Lecture 2: Neoclassical Growth Theory

(Acemoglu 2009, Chapter 8), adapted from Fabrizio

Zilibotti

Kjetil Storesletten

August 27, 2013

Kjetil Storesletten () Lecture 2 August 27, 2013 1 / 57

Introduction

Ramsey or Cass-Koopmans model:

di¤ers from the Solow model insofar as it explicitly
models the consumer side and endogenizes savings
Beyond its use as a basic growth model,
it is also a workhorse for many areas of macroeconomics
Example: real and monetary business cycle theory

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 2 / 57

Preferences, Technology and Demographics I
In…nite-horizon, continuous time.
Representative household with instantaneous utility function

u (c (t )) ,

Assumption u (c ) is strictly increasing, concave, twice continuously

di¤erentiable with derivatives u 0 and u 00 , and satis…es
the following Inada type assumptions:

lim u 0 (c ) = ∞ and lim u 0 (c ) = 0.

c !0 c !∞

Suppose representative household represents

set of identical households (normalized to 1)
Each household has an instantaneous utility function u (c (t ))
L (0) = 1 and
L (t ) = exp (nt )
Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 3 / 57
Preferences, Technology and Demographics II
All members of the household supply their labor inelastically
Objective function of the representative household at t = 0:
Z ∞
U (0) exp ( ρt ) L (t ) u (c (t )) dt (1)
Z0 ∞
= exp ( (ρ n ) t ) u (c (t )) dt,
0

where
I c (t )=consumption per capita at t,
I ρ=subjective discount rate, so that e¤ective discount rate is ρ n.
Objective function (1) embeds:
I Household is fully altruistic towards all of its future members,
and makes allocations of consumption
(among household members) cooperatively
I Strict concavity of U ( )
C (t )
Thus each household member has an equal consumption, c (t ) L (t )

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 4 / 57

Preferences, Technology and Demographics III

Assumption: ρ > n
Benchmark model without any technological progress
Factor and product markets are competitive
Production possibilities set of the economy is represented by

Y (t ) = F [K (t ) , L (t )] ,

F features constant returns to scale and Inada conditions, i.e.,

Y = FK K + FL L (Euler Theorem) and
limK !0 FK = limL !0 FL = ∞, limK !∞ FK = limL !∞ FL = 0

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 5 / 57

Preferences, Technology and Demographics IV

De…ne variables in p.c. terms, x (t ) X (t ) /L (t )

Per capita production function f ( )

K (t )
y (t ) = F ,1 f (k (t )) ,
L (t )

Competitive factor markets imply:

R (t ) = FK [K (t ), L(t )] = f 0 (k (t )).

and (from the Euler theorem)

F [K (t ), L(t )] K (t )
w (t ) = FL [K (t ), L(t )] = FK [K (t ), L(t )]
L (t ) L (t )
0
= f (k (t )) k (t ) f (k (t )).

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 6 / 57

Preferences, Technology and Demographics V
Denote asset holdings of the representative household at time t by
A (t ). Then,

Ȧ (t ) = r (t ) A (t ) + w (t ) L (t ) c (t ) L (t )

r (t ) is the market ‡ow rate of return on assets, and w (t ) L (t ) is the

‡ow of labor income earnings of the household.
De…ning per capita assets as
A (t )
a (t ) ,
L (t )
we obtain:

ȧ (t ) = (r (t ) n ) a (t ) + w (t ) c (t )

Household assets can consist of capital stock, K (t ),

which they rent to …rms and bonds in zero net supply, B (t ).
Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 7 / 57
Preferences, Technology and Demographics VI

Given no uncertainty, arbitrage implies that the rate

of return on bonds must equal the net return on capital
(after depreciation at the rate δ).
Both returns must equal r (t ) )

r (t ) = R (t ) δ

Moreover, market clearing )

a (t ) = k (t )

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 8 / 57

The Budget Constraint I

Let us return to the ‡ow (or dynamic) budget constraint

ȧ (t ) = (r (t ) n ) a (t ) + w (t ) c (t )

Imposing that the ‡ow constraint holds for all t 2 [0, ∞[

is not su¢ cient to ensure that a proper budget constraint
hold unless we impose a lower bound on assets
A dynasty could increase its consumption
by running an ever growing debt

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 9 / 57

The Budget Constraint II
No-Ponzi Game Condition
Total debt cannot grow at a rate exceeding the interest rate;
Z t
lim A (t ) exp r (s ) ds 0.
t !∞ 0

Equivalently, debt per capita cannot grow at a rate higher than r n:

Z t
lim a (t ) exp (r (s ) n ) ds 0.
t !∞ 0

Since it will never be optimal to have positive wealth asymptotically

(formally, this will be captured by a Transversality Condition, TVC)
the no-Ponzi-game condition can in fact be strengthened to:
Z t
lim a (t ) exp (r (s ) n ) ds = 0.
t !∞ 0

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 10 / 57

The Budget Constraint III

The no-Ponzi Game rules out the possibility for agents

to borrow to …nance present consumption and then use
future borrowings to roll over the debt and pay the interest
It can be shown formally (see textbook for a proof) that
the no-Ponzi-game condition + period budget constraint
ensures that the individual’s lifetime budget constraint holds in
in…nite horizon
Z ∞ Z t
c (t ) exp (r (s ) n ) ds dt
0 0
Z ∞ Z t
= a (0) + w (t ) exp (r (s ) n ) ds dt
0 0

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 11 / 57

De…nition of Equilibrium

De…nition A competitive equilibrium of the Ramsey economy consists

of paths [C (t ) , K (t ) , w (t ) , R (t )]t∞=0 , such that the
representative household maximizes its utility given initial
capital stock K (0) and the time path of prices
[w (t ) , R (t )]t∞=0 , and all markets clear.

Notice:
the de…nition refers to the entire path of quantities and prices,
not just steady-state equilibrium.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 12 / 57

Household Maximization I
Set up the current-value Hamiltonian:

Ĥ (a, c, µ) = u (c (t )) + µ (t ) [w (t ) + (r (t ) n ) a (t ) c (t )] ,

The solution must satisfy

(1) FOC : Ĥc (a, c, µ) = 0

0
, u (c (t )) = µ (t )

(2)EE : Ĥa (a, c, µ) = µ̇ (t ) + (ρ n ) µ (t ) = (r (t ) n ) µ (t )

µ̇ (t )
, = (r (t ) ρ )
µ (t )

(3) BC: ȧ (t ) = (r (t ) n ) a (t ) + w (t ) c (t )
(4) TVC: lim [exp ( (ρ n ) t ) µ (t ) a (t )] = 0
t !∞

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 13 / 57

Household Maximization II

Take logarithms in the FOC and di¤erentiate with respect to time

u 00 (c (t )) c (t ) ċ (t ) µ̇ (t )
= .
u 0 (c (t )) c (t ) µ (t )

Substituting into EE, obtain another form of the consumer Euler

equation:
ċ (t ) 1
= (r (t ) ρ )
c (t ) εu (c (t ))
where
u 00 (c (t )) c (t )
εu (c (t ))
u 0 (c (t ))
is the elasticity of the marginal utility u 0 (c (t )).
Consumption will grow over time when the discount rate is less than
the rate of return on assets.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 14 / 57

Household Maximization III

Speed at which consumption will grow is related to the IES,

elasticity of marginal utility of consumption, εu (c (t )).
εu (c (t )) can also be interpreted (see book) as the inverse
of the intertemporal elasticity of substitution (IES):
I regulates willingness to substitute consumption over time.
Suppose (
c1 θ 1
if θ 6= 1 and θ 0
u (c ) = 1 θ ,
ln c if θ = 1
This utility function (CRRA) induces a constant IES.
In particular, εu (c (t )) = θ, so 1/θ is the constant IES.
CRRA is necessary to have balanced growth.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 15 / 57

Household Maximization IV

Under CRRA utility,

θ
µ (t ) = c (t )
and the consumer Euler equation yields:

µ̇ (t ) ċ (t ) ċ (t ) r (t ) ρ
= (r (t ) ρ) = θ =) =
µ (t ) c (t ) c (t ) θ

Thus, integrating,
Z t
µ (t ) = µ (0) exp (r (s ) ρ) ds ,
0

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 16 / 57

Household Maximization V

Consider the TVC

0 = lim [exp ( (ρ n ) t ) a (t ) µ (t )]
t !∞

Z t
= lim exp ( (ρ n ) t ) a (t ) µ (0) exp (r (s ) ρ) ds
t !∞ 0
Z t
θ
= lim a (t ) exp (r (s ) n ) ds c (0) .
t !∞ 0
h Rt i
Thus, limt !∞ a (t ) exp 0 (r (s ) n ) ds =0
We can now provide an interpretation of the TVC

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 17 / 57

Transversality Condition I

The transversality condition is a complementary condition that must

hold (in standard problems) in order for the consumption/savings
plan of the individual agents to be optimal.
In a …nite-horizon problem,
the TVC has a straightforward interpretation:
the discounted value of the stock of capital left
at the end of the planning period (T ) must be zero
RT
(r ( ν ) n ) d ν
a (T ) e 0 =0

As long as the interest rate is …nite, the second term is positive,

which reduces itself to the intuitive condition that aT = 0.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 18 / 57

Transversality Condition II

In the in…nite horizon, we take the limit of

the …nite-horizon condition as T tends to in…nity:
h RT i
lim a (T ) e 0 (r (ν) n ) d ν = 0
T !∞

Interpretation: the PDV of assets at the “end of life” (in…nity)

must be zero. However, now a (t ) needs not converge to zero.
A simple case in which the TVC holds is an economy converging
to a steady-state where both a (t ) and r (t ) are constant

However, the TVC can also hold if a (t ) ! ∞

as long as the second term goes to zero "su¢ ciently fast"

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 19 / 57

Equilibrium Prices I
Equilibrium prices are given by

R (t ) = f 0 (k (t )) and w (t ) = f (k (t )) k (t ) f 0 (k (t )).
Since r (t ) = R (t ) δ, then
r (t ) = f 0 (k (t )) δ.
Substituting this into the consumer’s EE, we have
ċ (t ) f 0 (k (t )) δ ρ
=
c (t ) θ
θ
Moreover, since a (t ) = k (t ) and µ (t ) = c (t ) , the TVC can be
written as
lim [exp ( (ρ n ) t ) µ (t ) a (t )] =
t !∞
h i
θ
lim exp ( (ρ n ) t ) c (t ) k (t ) = 0
t !∞

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 20 / 57

Equilibrium Prices II

Finally, let us go back to the individual budget constraint

ȧ (t ) = (r (t ) n ) a (t ) + w (t ) c (t )
And using the equilibrium conditions

a (t ) = k (t )
r (t ) = f 0 (k (t )) δ
w (t ) = f (k (t )) k (t ) f 0 (k (t ))

We conclude that

k̇ (t ) = f (k (t )) (n + δ )k (t ) c (t ) ,

that can be interpreted as an aggregate resource constraint.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 21 / 57

Optimal Growth I

In an economy that admits a representative household, optimal

growth involves maximization of utility of representative household
subject to technology and feasibility constraints:
Z ∞
max exp ( (ρ n ) t ) u (c (t )) dt,
[k (t ),c (t )]t∞=0 0

subject to
k̇ (t ) = f (k (t )) (n + δ )k (t ) c (t ) ,
and k (0) > 0.
Versions of the First and Second Welfare Theorems for economies
with a continuum of commodities: solution to this problem should be
the same as the equilibrium growth problem.
Let us show the equivalence directly.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 22 / 57

Optimal Growth II
Again set up the current-value Hamiltonian:

H̃ (k, c, µ) = u (c (t )) + µ (t ) [f (k (t )) (n + δ )k (t ) c (t )] ,

The solution must satisfy

FOCPL : H̃c (k, c, µ) = 0

0
, u (c (t )) = µ (t )

and EEPL :

H̃k (k, c, µ) = µ̇ (t ) + (ρ n ) µ (t ) = f 0 (k (t )) δ n µ (t )
µ̇ (t )
, = f 0 (k (t )) δ ρ
µ (t )

RC : k̇ (t ) = f (k (t )) (n + δ )k (t ) c (t )
TVCPL : lim [exp ( (ρ n ) t ) µ (t ) k (t )] = 0
t !∞

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 23 / 57

Optimal Growth III

Assume CRRA. Repeating the same steps as before,

ċ (t ) f 0 (k (t )) δ ρ
= ,
c (t ) θ

k̇ (t ) = f (k (t )) (n + δ)k (t ) c (t ) ,
h i
lim exp ( (ρ n ) t ) c (t ) θ k (t ) = 0
t !∞

which are identical to the laissez-faire equilibrium conditions.

Thus the competitive equilibrium is a Pareto optimum and the Pareto
allocation can be decentralized as a competitive equilibrium.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 24 / 57

Steady-State Equilibrium I
Steady-state equilibrium is de…ned as an equilibrium path in which
capital-labor ratio, consumption and output are constant. Thus:

ċ (t ) f 0 (k ) δ ρ
= =0
c (t ) θ

() f 0 (k ) = ρ + δ

Pins down the steady-state capital-labor ratio only as a function of

the production function, the discount rate and the depreciation rate.
Then, the resource constraint pins down the steady-state
consumption level:

k̇ (t ) = f (k (t )) (n + δ )k (t ) c (t ) = 0

() c = f (k ) (n + δ ) k .

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 25 / 57

Steady-State Equilibrium II

A steady state where the capital-labor ratio and thus output are
constant necessarily satis…es the TVC:
h i
lim exp ( (ρ n ) t ) k (c ) θ = 0
t !∞

which is true as long as ρ > n.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 26 / 57

Transitional Dynamics I

Equilibrium is determined by two di¤erential equations:

k̇ (t ) = f (k (t )) (n + δ )k (t ) c (t )

ċ (t ) f 0 (k (t )) δ ρ
=
c (t ) θ
plus an initial condition, k (0) > 0, and a terminal condition:
h i
lim exp ( (ρ n ) t ) k (t ) (c (t )) θ = 0.
t !∞

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 27 / 57

Transitional Dynamics II

Appropriate notion of saddle-path stability:

I c (or, equivalently, µ) is the control variable, and c (0) (or µ (0)) is
free: it has to adjust to satisfy transversality condition
I If there were more than one path equilibrium would be indeterminate.
Economic forces are such that indeed there will be a one-dimensional
manifold of stable solutions tending to the unique steady state.
See Figure.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 28 / 57

 c(t)=0
c(t)

c*

k(t)=0
c’(0)

c(0)
c’’(0)
k(t)
0 k(0) k* kgold k

Figure: Transitional dynamics in the baseline neoclassical growth model

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 29 / 57

Transitional Dynamics: Global Stability Analysis

Intuitive argument:
I if c (0) started below it, say c 00 (0), consumption would reach zero,
thus capital would accumulate continuously until the maximum level of
capital (reached with zero consumption). This would violate the
transversality condition.
I if c (0) started above this stable arm, say at c 0 (0), the capital stock
would reach 0 in …nite time, while consumption would remain positive.
But this would violate feasibility (a little care is necessary with this
argument, since necessary conditions do not apply at the boundary).

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 30 / 57

Technological Change and the Canonical Neoclassical
Model I

Extend the production function to:

Y (t ) = F [K (t ) , A (t ) L (t )] ,

where
A (t ) = exp (gt ) A (0) .
Note: we assume labor-augmenting technological change.
Else, there would be no balanced growth equilibrium

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 31 / 57

Technological Change and the Canonical Neoclassical
Model II

De…ne x̂ (t ) X (t ) / (A (t ) L (t ))

K (t )
ŷ (t ) = F ,1 f k̂ (t ) ,
A (t ) L (t )

Assume CRRA preferences

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 32 / 57

Equilibrium

The equilibrium is now fully characterized by the following dynamic

equations
ĉ (t ) 1 0
= f k̂ (t ) δ ρ θg ,
ĉ (t ) θ

k̂ (t ) = f k̂ (t ) (n + g + δ) k̂ (t ) ĉ (t ) ,
plus an initial condition, k̂ (0) > 0, and a terminal condition (TVC)
n o
= lim exp ( (ρ n (1 θ ) g ) t ) k̂ (t ) (ĉ (t )) θ = 0.
t !∞

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 33 / 57

Equilibrium (derivation EE, see book for more)

r (t ) = f 0 k̂ (t ) δ

c (t ) ĉ (t )
Since c (t ) = A (t ) ĉ (t ) , then c (t )
= ĉ (t )
+g
Then:

c (t ) 1 ĉ (t ) 1 0
= (r (t ) ρ) () = f k̂ (t ) δ ρ θg
c (t ) θ ĉ (t ) θ

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 34 / 57

Steady-State

In steady state, f 0 k̂ = ρ + δ + θg .
Pins down the steady-state value of
the normalized capital ratio k̂ uniquely.
Normalized consumption level is then given by

ĉ = f k̂ (n + g + δ) k̂ ,

Per capita consumption grows at the rate g .

The TVC now requires ρ n > (1 θ) g .

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 35 / 57

Technological Change and the Canonical Neoclassical
Model XI

Proposition Consider the neoclassical growth model with labor

augmenting technological progress at the rate g and CRRA
preferences. Suppose that ρ n > (1 θ ) g . Then there
exists a unique balanced growth path with a normalized
capital to e¤ective labor ratio of k̂ , given by
f 0 k̂ = ρ + δ + θg , and output per capita and
consumption per capita grow at the rate g .

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 36 / 57

Transitional dynamics
c(t)=0
c(t)

c*

k(t)=0
c’(0)

c(0)
c’’(0)
k(t)
0 k(0) k* kgold k

Figure: When g>0, simply replace k and c by k̂ and ĉ

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 37 / 57
Comparative Dynamics I

Comparative statics: changes in steady state in response to changes

in parameters.
Comparative dynamics look at how the entire equilibrium path of
variables changes in response to a change in policy or parameters.
E.g.: Initial steady state represented by (k , c ). Unexpectedly,
discount rate declines to ρ0 < ρ.
Following the decline ĉ is above the stable arm of the new dynamic
system: consumption must drop immediately

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 38 / 57

 c(t)=0
c(t)

c**

c*

k(t)=0

k(t)
0 k* k** kgold k

Figure: The dynamic response of capital and consumption to a decline in the

discount rate from ρ to ρ0 < ρ.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 39 / 57

The Role of Policy I

Introduce linear tax policy: returns on capital net of depreciation are

taxed at the rate τ and the proceeds of this are redistributed back to
the consumers.
Capital accumulation equation remains as above:

k̂ (t ) = f k̂ (t ) ĉ (t ) (n + g + δ) k̂ (t ) ,

But interest rate faced by households changes to:

r (t ) = (1 τ ) f 0 k̂ (t ) δ ,

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 40 / 57

The Role of Policy II

Growth rate of normalized consumption is then obtained from the

consumer Euler equation

ĉ (t ) 1
= (r (t ) ρ θg ) .
ĉ (t ) θ
1
= (1 τ ) f 0 k̂ (t ) δ ρ θg .
θ
This implies
ρ + θg
f 0 k̂ = δ+ .
1 τ
Since f 0 ( ) is decreasing, higher τ, reduces k̂ .
Higher taxes on capital have the e¤ect of depressing capital
accumulation and reducing income per capita.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 41 / 57

Appraisal neoclassical model

Major contribution: open the black box of capital accumulation by

specifying the preferences of consumers.
Also by specifying individual preferences we can explicitly compare
equilibrium and optimal growth.
Paves the way for further analysis of capital accumulation, human
capital and endogenous technological progress.
However, this model, by itself, does not enable us to answer questions
about the fundamental causes of economic growth.
But it clari…es the nature of the economic decisions so that we are in
a better position to ask such questions.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 42 / 57

AK model I

Neoclassical model: no autonomous engine of growth. In the absence

of exogenous trend, growth dies o¤ in the long-run.
1 No theory of determinants of long-run growth;
2 No theory of determinants of long-run cross-country di¤erences in
growth rates;
3 Policies do not a¤ect long-run growth.
The AK-Model is a very simple model that can be viewed as the "limit
case" of the neoclassical growth model. It provides the common
analytical framework for a number of more interesting applications.

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 43 / 57

AK model II

Production technology (g=0):

f (k ) = Ak

Equilibrium interest rate is r (t ) = A δ.

Assume CRRA utility
Given k (0) , a competitive equilibrium is determined by

δ ρA
ċ (t ) = c (t ) (EE)
θ
k̇ (t ) = Ak (t ) c (t ) (δ + n ) k (t ) (BC)
h i
lim exp ( (ρ n ) t ) k (t ) (c (t )) θ = 0 (TVC)
t !∞

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 44 / 57

AK model III
We can obtain an explicit analytical solution:
(a) Guess a steady-state solution such that c/k is constant
(assume A > d + ρ).
ċ (t ) k̇ (t ) A δ ρ
= = =γ
c (t ) k (t ) θ
(b) Use (BC)
k̇ (t ) c (t )
= (A δ n) .
k (t ) k (t )
(c) From (a)+(b):
c (t ) c 1 θ
= =ρ n [A δ ρ]
k (t ) k θ
In particular:
1 θ
c (0) = ρ n θ [A δ ρ ] k (0) .
Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 45 / 57
AK model IV
Hence (a) and the solution for c (0) , we obtain analytical solutions:
A δ ρ
c (t ) = c (0) exp t
θ
A δ ρ
k (t ) = k (0) exp t
θ
TVC (after replacing k (t ) by its solution):
k (t ) θ
0 = lim exp ( (ρ n) t ) c (t ) (c (t ))
t !∞ c (t )
c 1
= lim exp ( (ρ n) t ) (c (0))1 θ
t !∞ k
1 θ
exp (A δ ρ) t
θ
provided that the following condition (bounded utility) holds:
ρ > n + (1 θ ) (A δ n)
No transitional dynamics.
Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 46 / 57
AK model V

In this model, policies have permanent e¤ects on growth.

Consider again the introduction of a permanent tax on the returns on
capital. The proceeds are rebated as lump-sums.
The equilibrium interest rate is r = (1 τ ) (A δ), and the
equilibrium growth rate is:

(1 τ ) (A δ) ρ
γτ =
θ

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 47 / 57

Two Simple AK Models

Two simple models that deliver AK dynamics

Assume n=g=0
Basic human capital and knowledge spillovers

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 48 / 57

Basic Human Capital Model I

Suppose agents can accumulate both physical and human capital.

Let the technology be

Y = F (K , H ) = AK α H 1 α
= AK (H/K )1 α

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 49 / 57

Basic Human Capital Model II

Assume (unrealistically):
1 Physical capital, human capital and consumption goods are produced
with the same technology: One unit of …nal output can be used for
consumption, investment in physical capital and investment in human
capital.
2 All investments are fully reversible.
3 Same depreciation (rate δ) for both types of capital (unimportant).

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 50 / 57

Basic Human Capital Model III

No arbitrage implies: RK = RH = r + δ.
Firms’pro…t-maximization:

RK = αA (H/K )1 α
= (1 α) A (H/K ) α
= RH

Solving for human-to-physical ratio yields:

H/K = (1 α)/α

Hence, substituting away H/K :

r = α α (1 α )1 α
A δ,
1 α
1 α
Y = AK
α

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 51 / 57

Basic Human Capital Model IV

The equilibrium features

α α (1 α )1 α
A δ ρ
ċ (t ) = c (t ) ,
θ
1 α
1 α
k̇ (t ) = Ak (t ) c (t ) δk (t ) ,
α

plus a TVC
In equilibrium, the economy grows at the constant rate

α α (1 α )1 α
A δ ρ
γ= .
θ

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 52 / 57

Learning-by-doing Externalities I

External e¤ects of capital accumulation on productivity.

As capital in …rm i accumulates, it has a productivity-enhancing
e¤ect on the capital installed in all …rms.
It becomes important to distinguish between …rm-level and aggregate
variables.
Firm-level technology:

Yi = F Ki , ÃLi

Labor-augmenting technical progress (Ã) is not …rm-speci…c. We

assume
à = φK

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 53 / 57

Learning-by-doing Externalities II

à can be interpreted as public knowledge. Knowledge is assumed to

have a non-rival character: when a …rm adds to the stock of
knowledge, all …rms in the economy can bene…t from this addition.
Knowledge accumulation is assumed to be a pure spillover.
For simplicity, we restrict attention to Cobb-Douglas technology:
1 α
F Ki , ÃLi = Kiα ÃLi = AKiα (KLi )1 α

where A φ1 α

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 54 / 57

Learning-by-doing Externalities III

Firms takes K̄ as parametric. Thus, the equilibrium rates of return

are:
1 α
KLi
R = FK i Ki , ÃLi = αA (LBD-FOC1)
Ki
(1 α) AKiα (KLi )1 α
w = FLi Ki , ÃLi = (LBD-FOC2)
Li
Assume a continuum of identical …rms with total measure equal to
M. Thus, in a symmetric equilibrium,

MKi = K and MLi = L. (LBD-EQ)

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 55 / 57

Learning-by-doing Externalities IV

IMPORTANT: to characterize the competitive equilibrium, one must

substitute (LBD-EQ) into (LBD-FOC1)-(LBD-FOC2) (i.e., atomistic
…rms ignore the e¤ect of their investments on aggregate productivity).
I.e., …rms act in an uncoordinated fashion. So, using (LBD-EQ) to
eliminate Ki and Li , leads to:

r= R δ = αAL1 α δ
(1 α) AK
w = = (1 α) A k L1 α
Lα

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 56 / 57

Learning-by-doing Externalities V

The equilibrium conditions are

αAL1 α δ ρ
ċ (t ) = c (t )
θ
k̇ (t ) = AL1 α
δ k (t ) c (t )

plus a TVC
The dynamics of this model are isomorphic to those of the AK model.
But there are two di¤erences:
1 Scale e¤ects
2 Equilibrium is not Pareto optimal
(discussed as an exercise).

Kjetil Storesletten (University of Oslo) Lecture 2 August 27, 2013 57 / 57