Professional Documents
Culture Documents
Sergio Parrinello*
University of Rome La Sapienza
(March 2004; revised August 2004)
ABSTRACT
This paper reconsiders a recent criticism which contends that the theory of general intertemporal equi-
librium, formulated by taking the physical endowments of capital goods as given, is not protected from
the problem of capital at the centre of the two Cambridges debate of the 1960s. The author confirms
such a criticism following a different approach. He argues that the stability analysis of an intertem-
poral equilibrium via ttonnement must be consistent with a uniform rate of return on capital. He
shows that the resulting non-orthodox ttonnement subverts the traditional analysis of equilibrium
stability.
1. INTRODUCTION
* Thanks to two referees and to Enrico Bellino, Pierangelo Garegnani, Michael Mandler and
Bertram Schefold for useful discussion, criticism and comments at different stages of elabora-
tion of this paper, under the usual exemption from responsibility. This article is a revised and
extended version of the authors Working Paper no. 67 published by the Dipartimento di Econo-
mia Pubblica, Universit di Roma La Sapienza, April 2004, and has benefited from the finan-
cial support of the University itself. A preliminary outline of section 4 was presented by the
author at the Conference Sraffa o unaltra Economia, organized by the Dipartimento di
Innovazione e Societ of the University of Rome La Sapienza, Rome, 1213 December 2003.
Blackwell Publishing Ltd 2005, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main
Street, Malden, MA 02148, USA.
Ttonnement Pricing Revisited 515
where la, lb are given positive labour coefficients, aa, ba, ab, bb are given posi-
tive coefficients of goods a, b used as circulating capital. The technology is
assumed to be viable. Let Djt(), j = a, b, denote the demand function for con-
sumption of goods a, b available at time t = 0, 1 and () the relation with the
independent variables (Pa0, Pb0, Pa1, Pb1, W1).
Market clearing equations for commodities available at t = 0:
A0 = Da 0 () + (aa A1 + a b B1 ) (2)
B0 = Db0 () + (ba A1 + bb B1 )
where A0, B0 are given total endowments of goods a, b at t = 0 and which are
supposed to be allocated according to given individual property rights; A1,
B1 are quantities of goods a, b produced during the period and available for
consumption at t = 1.
Labour market clearing:
l a A1 + l b B1 = L (3)
A1 = Da1 ()
(4)
B1 = Db1 ()
A0 Pa 0 + B0 Pb0 + LW0 Da 0 ()Pa 0 + Db0 ()Pb0 + Da1 ()Pa1 + Db1 ()Pb1 (5)
Let us take good b available at t = 1 as the standard of value and have the
equation of price normalization as
Pb1 = 1. (6)
S0(), I0() can serve for the interpretation of an equilibrium, but they seem
to play no distinct causal role for the determination of the equilibrium itself,
in comparison with the demand and supply functions of the individual phys-
ical commodities.1
Let us replace the quantities A1, B1 in (2), (3) by the equations (4) and the
prices Pa1, Pb1 in the demand functions, Djt(), j = a, b; t = 0, 1, by the price
equations (1). Let P = (Pa0, Pb0, W1) denote the price vector and djt(P) the
1
This negative remark has been already put forward by Schefold (2004).
dPa 0
= H a (E a 0 (P))
dt
dPb0
= H b (E b0 (P)) (9)
dt
dW1
= H L (E L (P))
dt
2
At each iteration the auctioneer is supposed to call only the prices of the initial endowments
and to receive back from the producers the information on the prices of goods a, b at t = 1,
which satisfy (1) and, in a more general model with alternative methods of production, are asso-
ciated with the choice of the cost minimizing techniques; next the auctioneer receives the infor-
mation on the individual net demands and calculates the corresponding aggregate excess
demands in order to call new prices of the initial endowments.
where the prices satisfy the equations Pb1 = lbW1 + abPa0 + bbPb0 = 1 and the
functions Ha(), Hb(), HL() satisfy the identity abHa() + bbHb() + lbHL() 0.
Then, given the initial call of prices, P(0), a path of subsequent calls P(t),
t > 0, can be determined by any two differential equations chosen from (9)
and by the numeraire equation.
Therefore, the determination not only of an equilibrium solution, but also
of the stability properties of such a ttonnement rule leaves no causal role
to the functions S0(), I0(), of aggregate saving and investment. We may
concede that, granted the validity of the Walrass law (5), we could replace
one equation chosen from (1)(4) with the equation S0() = I0() to calculate
an equilibrium solution. This substitution does not leads us far, because it
does not assign to S0(), I0() any special role in the adjustment mechanism.
However the auctioneer, instead of calling prices according to equations (9),
is compelled to follow a different rule, if the theory has to be consistent with
the extension of Jevonss law to the prices of capital goods.
Assume that in equilibrium Pa0 > 0, Pb0 > 0, Pa1 > 0, Pb1 > 0, W1 > 0 and define
P P P P
the own rates of interest ra a 0 - 1, rb b0 - 1. Let pa 0 = a 0 , pa1 = a1 ,
Pa1 Pb1 Pb0 Pb1
Pb0 P W
pb0 = = 1, pb1 = b1 = 1, w1 = 1 denote current relative prices with good
Pb0 Pb1 Pb1
b chosen as the numeraire at each time. The price equations at current prices
are
pa1 - l a w1 pa 0 - l a w0
= 1 + rb
pa 0 - l a w0 aa pa 0 + ba pb0
(10)
pb1 - l bw1 pb0 - l bw0
= 1 + rb
pb0 - l bw0 a b pa 0 + bb pb0
3
It should be noticed that the same rate rb, the own rate of interest in the numeraire, applies to
both equations (10). In particular, the first equation, which pertains to the industry of good a,
can be written
w1
where w0 is the discounted wage rate. Equation (10) sets that the
1 + rb
factor of return, received from investing in the production of a certain good,
is uniform and equal to 1 + rb, the own factor of interest on the numeraire.
The effectiveness of each return results from the multiplication of two terms
in brackets: (1) the factor of appreciation of a bundle of a good and of a
bad (i.e. a labour coefficient); (2) the own factor of profit calculated at
contemporary prices. It should be stressed the fact that the assumption of
constant returns to scale, adopted in this paper, is useful to maintain
as a benchmark the model used by Hahn and Garegnani for different pur-
poses, but it merges a general equalization between rates of returns
and a particular equalization which is implied by the price equations derived
from that assumption. If the assumption of constant returns to scale
should be replaced with the assumption of decreasing returns, the corre-
sponding price equations would not imply that the average rate of return on
capital invested in each industry is equal to the own rate of interest on the
numeraire. However, the following relation must be satisfied in force of
Jevonss law (the law of unique price) independently of the type of returns
to scale:
pa1 p
(1 + ra ) = b1 (1 + rb ) (11)
pa 0 pb0
Equation (11) sets a relation between the effective factors of return received
from saving and lending the goods a, b. The effective rate of return on each
good is calculated by multiplying its own factor of interest for the factor of
appreciation of the good itself. It should be noted that (11) pertains to the
sphere of exchange (not to the sphere of production) and must be interpreted
as an equation, although its mathematical form resembles a tautology.4 The
usual notation of the relative prices, written as ratios between nominal prices,
can be indeed misleading. In fact (11) can be written
Pa1 W P P P
pa1 = = la w1 + (aa pa 0 + ba pb0 )(1 + rb ) = la 1 + aa a 0 + ba b0 bo
Pb1 Pb1 Pb0 Pb0 Pb1
which shows that the equation at current prices is consistent with the equation at discounted
prices, Pa1 = laW1 + aaPa0 + baPb0, i.e. with the first equation (10). This consistency would be
missing if rb should be replaced with ra in the first equation (10).
4
Schefold (2004, p. 11) seems to interpret an equation like (11) as a tautology.
pa1
(1 + ra ) = 1 + rb . (11)
pa 0
It follows from (10), (11) that rb represents the uniform effective rate of
return, which in the model applies to saving, lending and productive invest-
ment. The recognition of a uniform rate of return for an economy, which is
not in a long-period equilibrium, is a crucial step of the argument.
Garegnani asserts (1) that capital goods are perfect substitutes for the saver
and (2) that the properties of the relative prices of commodities available
at different times (intertemporal relative prices) are different from those
pertaining to the relative prices of commodities available at the same time
(contemporary or current relative prices). In the theory of intertemporal
equilibrium (1) and (2) must be grounded on an explicit model of individual
choice. We shall specify such a model through some intermediate steps, which
aim to avoid certain possible misunderstandings.
5
For example, Pa1/Pb1 = 4/1 and Pb0/Pa0 = 1/2 mean that 1 unit of a1 is exchanged for 4 units of
b1 and 2 units of b0 are exchanged for 1 unit of a0. However, it would be logically possible that
(i) 2 units of a0 are not exchanged for 1 unit of a1 and (ii) 1 unit of b0 is not exchanged for 1
unit of b1. Instead market equilibrium via arbitrage rules out (i), (ii) and therefore the equations
Pa0/Pa1 = 1/2, Pb0/Pb1 = 1/1 must hold.
model (1)(6) are derived from the rational choices of the individual con-
sumer, who is supposed to solve the problem:
max u()
s.t. the intertemporal budget constraint (12)
a0 Pa 0 + b0 Pb0 + lW1 = ca 0 Pa 0 + cb0 Pb0 + ca1Pa1 + cb1Pb1
where l is the labour endowment; a0, b0 are the initial endowments of goods
a, b and cjt is the dated consumption of good j, j = a, b, t = 0, 1; and u() is
the utility function, all notations being referred to the consumer.
In microeconomics the attribute perfect substitutes has a definite meaning
in the following cases:
(i) if we specify u() = u(ca0, cb0, ca1, cb1) and we assume that the marginal
rate of substitution between two consumption goods is constant (perfect
substitutes);
(ii) if we define the indirect utility function6 as the maximum direct utility
achievable at given prices and income: f(Pa0, Pb0, Pa1, Pb1, W1, y)
max u(ca0, cb0, ca1, cb1) s.t. y = ca0Pa0 + cb0Pb0 + ca1Pa1 + cb1Pb1, where y =
a0Pa0 + b0Pb0 + lW1, and we say that the physical constituents of y are
perfect substitutes;7
(iii) if we assume that the consumer postpones at t = 1 the choice of goods
to be consumed at t = 1 and we specify u() = u(ca0, cb0, s), where s =
(a0 - ca0)Pa0 + (b0 - cb0)Pb0 is his total saving at discounted prices. In
this case we can say that the physical constituents of s are perfect
substitutes.
6
See Varian (1992, section 7.2).
7
For example, the marginal rate of substitution of b0 with a0 is equal to fa0/fb0, where fa0, fb0 are
the partial derivatives of f() with respect a0, b0. As f() is a function of the function y(), we have
fa0/fb0 = pa0/pb0.
perfect substitutes, even though they do not enter into the (direct or indirect)
utility function of the consumer. In fact, we may distinguish different facets,
roles, functions of the same decision maker: he is a consumer, a saver, an
investor and a worker at the same time. Each facet can be supposed to max-
imize some objective function and the (non-schizophrenic) result is an
optimal plan of consumption c*a0, c*b0, c*a1, c*b1, as a solution to problem (12).
Besides the signals of the net demands for goods and of the supply of labour
services which are sent by the consumer
the saver sends the signal of his optimal saving plan at t = 0 (with no saving
at t = 1):
( ) (
s* = a0 - ca*0 Pa 0 + b0 - cb*0 Pb0 )
It remains to explain why s* can be an independent effective signal. Why are
six signals, instead of five, received by the auctioneer as distinct effective
market signals?
We can imagine that the saver receives from the consumer (the same indi-
vidual) the quantities not consumed (a0 - ca0), (b0 - cb0) and the purpose of
the former is to transform their value s = (a0 - ca0)Pa0 + (b0 - cb0)Pb0 into the
maximum purchasing power available at t = 1. As in case (ii), the physical
constituents of s are perfect substitutes for him, at the given prices Pa0, Pb0.
He would change the physical composition (a0 - ca0), (b0 - cb0) of s before
lending the goods, if all contemporary arbitrages were not fully exploited;
otherwise he is indifferent to the basket of goods contained in s. For the same
reason the physical constituents of the income that he received from the bor-
rowers at the end of the period are perfect substitutes for the saver. He would
exploit all possible spot-forward arbitrages if equation (11) were not initially
satisfied, otherwise he is indifferent to the physical composition of the income
that he receives from the borrowers at t = 1. We should also note that for the
saver perfect substitutability implies indifference in the choice of the physi-
cal mix of saving only if the prices given to him satisfy Jevonss law expressed
by equation (11), which concerns the effective own rates of return. Of
course, saying that the goods a, b are perfect substitutes for the saver does
not mean that they are such also for the consumer. Each dated good a, b is
physically homogeneous; yet, from the point of view of the two facets of the
individual, ca0, cb0, ca1, cb1 are consumption goods, whereas (a0 - ca0), (b0 - cb0)
are capital goods.
The argument presented in the previous section points out that the auction-
eer can call only certain relative prices and must follow different rules for
calling the relative prices of contemporary commodities (contemporary
prices) and the relative prices of commodities available at different times
(intertemporal prices).
In the model the degrees of freedom for calling prices are limited by two
distinct limitations.
A first limitation is inherent in the assumption that we borrowed from
Schefold in order to apply a ttonnement pricing to a production economy
which is subjected to constant returns. At first sight the assumption seems to
prescribe that the auctioneer can call only the nominal prices of the three
inputs, Pa0, Pb0, W1 and then the cost-minimizing choices under perfect com-
petition will determine the prices of the commodities available at t = 1, Pa1,
Pb1, as from the price equations (1). However, labour is not a stock available
at t = 0 but a flow variable, and wages are supposed to be paid at t = 1, when
the products will be available. Therefore, it is just as possible for the auc-
tioneer to call the nominal prices of two inputs Pa0, Pb0 and the nominal price
of one output, say Pb1, and leave the price equations to determine the wage
rate W1 and the price Pa1. This is what we assumed in sections 2 and 3. In
each case the auctioneer can call only three nominal prices: the prices of two
contemporary commodities and the price of one commodity available at a
different time.
The second kind of limitation for calling prices is central to our argument
and is independent from the assumption of constant returns to scale reflected
by the price equations (1), (10). It rests on the fact that Jevons law prevents
the auctioneer from controlling each own rate of return independently from
the others. At each iteration he cannot call prices which violate equations
(11) or (11). A change in rb drags up or down all equalized own effective
rates of return. The ttonnement pricing must obey this second constraint.
We shall formalize such a ttonnement in the next section, but let us first
generalize the second characterization for a multicommodity and many
period economy.
Let us assume an economy in which n produced commodities are available
at dates 0, 1, . . . , T and labour is performed in periods [0, 1], [1, 2], . . .
[T - 1, T]. For the sake of the argument let us disregard the first constraint
It should be noticed that the total number of relative prices which the auc-
tioneer can call is not at issue. In the absence of the constraint related to con-
stant returns, the traditional auctioneer can call n(T + 1) + T - 1 relative
prices on the whole. The second limitation allows him to call the same total
number of prices, n(T + 1) + T - 1, but it prescribes a different selection of
relative prices on the basis of the distinction between contemporary and
intertemporal relative prices.
Which rules does the auctioneer follow to adjust the prices under his control?
The auctioneer must follow different rules for changing prices, according to
the distinction between current (contemporary) relative prices and the
intertemporal relative price of the numeraire. A change in a current relative
price is governed by the traditional rule which prescribes that the sign of the
change in the price is equal to the sign of the excess demand of that good.
8
Alternatively, he might take a basket made with one unit of each good available at time t
as the standard of value for each date t and period t, t + 1 and fix: P1t + P2t + . . . + Pnt = 1,
t = 0, 1, . . . , T.
7. HETERODOX TTONNEMENT
lw1 c p c
aa pa 0 + b0 + = ca 0 pa 0 + cb0 + a1 a1 + b1 (13)
1 + rb 1 + rb 1 + rb
The aggregate demand function for commodity j at time t is: Djt(pa0, pa1, w1,
rb). We can replace the prices pa1, w1 in Djt() with a solution to the price equa-
tions (10), provided that rb falls within its feasible range. We obtain the
demand function djt(pa0, rb). Let us define the price vector p = (pa0, rb) and
the excess demand functions:
9
The causal role assigned to the aggregate capital flows (i) does not presuppose a monetary
economy under uncertainty and (ii) does not pose an additional threat to methodological indi-
vidualism, in addition to that which is already implicit in the rudimentary price dynamics based
on the assumption of the auctioneer who calls unique prices. On (i) and (ii) we do not share the
criticisms that Schefold (2004) addresses to Garegnanis approach, although the former correctly
points out a similarity between the latter and Clowers (1969) formulation of the microeconomic
foundations of Keyness aggregate demand function.
This reduced form of model (1)(6) has only one intertemporal relative price,
the rate rb. The auctioneer will call a higher (lower) contemporary price, pa0,
if and only if he finds a positive (negative) excess physical demand E a0(p).
Instead he will call a higher (lower) rate rb if and only if he finds that the
value of the aggregate demand for investment exceeds (falls short of) the
value of the aggregate supply of saving. That aggregate excess demand at
current values is paoE a0(p) + pboE b0(p).
Let H a, H b be smooth sign-preserving functions of excess demands, with
H a(0) = H b(0) = 0. Then the following differential equations determine the
ttonnement dynamics for the whole economy:
dpa 0
= H a (E a 0 (p))
dt
(15)
drb
= H b ( pa 0 E a 0 (p) + E b0 (p))
dt
Given the initial prices, p(t), t = 0, a path p(t), t > 0, is determined by the
differential equations (15). The paths of the remaining current and intertem-
poral prices follow from the relations between current prices and discounted
prices and from the price equations (1), (10) with the numeraire equations
pb1 = 1 and Pb1 = 1. The corresponding path of the excess demand for labour
E L( P) follows from Walrass law (14).
The heterodox ttonnement of equations (15) can be compared with the
orthodox ttonnement of equations (9). The excess of aggregate investment
over aggregate saving plays a causal role in (15) and ultimately reflects the
role of the demand for a value of capital (flow) even in a model with het-
erogeneous physical capital goods, whereas it does not intervene in (9).10 As
a consequence, the proofs of existence and uniqueness of equilibrium are not
affected, but we should abandon the standard proofs of stability because
Jevons law imposes (15) instead of (9). In the next section we shall present
an example in which a well-established method of proving global stability
fails if we move from (9) to (15).
10
Assuming that the price of a certain commodity reacts not only to the excess demand for that
commodity but also to the excess demand for other commodities is not a novel approach in sta-
bility analysis. In particular, the application of Newtons method of numerical analysis pre-
scribes that the price of each commodity reacts to the excess demand for all commodities by a
certain uniform coefficient and brings about a proportional decrease in all excess demands (see
Smale, 1976). However, the specific feature of the adjustment described by (14) is the fact that
an intertemporal price reacts to the value of the excess demands for capital flows.
Assume that the sign-preserving functions of excess demands are the iden-
tity function. The complete system of differential equations is:
dV dp dr dw
= 2 ( pa 0 - pae 0 ) a 0 + (rb - rbe ) b + (w1 - w1e ) 1
dt dt dt dt
dV
= 2[( pa 0 - pae 0 )E a 0 (p) + (rb - rbe )( pa 0 E a 0 (p) + E b0 (p)) + (w1 - w1e )E L (p)]
dt
dV
= -2[( pae 0 + rbe pa 0 )E a 0 (p) + (1 + rbe )E b0 (p) + w1e E L (p)]
dt
dV (p e )
If p = pe is an equilibrium price vector, then = 0 . The expression in
dt
square brackets can be written
dV (p)
If Z is positive, < 0 and the equilibrium is stable.
dt
Let us assume that the demand functions satisfy the weak axiom of
revealed preferences (in particular that the goods are gross substitutes). Then
the term in square brackets in (16) is positive for any non-equilibrium price
vector. The axiom assures11 the uniqueness of equilibrium and its global sta-
bility within the orthodox ttonnement (9). However, the same axiom does
not imply that Z is positive, because the term rbe(pa0 - pea0)Ea0(p) in (16) may
be negative and its sign may prevail. In two cases the axiom would imply a
positive Z. First, if we assume pa0 = pea0, i.e. if the price of commodity a avail-
able at t = 0 is kept at its equilibrium level, the possible negative term disap-
pears. Second, assuming reb > 0, if (pa0 - pea0) and Ea0(p) have the same sign,
reb(pa0 - pea0)Ea0(p) is positive. The first case is uninteresting; the second is not
plausible, because it requires that any price above (below) its equilibrium level
is associated with a positive (negative) excess demand for the corresponding
good.
The analytical case examined in this section does not furnish proofs of sta-
bility or instability, but it helps to locate the specific difficulty for the theory
of intertemporal equilibrium. We have not proved a case of unstable equi-
librium, relatively to the heterodox ttonnement, and we have not excluded
that a different Liapunov function can be found in order to prove stability.
We have not proved either that an equilibrium, which is unstable under the
orthodox ttonnement, a fortiori is unstable under the heterodox one.
Furthermore we should be aware that more complicated adjustment
processes (e.g. processes with trading at false prices) may be stable and some
counter-intuitive results may turn up, as Franklin Fisher (1983) has reminded
us with regard to the passage from the traditional ttonnement to more real-
istic stability processes, where the convergence depends not only on the prop-
erties of the excess demand functions, E, but also on those of the reactions
functions, H. However, we have shown that the weak axiom of revealed pref-
erences, combined with the usual Liapunov function, does not allow us to
prove that an intertemporal equilibrium is stable within the adjustment
context which is consistent with Jevonss law. This negative result depends on
the fact that certain properties, which the excess demand functions for
individual commodities are supposed to satisfy (in our case gross substi-
tutability), do not imply definite properties of the excess demand function
for an aggregate capital flow. We shall expand on this issue in the final
section.
11
See Arrow et al. (1959).
12
I owe this remark to the comment of a referee.
REFERENCES
Arrow, K. J., Block, H. D., Hurwicz, L. (1959): On the stability of competitive equilibrium, II,
Econometrica, 27, pp. 82109.
Clower, R. W. (1969): The Keynesian counterrevolution: a theoretical appraisal, in Clower, R.
W. (ed.): Monetary Theory. Penguin, Harmondsworth.
Fisher, F. (1983): Disequilibrium Foundations of Equilibrium Economics, Econometric Society
Publication no. 6, Cambridge University Press.
Garegnani, P. (2003): Savings, investment and capital in a system of general intertemporal equi-
librium (with 2 appendices and a mathematical note by M. Tucci, in Petri, F., Hahn, F. H.
(eds): General Equilibrium. Problems and Prospects. Routledge, London (2003)). Earlier
version in Kurz, H. D. (ed.): Critical Essays on Sraffas Legacy in Economics. University
Press, Cambridge (2000).
Hahn, F. H. (1982): The neo-Ricardians, Cambridge Journal of Economics, 6, pp. 35374.
Schefold, B. (2003): Reswitching as a cause of instability of intertemporal equilibrium, Metroe-
conomica, 56 (4), pp. 43876.
Schefold, B. (2004): Saving, investment and capital in a system of general intertemporal equi-
libriuma comment on Garegnani, January, 4th version, mimeo.
Smale, S. (1976): A convergent process of price adjustments and global Newton Methods,
Journal of Mathematical Economics, 3, pp. 114.
Varian, H. (1992): Microeconomic Analysis, 3rd edn, Norton Co., New York, NY.