Inconsistency and Indeterminacy in Classical Economics

Inconsistency and Indeterminacy in Classical Economics Author(s): Karl Brunner Reviewed work(s): Source: Econometrica, Vol. 19, No.
2 (Apr., 1951), pp. 152-173 Published by: The Econometric Society Stable URL: http://www.jstor.org/stable/1905731 . Accessed: 15/02/2013 18:09
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp
.
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.
The Econometric Society is collaborating with JSTOR to digitize, preserve and extend access to Econometrica.
http://www.jstor.org
This content downloaded on Fri, 15 Feb 2013 18:09:41 PM All use subject to JSTOR Terms and Conditions
INCONSISTENCY AND INDETERMINACY CLASSICAL ECONOMICS*

BY KARL BRUNNER
I. THE ISSUE
IN
1. THE CENTRAL problem raised by Patinkin, in his articles on the properties of classical systems, is the proper construction of a theory of a monetary system, by which we mean a set of equations explaining relative prices, finite money prices, and nonzero money stocks.' Before we evaluate the members of the set of all monetary systems from the point of view of their "degree of realism," we want to know if there exists a subset of systems which might be discarded on purely analytical considerations as a scientific hypothesis. If such a subset can effectively be described, the analytical basis for empirical work will be the more definite after its elimination.
* ED. NOTE-Limitations of space have prevented ECONOMETRICA from publishing numerous critical contributions which have been received in comment upon the articles cited below. The editors therefore asked the writer of the present article, who had become concerned with this problem earlier, to attempt a clarification of the issues involved and to write a commentary on the points advanced by the various participants to the controversy. The article, however, should not be regarded as necessarily reflecting the views of persons other than the author. 1 Don Patinkin, "Relative Prices, Say's Law, and the Demand for Money," EcONOMETRICA, Vol. 16, April, 1948, pp. 135-154; "The Indeterminacy of Absolute Prices in Classical Economic Theory," ECONOMETRICA, Vol. 17, January, 1949, pp. 1-27; and "The Invalidity of Classical Monetary Theory," the paper which appears immediately preceding this one in the present issue of ECONOThese articles will henceforth be referred to as Patinkin I, II, and III, METRICA. respectively. Reference should also be made to the following three critical contributions W. Braddock which appeared in the January, 1950, issue of ECONOMETRICA: Hickman, "The Determinacy of Absolute Prices in Classical Economic Theory," EcONOMETRICA, Vol. 18, January, 1950, pp. 9-20; Wassily Leontief, "The Consistency of the Classical Theory of Money and Prices," ibid., pp. 21-24; Cecil G. Phipps, "A Note on Patinkin's Relative 'Prices,"' ibid., pp. 25-26. See also pp. 272-274, for an abstract and discusthe July, 1950, issue of EcONOMETRICA, sion of a paper by Jacob Marschak, "The Rationale of Money Demand and of Money Illusion," presented at a session devoted to this controversy at the New York meeting of the Econometric Society in December, 1949. The full text of the Marschak paper appears in Metroeconomica,Vol. 2, August, 1950, pp. 71-100. The present author has had the opportunity of studying unpublished manuscripts by Kenneth J. Arrow, Gary S. Becker, Donald Fort, Cecil G. Phipps, and David Rosenblatt, and these have to be regarded as an important basis of the present paper. I have to acknowledge further indebtedness to K. J. Arrow, C. Christ, D. Fort, C. Hildreth, L. Hurwicz, L. R. Klein, T. C. Koopmans, H. Markowitz, J. Marschak, C. G. Phipps, R. Radner, and M. Slater for discussion of
152
INCONSISTENCY
AND INDETERMINACY
153
2. The "Patinkin discussion" is not concerned with classifying each and every conceivable monetary system. Its purpose is more limited. There exists a well defined subset of monetary systems which forms the material under consideration. The feature uniting the group shall be termed the "complementary property." The following meaning will be attached to this term: Economic theory traditionally started with a study of market behavior and market processes. Thus was developed an analysis of price formation on commodity markets-and to focus the analysis, money problems were completely eliminated. In this way classical theory built up a (conceivably) consistent and determinate theory of a real economy. To proceed from such a real economy to a monetary system it was thought sufficient to "complement" the real system with an equation relating the monetary variables with the real variables, which were incorporated as parameters. Thus the total system was composed of two subsystems-a real and a monetary-each one closed with respect to its relevant variables. The simultaneous existence of these two subsystems gives rise to what has been termed the dichotomization of the pricing process. Behavior in the real part depends solely on the real variables, which are (if at all) uniquely determined by these behavior patterns. In the monetary sphere behavior is linked to money values, which result in this way from behavior patterns relevant only for this sector of the economy. The "complementary property" is thus equivalent to the logical dichotomy of the pricing process. The analytical value of both depends on the consistency of two postulates: namely, the postulate that market functions in the real subsystem are homogeneous of degree zero in the prices and the postulate that there exists a meaningful money equation.2 3. An objective of the present paper is to prove that the monetary systems characterized by the complementary property canniot be utilized as explanatory devices of observable phenomena. It will be shown that the systems under consideration will not satisfy the minimum requirevarious aspects of the problem. Of especial importance has been discussion with D. Patinkin, to whose patient clarification of moot points I feel very much indebted. Needless to say, the responsibility for all errors remains, unfortunately,
mine.
ED. NOTE-A further manuscript, "On the Dichotomy in the Theory of Price" by Yukichi Kurimura, relates to the present topic but was not received until after the above articles were completed. 2 By this I mean an equation involving finite money prices and a positive stock of money. It has to be added that the formulation above is valid only for macroanalysis. In the case of microanalysis, homogeneity is not a postulate, it is a theorem. For this case we may say that the complementary property implies that a meaningful money equation may be superimposed on the equilibrium conditions without affecting the properties of these conditions.
154
KARL BRUNNER
ments of a theory of a monetary system, which are that the theory must explain both nonzero money stocks and a unique set of finite money prices without producing inconsistent propositions.3 Since the controversy around the "Patinkin problem" started with a macroeconomic variant out of the subset of dichotomized monetary systems (the "Casselian system"), the analysis will proceed at first on a macro level. It will then be shown by a macroanalytical argument that systems which include the complementary property imply serious contradictions. But in order to clarify the nature of these inconsistencies it is necessary to resort to microanalysis. As a result we shall be able to formulate the conditions for the proper construction of a theory of a monetary system.
II. EXPLICIT AND IMPLICIT MONEY FUNCTIONS OF A MONETARY SYSTEM
1. The money equation of the monetary subsystem formulates a relation between money prices and the quantities of money demanded and supplied. This equation is usually developed as a relation between stocks-and, if presented as a flow relation, an elementary transformation gives us the stock equation. Thus the two formulations of the monetary subsystem are equivalent, and we may concentrate on the form:
(2.1) MD(P1, *l
,pn-1)
Ms(pi, *
Pn-)
M.(p,
pn-1)
where MD is the stock of money desired by economic units; Ms, amount of money outstanding;pi, - * , pn-l commodityprices;M=, excess demand for money in stock terms. This relation implies that, for any price vector p such that MD > MB, the totality of economic units desires to substitute money for commodities. This decision will be reflected in the "commodity" markets. On the other hand, money flows are generated by the demand and supply decisions in these markets, and the net result of these flows must affect the stock relationships. This fact suggests that the explicitly formulated money equation of the monetary subsystem finds its counterpart in a money equation that is implied by the working of the real subsystem: n-1 n-1
(2.2)
(2.3)
Dn-
i=i
piSi
Sn--
i=l
pi Di,
Dn-Sn-Xn,
3 It may be well to emphasize that the existenceof a solutionvector is not a is not involved in inconsistent sufficienttest that a system underconsideration This statementis strictly valid subjectonly to the restrictionsimpropositions. posedby the natureof the issue.
INCONSISTENCY
AND INDETERMINACY
155
where Si and Di are the demand and supply functions, respectively, for the ith commodity, D, and S. are the flows of money generated by the decisions to supply and demand commodities, and X. is the excess demand for money in terms of flows. 2. The logical relation between the two money functions has to be formulated. A simple budget equation for a typical unit a of a monetary system may be represented by (2.4) p[Za(P, Za) - Za] = 0O
where p denotes a row vector of prices; the bracket, the difference between two column vectors, Z,a and Z , of the desired and initial quanticommodities; and Zn, initial ties, respectively, of l,aa, Zn-i,quantities of money. The following quantified statement4 is in general tacitly attributed to the budget equation by the context of the economic analysis: (2.5)
Ap [p[Za(p, Za) -Za] =
0o,
p e Ra,
Za
Ma,
where Rat and Ma!tare the universe of discourse of the vectors p and Zar. This implies, by appropriate rearrangement,
A A
[Zn,a(py Znla)
-Zna
(2.6)
Zn,a P
Dn,a(p,
Zn,a) -
Sn,a(P,
Zn,a)I]
p e Ra.,
Zn, a Na C MaX
where 2,t,a is the money-supply vector. This proposition states that the net inflow into the axth individual's cash balance on the right will
4Let P(x) be a sentence involving x. Then, by A [P(x)], x x e U,
we mean that, for all x such that x e U, it is true that P(x). By

E [P(x)], x e U,
we mean that there is at least one x such that P(x) is true, where x e U. By
El [P(x)], x e U,
we mean that there exists a unique x such that P(x) is true, where x e U. For a discussion of this method, see Fundamental Mathematics, Vol. I (third edition), Chicago: University of Chicago Press, 1948, Chapter 6.
156
KARL BRUNNER
always be equal to the change in the cash balance, provided both quantities are measured for the same time interval.5 By summing over all a (a = 1, .* * , m) we obtain a proposition related to the whole of the money-utilizing sector P, such as
(2.7)
[M
p,
) -Ms
= D(p,
Zn,a) -
Sn(p, Zn], Ds ,
wherep eR D Ra, Zn,a EN

are the sums of Zn,a,
N, and whereMP f
Dn and S
n,a , Dn,a,, and Sn,a.
3. To proceed we have to consider the money-creating sector B. This sector's activities affect P by the net outflow created, which is S - DB. On the basis of simple economic considerations we may formulate the proposition
A [M,(2.8 )
-(
Znn,a)
S(p,
- DB(p, Zn,a)]
R,
Zn,a EN,
where Ms is the amount of money to be held by P as planned by B, SB and DB are the out- and inflow components of the net outflow of B. These quantities may be generated in two essentially different ways: i) By fiat: The money flow induces no counterflow of claims or commodities. The "fiat components" are SB,X
DB/
. If DBB
_ Slf,
then the in- and outflow quantities of the P-sector must be adjusted to B DP + SCB S nf X Sp Sn + + D Df Dn + ii) By market transaction: There exists a counterflow to the money flow. To simplify, we suppose it to consist only of claims. The "claims component" is made up of S o, Dc. By appropriate combination of (2.7) and (2.8) we derive A A -AB P A A [MD(Pr Zn,a p (2.9)
+
n) ) , -M
(p, Zn) ]S n,n,a)) , Z ,
DP(p,
Zn,a)
Dnc(p,Zn,)
S nS(p
E] p
R,
Z,a
N.
The terms of the "fiat components" of B's net outflow are canceled in this process. By eliminating the quantifiers and designating the lefthand side of the quantified sentence by MX and the right-hand side by Xn, we obtain
(2.10) M, = Xn.
is defined by Dn,a, 5 The inflow component D,,, Z,a), ilpi(Zi,a where s < n - 1 number of nonmoney objects of transaction and Zi,, - Zi, < 0, and the outflow component is defined by Sn,a = I2+_,+Pi(Zj,a Zia), where Zj,a - Zij, > 0.
INCONSISTENCY
AND
INDETERMINACY
157
Equation (2.10) will be termed the monetary relation. We may then state that this monetary relation will always be satisfied for any vector of prices and initial quantities included in the universe of discourse given within the context of some economic analysis. A simple rearrangement of terms shows that the monetary relation and Walras's law are equivalent propositions:
(2.11)
Ml
-Xn
= Ml + EpiXi = O. i1
n-1
Here Xn-l designates excess demand for bonds and is constituted by XB the two terms, Xp_1 and XB1, = xB= such that -Pn-i
S eno
-D Jno
Remembering the appropriate quantifiers, we observe that Walras's law states an obvious limitation of the economic universe: If all objects of transaction (including money) are accounted for, it follows that the sum total of the excess demand for all objects is always zero-whatever the price vector. By the nature of the market problem we cannot conceive of the contrary situation.6 Our main instruments in the analysis of macrosystems which have a complementary property are the concepts presented in this section. The relationship between the explicit and implicit money functions of a system and some implications of Walras's law furnish us the criterion for the analytical test.
III. THE LOGICAL CHARACTER OF MONETARY PROPERTY SYSTEMS HAVING A COMPLEMENTARY
1. We classify the monetary systems under consideration according to the nature of the postulates describing the money function into two groups: the Me-systems and the Xn-systems. The monetary subsystems of the members in the first group contain an M.,-function, while the monetary parts of systems in the second group contain an Xn-function. The M.-function is assumed to be nh in the commodity prices and hdl in this same price vector and in the money supply.7 These properties are satisfied by the Cambridge function,
fl-
/
8K Si
...
Pn2
-
(3.1)
6
M* -Kpn-1 i
M,
It is necessary to remind the reader that the whole discussion is kept within a static framework. That is, we suppose that the relevant phenomena can be analyzed as though they were solutions of static systems. It is assumed especially that the timing structure of market behavior, actual transactions, and payments can be taken account of implicitly or explicitly by some sort of compound variable. 7 To abbreviate the exposition the following notation is introduced: hdt represents homogeneity of degree t; h in p represents homogeneity in money prices; nh in p represents nonhomogeneity in money prices.
158
KARL BRUNNER
where K and M are given constants, Si denotes the supply functions of commodities, and M* is a member of the set M,. The Xn-function is defined as a linear combination of all nonmoney functions:
n-1
(3.2)
Xn =
-Epix.
i=1
The M.-Systems 2. The discussion of the Patinkin problem has produced a number of diverse systems with the uniting feature of the complementary property. Among the M,-systems two classes can be distinguished: (a) the class of all systems containing only one function nh in all or any proper a class of all systems containing subset of the commodity prices, and (fA) two functions nh in all or any proper subset of the commodity prices. All systems with more than two functions nh in p are excluded by the nature of the issue. The following sections will prove that the a and d classes of systems unavoidably produce inconsistent propositions. Argumentfor the a Class 3. Three sets of postulates have to be considered: CASE A: (Al) The commodityfunctions Xi (i = 1, ... , n- 1) are hdOin the commodityprices pi (i = 1, *- *, n -1). (A2) The M--function is nh in pi and hdl in pi and M. (A3) Walras's law is given. CASE B: (B1) The commodityfunctions Xi are hdOin pi, **, Pn-2i (B2) The bondfunction Xn-l is hdl in pi, - - , P.-2 , Pn-2 and in M (B3) The M.-function is hdl in pi,i
andnhinpi,
,Pn-2-
(B4) Walras's law is given. CASEC: (Cl) The commodityfunctions Xi are hdO in the commodity , I* pn-l and nh in pi and pi, ***, pn-l . prices P2 .. P and in pi,*, (C2) The M.-function is nh in P2 , ... Pn-1 pn-1 and hdl in Pi, **-, Pn-l and in M. (C3) Walras's law is given.
Moreover, for all three cases we have pi (i = 1,
*.-,
n -
1) 5
0. Further, in the case where Xn-I is an excess-demand function for bonds, it is considered to be a member of the monetary subsystem.8
8 Case A corresponds to the system presented by Hickman and to the system presented in Section 10 of Patinkin II. Case B corresponds to the system presented in Section 12 of Patinkin II after inserting an M,-function. Case C is the system presented in Section 11 of Patinkin II after modifying Xn into M. .
INCONSISTENCY
AND INDETERMINACY
159
Further, in economic analysis an important property is tacitly attributed to the M.X-function,namely,
ZA
l [M4(p,
Zn,)
01,
p e R,
Zn,a
e N.
4. In Section II of this paper it was shown that the explicit and implicit money functions of a monetary system have to be identical. Thus, the propositions produced by a monetary system are invariant with respect to the choice between an M.- and an Xn-function. It can be shown that for all three cases above this statement does not holdthat, in particular, by utilizing the explicit and implicit money functions, divergent results are obtained. Let us discuss the following two situations for all three cases: Situation 1: Using (A3), (B4), and (C3), X1 is expressed as a linear combination of all other functions. The fundamental set of equations in all three cases will then be
X2 =
X3 = ...
= Xn_l = Miw =
0.
Situation 2: Using (A3), (B4), and (C8), M. is expressed as a linear combination of all other functions. The fundamental set of equations will then be X1
= X2=
*
= Xn-1
0.
For Situation 1 we then obtain the result that the real parts of the fundamental equations are in general consistent, while the monetary parts are determinate in the money prices of the numeraire commodities. For Situation 2 we establish that the fundamental equations are either inconsistent in the real part or indeterminate (in money prices). This divergence reflects the fact that M-, the explicit, and X., the implicit, money functions cannot be identical in the cases under consideration-which fact implies that the various postulates in the three cases are incompatible. It especially implies that (A3), (B4), and (C3) are incompatible with the other postulates of the corresponding set.9 9The creation of inconsistentpropositionsby the sets of postulates under can be seen in the followingway: consideration Walras'slaw is definedby
(1)
A A [M.(p, Zn.a)
-
X.(p,
2n,a)
= 01,
p e R,
Zn,a e N,
or, in equivalentterms,
(2) A
A
[A[M(P,
2n..)
Xn(p,
Z,n,)
0]]*
160
KARL BRUNNER
5. The result established is independent of any considerations of consistency or inconsistency of the real parts. The analysis of consistent systems provides additional analytical insight into the properties of the postulational structures under investigation. The two following postulates will be useful to assure consistency in the real subsystem.1
and, in terms of economics, it states that for any money-supply and price vectors the sum of all excess-demand quantities will always be zero-in other words, that for any money-supply and price vectors excess demand for money in terms of flows and excess demand in terms of stocks will always be equal; or, in still other terms, that the excess demand for nonmoney objects generated by the money mechanism will always coincide at any given price and money-supply vector with the excess-demand flow for nonmoney objects generated by the nonmoney functions. From Postulates (Al), (B1) + (B2), and (C1) we derive A[Xn(p) = O], p p K CR.
The set K consists of all price vectors linearly dependent on the relative pricesolution vector. From Postulates (A2), (B3), and (C2), A El [M(p, Z,0,. p From these statements
Z,na)
= O],
p eK,
Zn,a eN.
we derive - Xn(p,
Zn.,)
(3)
Zn,a P
[M,(p, Zn.,)
o 0],
p e K,
Zn,a E N.
Propositions 2 and 3 are patently inconsistent. Thus, it follows that the subsets of Postulates (Al), (A2), (B1), (B2), (B3), (C1), (C2) are inconsistent with the corresponding postulates (A3), (B4), (C3). This result also covers the following somewhat special interpretation of Hickman's system: Suppose X, and M* always to be equal in the given universe of discourse of relevant variables and parameters. Further, let M be a monetary C M . Then we obtain the proposiconstraint imposed on the system, where tion
(4)
Zn.,a p
[Mo(p, 2n,a) - M*(p, 2n.a)
O],
p e K,
Zn,a e N.
Proposition 4 implies that there exist price vectors such that the excess-demand flows of commodities generated by the money constraint M (what else would 0 be?) will not be equal to the excess flows of commodities generated by M the commodity functions. Thus, the system implies that the same physical quantity may actually have, within a well defined matrix of conditions, two different values simultaneously. 10The development of the argument in Patinkin II and III and in Hickman's paper implies very clearly the thesis that consistency cannot be derived from homogeneity in macroanalysis. G. Becker provided, in a written discussion, a very interesting proof of this statement. We see that most participants to the discussion of the "Patinkin problem" agree that in macroanalysis some additional postulates are required.
INCONSISTENCY
AND INDETERMINACY
161
POSTULATE (AA): The particular systems under consideration have a unique solution if and only if the number of variables and independent equations coincide. (BB): For the particular systems under consideration, the POSTULATE functions of the real subsystem are related in such a way that , .(Xi, = 0 has theform 4V(0, **, 0) = O.11 X,jl) A variety of forms will satisfy the second postulate; so will any linear combination of commodity functions. As has been shown by Hickman,'2 the Jacobian of (n - 1) excessdemand functions for commodities hdO in p vanishes identically, thus implying a function (3.3) Vt(X,, * *
, Xn-1) =
0-
But this result is not sufficient to decide whether the functional dependence defined by (3.3) has the special properties required by Postulate (BB), which are essential in order to establish consistency. The second postulate above successfully eliminates a ,6-function of the form O. 0, **,0) = 0, where X1 (X, r
11 These two postulates will eliminate what corresponds to Cases (a) and (b) in Patinkin III, p. 141 of the current issue. It must be noted that Patinkin's two invalidity theorems are independent of such postulates. He achieved in this way a major improvement compared with Patinkin II. 12 Op. cit., p. 16. Actually, as was acknowledged by Hickman, this result was already stated in Patinkin II, p. 15. 13 Hickman's paper is very clear on this point. He states on page 19 that a relation of the type t(X1 , . X._.) = 0 is implied by the homogeneity postulate. If a solution vector exists for X2 = =X X.- = 0, then ,6(X1, 0, ... 0) = 0, where X* is the value of X1 for the given price vector. Then he states: "For a wide class of functions the equation of dependence will imply that X* 0, ** ,0) = 0, Hickman is zero." Because a class of functions exists such that W(0, thinks it "unwise to assume a priori that the commodity equations are independent and inconsistent." As was already mentioned, no such assumption is involved in establishing Patinkin's central theorems. There is actually no need for a dispute between Patinkin and Hickman at this point. Both state (consider Section 2 of Patinkin III) that, given the homogeneity postulate, there is a class 0) = 0, X* # 0, and another class of X1of functions such that ,6(X1, 0, ... = 0. It is obvious that Patinkin and Hickman functions such that ,p(0, ..-,0) consider that an additional assumption is needed, independent of the homogeneity postulate, in order to restrict analysis to the second class of functions, so that
.A
.
0) = 0. This is the situation
so far as functional
dependence is con-
cerned. In footnote 5, Hickman introduced the concept of "equational dependence." According to his definition, "equational dependence" exists if functional dependence assumes the special form ,6(0,..., 0) = 0. Thus Hickman is justified in claiming that both are clearly distinct concepts and that there is no relation of implication between them. Interpreted in this way, two apparently inconsistent
statements by Hickman are entirely valid: On page 19 he states: ". . . the de-
162
KARL BRUNNER
Classical theory provides both of these postulates. Although the first was rigorously introduced by Walras, the second goes back to the founders of classical theory. Say's law formulates a dependence of the excess-demand functions for commodities. It states
n--
(3.4)
i=l
piXi - 0.
(p)
This relation has to be considered as an additional postulate of Cases A, B, and C above. It will be seen that in this way a new, independent contradiction has been introduced into the postulational structure. From (3.4) we obtain (3.5) E
Zn,.
E
p
[Xn(p, Z,,,) # O],
p ER,
Zn,a EN,
and, by (A2), (B3), and (C2), we derive (3.6)

Zn,a
` 0], [MX (, Zn,a)
p ER,
Zn,, E N.
According to (3.5) and (3.6) the two propositional forms M* and Xn are different. This result contradicts Walras's law.4' 15 Combining (3.5) and
pendence of classical functions by no means eliminates the possibility that the equations may be inconsistent." This statement refers to the situation where 1(Xi, ... , Xn_) = 0 but a restriction (0O,*.. , 0) = 0 has not yet been imposed. On page 17 he states: "Dependent equations are always consistent ...." Since "equational dependence" consists of two parts--t(X 1, *, Xn_-) = 0 and 4(0, * - ,0) = 0-no contradiction is involved. 14 The fact that a system which implies Say's law provides a solution of relative prices in the real part and of money prices in the monetary subsystem seems to indicate its analytical soundness. But we have to consider in addition the logical propositions by which a solution is realized. Consider the procedure of the system. First we have the proposition (1) Determinacy A [X.(p) p = 0], p R.
can then be introduced principally in two ways:
(2)
El[p
pO], p R,
where p represents some given quantity, and A El E [M*(p, Zn.a) - 0], [M*(p, n,,,) O], p e R, Zn,a
E
(3)
N.
continued.) (Footnote continued.)
Zn,(Footnote
INCONSISTENCY AND INDETERMINACY
163
(3.6) we derive (3.7) E A , A Z,n) E [M*(p

Zn,ta P - Xn(p) $ 0].
Let us note first that determinacy means nothing else but that a unique money price has been selected from the infinite set consistent with Proposition (1). But the fact of having selected a unique price out of the infinite number of admissible prices tells us nothing about the logical structure of the selection process. Actually, the nature of such a selection process is essential for evaluating the system: While (2) above is evidently consistent with (1), (3) is obviously not. In this context Leontief's contribution to the discussion has to be considered. Patinkin stated that the classical systems with a complementary property are either inconsistent or indeterminate (after the money function has been adjusted to the homogeneity postulate). Leontief retorted by stating that there is consistency (in the real subsystems) as soon as Say's law is introduced. His case has been fully recognized in Patinkin II (Section 13). But Patinkin's theorem states further that, given Say's law, indeterminacy cannot be removed. On this Leontief remains silent. Thus, his statement is correct but irrelevant. Why does Leontief entitle his contribution "The Consistency of the Classical Theory of Money and Prices," while he analyzes a consistent theory of a real economy? If I have to interpret him as implying that determinacy is introduced "simply" by complementing the real system with a money equation, then the above criticism applies. In this case his proof of the consistency of the "classical theory" fails. And in case he removes indeterminacy by some device such as p = p0 above, then the system presented fails to satisfy the "minimum requirements" of a monetary theory. I want to acknowledge that G. Becker's paper contained a very thoughtful and elegant criticism of Leontief's contribution to the discussion. It is further necessary to remind the reader of Oscar Lange's contribution to the present issue in his paper, "Say's Law: A Restatement and Criticism," in Studies in Mathematical Economics and Econometrics, O. Lange, F. McIntyre, and T. 0. Yntema, eds., Chicago: University of Chicago Press, 1942, pp. 49-68. 16 Say's law is only one among many ways to create consistency. As shown by Hickman, we may construct a system consistent in its real part without implying Say's law. If we denote by X some well defined linear combination of the commodity functions, then we may state that Hickman assumes
0,
Xn = 0,
M = 0,
where p e R, and 3 p e R, such that Xn $ 0 $ M . But for Hickman's system the condition, -MO + Xn p 0, holds. It may be worthwhile to express this fact in concrete terms. On page 14 of Hickman's paper, Xn and M2 are defined in such a way that the proposition A
P Zn,a
[M (p, Zn.a) - X (p) = 0] is valid if and only if the following polynomial, K)]p3 - d2. p2.p2 dl.p .p2 + [S2(1 - K)]p
-
[Si(1 -
M .pl
p2 = 0,
vanishes identically. (1) or
Sufficient conditions K = 1,
for this to happen are
d2 = di = M = 0
(2)
S1 = di = d2== S = M = O.
(Footnote continued.)
164
KARLBRUNNER Argumentfor the ,3 Class
6. The following set of postulates will be subjected to analysis: n - 2) POSTULATE(A): The commodityfunctions Xi (i = 1, *., = in n-2). 1, i i , are hdO pi (i POSTULATE (B): The bond function X.-1 is nh in pi and hdl in pi and M. and in M (C): The M.-function is hdl in pi, P, p-2 POSTULATE and nh in pi, ***, Pn-2 (D): Walras's law is given. POSTULATE Postulates (B) and (C) together form the monetary subsystem. So far as the issue of dichotomization is concerned, we are not restricted in our choice of a second nh-function. Actually, a good case can be constructed in favor of Postulate (B) by utilizing either one of the following assumptions. POSTULATE (E): The net injection SB - D does not vanish; the relations SB = Sn0, Dn = D., are assumed; and MsB(p,Zno) is hdl in p and Zn,a X where Ms is the money supplied by the B-sector. POSTULATE(F): There exist imperfections in the capital market. They may take the form of special costs of transactions. The first rationalization seems to be preferable. It arises from a connection between the two monetary functions, Xn_l and Mx, in the form of transactions between the B- and P-sectors. Forgetting the quantifiers, we may write the relation
(3.8)
M
-S
S = S-
XB
and, by (E) above, the left-hand term is hdl i p and in ]Is =
This property is transmitted to relation (3.9)
XB
al Zn, . . Furthermore, by using the obvious

n-X
pn-
the same properties are further attributed to XB 1 (the B-component of the general bond equation). In this way X. is defined by the expression
n-2
(3.10)
Xn
i=l
pi *Xi +
pn-1
Xn1
pn-lXpn-1.
pn-2
This combination
is nh in pi,
P pn-2,
but hdl in pi, i,
and
M. Thus the properties of Xn and M- coincide.

It is interesting to observe further that the system presented by Hickman is equivalentto a system the structure of which implies Say's law. This assertion will not be proven here.
165
Again forgetting the appropriate quantifiers, we state

n-2
(3.11)
MP -
XIX= - Ypi*X,
i=1
where Xp-1 is the P-component of the general bond function (hdl in P, ***, Pn-2). Combining (3.8) and (3.9), we obtain
(3.12)
S-
1s
pn-1
n-1
If we subtract both sides of (3.12) from the corresponding sides of (3.11), we find, after reconsidering the quantifiers, (3.13) A A [Mx(p, 2n,a)
=
Xn(p
Zn,)],
R,
Zn, a e
N.
Thus we conclude that, in contrast to the class a, there is no contradiction within the set of propositions formulating the postulates. Thus, whatever equation is solved in terms of the others, the solution vector will remain unchanged. But closer inquiry reveals a curious type of inconsistency in the system. By eliminating either X-n1 = 0 or M.,= 0 with the help of Walras's law, we obtain a straightforward dichotomy with respect to the real variables and money variables. Now suppose that a commodity equation has been eliminated. Then we are left with (n - 3) equations in the real subsystem in (n - 2) variables. We can at best solve for all the real commodity prices in terms of r, the rate of interest. Inserting these solutions in Xn-1 = M- = 0, we have two equations, one each in r and pn-3. The first elimination procedurevia the monetary subsystem thus results in a dichotomizedprocess, while the elimination procedurevia the real part defines a system without any dichotomy. The system requires all relations to find a solution for any proper subset of variables. Thus the two ways of eliminating the redundant equation picture two essentially different notions about the working of the economic mechanism. This divergence becomes very marked when we consider the properties of the interest mechanism in the "modified Lange system." If we choose to eliminate Xn-1 = 0 or M. = 0, then we are bound to state that r is determined within the real subsystem. On the other hand, to
delete some X8 = 0 (s being any one of 1, - - *, n
-
2) implies that r is
determined by the operations of the monetary mechanisms. These statements reveal inconsistent economic conceptions about the working of the interest mechanism. In contrast to the systems in class a, the "asymmetry in the elimination process" does not show up in divergent solution vectors. We actually obtain the same solutions, but we obtain divergent propositions
166
KARL BRUNNER
about the working of the economic mechanism, about the very essence of our issue-namely, dichotomization.'6 The Xn-System 7. The group of Xn-systems will be derived if we commit ourselves once and for all to utilizing Walras's law to express the money function in terms of all the other functions. The transformation is achieved by replacing Postulates (A2), (B3), and (C2) above by utilizing Postulates (A3), (B4), and (C4) to define the Xn-function. The following cases have to be considered: a) the Xi-functions (i = 1, * , n - 1) are hdO in pi (i = 1, **, n - 1); b) the Xi-functions are hdOin p8, where ps C pi; , n - 2) are hdO in Pr (r = 1, **, c) the Xr-functions (r = 1, n - 2), while X1-, is hdl in pr; d) the Xr are hdOin pr, and Xn_1 is nh in Pr 17 Let us arrange the possible outcomes according to the homogeneity properties of X, . Then we find: 1) Cases (a) and (c) imply Xn hdl in a set of p's. 2) Cases (b) and (d) imply Xn nh in p. The Xn-systems are essentially systems in (n - 1) equations and (n - 1) variables. For each single system of this type there is a fundamental set of (n - 1) equations, which is regularly obtained whatever equation is deleted. Our problern is thus an evaluation of the fundamental set of equations of each Xn-system. The criterion will be that the set of relationships must present a consistent and determinate theory of a monetary system. An investigation of the two situations will furnish the following results: 1) Without postulating the property F(O, - --, 0) = 0 for = 0, we state with Patinkin that the system is either *... , Xn-1) F(xl, inconsistent or indeterminate. If F(0, -..-, 0) = 0 is assumed, the system is consistent but will be satisfied by an infinite set of money prices.
16 The system developed on page 22 in Patinkin II corresponds to this system with the difference (possibly) that he presented an Xn-system. In Patinkin II we find no analytical objection to this system, and in Patinkin III there is an implied criticism by his resorting to microanalysis. Patinkin argues further (Section 5) that the (implicit) money function of the system cannot be hdl in p and in the stock of money as it is in the Cambridge function or as in a system derived from microanalysis. As shown in the text, it is not true for this case. It seems important to build a direct argument against a system which is not consistent with microanalysis. 17 Case a has been presented on pages 5-6 and 14 of Patinkin II; Case b, on pages 16-17; Case c, on pages 18-19; Case d, on page 22.
167
2) Case (b) adds nothing to those already treated under (1). But an interesting feature is presented by case (d), which we see describes a dichotomized pricing process and which is subject to the following objections. First, the theory of interest involved states that r is an essential adjustment variable in one of the "real" relations. Thus, if there are an such that the corresponding X8 = 0 (s = 1, * , r and a z (= pr/pn-2) n 2), then r is unaffected by Xn_1 # 0. Such a conception of the interest mechanism strains our imagination. Second, the set of postulates under consideration does not satisfy the minimum requirements of a theory of a monetary system. It is able to account for finite prices, but fails completely to account for a nonzero stock of money. As the system stands, any stock is compatible with the value flow generated by the system. And, if the system is improved to take account of this, we are back at the modified Lange system described above. Thus, we conclude our review of the Xn-systems by stating that this set fails, even if consistency is assured, becauseit fails to meet the minimum requirementsof a theory of a monetary system.'8
IV. ASSETS AND HOMOGENEITY OF MARKET FUNCTIONS
1. A macroanalytical argument has shown us that the subset of monetary systems characterized by a complementary property is fallacious. It is possible to clarify further the nature of the fallacy by utilizing the possibilities inherent in microanalysis. In this way we can state the fundamental theorem of monetary systems as follows:19 The demand and supply functions of a monetary system with nonzero assets (including money stocks) and finite money prices will be nonhomogeneous in the set of money prices. 2. It has been stated in Patinkin I that a utility function with money included is a necessary condition for a solution in money prices to exist.20Actually, as conceded in Patinkin III, the inclusion of money in the utility function is not a necessary condition.
18 It may be necessary to state that the system (5.1)-(5.3) presented (Section Patinkin III is essentially different from the one discussed above. It is able in 5) to account for a definite nonzero stock of money and finite p's without creating an analytical impasse because the complementary property is not involved. 19We skip the proof of the fundamental theorem. Interested readers are referred to Patinkin III, Section 5. 20 A clarification of some issuse which have arisen in the context of Patinkin I may be appropriate here. Patinkin analyzed the necessary and sufficient conditions for the consistency of a system the utility functions of which exclude money and for which pn = 1 holds. The result was that the stock of money must be zero. Phipp's (op. cit.) objected, arguing that to exclude money from the utility function and to postulate pn = 1 involves a contradiction, while Patinkin
168
KARL BRUNNER
Using the results of Patinkin I, it is easy to prove that a system in which utility functions do not include money and which are not subject to any side relation in addition to the budget restraint will have, for any nonzero money stock, infinite money prices-with a corresponding infinite circulation velocity. If we start our analysis with a nonzero stock of money and wish to derive finite prices, then the system has to provide some mechanism which keeps velocity at some finite level. This can be achieved by various processes. Two versions of the analytical possibilities have been worked out in Patinkin I and III. Both versions are characterized by the fact that positive total and marginal utility is
retorted that Phipps is speaking about a nonexisting entity. The dispute can be organized in the following way: Let U" be a utility function Ua(Zla , *. , Zn-,,a) maximized subject to Zi, piZi.a = Z_l piZi,a . The form of this budget restriction was suggested by Patinkin I. The first-order conditions are Ul7/Ua_i = pr/pn-l (r = 1, *. , n - 2). We have, further, Z,,, = 0 and the budget equation, which might be rewritten in the following way:
n-2 p. n-2
z i-l1 Pn-l
-Zia Zn+
-l.a
2 il
Zn.l .-Z,a Pn-1
a -=
P Pn-l
0,
where pn is the price of money in terms of some num6raire, which may be a commodity or an abstract entity. If we solve for the dependent variables Zi, in terms of p and 2i,, and aggregate the resulting functions over all a's, we obtain the system of market excessdemand functions:
X Pi Pn IPn-2 Pn-1 .g2n_ = = y Z_== (i = 1. ... n -
I),
pn-l
Pn-1
Pn-1
pn X
pn
Pn-i
- O (Zn -Zn) =
P Z
Pn-1
= 0.
It is evident that, for this system to be consistent, the last term in the excessdemand functions for commodities must vanish, giving Zn = 0.
Pn-I
This condition can be fulfilled by the following three situations: (al) pn = 1, (bl) pn = 0, (cl) pn - 1, (a2) 0 < pn-1 < oo, (b2) 0 < pn-_ < o, o, (c2) pn-1 (a3) Zn = 0, (b3) Zn i 0, (c3) ,Zn 0.
System (a) is Patinkin's solution and (b) is Phipps's. Both of their contentions are correct, granted that the situations not considered by them are excluded. The three situations are actually equivalent properties of a "money-rejecting" system, where money has no function whatsoever. I do not think it advisable to exclude any one of the situations, nor am I convinced that any one way of presenting the logic of Patinkin's system B in Patinkin I is analytically unsound. I agree essentially with Marschak's statement that there is no need for a dispute here.
INCONSISTENCY
AND INDETERMINACY
169
attributed to cash balances. This postulate is considered to reflect some significant economic facts. The first variant includes nominal cash balances, while the second considers real cash balances. In this way a mechanism is introduced which induces economic units to hold circulation velocity at a finite level. 3. Marschak has shown by an interesting argument2' that we can derive a demand for (nonzero) money stocks without postulating money as part of the utility function. If the units have a horizon extending over a multiplicity of marketing periods, it will be rational to hold money balances at any given moment as soon as money prices are expected to be lowered. Therefore, in such a system circulation velocity will be kept at a finite level as long as at least some units have this type of expectation. Marschak was able to develop still another mechanism for trapping velocity without putting money into the utility function. The starting point was given by the important notion of the "cost of transaction." This phenomenon was expressed by a very ingenious device. While keeping the assumption of perfectly competitive markets, in the sense that the exchange price is considered by the individual to be unaffected by his supply and demand decisions, the net price is regarded as a decreasing function of the amount transacted. The net price differs from the exchange price by the costs of transactions. By this construction Marschak is able to utilize some parts of the apparatus of imperfect competition to prove that velocity will be kept at a finite level without the insertion of money into the utility function. 4. Let us suppose that money is not in the utility function, but that there exists, apart from the budget restraint, another equation relating money to the objects actually entering the utility function. Such an equation has to be considered as a second restraint subject to which utility will be maximized. This procedure allows different variants with varying degrees of empirical content and usefulness. At this point, the simplest procedure will be chosen in order to clarify the essential structure of this type of argument.22 The utility function contains all objects with the exception of money:
(4.1)
ua = Ua(Zl,aa *
Zn-1,a)I
and the budget equation is given by
(4.2)
E
s=l
ps(Zi.,-
i,a) = 0
pn
1.
21Marschak, op. cit.,

22 I had a highly stimulating discussion on this point with Donald Fort. Further, I am indebted to C. Hildreth for important suggestions as to the analytical argument.
170
KARL
BRUNNER
To realize our objective, U must be maximized, not only subject to (4.2), but subject as well to some restriction with respect to the rate of utilization of given stocks of money. Thus we postulate (4.3)
Zn,c = kc r-1
X pr(Zr,ec-Zr,a),
> 0. The summation on the right defines where s < n - 1, Zr,a-Zra the total value of the individual's purchases of all commodities for which he has a net demand. ka is a given constant determined outside the economic nexus.23 Solving the budget equation for Zn,a, and combining the resulting expression with (4.3), we may establish
(4.4) where s Zn,e n
-
n-1 Ej pj(Zj,,
j=i
- Zj,a) = ka
pr(Zr,a
-Zr,a),
r=1 -
Zr, > 0. Relation (4.2) has to be maximized subject to (4.4). The first order conditions are
1, Z, U
+ Xpr(1 + kc)
pk =0
(r < s (k = s + 1,*,
n-
1),
Ub,+
n -1) Zr,a > 0
Let us order the commodities in such a way that all cases Zr,a
comprise the first s of n
Zna n-1 -
1. Then let us choose (n

pn-1.
-Zj,)
1) as numgraire
and divide (4.4) through by

(4.6)
Pn-1
Thus we obtain
= kg
_P
jp1 Pn-1
(Zj
X P' r=1 Pn-1

(k =
(Zr,a -
Zr))
and from (4.5) we derive

47) U/n_ Uka/Un-1 = (Pr/Pn-1)(l + ka) (r=
1,
...
s), -
= pk/pn-I
s + 1,
,n
2).
23 In an unpublished paper by Donald Fort there was an able discussion of the economic significance of the timing pattern of the flow of payments through a cash balance. We may conceive of the k as a simple characterization of such timing patterns. Donald Fort's arguments were essentially organized around the problem of the timing patterns of the flow of money in and out of cash balances. If we wish to acknowledge these facts but treat them by a short cut-without incorporating them as an additional variables-we proceed as above within a static framework. But within static theory it is already possible to do much better and to have velocity as a variable emanating from the solution of the whole system. The essential contention is that, independent of the analytical treatment given to the problem of timing structures raised by Fort, a properly constructed monetary system will not possess the complementary property. Whether timing structures are indicated in the form of a compound variable within static analysis or dealt with in terms of dynamic analysis has no relevance for the issue at hand.
171
Relations (4.6) and (4.7) together form (n - 1) equations in (n - 1)
unknownsZi,a (i = 1, - * , n - 1) for the ath individual.By inspection of the system it follows that pi (j = 1, * , n - 2), pn-l, and Zn,c
.
always appear in ratios. Clearly, the solutions of Zi,, in terms of pj, and the money stock Zj, a) and Zn, , are hdOin the prices pj, pn-i, pn-1X Zn,a )but nh in p alone. Net demand and supply are then defined by (4.8)
Dr,a
= Zr, a Zk,a -
Zr,a (r = 1, ... , s < n - 1;

Zk,a (k =
Zr,a
2r,a Zk,a
>
0),
(4.9)
Sk,a =
s + 1, *,n
1; Zk,a
> 0).
By summing over all a's and all commodities, we derive the market (net) demand and (net) supply functions. By summing over all money restrictions, the market function of money is found. These results are organized in the following system: (4.10)
(4.11) 19k = D; (p-, *, p-,
p
n
(i = 1,***
1);
Si=
Si (pn-2
(4.12)
(4.13)
Di =Si;
n-1 Zn = K-pn
E i=1
Pn-i
Si;
(4.14)
Zn = Zn.
This system eliminates any homogeneity in prices, together with the dichotomy in the pricing process. The fact of a "meaningful" money equation is sufficient to make the commodity functions, if properly derived, dependent on the real values of assets. Systems of this type are free from both invalidities; they are consistent with Walras's law; and, furthermore, they do not present conflicting conceptions about the basic mechanisms of the economic process.24
V. CONCLUSION
Patinkin raised an interesting and important problem affecting our fundamental notions of the pricing mechanism. The critical reactions to his analysis showed that there was some need for inquiry into these
24 We may state, on the basis of these results, that Hickman's mistake was to impose a constraint on the system which was not reflected in the behavior equations for the commodity markets. His postulate of M' = 0 asserts that there exists a constraint, while in his Xi-functions he asserts just the contrary. This contradiction is avoided above by the way in which the market functions have been derived; but then dichotomization disappears.
172
KARL BRUNNER
problems. Patinkin himself contributed a very thoughtful and compact restatement of his position. The following summary of the arguments developed in this paper is to be viewed essentially as a "complementary commodity" to his main thesis. We observed that dichotomization arises in systems combining one or two functions nh in the p's (one of which is the explicit excess-demand function for money) with functions h in some subset of the p's. For the set of systems combining (n - 1) excess functions h in some subset of p's with one function nh in p, we observed that the postulates were inconsistent with Walras's law. Systems combining (n - 2) excess-demand functions h in some subset of p with two nh functions were free of this type of logical fallacy. But they imply at one and the same time that the pricing mechanism is dichotomized and that there exists no dichotomy. Thus, this type creates inconsistent propositions about the very essence of dichotomization. Such a conflict in basic notions about the nature of the economic process within one system is not conducive to satisfactory analysis. On the other hand, a monetary theory cannot be based upon an Xn-system. Hence, we may formulate the general statement: Monetary systems with a complementaryproperty cannot be accepted as explanatory devicesfor the following two reasons: (a) either they are inconsistent with Walras's law, or (b) they do not satisfy the minimum requirements of a theory of a monetary system. Macroanalysis is able to prove the inadequacy of the systems under consideration. But we have to resort to microanalysis to derive proper guiding principles for the construction of monetary systems. We realize that there is no need for inserting money into the utility function in order to build up a consistent monetary theory. Considerably more investigation is required to clarify the relative usefulness of various approaches and their economic implications. If any properly constructed monetary system is considered, it will be seen that the explicit and implicit money functions will coincide. This is a necessary condition of proper construction.25 The clarification of this fact is an important contribution of microanalysis.
25 It is not sufficient. For instance, we may define a macrosystem its functions nh in p and then postulate Walras's law, so that Xi(pl,
-, pn-l)
with all of n 1).
= 0
(i == 1, .
But, from the point of view of economic analysis, this procedure is thoroughly unsatisfactory. By resorting to microanalysis we can do better than just stating nh in p. An important economic meaning will be attached to the resulting nhfunctions. Thus, we may state as a construction principle of monetary theories that any macrosystem ought to be consistent with microanalysis. Such a procedure will eliminate all types of inconsistencies mentioned. Artificial and arbi-
173
How is classical theory affected by our results? We have to realize that classical theory is composed of two unrelated segments. One contains a theory of a real economy and the other, a theory of a monetary economy-and there is no simple step from one to the other. They are two completely different theories with different implications, each one appropiate for its specific range of phenomena. Handelshochschule,St. Gallen, Switzerland
trary assumptions will be avoided, and a maximum of economic meaning can be gained in this way. The insufficiency of the coincidence of the explicit and implicit money functions is further clearly evidenced by the modified Lange system described in Section III-6.

Inconsistency and Indeterminacy in Classical Economics

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Inconsistency and Indeterminacy in Classical Economics

Uploaded by

Copyright:

Available Formats

Inconsistency and Indeterminacy in Classical Economics Author(s): Karl Brunner Reviewed work(s): Source: Econometrica, Vol. 19, No.

INCONSISTENCY AND INDETERMINACY CLASSICAL ECONOMICS*

we mean that, for all x such that x e U, it is true that P(x). By

wherep eR D Ra, Zn,a EN

n,a , Dn,a,, and Sn,a.

(p, Zn) ]S n,n,a)) , Z ,

Further, in economic analysis an important property is tacitly attributed to the M.X-function,namely,

[Mo(p, 2n,a) - M*(p, 2n.a)

0) = 0. This is the situation

[Xn(p, Z,,,) # O],

and, by (A2), (B3), and (C2), we derive (3.6)

` 0], [MX (, Zn,a)

can then be introduced principally in two ways:

INCONSISTENCY AND INDETERMINACY

(3.6) we derive (3.7) E A , A Z,n) E [M*(p

vanishes identically. (1) or

for this to happen are

KARLBRUNNER Argumentfor the ,3 Class

and, by (E) above, the left-hand term is hdl i p and in ]Is =

This property is transmitted to relation (3.9)

al Zn, . . Furthermore, by using the obvious

but hdl in pi, i,

M. Thus the properties of Xn and M- coincide.

INCONSISTENCY AND INDETERMINACY

Again forgetting the appropriate quantifiers, we state

INCONSISTENCY AND INDETERMINACY

Zn.l .-Z,a Pn-1

and the budget equation is given by

21Marschak, op. cit.,

n -1) Zr,a > 0

1. Then let us choose (n

and divide (4.4) through by

X P' r=1 Pn-1

and from (4.5) we derive

INCONSISTENCY AND INDETERMINACY

Relations (4.6) and (4.7) together form (n - 1) equations in (n - 1)

Zr,a (r = 1, ... , s < n - 1;

with all of n 1).

INCONSISTENCY AND INDETERMINACY

You might also like