Professional Documents
Culture Documents
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.
Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to The Quarterly
Journal of Economics.
http://www.jstor.org
I. INTRODUCTION
FUNCTIONS
II. PRODUCTION EQUILIBRIUM
ANDPRODUCER
The reason for requiring pi>O will be discussed presently. If, now,
(2.4) is substituted into (2.3), we obtain
(2.5) yj=KiZiPt (i=1, 2, . . . , n).
This will be called the objective production function in the ith in-
dustry, whereas (2.3) will be called the subjective production func-
tion. (In the dynamic analysis of Section VIII, (2.5) will be identi-
fied with the long-run production function, and (2.3) with the short-
run production function.) In analyzing the criteria for economic
welfare, it is of course the objective production function (2.5) that
must be used; on the other hand, in order to explain the behavior of
the firms in the industry, it is the subjective production function
that is relevant. Both are relevant to the analysis of the determi-
nateness of competitive equilibrium. The essential feature of the
model is that ki will be treated as a parameter by each firm, and
consequently the collective behavior of the firms in industry i will
not take the relation (2.4) into account.
As an alternative to expressing the scale effect ki in terms of
the amount of employment as in (2.4), we may equally well - and
more in the tradition of Pigou 3- express it in terms of industrial
output as follows: 4
(2.6) k, = caky.; -oo < ni<l
(the restriction on vi will be discussed presently). This formulation
is also more appropriate for extending the analysis to more than one
factor. Substitution of (2.6) in (2.3) gives
1 1
(2.7) 1-? o Zi
Y'=/Xi 1-Z a,
which is the same as (2.4) if we set pi=1/(1-pi) and Ki=/Le i. The
reason for imposing the condition - oo <i <1 is now evident: if
v= 1, output would be unbounded, and at the other extreme, vi=
-00, it would be a constant, independent of the quantity of labor.
Likewise, the equivalent condition O<pi< oo in (2.4) expresses the
assumption that the objective marginal product of labor
(2.8) dy-= pJKJZ4
dzi
3. "An Analysis of Supply," op. cit., and the 3rd and 4th editions of
The Economics of Welfare, Ch. XI and App. III.
4. Professor Herbert A. Simon has kindly brought my attention to the
fact that he introduced a similar formulation in his unpublished paper, "Some
Models for the Study of the Economic Effects of Technological Change,"
Cowles Commission Discussion Papers, Economics No. 213, University of
Chicago, Dec. 16, 1947, pp. 12-13. He further noted (p. 13): "We see that
stability from the standpoint of the individual entrepreneur . . . is entirely
compatible with increasing returns from the standpoint of the entire econ-
omy. . . ." (The word "stability" was used here in the sense of a stationary
maximum point, rather than of dynamic stability.)
(2.12b) ( )
where (2.12b) is obtainedfrom (2.4) and (2.5) since
(2.13) ki= ICZiP4= Ki) (P 1)/PI = KS1/P4y.(Pi 1)/P4.
(2.17)
w wdy l, =
_= 1y Kc'/Pjyi(lp4)/p=S4(y )
where the pi are the market prices. Since prices and quantities are
intrinsically nonnegative, in view of the utility function (3.7), ob-
viously equality must hold in (3.9). Setting marginal utilities pro-
portional to prices, we arrive at the well-known rule that the pro-
portion of each consumer's income that is devoted to the ith com-
modity is equal to Gi:
(3.10) piyi=iwZ (i=1, 2, . . ., n).
The demand curve for the ith product is defined (for positive w)
by
(3.11) Pi (yi)
w yi
These are shown below in the two diagrams of Figure I (for w=
Z=1).
91i.5; PIr.8
2~~~~~~W1
, IsZ -
PI Zi\
\ t ' .304 or
\ \ Xo",ro.13, rl..., 2
* I
.S*II
Ct _+ _
_._..i I I~~~~~~~1
FIGUIM
e~-.5 e2 -1.5
-y T -. 304 or
P2+<T2e' os3
S2
P2*
I
~~~~~~M2
yi . . XI
0 Y2' Y2 Y2
P~~~~~~~~~~~~~~~~~2
FIGURE lb N
and from (3.10) and (4.2) we obtain as in Section III the demand
curves
OZZ
These Pi- =_ (y.Di ) 1(i=1 , 2, . . . n)
rium"solutionsin the sensethat they are valid only for given values
of the remainingvariables;on the contrary,for each i, the two equa-
tions (2.17) and (3.11) fully determine the equilibriumquantity
yMtproduced and consumedof the ith commodity, as well as its
equilibriumreal price pit/wt, independentlyof the remainingquan-
tities and prices. These of course are the same solutions as given
by (4.7) and (4.8).
OUTPUT
V. IDEAL
In view of the assumptionthat all consumershave the same
homogeneousutility function (3.6), the correspondingaggregate
function (3.7) coincides (assumingthat the appropriatelump sum
transferscan be made without cost) with the social utility function
in Samuelson'ssense-3 whatever the particularform of the social
welfare function may be. A situation will thereforebe Pareto opti-
mal if and only if (3.7) is maximized. It is to be maximizedsubject
to the productionpossibility set which, from (2.5) and (3.8), is
given by
(5.1) Y= { (Yl, Y2, . .. , Yn) Zi y_);ps !t
yoyo.
i=1 KjX
both in log Y, we must show that for any A, O<X<1, the convex
combination XY+ (1 - x)Y' is in the interior of log Y. From (5.2)
and (5.4), if Y, Y'Elog Y, then
Y2
Y2'>Xu
0 Y,*
FIGURE Ha
(5.6)
i-l Kj i-1 Ki )i Ki i= Ki
equality holding if and only if Y= Y'. Now the fact that (5.7) fol-
lows from (5.6) is a simple consequence of the well-known inequal-
ity between the geometric and arithmetic mean: 4
(5.8) (Yl/p,)X\(y."/Pt) I-X.< xyil/pt+(-)8t/i
4. Cf., e.g., G. H. Hardy, J. E. Littlewood, and G. P61ya, Inequalities
(Cambridge: at the University Press, 1934), p. 17,
log Y2
log y1Is
T
logy1
<---w?9Y2~~~- -- *0Y2
As equality holds in (5.8) if and only if ye= yi', the set log Y defined
by (5.4) is strictly convex.
The "ideal output"5 may now be found by maximizing (3.7)
subject to (5.1). Define the Lagrangeanfunction
(5.10) L (i, Y2, , Yn; A) = i1 Yi KX j= 1 I/Pi-Z'
Then
(5.11) =-u-- / =0,
D~yi yi Pi i
whence
Y
(5.12) =ipi AX ) =Azi
j-1
or, from (4.4),
(5.15) o*= (i=1, 2, n),
j=p1
which may be contrastedwith the laissez-faire solutions (4.3) and
(4.5).
The optimal outputs, or ideal outputs in Pigou's terminology,
are now determinedfrom (5.14) and the objective function (2.5):
pi
(5.16) yi* = Kj(OZ)P'.
Comparisonof (5.16) and (4.7) shows that the optimal and laissez-
faire outputs differ by a factor of (pi/ I Ojpj)Pi. Thus we arrive at
j. 1
the simple rule:
where
n
(5.18) P., X Aps.
j=1
That is, optimal output of the ith product is greater than, equal to,
or less than laissez-faire output according as the degree of homogene-
ity of the production function for the ith commodityis greater than,
equal to, or less than the weighted average of degrees of homogeneity
,
( (adsGu)=*
i
yi
= (0Zgi yi
* ,,
i.e., satisfying
(6.6) gt(W) = ($-It) tps( -It)CO*pt O
togetherwith the condition
(6.7) i=
Ethe I
For pi?C>iY9Ojpj it is easily verified from (6.6) that gi(gt) has ex-
actly one positive root, but it could have several for pi> Y, lmjpj. .
(6.12) =
i. lipyi 1TirAWZ = 0,
i.e., subsidies paid are balanced by taxes collected. The optimal ad
valorem tax rates are obtained by substituting (5.15) in (6.9):
(6.13) ri* = 1- = 1- P Pi
di n P
j= ioPj
Let us turn to the optimality criteria developed by Pigou and
Kahn. Denoting the objective marginal cost by
(6-14) x, __dC= Wp Ki- /p, (1-P)
dy W
and the market value of the objective marginal product by
(6.15) =
d= iPypiKiZi -
(see eqs. (2.15) and (2.8), respectively), and substituting the pro-
duction function (2.5) in (6.14), we find upon making use of (2.5)
again, as well as (3.10), (4.4), and (6.8), that
Pi As dsps Pi
(6.16) -=-= Pi.
7ri IV
qW
Under a regime of laissez-faire this becomes, for any fixed w >0,
(6.17) Pit =t
Wit w
in view of (4.5), whereas optimality requires
(6.18) * = w -=e=1JPJ=P'
other hand, they have failed to provide much in the way of precise
fiscal proposals for the implementation of marginal cost pricing,
leaving some doubt as to where the issue rests.3
The case against the proportionality rule when interindustrial
relationships are taken into account was argued extensively by
L. W. McKenzie.4 A succinct argument was presented by J. de V.
Graaff,5 which may be summarized as follows. Let the objective
production functions (2.5) be replaced by
(6.19) Xi = hi (xl, XQ2, . X. ., Gino ZO; YJ = Xi -:4 i= 1
where xij is the amount of the jth input used to produce the ith out-
put, and xi is the gross output of the ith product. When one assumes
ft to be continuously differentiable, minimization of cost
(6.20) Ci = P1Xi1+P2Xi2+. . . +PnXin+WZi
happens to be unity). On the other hand a problem does arise in the case of
self-service or leisure (to be discussed presently) since these are activities that
take place outside the market. Similar difficulties beset the arguments pre-
sented by I. M. D. Little, A Critique of Welfare Economics (Oxford: at the
Clarendon Press, 1950), pp. 135-36; 2nd ed., 1957, pp. 143-45.
7. The necessity of introducing such a tax was perceived by Aoki, op.
cit., in his extension of the present model to allow for interindustrial relation-
ships and variable supply of factors of production. The discussion presented
in the following paragraphshas been influenced by his work, in which the tax
rates given by (6.30) and (6.31) below are shown to apply to the general case,
provided to 2 p.*y,* is levied in lump sum fashion as in (6.35).
i.e., the price of leisure would have to exceed the wage rate. To the
extent that leisure is a private activity, this obviously cannot be
Marshall's consumer's surplus notions, may be said at best to have been in-
complete derivations, and at worst may be said to be absolutely incorrect
statements which, by a pun or play on words, seem to resemble the Pigouvian
doctrine concerning industries with external economies and diseconomies. In
its earlier form the Pigouvian doctrine is close to that of Marshall, but from
the writings of Knight and Pigou himself, we know that earlier form to be
quite wrong." (Foundations, p. 196.) See also the skeptical remarks by Milton
Friedman in "The 'Welfare' Effects of an Income Tax and an Excise Tax,"
Journal of Political Economy, Vol. 60 (Feb. 1952), pp. 25-33 (p. 33), reprinted
in Milton Friedman's Essays in Positive Economics (Chicago: University of
Chicago Press, 1953), pp. 100-13 (p. 113).
4. Principles, Mathematical Appendix, Note XII bis, p. 845.
5. In the Principles, p. 100, he recognizes that, for instance, the demand
price for tea will be a function of the quantity of coffee, but observes in a
footnote that it is convenient to group commodities together so as to eliminate
such cross-elasticities.
6. If the social welfare function U=F(U1, U2, . . . , Ur) is symmetric,
then from the assumption that all people are identical in their capacities and
tastes, their incomes will be equal in equilibrium, and consequently so will
their marginal utility of income dUt/dI,=mu; further, no lump-sum transfers
would be needed to bring this about. More generally, however, it can only
be said that dU/dIi is the same for all individuals, since the dU4/dI can be
different; that is, it is only the "marginal social significance of every dollar"
in Samuelson's terminology ("Social Indifference Curves," op. cit., p. 11) that
yields 0j (yi) = coj+ cli log yj where coi and cl0>0 are constants.
Sincethe same argumentappliesto each j, (7.1) must be an increas-
ing linear transformationof
(7.5) U(y1, Y2, Y,y) =E- l6 log y, (6>0, Es 1j= 1),
for which = 1/I. The inverse demand function correspondingto
(7.3) is then
(7.6) Pi
I Yi
A natural definition for the consumers'surplus obtained from
the ith commoditywould be the contribution8,q,(yi) = oi log yj of
this commodityto total utility, which could be written as
Marshall'sdefinitionwas, instead,7
(7.8) L'dy,
0
Unfortunately,the integral in (7.8) diverges;however,a close scru-
tiny of Marshall'suse of the concept reveals that only increments
in consumers'surplus enter in any essential way into his analysis,
and since the secondterm of (7.8) is the constant O,,we may just as
well choose the definition (7.7). Equivalently we may define con-
sumers'surplusas
(7.9) s (y,) =- f/
yid(Pj) = J P4/' d(pi/I)
= log PI-log es)=o log y,
is equalized. To the extent that maximization of social welfare required a
system of lump-sum interpersonal transfers, this of course would be quite
apart from and supplementary to the taxes and subsidies being discussed in
this paper.
7. In his "The Rehabilitation of Consumers' Surplus," Review of Eco-
nomic Studies, Vol. 8 (Feb. 1941), pp. 108-16, J. R. Hicks made much of the
distinction between Marshall'sverbal and mathematical definitions (Principles,
pp. 124 and 128n); according to his interpretation, they coincide when the
marginal utility of income is constant. As Samuelson showed in the Schultz
memorial volume (op. cit.), 1A cannot be invariant with respect to all changes,
and in the present case it is invariant with respect to quantities consumed, for
fixed income and utility scale, so presumably Hicks's distinction does not
apply. A related distinction has been emphasized by Milton Friedman, "The
Marshallian Demand Curve," Journal of Political Economy, Vol. 57 (Dec.
1949), reprinted in his Essays in Positive Economics, op. cit., pp. 47-99.
Friedman's interpretation is that Marshall's demand curve is drawn on the
assumption that other prices adjust so as to keep "real income" (a euphemism
for utility or welfare) constant. However, it is hard to believe that in analyz-
ing the effects of taxation on welfare, Marshall would have employed a concept
that assumed welfare to be held constant throughout.
or
(7.16b) Tri<O0 and lo lTj <Pi.
l og (11-Ti)
Define the function
I -T
(7.17) g(T)= log (1-T)i
li if T0.
proportionality rule, and optimal excise taxes of T1*= 20%o,72* =-50% to-
gether with an income tax of To*=13% under marginal cost pricing.
9. Differentiate g(r); the result is negative if and only if
log (1-r)>-r/(1-T).
To prove this inequality for r<O, integrate both sides of the inequality
g (T) increases without limit as r-* - oo, and also g (1) = 0. Therefore
there exist unique T +>0 and T4 <0 such that g(Th)=pi and
t - -1.145 0 Tl * 371 l
FIGUREIII
I~~~~~~
(l2 -t).P,~~~~~~~~~
(1-,r2)GA#
01 .25 -.75
1-Ti- _P -
*.8 P2- 1.5
It has already been shown that the Marshallian criterion for sta-
bility is met, that the supply curve always cuts the demand curve
from below. This will now be shown to carry over to dynamic sta-
bility in a general equilibriumframework.
The distinctionmade in Section II between subjective and ob-
jective productionfunctions actually becomesmore plausible when
put in a dynamic context. Instead of assuming that the objective
productionfunction (2.5) holds for both variables simultaneously,
let us replace (2.4) by
(8.1) dt) =a4[k'(t)-k4(t)]; aj>O (i=l, 2, . . ., n)
(8.10) E ~z = Z.
Now we naturally assume
(8.11) Y.".izi(t) -<Z,
so that (8.8) in conjunction with these two conditions forces equality
in the second, i.e.,8
{s.12 nr izi (t =Z.
UNIVERSITY OF MINNESOTA