Macroeconomics I

Problem Set # 3

This is due Thursday, February 3rd .

Exercise # 1 CRRA utility function

(a) Show that the CRRA utility function, u(c) = c1−θ , converges to log-utility when
1−θ

θ = 1. (Take the limit θ → 1 and use L’Hospital’s rule).

−c1−θ log c
lim u(c) = lim = log c
θ→1 θ→1 −1

Consider the following inter-temporal maximization problem:

max u(ct ) + βu(ct+1 )

ct ,ct+1
ct+1
s.t. ct + = w̄
R

(b) Using the optimality conditions, write down the Euler Equation and interpret it.
Think what the consumer would do if the equation didn’t hold with equality.
Set the Lagrangean

ct+1
L = u(ct ) + βu(ct+1 ) + λ (w̄ − ct − )
R

1
 FOC:

[ct ] ∶ u′ (ct ) = λ
1
[ct+1 ] ∶ u′ (ct+1 ) = λ
βR
ct+1
[λ] ∶ ct + = w̄
R

Combining the FOCs , we get the Euler equation:

u′ (ct ) = βRu′ (ct+1 )

The Euler Equation is an inter-temporal optimality condition. Intuitively, if we

reduce consumption today by one unit, we would forego u′ (ct ) utils. We could then
invest this 1 unit of consumption and receive R units of consumption tomorrow,
which would yield Ru′ (ct+1 ) utils. But since we discount future utility by β, this is
equivalent to βRu′ (ct+1 ) utils today.
If, for instance, the left-hand side of the equation was greater than the right-hand
side, the consumer would be better off by increasing consumption today (and saving
less).
In the Ramsey model, we have a similar condition that must hold in every period
and which determines the consumption path.

(c) Rewrite the Euler Equation for the CRRA utility case. calculate the inter-temporal
elasticity of substitution (IES). What is the intuition of a higher/lower IES?
The IES is given by
−1
u′′ (c)c
σ = −( ′ )
u (c)
−1
−θc−θ−1 c 1
= −( ) =
c−θ θ

The greater the IES, the more the consumer is willing to substitute consumption

2
over time. Note that from the Euler equation, we can derive the following elasticity

ct+1 1
= (βR) θ
ct
ct+1 1 1
log ( ) = log( R ) + log β
ct θ ® θ
1+r
´¹¹ ¹ ¹ ¹ ¹ ¸ ¹ ¹ ¹ ¹ ¹ ¶
≈r
∂ log ( ct+1 ) 1
Ô⇒ ct
=
∂r θ

Which is a more intuitive way of interpreting the IES. The higher the IES, the more
sensitive the consumer is to a change in the relative price of consumption today to
consumption tomorrow, which is given by the interest rate.

3
Exercise # 2 The Ramsey Model

In the lecture, we solved the Ramsey model for the decentralized economy. We are now
going to solve the Social Planner (SP) problem.

(a) Why the SP problem is equivalent to the decentralized economy problem?

The first welfare theorem tells us that if an allocation and a set of prices consti-
tute a competitive equilibrium, then the equilibrium is Pareto Optimal. Moreover,
every Pareto optimal allocation is a social welfare optimal for some welfare weights
(you can check MWG chapter 16).
Consider the case where households grow at rate n and without technological growth.
Assume that ρ > n. The SP problem is:
∞
max ∫ u(ct )e−(ρ−n)t dt
{c }∞
t t=0 0

subject to the resource constraint

K̇t = Yt − Ct − δKt = F (Kt , Lt ) − Ct − δKt

where F (Kt , Lt ) is a neoclassical production function.

(b) Write the resource constraint in per capita terms.

K̇t = F (Kt , Lt ) − Ct − δKt

K̇t
= f (kt ) − ct − δkt
Lt

Where kt ≡ KLt , ct ≡ Lt and f (kt ) ≡ F ( Lt , 1).

t Ct Kt

Note that log kt = log Kt − log Lt . Differentiating w.r.t time we get

k̇t K̇t L̇t K̇t

= − = −n
kt Kt Lt Kt
K̇t
Ô⇒ = k̇t − nkt
Lt
Ô⇒ k̇t = f (kt ) − ct − (δ + n)kt

Note that this equation is similar to the one we had in the Solow model. However,
now households choose consumption in every period (and therefore savings, yt − ct )
instead of consuming (saving) a fixed share of their income as in the Solow model.

4
(c) Write the current-value Hamiltonian.

H = u(ct )ent + λt [f (kt ) − ct − (δ + n)kt ]

(d) Write the FOCs.

Hc = 0 Ô⇒ u′ (ct )ent = λt
Hk = ρλt − λ̇t Ô⇒ λt [f ′ (kt ) − (δ + n)] = ρλt − λ̇t

(e) Using the FOCs, write the Euler equation (the equation that determines cċ .
HINTS:

• use the FOC w.r.t. k to find an expression for λλ̇ .

• time differentiate the FOC w.r.t. c, and divide this expression by the FOC
′ (c)
w.r.t. c. Define σ ≡ − uu′′ (c)c .

From the Hk FOC we have:

λ̇t
= ρ − f ′ (kt ) + δ + n
λt

Differentiating the Hc w.r.t. time we have:

u′′ (ct )c˙t ent + u′ (ct )ent n = λ̇t

Dividing the last expression by the Hc FOC:

u′′ (ct ) λ̇t

′
c˙t + n =
u (ct ) λT
u′′ (ct )ct c˙t λ̇t
Ô⇒ ′
+n=
u (ct ) ct λT
c˙t
Ô⇒ = σ[f ′ (kt ) − δ − ρ]
ct

(f) What conditions kk̇ and ċc satisfy on the BGP? What are the two expressions that
determine capital and consumption on the BGP?

5
 On the BGP we must have

k̇
=0
k
ċ
=0
c

Which implies the steady-state equations:

k̇
= 0 Ô⇒ c∗ = f (k ∗ ) − (δ + n)kt∗
k
ċ
= 0 Ô⇒ f ′ (k ∗ ) = δ + ρ
c

(g) How does capital per capita on the BGP compares to the golden rule capital per
capita of the Solow model? Discuss.
Differentiating the steady-state equation relative to k̇
k = 0 w.r.t. kt∗ we get

∂c∗
= f ′ (k ∗ ) − (δ + n)
∂k ∗
Ô⇒ f ′ (kgr ∗
)=δ+n

Since we assumed that ρ > n, we have that f ′ (k ∗ ) > f ′ (kgr

∗ ). And since f ′ (k)

is decreasing in k, k ∗ < kgr

∗ . The optimal BGP level of capital is lower than the

golden rule from the Solow model. Because future consumption is discounted, it is
not optimal to save so much that BGP consumption is maximized - it is best to
consume more along the transition to the BGP.

(h) Plot the phase diagram of the system in (k, c) space. For a given initial level of
capital, k0 , below the BGP level, discuss the possible initial levels of consumption
that are not in the saddle path. What conditions do they violate?
For the phase diagram refer to figure 1.
For the different initial consumption possible cases refer to figure 2. Note that all
these paths satisfy the Euler equation and the law of motion of capital. However,
some cases will not satisfy either the transversality condition (which is equivalent
to the household lifetime budget constraint) or the constraint that capital must
be positive. In a point like A, consumption is increasing and capital is falling
hence at some point capital will be zero and consumption will still be increasing,
which is not possible since this would imply that capital would be negative in the

6
 (infinitesimally) next period. In a point like B, capital is initially constant and
consumption is increasing, hence we would necessarily move to a situation similar
to point A.
For the other cases, it is important to take into account that both k̇ and ċ are
continuous functions of k and c. In a point like C, slightly below the k̇ = 0 locus, k̇
is initially positive but small (because of continuity!), hence capital and consumption
are increasing, but consumption is increasing faster so that we cross at some point
the k̇ = 0 locus and we are back to the case of point A. Similarly, in a point like
D, with low initial consumption. ċ and k̇ are both initially positive. From the
Euler equation, we know that ċ is proportional to c; when c is small, ċ is therefore
small. Thus c remains low, and so the economy eventually crosses the ċ = 0 line.
After this point ċ becomes negative and k̇ remains positive. At some point, capital
is going to be greater than kgr and therefore r(t) = f ′ (kt ) − δ < n, which implies
that limt→∞ k exp (− ∫0 r(s) − nds) → ∞. This is equivalent to the statement that the
t

present value of household’s lifetime wealth is infinitely larger than the present value
of their lifetime consumption. Thus each household can afford to raise consumption
at each point in time, and so can attain higher utility. Hence, such a path cannot be
an equilibrium. Finally, because of continuity, there must exist a point F, between
C and D, in order that the rate of growth of capital and consumption are such that
the economy is in the saddle path and converges to E.

(i) What is the effect on the BGP and on impact of an (unexpected) decrease in the
discount rate, ρ?
Refer to figure 3. A fall in ρ only affects the Euler equation. Since f ′ (k) is dimin-
ishing in k, this will shift the ċ = 0 locus to the right and thus the new BGP level
of capital and consumption will increase. If initially, we are on the BGP (point E),
since capital was determined prior to the shock it cannot immediately change, thus
consumption should fall on impact so that we are in the new saddle path which is
below the BGP by the reasons we discussed on the prior question.

(j) Linearize the system around the BGP. (Do a first-order Taylor expansion of k̇ and
ċ around the SS).
The two equations that we need to linearize are

k̇t = f (kt ) − ct − (δ + n)kt

c˙t = σ[f ′ (kt ) − δ − ρ]ct

7
 Figure 1: Phase Diagram: The dynamics of k and c

The first-order Taylor expansion of this system of equations around the BGP,
(k ∗ , c∗ ), is

∂ k̇t ∂ k̇t
k̇t = k̇t ∣SS + ∣ (kt − k ∗ ) + ∣ (ct − c∗ )
∂kt SS ∂ct SS
∂ c˙t ∂ c˙t
c˙t = c˙t ∣SS + ∣ (kt − k ∗ ) + ∣ (ct − c∗ )
∂kt SS ∂ct SS

where ∣SS means that the expression is being evaluated at the steady-state (on the
BGP).

8
 Figure 2: The behaviour of k and c for various initial levels of c

Moreover

∂ k̇t
∣ = f ′ (k ∗ ) − δ − n = ρ − n
∂k SS
∂ k̇t
∣ = −1
∂ct SS
∂ c˙t
∣ = σc∗ f ′′ (k ∗ )
∂kt SS
∂ c˙t
∣ = σ[f ′ (k ∗ ) − δ − ρ] = 0
∂ct SS

9
 Figure 3: The effects of a fall in the discount rate

Finally

k̇ = (ρ − n)(kt − k∗) − (ct − c∗ )

ċ = σc∗ f ′′ (k ∗ )(kt − k∗)

10