Professional Documents
Culture Documents
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Wiley and Royal Economic Society are collaborating with JSTOR to digitize, preserve and extend access to The
Economic Journal.
http://www.jstor.org
MACROECONOMIC
MODELS
RATIONING*
QUANTITY
WITH
I. INTRODUCTION
Our objective here is to clarify the structure and behaviour of a simple macroeconomic model with rationing, emphasising the interaction between the
markets for goods and labour and the role of quantity signals when markets
do not clear. Many features of the model are common to a wide range of recent
contributions to the growing "disequilibrium macro" literature. The major
novelties here are our explicit treatment of inventories, our "double wedge"
diagrammatic apparatus, and our comparative statics results. The buffer stock
role of inventories gives a valuable insight into the nature of quantity signals
in the goods market. We emphasise the intertemporal character of decisions
and the importance of expectations about future states of markets.
Our plan is as follows: Section 2 gives definitions, assumptions, and a simple
introduction to the basic macro model. Section 3 lays out the micro models of
the household and firm, and Section 4 discusses different types of quantityconstrained macroeconomic equilibria which these generate. Section 5 deals
with comparative statics and dynamic problems. Section 6 suggests some further
extensions.
Even without a market for financial assets, the structure of this class of
models is in some respects more complex than that of IS-LM. If it can be
made equally familiar and easy to manipulate, however, it might be more
likely to serve as an alternative framework. Once the basic principles of the
interaction of markets in non-Walrasian equilibria are understood, one is on
more secure ground when turning to the fascinating but difficult problems
raised by the recent reappraisal of Keynes.
It was this reappraisal, associated initially with Clower (I965) and Leijonhufvud (I968), which began the development of a general equilibrium macroeconomics without market clearing.' Their stress on Keynes's conception of
unemployment as a chronic state, emergence from which is at best slow and
imperfect. reauired the denial of Walras's Law. If the values of excess demands
* We owe to our colleague Bertie Hines our original interest in the reappraisal of Keynes and macroeconomic models with quantity rationing (see Hines, I97I). We have benefited from discussions with
him and with Volker Bohm, Dwight Jaffee, Richard Quandt, and David Winter. The Institute for
International Economic Studies of the University of Stockholm provided Portes with a stimulating
environment and support in September I976 during the writing of a first version of this paper, which
Muellbauer benefited from
arose out of Portes's project supported by SSRC (UK) grant HR 3309/I.
participation in Jacques Dreze's seminars at CORE in Spring I976 and I977. Jean-Pascal Benassy,
John Black, Bill Branson, Angus Deaton, Avinash Dixit, Miles Fleming, John Flemming, Bertie Hines,
Takatoshi Ito, Gerald Kennally, and Lars Svensson offered helpful comments on various versions of
the paper; all are exempt from responsibility for remaining flaws.
1 There are of course further antecedents, aside from Keynes himself. Especially important, midway
in time between Keynes and Clower, are Hansen (I95I), which laid the groundwork for the attack
on Walras's Law, and Patinkin (I956), which analysed the effect on the labour market of excess supply
in the goods market.
[ 788 ]
[DECEMBER
1978]
MODELS
MACROECONOMIC
WITH
789
RATIONING
do sum to zero, so that negative excess demand in one market must have a
counterpart in positive excess demand in some other market, there will be a
tendency for relative price adjustment in the direction of eliminating the excess
supplies and demands. If there is a Walrasian auctioneer who knows all the
"notional" (i.e. determined only by initial endowments and prices) excess
demands and can change the prices accordingly before any transactions are
executed, this adjustment might well take place (with some further strong
).
Ecs 88
27
THE
790
ECONOMIC
JOURNAL
[DECEMBER
auctioneer, no tdtonnementin prices or recontracting, and non-competitive conditions facing agents - but they do all represent an interrelated system of
markets in which actions are made consistent within the unit period. Their
general equilibrium character may be a corrective to the regrettable tendency
among critics of "neoclassical orthodoxy" to identify general equilibrium with
competitive equilibrium or the specific Arrow-Debreu formalisation and dismiss it as hopelessly unrealistic. Such confusion discredits more justifiable
criticism, while ignoring attempts to adapt the general equilibrium framework
so it can deal with real economies.
A major part of the difficulty of models with quantity rationing is that one
is so accustomed to assuming that markets do clear through price adjustment.
When they do not, one has to juggle different combinations of excess supplies
and demands, and this becomes rather complicated. Our "double wedge"
diagram in Section 4, whose basis is derived in Section 3, makes this juggling
quite simple.
Our intention throughout is to give an analytically manageable macro
model with quantity rationing, which can be used in various ways: for exposition; to analyse standard theoretical macroeconomic propositions (the effects
(I977)
;1
as
2.
DEFINITIONS
AND
ASSUMPTIONS
See Portes (2976) for an application to centrally planned economies. Fair (I974) takes some important initial steps towards allowing quantity constraints in a large macro model.
2
I978]
MACROECONOMIC
MODELS
WITH
QUANTITY
RATIONING
79I
strained, then id
ld(.);
if he is, then Id
ld(.).2
(I976).
27-2
792
THE
ECONOMIC
JOURNAL
[DECEMBER
(I975) and Malinvaud (I977) stress, these conditions will not determine an
equilibrium uniquely. We require in addition a (typically "informal")
rationing scheme for each market, specifying how the commodity is to be
distributed among those on the long side of the market. Benassy's formulation
is quite general, whereas Malinvaud chooses particular rationing schemes for
concreteness (Grandmont, I977a, surveys this question).
We discuss rationing schemes and aggregation in Section 4. A quantityconstrained equilibrium is a state in which each agent's realised transactions
maximise his objective function (utility, profit, or whatever): they are for the
agent the best he can attain, subject to all constraints facing him. A formal
definition is provided by Benassy (I975). We suppose that agents face given
prices in the unit period, but prices can change between periods; this is the
"fix-price" method (see Grandmont, I977b). At any instant, agents view the
quantity "rations" they face as exogenous parameters.' Thus the consistency
of all agents' transactions in a quantity-constrained equilibrium is achieved
through the adjustment of the quantity "rations" without any price changes.
We deal with a capitalist market economy. There are three types of decisionmakers: households, firms, and the government. The government enters the
model through fixing the volume of real government expenditure, which is a
prior claim on output (the government is never rationed2). Government expenditure is entirely exogenous and is taken to include the fixed investment
demand of the private sector. It is assumed to be unproductive in the short run,
although in the longer run the accumulation of fixed assets will shift the production function.
There are three commodities: money, output (goods), and labour, each
homogeneous. Thus there are two markets. Because production is in reality
specialised, firms do not pay workers in output which they consume, so exchange takes place through money. As Hahn (I977a) stresses, it is not the
presence of money in itself which under-lies the failure of markets to clear in
these models, but rather the impossibility of exchanging labour directly for
consumption goods. Money is fiat money. The stock of money is exogenous,
and its total increment during the period is simply the volume of government
expenditure (including fixed investment) during the period; there is no taxation. The excess demand for money is purely an asset demand, unaffected by
transactions requirements (or interest rate uncertainty, since there is no return
on holding money balances), and the corresponding excess demand function
is therefore best viewed as a savings function (see Section 3). For firms, the
analogous asset is inventories. Firms do accumulate money (undistributed
profits) during the period but pay it all out to households as dividends at the
1 As Arrow (I959)
pointed out, an agent in a market with excess supply or demand must perceive
a demand or supply curve which cannot be unbounded and horizontal, and the fix-price assumption
is dropped in recent papers by Hahn (I978) and Negishi (1976).
2 In reality, of course, the expenditure plans of government and investors may also be partly frustrated if there is excess demand for goods. The simplifying assumption here is not essential, but it
facilitates the analysis of effects of changes in government expenditure and allows us to avoid discussing
its determinants. Even when Fair (1974) introduces goods rationing into his much more complex
model, he confines its impact to the household sector so as to simplify consideration of the labour
supply effect, which is our primary concern here.
I978]
MODELS
MACROECONOMIC
WITH
QUANTITY
RATIONING
793
beginning of the next period. Neither households nor firms may borrow.
Clearly, this simple model will omit many of the interactions between shortand long-run expectations, liquidity preference and investment decisions,
which have been such a major theme in the reappraisal of Keynes (see Leijon-
I97I).
Labour services are the only variable input. Households supply labour,
demand goods and money balances, and have stable preferences represented
by a utility function which they maximise. They may receive profits, but they
take all non-labour incomes to be exogenous, independent in the slhort run of
labour services sold. Firms demand labour and inventories, supply goods, and
maximise profit or a more general utility function.
We use the following notation:
Yt
Xt
it
Ct
gt
it
Tt- It
mt
Wt
Pt
St
iTt
dt
I)
xt + Ait,
()
+gt)
(2)
St -dt +wtlt-ptct
Tt
=Ptxt-
- Amt,
wt It,
(4)
ptxt-ptct
(3)
Am+Amf,
Irt = dt+?.
(5)
(6)
(7)
The household savings flow (income less consumption) equals the increment
to household money balances. Firms' profits this period (sales less wages) less
dividends (last period's profits) equal the increment to firms' money balances.
Government expenditure equals output less consumption and inventory accumulation, or profits plus savings out of labour income, i.e. the total increment
794
THE
ECONOMIC
JOURNAL
[DECEMBER
c = min (Cd,cS),
(8)
(9)
The real complications arise because the effective demand and supply in
each market depend on what is happening in the other market. Indeed, because
the functionalform of a quantity-constrained demand differs from that of the
corresponding notional (entirely price-determined) demand, if excess demand or supply appears in one market and either buyers or sellers there become rationed, we must switch from the unconstrained to the constrained
functional form to represent their behaviour on the other market. Thus not
only does effective demand on the goods market depend on the quantity traded
in the labour market: the effective demand function for goods depends on
whether there is excess supply or demand for labour. If there is non-negative
excess demand for labour, then the effective demand for goods is the notional,
unconstrained demand, so the demand function for goods has only prices and
initial assets as arguments: cd-- cd(.). But if there is excess supply of labour,
then we switch to the quantity-constrained effective demand for goods, one
argument of which is the quantity of labour actually sold: cd =d(l;
. ). This
switching between functional forms occurs not only for the demand for goods,
but also for the supply, and for the demand for and supply of labour. The
are indeed the essence of the Clower-Patinkin "dual decision
switchingconditions
hypothesis".
Thus, disregarding the borderline cases of market clearing, we have four
possible constraint regimes (the labels K, R, C and U will be explained in
Section 4):
(K) cd <
cS,
Id < IS,
so that
C = Cd = ed(l;
cs =
cS(
I =
Is =
),
ld(C;
(IO)
is(.)
(R) cd >
Cs, Id
C = cs = es(l;
Cd
Cd (.
)
))
= Is = 1S(C;)
Id
(I)
Id()
I978]
(C)
MACROECONOMIC
cd > C8,
Id <
ls,
MODELS
so that
c = c
cd
= cs(.),
=
jd
Y;
795
RATIONING
QUANTITY
WITH
Id =
(I 2)
Id(.)
=
IS(C;*)
(U)
cd < cs,
Id
I = Is = Is)
C = Cd = Cd(.),
cs
.)
j(l;
Id=
(I 3)
d(C; . )
c+g
yC
Fig.
I.
~ ~
~ ~~~~~~~~YMl-g
C
The intersection of the 450 line and the total expenditure function determines
equilibrium output. Given the interaction between the goods and labour
maikets, there is a story about the labour market implicit in this diagram. The
first step in making the market interaction explicit is to translate the axes of
the diagram into consumption-labour space, as in Fig. 2. Now equilibrium is
determined by the intersection of the aggregate supply curve for consumption
goodls (the production function shifted downwards by the amount of
government demand - recall that the government is never rationed) and the
aggregate demand curve for consumption goods. This shows consumption
expenditure as a function of labour services sold. In Section 4 we elucidate the
using a
relations between goods and labour markets expressed by (IO)-(I3),
suitably extended and reinterpreted version of Fig. 2. First, however, we must
provide a microeconomic basis for the supply and demand functions of firms
and households in alternative constraint regimes.
796
3.
AN
JOURNAL
ECONOMIC
THE
VIEW
INTERTEMPORAL
OF
THE
HOUSEHOLD
[DECEMBER
AND
FIRM
For the household we introduce money balances, and for the firm, inventories
of finished goods; each has the function of a store of value. We shall make
explicit- more so than has been done by Barroand Grossman(I97I, I976),
Clower (I965), Malinvaud (I977) - that this role of a store of value is derived
from the future part of the objective functions of households and of firms.
Therefore the circumstancesexpected to prevail in the future can make this
role qualitatively as well as quantitatively different. For example, a household
which expects that it cannot sell labour next period clearly has a quite different
need for money balances than one which expects to be rationed in the goods
market. Analogously, a firm that expects not to be able to obtain all the labour
it would like next period has a quite different need for inventories of finished
goods than one that expects not to be able to sell all it would like next period.
Households have in reality a variety of stores of value by which to transfer
purchasing power from one period to the next. In wrapping them up in money
balances we are giving a stylised version of a more general theory of savings
behaviour. Similarly, firms can transfer purchasing or selling power through
time not only through inventories but money balances, other financial assets
and fixed capital. By considering only inventories we are nevertheless giving
the core element of a more general theory of the financial and investment
policy of firms.
For both households and firms we shall discuss a two-period model which,
given the recursive structure of intertemporal problems, extends to many
periods with no loss of generality.
(a) Households
We begin with a two-period model of a household which has expectations held
with confidence about the second period. Let us suppose that the utility
function is defined over consumption ct and leisure Tt-It, in the two periods
t =
I, 2:
UH =
U(C1, T,-
l,
C2, T2-12),
(I4)
(I5)
where w2 is the expected wage rate, p2 the expected price level and d4is exogenous non-labour income (either dividend payments or a government transfer
payment).
The household could in principle expect to be in any of four alternative
states in period 2: it might be unrationed, so that (I5) alone would be the
entire constraint; it might be rationed in the labour market only, so that (I5)
is supplemented by 124= ; it might be rationed in the goods market only, so
that (I5) is supplemented by c2 = c2; finally, it might be rationed in both.
I978]
MACROECONOMIC
MODELS
WITH
QUANTITY
RATIONING
797
However, if nothing beyond period 2 matters, and given that rationing must
always be an upper bound on consumption purchases c2 and labour sold 12,
constraints on both cannot be simultaneously binding. If a household were
rationed in the goods market, so that c2(.) > c-2, and had no use for endperiod money balances, it would not make sense to supply more labour than
12 =
(P2j2-
-Md2)
/w2. If the labour ration were below this level, then the
goods ration ?j would not be binding, since as much as C2could not be bought.
This allows us to neglect the possibility of simultaneous rationing in period 2
in both labour and goods markets. But notice that in a model with three or
more periods this possibility could re-emerge, and as we shall see, it could
certainly arise in the first period.
Let us solve the household's problem of maximising utility for period 2,
conditional upon cl and 1,. In the unrationed case the optimal levels of consumption c2 and labour supply 12will depend upon (ml+ d2)/p2, w2/p2and c,
and T, - 11.Substituting these levels gives the solved utility function:
UH
(i 6)
The fact that ml/p2 enters (i6) indirectlythrough the optimal levels of c2 and
'2makes explicit that the utility of money balances ml is derivedutility.
Analogously, when only labour is rationed by 12 < 12 and this constraint is
binding because 12 < 12(.), the budget constraint implies
C2= (ml+ d2+ W212)/P2,
so that
UH = U[CD,7T-lID (ml+d2+W212)/p2, 7 -12]
(I7)
When only goods are rationed by c2 < c2, and this constraintis binding because
-d2) Iw2, as long as
c2 < cll(.), the budget constraint implies m2= (P2c2>
o. Then
12
2 -ml d2)/wi2
(I8)
UH = U[Cl, TX-11 j2 T-(P2
The three functions (I6), (I7) and (I8) are sometimes called conditional or
mixed indirect utility functions' and make plain the different role played by
money balances, given the three alternative states of the world which can be
expected.
We are interested primarily in analysing decisions in the current period,
and to have to worry about three alternative utility functions with arguments
cl, 11and ml does not make for analytical simplicity: if there are four distinct
constraint regimes in period I, then there are I2 possible behaviour regimes
to consider for period I! This is probably why various authors have preferred
to write simply
UH = V(c,1 T,-1
('9)
Ml/P1)P
1 Such utility functions have been considered by Samuelson (1969), Fama (1970),
Epstein (1975),
Benassy (1975), and a number of other writers. Basically, they stem from the recursive solution structure
of dynamic programming problems -see Bellman (1953) and the discussion in Grandmont (1977a,
section2).
798
THE
ECONOMIC
JOURNAL
[DECEMBER
However, this obscures the important role of expectations and assumes that
next period's price level is expected to be the same as this period's and that
the regime which actually prevails in period I does not affect expectations
about the regime in period 2.
The analytical simplicity of only one instead of three types of behaviour in
each of the four constraint regimes in period I can be obtained consistently in
two ways. In the first, one assumes that, for each regime in period I, there is
a unique expectation of which constraint regime will be operative in period 2.
Perhaps the most natural expectational assumption is that period 2 will have
the same constraint regime as period i. However, a second approach which
relies on an explicit, probabilistic Von Neumann-Morgenstern formulation is
rather more attractive. This involves the choice of a transformation of the
utility function whose expected value will be the new utility indicator to be
maximised. Because this will be an appropriately weighted average of (i6),
(I 7) and (i 8), the resulting utility function has as arguments cl, T1-11, mIn, T2
and the parameterswhich govern the (probabilistic) expectations of d2,P2, W2,
c2 and 12. If we absorb into the utility function the part of these parameters
which depends upon cl and 11,and let the rest of these parameters be represented by the vector 0, we can write
UH =
V(C1, T1-1l,
ml,
T2, 0),
(20)
= plcl+ml,
(2I)
into
(20),
to obtain
UH
T2,
6).
(22)
Having thus absorbed the constraint, draw a constant utility contour map in
consumption-labourspace as in Fig. 3. The point H is the optimal point when
no rationing is imposed and represents the highest attainable utility. The coordinates of H are the unrationed goods demand and labour supply. These
must be functions of the vector of exogenous arguments of V(.) in (22):
(T1,MOn,w, PT2,
p1,
), which we denote by AH. So we can write
c
C-
11-
(AH) = C
I (AH)
23)
I978]
MACROECONOMIC
MODELS
WITH
QUANTITY
RATIONING
799
where
GCD
(24)
'
lS(c)X
/1 =Ti;
Fig. 3. The household's labour-rationed consumption function
and goods-rationed labour supply function.
Similarly, by shifting the goods ration level cl <s cl, and tracing out points
of tangency with the indifference curves such as B, one obtains the goodsrationed labour supply function
where
(25)
Finally, if the goods ration is cl < cl and the labour ration is 11 < 11, the
household is at a point such as C., Such a point is quite feasible. If both constraints are binding the household is accumulating money balances in the
amount
(26)
ml - MO= W, 11-P1i C.
Given 11 and cl, the household accumulates money balances given by (26),
apparently as a residual. But note that these are certainly not without utility:
1 Note that the ration 1, enters d(. ) not only as part of the budget (mo+ w, 11) but also by influencing
expectations and (unless the utility function is separable in 1) by affecting the rate of substitution between other arguments of the utility function.
800
THE
JOURNAL
ECONOMIC
[DECEMBER
if they were, labour supply would be cut so that the labour ration would no
longer be effective. Formally,
11 <
11
C=
NA(H)
el)
lsb
A27
The functions (23), (24), (25) and (27) express the effectivedemands and
supplies in each of the four constraint regimes respectively. When rationing
constraints are binding in one market, the effective demand or supply expressed on the other market, instead of being the notionaldemand and supply
(23), becomes the rationed demand (24) or rationed supply (25). The labourrationed consumption function and the goods-rationed labour supply function
meet at H in Fig. 3. Thus, the value of the labour-rationed consumption function (24) coincides with that of the unrationed function from (23) when the
labour ration in it, 11,is replaced by the unrationed labour supply from (23).
Similarly, the value of the goods-rationedlabour supply function (25) coincides
with that of the unrationed function from (23) when the goods ration in it, cl,
is replaced by the unrationed goods demand from (23). This is a fundamental
property of conditional demand functions (see Pollak, I969).
This consistency condition and indeed the simple form in which the effects
of rationing are representedin Fig. 3 rest on a continuity assumptionconcerning
expectations. This is best explained by an example in which the consistency
condition is violated. Suppose that the same rationing regime is always expected in period 2 as in period I. With labour rationing in period I, the relevant
utility function is (I 7), where 1 may depend upon 11. Replacing 11 by the
level of unrationed supply 1' does not result in the unrationed function (I6),
as is easily illustrated with a Cobb-Douglas utility function. The reason is
that the way 12 enters is conditioned not by the level of 11 as such but by the
regime in period I, which determines that of period 2. Thus there is a discontinuity: a small change in 11 can lead to a switch in regime and hence a
complete switch in the functional form of the relevant utility function.' To
avoid it we need to assume that expectations of period 2's regime are determined
as continuous functions of cl, 11 and prices, etc., but not of period I's regime.
This is quite a special assumption, which we shall make throughout.
The distinction between the form of the consumption function in (23) and
(24) is what Clower (I965) calls the "dual decision hypothesis". If the labour
market clears, consumption does not depend upon current income from the
sale of labour services - this is "solved out" in the household utility maximisation. The Keynesian consumption function (24) operates when households
encounter a quantity constraint in the labour market.
As discussed in Section 5, existence, uniqueness and stability of the quantityconstrained equilibrium require (among other conditions) that the consumption function C-d(l) and the labour supply function ls(c) have positive and finite
1 This is analogous to a discontinuity as the regime changes which can arise in the effective demands.
These also are conditional not merely on the ration level in the other market but on the regime as well.
I978]
MACROECONOMIC
MODELS
WITH
QUANTITY
RATIONING
801
slopes, with the latter steeper than the former.1 Given goods rationing to
cl = cl, one would expect labour supply to be less than that amount of rationed labour sales which would give rise to the same amount of consumption
demand. With labour rationing, one would expect the household to be running down its money balances unless it anticipated even more severe unemployment next period. With goods rationing, one would expect it to be accumulating
money balances unless it anticipated even more severe goods rationing next
period.
Leland (i 968) and Levhari and Srinivasan (i 969) replace the TobinMarkowitz theory of the precautionary demand for money, based on interest
rate uncertainty, by the precautionary demand for saving based on income
uncertainty. We have sketched a further generalisation suggested by Benassy
(I975) and Grandmont and Laroque (I976), introducing uncertainty about
rationing levels explicitly in both labour and goods markets. This is one reason
why we do not write "demand for money" functions. If we did, then as suggested above, under goods rationing we might expect 9md/Ic < o, which
makes sense when thinking of money as an asset but not in its pure transactions
role.
(b) Firms
In many ways, our theory of the firm parallels that of the household, with the
demand for inventory holding analogous to the demand for money balances or
financial assets more generally. One way of representing inventory deterioration or holding costs is
it - ht(it1) +Yt -xt,
Profit in period
where
ht(it-l)
< it-l
(28)
is given by
7Tt=txt-Wtl(Yt)3
where l(yt) is the inverse of the production function, the labour requirements
function.
As with the household, two approaches are possible. In the first, expectations
of next period's circumstances are held with complete confidence. The firm
is assumed to maximise the sum of the two periods' profits 7rT+7T2. Assuming
that inventory holdings at the end of period 2 have no role in the firm's objectives, (29) gives the analogue of the household's period 2 budget constraint:
(29)
THE
802
ECONOMIC
[DECEMBER
JOURNAL
conditional profit functions for period 2, each a function of the level of inventory holding il. For analytical simplicity one could assume that given the
constraint regime of period I, there is a unique expectation of the period 2
constraint regime. For example, one could assume that the firm never expects
to be rationed in period 2. In a brief appendix, the period I behaviour which
results on this assumption, making specific assumptions about the form of
h(it) and l(yt), is worked out for illustrative purposes. But as for households,
the explicitly probabilistic approach is perhaps more attractive. Letf2(&T2) be
a concave function, so that the firm is risk averse, and make the objective
function the expectation of an additive function of iT and T2:
max
conditioning variables)]}).
(30)
max UF
subject to
= f1[pixi
il = hk(io) +y1-xx
+f2
WI I(yi)]
[ii, I(yi),
X1, 4]
(3I)
l1
xs(AF)
xl,]
i1.J
(AF) =
()
I978]
MACROECONOMIC
MODELS
WITH
QUADNTITY
RATIONING
803
If the amount of labour the firm can obtain is rationed but sales are not, the
labour-rationed supply function of the firm is
xi =
where
11 = 11 <
(34)
Id.J
Id (AF)
x1
OX,
/ (Xl)
/l
If the amount of sales the firm can make is rationed but labour is not, the sales
rationed labour demand function of the firm is
11 = I X(A, x =)
where
xi
x-
<
ld
xS(AF) = xl.J
Finally, the last possible constraint regime arises when the firm is rationed in
both markets. Then
1| = 1 < Id(A?, X1) = d1,l
(36)
and
x=
= Yi(ti)
-xi
804
THE
ECONOMIC
JOURNAL
[DECEMBER
represents the inventory accumulation of the firm. Although this looks like a
residual, the firm could in fact choose to produce less and accumulate fewer
inventories. If the firm did not believe that they could be sold profitably next
period, it would cut production and no longer find itself rationed in
labour.
We have not analysed the shut-down decision of the firm explicitly or the
possibility of corner solutions which can arise if one of the restrictionsil > o,
y, > ?, xl > o is binding. These can be understood to be included in the
definition of the various demand or supply functions above, which may therefore have kinks.
We have treated money balances and dividend payments as exogenous:
although they enter the accounts of the firm they do not enter into the decisions
of the firm which we have analysed. Implicitly, we have assumed that firms
always have sufficientretained or current profitsto be able to finance inventory
accumulation. While in principle one could write down an indirect utility
function for the firm which would include money balances and dividends, this
derived utility stems from transactions in financial markets and markets for
fixed investment which are outside the scope of our simple model.
Finally, it must be pointed out that saving (and particularly the money
component in saving) acts as a buffer stock, absorbing unanticipated shocks.
This role is in part already implicit in the theory: each period the household
replans in the light of new information and consequently takes correcting
action for past mistakes. The concept of equilibrium here can accommodate
such bufferstock behaviourwithin the period, as long as we take the equilibrium
to be reached at the end of the period, after a process of adjustment. Our
concept of equilibrium requires that the perceived ration coincides with the
actual ration so that neither buyers nor sellers have any incentive to change.
However, we can think of a sequence of very short periods in which the current
ration levels are interpreted as perceived levels which may start different
from the actual ration levels but converge to them by the end of the sequence.
This tdtonnement
on quantities is part of the familiar multiplier process.
Inventories play a similar buffer stock role in this process. If households
suddenly decide to increase their expenditure, the increase causes an unexpected drop in inventories as sellers meet the unexpected demand increase by
selling from inventories. If we interpret xl, for example, as the perceived sales
ration in the current period, the theory of the firm as formulated can accommodate an error in this perception: the firm will then have made an error in
its employment and inventory plans. Information on this error materialises
from the inventory levels which, if xl underestimated demand, will be unexpectedly low. In the next short period, plans will be made in the light of
actual inventory holdings and of revised views on the sales ration. Such a
process is likely to converge quickly to an equilibrium in which the perceived
sales ration is the true one. While such a process is working itself out in the
very short run, we can say that the economy is "in disequilibrium", and this
is really the only sense in which this class of models can be said to be about
disequilibrium. The information generating role of buffer stocks suggests that
I978]
MODELS
MACROECONOMIC
WITH
QUANTITY
RATIONING
805
models with quantity rationing do not, after all, require as much centralised
co-ordination as Grandmont (I977b), for example, argues.
4.
THE
COMPLETE
MODEL
8o6
THE
ECONOMIC
JOURNAL
[DECEMBER
ed(AH,'
if
1) if
xs(AF) -g
Xs(AF l)g
c
= min (Cd,
Is > ld
if
if
Is
1=
Id = I}
|s
(38)
<ld
(39)
Cs);
ltd(AF)
if
C =
cs <
Id (AF, C+g)
if
Cs >
Cd =
l=S(AH)
{S(AH,c)
1=
(37)
(37
1 = Is
if cs
if
Cd
C(
cd =c,
(4)
C = Cs < Cd,
(42)
1978]
MACROECONOMIC
MODELS
WITH
QUANTITY
RATIONING
w
-d
C
-d
la
cs
-----
xd
H
ed
/s
d~~~~~I-
C
_
__ C 6. Keynesian
unemployment.
Fig.
xd
cc
-d
I_Ce
S
Fig.6. Keynesicanxdunemployment.
unemployment.
Fig. 8./Classical
S
xd
d
Fig. 7. Rnepreossemifltion.
Fig. 9.
807
8o8
THE
ECONOMIC
JOURNAL
[DECEMBER
formed by
&d( .)
es(.)
and Id(.). As noted in Section 3, to be within the wedge means a double constraint on the relevant set of agents, and this occurs in C and U. This is an
important difference between R and C for households, and between K and U
for firms.
In R, households facing goods rationing adjust their labour supply and saving
behaviour. They move along 15(.), substituting leisure and future consumption
for the unobtainable present consumption. The "supply multiplier" process
(involving a fall in output as labour supply falls, and further adjustment by
households) is assumed to work itself out in the unit period, and in equilibrium
they are supplying the amount of labour and accumulating the amount of
assets which they desire, given the constraint facing them. Although there may
be "forced saving" in the sense that they are saving more than they would in
the absence of the constraint, this saving is desired to finance future consumption, given their expectations about future constraints in the goods market; if
it were not, they would reduce labour supply still furtber. In C, however, this
option is not open: they are already constrained to sell less labour services than
they would like, even taking into account the rationing in the goods market.
Given the level of employment, they cannot consume as much as they would
like and are accumulating some money balances which they would rather
spend this period than next but cannot. Note, however, that they would rather
have them for next period than work even less.
For firms, the contrast between K and U is analogous. In the former, they
can adjust employment downwards to reach their desired level of inventory
accumulation, given the sales constraint facing them (the "demand multiplier"
process is assumed to work itself out). In the latter, they cannot, but given the
labour supply available, they are selling less this period than they would like
(though it is nevertheless profitable to produce for sale next period).
1 Note also that in Fig. 6, there is no need for F to lie northwest of H; there could also be a Keynesian
equilibrium with F southwest or northeast of H (but not southeast - there cannot be excess effective
supply in both markets with excess notional demand in both markets). Similarly, we could have a
repressed inflation equilibrium with F southwest or northeast of H.
I978]
MACROECONOMIC
MODELS
WITH
QUANTITY
RATIONING
809
I97I;
Portes,
I976,
I977).
In so far as
they are, the endogenous forces for relative price adjustment to clear markets
may be very weak. Walras's Law applies for the notional demands, but it
need not hold for the effective demands in a quantity-constrained equilibrium.
In the simplest case, with g = Am = o and no inventory accumulation, the
sum of the excess effective demands is just W(ld -Is) +p(cd -Cs), since
Amh + Amf = Am = 0.
(.)=Is
I
( .)
I+
11/
IV
p
Fig. Io. Loci of notional equilibria in wage-price space.
Fig. IO shows loci of notional equilibria. Assuming the normal signs for the
relevant partial derivatives given in Section 5, it is straightforward to show
that dw/dp > o along each locus. The signs on either side of each locus indicate
regions of excess demand (+ ) and supply (-). To see why the curve Cd(.) =Cs
is steeper, note that this configuration makes region I have general excess
1 In the general case with g = Am > o, we reach the same conclusions from the version of Walras's
Law which does hold in quantity-constrained equilibrium: the sum of excess effective demand for
unrationed goods is zero (the only unrationed good here being money).
8io
THE
JOURNAL
ECONOMIC
[DECEMBER
supply and region II general excess demand, so that low w and p have a positive
real balance effect on demand.' Now consider the efective demands and supplies. When there is excess supply in one market, that lowers the effective
demand in the other relative to the notional demand there; thus the region of
general excess effective supply (which we denote by K) must contain that of
general excess notional supply. An analogous argument applies for general
excess demand. Hence when we go from notional to effective demands, regions
I and II expand, so regions III and IV must contract, and we obtain Fig. II.2
The curves are now those on which effective demand equals effective supply,
and therefore they divide wage-price space into regions corresponding to those
defined by (i o) to (I 3).
w
d s
C =C
~ ~~~~~~~~~
/=1
d s
Fig.
i I.
1978]
MACROECONOMIC
MODELS
WITH
8I I
RATIONING
QUANTITY
in which the existing (w, p) deviates from that at W is only slightly less so.
Even in dynamics, to focus on wage-price adjustment instead of quantity
changes in this context seems misplaced. We therefore prefer for most purposes
to work directly with the endogenous variables.
Time plays an essential role in the macro model we have set out in this
section. We could have proceeded with an atemporal model of the firm, the
only asset then being household money balances. In this case, there can be no
equilibrium of type U. For if firms cannot carry inventories forward for sale
in the future, and they face excess supply of goods, they will not seek to hire
more labour than they already employ, merely to accumulate more stocks of
unsaleable goods. Rather, they will cut back on labour use until production
is no greater than what they can sell in the current (the only) period. Consequently, the goods-constrained demand for labour function becomes the inverse
of the labour-constrained supply of goods function, which makes the distinction
between them less obvious. The F-wedge collapses to a single curve representing
both &s(l) and ld(c), the production function shifted downwards by g. Disregarding borderline cases, we have three possible equilibria, seen in Figs. I2I4. These cases are analysed in detail by Malinvaud (I977).' Note that Fig. I2
H
K
Fig.
I12.
Keynesian unemployment.
Fig.
I 3.
H
F
Repressedtinflation.
Fig.
I 4.
Classical unemployment.
is effectively Fig. 2 above with the addition of ls(C), which is nlot relevant in
the Keynesian regime. This means that the production function in Fig. 2
which is the translation of the 450 line in Fig. I iS the goods-constrained labour
demand function.
If we now consider a fully atemporal model, we find that the interaction of
markets through quantity constraints generates inconsistencies. Analogously to
the argument above, we can rule out the possibility of regime C. Suppose
households hold no money balances at the end of the period, because there is
no future in which they could be used. If they face excess demand for goods,
they will not seek to sell more labour merely to accumulate unusable money
balances; they will instead withdraw labour until they are earning no more
than they can spend. Their goods-constrained labour supply function is now
u s(C) cothe inverse of their employment-constrained goods demand function;
The intersection of this single H curve with the F curve of
incides withed(l).
1 Malinvaud (1977, pp. 123-5)
also considers an ad hoc objective function for firms incorporating
inventories, but it is not based on any intertemporal model of the firm.
8I2
THE
ECONOMIC
[DECEMBER
JOURNAL
Figs. I2-I4 would give a picture identical to Fig. 2, which would now appear
to represent simultaneous determination of interacting quantity-constrained
equilibria in both the goods and labour markets.
On this interpretation, however, the intersection of Cd(l) and ld(c) would
coincide with that of es(l) and is(c), and we would have equilibria of both
type K and type R with the same data at the same (c, 1). This is the paradox
mentioned earlier. If the labour supply function is the inverse of the demand
function for goods, and the labour demand function is the inverse of the supply
function for goods, the system is underdetermined. We have two markets and
only one independent equilibrium condition. The static system is underdetermined because behaviour is overdetermined: neither households nor firms
have the extra degree of freedom offered by the opportunity to hold an asset
for future use. This is not to say that static models are necessarily inconsistent
when wages and prices are flexible, but under sluggish wage and price adjustment, they cannot generate Keynesian effects.
5. COMPARATIVE
STATICS
DYNAMICS
AND
(adl|)
(aldOaC)] dg
Qjl]
dg,
(43)
dl
[I - (@cd/9l)
(cd/@9c)j
dg
&ld/&]dgc
Notice that the marginal propensity to consume, aCd/lc, consists of the consumption demand response to sales of labour and the labour demand response
to sales of goods. The reader can easily verify that excess effective supply falls
in both the goods and labour markets. One can similarly investigate the effects
of a wide variety of such parameter changes, starting in each of the different
regimes, in the manner of Barro and Grossman (1976) and Malinvaud (1977).
The analysis of Section 3, with its explicit treatment of intertemporal and
expectational effects, should however suggest considerable reservations about
the value of such exercises. Here we have expectations not only on prices but
also on quantities, and they may often dominate the simple contemporaneous
effects. Thus in the example above, falling excess supply of labour may make
households revise their expectations about future unemployment optimistically,
and therefore feel less need for real balances to carry forward, and hence
increase current demand for goods. Meanwhile producers are also revising
1 Note that with inventories the multiplier is smaller than it would be without them. The expressions
in (43) are increasing in a4/lac, which is for any c normally less than its analogue in the case without
inventories, i.e. the slope of the labour requirements function (inverse of the production function).
I978]
MACROECONOMIC
MODELS
WITH
QUANTITY
RATIONING
8I3
their expectations about future demand for goods optimistically, so they increase
their demand for labour. If expectations are fairly sensitive to observed rationing, this may generate a good deal of instability. Indeed, the power of quantity
expectations and their interactions with each other and with current variables
tends to give the analysis much more of a "bootstraps" character than in
models with only price expectations.'
The "bootstraps" element becomes even clearer when we realise that although each of the constrained and unconstrained demand functions derived
in Section 3 does incorporate expectations, these are defined with respect to
the agent's initial position. There is no ahistorical, abstract consumption function. Each function must be referenced by the starting point. Thus in Fig. 15,
Cd(l) is the labour-constrained consumption demand function when the goodsconstrained demand for labour function is tj(c) and the equilibrium is K,.
c
F2
K,
C'E
Fig.
I5.
When the demand for labour function shifts to t2(c), the system moves along
cld(l) to K2. But having arrived there (accepting this arbitrary periodisation),
expectations are affected by being in K2, and the new consumption demand
function is (say) Cd(l) . The same point of course applies to the F-system.
In principle, therefore - and certainly in practice - we ought to specify the
expectations side of the model much more closely before trying to make any
statements in comparative statics. Nevertheless, we might take results derived
by assuming inelastic expectations as some indication of directional effects, and
also as giving us some insight into the different conditions generating different
types of equilibrium.
Thus if we start from Walrasian equilibrium and increase mo (a helicopter
drop) or g, we move into R in two different ways: if mo is higher than that
I This is apparent in the model of
Varian (1977), which includes an explicit quantity expectations
mechanism. The author remarks, "In the effective demand model, this type of inefficient, 'selffulfilling expectations' is the rule" (p. 579).
814
THE
ECONOMIC
JOURNAL
[DECEMBER
consistent with W at given wages, prices and other data, then the H-system
will be to the northwest of W (both consumption demand and labour supply
effects); whereas if g is higher, the F-system is shifted downward (see Figs. I 6I7). In either case, we have an intersection of ls(c) and es(l), and in the latter
we could represent the "process" of contraction as the intersection moves
along the curves by the "supply multiplier" (see Barro and Grossman, I974):
dc
s/ l) ( )i
Cs/
s/(a
- I[-
] dg
aisl ac
I (a-s/ac) dg.
_islac
dl = - [I-(ae&/al)
dg (ais/@c)1
dg,
(&s/Cl/)]
(44)
H'(mn'0)
-d
RI
(inces
T~
Wn
/F(gi)ras n)
T
~~
Ac
/8
whichrequinretatste ingood-ontaie
/s
@)
demad
nfrelaseu
g curv
nerec
the labour-constraineddemand for goods curve from below, and that the goodsconstrained supply of labour curve intersect the labour-constrainedsupply of
goods curve from below. To these we add the assumptions we made in constructing each "wedge":
O <
5sd/@l <
I /(01
/@C),)
(46)
O < aC/l<
Il(alcllaC).
It will be apparent that if we impose these conditions on the slopes over the
whole of the relevant range of variation of and c andTadd some global boundary conditions, we can get an existence-uniqueness result. Bfhm (I976) does
this for the case with no inventories. Given that Cd(.) has a positive intercept
and os(.) a non-positive intercept on the c-axis (the latter is not trivial when
I978]
MACROECONOMIC
MODELS
WITH
QUANTITY
RATIONING
8I5
there are inventories), the most important additional requirement is that is(.)
intersect the i-axis to the left of &s(.). This means that government cannot take
an excessive share of output, and household real balances cannot be so high
as to cut back labour supply substantially.
Given the signs of the partial derivatives of the various functions, one may
explore the comparative statics properties of the model by shifting the H- and
F-systems appropriately. We make the following not implausible assumptions:
aCdlaMo
>
o,
aCd/lw
acdlaMo
>
o,
ajdl
aW
ais/
>
o,
aCd/lp
< o,
>
o)
ajdlap
<
0o,
aisl/p
0,
4ac-s/ap>0
ald/ aW < o,
adl ap>
ais
,a6l
,
> 0o
= 0)
acd/@l
ais/
> 0,
> O,
>
An upward shift in the production function or an improvement in the terms of trade would have
the same effects as a fall in g.
3 This is essentially the point made by Ehemann (I974).
2
8i6
THE
[DECEMBER
JOURNAL
ECONOMIC
s)
(Cd-C
dw
dg, dm0
dw
dg, dn0
dp
dp
dw>O,dp<O
dg>O,dm0>O
dg<O,dmo<O
dp>O,dw<O
dw
dg, dno
dw
dg, dm0
U
dp
dp
Table i
and Output
StaticEffectson Employment
Comparative
K
dw
dp
+?
+
o
o
dm
dg
+
+
U
?
o?
-
In going fromone periodto the next, the governmentmay have new productiveassetsand new liabilities(Am),while firms'inventorylevelsand households' money balances will in general have changed (Ai, Amh), and this
period's dividend payments may differ from those of last period. These are
endogenously generated changes in quantities; to them, one might wish to add
price and wage adjustment, if we believe that this will follow upon quantity
rationing. Again, we may represent these dynamic phenomena in (c, 1) space
by shifting the H and F systems, or moving in (w, m) or (w,p) space when
appropriate. Barro and Grossman (I976), Bohm (I976) and Varian (I977)
work out some dynamic paths, and less formally, Malinvaud (I977) puts some
arguments about long-run behaviour of the system. Any such analysis, however, should take account of the full range of intertemporal changes in asset
I978]
MACROECONOMIC
MODELS
WITH
QUANTITY
RATIONING
8I7
EXTENSIONS
Finally, we discuss briefly some important respects in which these models need
to be qualified or extended. First, markets for fixed investment and financial
assets should be added. Barro and Grossman (I976) take a significant step in
this direction. Assuming (as they do) that financial markets always clear and
that investment goods are perfect substitutes in supply for consumption goods,
there is only one significant extra dimension, the endogenous interest rate. A
more complicated alternative would be a two-sector model with the possibility
of rationing in both consumption and investment goods markets.
More fundamental in many ways, because it may change the very notion of
equilibrium with which one works, is the question: who sets prices and wages?
One possible answer is: unions, industry price leaders, government; and this
direction has been pursued by Benassy (I978). Giving these decisions to a
different set of agents leaves the microeconomics of our paper valid though incomplete. An alternative possibility is to pursue Arrow's (I959) recognition that
markets which do not clear are not perfectly competitive and attempt to model
the conjectural demand and supply schedules of households and firms explicitly.
Such an analysis suggests some insights into the prevalence of different regimes.
Under excess demand in the goods market, sellers' temporary monopoly power
is likely to be reflected in relatively price inelastic conjectured demand curves,
and the resulting tendency to raise prices suggests that excess demand episodes
(repressed inflation or classical unemployment regimes) are unlikely to endure.
Under excess supply in the goods market, sellers are weak, their competitors
being flush with inventories. Then the theory of kinked oligopoly demand will
apply as Negishi (I976) argues and downward price rigidities will be the rule,
at least in the early part of recessions. Together with similar asymmetries in
though the presence of unions is
the labour market (see Flemming, I975),
to excess demand for labour, this
sensitive
less
increases
wage
make
likely to
suggests that Keynesian unemployment will be the most commonly observed
regime, with underconsumption essentially incompatible with the existence
of powerful unions. A proper microeconomic analysis of conjectural equilibria
will need to combine the insights of the Negishi (I960) model of monopolistic
competition (see Hahn, I977 b) - with those of the recent literature on market
8I8
THE
ECONOMIC
[DECEMBER
JOURNAL
even a small open economy need not face perfectly elastic supply and demand
schedules for tradeables if the goods market does not clear at the level of the
world economy. Arrow's argument then suggests that a more general model is
needed to analyse open economies.
Even in the international economics literature, where the extreme small open
economy assumption has been so convenient, an asymmetry is increasingly recognised. Although a small country may well face a very elastic supply of goods
from abroad, it may be fairly specialised as an exporter and thereforeunable to
sell all it wishes. This may help to explain why we observe Keynesian unemployment more often than classical unemployment. Similarly, the possibility of meeting excess demand for goods through imports makes repressed
inflation fairly rare, except when imports are restricted (as in wartime or in
centrally planned economies). Clearly, introducing foreign trade in a more
realistic way would increase the applicability of the quantity-constrainedmacro
model considerably.
Finally, we come to questions of aggregation and spillovers. Taken literally,
the model we have presented can be thought of as a macroeconometricmodel
with discrete switching between distinct regimes, and hence as a rather complicated extension of the recent literature we have mentioned on the econometrics of non-clearing markets. One difficulty with this approach arises in
aggregating, as for example over labour markets,where it seems more plausible
that different submarkets may be in different regimes but that the relative
dominance of the different regimes can change. Another arises when there is
excess demand for "the" consumer good. In reality with a range of consumer
goods, there will be "forced substitution" (Kornai, I97I) from those in excess
demand to those in excess supply at the given non-market-clearingprice vector.
Thus there will be differentintensities of rationing, depending on the elasticities
of substitution and how far consurnersare forced to switch, and the very concept of aggregate excess demand becomes ill-defined (Portes, I976).
These phenomena of aggregation and spillovers, as suggested by Muellbauer
(1978), can have a fundamental impact on the empirical and statistical forms
taken by macroeconomic models constructed in the spirit of the theory we
have presented. Such models are likely to be rather different from the macroeconometric models which are currently popular, whether one represents the
theory in the form of discrete switching between regimes, as we have done here,
or attempts to aggregate explicitly over submarketsin different regimes.
BirkbeckCollege
BirkbeckCollege
HarvardUniversity
JOHN
MUELLBAUER
RICHARD
January1978
Date of receiptoffinal typescript:
PORTES
I978]
MODELS
MACROECONOMIC
APPENDIX:
Malinvaud
A SIMPLE
WITH
MODEL
INVENTORY
8I9
RATIONING
QUANTITY
OF
THE
FIRM
(I977),
(I976),
max 72=
+ 21Y22)
P2X2-W2(,2
subject to
(Y2 > o,X2
o?=il1-jcxi1+y2-X2
" o).
Y2= P2/W2-ft.
The unrationed period
2-
(A 2)
= 7r2(i1,P2, w2).
- wl(,yl+ + I+
y
(4 >
i1 i0--< 2+y1-x2
r2(i,
p2, w2)
(3I)
is
(A 3)
io-__iCi2+
(P1-
(A 4)
11=
(A 5)
820
THE
ECONOMIC
JOURNAL
[DECEMBER
io]-wl}/(w1
+p2x)
from which sales rationed labour (35) can be computed through 11 = /Y + 2y2
The resulting function and (A 6) are easily seen to give something like Fig. 4.
Note that for the sales constraint to be binding, xl is below the level which
(A 4) says would have been supplied in the absence of rationing.
Rationed both in sales and labour. If given the sales ration, the labour ration is
effective (less than the sales-rationed labour demand), and if given the labour
ration, the sales ration is effective (less than the labour-rationed supply of
goods), then the firm is rationed doubly. Inventories are then given by
il=j
1'2 + l (7 ) _j
REFERENCES
of EconomicStudies(forthcoming).
Dreze, J. (1975))." Existence of an Exchange Equilibrium under Price Rigidities." InternationalEconomic
Review,vol. i6, pp. 301-20.
Ehemann, C. (1974). "General Disequilibrium, Fiscal Policy, and a Wage-Price Freeze." EconomicInquiry, vol. 12, pp. 35-52.
Epstein, L. (1975). "A Disaggregate Analysis of Consumer Choice under Uncertainty." Econometrica,
vol. 43,
pp. 877-92.
pp. 497-514.
Fama, E.
(I970).
pp. I63-74.
I978]
MACROECONOMIC
MODELS
WITH
QUANTITY
RATIONING
82I
28