0.

1 Ramsey problem

The production side is like in the Solow model. Output per capita

yt = f (kt )

simplify n = 0 and g = 0 so the law of motion for capital per capita is

kt+1 = (1 − δ)kt + it

ct + it = yt

=⇒ kt+1 = (1 − δ)kt + f (kt ) − ct

ct = (1 − δ)kt + f (kt ) − kt+1 (1)

with the constraints ct ≥ 0 and kt+1 ≥ 0, and k0 given.

In the Solow model, ct = (1 − s)f (kt ). Now instead we consider the problem of how

much the planner would consume/invest. The people in the economy derive utility from

consuming, and so does the planner. For a consumption stream c = {ct }Tt=0

T
X
β t u(ct )
t=0

where β ∈ (0, 1) is the discount factor an captures impatience and we have an infinite

horizon T = ∞. u(ct ) is the per period utility function and we assume it is

• increasing

• concave

• Inada conditions limc→0 u0 (c) = ∞ and limc→∞ u0 (c) = 0

1
Example 1. For example, the CEIS function

1
c1− θ
u(c) =
1 − 1θ

where θ > 0 is the EIS and controls how much the agent is willing to let his consumption

vary across periods.

The Ramsey problem is

t c
T
X z }| {
t
max β u (1 − δ)kt + f (kt ) − kt+1
{kt }∞
t=0 t=0

st : 0 ≤ kt+1 ≤ (1 − δ)kt + f (kt )

k0 given

Where if we have a solution {kt∗ } then we can rebuild the sequence of consumption {ct }

from (1) (and output {yt } and investment {it }).

Finite Horizon T < ∞. We can solve this problem in several ways. First imagine we

have a finite horizon problem: t = 1, ..., T .

Then we know how to solve this problem. Ignore the non-negativity constraints (we can

check them later), taking FOC (because of the concavity, FOC will be sufficient) we get

β t u0 (ct )(−1) + β t+1 u0 (ct+1 ) (1 − δ) + f 0 (kt+1 ) = 0 ∀t = 1...T − 1

u0 (ct ) = βu0 (ct+1 ) (1 − δ) + f 0 (kt+1 )

The idea is that if we have a plan {kt } and we decide to reduce consumption in period t

by a small  and use that to invest and accumulate capital for next period kt+1 +  then we

2
can increase consumption next period by (1 − δ) + f 0 (kt+1 ) and keep kt+2 and the whole

subsequent plan unchanged:

ĉt+1 = (1 − δ)kt+1 + f (kt+1 ) + (1 − δ) + f 0 (kt+1 ) − kt+2

The reduction in consumption at time t has a cost in utility u0 (ct ) and from the increase

in consumption in period t + 1 we get βu0 (ct+1 ) ((1 − δ) + f 0 (kt+1 )) . It better be the case

that we cannot improve by picking a small  (greater or smaller than zero), and this is what

the FOC condition captures: “local deviations”

So we have a second order diference equation for {kt }:

tc t+1 c
z }| { z }| {
u (1 − δ)kt + f (kt ) − kt+1 = βu (1 − δ)kt+1 + f (kt+1 ) − kt+2 (1 − δ) + f 0 (kt+1 )
0 0

with an initial condition k0 given. We need a second “boundary condition”. For the

last period T we have kT +1 = 0, so this has a unique solution {kt∗ }. To understand this

condition take the FOC for kT +1 :

β T u0 (cT )(−1) ≤ 0 and kT +1 ≥ 0

because the non-negativity constraint here can be binding (this is the Kuhn Tucker condi-

tion), and with complementary slackness:

β T u0 (cT )(−1)kT +1 = 0

So if kT +1 > 0, then β T +1 u0 (cT )(−1) = 0 which cannot be. Hence, kT +1 = 0. Intuitively,

capital is worthless since the economy ends and we can’t use it to produce consumption

goods.

3
 If this condition failed, we could at some period t < T consume a little more ct +  and

obtain a little bit less capital next period but not make up for it with less consumption

next period (keep the same consumption for every consecutive period), so that instead of

keeping kt+2 unchanged, it would go down a little, and so would all the consecutive ks for

s = t + 2...T . If the original consecutive capital levels ks were strictly positive (and this

will be the case for some t) this plan is still feasible, but better because we consumed more

in one period and the same in all others! This is a “global” deviation.

Becuase of the concavity, the FOC - including the Kuhn Tucker inequality - are sufficient.

Finally, after building the solution, check that the non-negativity constraints we ignored

are actually satisfied and we are done.

Infinite Horizon T = ∞. With an infinite horizon T = ∞, we don’t have a “last period”

and so we never want to have no capital. The second boundary condition becomes instead

a “tranversality condition”:

lim β t u0 (ct )kt+1 = 0

Proving this is beyond the scope of this course, but the intuition is similar to the finite

horizon case: we don’t want to accumulate capital for its own sake. The FOC conditions

make sure there are no “local” deviations (consuming a little bit less today and a little bit

more tomorrow), the “transversality condition” makes sure there is no “global” deviation,

like simply consuming more today (without reducing consumption in the future) and having

a little less capital in every consecutive period.

4
MIT OpenCourseWare
http://ocw.mit.edu

14.05 Intermediate Macroeconomics

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.