Master Economics and Public Policy

ECO 553. Economic Growth1

Lecture 4
The neoclassical growth model

Pierre Cahuc

Winter 2012-2013

1 http://sites.google.com/site/eco553x/
1 / 65
Introduction

I Solow model: saving behavior is exogenous

I Neoclassical model: saving is determined by household
optimization
I Better description of the economy
I Neoclassical model useful for the analysis of
I the optimality of growth
I …scal, taxation policies
I business cycles (2nd quarter)

2 / 65
Introduction

I The Neoclassical model of growth is one of the most

important model in macroeconomics
I It has been developed by
I Franck Ramsey (1928) (Ramsey’s model)
I John von Neumann (1945)
I David Cass (1965), Tjalling Koopmans (1965)
I Presentations in macroeconomic textbooks
I Blanchard and Fischer (1989), Barro and Sala-i-Martin (2004),
Acemoglu (2009)

3 / 65
Introduction

I Outline
1. Saving behavior
2. The competitive equilibrium
I Then, next lecture:
1. The optimality of growth
2. Public debt
3. Distortionary taxation

4 / 65
1. Saving behavior

I We proceed in 2 steps

1. Simple 2 period model

2. In…nite horizon model

5 / 65
1. Saving behavior
1.1. 2 period model

I period t = 0, 1
I ct consumption in period t
I Preferences
U (c0 , c1 ) = u (c0 ) + βu (c1 )
I u ( ) is the instantaneous utility function or the ‘felicity
function’, u 0 (c ) > 0, u 00 (c ) 0
I β 2 (0, 1) is the discount factor, or

1
β=
1+ρ

ρ > 0, discount rate

6 / 65
1. Saving behavior
1.1. 2 period model

I Income in period t, denoted by wt

I Perfect …nancial market, interest rate r
I Instantaneous budget constraints

c0 + s = w0
c1 = ( 1 + r ) s + w1

I s : saving, positive or negative

7 / 65
1. Saving behavior
1.1. 2 period model

I Elimination of s in the instantaneous budget constraints yields

the intertemporal budget constraint
c1 w1
c0 + = w0 + =W
1+r 1+r
I The discounted sum of consumptions is equal to the
discounted sum of incomes
I By de…nition, the discounted sum of incomes is the wealth,
denoted by W

8 / 65
1. Saving behavior
1.1. 2 period model

I The optimal choice of the household is

max u (c0 ) + βu (c1 )

(c0 ,c1 )

subject to
c1 w1
c0 + = w0 + =W
1+r 1+r
I Or:
max u (c0 ) + βu ((1 + r )(W c0 ))
c0

I First order condition:

u 0 ( c0 ) = β ( 1 + r ) u 0 ( c1 )

9 / 65
1. Saving behavior
1.1. 2 period model

I Or, with β = 1/(1 + ρ)

1+r 0
u 0 ( c0 ) = u ( c1 )
1+ρ

The marginal utility of consumption in period 0 is equal to the

present discounted value of the marginal utility of
consumption in period 1 times the gross interest rate
I This is the Euler equation, or the Keynes-Ramsey rule
I When r increases
I more costly to consume today with respect to tomorrow
I substitute future consumption to present consumption

10 / 65
1. Saving behavior
1.1. 2 period model

I The optimal consumption path (c0 , c1 ) is de…ned by the

Keynes-Ramsey rule and the budget constraint
1+r 0
u 0 ( c0 ) = u ( c1 )
1+ρ
c1
c0 + = W
1+r
I These 2 equations de…ne functions

ct (ρ, r , W ), t = 0, 1.

I Consumptions depend on wealth, and not on the sequence of

incomes, because the …nancial market is perfect.

11 / 65
1. Saving behavior
1.1. 2 period model

I Example: Constant relative risk aversion utility function

(CRRA)
I Arrow-Pratt coe¢ cient of relative risk aversion
u 00 (c )c
σ (c ) = 0
u 0 (c )
I Family of CRRA utility functions:
( 1 σ
c
1 σ if σ 6= 1, σ 0
u (c ) =
log(c ) if σ = 1

I Risk neutrality: σ = 0

12 / 65
1. Saving behavior
1.1. 2 period model
I With a CRRA utility function, the Keynes-Ramsey rule yields
1
c0 1+r σ
= , when σ > 0
c1 1+ρ
I An increase in r :
I more costly to borrow, higher returns of lending
I future consumption increases relative to present consumption
I This e¤ect is larger if the coe¢ cient of relative risk aversion,
σ, is smaller
I 1/σ is the intertemporal elasticity of substitution:
d(c0 /c1 )/(c0 /c1 ) d log(c0 /c1 ) 1
= =
d(1 + r ) / (1 + r ) d log(1 + r ) σ
I Interpretation: one percent increase in the gross returns of
savings induces a 1/σ percent decrease in the ratio c0 /c1
13 / 65
1. Saving behavior
1.1. 2 period model

I When σ = 0 (risk neutrality) the Keynes-Ramsey rule reads

8
< c0 = W , c1 = 0 if r < ρ
c0 = 0, c1 = (1 + r )W if r > ρ
:
c0 2 (0, W ) and c1 = (1 + r )(W c0 ) if r = ρ

I The intertemporal elasticity of substitution is in…nite

14 / 65
1. Saving behavior
1.1. 2 period model

I The 2 period model shows that the optimal consumption path

is detemined by
I the Keynes-Ramsey rule
I the intertemporal budget constraint
I These results are also satis…ed when
I the time horizon has more than 2 periods, even an in…nite
number of periods
I time is continuous

15 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I Justi…cation of in…nite horizon

I uncertainty about the date of death
I intergenerational altruism
I Time t 0 is continuous
I Let us consider any horizon, T , that can be either …nite or
in…nite
I The consumer maximizes
Z T
ρt
u (c (t ))e dt
0

16 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I Let us look at the budget constraints

I In the 2 period model
I the consumer has no debt at the terminal date
I the value of his assets is equal to zero at the terminal date
I Suppose that the value of the assets of the consumer in the 2
period model at the end of period 1 is a1 :

a 1 = w0 ( 1 + r ) + w1 c0 ( 1 + r ) c1 (1)

I We assumed that a1 = 0, which implies that

c1 w1
c0 + = w0 + =W
1+r 1+r

17 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I Same logic should apply when the horizon is in…nite: the

present value of the assets should go to zero2
Z t
lim a(t ) exp r (x )dx =0 (2)
t !∞ 0

I This is the no-Ponzi (no-Mado¤) game condition

I Remark: imposing simply limt !∞ a(t ) = 0 cannot be
consistent with steady state solution with strictly positive
assets

2 See Appendix A.
18 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I The instantaneous budget constraint at time t can be

represented by the law of motion of the value of the assets of
the consumer
I Let us consider a small interval of time dt ! 0
I The law of motion of a(t ) reads

a(t + dt ) = a(t ) + r (t )a(t )dt + w (t )dt c (t )dt

I Notice that a(t +dt ) a(t ) corresponds to instantaneous

saving at time t

19 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I Let us re-write the law of motion of a(t ) as

a(t + dt ) a (t )
= r (t )a (t ) + w (t ) c (t )
dt
I Since, by de…nition the derivative with respect to time of a(t )
is
a(t + dt ) a(t )
ȧ(t ) = lim
dt !0 dt
we get
ȧ(t ) = r (t )a(t ) + w (t ) c (t ) (3)
I ȧ(t ) is the instantaneous saving at time t

20 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I The consumer chooses the consumption path

fc (t ) 0, t 0g which maximizes the discounted utility
subject to the budget constraint; i.e.
Z ∞
ρt
max u (c (t ))e dt
fc (t ) 0,t 0 g 0

subject to
ȧ(t ) = r (t )a(t ) + w (t ) c (t )
Z t
lim a(t ) exp r (x )dx =0
t !∞ 0

21 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I We can characterize the solution using the maximum principle

I The maximum principle is applied by proceeding in 2 steps
1. Write the Hamiltonian function
2. Use the necessary conditions for a path to be optimal

22 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

1. The Hamiltonian function reads

ρt
H (t ) = u (c (t ))e + µ (t ) [r (t )a (t ) + w (t ) c (t )]

The variable µ(t ) is called the costate variable associated with

the state variable a(t )
2. The necessary conditions for a path to be optimal are
∂H (t )
= 0,
∂c (t )
∂H (t )
= µ̇(t )
∂a(t )
lim a(t )µ(t ) = 0 transversality condition
t !∞

I Transversality condition is equivalent to the no-Ponzi game

condition (Appendix B)
23 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I Let us apply the necessary conditions

∂H (t )
= u 0 (c (t ))e ρt µ(t ) = 0 (4)
∂c (t )
∂H (t )
= r (t )µ(t ) = µ̇(t ) (5)
∂a(t )
lim a(t )µ(t ) = 0 (6)
t !∞

I Eq. (4) shows that the costate variable µ(t ) is equal to the
present value of the marginal utility of consumption c (t )

24 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I The transversality condition, which reads, using (4) and (6),

lim a(t )u 0 (c (t ))e ρt

=0
t !∞

means that the present value of the assets when t ! 0 goes

to zero
I Otherwise, the solution would not be optimal: it is not
optimal to keep assets beyond the terminal date if the
marginal value of consumption is positive
I The necessary conditions are su¢ cient when the utility
function is concave

25 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time
I The derivative with respect to time of condition (4),
∂H (t )
= u 0 (c (t ))e ρt
µ (t ) = 0
∂c (t )
yields
µ̇(t ) = e ρt
u 00 (c (t ))ċ (t ) ρu 0 (c (t ))
I Substituting into condition (5) gives the Euler equation, or
the Keynes-Ramsey rule
ċ (t ) u 0 (c (t ))
= [r (t ) ρ]
c (t ) u 00 (c (t ))c (t )
I Or, using the de…nition of the Arrow-Pratt coe¢ cient of
relative risk aversion,
u 00 (c (t ))c (t )
σ(c (t )) =
u 0 (c (t ))
26 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I We get
ċ (t ) 1
= [r (t ) ρ]
c (t ) σ(c (t ))
where 1/σ(c (t )) is the intertemporal elasticity of substitution
I Interpretation: increase in future consumption relative to
present consumption when the interest rate is increased
I The increase is larger when the intertemporal elasticity of
substitution is larger
I The Keynes Ramsey rules in discrete time and in continuous
time are in fact identical (Appendix C)

27 / 65
1. Saving behavior
1.2. In…nite horizon model in continuous time

I Finally, the optimal consumption path chosen by the

consumer is de…ned by
1. the Keynes-Ramsey rule

ċ (t ) 1
= [r (t ) ρ]
c (t ) σ(c (t ))

2. the law of motion of a(t )

ȧ(t ) = r (t )a(t ) + w (t ) c (t )

3. and the no-Ponzi game condition

ρt
lim a(t )u 0 (c (t ))e =0
t !∞

I Note that we assumed a(0) = 0 until now

28 / 65
2. The competitive equilibrium

I Let us analyze the competitive equilibrium of the economy

represented by the Solow model, but where savings are
determined by optimal behavior rather than by a constant
saving rate
I We proceed as follows
1. De…nition of equilibrium
2. Behavior of households
3. Behavior of …rms
4. Characterization of the competitive equilibrium
5. Steady state equilibrium
6. Transitional dynamics

29 / 65
2. The competitive equilibrium
2.1. De…nition of equilibrium
I Model with labor and an output, produced, consumed,
invested
I Output is the numeraire
I There are many in…nite live identical households, continuum,
size = 1
I Their welfare function is given by
Z ∞
ρt
u (c (t ))e dt
0

I Each household decides at any date t how much labor and

capital to rent to …rms and how much to save and consume
I Each household o¤ers one unit of labor per unit of time, the
wage rate is denoted by w (t )
I Each household can borrow and lend to …rms or other
households at the interest rate r (t )
30 / 65
2. The competitive equilibrium
2.1. De…nition of equilibrium

I There are many identical …rms, with the same constant return
to scale technology

F (K (t ), L(t )),

where K (t ) and L(t ) denote capital and labor inputs at date

t, respectively.
I Let us denote

f (k ) = F (k, 1), k = K /L

I It is assumed that f is strictly concave and satis…es the Inada

conditions:

f (0) = 0, f 0 (0) = ∞, f 0 (∞) = 0

31 / 65
2. The competitive equilibrium
2.1. De…nition of equilibrium

I The initial stock of capital is given

k (0) = k0 > 0

I A competitive equilibrium consists of paths of consumptions,

employment, capital stocks, wage rates and rental price of
capital fc (t ), L(t ), K (t ), w (t ), r (t ), t 0g such that
I each household maximizes its welfare function given initial
assets holding and taking the time path of prices
fw (t ), r (t ), t 0g as given
I each …rm maximizes pro…ts taking the time path of prices
fw (t ), r (t ), t 0g as given
I fw (t ), r (t ), t 0g clear all markets

32 / 65
2. The competitive equilibrium
2.2. Behavior of households

I Each household chooses the consumption path

fc (t ) 0, t 0g which maximizes the discounted utility
subject to the budget constraint:
Z ∞
ρt
max u (c (t ))e dt
fc (t ) 0,t 0 g 0

subject to
ȧ(t ) = r (t )a(t ) + w (t ) c (t ),
Z t
lim a(t ) exp r (x )dx =0
t !∞ 0

with a(0) = a0 (assumed equal to zero when we studied

saving behavior)

33 / 65
2. The competitive equilibrium
2.2. Behavior of households

I The solution is de…ned by

1. the Keynes-Ramsey rule

ċ (t ) 1
= [r (t ) ρ] (7)
c (t ) σ(c (t ))

2. the law of motion of a(t )

ȧ(t ) = r (t )a(t ) + w (t ) c (t )

3. the no-Ponzi game condition

ρt
lim a(t )u 0 (c (t ))e =0
t !∞

4. the initial condition a(0) = a0

34 / 65
2. The competitive equilibrium
2.3. Behavior of …rms

I Firms maximize pro…ts

max F (K (t ), L(t )) w (t )L(t ) [r (t ) + δ ]K (t )

fK (t ),L (t )g

where δ denotes the rate of depreciation of capital:

I One unit of capital borrowed at date t for a small period dt
I rental costs r (t )dt
I depreciation is δdt
I The user cost of capital is r (t ) + δ

35 / 65
2. The competitive equilibrium
2.3. Behavior of …rms

I The …rst order conditions yield

f 0 (k (t )) = r (t ) + δ (8)
0
f (k (t )) k (t )f (k (t )) = w (t ) (9)

I The interest rate is equal to the marginal productivity of

capital (net of the depreciation cost)
I The wage is equal to the marginal productivity of labor

36 / 65
2. The competitive equilibrium
2.4. Characterization of equilibrium
I There are 2 factor markets: labor market, capital market
I Labor market equilibrium:

L(t ) = 1, for all t 0

so that k (t ) = K (t ) in this simple model

I Capital market equilibrium

a (t ) = k (t )

I Total assets of households = capital stock because there is

zero excess supply of bonds
I Then, the law of motion of a(t ) implies (see eq. (18)) the
following law of motion of k (t )

k̇ (t ) = r (t )k (t ) + w (t ) c (t )
37 / 65
2. The competitive equilibrium
2.4. Characterization of equilibrium

I Since f (k (t )) = [r (t ) + δ]k (t ) + w (t ), the law of motion of

k (t )
k̇ (t ) = r (t )k (t ) + w (t ) c (t )
can be written

k̇ (t ) = f (k (t )) δk (t ) c (t )

I This equation shows that change in the capital stock at time

t is equal to production, minus the depreciation of capital,
minus consumption at time t

38 / 65
2. The competitive equilibrium
2.4. Characterization of equilibrium
I Using eq. (8) and (7), we …nd that an equilibrium path
fc (t ), k (t ), t 0g satis…es
ċ (t ) 1
= [f 0 (k (t )) δ ρ] (10)
c (t ) σ(c (t ))
k̇ (t ) = f (k (t )) δk (t ) c (t ) (11)

with the initial condition

k (0) = k0 > 0 (12)

and the terminal condition

lim a(t )u 0 (c (t ))e ρt

=0 (13)
t !∞

I Once fc (t ), k (t ), t 0g is known, eq. (8) and (9) yield

fr (t ), w (t ), t 0g
39 / 65
2. The competitive equilibrium
2.5. Steady state equilibrium

I In steady state: c, k, w , r are constant over time,

I Eq. (7) and (8) imply that the steaty state equilibrium values
of r and k are de…ned by

r = ρ (14)
0
f (k ) = ρ + δ (15)

I Remark: the steady state is not unique

I In particular, the situation k = c = 0 is a steady state
equilibrium, like in the Solow model
I Steady state equilibrium consumption is

c = f (k ) δk

40 / 65
2. The competitive equilibrium
2.5. Steady state equilibrium

I Eq. (15) is the modi…ed golden rule: the level of capital does
not maximize steady stated consumption, as with the golden
rule, because future consumption is discounted at rate ρ
I The steady state level of capital decreases with the discount
rate ρ
I The steady steady state saving rate δk /f (k ) decreases with
ρ
I Is this solution socially e¢ cient ??

41 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

I Let us analyze the properties of the equilibrium values of

fc (t ), k (t ), t 0g
I The uniqueness of steady state equilibrium does not imply the
uniqueness of the equilibrium path fc (t ), k (t ), t 0g ,
because several paths can converge toward the same steady
state
I The analysis of the dynamics of fc (t ), k (t ), t 0g must take
into account that
I c (t ) is a control variable which can adjust instantaneously
I k (t ) is a state variable, whose value is predetermined at t and
whose dynamics is de…ned by the law of motion

k̇ (t ) = f (k (t )) δk (t ) c (t )

42 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

I Phase diagram: graphical analysis of the system of eq. (10),

(11) in the (c (t ), k (t )) plane
I Eq. (10),

ċ (t ) 1
= [f 0 (k (t )) δ ρ]
c (t ) σ(c (t ))

implies that 8
< ċ (t ) > 0 if k (t ) < k
ċ (t ) = 0 if k (t ) = k (16)
:
ċ (t ) < 0 if k (t ) > k
I These properties are displayed in the next …gure

43 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

k* k

44 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

I Eq. (11),
k̇ (t ) = f (k (t )) δk (t ) c (t )
implies that
8
< k̇ (t ) > 0 if c (t ) < f (k (t )) δk (t )
k̇ (t ) = 0 if c (t ) = f (k (t )) δk (t ) (17)
:
k̇ (t ) < 0 if c (t ) > f (k (t )) δk (t )

I These properties are displayed in the next …gure

45 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

kgold k

46 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

I Using the 2 previous …gures

c

c*

k* k

47 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

I The …gure shows there is a single path of fc (t ), k (t ), t 0g

that converges toward the steady state
I The path is such that
I c (0) jumps instantaneously on the stable arm and c (t ), t > 0
go monotically toward the steady state value c
I k (0) = k0 and k (t ), t > 0 go monotically toward the steady
state value k

48 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

c*

k* kgold k

49 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

I The equilibrium path is unique, because

I If c (0) is above the stable arm, c (∞) > 0 and k (∞) = 0, do
not satisfy the feasibilty constraint: impossible
I If c (0) is below the stable arm, c (∞) = 0 and k (∞) > kgold ,
such that f 0 (kgold ) = δ. Using the necessary conditions (4) and
(5) and the demand for capital r (t ) + δ = f 0 (k (t )) we have

du 0 (c (t ))/dt
=ρ r (t ) = ρ f 0 (k (t )) + δ > ρ
u 0 (c (t ))

when t ! ∞ because k (∞) > kgold entails that

f 0 (k (∞)) < δ. Therefore, u 0 (c (t )) grows at a larger rate than
ρ, which is not compatible with the transversality condition
ρt
lim k (t )u 0 (c (t ))e = 0.
t !∞

50 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

I Example (Barro and Salai-i-Martin, 2004)

I Economy with Cobb-Douglas technology

K α (AL)1 α

and CRRA preferences, σ = 3 and

I δ = 0.05, ρ = 0.02
I α = 0.3 or α = 0.75
I Ȧ/A = 0.02, L̇/L = 0.01,
I Behavior of an economy where k (0) = 0.1k , k = K /AL.

51 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

Output per e¢ cient unit of labor: y = Y /AL, (α = 0.3 =

continuous line, α = 0.74 = dotted line)

52 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

Saving rate (α = 0.3 = continuous line, α = 0.74 = dotted line)

53 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

Capital output ratio (α = 0.3 = continuous line, α = 0.74 =

dotted line)

54 / 65
2. The competitive equilibrium
2.6. Transitional dynamics

Interest rate (α = 0.3 = continuous line, α = 0.74 = dotted line)

55 / 65
Concluding comments

I We have studied one of the most important model in

macroeconomics
I Allows us to link savings to preferences, technology and prices
I In the next lecture we will see how this model sheds light on
1. The optimality of growth
2. Fiscal policy: the neutrality of public debt
3. distortionary taxation

56 / 65
APPENDIX
A. The intertemporal budget constraint and the no-Ponzi game condition

I The no-Ponzi game condition allows us to write the

intertemporal budget constraint, when T ! ∞ as follows
Z ∞ Z t Z ∞ Z t
c (t ) exp r (x )dx dt = w (t ) exp r (x )dx dt
0 0 0 0
(18)
I This intertemporal budget constraint shows the equality
between:
I the present value of all consumptions over t 2 [0, ∞)
(left-hand side)
I the present value of incomes over t 2 [0, ∞), equal to the
wealth

57 / 65
APPENDIX
A. The intertemporal budget constraint and the no-Ponzi game condition

I The intertemporal budget constraint over the dates [0, T ],

can be written in the same way as in the 2 period model (see
eq (1)):

a (T ) =
value of the assets at date T
Z T Z T
w (t ) exp r (x )dx dt
0 t
Value of the sum of incomes at date T
Z T Z T
c (t ) exp r (x )dx dt
0 t
Value of the sum of consumptions at date T

58 / 65
APPENDIX
A. The intertemporal budget constraint and the no-Ponzi game condition

I Let us multiply each side of the previous equality by

RT
exp 0
r (x )dx dt to get an evaluation at date 0 :
Z T Z t
c (t ) exp r (x )dx dt = (19)
0 0
Z T Z t
w (t ) exp r (x )dx dt
0 0
Z T
a(T ) exp r (x )dx dt
0

I Using the
h no-Ponzi game
Rt
condition,
i
limt !∞ a(t ) exp 0
r ( x ) dx = 0 we get eq (18).

59 / 65
APPENDIX
B. Transversality condition and no-Ponzi game condition

I Let us show that the transversality condition implies the

no-Ponzi game condition
I The no-Ponzi game condition is
Z t
lim a(t ) exp r (x )dx =0
t !∞ 0

I The transversality condition is

lim a(t )µ(t ) = 0

t !∞

60 / 65
APPENDIX
B. Transversality condition and no-Ponzi game condition

∂H (t )
I The necessary condition = r (t ) µ (t ) = µ̇(t ) implies
∂a (t )

Z t
µ(t ) = µ(0) exp r (x )dx
0

∂H (t )
I The necessary condition ∂c (t )
= u 0 (c (t ))e ρt µ (t ) = 0
yieds Z t
µ(t ) = u 0 (c (0)) exp r (x )dx
0

61 / 65
APPENDIX
B. Transversality condition and no-Ponzi game condition
I Substituting
Z t
µ(t ) = u 0 (c (0)) exp r (x )dx
0

in the transversality condition

lim a(t )µ(t ) = 0
t !∞

yields
Z t
0
lim a(t )u (c (0)) exp r (x )dx =0
t !∞ 0

I Or, since u 0 (c (0)) is positive and …nite

Z t
lim a(t ) exp r (x )dx =0
t !∞ 0

which is the no-Ponzy game condition

62 / 65
APPENDIX
C. Keynes-Ramsey rule in discrete time and in continuous time

I Let us show that the Keynes-Ramsey rules in discrete time

and in continuous time are actually identical
I In discrete time (the 2 period model), the Keynes-Ramsey is

1+r 0
u 0 ( c0 ) = u ( c1 )
1+ρ
I In continuous time, the Keynes-Ramsey is

ċ (t ) u 0 (c (t ))
= [r (t ) ρ]
c (t ) u 00 (c (t ))c (t )

63 / 65
APPENDIX
C. Keynes-Ramsey rule in discrete time and in continuous time

I Let us write the formula in discrete time over a small interval

of time dt ! 0 :
1 + r dt 0
u 0 (c (t )) = u (c (t + dt )) (20)
1 + ρdt
I By de…nition

u 0 (c (t + dt )) u 0 (c (t ))
u 00 (c (t )) =
c (t + dt ) c (t )
or

u 0 (c (t )) = u 0 (c (t + dt )) [c (t + dt ) c (t )] u 00 (c (t ))

64 / 65
APPENDIX
C. Keynes-Ramsey rule in discrete time and in continuous time

I Substituting this expression of u 0 (c (t )) into (20), and using

the approximation
1 + r dt
lim = 1 + (r ρ)dt,
dt !0 1 + ρdt

we get

[c (t + dt ) c (t )]
u 00 (c (t )) = u 0 (c (t + dt ))(r ρ)
dt
which yields

ċ (t ) u 0 (c (t ))
= [r (t ) ρ]
c (t ) u 00 (c (t ))c (t )

65 / 65