UNIT8 THE MSEY MODEL

Structure
Objectives
Introduction
Planner's Problem
Decentralized Households' Problem
Governmentin Ramsey Model and Ricardian Equivalence
Let Us Sum Up
Key Words
Some Usefhl Books
AnswerMints to Check Your Progress Exercises
8.9 Mathematical Appendix

8.0 OBJECTIVES
After going through this Unit you should be able to:
explain the Ramsey problem of optimal growth;
compare the optimal growth problem of the central planner and that of the
decentralized perfectly competitive economy;
0 explain the role of government in the optimal growth framework and the
correspondingconcept of Ricardian Equivalence;and
if you gothroughthe MathematicalAppendix cmfully you should be in a position
to solve any standard dynamic optimization exercise.

8.1 INTRODUCTION
In our discussion of household's intertemporal consumption and saving decisions in
the previous unit have, we consideredoptimization problems where the time horizon
was fuzite. While this is a valid assumption for an individual,a household (or for that
matter, the society as a whole) is generally infinitely lived. In other words, one can
think of a household as consisting of individual members, each of whom are fulitely
lived, but new members are born in each time period who are an exact replica ofthe
older members - so that the household as a dynasty continues to live forever.

An important economic question that is often asked in the context of an infinitelylived

household is the following: how should the household allocate its resources between
consumption and savings in every period so that the utility of each of its member are
- .
maxlrmzed?In other words, wfiat are the optimal consumptionpath and optimal savings
path forthis household?To put it more generally, in an economy consistingof identical
hfmitelylivedhousehold, whaf should be the optimal savingslaccumulationpatternthat
maximizesthe utility ofeach of dynastic household? This latter question was fmt tackled
by Frank Ramsey (1928), which was later generalized by Cass (1 965) and Koopmm
(1 965). The model has subsequently been known as the Ramsey-Cass-Koopmans
optimal growth model.
 The Rarnsey-Cass-Koopmansoptimal growth model essentially involves a dynamic The Ramsey ModBl
optimizationexercise whereby an agent maximizes its objective function defined over
aperiod of time (finite or infinite) subject to some constraints. The constmintsare also
d h i c in nature in the sense that they relate to values of the variables for different
time periods. The technique for solving such dynamicoptimization exercises lxisbeen
specified in the Mathematical Appendix at the end of the Unit. In the text w e shall
simply state the conditionsfor optimization(withoutgetting intothe underlyingtechnique)
and explain the ecdnomic intuition behind each of these conditions.
I

8.2 PLANNER'S PROBLEM

In the original Rarnsey formulation, the optimal growth problem was postulated as the
problem of a social planner who seeks to maximize the welfare of each household
subject to the economy's resource constraint at each point of time. The households
are all assumed to be identical in terms of preferences and their welfare is represented

by an infinite horizon utility fimction of the form = ["(cI ) ~ X P - ~dt' ,where c, is

0

the per capita consurnptwn in period 1, u(c,) is the associated instantaneous utility
in period t, and p is positive constant term which represents the subjective discount
rate of the household, or its rate oftimepreference. Before we proceed further, it is
important to explain the presence of the discount factor in the household's welfare
function. The positive discount factor reflects the household's preference for present
over future. In otha words, it impliesthat a household puts more weight on consumption
today vis-a-vis consumption tomorrow. Note that when t = 0 (i.e., the initial time
period) exp-P' = 1 . Subsequently when t = 1 , exp-P' = exp-P; when
t = 2 , exp-P' = exp-2pand so on, such that 1 > exp-P > exp-2~..... ... Therefore
+

in the infinite-tie utility function W, current utility at period zero is associated with the
highest weight (unity), and each subsequent utility term is associated with lower and
lower weights.

The social planner maximizes the integral W, but he is constrained by the fact that at
;r each point oftime,the economy's total consumption and total investment cannot exceed
dK
1 its total output.' To put it formally, c,+ -
dt
= ,where C, is aggregateconsumption

I dK
in period 1; - is the amount of investment in period I which augments the capital
dt
I

stock (K)of the economy and Y, is the total output produced in period 1.

Total population at time t is given by L, , which is assumed to grow at a positive

1 dL
exogenous rate n: - -= n .
L dl

Output at any point of time is produced using the existing capital and labour at that
point of time. Technology is represented by a neoclassical production fbnction
Y, = F ( K ,, L,) ,where Fis continuous, concave and exhibits constant returns to scale

' We rule out international borrowing here.

Intertemporal (CRS). We have explained the neoclassical production function in Unit 3, while
Decision-Ma king discussing Solow growth model. The CRS property ofthe production fiznction implies
that per capita output (y) can be written as the function of the capital-labour ratio (k)
in the following way:

Moreover the m a r d products of capital and labour can also written as the following
functions of the capital-labour ratio:

a~ a~
-= f'(k) ; -= f (k)- W ' ( k ).
aK aL

The continuity and concavity propertiesof F(L, K) ensure that f ( k ) is also continuous
and concave. Additionally we assume that f( 0 ) = 0 ; f ' ( 0 )= w, ;f '(a)= 0 . The
first of these assumptions implies that no production is possible with zero capital-
labour ratio. The last two assumptions are known as the Inada conditions which state
that when the capital-labour ratio tends to zero, the marginal product of capital tends,
to infinity, while an infinitely high capital-labour ratio implies that the corresponding
marginal product of capital approaches zero.

dK
The aggregate resource constraint for the economy is, C, + -= Y( . Dividing both
dt

sides by L, we can write it as 2C +

1dK
=
Y, dk
L,,i.e.,c, + -dr+ n k , = . f ( k , )where
the last equation denotes the resource constraint faced by the social planner in per
capita terms.

The economy starts with a given amount of capital stock and population; hence the
initial capital-labour ratio is given, denoted by k, . At every point of time the existing
capital and labour stocks are fully employed. Hence the capital-labour ratio k also
denotes the per capitacapital stock. While referring to k we shall use these two terms
interchangeably.

The complete dynamic optimization problem for the social planner can now be written
in terms ofthe two time dependent variables per capita consumption (c, ) and per
I
,
'
I

capita capital stock ( k, ):

M- W= F(c,)exp-''
0
dk
d t subjectto - = f (k,) - nk, - c,;k, given.
dt 1
>
i I
The correspondingHamiltonian function and the first order conditions in tern~softhe 1 ,

I I
See the Mathematical Appendix for the definitions of the Hamiltonian function, control,
state and co-state variables in a dynamic optimization problem. j !
I ;
iI
 control variable ( c, ), state variable ( k, ) and co-state variable (A, ) are given by:3 The Rarnsey dbael

H = ~ ( c , ) e x p - ~A,' +b ( k , ) - nk, - c,]

ia) ul(c,)e ~ p - =~ A,'

dA
iia) - = -A, (.f '(k, ) - n)
dt

Notice that A, is the shadow price of capital (as explained in the Appendix) or the
value of capital in utility terms. Hence condition (ia) states that along the optimal path
the present discounted value of marginal utility from consumption would be equal to
the shadow price ofcapital. The economic significanceof this condition would become
clear if you note that at any point of time one unit of output can be put to two different
uses - one can either consume it or invest it which augments the capital stock. Now
optimality condition requires that the returns in terms of utility from these two uses
should be equal, which is preciselywhat condition (ia) states. Conditions(iia) and (iiia)
respectively show how the co-state and the state variables change over time along the
optimal path. The fourth condition, known as the transversality condition, implies
that as the economy approachesits terminal time (which in this case is infmity), either
the'value of capital goes to zero (in which case it does not matter for the economy if it
leaves a I;ositive capital stock at the end), or, ifthe value ofthe capitai stock is positive,
then the economy uses up all its capital stock (i.e., kgoes to zero).

ith respect to t and using (iia) to eliminate A, ,we get the folowing
Differentiating(ia) w

differential equation: dI df '(kt - n - P ) . Noting that the

- cuW(c)-
term = * denotes the elasticity of marginal utility with respect to consumption,
u'(c>
the above equation can be written as:

Equations (iiia) and (va) together form a system of differential equations which, along
with the transversalitycondition, determine the movements of per capita consumption
and per capita capital stock of the economy along the optimal path. The qualitative
characterizationof this path is shown in the phase diagram given at Fig. 8.1.

dc dk
The phase diagram traces the litledcurvesalong which - = 0 and - = 0. The point
dt dt
of intersection of these two lines is called the steady slale.A steady state is equivalent
to long run equilibrium whereby the values of the variable remain constant overtime.
Intertemporal
dc
Decision-Making From (va),- = 0 implies eitherc = 0 or f '(k) = n + p . On the other hand, from
dt
dk
(iiia),- = 0 implies c = f (k) - nk . Plotting all these lines and curves in the c-k
dt
plane with c along the vertical axis and kalong the horizontal axis, we get a diagram as
shown in Fig. 8.1.

Fig. 8.1: Phase Diagram

The direction of arrows in the phase diagram shows the direction of movements of c
dc dk
and k when they are not on the- = 0 and - = 0 lines respectively. From Fig. 8.1 it
dt dt
is clear that there exits one steady state, denoted by point E in the diagram, which is
associated with positive values of c and k. It can be shown (though we make no
attempt to prove it here) that this steady state point is a 'saddle point' such that there
exits a unique path which approaches the steady state point; all the other paths move
away tiom the steady state point. This unique path is denoted by the dotted line in the
diagram.This is the only path which satisfies all the first order conditionsof optimality
(conditions (ia)-(iva)given above) including the transversalitycondition.This path is
therefore the optimal path which satisfies the social planner's dynamic optimization
problem. Given any initial capital-labourratio, the social planner will choose the initial
per capitaconsumptionso as to be on this optimal path, which will eventually take the
economy to its long runequilibrium (or steady state) point E.

At this juncture it is important to point out certain important features of the steady state
point E. As was mentioned before, at the steady state both the per capita capital
stock, k, and per capita consumption, c, are constant. Also, at this non-zero steady
state, the k-value and the corresponding c-value are defined respectively by the
equations: f'(k) = n + p and c = f(k) - nk . The condition that f '(k) = n + p is
called the modijledgolden rule. It states that the steady state value of the capital-
labour ratio is such that the marginal product of capital is eq* to the sum of the rate of
population growth and the rate of time preference. In any dynamic model of capital
accumulationand growth the concept of golden rule plays an important role (see Unit
3). Thegoldenrule itself refers to that steady state value of capital-labour ratio where
f'(k) = n , i.e., the rnarghlproduct of capital is exactly equal to the rate of growth of
I
population.The signiscam of tlae golden rule obtains h m the fact that it is that steady
I

1
1
state value of capital-labour ratio which corresponds to the maximum steady state MA+
value ofper capita consumption. Note that when households' rate oftime preference
is zero, the modified golden rule coincides with the golden rule and the steady state of
the Rarnsey model is attained at the golden rule point. However, when people have
positive rate of time preference, they are unwilling to sacrificecurrent consumptionfor
future consumption beyond a point; as a result they accumulate less and reach a lower
level of steady state consumption than in the golden rule.

8.3 DECENTRALISED HOUSEHOLDS' PROBLEM

The Rarnsey model described above can be easily transformed into the problem of a
decentralized competitive market economy consistingof many households -all with
identical preferences. We shall see that under certain assumptionsthe'optimal solution
path in the two cases turns out to be identical.

Instead of assumingthe existenceof a central planner, let us consider amarket economy

where the households are price takers. There are two competitive factor markets
which determinethe wage rate ( w,) and the inte~st rate or the rate of returnon capital
( r,) at each point of time.

the^ exist many identical competitivefirms whichhire in labour and capital h mthe
households at the above mentioned wage rate and rental rate. Using labourand capital
these firms produce the final output, using the same technology as the central planner.
Competition ensures that the wage rate andthe rental rate are equal to the respective
marginal products of labour and capital at full employment, i.e., r, = f '(k,) and

0 0 .

Each household maximizesits welfare givenas before by = J~(c

)t ex^"' dt. Esch
0

household starts with a certain amount of capital stock and labour stock. Additionally
householdscan borrow fiom one another. Let a, = k, - b, denote the per capita asset
stock of anhousehold at period t,which consists of the per capita capital stock owned
by the household at period t minus the amount of debt (per capita) at period t. (If the
household is a net lender, then b, would be negative).Arbitrage condition in the asset
market ensures that capital and lending earns the same rate of return. Hence
the per capita income of the household at period t is w, + r,a,,which the household
spends on consumption and fiuther asset accumulation. Thus the budget constraint
da
faced by the household in per capita terms is given by: Ct + -+ flat = W, + Cad .
dt
The completedynamic optimization problem for the household can now be written in
terms of the two time dependent variables per capita consumption( c,) and per capita
asset stock (a,):

W=
~awimize IU(C,)exp-"
o
da
subject to - = W , + (q - n)a, - C,;a, given.
dt 23
r-
Int./ertemporal The corresponding Hamiltonianfunctionand the first order conditions in tenns of @e
Fdecision-Making
control variable ( c, ), state variable (a, ) and co-state variable( il,) are given by:

ib) ur(c,) e ~ p -=~ A,

'

Note that these conditions look quite similarto the first order conditions that we obtained
for the central planner's problem (conditions (ia)-(iva) in the previoussection).However,
for the household's problemthereis an additional condition which has to be satisfied if
we want to get ameaningfhlsolution. This condition is called the No-Ponzi-Game
condition which we elaborate below. First note that we have allowed for intra-
household bo~~0wj.n~ which mea& that a householdcanmaintain a consumptionstream
above its income by borrowing. Now if a household could go on borrowing indefinitely
then it will be optimal for the hausehold to alwaysmaintain an idnitely high(or rmxhum
possible)consumption stream and finance such high level of consumption simply by
borrowing more and more. Of course, such a strategy would also imply that the net
per capitabornwingof the household would inc- exponentiallyat the rate( r, - n)
(since the household kcts to borrow not only for consumption, but also to pay back the
interest rate as well) and the present discounted value of the net debt of the household
will approach infinity.Such a financing scheme is called Ponzi-Gamefinancing4. In
order to rule o~lksuch idbite indebtedness by any family we specifLthe condition that
the even thqugh the household can be temporarily a net debtor (i.e., the present
discounfBCI~value of its asset is temporarilynegative), over a sufficientlylonger horizon,
it must eventually repay all its debt and hold non-negative asset stock. Formally, as t
goes to infinity, the present discounted value of the (per capita) asset stock of the
household must be non-negative:
I
-J(rr-n)dr
vb) limq exp
1-0
" 2 0.

condition (vb) is called the No-Ponzi-Gamecondition. For a decentralizedhousehold

conditions (ib)-(vb) specify the optimal path the solves the household's dynamic
optimizationproblem.

As before, differentiating(ib) with respect to t and using (iib) to eliminate A, , and 1

- cuR(c)-
noting that the term U,(C) = 0 denotes the elasticity of marginal utility with respect
..
Named afterCharles Ponzi (1 877-1 949) who raised considerable amount of money promising a
high rate of interest (50% for 45 days, 100% for 90 days). As long as new lenders were
attracted to these returns, he was able to repay previous debt. In eight months he ended up
with 10 milliondollars of certificates and 14 million of debt. Criminal charges were fulally levied
on him in the US and died a pauper.