202: Dynamic Macroeconomics

Neoclassical Growth with Optimizing Agents (Ramsey-Cass-Koopmans

Model)

Mausumi Das

Delhi School of Economics

Jan 22-28-29 & Feb 2-9-13 &18, 2015

Jan 22-28-29 & Feb 2-9-13 &18, 2015 1

Das (Delhi School of Economics) Dynamic Macro / 86
Critique of Solow Model and Subsequent Extensions:

Recall the two basic critisisms of the Solow Model:

Even though it is supposed to be a growth model - it cannot really
explain long run growth:
The steady state in the Solow model may be dynamically ine¢ cient.
The basic Solow growth model has subsequently been extended to
counter some of these critisisms.
We have already looked at one such extension: Solow Model with
Exogenous Technological Progress
In today’s class we shall look at the other extension: Neoclassical
Growth Model with Optimizing Agents

Jan 22-28-29 & Feb 2-9-13 &18, 2015 2

Das (Delhi School of Economics) Dynamic Macro / 86
Neoclassical Growth with Optimizing Agents:

Let us now extend the Solow model to allow for optimizing agents.
There are two frameworks which allow for optimizing
consumption/savings behaviour by households:
1 The Ramsey-Cass-Koopmans Ini…nite Horizon Framework (henceforth
R-C-K);
2 The Samuelson-Diamond Overlapping Generations Framework
(henceforth OLG).
The basic di¤erence between the two is that in the R-C-K model
agents optimize over in…nite horizon; while in the OLG model, agents
optimize over a …nite time horizon (usually 2 periods).
As we shall see later, this apparently innocuous di¤erence in terms of
time horizon spells out very di¤erent growth trajectories for the two
models.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 3

Das (Delhi School of Economics) Dynamic Macro / 86
Neoclassical Growth with Optimizing Agents: The R-C-K
Model
We start with the R-C-K model. This model is still Neoclassical -
beacuse it retains all the assumptions of the Neoclassical production
function (including the diminishing returns property and the Inada
conditions.)
In fact the production side story is exactly identical to Solow.
As before, the economy starts with a given stock of capital (Kt ) and
a given level of population (Nt ) at time t. (We are ignoring
technological progress for now).
Since the production side story is identical to Solow, we know that
the …rm-speci…c production functions can be aggregated to generate
an aggregate production function:Yt = F (Kt , Nt ).
And at every point of time the market clearing wage rate and the
rental rate of capital are given by:
wt = FN (Kt , Nt ); rt = FK (Kt , Nt ).
Jan 22-28-29 & Feb 2-9-13 &18, 2015 4
Das (Delhi School of Economics) Dynamic Macro / 86
The R-C-K Model: The Household Side Story

There are H identical households indexed by h.

Each household consists of a single in…nitely lived member to begin
with (at t = 0). However population within a household increases
over time at a constant rate n. (And each newly born member is
in…nitely lived too!) This implies that total population also increases
at the rate n.
At any point of time t, the total capital stock and the total labour
force in the economy are equally distributed across all the households,
which are supplied inelastically to the market at the market wage rate
wt and the market rental rate rt .
Thus total earning of a household at time t: wt Nth + rt Kth .
Corresponding per member earning: yth = wt + rt kth ,
where kth is the per member capital stock in household h,
which is also the per capita capital stock (or the capital-labour
ratio, kt ) in the economy.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 5
Das (Delhi School of Economics) Dynamic Macro / 86
The Household Side Story (Contd.):

In every time period, the instantaneous utility of the household

depends on its per member consumption:

ut = u cth ; u 0 > 0; u 00 < 0; lim u 0 (c h ) = ∞; lim u 0 (c h ) = 0.

c h !0 c h !∞

The household at time 0 chooses its entire consumption pro…le

∞
cth t =0 so as to maximise the discounted sum of its life-time utility:

Z∞
U0h = u cth exp ρt
dt; ρ > 0,
t =0

subject to the household’s budget constraint in every time period.

Notice that identical households implied that per member
consumption (cth ) of any household is also equal to the per capita
consumption (ct ) in the economy at time t.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 6
Das (Delhi School of Economics) Dynamic Macro / 86
Interpretation of the Discount Rate:

Notice that the objective function of the household is an integral

de…ned over in…nite horizon, where future utilities are discounted at a
constant rate ρ. The discount rate (ρ) may have three possible
interpretations.
It is easier to understand these interpretations if we write down the
discrete time counterpart of the above objective function:

u c1h u c2h
U0h ' u c0h + + + .........
1+ρ (1 + ρ )2

1. One interpretation of the above ini…nte horizon utility function is that

agents are immortal (live for ever), but they have an innate
(psychological) tendency to prefer current consumption over future
consumption; hence they discount utilities from consumption that
happen in future dates. In this case ρ is interpreted as the "pure"
rate of time preference of an agent.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 7
Das (Delhi School of Economics) Dynamic Macro / 86
Interpretation of the Discount Rate (Contd.):

Notice that the curvature of the utility function itself to some extent
captures the preference of an agent over consumptions at two dates:
t and t + 1. But this measure is ‘impure’in the sense that it depends
on the precise amounts of cth and cth+1 .
If I already have too much of cth and too little of cth+1 , then I might
prefer an extra unit of future consumption more than an extra unit of
today’s consumption; and it would be the other way round if my cth is
too low compared to cth+1 .
This happens simply because the utility function is concave in c, which
induces this kind on ‘consumption smoothing’.
A ‘pure’rate of time preference measures the agent’s preference for
current consumption over future consumption even when the actual
consumption at the two time periods are exactly equal (i.e., cth =
cth+1 ). This way it neutralizes the e¤ect of concavity of the utility
function and looks at the pure psychological preference for today
vis-a-vis tomorrow - which is independent of consumption smoothing.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 8
Das (Delhi School of Economics) Dynamic Macro / 86
Interpretation of the Discount Rate (Contd.):

In a two period set up where the total lifetime utility is de…ned by

U (ct , ct +1 ), the ‘pure’rate time preference is de…ned as:
∂U ∂U
∂ct ∂ct +1
∂U
∂ct +1 ct =ct +1 =c̄

In other words, it measures the rate of change in the marginal

valuation on an extra unit of consumption available today vis-a-vis
available tomorrow - along a constant consumption path.
It is easy to see that in our additive utility speci…cation where

u ( ct + 1 )
U ( ct , ct + 1 ) = u ( ct ) +
1+ρ

the above de…nition coincides with the discount rate ρ.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 9
Das (Delhi School of Economics) Dynamic Macro / 86
Interpretation of the Discount Rate (Contd.):

2. Another interpretation of the discount rate ρ follows if we read the

in…nite horizon utility function of the household as the sum of utilities
of successive generations of agents who themselves are …nitely lived,
but who care for their future generations.

To understand the idea better, suppose each member of the

household lives exactly for one period. But in every successive period
(1 + n ) proprtion of new members are born (also with a life-time of
exactly one period), each of whom are an exact replica of the previous
set of agents (i.e, have identical tastes and preferences).
Each agent cares for the utility of her child, who in turn cares for the
utility of her child and so on....In other words, the agents are altruistic
towards their children. But the altruism is ‘imperfect’in the sense
that they care a little less for their children than they do for
themselves.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 10
Das (Delhi School of Economics) Dynamic Macro / 86
Interpretation of the Discount Rate (Contd.):

By this de…nition then, utlity of an agent belonging to genertaion t :

1
U t = u ( ct ) + Ut + 1 .
1+ρ
If we now expand the successive values of Ut , then we shall get back
the utility function of an agent belonging to generation 0 as:
1 1
U 0 = u ( ct ) + u ( ct + 1 ) + u (ct +2 ) + ........
1+ρ (1 + ρ )2
In other words, we shall get back the in…nite horizon utility function
as had been de…ned earlier, except that the term ρ now measures the
‘degree of parental altruism’.
The lower is ρ, the higher is the parental altruism.
When ρ = 0, there is ‘perfect’altruism (i.e., parents care as much for
their children as they care for themselves).
Jan 22-28-29 & Feb 2-9-13 &18, 2015 11
Das (Delhi School of Economics) Dynamic Macro / 86
Interpretation of the Discount Rate (Contd.):
3. A third interpretation of the discount rate ρ follows if we allow each
agent to potentially live forever, but introduce a constant
(age-independent) mortality risk at every time period.
Suppose an agent lives for sure in the …rst period of his life (when he
is born); but at every subsequent period he faces a constant
probablility of dealth, denoted by p.
If the agent is alive in any time period t, then he can enjoy utility
from consumption at that point of time, given by u (ct ). But if he
dies then he gets zero utility.
Thus beginning at time 0, the expected life-time utility of the agent
will be given by:
U0 = u (ct ) + pu (ct +1 ) + p 2 u (ct +2 ) + ........
1
Without any loss of generality, replace p by and we shall get
1+ρ
back the in…nite horizon utility function as had been de…ned earlier.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 12
Das (Delhi School of Economics) Dynamic Macro / 86
The R-C-K Model: Centralized Version (Optimal Growth)
There are two version of the R-C-K model:
A centralized version - which analyses the problem from the perspective
of a social planner.
A decentralized version - which analyses the problem from the
perspective of a perfectly competitive market economy where
‘atomistic’households and …rms take optimal decisions in their
respective individual spheres.
The centralized version was developed by Ramsey (way back in 1928)
and is oftem referred to as the ‘optimal growth’problem.
It is assumed that there exists an omniscient, omnipotent,
benevolent social planner who wants to maximise the citizens’
welfare.
Since all households are identical, the objective function of the social
planner is identical to that of the households:
Z∞
ρt
U0 = u (ct ) exp dt. (1)
t =0 Jan 22-28-29 & Feb 2-9-13 &18, 2015 13
Das (Delhi School of Economics) Dynamic Macro / 86
The R-C-K Model: Centralized Version (Contd.)

The social planner maximises (1) subject to the planner’s budget

constraint in every period.
Notice that in a centrally planned economy there are no markets
(hence no market wage rate or market rental rate), and there is no
private ownership of assets (capital) and no personalized income.
The social planner employs the existing capital stock in the economy
(either collectively owned or owned by the government) and the
existing labour force to produce the …nal output -using the aggregate
production technology.
After production it distributes a part of the total output among its
citizens for consumption puoposes and invests the rest.
Thus the budget constraint faced by the planner in period t is
nothing but the aggregate resource constraint:

Ct + It = Yt = F (Kt , Nt ).
Jan 22-28-29 & Feb 2-9-13 &18, 2015 14
Das (Delhi School of Economics) Dynamic Macro / 86
The R-C-K Model: Centralized Version (Contd.)
dK
Investment augments next period’s capital stock: = It .
dt
Thus the budget constraint faced by the planner in period t is given
by:
dK
Ct + = F (Kt , Nt ) δKt .
dt
Writing in per capita terms:
dk
ct + = f (kt ) δkt nkt .
dt
Thus the dynamic optimization problem of the social planner is:
Z∞
ρt
u (ct ) exp dt (I)
t =0
subject to
dk
= f (kt ) (δ + n)kt ct ; kt = 0 for all t; k0 given.
dt
Jan 22-28-29 & Feb 2-9-13 &18, 2015 15
Das (Delhi School of Economics) Dynamic Macro / 86
A Digression: Dynamic Optimization in Continuous Time
(Optimal Control)

Consider the following optimization problem which is de…ned over a

…nite time horizon from 0 to T :

ZT
W = F (ut , xt , t ) dt (2)
t =0

subject to
dx
= g (ut , xt , t ); ut 2 U; x0 given.
(i)
dt
Here ut is called the control variable; xt is called the state variable; F
represents the instantaneous payo¤ function, or the felicity function.
(i) speci…es the evolution of the state variable as a function of the
state and control variables.
It is called the equation of motion or the state transition equation.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 16
Das (Delhi School of Economics) Dynamic Macro / 86
Optimal Control (Contd.):

The objective function here is an integral, and our task is to …nd out
a time path of the time dependent variable u from the corresponding
choice set U, (i.e., to choose a u 2 U for each point of time t starting
from 0 to T ) such that the value of this integral is maximized.
But our choice is not unconstrained. (Had it been so, a simple
point-by-point static optimization exercise would have given us the
required solution path).
Note that the F function depends not only on u but also on another
time dependent variable x. And our choice of u at each point of time
a¤ects the next period’s value of x through the given di¤erential
equation.
Thus our choice of u a¤ects the objective function directly, as well as
indirectly through x.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 17

Das (Delhi School of Economics) Dynamic Macro / 86
Optimal Control (Contd.):

ut is called the control variable because we choosing its value directly.

Once the value of ut is chosen in any time period t, the value of the
state variable evolves automatically through the state transistion
equation.
Notice that when we are considering the problem at time 0, the initial
value of the state variable is given to us, but not that of the control
variable.
The initial value of the control variable will also be optimally
chosen, along with all its subsequent values.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 18

Das (Delhi School of Economics) Dynamic Macro / 86
Optimal Control: Pontryagin’s Maximum Principle

Let ut be a solution path to the problem speci…ed in (2), and let xt

be the associated path for the state variable, where ut is a piece-wise
continuous function of t and xt is a strictly continuous but piece-wise
di¤erentiable in t. Then there exist a strictly continuous and
piece-wise di¤erentiable variable λt , and a function H de…ned as:

H (u, x, λ, t ) F (ut , xt , t ) + λt g (ut , xt , t ) ,

such that
1 H is maximized with respect to u at ut for all t 2 [0, T ] ;
∂H dλ
2 = ;
∂x (u ,x ,λ,t ) dt
t t
∂H dx
3 = ;
∂λ (u ,x ,λ,t ) dt
t t
4 λT = 0

Jan 22-28-29 & Feb 2-9-13 &18, 2015 19

Das (Delhi School of Economics) Dynamic Macro / 86
Optimal Control (Contd.):

The function H is called the Hamiltonian Function associated with

the given dynamic optimizations problem.
The newly introduced time dependent variable λt is called the
co-state variable associated with the state variable xt .
The co-state variable λt measures the change in the value of the
objective function W associated with an in…nitesimal change in the
dx
state variable x at time t (which is the same a change in the
dt
function or the constraint function.
If there were an exogenous tiny increment to the state variable at
time t, and if the problem were modi…ed optimally thereafter, then
the increment in the total value of the objective would be λt . Thus it
is the marginal valuation of an incremental change in the state
variable at time t
λt is therefore often referred to as the shadow price of the state
variable at time t.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 20
Das (Delhi School of Economics) Dynamic Macro / 86
Optimal Control (Contd.):
Pontryagin’s Maximum Principle gives us four …rst order necessary
conditions for the optimization problem de…ned in (2).
These necessary conditions are also su¢ cient if additionally the
following conditions (due to Mangasarian) hold:
the functions F and f are concave in (u, x );
λt = 0 for all t whenever f is nonlinear in either u or x.
The …rst three F.O.N.C’s are de…ned in terms of the Hamiltonian
function.
Note that if the Hamiltonian function is non-linear in u, then (1) can
be replaced by the condition
∂H
= 0,
∂u (u t ,xt ,λ,t )

provided the second order check is veri…ed.

The last condition of the Maximum Principle, which speci…es a
terminal condition for λt , is called the Transversality Condition.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 21
Das (Delhi School of Economics) Dynamic Macro / 86
Optimal Control (Contd.):

Sometimes depending on the speci…cation of the problem the

transversality condition may change.
In the above problem we are given an initial condition about the state
state variable, but nothing has be speci…ed about the terminal value
of the state. This type of problems are called problems with a free
terminal state, and the relevant transversality condition for this set
of problems are given by (4).
Alternatively you may have an optimization problem where not only
the initial state value, but the terminal value of the state is also
given: xT = x̄ (given).
This is a problem with a …xed terminal state. In this case the …rst
three F.O.N.C.s will again be given by (1) – (3). Only condition (4)
will be replaced by a new transversality condition now, given by

xT = x̄.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 22
Das (Delhi School of Economics) Dynamic Macro / 86
Optimal Control (Contd.):

Yet another type of problem speci…es a terminal condition on the

state variable in the form of an inequality. These are problems with a
truncated vertical terminal line. : xT = x̄ (given).
In this case once again the …rst three F.O.N.C.s will again be given by
(1) – (3). But condition (4) will be replaced by a new transversality
condition now, given by the following Complementray Slackness
condition:
λT = 0; xT = x̄; λT (xT x̄ ) = 0.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 23

Das (Delhi School of Economics) Dynamic Macro / 86
The R-C-K Model Revisited:
Let us now go back to the centralized version of the R-C-K model.
Recall that the social planner’s problem is given by:
Z∞
ρt
u (ct ) exp dt (I)
t =0
subject to
dk
= f (kt ) (δ + n)kt ct ; kt = 0 for all t; k0 given.
dt
Notice that the choice set for the control variable is: ct 2 R+ .
Also the terminal condition on the state variable can be written as:
limt !∞ kt = 0.
Corresponding Hamiltonian Function:
ρt
Ht = u (ct ) exp +λt [f (kt ) (δ + n)kt ct ]
Jan 22-28-29 & Feb 2-9-13 &18, 2015 24
Das (Delhi School of Economics) Dynamic Macro / 86
The R-C-K Model: Centralized Version (Contd.)

The Corresponding FONCs (which are also su¢ cient in this case):

∂H
H is maximixed with respect to ct ) = 0 for all t; (i)
∂ct
∂2 H
verify: 2 < 0
∂ct

∂H dλ
= ; (ii)
∂kt dt
∂H dk
= ; (iii)
∂λt dt

TVC: lim λt kt = 0. (iv)

t !∞

Jan 22-28-29 & Feb 2-9-13 &18, 2015 25

Das (Delhi School of Economics) Dynamic Macro / 86
The R-C-K Model: Centralized Version (Contd.)
Sometimes instead of the Hamiltonian Function, we use the
Current-value Hamiltonian Function, de…ned as:
Ĥt = Ht expρt
= u (ct ) + µt [f (kt ) (δ + n)kt ct ] ,
where µt = λt expρt is called the Current-value co-state variable.
FONCs in terms of the Current-value Hamiltonian:
∂Ĥ
Ĥ is maximixed with respect to ct ) = 0 for all t; (i)
∂ct
∂Ĥ dµ
= + µρ; (ii)
∂kt dt
∂Ĥ dk
= ; (iii)
∂µt dt
ρt
TVC: lim µt exp kt = 0. (iv)
t !∞
Jan 22-28-29 & Feb 2-9-13 &18, 2015 26
Das (Delhi School of Economics) Dynamic Macro / 86
Interpretation of the Optimality Conditions:

FONC (i):
∂Ĥ
= 0 ) u 0 (ct ) = µt for all t
∂ct
implies that the marginal utility from consumption at every point of
time must be equal to the shadow price of capital (i.e., the
incremental utility associated with a unit increase in capital stock).
FONC (ii):

∂Ĥ dµ 1 dµ
= + ρµt ) f 0 (kt ) δ n + =ρ
∂kt dt µt dt

implies that the ‘net’rate of return (inclusive of capital gains/losses)

on savings must be equal to the minimum compensation required to
induce people to forego a unit of current consumption for the sake of
tomorrow (i.e., the agents’subjective rate of time preference).
Jan 22-28-29 & Feb 2-9-13 &18, 2015 27
Das (Delhi School of Economics) Dynamic Macro / 86
Interpretation of the Optimality Conditions (Contd.):
FONC (iii):

∂Ĥ dk dk
= ) = f (kt ) (δ + n)kt ct
∂µt dt dt
denotes the per capita budget constraint of the social planner.
Finally, the Transversality Condition:
ρt
lim µt exp kt = 0
t !∞
implies that at the terminal time
if the shadow price of capital is positive, no capital stock should be left
unused (unconsumed) and the economy must end up with zero capital
stock (µT > 0 ) kT = 0);
on the other hand, if some capital stock is indeed left unused then it
must be the case that the corresponding shadow price is zero (i.e.,
consuming further generates no utility value) (kT > 0 ) µT = 0).
Needless to say in this in…nite horizon problem, the above conditions
hold in a limiting sense.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 28
Das (Delhi School of Economics) Dynamic Macro / 86
Interpretation of the Current Value Hamiltonian Function:
The Current-value Hamiltonian Function:
Ĥt = u (ct ) + µt [f (kt ) (δ + n)kt ct ]
measures the utility valuation of the per capita GDP at any point of
time t.
Note that the per capita output at any time period f (kt ) can be used
for two purposes: to be enjoyed as consumption (ct ) and to augment
dk
the capital stock .
dt
The part that is consumed generates direct utility given by u (ct ) .
That part that is used for investment generates potential future
consumption and associated with an utility valuation of µt .
Thus the Current Value Hamiltonian measures the direct as well as
the indirect utility associated with the per capita output at any time
period t.
The Hamitonian (or the Present-value Hamiltonian), Ht , measures
the present discounted utility value of the per capita output at time t.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 29
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (Centralized Version): Characterization of
the Optimal Path
To summarise, the optimal trajectories of ct , kt and µt must satisfy
the following set of equations at every point of time t:
u 0 ( ct ) = µt ; (i)
1 dµ
= ρ f 0 (kt ) δ n ; (ii)
µt dt
dk
= f (kt ) (δ + n)kt ct ; (iii)
dt
ρt
lim µt exp kt = 0. (iv)
t !∞

Notice that even though we have three time-dependent variables (ct ,

kt and µt ), ct and µt are always tied to each other by virtue of
equation (i); hence their dynamic paths are also inter-dependent.
Thus we can eliminate one of them to get a system of di¤erential
equations either in (ct and kt ) or in (kt and µt ).
Jan 22-28-29 & Feb 2-9-13 &18, 2015 30
Das (Delhi School of Economics) Dynamic Macro / 86
Characterization of the Optimal Path (Contd.):

Here we shall eliminate µt and work with ct .

(Verify that you reach the same conclusions when you eliminate
ct and work with µt instead).
Log-di¤erentiating (i), and using (ii):
u 00 (ct ) dc 1 dµ
= =ρ f 0 (kt ) δ n
u 0 (ct ) dt µt dt
dc ct
) = f 0 (kt ) δ n ρ . (v)
dt ct u 00 (ct )
u 0 (c t)

We also know:
dk
= f (kt ) (δ + n)kt ct . (iii)
dt
Equations (iii) & (v) represent a 2 2 system of di¤erential equations
which along with the Transversality Condition characterize the
optimal path of the economy under the centralized R-C-K model.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 31
Das (Delhi School of Economics) Dynamic Macro / 86
Intertemporal Elasticity of Substitution:

Notice that in equation (v), there is a term:

ct u 00 (ct )
σ ( ct ) .
u 0 ( ct )

This term has multiple interpretations.

1 The most obvious interpretation is that it is the elasticity of

marginal utility with respect to consumption.
2 In choices under uncertainty, σ(ct ) coincides with the Arrow-Pratt
measure of relative risk aversion.
3 The σ(ct ) terms is also the inverse of the elasticity of
substitution between current and future consumption.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 32

Das (Delhi School of Economics) Dynamic Macro / 86
Intertemporal Elasticity of Substitution (Contd.):

Note that elasticity of substitution between consumption at date t

and consumption at date t + ∆t is de…ned as
ct ct
d ct +∆t / ct +∆t
e= .
u 0 (c t ) u 0 (c t )
d u 0 (ct +∆t )
/ u 0 (ct +∆t )

It can be shown that as ∆t ! 0, e ! σ1 . (Verify this.)

It is sometimes convenient to work with utility functions where σ (ct )
is a constant.
Examples:
Log utility function: u (ct ) = log ct
(ct )1 θ
CRRA utility function: u (ct ) = ; θ 6= 1.
1 θ
For the time being, however, we shall work with a general utility
function where σ(ct ) need not be a constant.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 33
Das (Delhi School of Economics) Dynamic Macro / 86
Characterization of the Optimal Path (Contd.):

The 2 2 non-linear and autonomous system of equations for the

centralized economy are given by:

dc ct
= f 0 (kt ) δ n ρ ; (v)
dt σ ( ct )
and
dk
= f (kt ) (δ + n)kt ct . (iii)
dt
Both (iii) and (v) are non-linear di¤erential equations; so we have to
use phase diagram technique to qualitatively characterize the optimal
path.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 34

Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (Centralized Version): Steady State(s)
First let us identify the possible staedy state(s) of the dynamic system.
The steady state is now de…ned as pair of values (k, c ) such that
neither values change over time.
In other words, the steady states are de…ned by the following two
equation:

ct
f 0 (kt ) δ n ρ = 0;
σ ( ct )
f (kt ) (δ + n )kt ct = 0.
Notice that σ (ct ) > 0. Hence from the above equations we can
identify three possbile steady states of the system:
Trivial steady state : c = 0; k = 0;
Semi-trivial steady state : c = 0; k = k̄ such that f (k̄ ) = δ + n;
Non-trivial steady state: c = c > 0; k = k > 0 such that
0
f (k ) = δ + n + ρ; c = f (k ) ( δ + n )k .
Jan 22-28-29 & Feb 2-9-13 &18, 2015 35
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (Centralized Version): Constrcution of the
Phase Diagram
From equation (v):

dc
T 0 according as
dt
either ct = 0 or f 0 (kt ) T δ + n + ρ.
On the other hand, from equation (iii):
dk
T 0 according as
dt
ct S f (kt ) (δ + n )kt .
dc dk
Now we can trace the level curves = 0 and = 0 in the (kt , ct )
dt dt
plane and draw the coresponding directional arrows to get the
corresponding phase diagram.
(How should the Phase Diagram look?)
Jan 22-28-29 & Feb 2-9-13 &18, 2015 36
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (Centralized Version): Phase Diagram

Jan 22-28-29 & Feb 2-9-13 &18, 2015 37

Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (Centralized Version): Characterization of
the Optimal Path (Contd.)
Notice that any pair of directional arrows in the phase diagram satisfy
(by construction) the dynamic equations (iii) and (v) and therefore
satisfy the …rst three FONCs of the given dynamic optimization
problem .
Also note that k0 is given, but c0 is not. In fact our choice of c0
would generate multiple possible time paths of ct and kt - all
satisfying the …rst three FONCs.
One can classify these multiple trajectories in three broad categories:
Category I: Trajectories that move towards the horizontal axis over
time and eventually approach the point (k̄, 0) as t ! ∞;
Category II: Trajectories that approach the vertical axis over time;
Category III: A unique trajectory that represents the stable arm of
the saddle point (k , c ) and approaches the non-trivial steady state
point (k , c ) over time.
Which one of these is the optimal trajectory?
Here the transversility condition comes to our Jan
rescue.
22-28-29 & Feb 2-9-13 &18, 2015 38
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (Centralized Version): Characterization of
the Optimal Path (Contd.)

Recall that the TVC is part of the necessary (and su¢ cinet)
conditions for optimality.
So among all these trajectories, the one which satis…es the TVC will
indeed be the optimal path. (What if there are multiple such
trajectories?)
As it turns out, only the unique trajectory belonging to Category III
(represented by the line SS 0 in the diagram) satis…es all the four
FONCs including the transversility condition.
We now provide heuristic arguments as to why trajectories belonging
to the other two categories cannot be optimal.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 39

Das (Delhi School of Economics) Dynamic Macro / 86
Proof that Trajectories of Type I cannot be Optimal:

Recall that the Transversality condition for the central planner’s

problem is speci…ed as:
ρt
TVC: lim µt exp kt = 0. (iv)
t !∞

Also noting that along the optimal path, µt = u 0 (ct ), we can write
the TVC as:
TVC: lim u 0 (ct ) exp ρt kt = 0. (iv0)
t !∞

Now let us check whether trajectories belonging to Category I satisfy

condition (iv0).
Notice that along a trajectory of type I, limt !∞ kt = k̄ > 0.
Hence condition (iv0 ) will be satis…ed along these trajectories if and
only if
TVC: lim u 0 (ct ) exp ρt = 0. (iv00 )
t !∞
Jan 22-28-29 & Feb 2-9-13 &18, 2015 40
Das (Delhi School of Economics) Dynamic Macro / 86
Proof that Trajectories of Type I cannot be Optimal:
(Contd.)
Now along a trajectory of type I, limt !∞ ct = 0 which implies that
limt !∞ u 0 (ct ) ! ∞.
At the same time, limt !∞ exp ρt = 0.
Thus it is not immediately clear whether condition (iv0 ) will be
satis…ed or not.
It depends on whether in the neighbourhood of (k̄, 0), u 0 (ct ) is
increasing at a faster/slower rate than the rate of fall of exp ρt .
The exponential term exp ρt is of course decreasing at a constant
rate ρ.
On the other hand, the u 0 (ct ) term is increasing at the rate
[(n + δ f 0 (k )) + ρ] .
In the neighbourhood of (k̄, 0), [(n + δ f 0 (k )) + ρ] > ρ. (Why?)
Thus u 0 (ct ) is increasing at a faster rate than the rate of fall of
exp ρt .
Jan 22-28-29 & Feb 2-9-13 &18, 2015 41
Das (Delhi School of Economics) Dynamic Macro / 86
Proof that Trajectories of Type I cannot be Optimal:
(Contd.)

Therefore along any trajectory of type I,

lim u 0 (ct ) exp ρt

! ∞,
t !∞

which violates the TVC (iv00 ).

Hence trajectories of type I cannot be optimal.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 42

Das (Delhi School of Economics) Dynamic Macro / 86
Proof that Trajectories of Type II cannot be Optimal:
The trajectories of type II approach the vertical axis over time.
If they approach the vertical axis asympototically (never actually
hitting it at any …nite point of time) then indeed the TVC (iv) will be
satis…ed and such trajectories would be optimal.
However we now argue that this is not possible. Indeed every
trajectory belonging to category II actually hit the vertical axis
within a …nite point of time.
We prove this by contradiction.
Suppose, if possible, that trajectories of type II approach but do not
hit the vertical axis within …nite time.
In other words, suppose, if possible, that trajectories of type II are
asymptotic either to the vertical axis or some line parallel to the
vertical axis.
If that is true then starting form any given …nite initial value, k0 ,
lim kt = M (where M is a non-negative constant < k0 ).
t !∞
Jan 22-28-29 & Feb 2-9-13 &18, 2015 43
Das (Delhi School of Economics) Dynamic Macro / 86
Proof that Trajectories of Type II cannot be Optimal:
(Contd.)

Now from FONC (iii), we know that,

dk
= f (kt ) (δ + n)kt ct .
dt
Therefore,
Zt
kt = k0 + [ f (kτ ) ( δ + n )kτ cτ ] d τ.
0

Hence
Z∞
lim kt = k0 + [ f (kτ ) ( δ + n )kτ cτ ] d τ = M.
t !∞
0

Jan 22-28-29 & Feb 2-9-13 &18, 2015 44

Das (Delhi School of Economics) Dynamic Macro / 86
Proof that Trajectories of Type II cannot be Optimal:
(Contd.)
Rearranging terms:

Z∞
[ cτ f f (kτ ) (δ + n)kτ g] d τ = k0 M N (a …nite constant).
0

We now argue that along any trajectory of type II, the integral
de…ned by the LHS above will diverge away to +∞ and therefore can
never converge to …nite constant N.
(Prove this yourself. Hint: A necessary condition for an in…nite
R∞ daτ
integral I aτ d τ to converge is that < 0. De…ne
0 dτ
aτ [cτ ff (kτ ) (δ + n )kτ g] here and show that along any
trajectory of type II this necessary condition is violated.)
Hence there is a contradiction.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 45
Das (Delhi School of Economics) Dynamic Macro / 86
Proof that Trajectories of Type II cannot be Optimal:
(Contd.)
We have just proved that any trajectory belonging to category II
cannot be asymptotic to the vertical axis; it must hit the
vertical axis within a …nite period of time.
Now take any such trajectory. Can it still be optimal?
The answer is "No".
The reason is as follows:
Suppose the trajectory hits the vertical axis precisely at time T .
then exactly at time T , kt reaches zero;
Consequently, ct falls from a …nite value to zero (since a positive value
of consumption cannot be sustained with zero capital stock);
This implies that precisely at time T , µt jumps from a …nite value to
in…nity.
Such a discrete jump of µt violates FONC (ii) - which presupposes
continuity of µt .
Hence trajectories of type II cannot be optimal.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 46
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (Centralized Version): Identi…cation of the
Optimal Trajectory
We have now seen that trejectories belonging to either category I or
category II cannot be optimal because they violate one of the FONCs
(i)-(iv).
That leaves the unique trajectory beloging to category III, which is
the stable arm of the saddle point (k , c ) and represented by the line
SS 0 in the diagram.
Along this trajectory, as t ! ∞, ct and kt approach c and k
respectively.
It is easy to verify that this trajectory satis…es all the four FONCs,
including the Transversality Condition.
Hence this is the unique optimal trajectory for the social planner’s
problem.
Thus, given k0 , it is optimal for the social planner to choose the
corresponding c0 that lies on trajectory III and then let the economy
evolve according to the two dynamic equations (iii) & (v).
Jan 22-28-29 & Feb 2-9-13 &18, 2015 47
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (Centralized Version): Growth Implications

In the centralized version of the R-C-K model (without technical

progress), once again the economy goes to a steady state in the long
run, where per capita income becomes constant.
So there is no long run growth of per capita income; aggreagte
income in the long run grows at the constant rate n.
Thus the growth conclusions of the centralized R-C-K model are
exactly identical to that of the Solow model (without technical
progress).
Excercise: Introduce exgoneous technical progress in this R-C-K
model as we did in Solow; show that even then the growth
conclusions would be the same - except that you now need more
restrictions on the utility function so that the TVC holds.
What happens during transition? Does conditional convergence
hold? (Verify using a log utility function and a Cobb-Douglas
production function.)
Jan 22-28-29 & Feb 2-9-13 &18, 2015 48
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (Centralized Version): Implications for
Dynamic E¢ ciency
Although the growth implications of the centralized R-C-K model are
exactly identical to that of Solow, there is a big di¤erence.
The steady state of the R-C-K model is always dynamically e¢ cient.
(Why?)
In fact with a positive rate of time preference (so that ρ > 0), the
steady state point (k : f 0 (k ) = n + δ + ρ) represents the "best
point’in the sense that it maximises the welfare of the household.
This point is called the ‘modi…ed golden rule’which is di¤erent from
the ‘golden rule’capital stock de…ned earlier (kg : f 0 (kg ) = n + δ)
Indeed the ‘golden rule’point is no longer the ‘best point’for the
agents (unless they have zero rate of time preference.) Even if I start
with an initial capital stock k0 = kg , it is not optimal for me to stay
at (kg , cg ) forever. I would like to move to the corresponding point of
the saddle path SS 0 and evetually approach (k , c ) because that, by
construction, maximizes my welfare (life-time utility).
Jan 22-28-29 & Feb 2-9-13 &18, 2015 49
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model: the De-centralized Version

So far we have analysed the centralized version of the R-C-K model,

where a benevolent social planner makes all the production and
savings/investment decisions on behalf of the households.
We now turn to the corresponding problem for a decentralized market
economy, where all these economic decisions are undertaken by
‘atomistic’…rms and households.
Perfect competition prevails - which means that while optimizing, the
…rms and households take all the market prices (in the commodity
market as well as in the factor markets) as given.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 50

Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (De-centralized Version): Production Side
Story
Recall that the production side story in the R-C-K model is identical
to that of Solow.
Thus the economy starts with a given stock of capital (Kt ) and a
given level of population (Nt ) at time t. (We are ignoring
technological progress for now).
From previous analysis, we also know that all …rms have access to an
identical production technology - which satis…es all standard
neoclassical properties.
The …rm-speci…c production functions can be aggregated to generate
an aggregate production function such that :
Yt = F (Kt , Nt ).
At every point of time the market clearing wage rate and the rental
rate of capital are given by:
wt = FN (Kt , Nt ); rt = FK (Kt , Nt ).
Jan 22-28-29 & Feb 2-9-13 &18, 2015 51
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (De-centralized Version): Household Side
Story
The household at time 0 chooses its entire consumption pro…le
∞
cth t =0 so as to maximise the discounted sum of its life-time utility:
Z∞
U0h = u cth exp ρt
dt; ρ > 0,
t =0
subject to the household’s budget constraint in every time period.
(u cth of course satis…es all the standard properties speci…ed earlier).
If we do not allow intra-household borrowing, then household h’s
budget constraint (in aggregate terms) would be given by:
dKth
Cth + Ith = wt Nth + rt Kth , where
= Ith δKth .
dt
In per member terms, household h’s budget constraint becomes:
dkth
= wt + rt kth (n + δ)kth cth .
dt
Jan 22-28-29 & Feb 2-9-13 &18, 2015 52
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model (De-centralized Version): Household Side
Story (Contd.)

Thus in the absence of intra-household borrowing the optimization

problem of the representative household h is given by:
Z∞
Max. ∞
U0h = u cth exp ρt
dt; ρ > 0,
fcth gt =0 t =0
subject to
dkth
= wt + rt kth (n + δ)kth cth ; kth = 0 for all t = 0; k0h given,
dt
where cth is the control variable and kth is the state variable.
Notice that in order to solve this problem the households would have
to have some expectation about the entire time paths of wt and rt .
We shall assume that households’have perfect foresight. So they
can correctly guess all the future values of the market wage rate and
rental rate. Jan 22-28-29 & Feb 2-9-13 &18, 2015 53
Das (Delhi School of Economics) Dynamic Macro / 86
De-centralized R-C-K Model: Household Side Story
(Contd.)

Note that the household’s problem is di¤erent from the social planner
problem in an important way: the households can borrow from one
another.
Allowing for intra-household borrowings means that a household’s
consumption at any point of time t need not be limited by its current
income and current capital stock.
The household can consume beyond its current income at any point
of time - by borrowing from others.
Allowing for intra-household borrowings also means that a household
now has two forms of assets that it can invest its savings into:
1 physical capital (Kth );
2 …nancial capital, i.e., lending to other households (Lht Bth ).

Jan 22-28-29 & Feb 2-9-13 &18, 2015 54

Das (Delhi School of Economics) Dynamic Macro / 86
De-centralized R-C-K Model: Household Side Story
(Contd.)
Let the interest rate on …nancial assets be denoted by r̂t .
We already know that the (net) interest rate on investment in
physical capital is given by (rt δ) .
Arbitrage in the asset market ensures that in equlibrium two interest
rates are the same :
r̂t = rt δ.
Hence we can de…ne the total asset holding by the household as
Aht Kth + Lht , where Lht, < 0 if the household is a net borrower.
Thus the aggregate budget constraint of the household now becomes:
dAht
Cth +
= wt Nth + r̂t Aht .
dt
Writing in per member terms:
dath
= wt + (r̂t n )ath cth .
dt
Jan 22-28-29 & Feb 2-9-13 &18, 2015 55
Das (Delhi School of Economics) Dynamic Macro / 86
Household Side Story: Ponzi Game
But allowing for intra-household borrowing brings in the possibility of
Ponzi game.
Consider the following plan by a household:
Suppose in period 0, the household borrows a huge amount B̄ - which
would allow him to maintain a very high level of consumption at all
subsequent points of time. Thus
B0h = B̄.
In the next period (period 1), he pays back his period 0 debt with
interest by borrowing again (presumably from a di¤erent lender). Thus
his period 1 borrowing would be:
B1h = (1 + r̂0 )B0 .
In period 2 he again pays back his period 1 debt with interest by
borrowing afresh:
B2h = (1 + r̂1 )B1h = (1 + r̂0 )(1 + r̂1 )B0h .
Jan 22-28-29 & Feb 2-9-13 &18, 2015 56
Das (Delhi School of Economics) Dynamic Macro / 86
Household Side Story: Ponzi Game (Contd.)
Proceeding this way,
Bth+1 = (1 + r̂t )Bth = (1 + r̂0 )(1 + r̂1 )...(1 + r̂t )B0h .
In other words, the household’s debt grows at the rate r̂t .
Notice that by playing this game, the household e¤ectively never pays
back its initial loan B̄; it is simply rolling it over period after period.
In the process it is able to maintain an arbitrarily high level of
consumption (over and above it’s current income).
This kind of …nancing scheme is called Ponzi …nance.
If a household is allowed to play such a Ponzi game, then the
household’s budget constraint becomes irrelevant. There is e¤ectively
no budget constraint for the household any more; it can maintain any
arbitrarily high consumption path by playing a Ponzi game.
To rule this out, we impose an additional constraint on the
household’s optimization problem - called the No-Ponzi Game
Condition.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 57
Das (Delhi School of Economics) Dynamic Macro / 86
Household Side Story: No-Ponzi Game Condition
Recall that when the the Household is playing the Ponzi game, it’s
debt grows at the rate r̂t :
1 dBth
= r̂t .
Bth dt
Hence the household’s debt in per capita (per member) terms
B th
bth Nh
grows at the rate r̂t n.
t
One Version of No-Ponzi Game (NPG) Condition, which rules out
this kind of behaviour is given below:
Rt
(r̂v n )dv
NPG Condition: lim ah exp 0 = 0.
t !∞ t

This Non-Ponzi Game condition states that as t ! ∞, the present

discounted value of the household’s asset (when the discount rate is
the population-adjusted interest rate) must be non-negative.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 58
Das (Delhi School of Economics) Dynamic Macro / 86
Implication of the No-Ponzi Game Condition:
Notice that one can always rule out Ponzi behaviour by banning
intra-household borrowing altogether.
But that is a stronger restriction than required.
In fact the above NPG condition is very ‡exible in the sense that it
allows households to borrow perpetually, but at the same time rules
out Ponzi behaviour.
How?
Let us …rst see how the NPG condition rules out Ponzi behaviour:
Recall that when the household is playing a Ponzi game, its per
member borrowing is increasing at the rate r̂t n, i.e.,
1 dbth
= r̂t n.
bth dt
Solving:
Rt
(r̂v n )dv
bth = b0h exp 0
Jan 22-28-29 & Feb 2-9-13 &18, 2015 59
Das (Delhi School of Economics) Dynamic Macro / 86
Implication of the No-Ponzi Game Condition: (Contd.)
Since the household is playing a Ponzi Game, b0h > 0. Also it must
have already consumed all its initial capital stock and its net asset
stock for any t > 0 must be given by: ath = bth .
Thus,
Rt
(r̂v n )dv
ath = b0h exp 0
Rt
(r̂v n )dv
) ath exp 0 = b0h
Taking the limiting value:
Rt
(r̂v n )dv
lim ath exp 0 = b0h < 0,
t !∞
which violates the given NPG condition.
In other words, playing a Ponzi game necessarily violates the NPG
condition.
Does it rule out borrowing altogether? The answer is "no".
Jan 22-28-29 & Feb 2-9-13 &18, 2015 60
Das (Delhi School of Economics) Dynamic Macro / 86
Implication of the No-Ponzi Game Condition: (Contd.)

Suppose the household follows a consumption path such that b0h > 0
and also any subequent time period beyond the initial point (i.e., for
any t > 0) ath = bth . In other words, suppose the household is once
again borrowing perpetually.
But now it’s borrowing grows at a rate gt < r̂t n.
(This implies that the household is paying at least part of the interest
payment in every period from its own pocket.)
Once again,
Rt
g v dv
bth = b0h exp 0
Rt
g v dv
) ath = bth = b0h exp 0
Rt Rt
(r̂v n )dv [(r̂v n ) gv ]dv
) ath exp 0 = b0h exp 0

Jan 22-28-29 & Feb 2-9-13 &18, 2015 61

Das (Delhi School of Economics) Dynamic Macro / 86
Implication of the No-Ponzi Game Condition: (Contd.)

If gt is low enough (in relation to (r̂t n )) then, as t ! ∞, the

Rt
intergral [(r̂v n ) gv ] dv would diverge to in…nity and hence the
0
Rt
[(r̂v n ) gv ]dv
exp 0 term will converge to zero.
Notice that in this case the NPG condition will be satis…ed, despite
the fact that the household is perpetually borrowing.
In other words, the NPG condition does not rule out borrowing (even
perpetual borrowing); it just requires the household to start paying
back its debt from its own pocket from some point of time onwards.
The implication of the NPG condition becomes clearer if we combine
it with the household’s ‡ow budget constraint.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 62

Das (Delhi School of Economics) Dynamic Macro / 86
Economic Implication of the NPG/TVC:

Recall that the household’s ‡ow budgent constraint (when

intra-household borrowing is allowed) is given by:

dath
= wt + r̂t ath nath cth
dt
dath
) (r̂t n ) ath = wt cth
dt
Rt Rt
(r̂v n )dv dath (r̂v n )d
) exp 0 (r̂t n ) ath = wt cth exp 0
dt
Rt Rt
d âth (r̂v n )dv (r̂v n )dv
) = wt exp 0 cth exp 0
dt
Rt
(r̂v n )dv
where âth ath exp 0 .

Jan 22-28-29 & Feb 2-9-13 &18, 2015 63

Das (Delhi School of Economics) Dynamic Macro / 86
Economic Implication of the NPG/TVC (Contd.):

Moving dt to the other side and intergrating both sides (from 0 to

∞):

Z∞ Z∞ Rt
(r̂v n )dv Z∞ Rt
(r̂v n )dv
d âth = wt exp 0 dt cth exp 0 dt
0 0 0
Z∞ Rt
(r̂v n )dv Z∞ Rt
(r̂v n )dv
) lim âh â0h = wt exp 0 dt cth exp 0
t !∞ t
0 0

Rt
(r̂v n )dv
Now recall that âth ath exp 0 .
Rt
(r̂v n )dv
Hence, â0h = a0h , and limt !∞ âth = lim ah exp 0 = 0 (by
t !∞ t
NPG).
Jan 22-28-29 & Feb 2-9-13 &18, 2015 64
Das (Delhi School of Economics) Dynamic Macro / 86
Economic Implication of the NPG/TVC (Contd.):

Thus,

Z∞ Rt
(r̂v n )dv Z∞ Rt
(r̂v n )dv
lim âth = â0h + wt exp 0 dt cth exp 0 dt = 0.
t !∞
0 0

By rearranging terms and substituting for â0h ,

Z∞ Rt
(r̂v n )dv Z∞ Rt
(r̂v n )dv
cth exp 0 dt 5 wt exp 0 dt + a0h .
0 0

In other words, the NPG condition implies that no matter what the
consumption path (and the consequent borrowing pattern) is for the
household, eventually the present value of the consumption stream
must be limited by the sum of its non-human and human wealth
(namely the discounted value of its labour earnings).
Jan 22-28-29 & Feb 2-9-13 &18, 2015 65
Das (Delhi School of Economics) Dynamic Macro / 86
Household’s Optimization Problem:

Imposing the No-Ponzi Game condition, the household’optimization

problem becomes:
Z∞
Max. ∞
U0h = u cth exp ρt
dt; ρ > 0,
fcth gt =0 t =0

subject to

dath
(i) = wt + r̂t ath nath cth ; a0h given.
dt
Rt
(r̂v n )dv
(ii) NPG Condition : lim ah exp 0 = 0.
t !∞ t

As before, we can write down the FONCs (which are also su¢ cient)
in terms of the corresponding Hamiltonian/Current-value Hamiltonian
function.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 66
Das (Delhi School of Economics) Dynamic Macro / 86
Household’s Problem: FONCs in terms of Hamiltonian

The Hamiltonian Function:

h i
Ht = u cth exp ρt
+λt wt + r̂t ath nath cth

Corresponding FONCs:

∂H
H is maximixed with respect to cth ) = 0 for all t
∂cth
i.e., u 0 cth exp ρt
= λt (i)

∂H dλ
=
∂ath dt
dλ
i.e., = λt [r̂t n] (ii)
dt
Jan 22-28-29 & Feb 2-9-13 &18, 2015 67
Das (Delhi School of Economics) Dynamic Macro / 86
Household’s Problem: FONCs in terms of Hamiltonian
(Contd.)

∂H dath
=
∂λt dt
dah
i.e., t = wt + r̂t ath nath cth (iii)
dt

TVC: lim λt ath = 0. (iv)

t !∞

In addition we have the NPG Condition:

Rt
(r̂v n )dv
lim ah exp 0 = 0. (v)
t !∞ t

(Too many boundary conditions?)

Jan 22-28-29 & Feb 2-9-13 &18, 2015 68
Das (Delhi School of Economics) Dynamic Macro / 86
Household’s Problem: FONCs in terms of Current-value
Hamiltonian
Current-value Hamiltonian Function:
h i
Ĥt = Ht expρt = u cth + µt wt + r̂t ath nath cth ,

where µt = λt expρt .
FONCs in terms of the Current-value Hamiltonian:
∂Ĥ
Ĥ is maximixed with respect to ct ) = 0 for all t
∂cth
i.e., u 0 cth = µt (i)

∂Ĥ dµ
= + µρ
∂ath dt
dµ
= µt [r̂t n ρ] (ii)
dt
Jan 22-28-29 & Feb 2-9-13 &18, 2015 69
Das (Delhi School of Economics) Dynamic Macro / 86
Household’s Problem: FONCs in terms of Current-value
Hamiltonian (Contd.)

∂Ĥ dath
=
∂µt dt
dath
i.e., = wt + r̂t ath nath cth (iii)
dt

TVC: lim µt exp ρt

ath = 0. (iv)
t !∞

And the NPG Condition:

Rt
(r̂v n )dv
lim ah exp 0 = 0. (v)
t !∞ t

Jan 22-28-29 & Feb 2-9-13 &18, 2015 70

Das (Delhi School of Economics) Dynamic Macro / 86
Household’s Problem: Optimal Solutions
From FONCs (i)-(iii) of the household’s optimization problem we get
the following dynamic equations:
dcth cth
= [r̂t n ρ] (1)
dt σ(cth )
dath
= wt + r̂t ath nath cth (2)
dt
Equations (1) and (2) represents a 2X 2 system of di¤erence equations
which implicitly de…nes the ‘optimal’trajectories of household h.
However we now have two dynamic equations, but three boundary
conditons: the initial condition a0h , the TVC and the NPG condition.
But note that along the optimal trajectory, when the TVC holds, then
the NPG condition is also satis…ed at the margin. Hence we can
combine the two together to get the following boundary condition:
Rt
(r̂v n )dv
lim ath exp 0 = 0. (vi)
t !∞
Jan 22-28-29 & Feb 2-9-13 &18, 2015 71
Das (Delhi School of Economics) Dynamic Macro / 86
De-centralized R-C-K Model: Solution Paths for the
Aggregate Economy
Given the optimal solutions to the household’s problem, we can now
aggregate over all households to get the corresponding time paths for
the economy-wide averages.
When households are all identical, then of course aggregation
becomes trivial:
ct1 = ct2 = ct3 = ..... = ctH = ct (average consumption);
at1 = at2 = at3 = ..... = atH = at (average asset holding).
Also, notice that average asset holding in the economy:
∑ Aht ∑ Kth + ∑ Lht ∑ Kth
at == = = kt ,
Nt Nt Nt
since aggregate borrowing (or lending) across all households must be
zero.
So when the households are identical, we can directly replace ath and
cth in equations (1) and (2) by kt and ct respectively.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 72
Das (Delhi School of Economics) Dynamic Macro / 86
De-centralized R-C-K Model: Solution Paths for the
Aggregate Economy (Contd.)
Thus,
dct ct
= [r̂t n ρ] (10 )
dt σ ( ct )
dkt
= wt + r̂t kt nkt ct (20 )
dt
Finally noting that in a competitive market economy:
wt = f (kt ) kt f 0 (kt ) and r̂t = rt δ = f 0 (kt ) δ, we get dynamics
of average consumption and per capita capita stock for the aggregate
economy as:
dct ct
= f 0 (kt ) δ n ρ (3)
dt σ ( ct )
dkt
= f (kt ) (n + δ)kt ct (4)
dt
Jan 22-28-29 & Feb 2-9-13 &18, 2015 73
Das (Delhi School of Economics) Dynamic Macro / 86
De-centralized R-C-K Model: Solution Paths for the
Aggregate Economy (Contd.)

Compare equations (3) and (4) with the dynamic equations derived
for the social planner earlier. Observe that they are exactly
identical!
This implies that under the R-C-K model, the optimal trajectories
for the decentralized market economy and the centralized
planning economy would be identical.
Since we have already proved that the steady state for the social
planner’s problem would be dynamically e¢ cient, so would be the
steady state for the market economy.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 74

Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model: Equivalence between Centralized and
De-centralized Economy

We have just seen that in the R-C-K model with identical households,
the solution paths of the social planner and that of the market
economy will coincide.
But in deriving this strong equivalence result, we have assumed the
households are identical in every respect.
In fact with identical household allowing for intra-household borrowing
and the consequent NPG condition become super‡uous because one
side of the lending/borrowing market will be always missing!
A more interesting questio is: will this strong equivalence result hold
even when households are heterogenous?
The answer is "yes", provided the utility function satis…es certain
additional properties.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 75

Das (Delhi School of Economics) Dynamic Macro / 86
Decentralized R-C-K Model with Heterogenous
Households:
Let us now introduce heterogenous households in the de-centralized
R-C-K model.
In particular let us assume that households have identical preferences
but they di¤er in terms of initial asset holdings.
Let the H households have initial asset holdings denoted by a01 , a02 , a03
.......and a0H respectively such that
a01 6= a02 6= a03 6= a04 .... 6= a0H .
We already know that along the optimal path, per member
consumption and asset stock of any household h will follow the
dynamic equations given below:
dcth cth
= [rt δ n ρ] (1)
dt σ(cth )
dath
= wt + (rt δ n ) ath cth (2)
dt Jan 22-28-29 & Feb 2-9-13 &18, 2015 76
Das (Delhi School of Economics) Dynamic Macro / 86
Decentralized R-C-K Model with Heterogenous Households
(Contd.):
Now aggregating over all households (which are no longer identical),
we can write the average asset holding in the economy as:
∑ Aht A1t A2 A3 AH
at = kt = = + t + t + ... + t
Nt Nt Nt Nt Nt
A1t Nt1 2 2
At N t 3
At Nt3 AH
t Nt
H
= + + + ... + .
Nt1 Nt 2
Nt Nt Nt3 Nt NtH Nt
Noting that population is equally divided across all households, we
get:
1
at = kt = a1 + at2 + at3 + ... + atH . (5)
H t
Likewise, aggregating over all households, we can write the average
consumption in the economy as:
∑ Cth 1
ct = = c 1 + ct2 + ct3 + ... + ctH . (6)
Nt H t
Jan 22-28-29 & Feb 2-9-13 &18, 2015 77
Das (Delhi School of Economics) Dynamic Macro / 86
Decentralized R-C-K Model with Heterogenous Households
(Contd.):

Di¤erentiating (5) with respect to t:

dkt dat 1 dat1 da2 da3 daH

= = + t + t + ... + t .
dt dt H dt dt dt dt

Then using (2):

dkt dat 1 dah

= = ∑ t
dt dt H dt
1 h i
H∑
h h
= wt + (rt δ n ) at ct
1 1
= wt + (rt δ n) ∑ ath
H∑ t
ch
H
= wt + (rt δ n) at ct

Jan 22-28-29 & Feb 2-9-13 &18, 2015 78

Das (Delhi School of Economics) Dynamic Macro / 86
Decentralized R-C-K Model with Heterogenous Households
(Contd.):
Once again, recognising that in this competitive market economy:
wt = f (kt ) kt f 0 (kt ) and r̂t = rt δ = f 0 (kt ) δ, we get dynamics
of average capital stock for the aggregate economy with heterogenous
households as:
dkt
= f (kt ) (δ + n) kt ct .
dt
However aggregating for the average consumption is not that easy.
Di¤erentiating (7) with respect to t:
dct 1 dct1 dc 2 dc 3 dc H
= + t + t + ... + t .
dt H dt dt dt dt
Then using (1):
dct 1 dcth 1 cth
dt
=
H ∑ dt
=
H ∑ σ (cth )
[rt δ n ρ] .

Jan 22-28-29 & Feb 2-9-13 &18, 2015 79

Das (Delhi School of Economics) Dynamic Macro / 86
Decentralized R-C-K Model with Heterogenous Households
(Contd.):
Notice however that if σ(cth ) is not a constant, then consumption of
di¤erent households will grow at di¤erent rates and therefore
aggregation for the entire economy becomes an issue.
The dynamics that hold for a household may not hold for the
economy-wide average.
This aggregation problem can however be avoided if the utility
function is of CRRA variety, so that σ is a constant.
In this case, along the optimal path, the rate of growth of
consumption for all households would be the same:
1 dcth 1
h
= [ rt δ n ρ ]
ct dt σ
and therefore so would be the rate of growth of average consumption:
1 dct 1
= [rt δ n ρ]
ct dt σ
Jan 22-28-29 & Feb 2-9-13 &18, 2015 80
Das (Delhi School of Economics) Dynamic Macro / 86
Decentralized R-C-K Model with Heterogenous Households
(Contd.):

Once again substituting for rt :

1 dct 1
= [f (kt ) δ n ρ]
ct dt σ
Thus the equivalence between the centralized and decentralized
solution prevails despite households being heterogenous in terms of
initial asset holding (provided of course their utility is of CRRA
variety)..

Jan 22-28-29 & Feb 2-9-13 &18, 2015 81

Das (Delhi School of Economics) Dynamic Macro / 86
Decentralized R-C-K Model with Heterogenous Households
(Contd.):

Finally, when households have di¤erent initial asset holding, we have

just seen that their rate of growth of consumption would be the same.
But how about the level of consumption?
Here the inital asset holding makes di¤erence.
In fact, the initially rich households will always maintain a higher level
of consumption than the initially poor households and the initial level
di¤erence will perpetuate in the long run.
To see this, note that for any household h :

dcth cth
= [rt δ n ρ]
dt σ
Rt (r v δ n ρ)
dv
cth c0h
σ
) = exp 0 .

Jan 22-28-29 & Feb 2-9-13 &18, 2015 82

Das (Delhi School of Economics) Dynamic Macro / 86
Decentralized R-C-K Model with Heterogenous Households
(Contd.):

Now, we have see earlier (from the household’s budget constraint and
the NPG condition) that:

Z∞ Rt
(r̂v n )dv Z∞ Rt
(r̂v n )dv
cth exp 0 dt = wt exp 0 dt + a0h
0 0
Z∞ Rt
(r̂v n )dv
) cth exp 0 = Ŵ0 + a0h
0

Plugging the solution for cth in the RHS above:

Z∞ Rt h (rv δ n ρ)
i
(r̂v n ) dv
c0h exp 0 = Ŵ0 + a0h
σ

Jan 22-28-29 & Feb 2-9-13 &18, 2015 83

Das (Delhi School of Economics) Dynamic Macro / 86
Decentralized R-C-K Model with Heterogenous Households
(Contd.):

Simplifying:

Z∞ Rt h (rv δ n ρ)
i
Ŵ0 + a0h (r̂v n ) dv
c0h
σ
= , where R0 exp 0 .
R0
0

Notice that Ŵ0 and R0 are the same for all households, but a0h are
not.
Thus a rich household will enjoy a higher level initial consumption
than a relatively poor households.
But the rate of growth of consumption for all households is the same.
This implies that the initial consumption di¤erence between the rich
and the poor will persist in the long run.

Jan 22-28-29 & Feb 2-9-13 &18, 2015 84

Das (Delhi School of Economics) Dynamic Macro / 86
Equivalence between a Planned Economy & the Market
Economy under R-C-K Model: Implications
The equivalence of outcomes between the socially planned economy
and the competitive market economy is a very strong result.
It implies that the actions of ‘atomistic’agents acting in their
individual spheres result in an outcome which is exactly identical to
that of the omniscient, omnipotent social planner (or governement).
Thus there is no logical scope for government intervention here -
either in terms of in‡uencing the long run growth rate (which it
cannot a¤ect anyway) or in terms of improving e¢ ciency (which is
super‡uous, because now the market economy is already e¢ cient)!
Note however that this equivalence result depends crucially on
the assumption that households have perfect foresight/rational
expectations.
Without this assumption, the equivalence result breaks down; the
optimal trajectory chosen by the market economy will no longer be
the socially optimal one.
Jan 22-28-29 & Feb 2-9-13 &18, 2015 85
Das (Delhi School of Economics) Dynamic Macro / 86
R-C-K Model: References

Reference for the R-C-K Model (Centralized & De-centralized

versions):
Barro & Sala-i-Martin, Economic Growth (2nd Edition), Chapter 2
Reference for Dynamic Optimization Technique in Continuous
Time (for students who are interested in diggging deeper in terms of
technique):
A.C. Chiang: Elements of Dynamic Optimization, Chapters 7,8 & 9
M. Kamien & N. Schwartz: Dynamic Optimization, Part II, Sections
1-9.
(The second book is more rigorous, but also more terse. Consult at
your own risk!)

Jan 22-28-29 & Feb 2-9-13 &18, 2015 86

Das (Delhi School of Economics) Dynamic Macro / 86