Macroeconomics II—Problem Set 3

Solutions

TAs 2013: George Lukyanov and Harry Di Pei

TAs 2014: Constance Blandin and Francois de Soyres
TA 2019: Miguel Zerecero

1 Decreasing Returns and Pricing Complementarities

...

2 Asset Pricing
This problem is taken from Ljungqvist and Sargent (2004); see Exercise 8.3 on p. 252.

a. By definition, a competitive equilibrium is a feasible allocation of consumption and asset

holdings {cit , ait+1 }∞ 0 ∞
t=0 for each agent i = 1, 2 and a price sequence {qt }t=0 such that
i) Given price, the allocation solves the household’s problem, for all i = 1, 2.
ii) The allocation is feasible: for all t = 0, 1, . . .
2
X 2
X 2
X
cit = yti and ait+1 = 0
i=1 i=1 i=1

In order to compute competitive equilibrium, we write down agent i’s maximization prob-
lem. He chooses the sequence {cit , ait+1 }∞
t=0 , solving
∞
X
max β t u(cit ) (2.1)
{cit ,ait+1 }
t=0

subject to the budget constraint1

∞
X ∞
X
qt0 cit ≤ qt0 yti (2.2)
t=0 t=0

and ai0 = 0. The Lagrangian is given by

∞ ∞ ∞
" #
X X X
L= β t u(cit ) + µi qt0 yti − qt0 cit (2.3)
t=0 t=0 t=0
1
There is a single budget constraint, because by assumption, the good is non-storable and all trade in Arrow
securities occurs at date
P∞ 0. The {ait+1 } P
terms effectively vanish from both sides of the constraint, since with
i
a0 = 0, we must have t=0 qt+1 0
ait+1 = ∞ 0 i
t=0 qt at .

1
 where µi is household i’s Lagrange multiplier (note that µi is time-invariant). The asso-
ciated FOC’s with respect to ci0 and cit are

u0 (ci0 ) = µi (2.4)

β t u0 (cit ) = µi qt0 (2.5)

where we used the normalization q00 = 1. Dividing (2.5) by (2.4), we get the condition

u0 (cit )
qt0 = β t (2.6)
u0 (ci0 )

Let us guess and verify that in the competitive equilibrium, each agent’s consumption will
be constant across time.2 Then we have u0 (cit ) = u0 (ci0 ) for all t, and from (2.6), qt0 = β t .
Agent i’s consumption can then be found from his budget constraint:
∞
X ∞
X ∞
X
β t ci = β t yti ⇐⇒ ci = (1 − β) β t yti
t=0 t=0 t=0

Thus agent 1’s consumption is given by

∞
X 1−β 1
c1 = (1 − β)(1 + β 3 + β 6 + . . .) = (1 − β) β 3t = =
1 − β3 1 + β + β2
t=0

and by market clearing, agent 2’s consumption is thus given by

1 β(1 + β)
c2 = 1 − c1 = 1 − 2
=
1+β+β 1 + β + β2

b. In the world of complete markets, this derivative security would be redundant. Since it
promises to pay perpetual dividend stream D = (.05, .05, . . .), it can be priced as
∞ ∞
X X .05
P0 = qt0 D = .05 βt = (2.7)
1−β
t=0 t=0

3 Ramsey Taxation
This exercise considers a simplified version of the Ramsey problem with complete markets
seen in class. The latter parts of the problem then modify the asset market structure to
conjecture what happens in incomplete market environments. The two simplifications are:
(i) no capital accumulation, only labor taxes and (ii) a specific process for uncertainty with a
single shock at date 1, and no further uncertainty afterwards. We saw in the generalized Ramsey
framework that with complete markets labor tax distortions should be smoothed across time,
and debt/assets be used to shift resources over time and across states. With power utilities,
this implied constant labor tax rates.
2
This is a reasonable guess, given that (i) aggregate endowment is non-stochastic and constant across time: we
have yt1 + yt2 = 1 for all t = 0, 1, . . . and (ii) each individual has preference for consumption smoothing, which
is captured by the concavity of u(·).

2
 The subsequent asset market structures alter the ability to shift resources over time. With
one period bonds, there is limited ability to shift resources across states in period 1, but perfect
smoothing across time, so the shock should lead to a fluctuation in labor taxes in period 1, but
not in future periods. With one period bonds and perpetuities, two securities are again sufficient
for perfect smoothing across time and states: after date 1, the second security is redundant. At
date 1, the gap in their one period returns, will provide full spanning of the spending risk, and
hence recover the full tax smoothing result.

a. The simplest way to analyze the problem is by working backwards. Let V (b) denote
the value of the deterministic Ramsey problem if there is an initial debt level b, and no
further government spending obligations – this is the problem starting from period 2 after
a disaster, and starting from period 1, if there is no disaster.
V (b) is defined as follows. In any period, for a given labor tax τ , the resulting labor supply
solves the first-order condition (1 − τ )u0 (n) = v 0 (n), where we have used the condition
that n = y = c. Let τ (n) = 1 − v 0 (n)/u0 (n) denote the tax rate that generates a labor
supply n, and let w(n) = u(n) − v(n) denote the resulting per period utility, in terms of
n. The problem is then defined as
∞
X
V (b) = max
∞
β t w(nt ) (3.1)
{nt }t=0
t=0

subject to
∞
X u0 (nt )
b≤ βt τ (nt )nt (3.2)
u0 (n0 )
t=0
It is immediate that the solution to this problem is stationary over time, and prescribes a
constant tax rate τ , such that b = τ (n)n/(1 − β). Writing the resulting value function in
terms of n, we have V (b(n)) = w(n)/(1 − β).
Likewise, the period 0 utility is w(n0 ). With this, we can write the planner’s objective in
terms of the labor supply in all periods as:
β β2
w(n0 ) + (1 − π)w(nN ) + πw(nD ) + βπ (u(n̂D − G) − v(n̂D ))
1−β 1−β
where n0 is the labor supply in period 0, nN is labor supply in periods 1, 2, . . . if no
disaster occurs, n̂D is labor supply in period 1 if a disaster occurs, and nD is labor supply
in periods 2, 3, . . . after a disaster has occurred.
In addition, we need to keep track of the first-order condition in period 1 in the disaster
state:
(1 − τ̂D )u0 (n̂D − G) = v 0 (n̂D ) (3.3)

The government’s revenue at any history is τ n, which has to be weighted by the state-
prices (discounted probability-weighted marginal utilities). This leads to the following
government budget constraint:

β β2
τ (n0 )n0 u0 (n0 ) + (1 − π)τ (nN )nN u0 (nN ) + πτ (nD )nD u0 (nD )
1−β 1−β

+ βπ (n̂D − G)u0 (n̂D − G) − n̂D v 0 (n̂D ) ≥ 0 (3.4)

3
 We have thus formulated the Ramsey planner’s objective and implementability constraint
just in terms of the labor allocations.

b. Now, from the stationarity of the problem above, it is immediate that at the optimum,
n0 = nN = nD , i.e. the labor supply and tax rate should be the same whenever G = 0.
Working out the first-order condition for n̂D and n0 , we find the same tax-smoothing
result as in class, i.e. that the tax distortions in the disaster and no-disaster states are
also the same, if preferences admit a power function representation. Notice that it implies
that the labor supply in the disaster state n̂D is higher than in the other states in this
case:
v 0 (nˆD ) v 0 (n0 )
τˆD = τ0 ⇒ 0 = 0
u (nˆD − G) u (n0 )

c. With uncontingent bonds, perfect smoothing between the disaster and no-disaster histories
is no longer possible. In particular, if the government saves a in uncontingent bonds at
date 0, then it needs to raise a revenue of G − a after the disaster state, while in the
no-disaster state it is zero – the interest payments on the bonds are simply rebated back
to the households. The resulting planner’s value function takes the same form as above,
but adds as a restriction that nN = n∗N (the first-best labor supply), and τN = 0 (no
revenue raised from no-disaster states). As a result, the no-disaster states can no longer
be used to reduce the distortions needed to finance the spending in the disaster state.

d. With uncontingent bonds and perpetuities, we show that perfect smoothing again becomes
possible. Now, for all periods from t > 1 onwards, there is no further uncertainty, and the
two assets are redundant, implying that the perpetuity will be priced at β/(1 − β) in all
future periods (ex dividend). At date 1, the state-contingent return for a perpetuity is
(
1
1−β , if there is no disaster
u0 (n̂D ) 1
1+ u0 (n̂D −G) 1−β , if there is a disaster

while the return from the bond is uncontingent.

0
Therefore, as long as u0u(n̂(n̂D)
D −G)
6= 1, the government can exploit the fluctuation in the
value of the perpetuity after a disaster, along with the uncontingent bonds, in order to
fully span the state space, and therefore perfectly hedge against the risk of government
spending. As a result, complete contingent markets are restored and the government can
implement the complete markets solution from (a).

4 Measuring Business Cycle Costs by Asset Prices

...

4
References
Ljungqvist, L. and T. J. Sargent (2004): Recursive Macroeconomic Theory, MIT Press, 2
ed.