TUTORIAL 3 – SOLUTIONS

JAN GROBOVŠEK

TOPICS IN MACROECONOMICS - GROWTH AND DEVELOPMENT

Question 1

a) In economics, the problem of the“social planner” consists of maximizing whatever

is the public’s welfare function, subject to the economy’s feasibility constraints. Feasibil-
ity constraints should be thought of as physical or technological constraints that no-one
(except possibly God) can overcome. In particular, the social planner is typically not
constrained by prices or taxes, because the latter are not set in stone. In our problem, the
1−θ
t Ct −1
planner maximizes U = ∞
P
t=0 β 1−θ
subject to the constraint Ct +Kt+1 −(1−δ)Kt = Yt ,
∀t = {0, 1, 2, ...}. Comparing this to the household’s problem, we see that the only dif-
ference is that the household’s constraint is Ct + Kt+1 − (1 − δ)Kt = (1 − τ )Yt + Tt ,
where Tt = τ Yt in equilibrium. Since the planner’s solution is optimal, in the sense that
the allocation {Ct , Kt+1 }∞
t=0 achieves the highest feasible U , we see immediately that the
optimal tax is such that the household’s problem is equal to the planner’s problem. And
this happens when τ = 0. In that case, we know that the household’s choice will be
identical to the planner’s, and hence setting τ = 0 is optimal.1

b) The household’s Lagrangian is

∞ ∞
X Ct1−θ − 1 X 
βt λt (1 − τ )Ktα (AL)1−α + Tt − Ct − Kt+1 + (1 − δ)Kt .

L= + (1)
t=0
1−θ t=0

where the λ’s are Lagrange multipliers on each period t budget constraint. As a note
on Lagrangians, remember that the term that comes after the λt could also have been
Ct + Kt+1 + (1 − δ)Kt − (1 − τ )Ktα (AL)1−α − Tt , since according to both formulations,
the resulting term in brackets is 0.

When solving a problem like this, we must first know what the choice variables are.
Here, the choices are Ct and Kt+1 , ∀t ∈ {0, 1, 2, ...}. Consider for a moment the meaning
of that choice. It implies that at period 0, the household predicts all the values of endoge-
nous variables that it does not choose (in this case, all values of Tt from t = 0 until ∞)
and then makes a plan about the choices Ct and Kt+1 , ∀t ∈ {0, 1, 2, ...}. The equilibrium
(we will get to this shortly) is then such that household’s prediction on Tt turns out to

1By the way, you may wonder whether some τ 6= 0 could also lead to an optimal allocation. This is in
principle possible, but we cannot see it at this stage because we would first need to solve the equilibrium
and then compare its allocation to the one obtained from the planner’s equilibrium (it turns out that in
this problem, the answer is that τ = 0 is unique in replicating the highest possible U achieved by the
planner).
1
2

be correct, and the plan is realized.2

∂L
The household’s first-order conditions with respect to Ct and Kt+1 are ∂Ct
= 0 and
Ct−θ
∂L
∂Kt+1
= 0. The first one is simply β t (1 − θ) 1−θ − λt = 0 and therefore
β t Ct−θ − λt = 0. (2)
The first-order condition with respect to Kt+1 is
α−1
(AL)1−α + 1 − δ = 0.
 
− λt + λt+1 (1 − τ )αKt+1 (3)
Note that the Lagrangian problem also posits an additional first-order condition, namely
∂L
∂λt
= 0. We see from (1) that this simply re-introduces our initial budget constraint:
(1 − τ )Ktα (AL)1−α + Tt − Ct − Kt+1 + (1 − δ)Kt = 0. (4)

The Euler equation can be found by replacing the Lagrange multipliers in equations (2)
and (3). The first one gives λt = β t Ct−θ and therefore also λt+1 = β t+1 Ct+1
−θ
. Replacing
these two expressions in (3) yields
− β t Ct−θ + β t+1 Ct+1
−θ α−1
(AL)1−α + 1 − δ = 0
 
(1 − τ )αKt+1
⇔β t Ct−θ = β t+1 Ct+1
−θ α−1
(AL)1−α + 1 − δ
 
(1 − τ )αKt+1
⇔Ct−θ = βCt+1 −θ α−1
(AL)1−α + 1 − δ
 
(1 − τ )αKt+1
θ
Ct−θ = β (1 − τ )αKt+1 α−1
(AL)1−α + 1 − δ
 
⇔Ct+1
which can finally be nicely expressed as.
 θ
Ct+1 α−1
(AL)1−α − δ .
 
= β 1 + (1 − τ )αKt+1 (5)
Ct

c) We are interested in the equilibrium allocations of the variables Ct , Kt+1 , Yt and

Tt , ∀t ∈ {0, 1, 2, ...}. Let us start with t = 0 and see if we can find a system of equations
which we could solve for C0 , K1 , Y0 and T0 . The first two-equations are the household’s
budget constraint (4) and optimality (“Euler”) condition (5):
(1 − τ )K0α (AL)1−α + T0 − C0 − K1 + (1 − δ)K0 = 0,
 θ
C1
= β 1 + (1 − τ )αK1α−1 (AL)1−α − δ .
 
C0
In addition, we have the equations for GDP and the transfers:
Y0 = K0α (AL)1−α ,
T0 = τ Y0 .
We see that we have 4 equations, and 5 unknowns (i.e., 5 variables). In addition to C0 ,
K1 , Y0 and T0 , there is also C1 (note that K0 is not a variable, because it is exogenously
given). In order to find an equilibrium, we require a fifth equation. We can add another
round of equations for t = 1 in the form of
(1 − τ )K1α (AL)1−α + T1 − C1 − K2 + (1 − δ)K1 = 0,
2Notethat at some later t0 > 0 it could happen that something in the economy changes, for instance a
parameter or a tax, or that there is an announcement that at some t00 > t0 something is going to change
(these are hard problems, and we do not consider them here). In that case, upon being informed of the
change or future change in t0 , the household would immediately re-optimize the allocation of current and
future Ct and Kt+1 for all t ≥ t0 .
 3
 θ
C2
= β 1 + (1 − τ )αK2α−1 (AL)1−α − δ ,
 
C1
Y1 = K1α (AL)1−α ,
T1 = τ Y1 .
Now we have a system of 8 equations, but 9 unknowns, so again we can’t compute the
equilibrium. This can be repeated up to infinity.3 What is of interest is to see that the
household’s problem is indeed very sophisticated. While in Solow, the household does
not optimize anything, in Ramsey, the household optimizes fully, which means that an
equilibrium consists of solving an infinite set of equations simultaneously!

d) For our purposes, we do not require all the variables but can focus on a subset.
Here we are asked to write down the conditions that characterize the equilibrium in
terms of consumption and capital only. Consider the household’s equations (4) and (5).
We see that there is still Tt floating around. It is now that we can replace it with
Tt = τ Yt = τ Ktα (AL)1−α . The budget constraint hence becomes
(1 − τ )Ktα (AL)1−α + τ Ktα (AL)1−α − Ct − Kt+1 + (1 − δ)Kt = 0
or simply
Ct + Kt+1 − (1 − δ)Kt = Ktα (AL)1−α . (6)
The second equation is of course the Euler equation (5):
 θ
Ct+1 α−1
(AL)1−α − δ .
 
= β 1 + (1 − τ )αKt+1 (7)
Ct
These two equations ∀t ∈ {0, 1, 2, ...} provide a system of equations with the variables Ct
and Kt+1 .

Note that Tt can only be replaced after finding the household’s optimality condition. The
reason for this is that the household fixes its choices without internalizing that Tt = τ Yt .
It is only in equilibrium, then, that Tt = τ Yt must hold.

e) Consider now equation (7). Since we define the growth rate of consumption as gtC ≡
Ct
Ct−1
− 1, we have CCt−1
t
= gtC + 1. Replacing this in (7) gives

θ
gtC + 1 = β 1 + (1 − τ )αKtα−1 (AL)1−α − δ .
 
(8)
There is an informal and a formal way of replying to the question. The informal way is
fine. We see that because α−1 < 0, the right-hand side is decreasing in Kt . The left-hand
side is an increasing function of gtC (because θ > 0), and hence it is immediate that an
increase in Kt requires a fall in gtC . In particular, this also means that if Kt+1 > Kt ,
C
then it must be that gt+1 < gtC . What is the intuition? Because production features
diminishing marginal returns to capital, an increase in capital tomorrow leads to a lower
return tomorrow. This lowers the implicit value of utillity of consuming tomorrow relative
to consumption today. As a result, CCt+1t
falls.

3You may wonder how the system can be solved if there is, even for t → ∞, always one equation missing.
We have not covered this because involves an additional mathematical step during optimization. At t
close to infinity there is another optimality condition, the so-called “transversality condition.” Basically,
it says that the household does not want to bring any capital into the period after infinity since the
household won’t be alive then. And this additional equation then allows to find the missing equation.
But this is really a technical matter that goes well beyond the requirement of this course.
4

Before continuing, let us also consider the formal derivation (this is not necessary, but
good practice of calculus). First, rewrite (8) as
  θ1
1−α
gtC + 1 = β 1 + (1 − τ )αKtα−1 (AL)
 
−δ
  θ1
1−α
⇔gtC τ )αKtα−1
 
= β 1 + (1 − (AL) −δ − 1.

∂gtC (Kt )
Now compute ∂Kt
. This yields
 θ1 −1
∂gtC (Kt )

1 1−α
β 1 + (1 − τ )αKtα−1 (AL) β(1 − τ )α(α − 1)Ktα−2 (AL)1−α .
 
= −δ
∂Kt θ
On the right-hand size, all terms are positive, except α − 1, which is negative. It follows
∂gtC (Kt )
that ∂K t
< 0. Thus, if capital increases from one period to the other, the rate of
consumption growth falls.

f) Steady state capital can be found easily from the Euler equation (7) when Kt+1 = K ?
and Ct+1 = Ct = C ? . Then we have that
 ? θ
C  ?α−1 1−α 
= β 1 + (1 − τ )αK (AL) − δ
C?
⇔ 1 = β 1 + (1 − τ )αK ?α−1 (AL)1−α − δ
 

1
⇔ = 1 + (1 − τ )αK ?α−1 (AL)1−α − δ
β
1
⇔ − (1 − δ) = (1 − τ )αK ?α−1 (AL)1−α
β
1 − β(1 − δ)
⇔ = (1 − τ )αK ?α−1 (AL)1−α
β
?1−α β(1 − τ )α (AL)1−α
⇔K =
1 − β(1 − δ)
and hence   1
? β(1 − τ )α 1−α
K = AL. (9)
1 − β(1 − δ)
Since Y ? = K ?α (AL)1−α , we simply plug in (9) for K ? :
" 1
 1−α #α
β(1 − τ )α
Y? = AL (AL)1−α
1 − β(1 − δ)
"  1 #α
β(1 − τ )α 1−α
= (AL)α (AL)1−α (10)
1 − β(1 − δ)
  α
β(1 − τ )α 1−α
= AL.
1 − β(1 − δ)

Notice that the question asked for K ? and Y ? as functions of exogenous parameters only.
Equations (9) and (10) do just that, because on the right-hand side there are no variables,
only exogenous parameters.
 5

g) In this model, investment is defined implicitly from Kt+1 = (1 − δ)Kt + It and therefore
It = Kt+1 − (1 − δ)Kt . In steady state, we have I ? = K ? − (1 − δ)K ? = δK ? . Using the
value for K ? from (9) gives
  1
? β(1 − τ )α 1−α
I =δ AL. (11)
1 − β(1 − δ)
What is the savings rate? Since we can define it implicitly (and generally, for any t) as
?
It = st Yt , it is st = YItt . In steady state it is therefore simply s? = YI ? . Replacing I ? from
(11) and Y ? from (10) yields
  1
β(1−τ )α 1−α  1−α
δ 1−β(1−δ) AL 
β(1 − τ )α 1−α δβ(1 − τ )α
?
s =  α =δ = . (12)
1 − β(1 − δ) 1 − β(1 − δ)
 1−α
β(1−τ )α
1−β(1−δ)
AL

h) We see that in steady state, the savings rate is constant and a function of exogenous
parameters. Remember, in the Solow model, a constant and exogenous savings rate is
assumed. Here, we also obtain a constant savings rate. This means that in the steady
state, the Ramsey model yields similar predictions to to the Solow model (just by observ-
ing the investment and capital in steady state in an economy following Ramsey, we could
easily conclude that it follows Solow). But we need to remember that this only holds in
steady state. Outside the steady state, the Solow savings rate is exogenous and constant
because this is the underlying model assumption. In Ramsey, outside the steady state, the
savings rate is not constant, it is complicated function of the household’s intertemporal
optimization problem.

In addition, note the even in steady state, the Ramsey savings rate is an endogenous
function of model parameters. It is therefore not assumed, but an outcome of the econ-
omy. This is much richer now. For example, we could analyze why exactly an economy
tends to have a long-run (≈ steady state) savings rate that is higher that that of another
economy. The parameters in (12) provide a clue. It could be that the household is more
patient (higher β), or that the production function is more intensive in capital (higher
α), or that the depreciation rate is higher, requiring more investment (higher δ) or that
there must be something else, a lower wedge τ , that makes it more lucrative to save and
invest.

i) Since Ct + It = Yt , Ct = Yt − It , and in steady state C ? = Y ? − I ? . Replacing Y ?

from (10) and I ? from (11) yields
  α   1
? β(1 − τ )α 1−α β(1 − τ )α 1−α
C = AL − δ AL
1 − β(1 − δ) 1 − β(1 − δ)
"  α   1 #
β(1 − τ )α 1−α β(1 − τ )α 1−α
= −δ AL.
1 − β(1 − δ) 1 − β(1 − δ)
For our purposes, it is more convenient to factor out the term with 1 − τ , so let us write
α 1
"  1−α   1−α #
βα α βα 1
C ? =AL (1 − τ ) 1−α − δ (1 − τ ) 1−α .
1 − β(1 − δ) 1 − β(1 − δ)
We see that different values of τ will lead to different steady state values of C ? . In par-
ticular, note that for τ = 1, C ? = 0. In this scenario the household’s output is completely
expropriated in each period, so the household does not perceive any return from investing
6

in capital. As a result, it drives capital to 0 in steady state so that in the long-run there
is no production and therefore also no consumption.
?
j) We are asked to find the value of τ that maximizes C ? . For this we compute ∂C∂τ(τ ) and
set it equal to 0.4 Hence:
α 
∂C ? (τ )
  1−α 
βα α α ∂(1 − τ )
=AL (1 − τ ) 1−α −1
∂τ 1 − β(1 − δ) 1−α ∂τ
 1
 1−α  
βα 1 1 ∂(1 − τ )
− ALδ (1 − τ ) 1−α −1 .
1 − β(1 − δ) 1−α ∂τ
∂(1−τ )
Given that ∂τ
= −1, the two terms hence change signs:
α 
∂C ? (τ )
  1−α 
βα α α
= − AL (1 − τ ) 1−α −1
∂τ 1 − β(1 − δ) 1−α
 1
 1−α  
βα 1 1
+ ALδ (1 − τ ) 1−α −1 .
1 − β(1 − δ) 1−α
?
The first-order condition is ∂C∂τ(τ ) = 0, so
 α 
 1−α 
βα α α
− AL (1 − τ ) 1−α −1
1 − β(1 − δ) 1−α
 1
 1−α  
βα 1 1
+ ALδ (1 − τ ) 1−α −1 = 0
1 − β(1 − δ) 1−α
 α 
 1−α 
βα α α
⇔AL (1 − τ ) 1−α −1
1 − β(1 − δ) 1−α
 1
 1−α  
βα 1 1
= ALδ (1 − τ ) 1−α −1
1 − β(1 − δ) 1−α
 α 
 1−α    1−α1  
βα α α
−1 βα 1 1
⇔ (1 − τ )
1−α =δ (1 − τ ) 1−α −1
1 − β(1 − δ) 1−α 1 − β(1 − δ) 1−α
 α
 1−α  1
 1−α
βα α βα 1
⇔ α(1 − τ ) 1−α = δ (1 − τ ) 1−α
1 − β(1 − δ) 1 − β(1 − δ)
  1−α
βα 1−α 1−α
⇔α = δ (1 − τ ) 1−α
1 − β(1 − δ)
 
βα
⇔α = δ (1 − τ )
1 − β(1 − δ)
[1 − β(1 − δ)]
⇔ = (1 − τ )
δβ
[1 − β(1 − δ)]
⇔τ = 1 −
δβ

4Notethat this procedure could in principle also lead to find the minimum possible C ? (or a local
minimum or maximum). It can be checked that the second-order condition is satisfied and that the
function has no local minima/maxima, so the first-order condition indeed yields a maximum.
 7

δβ [1 − β(1 − δ)]
⇔τ = −
δβ δβ
δβ − [1 − β(1 − δ)]
⇔τ =
δβ
δβ − (1 − β) − βδ
⇔τ =
δβ
1−β
⇔τ = − .
δβ
After all this algebra it’s convenient to stand up and stretch the fingers to avoid suffering
a cramp.

Still, the effort is worth it, because we now have a very nice expression. Because β ∈ (0, 1),
we see that the tax that maximizes C ? is negative, τ < 0. So a government that wishes to
achieve the highest possible steady-state consumption would subsidize output (and pay
for the subsidy via a lump-sum tax on households, meaning a negative transfer). Now,
in (a) we said that the optimal tax here is τ = 0. How can we square those two results?
The point is that the household’s objective is not to achieve the highest possible C ? , but
to maximize discounted lifetime utility. In particular, the household discounts time, and
hence maintaining the (maximum possible) steady state associated with τ < 0 is just too
costly in terms of the opportunity cost. The household would prefer to consume a bit
more early on and maintain a steady state consumption lower than the maximum one.
You can see that clearly in the limit for a household that is completely patient, namely
β → 1. In this case the expression τ = − 1−β δβ
→ 0. So the really, really patient household
?
maintains the maximum feasible C in absence of any subsidy. What the subsidy does is
to “correct” consumption for the households that are impatient.

In textbooks, the associated highest possible steady-state consumption is also referred

to as “golden” steady-state consumption, C ?gold . But note that this is not optimal from
the perspective of period 0. The household in our model would prefer the government to
leave it alone as opposed to intervene with a paternalistic subsidy.

k) In this exercise (and in general in the presentation of the Ramsey model in Lecture 4)
we do not focus much on how the variables of interest (K, C, and Y ) move dynamically
outside the steady state. Instead, the focus is more on outcomes in steady state. The rea-
son for this is that a proper analysis of the dynamics of the system is technically slightly
more challenging. Nonetheless, it is possible to provide an intuition for such movements.
Now, we are asked what happens to an economy that is in steady state and that suddenly
receives an influx of capital such that K becomes K > K ? . That is, we can think of how
the economy will evolve starting now from some K0 > K ? .

First, we know that the economy eventually converges back to K ? . So the likely scenario
is for the economy to start shedding capital, in the sense that K0 > K1 > K2 > ... = K ? .
Because output is an increasing function of capital, this implies that output would fall
over time as well, namely Y0 > Y1 > Y2 > ... = Y ? . Given what we know from (e), this
would also imply that as time goes on, the growth rate of consumption C would rise. Yes,
but that growth rate is negative to begin with, and remains negative until we reach the
steady state. Consumption, therefore, would also decrease, C0 > C1 > C2 > ... = C ? .
We do not prove that this pattern is necessary, but a more detailed analysis would reveal
that this is indeed the case. What is going on? Quite simply, the household in the initial
8

period all of a sudden has “too much” capital. Not “too much” in the sense that it is
unhappy about the gift, but too much in the sense that it has no interest in maintaining it.
Instead, the household immediately drives up its initial consumption C0 and then steadily
decreases it up to the point where it reaches C ? . So why not just have a huge increase in
C0 and then go back immediately to C ? ? Because the household’s period utility function
is concave and it values a more smooth consumption stream over time. Intuitively, throw-
ing a humongous party today and then return to “normal” tomorrow is just not worth it
because of diminishing returns to the party today. Instead, the household throws a nice
(but not excessive) party today, and then another, smaller one tomorrow, and another
even smaller one the day after tomorrow, etc.

Question 2

a) The firm’s profits are

Π(Kt , Lt ) = max{Yt (Kt , Lt ) − Rt Kt − Wt Lt } = max{Ktα (ALt )1−α − Rt Kt − Wt Lt }.
Kt ,Lt Kt ,Lt

The resulting first-order conditions with respect to Kt and Lt are:

αKtα−1 (ALt )1−α − Rt = 0
and
(1 − α)Ktα A1−α L−α
t − Wt = 0.
Hence, these can be written as
Rt = αKtα−1 (ALt )1−α
and
Wt = (1 − α)Ktα A1−α L−α
t .

We know that in equilibrium, the demand for labor Lt , must equal its supply, L. Demand
for capital also needs to equal its supply, but this is endogenously determined by the
household, Kt . The two equations are therefore
Rt = αKtα−1 (AL)1−α (13)
and
Wt = (1 − α)Ktα A1−α L−α . (14)

b) The household’s Lagrangian is

∞ 1−θ ∞
X
t Ct−1 X 
λt (1 − τ K )Rt Kt + (1 − τ L )Wt L + Tt − Ct − Kt+1 + (1 − δ)Kt .

L= β +
t=0
1−θ t=0
(15)
∂L
The household’s first-order conditions with respect to Ct and Kt+1 are ∂Ct
= 0 and
Ct−θ
∂L
∂Kt+1
= 0. The first one is simply β t (1 − θ) 1−θ − λt = 0 and therefore

β t Ct−θ − λt = 0. (16)
The first-order condition with respect to Kt+1 is
− λt + λt+1 (1 − τ K )Rt+1 + 1 − δ = 0.
 
(17)
 9

The Euler equation can be found by replacing the Lagrange multipliers in equations (16)
and (17). The first one gives λt = β t Ct−θ and therefore also λt+1 = β t+1 Ct+1
−θ
. Replacing
these two expressions in (3) yields
− β t Ct−θ + β t+1 Ct+1
−θ
(1 − τ K )Rt+1 + 1 − δ = 0
 

⇔β t Ct−θ = β t+1 Ct+1

−θ
(1 − τ K )Rt+1 + 1 − δ
 

⇔Ct−θ = βCt+1 −θ
(1 − τ K )Rt+1 + 1 − δ
 

θ
Ct−θ = β (1 − τ K )Rt+1 + 1 − δ
 
⇔Ct+1
or simply
 θ
Ct+1
= β 1 + (1 − τ K )Rt+1 − δ .
 
(18)
Ct
Because we are now in a decentralized economy with capital and labor markets, the
household’s consumption choice depends on the return  that it receives from
 investing in
K
capital, Rt+1 . Consumption will grow as long as β 1 + (1 − τ )Rt+1 − δ > 1, namely if
the (after-tax) return on capital tomorrow is sufficiently large to overcome the household’s
impatience, β < 1. Also, the rate at which consumption change depends on the size of θ,
i.e., the household’s desire to smooth consumption.

c) We start with the household’s key equations, namely the household’s budget constraint
Ct + Kt+1 − (1 − δ)Kt = (1 − τ K )Rt Kt + (1 − τ L )Wt L + Tt . (19)
and the household’s optimality condition (18):
 θ
Ct+1
= β 1 + (1 − τ K )Rt+1 − δ .
 
(20)
Ct
We are asked to summarize the equilibrium so that C and K are the only variables. In
the above, we therefore need to get rid of the additional endogenous variables, namely
W , R, and T . We are told that the transfer equals
Tt = τ K Rt Kt + τ L Wt L.
Replacing this in (19) gives
Ct + Kt+1 − (1 − δ)Kt =(1 − τ K )Rt Kt + (1 − τ L )Wt L + τ K Rt Kt + τ L Wt L
=Rt Kt + Wt L.
Next, we can replace Rt and Wt from (13) and (14), respectively:
Ct + Kt+1 − (1 − δ)Kt = αKtα−1 (AL)1−α × Kt + (1 − α)Ktα A1−α L−α × L
⇔Ct + Kt+1 − (1 − δ)Kt = αKtα (AL)1−α + (1 − α)Ktα (AL)1−α
and hence
Ct + Kt+1 − (1 − δ)Kt = Ktα (AL)1−α . (21)
As for the Euler equation (20), after replacing (13) at t + 1 it becomes
 θ
Ct+1
= β 1 + (1 − τ K )αKt+1
α−1
(AL)1−α − δ .
 
(22)
Ct

d) Compare the resulting equations (21) and (22) to (6) and (7), respectively. What
values of τ K and τ L are such the equations are identical? We see that (21) is identical
to (6). As for (22), it is identical to (7) if τ K = τ . In other words, the tax rate on
10

capital income must be identical to the tax rate on the overall economic activity in the
previous economy. In that case, assuming that both economies start off with the same
K0 , the allocations Ct and Kt+1 in any t = {0, 1, 2, ...} are identical in the two economies.
Finally, since the output function is Yt = Ktα (AL)1−α is in both cases, output in the two
economies will also equal as long as τ K = τ .

The other thing to notice is that in our decentralized economy, τ L has no impact on
C, K or Y . Why? The reason for this is that τ L is a tax on labor income. Labor is in
fixed supply in this economy (it always equals L, no matter what) so a tax on its income
will not change how much is supplied. This is different from capital, which is endogenous.
As a result, the tax rate on capital income, τ K , affects the return on capital and therefore
also how much capital the household wants to accumulate.

More generally, notice that for τ = 0 the economy in Question 1 is a standard centralized
Ramsey economy where the household makes all decisions, while the present model with
τ K = 0 and any τ L (but typically set to τ L = 0 in textbook treatments) is a standard
decentralized Ramsey economy with a capital and labor market. The two economies yield
identical equilibria. This is interesting, because it implies that simplifying the problem
by having the household make all decisions (as opposed to introducing capital and labor
markets) does not change the outcome with respect to a more sophisticated model. The
reason for this is that competitive markets basically reproduce the optimal allocation.
There are several properties at work here that make this possible. One of them is that
there are no externalities.

From now on, we suppose that

At = Ktγ (23)
with γ ∈ (0, 1). In particular, we suppose that the household and firm do not internalize
that relationship.

e) There are various ways of replying to this question. One is to think of At as the
level of know-how (or the abundance of ideas) in the economy. By linking A to K we
make the assumption that ideas are created as a spillover or by-product of capital invest-
ment. For example, this could capture the notion that during the process of investment,
firms create innovations that can be costlessly replicated in the economy and that raise
overall output. The parameter γ measures the strength of the spillover. In particular,
assuming γ ∈ (0, 1) implies that while an increase in Kt will necessarily positively impact
At , it also implies that there are diminishing marginal returns in Kt . In other words, At
grows more slowly than Kt .

f) Consider the Euler equation (22). Notice that we have to replace the A in that equation
γ
by At+1 , which in equilibrium is At+1 = Kt+1 , so:
 θ
Ct+1 h
γ

1−α i
= β 1 + (1 − τ K )αKt+1 α−1
Kt+1 L −δ
Ct
h i
K α−1 γ(1−α) 1−α
= β 1 + (1 − τ )αKt+1 Kt+1 L −δ (24)
h i
(α−1)(1−γ) 1−α
= β 1 + (1 − τ K )αKt+1 L −δ .
 11

In steady state we have that Ct+1 = Ct = C ? and Kt+1 = Kt = K ? , so the equation

becomes
1 = β 1 + (1 − τ K )αK ?(α−1)(1−γ) L1−α − δ .
 

Now, we can solve for K ? , namely:

1
= 1 + (1 − τ K )αK ?(1−γ)(α−1) L1−α − δ
β
1
⇔ − (1 − δ) = (1 − τ K )αK ?(1−γ)(α−1) L1−α
β
1 − β(1 − δ)
⇔ = (1 − τ K )αK ?(1−γ)(α−1) L1−α
β

1 − β(1 − δ) ?(1−γ)(1−α)
⇔ K = (1 − τ K )αL1−α
β
β(1 − τ K )α 1−α
⇔K ?(1−γ)(1−α) = L
1 − β(1 − δ)
1
β(1 − τ K )α (1−γ)(1−α) (1−γ)(1−α)
 
1−α
?
⇔K = L
1 − β(1 − δ)
and thus
1
β(1 − τ K )α
  (1−γ)(1−α)
1
?
K = L 1−γ . (25)
1 − β(1 − δ)
α+γ(1−α)
Next, since Yt = Ktα (At L)1−α = Ktα (Ktγ L)1−α we have in fact Yt = Kt L1−α . And
hence in steady state, Y ? = K ?α+γ(1−α) L1−α . Replacing K ? by its expression from (25)
obtains
" 1 #α+γ(1−α)
K
 (1−γ)(1−α)
β(1 − τ )α 1
Y? = L 1−γ L1−α
1 − β(1 − δ)
 α+γ(1−α)
β(1 − τ K )α (1−γ)(1−α) α+γ(1−α)

= L 1−γ L1−α
1 − β(1 − δ)
 α+γ(1−α)
β(1 − τ K )α (1−γ)(1−α) α+γ(1−α)

(1−α)(1−γ)
= L 1−γ L 1−γ
1 − β(1 − δ)

α+γ(1−α)
β(1 − τ K )α
  (1−γ)(1−α)
α+γ(1−α)+(1−α)(1−γ)
= L 1−γ
1 − β(1 − δ)
α+γ(1−α)
β(1 − τ K )α
  (1−γ)(1−α)
α+γ(1−α)+1−α−γ(1−α)
= L 1−γ
1 − β(1 − δ)
α+γ(1−α)
β(1 − τ K )α
  (1−γ)(1−α)
α+1−α
= L 1−γ
1 − β(1 − δ)
and therefore
α+γ(1−α)
β(1 − τ K )α
  (1−γ)(1−α)
1
?
Y = L 1−γ . (26)
1 − β(1 − δ)
12

?
g) Using (26), output per worker in steady state, y ? = YL , is therefore
 α+γ(1−α) 1
β(1 − τ K )α (1−γ)(1−α) L 1−γ

?
y =
1 − β(1 − δ) L
α+γ(1−α)
β(1 − τ K )α (1−γ)(1−α) 1−γ
 
1 1−γ
= L − 1−γ
1 − β(1 − δ)
and thus
 α+γ(1−α)
β(1 − τ K )α (1−γ)(1−α) 1−γ

γ
?
y = L .
1 − β(1 − δ)
γ
How does it depend on L? We see that since 1−γ > 0, an economy with more labor (which
can be thought of as simply reflecting a larger population), will have a higher steady state
GDP per worker ! Notice that this hinges on the fact that γ > 0. In contrast, with γ = 0
(which is the standard Ramsey model with no linkage between K and A) we would obtain
the usual result where L does not impact the long-run steady state output per worker. So
what is the intuition for this radically different outcome? It has to do with the fact that
technology, A, is assumed to be positively affected by aggregate capital K. An economy
with a larger population will accumulate more K, and this, in turn will have a positive
spillover on A and therefore long-run GDP per worker. This is actually a remarkable
result. If we believe indeed that technology reflects aggregate investment in K, then a
larger population (at least in steady state) turns out to be a good thing for the average
worker.

Next, we modify the setting slightly by assuming that γ = 1.

h) With γ = 1, technology is in fact linear in capital, At = Kt . Notice that as γ → 1

(i.e., approaches 1), the steady state values of K ? and Y ? in (25) and (26) in fact tend to
infinity. It turns out that now, the steady state, in fact, does not exist. We can see this
most clearly from the Euler equation after substituting in At = Kt , namely equation (24)
where γ = 1. It becomes
 θ
Ct+1 h
(α−1)(1−1) 1−α
i
= β 1 + (1 − τ K )αKt+1 L −δ
Ct
= β 1 + (1 − τ K )αL1−α − δ
 

and hence simply

 θ
Ct+1
= β 1 + (1 − τ K )αL1−α − δ
 
Ct
Notice that the right-hand side is constant at any t, so the growth rate is constant g C =
C
gt+1 = CCt+1
t
− 1, namely

1
g C = β 1 + (1 − τ K )αL1−α − δ θ − 1.

(27)
This is indeed remarkable. We see that C grows at a constant rate forever!!! In general,
there is no steady state. The only exception is the case where by coincidence Ct+1 = Ct ,
which is knife-edge case when the parameter constellation is exactly such that g C = 0:
1 = β 1 + (1 − τ K )αL1−α − δ .
 
 13

Other than that, there exists no steady state in this economy. Consumption just keeps
on growing forever, either at a positive or negative rate depending on whether the term
β 1 + (1 − τ K )αL1−α − δ is larger or smaller than 1. What is going on? Now, with
γ = 1, the externality from K to A is sufficiently strong so that in equilibrium, the
production function no longer exhibits diminishing marginal returns to K. Consider the
production function Yt = Ktα (At L)1−α . An increase in K should in principle lead to
diminishing marginal returns through its direct impact. But because At = Kt , the pro-
duction function actually becomes Yt = Kt L1−α , and therefore linear in K. As a result,
the marginal return to capital is always a constant, independently of how much capital
there is in the economy.

If you are interested, you can check from the resource constraint Ct +Kt+1 −(1−δ)Kt = Yt ,
which is now Ct + Kt+1 − (1 − δ)Kt = Kt L1−α , that a constant growth rate of Ct implies
that capital Kt grows at the same constant rate, and therefore also Yt . We have therefore
found an endogenous growth model where all variables of interest grow at the rate g C
given in expression (27). The reasons that this growth is called endogenous is because the
long-run growth rate of the economy is not assumed but the result of the model’s other
parameters as can be seen in (27).

i) Consider the growth rate (27) which is the growth rate rate not only of consumption
but also of output. We see clearly that it is increasing in β. A more patient household
saves more, which increases capital in the next period. Since this does not lower the
capital’s rate of return, the economy simply grows more quickly. Similarly, the growth
rate is decreasing in τ K . The larger is the household’s perceived return to investing in
capital, the more it invests, leading to a higher growth rate in the absence of diminishing
returns. Finally, note that a larger population in (27) also increases the growth rate. The
reason for this is that a larger population increases the marginal (and constant) rate of
return to capital. Larger economies, according to this model, grow more quickly.

Before concluding this exercise, note that there are a number of strong assumptions that
helped generate the above results. One is that A is a function of capital, the other one was
that A is a pure externality, and the third one is that the externality is sufficiently strong
(γ = 1) that we obtained long-term endogenous growth forever. Some of these assump-
tions are highly stylized and would, in this form, not stand up to more careful empirical
scrutiny. However, the model does represent a stepping stone in the direction of endoge-
nous growth theory which strives to provide a plausible explanation for why economies
(as observed over the last century or so) do exhibit fairly constant long-run growth rates.
In all those theories it is typically assumed that there exists some type of spillover from
economic activity to ideas (Y or K → A), that giving incentives to innovation (here, very
crudely, a low τ K ) fosters long-run growth, and that there are so-called“scale effects” by
which a larger population (L) ultimately generates more ideas that everyone can use.