rising growth due to serfdom
(medieval farm workers)
o Simon Scarcities caused by pop. growth can be handled by mkts. scarcities give P
rises which gives incentive to invest in new technologies giving cons. fall meaning
scarcities disappear. Also argues large pop. leads to longer tail (large kurtosis) so
increases human genius, which leads to tech. advance
Limitation human genius requires inputs of capital. Educ. and health
investments may be lower if pop. growing too fast
Endogenous growth models new ideas/knowledge the main sources of econ. growth; pop.
growth increases mkts. for goods and services
o Problem assumes tech. growth wont have more than finite effect on natural
resources
Bauer argument on peasant rationality - poor people in LDCs irrational? If bad for
development, must still explain why people in LDCs have high birth rates
o Bauer argued poor people in poor societies are rational and they believe high fertility
increases their well being development failure has other causes
Collection of reasoned decisions at indiv. level always optimal at collective level?
What are resource allocation failures in context of childbearing?
o Mkt. failure here that factors to decide if have kids or not are wrong dont take into
account externalities (social cost of having child may exceed private cost) - people
dont see links between indiv. behaviour and society level effects
o Mkt. failures households face wrong prices rel. P of goods and services that
influence decision to have children could be wrong
o True externalities cost of having children arent incurred by parents so they
overconsume
At household level, parents in LDCs may be perfectly rational in preferring large numbers of
children but at collective level, pop. growth may be problem
o Pecuniary externalities external effects resulting completely from trading in mkts.,
expressed completely in prices e.g. as pop. increases, so does lab. supply, which can
lead to higher rents and lower wages
Major sources of externalities from high fertility and fast pop. growth:
Savings and Investment (especially human K investment)
Savings
o Fast pop. growth leads to higher dependency ratio (more dependents per worker)
o Dependents cons., workers also prod.
o High dependency ratio leads to higher cons. at any given output which leads to lower
per cap savings
o Criticisms of this
MDCs show empirical link between pop. and savings rate but LDCs dont
Could be because of data problems or behavioural differences
Argument could hold but no empirical findings to suggest it does
Part IIB Revision, Stephen Ward 82 Development
K widening and deepening
o If pop. growth occurs then K has to rise just to maintain current levels of investment
per person makes hard to deepen K invested per person
o Pop. growth may harm productivity, output and income
o Criticisms
Empirically, hard to trace link between pop. growth and low investment
Could be case that fertility high as investment low rather than vice versa
Environmental externalities (especially well-defined property rights) and tragedy of the
commons
o Effects on nat. resources excessive use lack of indiv. incentive to restrict use
o Originally people thought that with pop. growth would run out of non-renewable
resources but current consensus that non-renewable resources with well-defined
property rights less of problem than renewable resources with poorly defined property
rights
Where well defined, P mechanism ensures change in scarcity reduces cons.
and/or impels search for substitutes
Faster pop. growth shortens period over which people consume resources but
not number of people that consume it
Greater use at time t may be good people poorer now and rising P will speed
discovery of substitutes
o Criticisms
Longer time period allows time for accidental tech. substitutes to take place
but in contrast, if inventions scarcity induced, shorter time period may be
beneficial
P may not reflect actual scarcity e.g. oil rich parts of 3
rd
world and Nigeria
where energy P subsidised. Pop. growth could make worse solution to target
pricing pol. not pop. growth
o Pop. growth interacts with renewability in several ways
Greater use at time t need not reduce use in time t + 1 unless use excessive
Excessive use = use that exceeds rate of regeneration
Fast pop. growth leads to higher risk of excessive use
Some borrowing of stock from future may be sensible as people currently poor
but irreversible over-use must have huge benefits no matter what discount
rate is, very serious to completely give up future
o How pop. growth interact with property rights?
Good property rights make owners resist excessive use to protect value bad
property rights leads to tragedy of commons
Abuse of commons exacerbated by pop. growth may create stronger incentive
to free-ride
o Criticisms
What is really fundamental problem? Is problem really higher fertility or is
higher fertility rational given circumstances? Seems problem not pop. growth
but absence of well-defined property rights
o Revisionist perspective pop. growth not the cause, but is exacerbating factor due to
lack of property rights
Pop. makes property rights develop but may cause degradation faster than
institutions develop depends on rel. speeds
2 policy dilemmas
More feasible to intro property rights or pop. limitation?
Could pop. limitation compensate fully for lack of property rights?
Social norms Marshallian atmospheric externalities = cultural externalities peer group
interactions/emulation
Part IIB Revision, Stephen Ward 83 Development
o Dasgupta (1993) Indiv. choices a function of av. level chosen by others social and
indiv. choices are a self-reinforcing coordination problem
o Argued reproduction is a social activity
o Social = individuals goal maximising level of action is an increasing function of av.
level chosen by others (norm) other examples = how much we work and how much
education to give daughters
o Fertility influenced by priv. costs and benefits plus shared values in community may
have self-sustaining equilibrium
o Woman wont break away from fertility norm when no-one else does but will when
others do
o Multiple social equilibria with associated norms of fertility
High fertility regimes atmospheric externalities that sustain social norms
(especially of female behaviour) that give rise to high pop. growth to begin
with
Leads to more high fertility eve when harms society as a whole = vicious cycle
Theories to explain pop. growth in developing economies
o Theories of demographic transition
o Malthusian theories
o Microeconomic household D models
o Models of social interactions
Demographic Transition
Part IIB Revision, Stephen Ward 84 Development
Theory that all economies pass through certain levels through their development
Stage 1 = high birth rates and death rates
o Typical birth rate 35 per 1000
o Typical death rate 30 per 1000
o Slow pop. growth of 0.5% p.a.
o Problem
Before 1950, LDCs had birth rates (43 per 1000) and death rates (37 per 1000)
much higher than Europe at same stage
Stage 2 = modernisation
o Death rates fall from 30 per 1000 to 20 per 1000
o LE rises from <40 t0 >60
o Birth rates stay high (35 per 1000)
o Fast pop. growth of 1.5%
o Problem
Death rate did begin to fall in most LDCs from 1950, but much faster than in
19
th
century LDCs to 10 or 15 per 1000, so instead of growing only at 1.5% p.a.,
20
th
century LDC pop. grew at 2-2.5% p.a. (due to medical advances?)
Stage 3 = further modernisation
o Birth rate falls from 35 per 1000 to 10 per 1000
o Death rate continues to fall to 10 per 1000
o Birth rate and death rate converge and pop. growth is zero or very slow
o Problem
Part IIB Revision, Stephen Ward 85 Development
Consider 2 cases A and B on diagram. Majority of LDCs (SubSahAf) still in case
B. Failure to experience entire mortality decline of stage 2 because of
widespread and persistent poverty (still at 40 deaths per 1000)
Problems showed some difference in transition for some present day LDCs from the Western
European experience
o LDCs live at home after marriage whereas in Europe, save up for kids and marriage
o Both birth and death rates higher for current LDCs than theory predicts
Marriage behaviour in Europe very late but in LDCs about 16-18
High celibacy rates in Europe
Marriage near universal in LDCs
Transition theory purely descriptive explaining fertility transitions in terms of black box of
modernisation
Malthusian Model
Assumes
o Pop. grows at geometric rate, doubling every 30-40 yrs
o Food supply grows at arithmetic rate
Fixed factor of land
Cannot keep pace with population increase
Very striking consequence per cap incomes fall to subsistence
o = idea of Malthusian trap = analogous to low level poverty trap
Avoid absolute poverty by indulging in moral restraint
Population trap theory
Diminishing returns to scale in prod.
o More lab. leads to productivity falls
Positive income effect on fertility
o As per cap income increases, fertility increases too
Reproductive behaviour homogeneous across economy
o Fertility responds to change in aggregate factors
See diagram below
o A = Malthusian pop. trap level of income attained, economy stuck at chronic low
income Y1, and is a stable equilibrium
o For poor nations to overcome pop. pressures, need
Malthusian preventive checks such as delayed marriage and birth control
Malthusian positive checks such as starvation/disease/wars will restrain pop.
Part IIB Revision, Stephen Ward 86 Development
Types of big push to get to B
o On P curve
By preventive checks e.g. family limitation programmes
o On Y curve
Investment and industrialisation programme
What Malthusian pop. trap model doesnt explain
o Ind. revolution
o Shape of Y curve: technological progress
o Shape of P curve: mortality and fertility dont relate to income in predicted way
o Poverty in India in 19
th
Century
Lessons from Malthusian Theory
Effect of pop. on standard of living
Influence on pol. making and model-building in economics
Malthusian (better Historian than Economist) theories flawed by assumptions of diminishing
returns and of positive income effect on fertility (both wrong) but model better than
demographic transition
o Contribution of possibility of effect vital and very important for all to follow up
Still need a better theory
Part IIB Revision, Stephen Ward 87 Development
But assumption of decreasing returns essentially wrong
In fact increasing returns
o So no point A where cannot escape
Also income effect on pop. curve questioned
o Several quantitative studies question
it
o Increase income not generally
associated with increase in fertility
Positive relationship between per cap and agg. income
growth until Ys where decreasing returns set in
A, B and C all equilibria
A = Malthusian pop. trap level of income set very low
level
Small level above or below returns economy to
point
Need big push strategy to get far away from
trap, to B
Can get big push on P or Y curve
Push P below Y increases per cap income (pop.
restriction strategy) or push Y above P
B moves on to C = stable
B ensures reaching C but C is much higher per
cap than B
Policy aim to push past B as quick as possible
Theory better than demographic transition
Lecture 2 The Household Demand Approach and its
Applications
Origins of neoclassical household D approach
Micro theories of D group of, not just 1 theory: Becker (1960,1981), Willis (1973), Becker and
Barro (1988)
o Based on household prod. function treat household as prod. unit like firm
4 Developments in economics that underlie the model
Investment in human K
o Benefits of kids different in rich and poor countries
o In HDM, households can use kids as form of human K parents make sacrifices to bear
and rear children and in return get services or benefits in future periods
o In rich countries, benefits primarily personal satisfaction
o In poor countries, may be additional benefits
Benefit from kids working in family household or farm
Kids mkt. earnings contribute to household income
Kids may support parents in old age
Allocation of human time
o Mkt. and household (non-mkt.) activities
o Money costs and time costs
o Bearing and rearing children is household activity as highly lab. intensive, particularly
in mothers time
Household prod. function
o Household services
o Child services highly lab. intensive in cost of mothers time
o Important goods that not traded in mkts.
Household as cons. or prod. unit
o Prod. function goods and time as inputs to prod. household goods e.g.
surviving/educated kids, cooked meals etc.
o Maximises utility in cons. and determines allocation of time and goods in household
prod.
o Assumptions similar to theory of firm
o Welfare of each family member combined into household U function
Utility function
( ) Z Q N U U , ,
N = number of children = quantity of children
Q = quality per child = household investment per child
Z = all sources of satisfaction to husband/wife other than those arising from children
Production function
( )
c c
x t f NQ C ,
C = child services = product of number of children and quality per child. Requires both money
and time
tc = vector of total amount of time parents devote to children during parents lifetime
xc = vector of total amount of goods parents devote to children during parents lifetime
o Trade off between money and time
Budget Constraint this is a full-income wealth constraint which includes both money and
time, and in which income is exhausted on child quantity N, child qual. Q, and all other
satisfactions Z
z Q N c
Z QP NP NQ I + + +
Part IIB Revision, Stephen Ward 88 Development
I = familys full income
i = cost minimising shadow prices (non-mkt. P reflecting OC of commodity/time)
c = shadow P of child services
Z = shadow P of all other commodities/satisfactions
PN = fixed P applying to component of child costs independent of level of Q chosen
PQ = fixed P applying to component of child costs independent of level of N chosen
Solving F.O.C. i.e derived demand for children
o N = demand for child quantity = ( )
Q N z c
P P I N , , , ,
o Q = demand for child quality = ( )
Q N z c
P P I Q , , , ,
o Z = demand for other satisfactions = ( )
Q N z c
P P I Z , , , ,
Model makes predictions about relationship between fertility and income
Child services are normal goods as incomes rise, parents demand more child services
Rising wages increase the value of time (OC of time to parents higher)
P of children rises relative to the price of less time-intensive satisfactions
P of children rises because womens time costs rise disproportionately
o Women affected more than men leads to substitution of quality for quantity of kids as
incomes rise
o Any factor that increases OC of womens time remaining out of LF essentially increases
cost of having children and reduces D
Means if rising incomes result in disproportionate rise in womens wages, this
will make children rel. more costly
Also predictions about trade-off between child qual. and quantity
Differing income elasticities of D for quantity and qual.
o D for child qual. may be more responsive to income gains than D for child quantity
Evidence that qual. elasticity of D for cons. goods more responsive than Q
elasticity
Interaction between qual. and Q
o Fall in shadow P of child qual. causes decrease in D for child quantity
Rel. cost of Q in model depends on N. C = NQ. If high quantity, cost of
increasing qual. higher (as like to give all same qual.)
Expected benefits from having kids
Shift from household to mkt. prod. makes child lab. input less valuable
o As income increases, new substitutes available more income from mkt. prod. more
inputs from outside household
Infrastructure reduces value of child labour to collect households fuel and water requirements
Structural change in economy reduces mkt. jobs for kids instead greater emphasis on skilled
work
Greater access to K mkts. reduce savings/investment value of children
Insurance, pensions, state welfare reduces old-age motive for having kids
Greater geographical mobility reduces certainty of old-age support from kids
Conceptual limitations of household D model
Assumes fertility maximisation
o Ignores possibility of change in rationality that may affect fertility decision making
o Ignores things such as influence of religion, politics, institutions etc.
Assumes complete information and certainty = weakest of all assumptions
o Essentially a 1 period static model
o Assumes fertility decisions made simultaneously, not periodically
Part IIB Revision, Stephen Ward 89 Development
o Assumes know life income, kids futures and costs forever etc.
Assumes household U function
o Single function embodies all households
o Women and men differ in wants of children bargaining models of utility
In general, women want to have less children than their husbands
o Single household U function does not look at intra-household bargaining
Explains changes at margin better than changes over time
o Only over time if income changing very generally
Despite this, still makes good predictions of fertility decisions (better than previously discussed
models)
Empirical evidence on P-income relationship
Mothers time is key component in P of a child
o If source of rising incomes in economy higher wages, then this going to increase value
of time cost of child
o If rising incomes simply associated with greater investment in womens efuc., agein
going to increase their wages and work opportunities
Consider time taken to bear and rear (each successful) child
o 1.25 yrs of pregnancy and breast feeding
Africa 6.6 kids average and LE = 50 33% of adult life spent in this, not even
including raising kids
Most reproductive costs differ between men and women
o Women incur far more of the cost
o Cost could be shared by wider extended family too
In Africa, fostering 1/3 live with kin at any time
Costs of children spread across extended family leads to reproductive free riding e.g. in India
and Africa
o = usually higher fertility where some help with raising
Evidence on relationship between womens educ. and fertility in poor economies
Increase in female educ. (by 1 yr) reduces fertility by 25% = big effect
o Mens educ. doesnt have same impact
Primary schooling may be exception in some contexts e.g. Africa
o Positive effect only with primary education, not with secondary and higher educ. has
negative effect on fertility
o Primary educ. in many cases increases fertility
Could be because little bit of educ. sometimes weakens traditional taboos such
as practice of total abstinence after becoming grandmother
Educ. lowers fertility other than through value of time
Increases efficiency of contraceptive use
Informs women better about non-household options (such as LF participation)
Increases efficiency in prod. high qual. kids
Increases performance of status prod. duties
o Women can manage household accounts, help with childrens homework, bargain with
dominant mother in law
Evidence on female economic activity and fertility
Universal negative relationship does not exist
o In general, women need to be in activities that incompatible with child bearing and
rearing for LF participation to have impact on fertility
But clear association can be seen between fertility, female econ. activity and female education
Part IIB Revision, Stephen Ward 90 Development
o But doesnt reveal any causal mechanisms could be that fertility, educ. and work
decisions endogenous
Evidence on Q-qual. relationship
Trade-off between child quantity and child qual. implies as income rises, D for qual. increases
faster than D for quantity
o Income elasticity for quant of children lower than for qual. so as income rises, D for
childe qual. rises much faster than D for child quant typical with most cons. durables
Income effect conceals hidden P effect shift to rising Q means PN will rise meaning demand
more Q, less demand for N
o As qual. rises, P of child essentially rises and rising D for child qual. makes each unit of
N (child quant.) more costly, which leads to lowering of D for child quant. N
o Pure income effect also has hidden P effect built into it
Empirically, as economies grow and develop the av. qual. of kids rises
Lab. mkt. changes also important increasing returns to skilled as opposed to unskilled lab.
Ev. from S India time spent by kids in school has significant negative effect on family size
(Kanbargi and Kulkarni 1986)
o Look at Karnataka region in India and look at time spent by kids in school, whether or
not it has some impact on total family size
o Effect of time that child spends in school (measure of child qual.) persists even after
other variables like childrens work being controlled for
How much kids go to school partly a function of child lab requirements (Iyer 2002)
Evidence from Progressa scheme in Mexico
o Controlled experiment of conditional transfers that lasted about 3 yrs
o Parents paid money to send kids to school for kids in 3
rd
-9
th
grade
o Found payments actually increased number of children going to school
Another important scheme mid-day meal scheme in Southern India food provided when child
attends school increases incentives for parents to send kids to school
Evidence on income-benefits of childbearing
Children contribute to current income
o In Egypt, lab. participation of 8yr olds about 50% and children actually bring in 1/3 of
family income
Children costless to rear by time they reach adolescence in some poor countries (Cain 1977)
o Bangladesh child in rural Bangladesh becomes net producer as early as 12 and by 15,
male kids basically compensated for own and one other siblings cumulative cons.
o Kids extremely important in incomes that may be providing to households but
essentially costless to rear by time reach adolescence
Lack of fuel and water infrastructure increases value of children as producer goods
o Kids collect fuel and water in many areas
Population-poverty-environment relationship (Dasgupta 1993, 2000)
o Nexus between pop., poverty and environment high fertility and pop. growth damages
environment
o Makes fuel and water collection more costly which leafs to greater incentives to have
more kids
o That further damages resource base and may lead to mortality (as in Somalia) or to
pol. intervention
o Argument has Malthusian outlook and similar to positive checks
How model Kids as investment goods?
Kids contribute to future income more important for insurance than as producers today
Part IIB Revision, Stephen Ward 91 Development
Lexicographic safety-first decision rules show prob. parents would have son for old age support
(May and Heer 1965, Cain 1985)
o Calculated what would be prob. that Indian couple at end of lives (65) would have one
healthy, loyal, surviving son to look after them in old age
In 1965, came up with number of 6.5 kids to have 1
TFR in India at time was 6.5 study replicated more recently and number was
3.6 at time, TFR was also 3.6
Offspring substitute for missing mkts. in social security
Social security in formal and informal sectors
o Life insurance, pensions etc. tend only be available in formal sector
Rise in incomes can change parents investment decisions
Increases cash surplus for investment, awareness of alternative investments, availability of
other methods of saving and insurance
Evidence on widows in Bangladesh (Cain 1977, 1984) and microcredit schemes
o Looked at whether sons would be important for widows in Bangladesh pointed out
widows who had able bodied sons were more able to hold onto land
o Today in Bangladesh, massive change because of microcredit schemes lead to far less
reliance on sons for women than was case in 1984
o Implies possible to intro alternative source of savings, insurance, investment for
women in particular that does not have them relying on sons will immediately have
decline in fertility
Evidence on pensions in Mexico (Nugent and Gillaspy 1983)
o Held lots of factors constant and divided all of Mexican provinces into those that had
intro a pension system and those that had not
o Found those who had intro system had immediate reduction in fertility
Parents could use pension in old age and did not have to rely on kids
Conclusions
Model explains fertility decisions in terms of money costs of children and OC of parental time
Empirical findings address relationship between income and P, Q and qual., income and
benefits of childbearing
o Income benefits of kids are lab. income and contributions to old-age security
Compared to previous approaches, explains better both current causes of high fertility in less-
developed countries and reasons behind fertility transition in historical experience of more
developed ones
o But essentially an econ. theory about couples decision making doesnt really tell
much beyond that
Extensions take into account externalities of childbearing and role of social norms and
interactions
Pol. implications of models have spurred debate about D for kids vs. S of family planning
Part IIB Revision, Stephen Ward 92 Development
Lecture 3 Externalities to Childbearing and Social Norms:
Demand for Children vs. Supply of Family Planning
Empirical trends in fertility the HDM cannot explain
Sharp falls in birth rates over short stretches of time (e.g. Spain, Italy, East Asia)
o Spain and Italy now below replacement level of fertility (about 1.2 kids) but were
previously above but still poorer than other countries
Puzzle of European demographic transition e.g. Italy and Spain vs. France and Germany
Contemporary fertility trends in South Asia (e.g. India and Pakistan)
o Rapid fertility declines why?
Income and education do not explain fully actual declines in fertility (Bongaarts and Watkins
1996)
o Dont explain it alone hold constant and fertility still declines
o They hold income and educ. constant and find fertility declines
o Fertility declines earlier if neighbouring countries experience declines
o Suggests neighbouring countries have some sort of impact on fertility behaviour may
be that there are social interactions or could be some common variable explaining both
Explanations for actual fall in fertility
Ideational approaches
o Explain change in fertility with a change in ideas
o Religious beliefs, secularism and individualism protestant ethic (Weber) encouraging
Protestant beliefs encouraged literacy as people had to read bible in Eastern Europe
traditional societies fertility socially controlled
Broad change in ideas to secular beliefs affects fertility
o Key idea that in trad. Societies, fertility socially controlled and with greater changes in
those ideas and greater tolerance of indiv. choice, can see fall in fertility
o But mechanisms behind these changes largely unexplained
Argue many ideational factors endogenous to fertile. Decline so how separate
effects of ideas from effects of fertility and see which of these factors
affecting and in which directions?
o Role of culture ambiguous no causal relations identified that could be relevant for
fertility decline
Institutional approaches
o Culture needs definition in terms of structural conditions or historical experiences (e.g.
family systems, social classes, political structure)
o Interdependence between population change and other aspects of development
Fertility may be determined by level of development but in turn affects
environment recognise the endogeneity
o Fertility decisions of couples in a community are interdependent
Economic approaches
o Fertility should be modelled as outcome of optimising household behaviour subject to
constraints imposed by social institutions, imperfect info and bounded rationality
o Social institutions may also be endogenous with potential to be transformed by
aggregate behaviour
o Common changes in unobserved variables
o Social interactions peer-group behaviour and neighbourhood networks e.g. Kenya
Explain any change in econ. behaviour in terms of friends, peers, family and
networks
o Social multiplier effects
Part IIB Revision, Stephen Ward 93 Development
o Role of reproductive externalities, macro and micro level, on fertility
Small changes can lead to big effects
Small changes e.g. replace broken windows, clean graffiti lead to fall in crime
Apply to fertility too size of multiplier very large for social networks
Macro and Micro level externalities to childbearing
Source of fertility related externalities
Macro level socially optimal number of kids differs from privately optimal number
Micro level externalities created by structure of family

Joint families create intra-family externalities
Fixed and variable costs of children
As structure changes from joint to nuclear, costs of kids borne more directly by couple lowering
fertility
Micro level externalities also created at level of couple husbands and wives have different
reproductive goals
Empirical evidence on externalities reflected in demographers estimates of unwanted fertility
Dasgupta (1984) diversion between private and social costs finiteness of space
environmental impacts, crowding and spread of pathogens (HIV), increased prevalence of large-
scale migrations arising from crop failures or war
Hajnal (1982) NW Europe has later ages of marriage and lower fertility due to need for
resources to set up new household in nuclear family framework explains low fertility prior to
family planning introduced
Goody (1976) SubSahAf responsibility for rearing children more diffuse within kinship group
not adoption, just institution for mutual insurance in semi-arid regions
o Means too many kids if parents share of benefits exceed share of costs
Part IIB Revision, Stephen Ward 94 Development
of additional child to couple
Optimal by max distance between costs and benefits
i.e. MC = MB
Externality caused by increased fertility at macro level
but could also occur at family level (micro) too
Knowing family bear some cost of kids lowers
private costs and increases fertility
Slopes, not levels, are key determinants of
fertility
Shift in total cost does nothing to affect
couples private cost remain n*
As structure moves to nuclear, costs bared by
couple tend to costs of family and bear own
costs instead
Men ignore cost of childbearing more than
women?
Many demographers measure unwanted
fertility generally in developing countries
Cost to given couple
o Reproductive externality made worse by fact fathers often dont bear costs of siring
kids
Caldwell and Caldwell (1990) patrilineality, weak conjugal bonds, command land tenure and
strong kinship support system of kids combine together as source of reproductive externalities
that increase fertility heightened by conformity
Effect of social norms and societal interactions affect fertility
Custom and conformism in fertility related decisions see lots of under social norms
o Conformist if all else equal, every familys desired size greater the larger is tha av. size
of family in the community. Deviation can lead to loss of status effects
o Conformism a reason for multiple reproductive equilibria
Models of strategic complementarities
Marshallian atmospheric externalities: couples fertility decisions may depend on average
level of decision making of all other couples in society
Existence of multiple equilibria
Role of expectations and history
Changes in fertility can be affected through coordinated changes in pol.
o Economists focus on D for kids
o Demographers focus on supply of family planning
To move from R to P
Requires sig. change in pol.
o Educated women (Farooq, Ekonmen and Ojelade 1987)
o Bangoorts and Watkins, Iyer (2000) influence of exposure to media can lead to slow
downward shift of function and thus reduction in TFR
At first will be slow, gradual decline before tipping point reached leading to
larger shift
Kohler (2000) models fertility decline as coordination problem suggests social networks can
help establish Pareto superior equilibrium
o Abil. to escape depends on form of interaction of social network and pop. size and
structure
o Info networks provide info about fertility intentions of other community members and
allow women to update expectation about fertility levels chosen in their generation
If already high fertility equilibrium, network has little effect in achieving
fertility decline but can increase speed of fertility change that already taking
place
o Coordination networks ties among indiv. in groups sufficiently strong to allow
collective action groups anticipate effect on institutions and can change pathe can
lead to earlier econ. transition
Part IIB Revision, Stephen Ward 95 Development
P, R,T = stable equilibria, Q, S = unstable equilibria
Equilibriums are non-cooperative Nash equilibria of
the social system
o Which selected depends on expectations
each household has about each other
o History of system
P would be much better than R household in too
much activity but no one household has the
incentive to deviate from point R society stuck in
low marriage age, high fertility even when
rationality changes
o How move from R to P?
Economists focus on factors
from household D model e.g.
educ.
Demographers say no, focus on
family planning programmes
= household
reaction curve
Evidence from Bangladesh suggest outcomes in other villages has effect on social norms of
fertility, but with diminishing effects as geographical proximity lessens
o 70 treatment villages put under massive treatment and birth control programme and 70
control villages
o Prevalence of contraceptive use increased from 7-45% in treatment group 1977-85
while in control group increase was just 16%
o Shows influence of other villages did have beneficial behavioural externality but fact
treatment villages not proximate to control villages meant they were not followed all
the way
How can state affect fertility behaviour?
LDCs different to historical populations in number of respects
o Current rate of natural population increases large
o Experience different to historical populations e.g. in Europe
o Unparalleled and remarkable decline in death rate
Watkins (1990) regional differences in fertility between countries during fertility transition
gradually eroded
o 1870 spatial clumps within each country, suggesting importance of local communities
on behaviour. By 1960, differences within each country were less, with this being put
down to increase in geographical reach of national governments and spread of national
languages
Programmes state role in providing?
Overall regulatory regime
Access and opportunity land reform?
Public sector expenditures and transfer payments age? Family status?
Symbols of national identity and cultural continuity
o Using tools, state has very important role to play in affecting fertility
2 key issues pop. pol. must address
Reducing desired fertility number of children couples want
Reducing unwanted fertility number of children couples dont want
How target D for children to reduce desired fertility?
Micro theories suggest pol. that raise costs of children and reduce benefits = significant
E.g. female educ., reduce requirements for child lab.
Demographers argue need to qualify predictions of household D model
o Fertility decline occurred in varied contexts e.g. Europe and North America
o Modernisation is sufficient but not necessary condition for fertility decline
o Diffusion and cultural factors may be important related to language?
o Fertility control occurred in absence of state-sponsored family planning programmes
e.g. France
Watkins role of gossip people who net around village spread their ideas for change. Cultural factors,
contiguous provinces that may have shared similar culture might have had similar experiences
suggests this what happened in Fr and Ger and W Europe
Fertility decline in England in 19
th
century
Fruits of philosophy and precipitous fall in fertility in England in mid 1870s Knowlton
Population question
o Social outcry against by media
o Sales skyrocketed due to publicity
More effect on birth control than any govt. scheme could
Part IIB Revision, Stephen Ward 96 Development
Knodel and De Walle (1979) European experience fertility transition occurred suggesting family
planning programs in developing world unlikely to contribute much to precipitation or acceleration of
fertility declines
With socioecon. development, motivations to reduce family size emerge and fertility decisions
taker care of themselves modern birth control methods therefore not necessary
Argue fertility transition took place in spite of large varieties of social, economic and
demographic conditions with family limitation largely absent
When family limitation and decline of marital fertility began, they were largely irreversible
cultural setting helped influence onset and spread of fertility decline independently of
socioeconomic conditions
What is unmet need for family planning?
Condition of wanting to avoid or postpone childbearing but not using contraceptive method
Huge shortfall in supply of family planning worldwide
Economists objections to concept of unmet need
o Concept may be illogical
o Concept may not be consistent with evidence; cross-country differences in fertility not
explained by unmet need (Pritchett 1994)
Advocates desired fertility view couples can achieve fertility targets but there
is high fertility because people want extra kids need to change choices that
determined by social, econ. and cultural conditions
Pol. that improve objective conditions for women, raising income, educ.,
encouraging empowerment are probably most important and sustainable way to
achieve necessary reductions in fertility to slow pop. growth
Excess or unwanted kids play minor role in explaining fertility differences
o Cross country difference are explained by trends in desired fertility
Demographers argue family planning not only about access, also about info
Media campaigns many have impact e.g. China, India, Tanzania, Jordan
Costs of contraception that may affect unmet need is concept of unmet need really useful?
Unmet need not only about access to contraception may include other costs
Lack of necessary knowledge about methods e.g. SubSahAf or Pakistan
o Not aware of modern methods even if aware of traditional
Social opposition to use of contraception (e.g. India, Egypt, Ghana)
Health concerns about side effects multidimensional (e.g. Nepal, Pakistan, Zambia)
Unmet need linked to D for children as joins contraceptive behaviour with fertility preferences
Unmet need joins family planning to other development goals of pop. pol.
Unmet need results in unwanted fertility; to 1/5 of all births still unwanted
Implications of reproductive externalities and social norms of pop. pol.
Externalities important for fertility-related decisions
Rural societies could be stuck in high-fertility equilibrium that does not max well-being
Changing this requires coordinated change in social pol.
o Key is getting to new equilibrium but this is very hard
Economists focus on targeting D for children in order to reduce desired fertility
Demographers focus on supply-side factors to reduce unwanted fertility
Persistent pol. mistake during past 3 decades been to equate pop. pol. only with improving
family planning services
Must pay more attention to reproductive externalities, social interactions and fertility transitions
Conclusions
Part IIB Revision, Stephen Ward 97 Development
Existence of coordination failures and multiple equilibria suggest role for govt. pol
Demographers focus on targeting supply of family planning
Economists focus on targeting desired fertility and D for children
Consider how to estimate effect of social interaction on fertility behaviour
Dasgupta to help lower fertility, desirable to offer family planning services and to empower
women but other pol. such as provision of infrastructural foods, changes to property rights,
means of communication and measures that directly increase econ. security of poor are also
desirable, as are pol. and civil liberties
Lecture 4 Social Interactions and Fertility Transactions
Why social interactions important for development theoretically?
Indiv. decision making in presence of social interactions
o Behaviour of 1 agent affects behaviour of other agents
Interdependencies between indiv. decisions and decisions and characteristics of others within
common group
Interdependencies that directly link indiv. (Arrow and Hahn 1971)
Classical and neoclassical economics operates with atomised and under-socialised conception
of human action. Theoretical arguments disallow by hypothesis any impact of social structure
and social relations (Granovetter 1985)
Household D model suggests fertility will decline with education and substitution towards child quality
Social interactions as explanation for fertility transactions: Household D model with social interactions
Womens actions based on average of everyone else in that societys actions
Social Interaction = public interaction between indiv. in society as they perceive each other and
observe each others fertility behaviour
Approach 1: Social interactions in predetermined groups (Akerlof 1997, Brock and Durlauf 1999,
2001)
o Social interactions affect intertemporal decision-making (Binder and Pesaran 2001)
o Social interactions within ethnic groups (Loury 1997, 2001)
Approach 2: Social interactions lead to group formation
o Neighbourhoods (Borjas 1995)
o Interactions beyond neighbourhoods (Aizer and Currie 2002, Conley and Topa 2002)
How measure empirically?
o Demography
Social learning women share experiences of new contraceptive technologies
etc. leads to greater use, leads to fall in cost of use = local interaction
Social Influence other ethnic/community members use contraception type
makes acceptable e.g. certain religious type
o Economics Microeconometrics of social interactions (Manski 1993)
Part IIB Revision, Stephen Ward 98 Development
Money Costs
Time Costs
Fertility
Social interactions or strategic
complementarities
Endogenous effects womans CEB (child ever born) tends to vary with average
CEB of ethnic group or locality i.e. with actions of reference group 1 woman
observes mean fertility in community then makes own decisions
Exogenous effects womans CEB affected by exogenous observed mean
characteristics of reference group e.g. education av. level may have some
impact
Correlated effects women in same group tend to behave similarly as face
similar institutional environments or have similar indiv. characteristics e.g.
priest in community says shouldnt use some contraceptive type
Social Interactions and fertility behaviour in Kenya (Weeks and Iyer 2005)
TFR (total fertility rate) fallen from 6 births per woman in 1948 to approx. 4.5 births today
(via. 6.7 1989, 5.4 1993 and 4.7 1998)
TFR = age between 15-49, how many kids expected to have on average
Drop in fertility considered one of most dramatic recorded anywhere in developed or
developing world fallen very sharply over very short period of time
Ethnic group identity important in Kenya
42 different ethnic African groups in Kenya
5 largest = Kikuyu, Kamba, Kalenjin, Luhya and Luo
Add Kisii, Mijikenda-Swahili and Meru-Embu groups, constitute 90% of population
Live in 7 distinct provinces of Kenya Nairobi, Central, Eastern, Rift Valley, Western, Nyanza,
Coast further divided into 47 districts
Anthropological studies
Desired family size smaller among Kamba, Kikuyu, Luhya compared to Luo
Luo exhibit strong preference for large families and use contraception much less than other
groups and dont engage in some traditional practices that may inhibit fertility relative to
other communities
Overall, massive variation in sample for CEB by region and ethnic group
o 2.5 per woman (Kikuyu) to 3.5 per woman (Luo)
o Luo in Nairobi, 1.6 per woman, Luo in Nyanza, 3.745 per woman, Kalenjin who live
near cost, 9 per woman!
MODEL: SOCIAL INTERACTION AS APPLIED TO FERTILITY BEHAVIOUR
Approach social interactions occur at number of specific levels
Each indiv. i (belonging to household h), resides in set of locations called clusters (village or
group of houses) j
Each indiv. also member of ethnic group s and can be described by vector of characteristics
Use notation and to denote pop. of groups s and j which exclude i
Assumption strategic complementarity
= MU accruing to fertility decision of woman i is increasing function of fertility choices of all
other indiv. in cluster j and other indiv. belonging to type s
So an increasing function of and (means) i.e. womans fertility is increasing function of
mean fertility in cluster they reside and mean fertility of their particular ethnic group
Incorporating social interactions into household D model
Point of departure = Beckerian U function
( )
i i i i i
x z n U U , , ,
where z = sources of satisfaction apart from kids, x = social demographic characteristics that
may affect preferences and = error from incomplete info
Part IIB Revision, Stephen Ward 99 Development
s.t. budget constraint
z i n i i
P z P n I +
Extend atomistic U model to include social interactions
( ) ( ) ( ) ( )
i
s
i i
s j
i i
j
i i i i i
n n S n n S x z n U V , , , , ; , ,

Can rewrite as additive in individual and social U terms


( ) ( ) ( )
i
s
i i
s j
i i
j
i i i i i
n n S n n S x z n U V + + +

, , , ,
o Interaction may be localised or diffused over wider area
Allows to test whether after controlled for effect of social interactions on account of
geography, is there some residual social interaction on account of ethnicity?
Local and global interactions
Cross partial effects will be key objects = S.O.C
( )
0
,
2
>


j
i i
j
i i
j
j
i
n n
n n S
k
( )
0
,
2
>


s
i i
s
i i
s
s
i
n n
n n S
k
Local interaction indiv. have incentives to conform to behaviour of small number of
neighbours whose actions they can observe
Global interactions indiv. have incentives to conform to expected behaviour of large common
reference group (e.g. religious or ethnic group) may be sanctions if deviate from certain sorts
of ethnic norm
Simultaneous interaction having more than one social utility term allows for multiple social
interactions
For each women, have count of total number of CEB and estimate Poisson Regression Model (as
children discrete, not continuous if use OLS, get bias) also use Poisson as dependent variable heaped
on certain values i.e. lots of women have between 2 and 5 children with very few having 10
Excess zeros
o 26% of women had no children but includes women who have no children and women
who cannot have children
o Zero-inflated models where inflation refers to excessive probability mass at zero
Must account for fact woman may have no kids as cannot or because doesnt
want to have children
Truncated fertility lifetimes
o Include age and age squared to account for fact women in previous generations may
have had more kids
Conditional Mean Specification
( ) [ ] [ ] ( ) ' ' ' ( ( exp , , ,
l
cl
c e l
x k p h yed cl i C f E e i C f E e e p h x C E + + + + + +
C is expected number of CEB
E denotes categorical variable denoting ethnic affiliation; cl is the cluster

[ ] e i C f E (
= some function f(.) of number of CEB at level of ethnic group
(constant within ethnic group but allowed to vary across them)

[ ] cl i C f E (
= some function f(.) of number of CEB at level of the cluster
Part IIB Revision, Stephen Ward 100 Development
Social U term for j = some function of
other womens choices of number of
children on this woman = local
interaction
Impact of every other womans
choice in s on this particular
woman = global interaction
represents measure of particular utility that accrues to I from not conforming to
everyone else (deviating) in particular location some incentive to be uniform
same for ethnicity if dont conform to norm, suffer some disutility from doing so
x
l
, h, p denote vectors of individual, household and cluster attributes
, k, are associated vectors of unknown control parameters

e
,
c
and are scalar parameters
i.e. CEB is function of several factors
Endogenous effects:
o Global effect of ethnic population
Based essentially on conformist social U function that exhibits strategic
complementarities but is constant within ethnic group but allowed to vary
across groups
o Conformist effects in clusters
Effect for geography
Exogenous effects:
o Education important in Kenya due to the presence of harambee (community self-help)
schools
o Family planning
Can test using this specification whether after controlling for indiv. attributes, cluster-level
attributes, household attributes are there endogenous and exogenous effects on a womans
fertility behaviour?
Data from Kenya Demographic and Health Survey 1998
Demographic and Health Survey (KDHS) conducted in Kenya 1998
Survey contains interview data for 7800 women aged 15-49
35 out of total of 42 districts sampled; 536 clusters of which 444 rural (only look at these)
Systematic sample drawn from clusters of on average 17 households in each rural cluster
Consider 5994 ever-married women living in rural clusters only
Women from all of Kenyas large ethnic groups and covers wide geographical area
Results
e

= parameter estimate proportional change in number of children born to change in


regressors
Endogenous effects
o Localised endogenous effects captured by parameter which weights difference
between mean level and individual CEB for women in same cluster
Significant only for Kalenjin (positive) and Kikuyu (negative)
o Endogenous effects with respect to ethnicity hugely significant for everybody
For Kikuyu, estimated effects considerably greater than for other ethnic groups
o So Kikuyu positive at ethnic level but negative at level of cluster
Says maybe Kikuyu women observing other women in area and seeing the
negative impact on pop. so at local level having neg. impact on their fertility
but Kikuyu norms overall may actually be norms that are increasing fertility
Exogenous effects
o Education effect significant for 2 ethnic groups: Kamba and Kisii size of multipliers
37% and 55% for Kamba and Kisii = limited evidence for Glaser-Shiteman hypothesis that
if social interactions are an important inclusion in metric models, then size of
multipliers going to be large
o Exogenous education effects are strong even after controlling for indiv. levels of
educational attainment
Other effects
o Cluster-level controls
o Household-level effects and controls
o Indiv.-level effects: significant ethnic effects
Part IIB Revision, Stephen Ward 101 Development
Inflation Sub-model
Inclusion of ethnic dummies here with Kikuyu as base ethnic group as infertility may vary across
ethnic groups due to various traditional practices
Findings show only group for which observe significant ethnic effect is Luo i.e. smaller number
of infertile women in this group
o Many practices Luo didnt practice compared to som other groups
With strong preference for large families, may explain why fertility is highest of any Kenyan
ethnic group, since they have lower levels of infertility
Conclusions
Multiplicity of social interactions for fertility transitions some societies essentially stuck in
high-fertility equilibrium that does not max welfare
Local and global interactions may be important in form of ethnicity and geography
o Notable that once included ethnicity in model, effect of geography was less significant
implies if dont include multiple levels of interactions, may be making wrong
inferences
Exercise caution when drawing inference from social interactions
Part IIB Revision, Stephen Ward 102 Development
Lecture 5 Mortality and Missing Women
Key trends in mortality and life expectancy
Mortality rates fallen dramatically over time
Mortality rates higher in poor countries
LE increased by about 30 yrs in past century
Difference in LE today between rich and poor countries of about 30 yrs
Changes in LE affected by HIV/AIDS epidemic and transition in Eastern Europe
Positive relationship between income and health within countries
Age-related declines in mortality due to decline in spread of infectious diseases
Much can be explained by huge fall in mortality but falls in mortality that underlies figures not
uniform by age
o Life expectancy for age group under 10 only increased by half the amount
Death from infectious disease high in 1948
Positive relation between income and health within countries
o 1980 US bottom 5% income dist. had 25% lower LE than top 5%
o Now in US, blacks have LE 4.5 yrs less than whites
Differences remain despite improvements in health etc.
Affected by epidemics like Aids etc
o SubSahAf LE now less than was in 1950s
o Russia, LE fell 7 yrs in the 1990s
Child mortality
10.5m children under 5 die every year in world (1950 was 17m)
Child mortality rate = probability of death under 5 yrs (Infant mortality rate under 1 yr)
Regional variation in child mortality rate (Africa = 170/1000 live births, Americas 25, SEAsia 80,
Europe 20, Eastern Med 90, West Pacific 30, World 80)
Nearly all child deaths occur in developing countries, almost half in Africa largely as result of
HIV/AIDS epidemic
o Many gains outweighed by HIV/AIDS
Of 20 countries in world with highest child mortality, 19 in Africa (exception Afghanistan)
Child in Sierra Leone 3.5x more likely to die than in India, 100x than in Iceland or Singapore
Generally higher among males than females
o 4 significant exceptions China, India, Nepal, Pakistan, where mortality of girls
exceeds boys = gender bias. China 33% higher
In India and China, while child mortality declining, most significant cause of child mortality
perinatal deaths = trauma around birth
Considerable variation in child mortality across different income groups within countries
o Kids from poor households have significantly higher risk of dying before 5 than richer
Current situation in Africa complex
o Surge in HIV/Aids in children in SubSahAf led to 332,000 deaths as consequence of
inheriting
Niger poor 13% points higher (at 34%). Most well off in Niger have double prob. of non-poor in
Americas (Bolivia) and still higher than the poor there
Link between child mortality and fertility
Child survival hypothesis idea that if parents expect to lose children to child death, going to
replace those children very quickly one reason why higher mortality may contribute to higher
fertility = Ex-ante response
Part IIB Revision, Stephen Ward 103 Development
Child replacement hypothesis if lose child to death, parents tend to replace child quickly =
Ex-post response
Both hypotheses lead to higher fertility if societys levels of child and infant mortality are high
o Uncertainty of survival affected by priv. & mkt. risks to women e.g. weather,
climate, services (health) etc.
Relationship significantly influenced by level of uncertainty and risk, both priv. and mkt. risks
Religion, Fertility and Child Mortality in India
Differences in fertility rates by religion in India very pronounced
Recent political controversy about 2001 Indian Census: 36% reported growth in Muslim fertility
between 1991 and 2001
o Obvious Hindus becoming more and more conscious of importance of small families.
Going to tilt pop. balance in favour of Muslims, who are merrily producing more and
more kids (Hindu politician 2004) True?
Hindus about 80% of Indias pop., Muslims about 15%
Muslims do on average bear 1 more child than any other group in India at all age groups from
15-49
o Average 2.8 for Hindus, 3.6 for Muslims
Pure Religion Effect or Particularised Theology Hypothesis
o Gods will for us to have children girls for the aarti (lighting the camphor flame for
prayer) and boys for keerti (family prestige) (Hindu mother of 4, India)
Characteristics Effect
o Gods will is all very well but who is to support the larger family? (Muslim mother of
5, India)
Son Preference and fertility hypothesis
o Son preference may explain religious differences in fertility
Paradox about religion and infant mortality in India
Infant mortality among Muslims (59 per 1000) much lower than infant mortality among Hindus
(77 per 1000)
Child mortality among Muslims (83 per 1000) much lower than child mortality among Hindus
(107 per 1000)
Fertility much higher among Muslims but mortality much lower = counter-intuitive and a huge
puzzle
Missing Women
Shown by sex ratio or number of females per 1000 males (Sen 2001)
o India: 933
o Europe and North America: 1050
o SubSahAf: 1022
Number of missing women in India estimated to be between 35 and 36m (Dreze and Sen 1996,
Klasen 1994) & 40m in China
Sex ratio at birth in sample differs considerably by religion
o Hindus: 976
o Muslims: 1026 look after girls better than Hindus?
2 Demographic features about India
Higher fertility rate among Muslim women than Hindu women
Gender bias due to son preference, leading to missing women phenomenon (Dreze and Sen
1996)
Part IIB Revision, Stephen Ward 104 Development
Model Assumptions
Extend son preference to complementary concept of daughter aversion
o Sons bring benefits and daughters impose costs
Complementing desire to have sons is desire not to have daughters
Desire for sons increases family size; fear of daughters limits it
Model where these 2 act to determine equilibrium family size and composition
Total number of children in family (N) composed of S sons (positive utility from) and D
daughters (negative utility from) so N = S + D
Let B(S) and C(D) represent benefit and cost functions associated with sons and daughters
where
0 >

S
B
, 0 >

D
C
0
2
2
>

S
B
, 0
2
2
>

D
C
Marginal expected utility associated with another child is positive (or negative) where
( ) ( ) ( ) D C S B EU ' 1 ' +
Assume , the probability of having a son is 0.5
Family decide to have another child iff
( ) ( ) D C S B ' ' >
And decide against another child iff
( ) ( ) D C S B ' ' <
An equilibrium number of children is one to which the family does not wish to add
Lower degree of daughter aversion and son preference will alter equilibrium family size and
composition
Part IIB Revision, Stephen Ward 105 Development
D*M
Sons, Daughters
MB/MC
SZ = DZ
MB
MC
No sons equilibrium
Z
B
X
DX
Parity Equilibrium
All other equilibrium = non-
parity
H
Sons, Daughters
MB/MC
MB
0 son
equilibrium
B
M
S** D**M
D**H
DH
DM D*H
H
Sons, Daughters
MB
0 son
equilibrium
B
M
DM
S*H
MB/MC
S*M
H
Sons, Daughters
MC
0 son
equilibrium
B
M
DH = DM S**M D**
S**H
No son same in both
cases
Parity equilibrium
greater with higher
son preference
For given
daughters,
equilibrium sons
higher with higher
son preference
Zero son
equilibrium
greater with
lower
daughter
aversion
Parity equilibrium
greater with lower
daughter aversion
too
Given sons, equilibrium
daughter greater with
lower daughter
aversion
Muslims always have family size
bigger higher daughter aversion
Daughter Aversion is represented by a shift in the MC curve
Son Preference is represented by movements of the MB curve
With similar son preference but lower daughter aversion, Muslims will have larger equilibrium family
size and more daughters than Hindus
Plotting all equilibrium combinations in this son-daughter space
Suppose HH is equilibrium locus for Hindus represents all combinations Hindu family will adopt
Suppose with daughter aversion, Muslims have same degree of son preference but lower daughter
aversion, then Muslim equilibrium locus represented by MH. In equilibrium, Muslim fertility greater
(OM) than Muslim fertility (OH)
If have situation where Muslims have same degree of daughter aversion but lower son preference than
Hindus, Hindu equilibrium locus represented by HH. In equilibrium Muslims going to have smaller
fertility (OM) compared to OH for Hindus
Data from 1993-94 Human Development Survey of India used
Unit record data on 33,000 households, 195,000 individuals, in 1765 villages, from 16 Indian
states. Of these, use data on 10,548 currently married women in sample who adopted terminal
method of contraception
4.6% of Hindu women rel. to 5.1% Muslim women terminated fertility without having any sons
19.6% of Hindu women re. to 13.4% Muslim women terminated fertility without any daughters
Control variables used: number of living sons/daughters; age at marriage; religion and caste;
educational attainment and occupations household characteristics; village-level facilities;
region
Method
Measurement
o Measure of son preference: 1 proportion of women who terminate fertility without
any sons
o Measure of daughter aversion: proportion of women who terminate fertility without any
daughters
o Male infant mortality rates: male infant deaths as % male live births
o Female infant mortality rates: female infant deaths as % female live births
Poisson Regression Models
o Explaining living sons and daughters
2 dependent variables: number of sons and daughters
Explain number of living sons and daughters in terms of personal, household
and region-specific characteristics
Part IIB Revision, Stephen Ward 106 Development
Sons
Daughters
M
H
H
Muslims have higher families as dont desire sons as much or
less worried of daughter consequences have less sons and
more daughters
Equilibrium Locus
M
0
o Explaining infant deaths
2 dependent variables: number of male and female infant deaths
Explain number of male and female infant deaths in terms of personal,
household and village characteristics
When reporting Poisson estimate, simply associated marginal effects when value of all other controls
being set to their means
Results
Degree of son preference and daughter aversion
o Difference in son preference not statistically different between Hindus (0.954) and
Muslims (0.949)
o Daughter aversion greater for Hindus (0.196) than for Muslims (0.134) and statistically
significant
Explaining number of living children
o Religion: Muslims have 0.27 more sons and 0.34 more daughters than Hindus
o Literacy: Women who literate have 0.22 fewer sons and 0.11 fewer daughters than
women who illiterate
o Region: living in S, E and W resulted in smaller number of sons than in N (default
region) or central region
o Occupation: Women who worked as labourers or cultivators had in equilibrium a
smaller number of sons, but larger number of daughters
Infant mortality rates if really want to buy theory, must also look at what going on with
respect to mortality. If Muslims less adverse to daughters than Hindus, must be taking better
care of them particularly expect that female infant mortality should be much lower form
Muslims than Hindus
o Male infant mortality rate not very different between Hindus (4.5%) and Muslims (4.7%)
and not significant
o Female infant mortality rate much higher for Hindus (6.3%) than Muslims (4.6%) and
highly significant
o Small gender gap in Muslim infant mortality rates but considerable gender gap for
Hindus
Even if Muslim parents have more kids, maybe looking after them better too
Explaining number of infant deaths
o Religion: Muslim women had less infant deaths
o Safe drinking water: reduced male and female infant deaths by respectively 4% and 8%
o Anganwadi schools: reduced male and female infant deaths by respectively 5% and 7%
o Poor housing quality: increased male and female infant deaths by respectively 7% and
15%
o Maternal occupation: labourers have on average 15% more male infant deaths and 9%
more female deaths
o Maternal literacy: literate mothers have on average 17% fewer male infant deaths and
23% fewer female infant deaths
o Paternal literacy: reductions of 4% and 7% respectively in male and female infant
deaths
o Even after controlling for all these factors: still Muslim women have lower infant
deaths If Muslim mother rather than Hindu, would have reduced male deaths per
woman by about 23% and female deaths by about 10%
There is statistical evidence that sex selection at birth among Hindus, and Muslim mothers may
actually be treating children better than Hindu mothers
Genetic explanations for missing women
Oster (2005) genetic explanations may explain missing women
o Carrier of hepatitis-B virus have sex ratios of 1.5 boys for each girl
Part IIB Revision, Stephen Ward 107 Development
o Since many countries with missing women also have relatively high hepatitis B carrier
prevalence, naturally occurring adverse sex ratio at birth could produce higher
population sex ratio even in absence of excess female mortality
o Virus especially common in China
o Hepatitis B can account for approximately 45% of missing women: 75% in China but very
little in India
Conclusions
Child mortality is a primary cause of high fertility
Child mortality varies by gender and income
For many developing countries like India and China, problem of missing women is key
demographic problem
Causes of missing women are due to greater son preference
Empirical evidence on infant/child mortality in India seems to suggest that sex selection at
birth is going on, and may in fact account for differences in fertility between different religious
groups
Genetic explanations for missing women are the other possibilities, especially in China
More research must be conducted on mortality and missing women
Skewed gender ratio in India/China
Biological and cultural explanations
o Strong culture of son preference variety of factors make females less valuable to
families
Households resorted to female infanticide and postnatal withholding of
healthcare
Since tech. available to determine babys sex pre-birth, shift towards prenatal
selection through induced abortion
o Biological (Oster 2005) women infected with hepatitis B virus give birth to male
children in higher proportions than norm
Das Gupta (2005) shows data consistent with cultural explanation rather than biological
o Oster study important as reminds of importance of biological factors in affecting sex
differentials in health outcomes
o However, ev. indicates parental preferences overwhelmingly shape female deficit in
South and East Asia
Sub-Saharan Africa more favourable ratio with 1,022 women per 1,000 men
Lower child mortality in Africa for daughters than sons naturally there
Less Hep. B than China
No Hinduism no daughter aversion
Culture different women very active in LF and many subsistence crops farmed by women
30% households headed by women and large amount (30%) of kids live away from parents
Cost of having daughter not significantly higher than son daughters not sold for marriage and
females dont leave family once married
Lack of easy and safe abortions that allow terminations of female foetuses
No abundance of tech. that allows couples to find out sex of baby before birth
Shows missing women not necessarily a result of a poor country
Part IIB Revision, Stephen Ward 108 Development
Lecture 6 Current Policy Debates about Demography in Poor
Countries
Economists focus on altering desired fertility i.e. number of children couples want
Demographers focus on reducing unwanted fertility i.e. number of children couples dont want
Characteristics of global pop. pol. over time
Pop.pol. initially only provided family planning services many countries pursued pop. pol. as if
only family planning involved
Other dev. Policies related to pop. largely neglected many countries only paid lip-service to
importance of female educ., reducing infant mortality, empowering women etc.
Shift in focus of pop. pol. since International Conference on Population and Development
(ICPD) in Cairo in 1994
o Shift focus from theories towards womens reproductive health
By mid 90s, 127 countries (94% of worlds pop.) officially supported family planning to some
degree
Successes and failures cases of China and India illustrate roles played by pop. pol.
Case of Sub-Saharan Africa illustrates HDM model and role of externalities
Case of MidEast illustrates role of social norms (e.g. religion)
India and China illustrate all models used so far
Between 50s and 90s, Chinese fertility been halved while India declined by
Chinese mortality less than 1/3 its previous level; India more than
Demographically, India will overtake China in mid 21
st
century Chinese pop. growth at
replacement level so pop. will be stable. Indian fertility shows rapid pop. growth and that is
going to keep continuing. So between 1990 and 2020, Indias pop. will increase by about bn.
Indian population policy 4 phases of evolution
Pop. pol between 1952-76
o First 2 Indian 5 yr plans (51-61) called for creation of family planning centres also
called for birth rate of around 25 per 1000 by 1984 this goal not met
o Today % couples using birth control still low (36%), main form of which is sterilisation
operation for women
o Success limited to South Indian states, upper classes and urban population
Coercive family planning 1976-80
o Declaration 16/4/76
o Decision to speed up the programme
Number of measures for family planning
Financial incentives for participants
Coercive sterilisation enforced sterilisation after 3
rd
child (never enforced
properly)
o Even today sterilisation most popular method of birth control most common age in
India is 25
o Disastrous consequences for family planning programme very large set-back and had
big impact on reputation of family planning programme
Renewal of demographic pol. in early 80s
o 1981 census revealed considerably larger pop. than expected
o 7
th
5yr plan called for replacement fertility by 2000
New approach to pop. pol. from 1997 to present day
o 1997 Reproductive Health Programme recognises pop. pol important
Part IIB Revision, Stephen Ward 109 Development
o Recognition that cookie-cutter approach to pop. pol. not ideal given Inisa ethnically
and linguistically diverse
o 2000 New Population Pol.
3 objectives Immediate (address unmet need for contraception and provide
basic reproductive and child health), medium-term (bring down fertility rate to
replacement level by 2010), long-term (stable pop. by 2045)
Strategic themes simultaneously: empowering women; health and nutrition;
adolescents; IEC (information, education and communication) very much
those suggested by HDM
Role for local self-govt. and voluntary sector (e.g. village panchayats)
Targets all aspects of social and econ. development (e.g. Anganwadi schools)
Population policy in China concern over pop. after results of 1953 Census published can identify 4
phases
Mao and 1
st
birth control programme
o Creation of assistance network
o Production of contraceptives
o Plan to encourage pop. to use birth control services
1958-59 Great Leap Forward
o High mortality and famine 1959-61
o 2
nd
birth control programme launched by dept. of family planning though suspended
during cultural revolution. Only with return of normality in 1971 that 3
rd
campaign in
China began
1971 the 3
rd
campaign
o later (marriage i.e. 23yrs), longer (birth intervals 4 yrs between 1
st
and 2
nd
birth),
fewer (children 2 in city, 3 in rural later in 1977, 2 in rural too)
o Diversity of methods used
Sterilisation but also IUDs and steroids
Abortion easily available, free and does not require husbands consent
Pop. pol. in 1980s and 1 child pol.
o Pop. goal set in Sept 1980 not to exceed 1.2bn by 2000
Chinese 1 child pol.
Incentives and disincentives introduced
1 child certificate, issued by local authorities, guarantees benefits for couples such as wage
and pension increases, larger dwellings, priority in schools and free medical care in exchange
for commitment to have one child
Disincentives in form of W cuts for >1 child
Until 1983, pressure increased as coercive methods implemented on vast scale
Subsequently, greater flexibility progressive extension of right of rural couples to have 2
nd
child when firstborn is a girl
Exceptions to 1 child rule introduced in 80s for ethnic minorities, mountainous areas, and
couples in special situations great flexibility with which pol. applied
Regional differences in Chinese fertility first declines were in Eastern provinces
Disadvantages of 1 child pol. local coercion and missing women (abortions due to target
setting and concern for old age security
Comparing Indian and Chinese demographic policies
Chinese social transformation proceeded more quickly particularly in public health care as
result seen large mortality declines much more rapidly than in India
Authority of Communist party allowed quick execution of demographic objectives
Efficient dist. networks employing variety of birth control methods, including abortion
Part IIB Revision, Stephen Ward 110 Development
Chinese society more receptive to fertility limitation take examples from Singapore and HK,
where ethnically linked social interactions?
Pop. trends in future in India and China
Consider age structures for China and India in 1950 and 2025
o 1950 China had larger population in all age groups
o 2025 projected (with current demographic pol.) China smaller in all age groups up to
30-35
Between 1950 and 2025, pop. of India will quadruple while China will increase by factor less
than 3
Key factors affecting fertility in Sub-Saharan Africa
High regional birth rate approx. 6 live births per woman growth rate of pop. about 3% p.a.
Fertility decline in evidence but not universal
Rural-urban dist. of pop. varied across countries in this region
Concept of household different on account of practice of polygamy marriage early and near
universal, widow remarriages commonplace very different to S Asia
Paradox of low female educ. and high levels of female work activity female literacy 53% but
LF participation 62% - HDM broken down as very low rates literacy but very high rates of econ.
activity
o Because primary responsibility for raising subsistence crops in this part of the world
resides with women and 30% of rural households in Saharan Africa actually headed by
women
Fosterage widely practised in much of W. Africa, 1.3 of children found to live with Kin at any
time and nephews and nieces have same rights as biological offspring
o Though costs of children shared, benefits not being diffused across kinship and may be
one reason why have very high fertility across sub-Saharan Africa
o Very strong strategic externalities may be powerful stimulus for higher fertility in this
region
Contraceptive prevalence and use low (only about 20%)
Significant demographic threat to region is incidence of HIV/AIDS and high child mortality
o Though been restricted to Eastern side of Sub-Saharan Africa, predicted to increase and
spread in coming decades have very important consequences both for mortality and
fertility
1960-99 Kenya and South Africa leading countries
Apart from SA, most people urban dwellers rates vary much across countries and time
12/23 countries have fertility less than 5 kids per woman
Key pol. variables to target to reduce fertility in Sub-Saharan Africa
Pol. that increase costs of having children e.g. education
o Few countries where less than girls and 1/3 boys make it past grade 4 HDM
important in reducing fertility here
Currently
o 40% 15-19 yr olds married
o Women allowed more than 1 relationship
o Female literacy very low but high labour force activity
Primary subsistence crops farmed by women
o 30% of households headed by women
o 1/3 kids live away from parents
o Nephews and nieces have same rights as offspring
o Effective family planning absent
Contraception availability only 20%
Part IIB Revision, Stephen Ward 111 Development
o High mortality and Aids in East Sub-Saharan Africa
Expect move West over nest few decades
Information about tech. of reproduction
o E.g. socially coordinated effort, using media more effectively
Factors that affect fertility behaviour in Middle East
Role of religion in fertility behaviour (also important in India)
o North Africa and West Asia = of worlds Muslim population
Fertility trends in Arab countries (about 4.2 today)
Diversity of Arab economies in income and educational attainment some own oil, some poor.
Educ. at primary level close to 100% but Somalia and Morocco still lag behind
Contraceptive use varies considerably across region (e.g. Egypt, Saudi Arabia) but tends to be
quite low. Egypt has widely developed FP programme
Unique model of fertility that exists in Arab world
o Arab fertility demonstrates positive correlation with development indicators
o Mortality trends and substantial health investments in Arab nations
o Evidence against household D model positive correlation between fertility and
development due to sub-ordinance of women in Muslim culture
Big health infrastructure improvements
Why fertility lagging behind mortality in Arab nations?
Pure religion effect: theology influences fertility but Obermeyer argues inadequate Islam
does permit family planning in certain situations
o Many allow abortion up to point en-souled could be 40
th
, 80
th
, 120
th
day
o Permitted if affects wifes health or beauty
Characteristics effect: socio-economic differences influence fertility (more likely)
o Women have restricted access to educ. and employment in some countries (Yemen)
Institutional context of fertility in Arab world (e.g. gender relations, political structures)
Region may be more important than religion
o Combo of econ. and politics determines if fertile stays high here
o Islamic explanation ignores diversity of region. Not fully sufficient
Trends in Arab fertility in future not rapid nor uniform
Influencing calculus of conscious choice with pol. (Cole)
o Tgt. Supply of family planning or D for children
Fertility behaviour related to econ. growth and human development
o Take care in old age
o Women increasingly sent to education and access to family planning improving plan
number of kids and the yrs between them
Conclusions
Targeting D for children vs. supply of family planning
Is pop. growth in 21
st
century still the principal difficulty in the way to societies improvement?
Lessons to be learnt from demography, economics, econometrics and history
Bottom line on pop. question is own judgement
Part IIB Revision, Stephen Ward 112 Development
Financial Liberalisation
Agents are intelligent and selfish, and operate in mkts. No govt. activity, simply make sure fair
Equilibrium Global = pareto optimal (or efficient) from perfectly competitive mkts.
Can have local equilibrium as well
Outcome of fin. mkt. could be crisis
o Could result from excess liquidity
o Borrowers and lenders accumulate excessive amount of risk privately which leads to
socially
Excess liquidity
1980 value of Int. Fin. stocks $12tn i.e. about world GDP
1990 - $43tn i.e. about 2x world GDP
2000 - $94tn i.e. about 3x GDP
2006 - $167tn i.e. about 3.5x GDP
1980 Definitely not characterised by shortage of liquidity (Mackinsie)
IMF annual meeting Countries used to queue at banks to negotiate terms. Now bank CEOs coming to
countries
K inflows characterised by long periods of low levels followed by short peaks of high inflows
Developing countries = mkt. of last resort stampede away when crisis, but now cycles
between crisis and access getting closer
Never been case where fin. mkts. want to lend to developing countries but cannot find a taker.
Other way round not always someone willing to lend = supply problem
Sudden shifts behind crises
o OPEC 1980s lots of money, dont know where to put it give to developing countries
Latin America first used only bank loans, then only FDI. Composition has shifted
1995 massive inflow, from 0-100 how absorb? In ER or interest rates?
o Then fell to zero adjust 1 way, then the other way again
How deal with? Open K account fully? Expose fully to raw cycles with no capacity to affect?
Policy?
Common characteristics undertaken most radical fin. liberalisation and had liberalised K accounts and
domestic fin. sectors at time of high liquidity in int. fin. mkts. i.e. when rapidly growing, highly
volatile and largely under-regulated int. fin. mkt. anxiously seeking new-yield investment opportunities
Part IIB Revision, Stephen Ward 113 Development
In Latin America, econ. reforms undertaken as response to perceived weakness in previous model of
development. Tried to reverse as many aspects as possible of previous system (neo-liberal or neo-
idiotic Franco Head of Brazillian CB)
In East Asia, reforms implemented as pragmatic mech. to strengthen existing development process
Surge in inflows
Common to all crises
Extraordinary turnaround from position of K outflows (due to dept repayments) to massive K
inflows
Brazil differences amounts to $220bn following fin. liberalisation. $150bn in Mx and Arg,
$260bn in E.Asia
Arg net inflows in 1993 at level equivalent to 154% X and 1999 (before 2
nd
crisis) net FDI
inflows equivalent to 50% of countrys investment
Chile net inflows reached levels similar to X
Some of these countries became players in newly developed derivatives mkt.
Kindelbergian view large surge in inflows and in particular its effect on domestic liquidity is key to
understanding all 3 routes of fin. crisis
Factors in K flows
Push = int. mkts. discovering developing countries excess liquidity in int. mkts., business
cycle conditions, changes to interest rates, rise and international diversification of institutional
investors, demographic forces in ind. countries
Pull = things happen in developing countries make them more attractive reforms since 1980
e.g. Financial Liberalisation and the opening up of the K account, stock mkt. liberalisation and
privatisation, high interest rate spreads and expectations of ER appreciation
o China more risky but not open K mkts. Could control who come in and out and timings
Optimism regarding source of econ. reforms in developing countries fuelled by spin of those
involved in Washington Consensus
Moral Hazard (pull) created by near certainty of fin. mkts. that vast international rescue
operations would bail out in occurrence of defaults
Big effect of opening K mkt. e.g. Brazil 300m = big amount of money. Similar for other
countries
Policy
Close K account?
Open fully? Partially as India, China and Taiwan have?
Open for foreigners and domestic?
o Open fully for them or just slightly?
Chile put tax on portfolio tax excess = P control
o Rogoff (IMF) says this the way forward previously never said anything for regulation
was always against
Summers (Harvard) open K mkt. = like building new power plant with no safety bars
o = only mainstream theory on side of regulation
Problem of financial mkts. unique and associated with liquidity
Banks handle affairs easily when lend to people who dont need to borrow can deal with it. When
lend to people who need to borrow, have problems e.g. sub-prime lend to people who know will
never be able to pay them back
= what financial mkts. are when excess liquidity
Part IIB Revision, Stephen Ward 114 Development
Why whenever liquidity wants to flow, someone there to receive it?
Kindleberger Nothing that lifts animal spirits more than availability of finance
o Availability of finance lifts expectations of future returns
o Self-fulfilling prophecy give credit, people borrow/spend, leads to increased
consumption and investment, increases wages and employment in SR but then
vulnerabilities rise. Balance of payments, inflation, ER etc. Effect of asset P equity P
for example
Sub-prime key rationality will work so long as house P increase forever. Low interest rate for 1
st
few
yrs. After 5 yrs for example, rate goes up lots. Asset P = driver key magnet that brings people into
finance
Excess liquidity in Int. Fin. Mkts.
IMF data value of assets of institutional investors increased 1988-1996 from $12-$26tn Anglo-
Saxon countries stand out most ratio of assets to GDP grew by 80% of GDP in UK
Palma (2003) exposure of institutional investors to LDCs so small proportionately they
believed it would not pay to invest adequately in information
Brady Plan
Became clear the negative transfer of resources of magnitude taking place from developing to
developed world could not continue much longer without causing social unrest and pol.
instability default and repayment not acceptable
Prompted implementation of Brady Plan scheme that allowed debtor countries to issue Brady
bonds in int. mkts. in order to re-fin. debt to commercial banks
How access to liquidity led developing countries to bring it in initial mechanism was credit to the
private sector
Banks/building societies make credit available to individuals
o Lower transaction cost of getting finance. Can get 0 transaction cost on internet etc. or
a telephone call. Used to be able to get max 75% mortgage. Now can get 120% (not
anymore)
o Use same stocks that given big return as collateral for future borrowing even more
problems when burst. Collaterals begin to collapse too
Latin American Story
Inflows volatile in size but also composition bank lending increased sig. up to 1982 debt crisis
then increases to even greater heights during 1990s mainly through portfolio investment and
speculation
In Arg., turnaround equivalent to 2.4x X 1989-93
Response to surge in inflows was to ride them out by unloading into economy via credit
expansion. Mexico, bank loan portfolio grew from $55bn in 1989 to $200bn in 1994. Brazils
response was opposite iron curtain around inflows to dampen expansionary effect undertook
high degrees of sterilisation and high interest rates in attempt to avoid type 1 crisis
3 routes to financial crisis
Part IIB Revision, Stephen Ward 115 Development
I
Cred
it
Interest
Rates
Cons./
Asset
Bubble
Route 1 - Mexico
I
Cred
it
Interest
Rates
Cons./
Asset
Bubble
I
Cred
it
Interest
Rates
Cons./
Asset
Bubble
Route 2 - Brazil Route 3 - Korea
Special vulnerabilities
Route 1 rapid revaluation of RER, explosion of credit to priv. sect., cons. booms and asset
bubbles and massive deterioration of CA e.g. Mexico
Route 2 high interest rates, mainly result of high levels of sterilisation to defend peg and as
poor sub. for public sect. reforms, helping create fragilities in banking system and in public
sector fin. and deterioration in term structure of debt
Route 3 corporate debt/equity ratios increased to remarkable heights
Route 1:
Mkts. know best
Let it be
Any interference counter-productive
Almost all liquidity want to consumers huge effect on the private sector
Mixture of endogenous pull, due to fact foreign K needed because of inability of LatAm
countries to generate sufficiently large trade surpluses so as to be able to service foreign debt
(led to Brady plan) and exogenous push exogenous movement of foreign K into these countries
Route 2:
Mkts. know best but something special about finance
Full sterilisation
o Give CB money (1m), increase reserves by 1m or increase money by 1m. If increase
money by 1m, gives liquidity to banks which is circulated
o Sterilisation CB further activity e.g. issue bonds for 1m and sell in open mkt. cancels
out the increase in the money supply
o Public expenditure effect must pay interest on bonds each yr some liquidity injected
o Brazil had zero credit expansion
Brazil had low credit to the priv. sector
o Almost declined
o Must think that Brazil opened up K account roughly at time of Mexico crisis of 94
Nothing happened but high interest rate no bubbles
Route 3:
Almost all inflows financed investment. Massive corporate debt
About 30% of priv. investment in Korea unsustainable
Crisis was over-investment did too much of a good thing
But Koreas private investment high but stable, so why now over invest?
o What changed was financing of investment not invest more than before. Massive
switch from own finance to external finance (from profits to loans)
Endogenous pull additional fin. required to sustain high levels of investment at time of rapidly
falling profits
Profits of Korean companies were collapsing, initiating switch to loans from profits due to the
fall in the price of microelectronics due to over supply and technological change
o With lower profits, why not decrease investment? Microelectronics must be at cutting
edge of mkt.
Require top investment (Taiwan gone from 4MHz to 16MHz processor) must
invest
Caught in mkt. and must invest what need, or go out of business
Part IIB Revision, Stephen Ward 116 Development
How make sustainable the resulting BoP problems?
Agents on household side accumulate excessive risk?
Why macro unworkable with massive asset bubbles and ER fluctuations?
Above 2 not mysterious what is mysterious is why Brazil and Korea ended up in the same crisis
but through a different mechanism
1
st
vulnerability of Brazil on banking side banks went bust
o Brazil began to rescue bank after bank
Took over and absorbed debt and sold on good will of bank
o Crucial issue not banking crisis problem was public sector. Domestic debt of bank
grew greatly from a single figure to about 60%
o Cost of debt and cost of rescuing led to service of domestic debt at about 10%
3 groups = same phenomenon massive liquidity due to opening of financial mkts. and K account
Once let in, no matter what did, had a crisis
Compare routes 1 and 3
Granger test in model, does S lead to I or does I lead to S? Does causality go from exogenous
to endogenous variables? Add lagged variables
o In fact tests time precedence
o What first, surge in influx or CA deficit?
EAsia deficit due to microelectronics dragged in inflows
LatAm inflows led to deficit Int. fin. mkts. not knowing what to do with flows
o Sent into LatAm (led to asset mkt. bubbles, money available for investment etc.)
International Finance = very much a supply side mkt. always demand, can always sell
Malaysia and Thailand followed aspects of route 1 and 3 both used inflows to invest in their
ambitious priv. investment programmes but had surplus which fuelled LatAm style asset bubble in stock
mkts. and real estate. However, in both, credit expansion associated with drop in share of cons in GDP
(49-46% in Malaysia)
Interest Rates
Route 1 real interest rates start high due to stabilisation programmes but soon after allow to
fall to int. levels
Route 2 Brazil RER not only set at much higher level during P stabilisation programme, but
due to sterilisation programme, never allowed to fall to level anywhere near East Asia or
Mexico
McKinnon and Pill (1997) - main cause of borrowing agents losing their capacity to assess and P risk
properly is that internal and external moral hazards lead to artificially low interest rates which in turn
Part IIB Revision, Stephen Ward 117 Development
Spreads
of bank system
Optimal Spread
above the qual. of
the portfolio begins
to deteriorate
give false incentives to agents to accumulate excessive amounts of risk In Brazil, though high interest
rates prevented route-1 style crisis, did so by creating different but equally damaging route-2 crisis
Consumption
Priv. cons. as %GDP is well over 65% for LatAm countries and in some cases (Chile) rises to >75%
For E. Asian countries (av. of Korea, Thailand, Malaysia), never above 55%
Imports of cons. goods follows similar pattern with route-1 countries characterised by
exceptional expansion in M of cons. goods
Stock Mkt.
Dow Jones, NASDAQ and index for Europe grew 1.4-3x 1975-1980, Chilean stock mkt. increases
22x whilst GDP growing at 3.2%
Self-fulfilling pull factor
Malaysian expansion 6x, Thailand 5.4x. Cannot be efficient mkt.
Real Estate P
Route-1 countries began to develop some micro & macro distortions and vulnerabilities
o Overvaluation of ER led to large CA deficits which reflected priv. imbalances (pub.
sect. largely in-line)
o ER overvaluation also distorting composition of what little investment there was
towards non-tradable sector
o While residential construction x2 1981-94 in Mexico, investment in machinery falls by
half and in infrastructure even further
o Distortion in rel. P, easy access to credit ans asset bubble in real estate set in motion a
high Kuznets cycle
Brazil crucial vulnerability
Excessively high interest rate lead straight into pub. sect. Ponzi finance
High cost of sterilisation evident in difference between i (nom. interest paid for public debt)
and r (income received for FX reserves) i.e. difference between what paid for paper sold to
sterilise and what recuperated from return on holding of FX reserves
Violates golden rule paying much higher rate for public debt than rate it managed to increase
pub. rev
Crisis more to do with weak public sect. than other 2 routes
Interest rates for working K peaked at 60% 1998 and 115% for cons. credit
o Hardly any priv. asset could perform increase in non-performing loans succession of
banking crises
o Indiscriminate bail-outs further increased public sect. debt
Poor banking sect. bought poor public paper (only asset that could perform at such rates) and
tried to increase profitability by self-defeating pol. of ever increasing spreads
East Asian story
High rate of priv. investment needed
Priv. investment rel. stable at 30% but corporate Sectoral surplus (S I) moved into 20% deficit
o Due to switch with almost all investment switching to debt finance
o One corporation has $80bn debt, > Columbia and Chile combined
Declining profitability due to collapse of micro-elec. P meant high but stable levels of
investment must be financed by debt
P of 1MB = $25 in 1993, $0.10 in 1996
IMF/World Bank say process of over-investment at this stage which was the real mkt. failure
But once country entered high-tech. X mkt., can only be comp. if able to prod. at cutting
edge of tech. and be able to remain doing so investment rates must be high
Part IIB Revision, Stephen Ward 118 Development
Once profitability collapsed, Korea had choice of stay in micro-electronics and invest or look
for different strategy all together
Day of Reckoning:
3 common probs.
Constant changing composition of priv. inflows made matter of dealing with inflows more
complex
Progressive shortening of term structure of debt adding fragility and heightened uncertainty in
already difficult situation
o Route 1 had lowest share of ST debt, especially Arg
o Early high share of Koreas ST debt remarkable happened much earlier than did in Lat
Am as logic suggests should have been other way around
o Reason was that in Korea, lots of red tape involved with LT debt that absent for ST so
Koreas govt. gave incentive to borrow short
o Contrasts with Brazil where int. fin. mkts. that imposed shortening of debt
Constant danger that in fin. liberalised economy, attack could also come from within
Ratio of ST debt to foreign reserves
Main weakness was ST structure with exceptionally low reserves could only cover ST
liabilities not even sufficient to cover debt with maturity of 90 days or less
Strange neglected this considering within real economy felt need for strong govt. intervention
o In areas relating to KA and mon. pol., only seem concerned with LT K movement, ER
stability and in keeping rates as low as possible
Led to over indebtedness of corporate sect. and CB had too few reserves.
Complacency thought too big to be attacked
Brazil
Mixed pol. on these issues
Against intervention in K account to reduce share of ST debt but did increase levels of reserves
o Under system, ST debt grew even faster and fundamentals deteriorated rapidly leaving
economy extremely vulnerable to sudden collapse in confidence and withdrawal of
external fin.
Kindelberger (1996)
Bubble always implode by def. involves non-sustainable pattern of P changes or cash flows
Virtually all Mex banks failed at end of 94 when Peso depreciated sharply
Most banks in Thai, Malay, S.Korea and several other Asian countries bankrupt after mid-1997
Asian Fin. Crisis
These crises resulted from implosion of asset bubbles or from sharp depreciations of nat.
currency
Losses incurred by Arg banks accounted for 50% GDP
Minsky Model of Fin. Crises.
o Used to model fin. crises in US, UK and other mkt. econ. but easily transferable
o Highlights procyclical changes in supply of credit. More optimistic in expansions
regarding returns so invest and borrow more
o Investors begin to buy goods/securities to profit from K gains associated with
anticipated increases in P of these products
o Herd behaviour more people join scramble for assets
o Banks over-borrow dont want to lose mkt. share to competitors who also over-
borrowing. Risk awareness becomes more lax
o Bubbles that created burst leaving many indiv. in debt, increase in non-performing
loans and bank sect. in poor position
Part IIB Revision, Stephen Ward 119 Development
o Int. propagation runs through trade links, psychological impact on investors
Criticisms
o Each crisis unique so general model not appropriate
o No longer relevant due to changes in business and econ. environment
o Efficient mkt. hypothesis when thinking about asset P bubbles
Relevance
o Growth of Mex, Br, Arg external debt bank loans to these countries increased at 30%
per yr, external debt increased 20% per yr
Appears banks giving these loans never asked question where money to repay
loans coming from should they stop lending
o Rational expectations assumption used in many models is that investors react to
changes in econ. variables as if fully aware of LT implications of changes
Mania suggests loss of touch with rationality closer to mass hysteria
o Mob. Psychology, herd behaviour rational indiv. but irrational group
o Speculation often develops in 2 stages
First rational response to shock
Second appetite for large gains
o Speculative manias gather pace through expansion of money and credit
o Qual. of debt an issue can lead to Ponzi fin.
Warnings
o Govt. publish reports warning over excitement in mkts.
[fig. 7.6]
What model can withstand these massive shifts in K flows?
Look at ER, i and KA. If have fully open KA, have to have endogenous ER and i pol. (Trilemma)
Cannot have i and ER pol. with KA closed
[fig. 10]
Interest rate spread i i
*
+ e vital for explaining inflows
K account pol. effectively shifts relationship between inflows and interest rate spread
downwards
o Can maintain spread by reduce inflows A to B or can maintain same level of inflows but
increase the interest rate spread A to C
Developing countries may wish to move A to C but if increase i, more incentive to foreign K to
come, to cope with inflating economy
o With open K account, fall in level policy available
o If i increases, self defeating pol.
Only way to increase i without bringing in more inflows is K-controls i.e. put tax on inflows
o If want to reduce inflows, go A to B
o If want to bring down domestic economy go A to C
Ocampo and Palma (2007)
Not obvious what ideal. Pol. mix for dealing with cycles
Fin. cycles the fault of govt. intervention?
What pol. recommendations have we given to developing countries?
Moral hazard probs. Created by govt. deposit insurance and bailouts by int. institutions, crony-
mechanisms
Chile, Columbia and Malaysia fine-tuned integration into int. fin. mkts. via reintroduction of K
controls to deal with 1990s surge in priv. K inflows (specifically ST)
K inflow structure
Part IIB Revision, Stephen Ward 120 Development
1990s cycle occurs only 20yrs after previous cycle of 70s. 70s was about 50yrs after boom
preceding it. In case of LatAm, rapid return of inflows due to USA decision to sort out problems
in 70s. Brady bonds
Bank lending, boom of 70s followed by extended period of net resource outflows that lasted
more than 2 decades
Lat. Am. Asymmetry
Unable to generate CA surpluses, even at times of domestic recession and net K outflows,
leading to FX reserve depletion, debt rescheduling and default problem generated by
composition of C and its endemic under-investment in X diversification
East Asia did not have this problem generated large CA surpluses when required
Problems with not generating CA surpluses is that inevitable affects capacity to repay
principal, increasing likelihood of Ponzi fin.
Timing
Lat Am changes in KA tend to precede (and useful predictors of) changes in CA in East Asia,
the opposite
o Large impact on pol. if primary source of Latin fin. cycle is induced change in K
account, then K controls more likely to be 1
st
best pol.
E. Asia CA changes precede and useful predictors of KA changes
o 1
st
best pol. would focus on dealing with domestic probs. leading to changes in CA
mainly large levels of deficit fin. needed to maintain high levels of investment in
development strategies
Regulation
K controls to help avoid pro-cyclical dynamics created by surges in inflows how to deal with
volatility of K flows. Also, open KA tend to create auto destabilisers
More selective path of participation in int. K mkts. proved more effective way of avoiding pro-
cyclical dynamics of unrestricted K flows
o 3 countries of interest, though initially opened KA fully, later decided to fine tune
integration using different forms of preventive reg. to control unwanted K flows
Price Based regulation (unremunerated reserve requirements) key method used in Chile and
Columbia penalised ST foreign currency liabilities
o Meant to have countercyclical role raised during periods of surges in K flows and
lowered when external conditions weak
o Palma URR dont affect volume of K flows at given interest rate spreads
In Columbia strong ev. these pol. reduced overall flows sig. amount of history
associated with closing loopholes
Quantity regulation minimum stay requirements. Chile and Columbia also used these too
(1yr including FDI in Chile). In Columbia, amount of funds that portfolio investors could bring
into country and their domestic use were regulated. Both countries required all loans to be
registered with CB.
Degrees of freedom in country with open K mkts. for pol. very much reduced
Stiglitz better to have auto-stabilisers than auto-destabilisers
o Imagine no K controls, try to free ER so to affect i
Suppose keep i fixed and i* fixed
Let e free leads to domestic destabilisation huge inflows lead to ER
re-values - people also get return to reflows from this
Increased return leads to increased inflows
Suppose i increases to domestic economy, bring down, to foreign,
increases return
Part IIB Revision, Stephen Ward 121 Development
Objectives
Macro pol. tool provide room for counter-cyclical pol., help to cool down AD and try to avoid
accumulation of unsustainable debt burdens
o Frees up mon. pol. can lean against wind during periods of fin. euphoria through
adoption of contractionary mon. pol.
Liability pol. avoid risky corporate balance sheet structures and maturity and currency
mismatches typical of priv. sect. fin. structures in developing countries
o KA regs. made with reference to fact mkts. usually reward sound external debt profiles
Loan and bond maturity profile that leans more towards LT external obligations
will reduce risk of BoP crisis
Help avoid asset bubbles
Surge in K inflows often generates over-optimism or irrational exuberance
which can result in excessively high valuation of stocks and real estate
Portfolio inflows may directly fuel stock P
Rising real estate P may result in excessive investment in construction
Collapses in asset P can result in non-performing loans, bank crisis etc.
First best solution may focus on deepening domestic fin. mkts.
Main reasons for imposing reg. in countries in question seem to be first 2, not alleviating asset
P bubbles
URR able to increase interest rates in Chile and Columbia ev. on ER more mixed but could reflect
difficulties in ER modelling
Given multiple channels through which URR can affect economy, effectiveness of regs. can best be
measured as broad index of mon. pressures that includes the 3 channels through which K inflows can
affect economy
Accumulation of int. reserves
ER appreciation
Reduction in interest rates
China Solution
Keep ER fixed removes 1 auto-de-stabiliser
Use i for what need in domestic economy
Handle inflows by controls (P or Q China more on Q side)
o Tax on inflows like new instrument of macro pol.
Often what want for domestic economy but has opposite effect on external economy
Big tax for K that came in and out within yr. Hardly taxed if in for long time
o Any inflow must deposit 1/3 in CB account with no return on it
Chile
K inflows s.t. flat rate foreign currency deposits in CB originally 2-% but raised to 30% in May
1992
o Deposits originally meant to be kept
Put in control inflows fell then increase again
o Same again with next control
Like speed bump works, but effect very short lived
o Sig. but ST effect priv. inflows bounced back after being affected briefly by
imposition of reg.
o Not simply learning to evade it, just returns so high tax should be higher
Results even clearer when just for equity securities
o In terms of volume, net equity securities and other investments went from generating
large neg. flows in 1988 to large positive flows in 1990
Part IIB Revision, Stephen Ward 122 Development
o However, in 1991 when reg. imposed and 1995 when strengthened, net equity
securities fell to very low levels but this only lasted for that yr
Intro P-based pol. not only to affect magnitude of K floes but also to allow interest rates to
remain at levels higher than parities and avoid further ER appreciation
Intro of regs. not accompanied by reversal of expansionary trend though those instituted in
1995 had more discernible effect
Effects of 1991 unclear but effects of 1992 allowed higher interest rate/devaluation mix but
were incapable of affecting reserve accumulation
Asset P index in 4 quarters preceding reg., index jumped more than 3x but reg. stopped trend
for about 7 quarters
o Further inflows caused another jump in 1994 but controls of 1995 had immediate effect
on new bubble, bringing index down considerably even more pronounced in real
estate mkt.
Columbia
Initially URR system applied to credits with maturities below specified term (initially 18
months) with amount of deposit originally inversely proportionate to term of credit
o Originally designed as liberalisation of previous regime which imposed reg. on final use
of loans
o Main difference from Chilean system that deposit had to be kept in local currency and
thus not immune to devaluation
In both countries, agents could opt to sub. unremunerated deposit with one-off payment to CB
(like a Tobin tax) i.e. a FC for external borrowing
Frequent changes in reg. mkt. make annual data unreliable to effect on K flows
o Palma (2004) uses priv. K flows involving cash transactions
Turning point in dynamics of priv. K flows in Columbia during 1990s followed
timing of pol. decisions closely
Pol. aimed at both opening up space for contractionary mon. pol. and reversing ER
appreciation that country had been experiencing
Macro effects stronger than in Chile as used more aggressive P based reg.
Allowed for undertaking of contractionary mon. pol. while avoiding ER depreciation (trilemma)
August 1994 controls and accompanying mon. pol. led to sharp reversal of stock mkt.
Malaysia
Implemented an effective Q control
Net priv. inflows in country reached 25% GDP by 1993. To stop surge, country adopted series of
drastic Q measures
Prohibited non-residents from buying wide-range of ST securities, placed limits on non-trade
related liabilities of commercial banks, prohibited commercial banks from making swaps and
forward transactions with foreigners
Deposit interest rates drastically reduced real deposit rates fell from 4.2% to -0.9% - hope to
stop arbitrage flows
Effect immediate and remarkable So effective went to lift it effects positive beyond doubt
o net priv. inflows by 18% of GDP in one yr particularly with ST flows which fell 13% of
GDP in 1 yr
o Some speed bump effect fell then grew afterwards though growth afterwards
nowhere near as big as before
Recovery of K flows following lifting of ref. relatively mild compared to Chile
o Net portfolio inflows still negative
o Low interest rates maintained as part of residual pol. package to act as disincentive to
possible rapid return of priv. K inflows after end of quantitative restrictions
o Maintained arbitrage flows but helped fuel real estate bubble
Part IIB Revision, Stephen Ward 123 Development
Controls had major effects on international reserve accumulation with additional moderate
effect on ER
Decision to reduce interest rates supported reduction of K inflows
Asset P index fell 30% under controls although aided in dampening on real estate mkt., low
interest rates and low mortgage rates fuelled another real estate b
Imagine want to do something to bubble e.g. in Brazil
Intervene via P or Q
Q always more arbitrary not dependent on size of thing = unfair?
P not discriminating more general affect all in same way but in K mkts., more difficult to
implement
Evidence that Q controls more effective at controlling inflows on whole
ST debt = key issue
o Chile relatively stable
o Thailand and Brazil who didnt apply K controls ST composition increased significantly
[fig. 13]
Malaysia
K controls imposed stock mkt. fell (bubble burst)
Lifted stock mkt. grew again
Then burst at a and fell completely
Palma all for K controls
Keynesians generally for
Classical economists against (monetarists)
80s Rogoff (Harvard) and Summers for fin. mkts., KA controls necessary
o Only thing they recommend constraints for
Developing countries so vulnerable to inflows and auto-de-stabilisers
o Need 1 mon. pol. for domestic and 1 mon. pol. to help external, with inflows
Brazil imploded due to high cost of interest rates
Ev. appears on side of Q controls but not to say should definitely do this
Sub-Prime Crisis
Long time ago, took 30-40yrs to have next crisis as when did have crisis, lost all money
o Now, not long as lose a bit, then saved
Good openings in developing countries Northern Hemisphere still billions of liquidity to
recycle but exhausted investment opportunities lower standards and take more risk
Sub-prime mortgage mkt. fine so long as P increasing. If cannot pay, can sell house, which
covers it
o ST logic takes over only when P stagnate do things get bad
Politics now interest group based had to have effects on K-mkt. controls
Developing countries = sub-prime mkts. themselves
o By bringing too much inflows in, makes them sub-prime (some sub-prime in own right)
o Excess inflows led Brazil, Korea, Thailand etc. to be sub-prime
Need pol. to stop inflows making country a sub-prime mkt.
o Only go to these when run out of other high return investments
o More herd behaviour with developing countries info much more expensive and scarce
No matter what do, end up in crisis and become sub-prime mkt. almost anything can happen
to bring things down
Part IIB Revision, Stephen Ward 124 Development
Korea had most sensible macro pol. on everything but this. Why CB of Korea not accumulate
reserves? Cost would have been very low (0.5% point above rate could issue at)
o Had they done this, would have been no crisis
o Only explanation thought unsinkable thought no one attack as have best
fundamentals
Part IIB Revision, Stephen Ward 125 Development

S-ar putea să vă placă și