Chapter 2: The Ramsey Growth Model

Yazid Dissou

Department of Economics
University of Ottawa

ECO 6120 Macroeconomic Theory IV

Winter 2020

1 / 67
Overview

I In the previous chapter, we analyzed a growth model in which

the saving rate of consumers was exogenous.

I We relax the assumption of exogenous saving rate in this

chapter by allowing consumers to determine optimally their

consumption/saving, hence their saving rate.

I Time is continuous in this model.

I A homogeneous good is produced in the economy using a

neoclassical production function and with capital and labor as

inputs.

2 / 67
Overview cont'

I There is a large number of identical rms and a large number

of identical households in the economy.

I Households supply labor to the rms and are the owners of the

capital used by the rms to produce.

I Firms sell the goods to the households for consumption and

investment.

I All markets are competitive.

I Firms maximize prots and determine the optimal levels of

capital and labor.

I Households maximize an intertemporal utility function derived

from a stream of consumption over time

3 / 67
Overview cont'

I As opposed to the Solow growth model, the saving rate of the

households is not xed:

I It is endogenous and determined by household preferences, i.e.

by the parameters of their utility function
I In contrast to the Solow growth model, we will see that there

is no possibility for dynamic ineciency in this model.

I As the saving rate is endogenously determined, the economy

cannot nd itself over-saving.

I In this model without distortion, using the First and Second

welfare theorems, we can use the decentralized and the social

planner approaches to analyze the growth behavior of the

economy.

I We will use both approaches in this chapter

4 / 67
Preferences

I There is a large number of identical households.

I The households are dynastic in the sense that generations are

connected through time by altruistic bequests.

I Each household has a size L(t), which grows at an exogenous

growth rate n
I L(t) = L(0)e nt ;
I Normalizing L(0) = 1, we have L(t) = e nt
I Let denote c(t) per capita consumption:

I c(t) = CL(t)
(t)

I C (t) is household total or aggregate consumption

5 / 67
Preferences cont'

I Households seek to maximize a time-separable intertemporal

utility function

I For the sake of simplity, households do not value leisure: their

labor supply is inelastic

I The discount rate is positive ρ>0

I In order to have a convergent integral, we assume that the

discount rate, is larger than the growth rate of the household

size: 0 <n<ρ

6 / 67
Preferences cont'

I The instantaneous utility function of the representative

household is: u (c(t)),

I u(c) is strictly increasing, concave, twice continuously
dierentiable with derivatives u' and u ,
I u 0 (c) > 0, and u 00 (c) < 0
I u (c) satises the Inada conditions :

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

c→0 c→∞

7 / 67
Preferences cont'

I At time t = 0, each household seeks to maximize a

time-separable intertemporal utility function

 ∞
U(0) = u (c(t)) .L(t).e −ρt dt (1)

t=∞0
U(0) = u (c(t)) .e nt .e −ρt dt (2)

t=∞0
U(0) = u (c(t)) .e −(ρ−n)t dt (3)
t=0

I ρ−n is the eective discount rate.

8 / 67
Preferences cont'

I U(0) is the present-value (discounted sum) of the stream of

instantaneous utilities of consumption of the household.

I In the specication of the intertemporal utility function in (3),

the assumption ρ>n , ensures that the discounted utility is

nite.

9 / 67
The Budget Constraint

I In each period, the household member supplies inelastically

one unit of labor to rms, and receives a wage rate of w (t).

I Households can borrow and lend at an interest rate given by

the market, r (t).

I Let denote the asset holdings of the representative household

at time t by A(t).

10 / 67
The Budget Constraint cont'

I The household ow (or period) budget constraint can be

stated as follows:

Ȧ(t) = r (t)A(t) + w (t)L(t) − c(t)L(t), ∀t (4)

I r (t) is the risk-free market rate of return on assets in each

period

I w (t)L(t) is the labor income earnings of the household in each

period

11 / 67
The Budget Constraint cont'

I The stock of net assets or nancial wealth per A(t) may be

negative or positive in any given period.

I A(0) is given (bequests from earlier generations).

A(t)
I Let now dene the per capita asset, a(t), by a(t) ≡ L(t)

12 / 67
The Budget Constraint cont'

I On a per capita basis the period budget constraint may be

written as follows:

Ȧ(t) A(t) L(t) L(t)

= r (t) + w (t) − c(t)
L(t) L(t) L(t) L(t)
Ȧ(t)
= r (t)a(t) + w (t) − c(t)
L(t)

Ȧ(t)
I Let's nd an expression for
L(t) in terms of a(t).

13 / 67
The Budget Constraint cont'

I We know that A(t) = a(t).L(t), it follows that:

Ȧ(t) = ȧ(t).L(t) + a(t).L̇(t)

Ȧ(t) L̇(t)
= ȧ(t) + a(t).
L(t) L(t)

L̇(t) Ȧ(t)
Noting that
L(t) = n, we have
L(t) = ȧ(t) + n.a(t)

14 / 67
The Budget Constraint cont'

I The budget constraint can then be rewritten as:

ȧ(t) = (r (t) − n) a(t) + w (t) − c(t) (5)

I As the household nancial assets can be negative, we rule out

a Ponzi game and impose a transversality condition by

requiring that in the long run, the limit of the present value of

the nancial assets must be non-negative:

t
  
 
lim a(t) exp − (r (s) − n) ds  ≥ 0 (6)
t→∞  
s=0

I Let's now nd the solution to the household problem

15 / 67
Solution to the household optimization problem

I Let's recapitulate the household optimization problem

 ∞
max∞ U(0) = u (c(t)) e −(ρ−n)t dt
{c(t)}t=0 t=0
s.t. ȧ(t) = (r (t) − n) a(t) + w (t) − c(t)
t
  
 
lim a(t) exp − (r (s) − n) ds  ≥ 0
t→∞  
s=0

I As time is continuous, we can solve the household problem

using the Hamiltonian approach by setting up the

present-value Hamiltonian.
1

I This is an optimal control problem in the continuous time

setting

1
Note that instead of using the present-value Hamiltonian, we could also
use the current-value Hamiltonian
16 / 67
Solution household problem cont'

I In general, the formulation of an optimal control problem can

be presented as follows:
 t=∞
max f (t, x(t), u(t)) dt
u(t) t=0

subject to the constraints

ẋ(t) = g (x(t), u(t)) , ∀t

x(0) = x0
 
lim x(t)b(t) ≥ x1
t→∞

where x0 and x1 are exogenous and f , g and b are given real-valued

functions. In particular, b is such that limt→∞ b(t) < ∞.

17 / 67
Solution household problem cont'

I In an optimal control problem, we have to identify the state

variable and the control variable

I In this general case, x(t) is the state variable and u(t) is the

control variable

I Using the optimal control approach, one will have to dene the

present-value Hamiltonian function, H, whis is specied as

follows:

H(0) = f (t, x(t), u(t)) + λ(t)g ((x(t), u(t)))

I where λ(t) are time Lagrangian multipliers related to the

constraints; they are called co-state variables

18 / 67
Solution household problem cont'

I The equations that describe the solution must satisfy the

Maximum principle of Pontryagin et al. (1962), which requires

the following FOCs

∂H ∂f ∂g
=0 or +λ = 0, (7)
∂u ∂u ∂u
∂H ∂f ∂g
= −λ̇ or +λ = −λ̇ (8)
∂x ∂x ∂x
∂H
= ẋ or ẋ = g (9)
∂λ

19 / 67
Solution household problem cont'

I Equations (7) to (9) implicitly determine the variables x(t),

u(t), and λ(t) as functions of the model's parameters.

I In our case, the present-value Hamiltonian is

H = u (c(t)) e −(ρ−n)t + µ(t). [(r (t) − n) a(t) + w (t) − c(t)]

(10)

I The variable µ(t) are the continuous time Lagrangian

multipliers

I They provide the present-value shadow prices of wealth, i.e.,

the value of an addition to assets at time t, measured in units

of time-zero utility.

20 / 67
Solution household problem cont'

I In our Hamiltonian problem

I c(t) is the control, choice, or decision variable

I a(t) is the state variable
I µ(t) is the co-state variable
I To nd the optimal solution of this problem we will use the

Maximum principle of Pontryagin et al. (1962), which requires

that following two FOCs and a transversality condition

condition be met.

21 / 67
Solution household problem cont'

I The rst FOC is:

∂H
= 0 , all t (11)
∂c(t)
u 0 (c (t)) .e −(ρ−n)t = µ (t) (12)

I It suggests that the optimal choice of consumption requires

that at time t, the discounted value of the marginal utility of

consumption be equal to the present value shadow price of a

marginal increase in the state variable (wealth), µ(t).

I The co-state variable is the shadow value of a marginal

increase in the state variable

I In other words, at the optimum, in every period, the

discounted marginal utility of consumption must be equal to

the marginal utility of the change in wealth induced by the

change in consumption.

22 / 67
Solution household problem cont'

I The second FOC is:

∂H µ̇(t)
+ µ̇(t) = 0 =⇒ = − (r (t) − n) (13)
∂a(t) µ(t)
I It is known as the Ramsey rule of optimal saving.

I It suggests that the change in the value of an increment to

time-t assets to time-zero marginal utility (the "capital gain"
in marginal utils) must be oset exactly by the time-zero
marginal utility value of the time-t return of an increment to
assets (the "income" in time- zero marginal utils).

23 / 67
Solution household problem cont'

I The transversality condition is

lim µ(t)a(t) = 0 (14)

t→∞

I The discounted value of the nancial assets must be equal to

zero in the long run.

I The transversality condition can be expressed dierently when

we replace the shadow price of wealth, µ(t), by its expression

derived from the second FOC (13).

24 / 67
Solution household problem cont'

I From (13), we have

I Integrating over time (13) will yield

t
 

µ(t) = µ(0) exp − (r (s) − n) ds  (15)

s=0

I Substituting µ(t) by the latter expression (15) into the

transversality condition, we get:

t
  

lim a(t) exp − (r (s) − n) ds  = 0 (16)

t→∞
s=0

25 / 67
Solution household problem cont'

I If we dene by r¯(t) the average interest between 0 and t

t
1
r¯(t) = r (s) ds (17)
t
s=0

I The transversality condition can be rewritten as

lim {a(t) exp − [(¯

r (t) − n) t]} = 0
t→∞
a(t)
lim = 0 (18)
t→∞ exp [(¯
r (t) − n) t]

26 / 67
Optimal consumption

I We'll use the FOCs (11 and 13 ) to characterize optimal

consumption

I Let's dierentiate (11) with respect to time

d u 0 (c(t)) .e −(ρ−n)t
 
dµ(t)
µ̇(t) ≡ =
dt dt
d e −(ρ−n)t
 
0
−(ρ−n)t [u (c(t))]
d 0
µ̇(t) = e + u (c(t))
dt dt
d e −(ρ−n)t
 
0
−(ρ−n)t d [u (c(t))] d (c(t)) 0
µ̇(t) = e . + u (c(t))
d (c(t)) dt dt
µ̇(t) = e −(ρ−n)t .u” (c(t)) .ċ(t) − (ρ − n) .u 0 (c(t)) .e −(ρ−n)t
µ̇(t) = e −(ρ−n)t . u 00 (c(t)) .ċ(t) − (ρ − n) .u 0 (c(t))
 

27 / 67
Optimal consumption cont'

I Dividing µ̇(t) by µ(t) we have while using (11):

µ̇(t) e −(ρ−n)t . [u 00 (c(t)) .ċ(t) − (ρ − n) .u 0 (c(t))]

=
µ(t) µ(t)
µ̇(t) e −(ρ−n)t . [u 00 (c(t)) .ċ(t) − (ρ − n) .u 0 (c(t))]
=
µ(t) u 0 (c(t)) .e −(ρ−n)t
µ̇(t) [u 00 (c(t)) .ċ(t) − (ρ − n) .u 0 (c(t))]
=
µ(t) u 0 (c(t))
µ̇(t) u 00 (c(t)) .ċ(t)
= − (ρ − n)
µ(t) u 0 (c(t))

28 / 67
Optimal consumption cont'

µ̇(t)
I Replacing
µ(t) by its expression in the second FOC (13) we

have:

u 00 (c(t)) .ċ(t)
− (r (t) − n) = − (ρ − n)
u 0 (c(t))
u 00 (c(t)) .ċ(t)
= ρ − r (t) (19)
u 0 (c(t))

I Multiplying and dividing the LHS of (19) by c(t), we have:

c(t).u 00 (c(t)) ċ(t)

. = ρ − r (t) (20)
u 0 (c(t)) c(t)

29 / 67
Optimal consumption cont'

I We can then nd an expression for the rate of change of

ċ(t) dc(t)/dt
consumption
c(t) = c(t)

ċ(t) u 0 (c(t))
= [ρ − r (t)]
c(t) c(t).u 00 (c(t))
ċ(t) 1
= [r (t) − ρ] (21)
c(t) εu (c(t))

where

c(t).u 00 (c(t))
εu (c(t)) ≡ − (22)
u 0 (c(t))
is the elasticity of the marginal utility of consumption, u 0 (c(t)), i.e.
the percentage change of the marginal utility following 1% change

in consumption.

30 / 67
Optimal consumption cont'

I Equation (21), is thus giving us the growth rate of each

household member's consumption:

I It's the continuous version of the consumption Euler equation

found in the discrete case.

I Consumption per household member grows when the rate of

return on assets is larger than the discount rate ρ,

I In other words, consumption per member rises when the return

on nancial assets exceed the rate at which the household

discounts future consumption.

31 / 67
Optimal consumption cont'

I In that case, the agent is relatively patient and he nds optimal

to reduce consumption in the short run so as to increase later

I The higher the rate of return on assets r (t), the higher the

growth rate of consumption.

I Besides, the larger the elasticity of the marginal utility of

consumption, the lower the growth rate of consumption.

I When the interest rate r (t) is equal to the the rate of time
preference, ρ, the growth rate of consumption is zero.

I Households will then choose a at consumption time path.

32 / 67
Optimal consumption cont'

I In reality, the elasticity of the marginal utility of consumption,

εu (c(t)), is the inverse of the intertemporal elasticity of

substitution, which is a local measure of the curvature of the

indierence curve.

I The intertemporal elasticity of substitution provides

information on the willingness of the household to substitute

consumption over time.

I The substitution elasticity between time t and s > t , σu (t, s)

can be dened as the percentage change of the ratio

c (s) /c (t), following 1 % change in the marginal rate of

substitution between t and s , MRSt,s .

33 / 67
Optimal consumption cont'

I Note that the marginal rate of substitution MRSt,s stands for

the relative price of consumption between the two periods,

pt /ps .
 
c(s)
dln c(t)
σu (t, s) =
dln (MRSt,s )
 
dln c(s)
c(t)
σu (t, s) =  0 
u (c(t))
dln u0 (c(s))
 
dln c(s)
c(t)
σu (t, s) = −  0 
u (c(s))
dln u0 (c(t))

34 / 67
Optimal consumption cont'

 
c(s)
d c(t)
c(s)
c(t)
σu (t, s) σu (t, s) = −  0
u (c(s))

d u 0 (c(t))
u 0 (c(s))
u 0 (c(t))
 
c(s) u 0 (c(s))
d c(t) u 0 (c(t))
σu (t, s) = −  
c(s) u 0 (c(s))
c(t) d u 0 (c(t))
 
c(s) u 0 (c(s))
d c(t) u 0 (c(t))
σu (t, s) = −   (23)
u 0 (c(s)) c(s)
d u 0 (c(t)) c(t)

35 / 67
Optimal consumption cont'

I We get the instantaneous intertemporal elasticity of

substitution, when we take the limit of σu (t, s) when s

converges to t

 
u 0 (c(s))
d c(s)
c(t) u 0 (c(t))
σu (t) ≡ lim −  0 
c(s)
s→t
d uu0 (c(s))
(c(t)) c(t)

u 0 (c (t)) 1
σu (t) ≡ − = (24)
u” (c (t)) .c (t) εu (c(t))

36 / 67
Optimal consumption cont'

I Assume that the interest rate r (t) is held constant at a given

value dierent from the rate of time preferences ( r (t) 6= ρ),

referring to the Euler equation (21), the growth rate of

consumption will be constant if εu (c(t)), is constant, or if the

intertemporal elasticity of substitution,σu (t), is constant.

ċ(t)
= σu (c(t)) [r (t) − ρ]
c(t)

I Hence, the popularity of functional forms of utility functions

that have a constant intertemporal elasticity of substitution.

I These utility functions are known in the literature as Constant

Intertemporal Elasticity of Substitution (CIES) function or

Constant Relative Risk Aversion (CRRA) utility function in the

presence of uncertainty.

37 / 67
Example with a CRRA utility function

1
(c (t))1− σ
u (c (t)) = 1 , σ > 0, σ 6= 1 (25)
1 −
σ
u (c (t)) = ln (c (t)) , σ=1 (26)

I The rst and second derivatives of the utility functions are:

1
u 0 (c (t)) = (c (t))− σ
1 1
u” (c (t)) = − (c (t))− σ −1
σ

38 / 67
Example CRRA utility function cont'

I We can easily nd that the elasticity of marginal utility,

εu (c(t)) is constant, and is equal to 1 /σ

 
c (t) 1 − σ1 −1
εu (c(t)) = − 1 − (c (t))
(c (t))− σ σ
1 1 1
c (t)1+ σ (c (t))−( σ +1)
h i h i
εu (c(t)) =
σ
1
εu (c(t)) = (27)
σ

I It follows that the intertemporal elasticity of substitution is

constant and independent of the level of consumption.

σu (c (t)) = 1/εu (c(t)) = σ

39 / 67
Example CRRA utility function cont'

I Consequently, the household's optimal consumption growth

rate is

ċ(t)
= σ [r (t) − ρ] (28)
c(t)

I The optimal growth rate of consumption is determined by the

dierence between the interest rate and the rate of the time

preference, weighted by the intertemporal elasticity of

substitution.

I The higher the elasticity of substitution, the larger the optimal

response of consumption to the gap between the interest rate

and the rate of time preference.

40 / 67
Example CRRA utility function cont'

I It is important to recall that the consumption Euler equation

only gives the growth rate of consumption. It does not give

the optimal level of consumption.

I In the case where the rate of interest rate is constant, and

equal to r, it is possible to integrate the Euler consumption

equation to get the the optimal level of consumption c(t)

c (t) = c (0) .e σ(r −ρ)t (29)

41 / 67
Firms

I Production possibilities set of the economy is represented by

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

I Firms use a technology that is characterized by constant

returns to scale and we assume that Inada assumptions hold.

FK , F L > 0 ; FKK , FLL < 0

lim FK = ∞; lim FL = ∞; lim FK = 0; limFL = 0
k→0 L→0 k→∞ L→∞

I We abstract from technological progress

42 / 67
Firms cont'
I As in the previous chapter, per capita production function can

be expressed as

Y (t)
y (t) =
L(t)
y (t) = f (k (t))
where, k(t) = K (t)/L(t)
I In a competitive setting, the optimal demands for labour and

capital imply
∂F [K (t), L(t)]
w (t) = = f (k(t)) − k(t)f 0 (k(t)) (30)
∂L(t)
and

∂F [K (t), L(t)]
R (t) = = f 0 (k (t)) (31)
∂K (t)
I where R(t) and w (t) are respectively, the rental rate of capital

and the wage rate

43 / 67
Firms cont'

I With constant-returns-to-scale technology, pure prots are

equal to zero. This implies that

F [K (t), L(t)] = w (t)L(t) + R(t)K (t)

L(t)y (t) = w (t)L(t) + R(t)L(t)k(t)
y (t) = w (t) + R(t)k(t)
f (k(t)) = w (t) + R(t)k(t) (32)

44 / 67
Equilibrium and dynamics

I We will now analyze the competitive equilibrium of this

economy by combining the behaviors of the households and

the rms.

I In this closed economy in equilibrium assets per capita must be

equal to the stock of physical capital per capita

a(t) = k(t) (33)

I In the presence of depreciation of physical capital, the rental

rate of capital must be able to cover the depreciation rate and

the interest rate paid to households who are the owners of the

capital stock

R(t) = r (t) + δ (34)

45 / 67
Equilibrium and dynamics cont'

I Replacing R(t) by its expression in the optimal demand for

capital in (31), we have:

r (t) = f 0 (k(t)) − δ (35)

I A competitive equilibrium of this economy consists of paths

{c(t), k(t), w (t), R(t)}∞

t=o , such that the representative

household maximizes their utility function (3) subject to their

budget constraint (5) and the transversality condition (6)

given initial capital-labor ratio k(0), the factor prices

{w (t), R(t}∞
t=0 as in (30) and (31), and the rate of return on

assets r (t) given as in (34).

46 / 67
Equilibrium and dynamics cont'

I Replacing the return to assets in the Euler equation, we have:

ċ(t)
= σ f 0 (k(t)) − δ − ρ
 
c(t)

I Using the zero-prot condition (34), assets equilibrium

condition, (33) and the (35), the household budget constraint

(5) can be expressed as:

k̇(t) = (r (t) − n) k(t) + w (t) − c(t)

k̇(t) = r (t)k(t) + w (t) − nk(t) − c(t)
k̇(t) = f (k(t)) − (δ + n) k(t) − c(t) (36)

47 / 67
Equilibrium and dynamics cont'

I The competitive equilibrium conditions can then be

summarized as follows:

ċ(t)
σ f 0 (k(t)) − δ − ρ
 
= (37)
c(t)
k̇(t) = f (k(t)) − (δ + n) k(t) − c(t) (38)
0
w (t) = f (k(t)) − k(t)f (k(t)) (39)
0
r (t) = f (k(t)) − δ (40)

t
  

f 0 (k(s)) − δ − n ds  (41)


 lim k(t) exp − =0
t→∞
s=0
k(0) > 0 given (42)

48 / 67
Equilibrium and dynamics cont'

I The dynamics of the economy is governed by the two ordinary

dierential equations (37) and (38), which provide the

dynamics of, respectively, consumption per capita and the

stock physical capital per capita.

49 / 67
Steady state

I These two dynamic equations will imply the following phase

diagram in the (k, c) space

I In the steady-state equilibrium physical capital per capita, and

consumption per capita are constant, i.e.,

ċ(t) = 0

k̇(t) = 0

I Plotting the loci of k(t) and c(t) such that ċ(t) = 0 and

k̇(t) = 0, we have the following graph: see Graph xx:

50 / 67
Steady state cont'

I The k̇ = 0 locus is given by (k, c) such that

k̇ = f (k) − δk − c = 0 ⇒ c ∗ = f (k ∗ ) − δk ∗
I The ċ = 0 locus is given by (k, c) such that

ċ = cθ[f 0(k) − δ − ρ] = 0 ⇒ c = 0 or f ´(k) = δ + ρ

f ´(k) = δ + ρ ⇒

− 1
k∗ = f 0 (δ + ρ)

51 / 67
Steady state cont'

I In the (k, c) space, the ċ = 0 locus is a vertical line,

I The k̇ = 0 locus is an inverse U-shaped curve

I From that curve, we can see that consumption per capita is

maximized at the golden-rule capital-labor ratio kgold
I The steady state will be found at the intersection of the two

loci:

k̇ = ċ = 0 such that {(c, k) = (0, 0) or (c, k) = (c ∗ , k ∗ )}

I We will ignore the obvious steady state (c, k) = (0, 0)

52 / 67
Steady state cont'

I The steady-state capital-labor ratio is determined by the

following relation and it is independent of the functional form

of the instantaneous utility function:

−1
k∗ = f 0 (δ + ρ) (43)

I The steady-state consumption level is found using

c ∗ = f (k ∗ ) − δk ∗ (44)

I This steady state corresponds to the modied golden rule,

rather than the golden rule in the Solow growth model.

53 / 67
Steady state cont'

I The modied golden rule is not the one that maximizes

steady-state consumption;

I It involves a capital stock that is lower than the level

corresponding to the golden rule in the Solow model. It

depends on the rate of time preferences, ρ

I The reason for this is that the optimizing households are

impatient and they discount future; they prefer earlier

consumption to later consumption.

I Their objective is not to maximize the steady-state level of

consumption; they give higher weight to early consumption.

I As a consequence, they do not want to defer enough current

consumption to obtain a higher level of consumption, cgold ,

which could be attained if consumption was not discounted.

54 / 67
Phase diagram

I Let us now study the dynamics toward the steady state

I Each of the two loci divides the (k, c) space in two parts

I For the the ċ = 0 locus, when

I k < k ∗ ⇒ f 0(k ∗ ) − δ − ρ > 0, hence ċ > 0: consumption

increases.
I k > k ∗ ⇒ f 0(k ∗ ) − δ − ρ < 0, hence ċ < 0: consumption
decreases.
I The direction of change of consumption is represented by the

vertical arrows on the graph.

55 / 67
Phase diagram cont'

I For the the k̇ = 0 locus, when:

I c < f (k) − δk⇒ , k̇ > 0, hence capital stock increases.

I c > f (k) − δk⇒ , k̇ < 0, hence capital stock decreases.
I The direction of change of capital is represented by the

horizontal arrows on the graph.

56 / 67
Phase diagram cont'

I The two loci together divide the (k, c) space into four regions.

I From Graph xx, we can draw the time paths of {k(t), c(t)}
that lead to the steady state from any starting point

(k(0), c(0)).
I We only have two paths that go through the steady state.

I The positively sloped one corresponds to the stable manifold or

saddle path.
I The sloped one represents the unstable manifold.
I Recall that the initial stock of capital, k(0) is given. The
consumption level, which is the control variable, is free.

I It has to adjust to a single value so as to satisfy the

transversality condition

57 / 67
Phase diagram cont'

I On the graph, we can say that, for any given k(0) > 0, the
corresponding equilibrium c(0) is to the one that puts the
economy on the saddle manifold;

I It is unique.
I The reason for this is that for a given k(0), if the economy

starts at consumption point, which is not of the stable

manifold, and follows the consumption path given by the Euler

equation, the system will end up violating the either the

feasibility condition, or the transversality condition

58 / 67
Phase diagram cont'

I Indeed:
I If the economy starts above the stable path at an initial
consumption level above c(0), the capital stock would end up
at zero in nite time, while consumption will still be positive.
I This is a contradiction with the feasibility condition, which
suggests that consumption cannot be positive with zero
capital (output).
I We must rule a consumption path above the stable path.
I If the economy starts with a consumption level below c(0), the
economy will be accumulating capital until it reaches the
maximum level, k,where consumption would be zero.

59 / 67
Phase diagram cont'

I It can easily be veried that at the maximum capital level,

k,f 0 k < δ + n.


It follows that the transversality condition

will be violated at , k

t
  

f 0 k − δ − n ds  > 0



 lim k exp −
t→∞
s=0

I Note that, once the economy has reached the golden rule of
capital-labor ratio, it will be wasting resources, if it continues
investing.

I It would be better to increase consumption

60 / 67
Optimal growth

I As mentioned earlier, in the current setting of the model with

no distortions, the decentralized solution will coincide with the

social planner's solution.

I Let consider the same closed economy in which are the

households and rms

I The social planner has a social welfare function, which is

identical to the household utility as dened in above.

I The social planner's objective is to choose the optimal level of

consumption that maximizes households' welfare with the

knowledge that more consumption today is equivalent to less

investment today and lower output in the next period and

hence less consumption in the future.

61 / 67
Optimal growth

I The social planner maximizes utility subject to the resource

and technological constraints

 ∞
Max∞ W (0) = u (c(t)) e −(ρ−n)t dt
{c(t)}t=0 t=0
k̇(t) = f (k(t)) − (δ + n) k(t) − c(t)
c(t) ≥ 0, k(0) > 0, k(t) ≥ 0 for all t≥0

I In this problem, k(t) is the state variable, c(t) is the choice

variable, and µ(t) is the co-state variable

62 / 67
Optimal growth cont'

I Setting up the present-value Hamiltonian of this problem, we

have

H(0) = u (c(t)) e −(ρ−n)t + µ(t). [f (k(t)) − (δ + n) k(t) − c(t)]

I From the maximum principle, the following FOCs:

∂H
= 0 ⇒ u 0 (c(t)) .e −(ρ−n)t = µ(t)
∂c(t)
∂H
+ µ̇(t) ⇒ µ̇(t) = −µ(t) f 0 (k(t)) − (δ + n)


∂k(t)
lim [µ(t)k (t)] = 0
t→∞

I Using the same method as in the previous case, we get

63 / 67
Optimal growth cont'

I The Euler equation:

ċ(t) 1  0 
= f (k (t)) − δ − ρ (45)
c(t) εu (c(t))

which suggests that consumption growth rate is proportional to the

dierence between the marginal product of capital, net of

depreciation, and the subjective discount rate.

I Equation (45) is similar to Equation (37) found in the

competitive equilibrium case

64 / 67
Optimal growth

I The transversality condition:

t
  

f 0 (k(s)) − δ − n ds  = 0


 lim k(t) exp −
t→∞
s=0

which is also identical to (41)

65 / 67
Optimal growth cont'

I It follows that with the standard assumptions on technology

and preferences, the competitive equilibrium is Pareto optimal,

and it coincides with the optimal growth path that maximizes

household utility.

I The saving rate and the associated steady-state capital stock

when consumption-saving choices are made by optimizing

households with a positive rate of time preference will be less

than the "golden-rule" saving rate and capital stock.

I Therefore we will never observe dynamically inecient

over-saving in steady-state.

66 / 67
References

I Barro, Robert J. and Xavier Sala-i-Martin, EconomicGrowth,

McGraw-Hill,1995.

I Lucas, Robert E.,On the Mechanics of Economic

Development,Journal of Monetary Economics,

1988,22(1),342.

I Rebelo, Sergio, Long-Run Policy Analysis and Long-Run

Growth,Journal of Political Economy, June1991,99, 500521.

I Romer, Paul M., Increasing Returns and

Long-RunGrowth,Journal of Political Economy, October1986,

94, 10021037.

I Romer, Paul M., Endogenous Technological Change, Journal

of Political Economy, October1990,98(5),S71S102.

I Romer, Paul M., Capital Accumulation in the Theory of Long

Run Growth, in Robert J. Barro, ed., Modern Business Cycle

Theory, Cambridge, MA: Harvard University Press, 1989.

67 / 67