A Paradox of Budgets: The

Postwar Stabilization of American

Neoclassical Demand Theory
Philip Mirowski and D. Wade Hands

I became an economist quite by chance, primarily because the analysis was

so interesting and easy—indeed so easy that at first I thought there must be
more to it than I was recognizing, else why were my older classmates making
such heavy weather over supply and demand? (Samuelson 1992, 236)

Heavy Weather over Supply and Demand

It is now generally accepted that the neoclassical hegemony in the
American economics profession dates from the period immediately fol-
lowing World War II. The standard story about the rise of this postwar
orthodoxy is a relatively simple tale about the smooth transatlantic
transmission of neoclassical ideas. It seems that Marshallian and Pare-
tian economics simply crossed the Atlantic and took root in American
soil with little or no transformation. For a variety of reasons, this sim-
ple story about the rise of the American orthodoxy — a kind of neo-
classical creation myth, if you will — seems to have been accepted
without any serious examination of the historical events that consti-
tuted the postwar parturition. Recently, the two of us have begun the

We would like to thank the many individuals who provided useful comments on this essay
during the 1997 HOPE conference and to give particular thanks to Ross Emmett, Dan Ham-
mond, Steve Medema, Perry Mehrling, and Mary Morgan for their remarks. This essay is an
abridged version of the original conference paper; the longer version is available from the
authors.

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 261

initial stages of a historical inquiry that will provide a more detailed

account of the rise of the American orthodoxy, and this essay is one
installment of this long-term project.
The first problem one encounters with this standard story is that the
period preceding the establishment of the neoclassical orthodoxy is not
a period dominated by a single price theory that might be progressively
displaced. It is a period of extraordinary theoretical diversity, with not
only a variety of non-neoclassical, particularly institutionalist-inspired,
approaches but also a number of different neoclassicisms. For this and
other reasons, we will argue that the secret to the stabilization of neo-
classicism in America can be found in its persistent inability to enforce
any monolithic orthodoxy in such a critical area as the formal treat-
ment of demand theory. Although this undoubtedly owes something to
broad cultural trends, we will temporarily resist such historiographic
temptations and focus on one seemingly narrow analytical issue: the
treatment of income constraint and income effects in standard “pure”
neoclassical consumer theory.
To reduce the motley of prestabilization American economics to
manageable proportions, we will focus on a particular subset of econ-
omists who shared some similar theoretical orientations. It seems that
most of the few destined to become the progenitors of the postwar sta-
bilization shared an initial fascination with the problem of the proper
reconciliation of what was called the “law of demand” with the Wal-
rasian-style maximization of individuals’ utility with interdependent
commodity prices. They each believed, to varying degrees, that the
British Marshallian tradition had ceased to provide an adequate scien-
tific foundation for the theory of demand and sought, each in their own
way, to find a more appropriate but still broadly neoclassical basis for
the “law.” One of the reasons they deemed this particular aspect of
price theory a significant obstacle to scientific progress was that they
were each deeply influenced by two individuals who would turn out, in
retrospect, to be the obligatory passage points for mathematical eco-
nomics in America in the 1930s: Harold Hotelling and Henry Schultz.
We have described this initial sequence of events in detail in an earlier
article (Hands and Mirowski 1998), but we will recapitulate their prob-
lem situation. Our task here is to pick up the narrative thread where our
previous article left off.
In short, we assert that in 1938 – 55 a number of economists inno-
vated an array of solutions to the (multiple) problems of the relation-

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 262 Philip Mirowski and D. Wade Hands

ship of the law of demand to neoclassical utility maximization and,

more to the point, managed to institutionalize those solutions at three
key centers of economic research: the postwar University of Chicago
Economics Department, the Cowles Commission while it was located
at the University of Chicago (1939 – 55), and Paul Samuelson’s Massa-
chusetts Institute of Technology (MIT). These innovations were note-
worthy in that, however much the individual solutions were rooted in
roughly the same broad intellectual tradition, they differed strikingly in
a number of ways, most significantly in their treatment of the interde-
pendent character of income constraints and income effects for the law
of demand. Far from standing as a minor technical differentia or the
quirks of expression that inevitably beset any communal intellectual
project, this speciation within neoclassicism is significant and can be
used to illuminate all sorts of other high-profile but seemingly unre-
lated controversies within the American orthodoxy. These divisions
also help us understand the sheer resilience, strength, and vibrancy of
the neoclassical school — to extend a biological analogy, it seems that
an interlocking competitive ecosystem proved more hardy than the var-
ious patches of monoculture it replaced.

The Hotelling-Schultz Impasse

In depression-era America, very few economists thought that the solu-
tion to the appalling array of economic problems — crop failures, wide-
spread unemployment, falling prices, the concentration of power in
large trusts, calls for national economic planning, and so on — would
come from a rigorous reexamination of the foundations of the utility
maximization – based theory of demand. In fact, a large contingent
believed precisely the opposite: that recent experience had effectively
repudiated any economic theory based on such abstract principles.
Nevertheless, this is the very project that Hotelling initiated in his 1932
article on the Edgeworth taxation paradox and prosecuted with singu-
lar intensity in tandem with Schultz from 1932 until Schultz’s death in
November 1938.
As we argued in Hands and Mirowski 1998, Hotelling and Schultz
discovered in each other the perfect complement to their independently
chosen research programs in economics. Schultz’s interest in the empir-
ical estimation of demand functions began under his mentor Henry
Ludwell Moore at Columbia University, but later at the University of

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 263

Chicago, he came to the position (opposed by Moore) that such func-

tions had to be based on the mathematical approach of Léon Walras and
Vilfredo Pareto (Mirowski 1990). Schultz was searching for a tractable
mathematical economics that would acknowledge pervasive interde-
pendence at the market level but also that could actually be used to
diagnose the real problems of agricultural production in the 1930s.
Schultz founded a statistical laboratory to conduct the empirical
research but also worked fervently to ground the derived empirical
relationships in the type of individual maximizing behavior he felt
would guarantee that there were underlying and dependable “laws”
behind the estimated functions.
Hotelling first became involved in estimating agricultural demand
functions during his work at Stanford’s Food Research Institute in the
late 1920s. His initial interest was in Fisherian hypothesis testing (an
enthusiasm shared with Schultz), but he also developed an appreciation
for the mathematical neoclassical theory that buttressed this effort. As
a member of the Columbia Economics Department in the early 1930s,
he not only became involved in the effort to produce a mathematical
theory that would underwrite demand functions but also argued that
such foundations would be relevant to practical issues like the evalu-
ation of welfare and diagnosing the causes of the Great Depression.
Nowhere was this dual commitment more evident than in his 1932 arti-
cle on the Edgeworth taxation paradox. Edgeworth’s demonstration
that the imposition of a tax on a particular good could actually lower
both its price and the price of a related good seemed to offer a serious
challenge to the theory of demand, and Hotelling’s article was an
attempt to propose an alternative neoclassical model that would pro-
vide better foundations and avoid these (and other) problems.
Schultz was the Journal of Political Economy (JPE) editor of Hotel-
ling’s article, and this contact initiated an intensive correspondence
about the statistical consequences of his demand theory. As we described
in detail in Hands and Mirowski 1998, Hotelling proposed two novel
ways to “derive” a demand curve: one (subsequently ignored) from a
cumulative normal density function and the other from an unconstrained
optimization of the quantity U-ȸpx. The latter, which Hotelling called
his “price potential” model, in direct analogy with the treatment of the
motion of a particle in classical mechanics, would guarantee downward-
sloping demand functions and straightforward welfare indices even in a
world of pervasive interdependence. Although it possessed a number of

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 264 Philip Mirowski and D. Wade Hands

other virtues, the Hotelling model had the drawback that the “income”
or “budget” term was not treated as a fixed entity independently of the
equilibrium outcome. This bothered Schultz, but he nonetheless chose to
collaborate with Hotelling to explore various justifications for the novel
treatment of the budget in the utility maximization problem.1 Schultz,
after all, was searching for a theory of interdependent demand curves,
and this seemed a viable formal candidate, even though it was not to be
found in Walras or Pareto. The novel treatment also gave rise to some
additional symmetry (or “integrability”) restrictions that Schultz imme-
diately tested and found to be violated in the agricultural demand func-
tions he estimated. Hotelling and Schultz then discussed the various
auxiliary hypotheses one might use to explain the failure, with
Hotelling focusing on pure theory, while Schultz tended to be con-
cerned with statistical and data problems. During this period Schultz
and Hotelling became aware of the now-standard Slutsky symmetry
conditions. Hotelling acknowledged the Slutsky equation but viewed it
as just another, more “Walrasian,” way to underwrite observed demand
curves. Although the Slutsky approach of maximizing an individual’s
utility function subject to a fixed income constraint had a certain appeal,
Hotelling found it unsatisfactory because it (unlike his own theory) had
only a limited ability to provide welfare theorems and did not guaran-
tee that demand curves sloped downward. When Schultz decided to test
the Slutsky and Hotelling symmetry conditions as rival hypotheses, he
found (much to his dismay) that both conditions appeared to be con-
tradicted by the data.
Here the saga of Hotelling and Schultz essentially draws to a close
with a denouement satisfying to neither of them. Schultz wrote up the
results of his decade-long search for the Walrasian principles underly-
ing demand curves in his 1938 Theory and Measurement of Demand. It
is seldom recognized that the book is essentially a swan song for empir-
ical Walrasian economics: Schultz bravely reported the empirical deba-
cle in detail and then produced a litany of excuses why things had not

1. Here we must signal that this was not simply an appeal to the special case of the Mar-
shallian constant marginal utility of money, nor was it straightforwardly a special case to be
rigidly restricted to some separate sphere of production. These issues are discussed in detail
in Hands and Mirowski 1998. For the present, it suffices to insist that Hotelling and Schultz
were engaged in a process of negotiation over the meaning and significance of the budget
term, which would imply revision of a whole array of other theoretical terms in tandem,
including but not restricted to the measurability of utility, the treatment of money, the signif-
icance of complementarity, the nature of interdependence, and so on.

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 265

worked out as hoped. The book ended with a promissory note that was
never redeemed, since Schultz died in a car accident in November
1938, just as the book appeared in print.
The reaction of Hotelling to the empirical disappointments and the
loss of Schultz was no less unexpected and presents more of an explana-
tory challenge to the historian. Initially, Hotelling showed every sign of
wishing to continue to pursue his initial price potential program.2 He
continued to teach a graduate course in mathematical economics at
Columbia, but the war and its own demands intervened. On 15 April
1943 he was formally appointed consultant to the Applied Mathemat-
ics Panel of the National Defense Research Council in recognition of
his work in putting together the Statistical Research Group (SRG) at
Columbia. Until it was dissolved in 1946, the SRG worked on the test-
ing of bombsights, the statistical effectiveness of various bombing
strategies, gun equilibration on ships, compressed flow through noz-
zles, the use of ordinary least squares (OLS) in antiaircraft fire control,
aerodynamic pursuit curves for overhead attacks, and numerous other
projects. This group included Allen Wallis, Milton Friedman, George
Stigler, Leonard Savage, and Abraham Wald.
The evidence suggests that Hotelling had essentially stopped reading
economics by 1940. Requests that Hotelling comment on subsequent
developments were met with reiterations of points he had made in the
1930s. In a letter of recommendation for Tjalling Koopmans’s move to
Yale University in 1955, he managed to avoid citing or discussing any
of his work in economics (Hotelling 1954). His class notes also suggest
that he restricted his lectures almost exclusively to problems tackled in
his 1930s works. It might be thought that with the loss of Schultz, his

2. Indeed, it seems that for Hotelling, the price potential model had become equated with
the very idea of rational action tout court. Evidence for this comes from Hotelling’s undated,
unpublished note titled “On the Nature of Demand and Supply Functions”: “ ‘Rational
Actions’ may be taken to mean a system of demand functions such that a ‘potential’ U exists
with pi = ȶU/ȶqi. Such demand & supply functions may well be taken as central, all others
being treated as more or less casual deviations, often of only temporary importance. But U
may be a function not only of the q’s but of their time — or space — derivatives. . . . Also, each
person’s U may depend upon the consumption of others (emulation; competitive display; but
also less wasteful types of activity, as when in intellectual cooperation a particular subject
occupying the focus of attention of a group may, advantageously to society, be pushed). The
statistical determination of ȶpi /ȶq j, which equals ȶpj /ȶqi for ‘rational action,’ involves a least-
squares solution & ideas of correlation which generalize ordinary calculus of correlation by
replacing individual variables by matrices. These matrices will, moreover, by [sic] symmet-
ric, giving rise to interesting theory.”

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 266 Philip Mirowski and D. Wade Hands

prime collaborator and interlocutor, he “had no one to talk to”— a char-

acterization that often crops up in the retrospective memoirs of oth-
ers — but it seems unlikely given the vast array of young talent he influ-
enced and had access to.
There is another possible explanation: Hotelling never entirely repu-
diated his 1932 price potential model. He taught it well into his career
and late in life conceived an interest in experimental work to directly
measure individual utility functions — an eventuality more in tune with
the 1932 model than with the behaviorism so rife during that period.
All signs point to a belief in a relatively more tangible and recoverable
version of utilitarianism, one more congruent with his own conceptions
of straightforward welfare indices and less bothered by “proper” treat-
ments of the budget constraint, which he regarded as an auxiliary con-
sideration. But after the war, there was simply nobody left who shared
this particular vision of a scientifically legitimate neoclassical price
theory.
Hotelling was of course aware of the Slutsky decomposition from at
least 1935, and his repeated postwar references to “measuring welfare”
seem ambiguous unless they are juxtaposed to his 1932 convictions
about the necessary character of the integrability conditions (1932b,
452). We can now see the postwar neoclassical consensus—that to be a
mathematical economist, one must start with Walras and Slutsky and
nowhere else—as a localized (and American) cultural prejudice highly
correlated with prior exposure to the Hotelling-Schultz dialogue of the
1930s. Once the community focused its attention definitively on the Slut-
sky equation, to the exclusion of all other versions of demand theory, the
only issues that could be addressed were ones that could be refracted
through the prism of substitution and income effects. The issues that had
captivated Schultz and Hotelling—issues like what was the proper role
of empiricism in demand theory? what could be said in the way of legit-
imate political statements about the welfare consequences of the mar-
ket? to what extent did this indicate a break with the formalism of the
potential field?—could now only be discussed in this idiom, if at all.

Chicago: They Do Things They

Won’t Do on Broadway
The first economist off the mark to respond to the Hotelling-Schultz
impasse was Paul Samuelson in 1938, but an understanding of the per-

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 267

sonalities of all three schools requires a narrative less concerned

with temporal precedence than with capturing local traditions. It thus
behooves us to begin with the University of Chicago Economics Depart-
ment, the site of most of the early activity in the American revision of
demand theory. Chicago not only contained Schultz’s statistical labo-
ratory but also was the home of the Cowles Commission from 1939 to
1955.
Those looking for the roots of the Chicago school often think they
can discern them in the 1930s, but in fact the department was quite
diverse during that period. Schultz himself was a conservative in poli-
tics but an advocate of the Walrasian brand of mathematical economics
combined with faith in the Fisherian theories of statistical inference.
There was Paul Douglas, a less technically inclined empiricist favoring
a mild form of technological determinism but a high-profile liberal in
political beliefs. There was Jacob Viner, who upheld an older style of
Marshallian economics, and there were some representatives from
the period of institutionalist strength. And finally there was the group
revolving around Frank Knight, including Aaron Director, Henry Simons,
and Lloyd Mints. Because it was predominantly Knight’s students and
supporters, including Friedman, Stigler, and Wallis, who subsequently
constituted the core of the postwar Chicago school, greater continuity
is often imputed to the school than actually existed. Nevertheless, some
direct affiliations can be drawn from Knight to Friedman and Stigler,
especially with regard to the issues that surround the Hotelling-Schultz
impasse.
One thing that helps us understand the postwar stabilization is an
appreciation of the changes that took place in the Chicago Economics
Department just prior to and during World War II. Knight’s group had
been treating Schultz with open contempt by the late 1930s, when
Schultz’s death in 1938 removed him from the scene. The election of
Douglas to the position of Chicago alderman in 1939 heralded the
beginnings of his political career and his absence from departmental
activities, neutralizing another Knight opponent. The retirement of
H. A. Millis in 1940 and the departure of Simeon Leland for Northwest-
ern University in 1946 (Reder 1982, 10) greatly attenuated the institu-
tionalist wing of the department. And Viner’s departure for Princeton
University in 1946 removed another competing brand of theory.
The pivotal figure in this wartime regime was Oskar Lange. Joining
the department in 1938, he quickly found himself the senior advocate of

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 268 Philip Mirowski and D. Wade Hands

the Walrasian approach to price theory and the sole local partisan of
“econometrics.” It is not irrelevant to our story that by 1938 he was the
prime defender of planning in the socialist calculation debate, an early
interpreter of Keynesianism, and a Marxist. His initial impact on the
Chicago scene was to polarize conceptions of formal economics in even
starker terms than one might find elsewhere. In the minds of many at
Chicago, Walrasian mathematical theory became conflated with social-
ism, crude numerical empiricism, and politically naive welfare eco-
nomics. Knight assumed proprietary rights over graduate price theory
during the latter part of the war, but it was Lange who taught John
Hicks’s Value and Capital. As Patinkin (1995, 372) notes, “it was the
socialist Oskar Lange who extolled the beauties of the Paretian opti-
mum achieved by a perfectly competitive market — and Frank Knight
who in effect taught us that the deeper welfare implications of the opti-
mum were indeed quite limited.”
The situation was further disturbed by the unanticipated arrival of
the Cowles Commission in September 1939. The move was arranged
largely by Alfred Cowles and Robert Hutchins, then president of the
University of Chicago; the Economics Department had little say in the
matter. Theodore Yntema of the School of Business was named direc-
tor of research, and the only members of the Economics Department to
be named to the Cowles staff were distinctly junior: Lange, Gregg
Lewis, and Jacob Mosak. In retrospect, it is clear that Cowles was rel-
atively weak just after the move. Some Cowles staff members declined
the move to Chicago, like Wald, who went to Columbia. Yntema was
largely an absentee research director from 1939 to 1942, and by 1942,
half the staff members were involved in war research (Christ 1952, 23).
The situation might have been tailor made for Lange to enter the void
and recast the Chicago department in his own image, but it never came
to pass; he was a visitor at Columbia in 1942 – 43, and by 1945 he
resigned from both the commission and the university to become Pol-
ish ambassador to the United States.
Despite Stigler’s (1988, 148) claim that in 1947 “there was no
Chicago School of Economics” (see also Bronfenbrenner 1962), it is
clear that Knight and his group were successful in advancing their
agenda during the war, and by 1944 they were engaged in open intel-
lectual warfare with the Cowles Commission (Reder 1982, 10; Ham-
mond 1993, 231). Friedman’s role as self-appointed tormentor of Cowles
and acknowledged leader of the Economics Department was para-

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 269

mount, beginning with his return to Chicago in 1946; thereafter the

Chicago school was consolidated with Wallis’s appointment as dean of
the Business School in 1955 and Stigler’s appointment to the Walgreen
professorship in 1958. Although none of these figures can be regarded
as carbon copies of Knight, the relationship of Chicago to the stabi-
lization of neoclassicism can be triangulated as a moving equilibrium
between Knight and Schultz, Knight and Friedman, Friedman and
Schultz, Friedman and Hotelling, and Friedman and Cowles.
Although there are, of course, many differences between Knight and
Friedman, one cannot make much headway in comprehending Chicago
price theory without beginning with Knight. Knight’s seductive appeal
for the stellar parade of graduate students in the 1930s and 1940s
cannot be underestimated, despite his rather Delphic style of lectures
and cynicism about the future of social science. Thus it is of profound
import that Knight chose the period of maximum disarray of his oppo-
nents to launch his attack on newer developments in demand theory. In
the signal JPE article of 1944, Knight explicitly criticized the “new”
school of Hicks, Eugen Slutsky, and Schultz and thereby implicitly crit-
icized their local surrogates, such as Lange and Jacob Marschak. The
general thrust of the article was that all the folderol concerning anom-
alous demand curves was a monumental red herring for economic
theory. “It is a fairly simple matter to explain the general shape — the
descending slope — of the demand curve” (300). Although primarily
philosophical, the attack was fairly subtle and foreshadowed much of
the later Chicago tradition in price theory.
Knight’s critique of the Slutsky approach was carried on simulta-
neously at different levels. At the most foundational level, Knight
ridiculed the idea that imitation of classical mechanics would serve to
reveal any hidden “laws” behind the law of demand: “Motives are not
analogous to forces” (310). His next level of attack concerned the
whole ordinalist self-denial of psychological commitments and the
associated cognitive agnosticism. According to Knight, all sorts of psy-
chological precepts were required for economic analysis, and his prag-
matism conflicted with the nascent orthodox line about the cognitive
impenetrability of other minds. Knight’s third level of attack was both
empirical and theoretical. On the empirical side, he credited his own
theory with having the “great advantage of eliminating the spurious
‘Giffen paradox’ ” (300), and on the theoretical side, he made an argu-
ment that echoed Hotelling’s own defense of his 1932 model: It had the

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 270 Philip Mirowski and D. Wade Hands

benefit of “making the theory of consumption logically parallel with

the theory of production” (300).
Finally, Knight made explicit reference to the Chicago tenet that sep-
arate income effects are politically dangerous:
The general point of view and habit of mind reflected in the Hicks-
Slutzky [sic] analysis has wide ramifications in the recent literature
and has led to utter confusion in the whole body of economic
thought. We refer, of course, to the huge corpus of discussion begin-
ning with Keynes’ General Theory and following the lead of that
work. J. R. Hicks’s book, Value and Capital, is especially interesting
as a general treatise which combines the theory of Keynes’s treat-
ment of unemployment and money with the Slutzky analysis of
demand and theory of utility. (300, n. 10)
Knight believed that the idea that one could analytically decompose
movements along a microeconomic demand curve into income and sub-
stitution effects would lead to the conviction that one could actually
factor out variations in macroeconomic incomes from price-theoretic
considerations and thus open the door to (insidious) Keynesianism.
Knight rarely made any argument elegantly or systematically, and
the 1944 JPE article was no exception. His philosophical objections
were sprinkled throughout an article that was ostensibly a technical cri-
tique of the Hicks-Slutsky approach to demand theory. Although the
fine points of the offensive were muddled, the overall thrust of the tech-
nical critique was that the “Slutzky school” had not adequately consid-
ered the implications of posting the invariance of income or the budget
constraint. By holding the prices of all other goods and money income
fixed when deriving the demand curve, Knight insisted, the Slutsky
approach had failed to recognize that real incomes continuously
change. Because he believed monetary theory had to be built into the
foundations of price theory, Knight argued for an alternative derivation
of demand that only expressed the effects of changes in relative prices
through the device of holding the value of money constant by compen-
sating changes in the prices of other goods. Knight’s disdain for math-
ematics essentially prevented him from specifying how this could actu-
ally be accomplished.
Knight’s article clearly exhibits the rough-hewn building blocks of
the Chicago orthodoxy that would ultimately be sculpted by Friedman,
Stigler, Gary Becker, and others. The approach began with an ontolog-

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 271

ical commitment: “The demand curve . . . [is] undoubtedly the most

solidly real of all the functional relations dealt with in economic the-
ory” (310). It followed that no “more fundamental” laws would be dis-
covered to buttress demand curves and that no additional legitimation
was needed. All applicable predictions and usable economic statements
would start and end with the theory of supply and demand — period.
Interestingly enough, this did not proscribe such exercises as the
derivation of the demand curve from constrained utility optimization,
and (slightly diverging from Knight) one might even entertain the Slut-
sky decomposition into income and substitution effects. But the ambi-
tion of testing the Slutsky symmetry restrictions was a dangerous delu-
sion resting on a serious misunderstanding of the very nature of price
theory. One should not believe for a minute that there was some “real”
invariant out there, like income, psychological motive, or technology;
nothing is really invariant in a dynamic economy. For Knight, the “com-
pensating variations” that would successfully isolate income effects
most likely could never be carried out in actual practice, much less
inferred from an observed sequence of actual exchanges; an income
effect was a virtual (and rather dangerous) notion. That is why Knight
and his group had become convinced that Schultz had been barking up
the wrong tree and why Friedman could not countenance the Cowles
Commission’s quest for the Holy Grail of “structure.”
The most influential statement of Chicago demand theory, Fried-
man’s 1949 article, “The Marshallian Demand Curve,” proudly displays
its intellectual heritage by insisting that its conclusion is “identical with
that reached by Frank Knight in a recent article” (135), namely, the
1944 article. Phalanxes of scholars have combed Friedman’s “method-
ology” article for philosophical precursors that render its theses more
comprehensible, just as smaller platoons have plundered and scoured
Alfred Marshall to explicate Friedman’s “Marshallian Demand
Curve.”3 Although these are interesting exercises, we would like to sug-

3. The most painstaking and perceptive case that Friedman’s demand curve cannot be
found in Marshall is Aldrich 1996. We can do no better than suggest that the reader interested
in the fine technical points of the argument begin there. We also view our argument as con-
sistent with that found in Hirsch and De Marchi 1990: “Looking to Marshall to help is just as
likely to confuse” (36); “it is only with the help of Knight, we suggest, that one can derive a
coherent political economy from Friedman’s writings” (285). What we add to their theses is
the historical context that can explain the genesis of the doctrine and the position it occupied
versus other constructions of orthodox neoclassical price theory.

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 272 Philip Mirowski and D. Wade Hands

gest another way to understand these texts. One need not search for
Friedman’s demand curve in Marshall, any more than one needs to find
his philosophy in Karl Popper or John Dewey. Friedman’s demand
curve grows out of Knight, via Schultz and Hotelling. It was the culmi-
nation of a whole sequence of reviews and articles, most notably Fried-
man 1941, Friedman and Wallis 1942, and Friedman 1946, seeking a
way out of the impasse bequeathed him by Schultz’s palpable failure.
These articles are the traces of a mighty struggle with the meaning and
significance of income for neoclassical price theory, and it is no coin-
cidence that the work for which Friedman is most famous, such as the
“permanent income hypothesis” and the mantra that “money matters,”
derives directly from his empirical work on the measurement of pro-
fessional incomes and the estimation of relevant demand functions in
the late 1930s (Friedman and Kuznets 1945).
What are the main precepts of Friedman’s theory of demand, which
became the foundations of the Chicago approach? First and foremost,
he agrees with Knight that the demand function is the primary entity
in price theory. Early on, he adopted the position that to “go behind”
the demand curve to its foundations in laws of utility or “indifference”
is largely a waste of time. It could not further the quest “to obtain exact
knowledge of the quantitative relation of consumer expenditures to
prices and incomes for the purpose of predicting the effect of changes
in economic conditions on the consumption of various commodities”
(Friedman and Wallis 1942, 188). Of course, the analyst should not repu-
diate something like utility altogether, since it “may be a useful expos-
itory device,” but one should not “mistake the scaffolding set up to
facilitate the logical analysis for the skeletal structure of the problem.”
By the time of his Price Theory (1966), he identifies utility as a conve-
nient fiction: “We shall suppose that the individual in making these
decisions acts as if he were pursuing and attempting to maximize a sin-
gle end” (37; emphasis in original). The divergence from Knight comes
in relinquishing all rights to criticize utility as a scientific concept and
instead linking its legitimacy to the uses it serves, namely, as a language
to motivate and discuss observable demand curves. Friedman the sta-
tistician here inverts Knight the “ideal type” philosopher: Good empir-
ical demand curves underwrite dubious quasi psychology rather than
vice versa.
The upshot of all these considerations for Friedman was emphati-
cally not the conclusion that demand theory was in a sorry state but

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 273

rather the endorsement of maximum freedom with regard to the treat-

ment of income and income effects. What he came to call “Marshallian
price theory” was perfectly fine as far as it went; the confusion over
income effects was taken as a symptom of what was still missing from
a truly comprehensive neoclassical theory (as in Knight), and that was
monetary theory: “The important point is that the existing theory of
relative prices does not really help to narrow appreciably the range of
admissible hypotheses about the dynamic forces at work. Monetary
theory, interpreted broadly, has somewhat more to offer” (Friedman
1951, 113; emphasis in original).
To those in other neoclassical camps, it seemed that Chicago was los-
ing all interest in “rigorous theory.” For example, Friedman’s Price
Theory simply ignored those like Martin Bailey (1954) who criticized
his method of aggregation, even though Bailey was on the faculty at
Chicago from 1955 to 1965. The first edition of Stigler’s Theory of
Price did not even see fit to mention the Slutsky equation, while the
second edition relegated it to a single footnote (1952, 303). Stigler
(1965, 151) did notice the Slutsky equation in his history of demand
theory, but only to try to summarily bury it: “Economists tacitly agreed
that it is better to have a poor, useful theory than a rich, useless one.” He
asked, “Can one say more about the demand functions if they are
derived from the utility functions?” (147). This process culminated in
Becker’s (1962, 4) assertion that “the fundamental theorem of tradi-
tional theory — that demand curves are negatively inclined — largely
results from a change in the opportunities set alone and is largely inde-
pendent of the decision rule.” In other words, in exploiting maximum
freedom to make “income” whatever we wish it to be, we need not be
held to any explicit conception of underlying intentional psychology or
functional interdependence at all in order to get a downward-sloping
demand curve. The moral of the story is to go forth and deploy demand
(and supply) functions wherever markets might potentially exist, and
income effects be damned (Reder 1982).
Curiously enough, in retrospect one can still discern how this all
began with Hotelling. What was needed were simple stories with clean
empirical implications that could be retailed in concise phrases to
clients of the economics profession. This, said Friedman, was the sine
qua non of a useful model. Hotelling’s price potential model would have
provided that as well, except that too many economists found the non-
conservation of income difficult to swallow. Consequently, they imposed

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 274 Philip Mirowski and D. Wade Hands

fixity of income, only to discover that one must not take it too seriously,
at least in Chicago. The next step might have been to double back to the
Hotelling model to reevaluate its attractions and drawbacks, but that
did not happen, primarily because the approach to demand theory now
became confused with tangential issues: prostatist versus antistatist
politics; the intrinsic attractions of mathematical formalism; the extent
of commitment to utilitarian psychology; the correct format for statis-
tical inference; the resonance or dissonance with Keynesian macro-
economics; the significance of imperfect competition for price theory;
and much more. But through all the blooming, buzzing confusion,
Chicago never lost sight of its most immediate rival in demand theory:
the Cowles Commission.

Out of the Cowles and into the Fire

Before the Cowles Commission came to Chicago, it was not the stan-
dard-bearer for general-equilibrium theory that it would become in the
1940s. Charles Roos and Harold Thayer Davis, the first two research
directors, were avid supporters of mathematical economics and econo-
metrics, but neither was enamored of Walras or Pareto. Before Chi-
cago, the commission’s entire funding derived from Alfred Cowles and
his family, and the benefactor’s inclinations tended toward practical
studies of financial assets and government policy. Yntema’s director-
ship from 1939 to 1942 had very little impact on the commission’s
research. The pivotal change came with Lange, who recruited Leonid
Hurwicz in 1942 and Marschak as research director in 1943.
The extent to which Marschak’s priorities came to dominate the
Cowles Commission in the 1940s is still not sufficiently appreciated.
Marschak was a Russian mechanical engineer and former Menshevik
who had studied economics at Berlin and Heidelberg, and his earliest
theoretical contributions, like those of Lange, were in the socialist cal-
culation debate. In the 1930s, he was regarded as one of the few experts
on the statistical estimation of demand functions, and he had partici-
pated in some widely noted controversies over pitfalls in demand esti-
mation with such major figures as Wassily Leontief and Ragnar Frisch
(Hendry and Morgan 1995). Thus he was viewed in Chicago as an heir-
apparent to the program of Schultz. In his 1939 review of Theory and
Measurement of Demand, Marschak (1939b, 487) seems to take up the
torch dropped by the fallen comrade. Although he credits Schultz with

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 275

“reviving and applying economic theorems of Pareto and Slutsky,” he

hints that Schultz may have been guilty of certain “errors or ambigui-
ties,” which were now set straight. Nevertheless, he makes it clear that
he intends to come up with multiple auxiliary hypotheses to append to
Schultz’s procedures, without once acknowledging the downbeat prog-
nosis that Schultz’s text tendered and that his own proposed hypotheses
were intended to explain away. Marschak never once openly entertained
the idea that the Walrasian neoclassical program might be at fault. For
him, the Walrasian program was always unreflectively equated with
science, and he surrounded himself with like-minded souls. Marschak’s
irrepressible — some would add premature — confidence that thorny
conceptual issues were on the cusp of being resolved by some new
mathematical formalism helped create a hothouse environment for the
cultivation of mathematical neoclassical economics at the commission.
Marschak’s work during the decade from roughly 1938 to 1948 was
primarily bent to this purpose. As he put it in the Schultz festschrift,
“To test theories describing the interdependence of economic variables,
empirical equations of the Schultz type are not a useful tool” (quoted in
Lange, McIntire, and Yntema 1942, 150). But Marschak’s inventive-
ness outran his stalwart purpose, since in this period he came up with
too many reasons why Schultz had to be revised, thus raising the dis-
turbing possibility that no particular empirical phenomenon could be
ruled out even by the most sophisticated techniques for statistical
demand estimation. Some auxilliary hypotheses proposed by Marschak
in this period, all strictly confined within the framework of the Slutsky
equation, included the following:
1. Changes in the distribution of income were not taken into account
in the equations (1939a, 1943).
2. The income constraint could be treated as something in between
Hotelling’s endogenous specification and a purely exogenous
invariant by portraying it as a random variable.
3. Income invariance could be loosened by the imposition of lags or
some deterministic dynamics.
4. Systematic shifts in supply and demand functions must be empir-
ically identified for purposes of analytical separation (1939a).
5. The analytical need to make a comprehensive accounting of all
other relevant factors in an interdependent system could be obvi-
ated by means of previously overlooked statistical criteria (Hendry
and Morgan 1995).

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 276 Philip Mirowski and D. Wade Hands

6. Another auxiliary hypothesis based on Mosak 1938, which Mar-

schak would promote into conventional textbook wisdom, was
that the Hotelling price potential model was in fact a theory of
production factor demands and not consumer demand and there-
fore had to be combined with the Slutsky model at one remove
rather than treated as a rival model.
The story that Marschak directed research along the path bequeathed
by Schultz and Hotelling is not part of the existing histories of the
Cowles Commission. Perhaps this is because most authors have focused
on the creation of an independent abstract discipline called “economet-
rics” as the predominant telos of the postwar commission (Christ, 1952,
1994; Epstein 1987; Hildreth 1986; Morgan 1990). Indeed the Cowles
Commission is usually identified with the probability approach to
econometrics, the identification problem, the correction of estimation
bias, and the genesis of Keynesian macroeconometrics rather than with
the theory of demand. We are suggesting an alternative reading that
shifts the center of gravity by looking at the whole record of the com-
mission during the 1940s. After Marschak, Cowles was perceived as
first and foremost the uncompromising partisan of Walrasian general-
equilibrium theory, including staunch adherence to the Slutsky equa-
tion, and the commission’s other formidable accomplishments were
often seen as subordinate to this larger purpose.
It seems that when Marschak moved to Chicago, he took advantage
of the disarray to recast the commission in his own image. He swept the
stables clean of any alternative methodological approaches to price the-
ory, most notably George Katona’s anthropological and psychological
approach to research on price controls. He revived the moribund
Cowles Papers series and instituted a series of Cowles seminars at
Chicago. He replaced the previous Advisory Council with an Advisory
Committee more to his liking (Christ 1952, 26). He had brought a
Rockefeller grant to study demand analysis with him from New York
and now sought to parlay this into long-term institutional funding for
the commission. Once at the commission, Marschak apparently had
settled on questions of statistical estimation as the primary source of
“protective belt” hypotheses surrounding prior attempts at demand esti-
mation, but at Rockefeller he ran into a brick wall, which he attributed
to the machinations of institutionalists at the National Bureau of Eco-
nomic Research (NBER). Hence, the strife between Cowles and the

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 277

NBER that culminated in the infamous 1947 “measurement without

theory” controversy between Koopmans and Rutledge Vining was in
fact initiated by Marschak and seemed to be as much about access to
funding and legitimacy as it was about fine points of statistical analysis
(Mirowski 1989a).
These attempts to capture outside funding had repercussions back in
Chicago. Rockefeller Foundation officer Joseph Willits complained to
University of Chicago dean of social sciences Robert Redfield that
Marschak’s proposal was confusing and ill motivated. Marschak (1944)
responded with a revealing letter to Redfield that justified his ambitious
plans and sought to differentiate Cowles’ research program from the
work of Schultz. He indicated that problems in price theory had origi-
nally justified “the adaptation of statistical method to the (not yet suf-
ficiently recognized) peculiarities of economic data.” Although Mar-
schak did manage to obtain $7,500 from Rockefeller, which he used to
hire Koopmans, this was nothing like the long-term institutional sup-
port he desired.
During these struggles, the commission had essentially arrived at its
position on the correct way to go about estimating a structural system
of demand equations, best represented by the famous Cowles Com-
mission Monographs 10 (Koopmans 1950) and 14 (Hood and Koop-
mans 1953). Although these works are often portrayed as definitive
statements about proper econometric technique (Christ 1994; Hildreth
1986), they met with opposition from many contemporaries. The prob-
lem was that although “structure,” “system,” and “interdependence”
were rallying cries at the commission, it did not take long to discover
how intractable true generality could be once researchers began to spell
out the implications of these themes. Those who had passed the Wal-
rasian litmus test did not give a second thought to utility or preferences
as vera causae, but the problem became salient in that pesky variable,
income. Koopmans (1950, 394) posed the obvious question: “When is
a system complete?”
The causal principle . . . regards as exogenous those variables which
influence the (remaining) endogenous variables but are not influ-
enced thereby. The causal principle is often used also if it applies
only approximately. . . . [An] example is found in the formation of
quantity and price of a consumers’ good that attracts only a small
fraction of consumers’ expenditure. In such cases, consumer income

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 278 Philip Mirowski and D. Wade Hands

is often taken as an exogenous variable, operating at the demand

side, although of course consumers’ income itself depends on the
demand for all commodities. . . . There is no sharp line of demarca-
tion between the application of the approximate causal principle and
what deserves mention as a third principle of consideration: the pur-
pose of exposition.
The significance of this admission is that it signals just how impre-
cise the phenomenon of the invariance of the budget constraint could
be from the viewpoint of the empirical economist; systems have arbi-
trary boundaries that are conditional on the uses to which they are put.
In pervasively interdependent systems, closure is a relative concept.
Conservation principles require closed systems, but closure had been
practically ruled out of bounds by the rejection of any experimental
protocol (Mirowski 1989b). A charitable reading might suggest that the
morals of this passage are not very different from those extracted from
Friedman in the previous section; but in fact, the Cowles approach is
quite different. Marschak and Koopmans repeatedly insisted on phras-
ing such questions as technical problems: problems of “causal identifi-
cation,” “statistical definitions of exogeneity,” vigilance against “obser-
vational equivalence,” “corrections for simultaneous equations bias,”
and so on. All this tough-minded technocratic language tended to sug-
gest that the data would ultimately settle these issues, which did res-
onate with the logical positivism coming to be embraced in America.
But experience with the actual regressions and getting one’s hands dirty
in the data tend to undermine this dream.
The number of logical elisions, leaps of faith, and undermotivated
compromises at the commission began to look daunting by the late
1940s. Suppose, for instance, that you proved willing to arbitrarily
truncate the full Walrasian system in order to arrive at a small system
of tractable demand equations, which then by some technical trick
were rendered causally identified. The first problem that must be con-
fronted is that Walrasian theory is thoroughly deterministic, but the
data appear stochastic. After flirting with the idea that randomness
should enter through measurement error, the commission members
rapidly backpedaled to the position that the appropriate specification
was “errors in equations” (Hildreth 1986, 31). Gaussian distributions
were assumed with very little justification. This could not straightfor-
wardly be chalked up to convenience or simplicity; it was the specifi-
cation that least deranged the mathematical structure of the Walrasian

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 279

system by keeping stochastic considerations fully quarantined from the

definition of equilibrium (Mirowski 1989c), but it also became the
stock excuse for the exclusion of any other potentially relevant causal
variables. Thus, development of the very class of statistical model was
rapidly being shaped by a priori images of the appropriate theory
underlying the phenomena.
But the story gets worse. Marschak, Trygve Haavelmo, Meyer
Girschick, and Lawrence Klein wanted to put the new techniques to
work verifying neoclassicism, but when they attempted to do so, the
results were uniformly disappointing. To estimate a demand system,
one had to confront the problems of aggregation, time trends, historical
change of structure, index numbers, and so on — in other words, all
those uneasy choices first tackled by Schultz. Any specification of the
relevant subset of commodities was bound to appear arbitrary to a
dyed-in-the-wool Walrasian, as would the choice of exogenous identi-
fying variables. Finally, if one plowed ahead anyway, what one discov-
ered was that some parameter signs were wrong, and many standard
errors were orders of magnitude larger than for old-fashioned single-
equation OLS models. Indeed, it was effectively impossible to point to
anything especially novel that was learned about demand functions
from the new estimates in this period; mostly they did not differ sub-
stantially from the parameters extracted by simpler means.
Two generalizations can be made about the Cowles program of the
1940s relative to the Hotelling-Schultz impasse of the 1930s. First, the
commission members grew so fascinated by their shiny new statistical
tools that most of their writings were absorbed with endless discussions
of the baroque complexities of identification, exogeneity, limited infor-
mation –maximum likelihood estimation, and all the rest; the estima-
tion techniques seemed to become the point of the whole exercise. The
result was that they generally lost sight of the whole program of testing
price theory; repeated ritual invocations of Neyman-Pearson language
could not repress the fact that pervasive freedom in model selection
and immunizing stratagems could never be adequately encompassed by
the framework of Type I/Type II errors. Henceforth, Schultz’s concerns
simply faded away. The integrability conditions so central to the valid-
ity of the neoclassical program in the 1930s slowly became downgraded
to the status of a mathematical curiosum in the 1950s at Cowles, not
because they passed any stringent tests but rather because the prolifer-
ating tests passed them by.

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 280 Philip Mirowski and D. Wade Hands

The second issue involves Keynesian macroeconomics; it is impor-

tant to understand why the commission’s structural economics became
much more closely associated with Keynesian macroeconomics than
with neoclassical price theory. As we have seen, Walrasian biases may
have shaped the format of simultaneous equations econometrics, but it
did not follow through to actually provide a plausible framework for
empirical practice. In this regard, two- and three-equation Keynesian
models were much more amenable to structural estimation than full-
blown Walrasian systems. They sported natural closure conditions; the
Keynesian synthesis interpretation rendered them modular, in that one
could always contemplate adding another sector; the definitions of the
variables were thought to better approximate the entities contemplated
by the theory; and policy variables seemed to capture exogenous inter-
vention much better than the vexed income variables in demand esti-
mation. Sometime between the first flush of enthusiasm in 1944 and
1947, structural econometrics at the commission stopped being about
price theory and switched allegiance to Keynesianism.
The forging of the unholy alliance between Slutsky and Keynes at
Cowles was of course Knight’s worst nightmare, and it was Friedman
who took it upon himself to carry the war back home to the commis-
sion. Friedman felt that the entire Cowles program of structural esti-
mation was a vast waste of time. Time and time again, he told Mar-
schak and Koopmans to their faces that their intricate statistical
procedures failed to solve any real scientific problems. Friedman
admitted later: “I was a major critic of the kind of thing they were
doing in Chicago. I introduced the idea of testing their work against
naive models, naive hypotheses, and so on. So I was very unsympa-
thetic to Koopmans from the beginning” (Hammond 1993, 231).
Although examination of the later reincarnations of the Cowles
Commission would take us beyond our allotted time frame, one devel-
opment does have profound implications for the stabilization of neo-
classical price theory in America. In brief, given the failure of the
Schultz program via Marschak, retention of the basic program would
require that the scientific credentials of neoclassical economics be val-
idated by some other authority. The redirection of the Cowles Com-
mission is poignantly captured by the revision of its motto in 1952 from
its previous “Science Is Measurement” to “Theory and Measurement.”
Actually, “Theory and Rigor and More Theory” would have been more
accurate, since the commission wholeheartedly embraced the Bour-

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 281

bakist program of axiomatization and formalization in order to bolster

the flagging fortunes of the Walrasian program (Weintraub and
Mirowski 1994). In caricature, if a concerted program of statistical test-
ing could not place demand theory on firm footing, then perhaps a com-
prehensive axiomatization of Walrasian general-equilibrium theory
could effectively do the job.
An exploration of the various twists and turns that led neoclassical
highbrow theory to circumvent the Hotelling-Schultz impasse at the
commission after 1950 would lead us too far afield. It would begin with
Koopmans’s search for a totally neutral and institution-free theory in
activity analysis; it would continue with the impact of John von Neu-
mann’s game theory on modern mathematical economics (Leonard
1995; Mirowski forthcoming a); it would marvel at the way in which
statisticians like Hurwicz, Koopmans, and Marschak abruptly turned
their efforts away from any constructive empiricism and toward a the-
oretical portrayal of the economic agent as a miniature econometrician;
it would track the increasing influence of the RAND Corporation and
the military on the commission’s research interests (Leonard 1991;
Mirowski forthcoming b); it would take into account the changing mean-
ing of “dynamics” within the Walrasian program (Weintraub 1991); and
much more. Yet no such list, however augmented and extended, would
be complete without a culmination in the infamous Sonnenschein-
Mantel-Debreu results of the 1970s, which settled the issue of the rela-
tionship of Walrasian general equilibrium to the theory of demand in a
sternly negative way: In general, Walrasian theory placed no restrictions
on excess demand functions except for homogeneity of degree zero and
Walras’s law (Sonnenschein 1972).4 But all that lay in the future; let us
now turn to the third version of postwar neoclassical price theory.

4. Here we must register out disagreement with the very interesting history of our period
by Ingrao and Israel (1990, esp. 284 – 88). Those authors present the history of price theory as
a single unified thread, with a single break in the 1930s– 40s, when mathematically sophisti-
cated authors renounced physical metaphor for the rigors of mathematical formalization.
They conclude that the Sonnenschein-Mantel-Debreu (S-M-D) results are the fruits of that
reorientation in that axiomatization has now definitively revealed the empirical emptiness and
conceptual limitations of the Walrasian program. This essay demonstrates that we would
challenge their theses on a number of counts: (1) Neoclassical theory, even within a single
nation, is not monolithic in its core doctrines but fractured and fragmented. (2) Many of the
main American protagonists in the 1930s and after persisted in taking their cues from physics,
though it tended to be from vintages contemporary with their experience. (3) The S-M-D
results would impact different versions of neoclassicism with differing degrees of damage,
depending on their variant attitudes toward the treatment of the income constraint, the Slut-

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 282 Philip Mirowski and D. Wade Hands

The Revealing Story of Paul Samuelson

Samuelson’s revealed preference approach to consumer choice theory
represents the third facet of our demanding triad. Entering the Univer-
sity of Chicago the same year that Hotelling’s paper appeared in the
JPE and just five years before the arrival of the Cowles Commission,
Samuelson found himself in the thick of the fray of the 1930s dispute
over the theory of demand. As he put it, “The year 1932 was a good
time to come to the study of economics” (Samuelson 1983, 4).
Although the Chicago and Cowles approaches involve (perhaps
shifting) epistemological visions of the legitimate foundations for eco-
nomic science, it is Samuelson’s approach that is most explicit about its
cognitive preferences. That economic concepts are scientifically legit-
imate only when they are operationally meaningful became Samuel-
son’s epistemic battle cry quite early in his career. Scientific concepts
are meaningful or significant only if they are defined by a set of inter-
subjectively observable procedures or operations, and the goal of sci-
entific theory is to redescribe, not to explain, that which is observed in
the empirical domain (Cohen 1995; Samuelson 1965; Wong 1978).
Although “empirical” is the watchword for Samuelson’s vision of
economic science, Schultz’s Theory and Measurement of Demand was
not the way to achieve it from his perspective. It is fair to say that
Samuelson has been as consistently critical of the efforts of Schultz and
others to empirically test the theory of demand as he has been of the
antipositivist approach of Knight’s “Swiss guards” (Samuelson 1983,
7). “Some of the skepticisms of Knight and Jacob Viner concerning the
empirical statistical studies that their colleagues Paul Douglas and
Henry Schultz were attempting, I readily admit, were well taken”
(Samuelson 1992, 241). In fact, Samuelson’s criticism has not been
restricted to efforts to empirically test the theory of demand; increas-

sky equation, and the centrality of the law of demand vis-à-vis Walrasian theory. (4) Most
denizens of the Cowles Commission did not find S-M-D very distressing, since they had
already renounced the laws of supply and demand.
Once these lessons are absorbed, it becomes clear that Ingrao and Israel’s narrative is just
an inversion of the more conventional Whig account, written from the constricted vantage
point of one subset of one school of American neoclassicism. To find these confusions in such
otherwise discerning authors reveals why a comprehensive history of demand theory in the
twentieth century is still desperately needed.

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 283

ingly over the last few decades, he has been critical of econometric test-
ing in general.
Let me make a confession. Back when I was 20 I could perceive the
great progress that was being made in econometric methods. Even
without foreseeing the onset of the computer age, with its cheapen-
ing of calculations, I expected that the new econometrics would
enable us to narrow down the uncertainties of our economic theories.
We would be able to reject false theories. We would be able to infer
new good theories. My confession is that this expectation has not
worked out. (Samuelson 1992, 243)
This puts the champion of positivist economics in a rather peculiar
position. The person who is often viewed as one of the staunchest advo-
cates of the empirical science of economics does not do, nor does he
have much faith in, the type of econometric work that dominates the
empirical practice of the contemporary mainstream he helped create.
This paradox has implications for Samuelson’s entire career, but we
will focus on his version of demand theory.
The problem for Samuelson was to find a way of formulating the the-
ory of demand that would be consistent with his positivist-operational-
ist methodology while simultaneously avoiding the type of economet-
ric testing associated with Schultz. How could this possibly be done?
How could one steer the ship of economic science between the Scylla
of apriorism and the Charybdis of Schultz’s economic estimates while
honoring the law of demand and accepting only neoclassical givens?
His answer, published in 1938, was to ground demand theory on what
later came to be called the theory of revealed preference. In essence his
answer was to change the place where the empiricism lived in the neo-
classical theory of demand. Instead of having empiricism enter at the
back end — by testing the empirical implications deduced from the
theory — the revealed preference approach would place empiricism
right up front at the beginning of the exercise. If the epistemologically
dubious notion of subjective utility could be replaced with a strictly
behaviorist — thus objective, observational, operational, and meaning-
ful — concept of consumer action, demand theory could be reconsti-
tuted on what Samuelson considered legitimate scientific foundations.
A demand theory based on a behaviorist-inspired notion of human
action (i.e., response) would already be empirical at the start, thus elim-

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 284 Philip Mirowski and D. Wade Hands

inating the need for after-the-fact econometric testing like that of

Schultz, while also being free of any residual subjectivism or apriorism.
Samuelson could kill two Chicago birds with one well-placed stone.
In the original 1938 presentation, Samuelson argued forcefully that
his goal was “to develop the theory of consumer’s behavior freed from
any vestigial traces of the utility concept” (71). He required three pos-
tulates for this result. The first two postulates had been standard in the
Slutsky approach: He assumed first the existence of a single-valued
demand function xi = xi ( p,l) = xi ( p1, . . . pn,l) for each good satisfying
the budget constraint ȸpx = l and, second, that demand functions are
homogeneous of degree zero in prices and money income. The third
postulate is the assumption that ultimately came to be called the
revealed preference condition:

ȸ x'p ” ȸ xp implies that ȸ xp' > ȸ x'p' (RP)

In this postulate, x is the bundle of goods at price vector p, and x' is the
bundle of goods purchased at the price vector p'. If bundle x' was afford-
able at p but was not selected, then x must have been more expensive at
p', since it was not selected. In the language of the later literature, since
x' was affordable at p but was not chosen, the bundle x was “revealed
preferred” to x'; thus if x' was chosen at p', it must be because x is more
expensive at these prices.
Samuelson was careful in the original article to avoid the word
“preference” or any other term that might imply subjective evaluation
or intentionality. The 1938 article strictly maintains the behaviorist
idiom of “selection” throughout; the theory was, after all, meant to
replace the intentional theory of subjective utility, preference, and
choice. From these three postulates, Samuelson was able to derive what
he considered the main result of the utility-based theory of demand: a
generalized version of the law of demand. According to Samuelson, he
started with the operational-observable condition RP and ended up
with an observable generalized law of demand; not only could eco-
nomics do without the nineteenth-century notion of cardinal utility, but
also now it could do without the concept of ordinal utility or well-
ordered preference. The shroud of utility metaphysics was cast away,
and economics was now indeed “part of the advancing army of sci-
ence” (Samuelson 1983, 10).
The substantive result of ordinal utility theory that did not follow

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 285

from RP was the Slutsky symmetry condition, known at the time to be

equivalent to the integrability of the demand functions:
ȶ—xi + x ȶ—xi = —
ȶ x j + x ȶ—x j
j i
ȶ pj ȶl ȶpi ȶl (S)
Although Samuelson admitted that RP did not generate the Slutsky
condition, he did not find the omission problematic, since he did not
find integrability very interesting (in 1938):
The only possible interest that integrability can have (except to those
who have an historical attachment to the utility concept) would be in
providing us with the additional knowledge concerning a certain recip-
rocal relation [S]. . . . But it is this very implication which makes it
doubtful and subject to refutation under ideal observational condi-
tions, although I have little faith in any attempts to verify this statis-
tically. (Samuelson, 1938, 68)
In Hands and Mirowski 1998, we suggested a number of reasons why
Samuelson’s interests — his vision of welfare economics, his reading of
the lesson from physics, and so on — might lead him to be uninterested
in integrability. There is no reason to repeat this discussion here, but
there is reason to point out at least two immediate ironies in Samuel-
son’s 1938 attitude toward integrability. First, as we will see below, by
1950 he had become quite interested in integrability and its relationship
to RP. In fact, by the publication of his Nobel lecture in 1972, he used
the (stronger) integrability conditions from Hotelling’s 1932 model as
his single “illustrative economic example” of the power of mathemat-
ics in economic analysis. It seems ironic that what was dismissed as an
anachronism in 1938 was thought to be the best example of successful
economic science for the Nobel audience. Second, there is the ongoing
irony of the outspoken positivist-empiricist who has little faith in sta-
tistical verification; worse yet is the apparent inconsistency of a person
who equates “empirical with “meaningful,” claiming that integrability
is not empirically true while simultaneously denying that it should be
tested. It seems the old Chicago demon still resisted exorcism.
Samuelson ([1948] 1968) continued the argument and introduced the
term “revealed preference” in his next major article on demand theory,
although his focus seemed to change substantially during the interven-
ing decade. The condition RP remains in the 1948 article, but it is now
a way of constructing a consumer’s indifference map; it is now a way of

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 286 Philip Mirowski and D. Wade Hands

revealing the consumer’s preferences. RP moves from a way of elimi-

nating preference from the theory of demand to a way of uncovering
preferences for the theory of demand. The behaviorist idiom disap-
pears, and the (now) standard mentalist language of preference, choice,
and ordinal utility returns with a vengeance.
A more significant change occurs in Samuelson’s 1950 article, in
which he attempts to provide an intuitive explanation of the integrabil-
ity condition. Hendrik Houthakker had argued earlier in 1950 that what
was needed was an RP-based condition that would be both necessary
and sufficient for all the implications of ordinal utility theory — includ-
ing integrability — and introduced the so-called Strong Axiom of
Revealed Preference to do the job. Samuelson admits in the 1950 arti-
cle that the problem of integrability “could not yield to this [Samuel-
son’s] weak axiom alone” (370) and that Houthakker’s stronger condi-
tion would indeed be needed to achieve the desired results. His general
tone is ebulliently consummatory: “We are now in a position to com-
plete the programme begun a dozen years ago of arriving at the full
empirical implications for demand behavior of the most general ordinal
utility analysis” (369; emphasis in original).
Despite the positive spin Samuelson puts on integrability, it seems a
bit strange that he would be so pleased that an RP-based theory has
been shown to be observationally equivalent to ordinal utility theory.
As Stanley Wong (1978, 121) points out, “It is puzzling how Samuelson
can consider revealed preference theory to be the logical equivalent to
ordinal utility theory and at the same time argue that the former theory
is observational while the latter is not.” For a positivist, if two theories
are logically and observationally equivalent, they are the same theory.
It appears that revealed preference no longer frees demand theory
“from any vestigial traces of the utility concept,” but rather it simply
provides an alternative way to theoretically describe the observational
domain of utility theory — an observational domain that Samuelson
earlier seemed to think was an empty set. As Houthakker (1983, 63) put
it: “The stone the builder rejected in 1938 seemed to have become a
cornerstone in 1950.”
It is also useful to point out that the issues raised by the Hotelling-
Schultz impasse continue to set the general problematic for Samuel-
son’s demand theory. By 1950, Hotelling’s 1932 work had been down-
graded to a model of a competitive firm, but the issues raised and
problems created by the work of Schultz and Hotelling remained
clearly in Samuelson’s mind.

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 287

As a partial digression, I should mention the 1932 work of Hotelling

on the related problem of profit-maximization by a firm not subject
to a budgetary constraint. Integrability conditions arise there which
are related to, but distinct from, those that arise in the case of a con-
sumer under budgetary constraint. In this connection, the reader can
be referred to Henry Schultz’ 1937 book which gives an extensive
discussion of the integrability question: while not itself an entirely
satisfactory resolution of the problem, Schultz’ pages give a very
good summary of the uncertainties that pervade the literature.
(Samuelson 1950, 357; emphasis in original)
Although Samuelson’s theory of revealed preference is proudly dis-
played in the demand theory chapter of every microeconomics text-
book, a number of recent commentators have concluded that the over-
all project has been a dismal failure. As Shira Lewin (1996, 1316) put it
in a recent survey of economics and psychology: “In spite of Paul
Samuelson’s (1938) high hopes, the revealed preference approach had
proved empirically useless. Like the indifference theory it was meant to
replace, it too would become an artificial theoretical construct of little,
if any, explanatory value.”
Of course, as damning or as interesting as such criticisms might be,
they do not detract from our thesis that Samuelson’s revealed prefer-
ence approach to the theory of demand represents a third effort to extri-
cate demand theory from the Hotelling-Schultz impasse.

How We Stopped Worrying and

Learned to Love Walras
Since our little chronicle has gone on for quite a while and has taken a
number of twists and turns, it is perhaps useful to review the central
story line. We have four main themes, or perhaps three main themes
and a point of departure. The point of departure is the existence of
diversity in 1930s demand theory. In the 1930s, there were many the-
ories of demand, differing primarily but not exclusively with respect to
what each viewed as the relevant foundational invariant; these include
Henry Moore’s “just a statistical relationship” view; the Cournot-Cas-
sel “start with demand functions” view; Hicks and Allen’s “marginal
rates of substitution” approach; Hotelling’s 1932 work; revealed pref-
erence; Slutsky; and other less influential approaches. The American
scene was one of hearty diversity and thus a far cry from its common

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 288 Philip Mirowski and D. Wade Hands

image as the passive repository for smooth transatlantic transmission

of Marshallian or Paretian ideas. Even when one narrows the field to
authors who were self-consciously “neoclassical”— those who started
from individual preference or utility, used mathematical optimization,
and so on — profound diversity remains. Although this point of depar-
ture is not standard in the historical literature, we do not think it very
controversial, and we have provided here very little documentation to
support its existence (Hands and Mirowski 1998 does better, but even it
is not complete).
The first of the three themes this article does document is that
although the multiple branches were pruned during the period of neo-
classical stabilization, they did not get pruned back to a single stem. A
better analogy than pruning off branches from a single trunk is to think
of many small threads being twisted together into a much tighter bun-
dle or skein but never forming a single homogeneous strand. The sec-
ond point is that the Chicago school, the Cowles approach, and Samuel-
son’s revealed preference theory constitute the three major distinct
strands of the postwar skein. These three programs vary with respect to
stated purposes; the importance of empirical testing; the underlying
vision of scientific method; ideological sensitivities; aggregation; and,
most relevant to our third theme, the role of income and income effects.
This third lesson, building on our earlier work, is that the Hotelling-
Schultz impasse was the obligatory passage point for all three of these
versions of demand theory; each was a response to the problems
encountered by Schultz and Hotelling in the early 1930s, and each is
distinctively defined by the way in which it tried to circumvent these
difficulties.
Even those persuaded by our narrative may still ask, So what? Even
if there are three strands of neoclassical demand theory — the Chicago,
Cowles, and revealed preference approaches — and even if they are in
fact three different responses to the Hotelling-Schultz impasse, so
what? Although we generally feel that interpretations are the responsi-
bility of readers and not authors, we are willing to close with two brief
comments regarding the overall direction and significance of our proj-
ect. The first is that it is a big project, and this is just one small part of
a much larger story that is yet to be told and that even in the earliest
stages seems to be a far different story than that available, explicitly or
implicitly, in any extant historical literature. Second, we believe that
our story helps us understand the strength of neoclassicism in a novel

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 289

way. Rather than saying it simply chased out the competition — which
it did, if by “competition” one means the institutionalists, Marxists, and
Austrians — and replaced diversity with a single monolithic homoge-
neous neoclassical strain, we say it transformed itself into a more
robust ensemble. Neoclassical demand theory gained hegemony by
going from patches of monoculture in the interwar period to an inter-
locking competitive ecosystem after World War II. Rather than present-
ing itself as a single, brittle, theoretical strand, neoclassicism offered a
more flexible, and thus resilient skein of three strands. Each subprogram
had the capacity to absorb certain forms of criticism and thus deflect
those criticisms away from the vulnerable areas in other subprograms,
where they might do the most damage. Attack general-equilibrium the-
ory for its lack of empirical relevance, and one is quickly directed
toward Becker, Stigler, and the grand Chicago tradition of applied
microeconomics. Go after Chicago for its loose use of subjective utility
and a priori categories, and one is quickly told how the relevant prefer-
ences are operationally defined and could be revealed by direct observa-
tions of consumers’ choices. Finally, go after revealed preference theory
for emptiness as a tautological definition of consistency in choice, and
one is immediately shown the way through the equivalence results and
right into the heart of the complete Walrasian general-equilibrium model.
Pace Harry Truman, the buck stops nowhere. We think this insight goes
a lot further toward explaining the postwar success of neoclassical eco-
nomics than any of the arguments presented in the standard internalist
or externalist histories.

References
Aldrich, John. 1996. The Course of Marshall’s Theorizing about Demand. HOPE
28:171 – 217.
Bailey, Martin. 1954. The Marshallian Demand Curve. Journal of Political Econ-
omy 62:255 – 61.
Becker, Gary. 1962. Irrational Behavior and Economic Theory. Journal of Political
Economy 70:1 – 13.
Bronfenbrenner, Martin. 1962. Observations on the Chicago School(s). Journal of
Political Economy 70:72 – 75.
Christ, Carl. 1952. History of the Cowles Commission. 1932 – 52. In Economic The-
ory and Measurement: A Twenty Year Research Report, 1932–1952. Chicago:
Cowles Commission.
———. 1994. The Cowles Commission’s Contributions to Econometrics at
Chicago, 1939 – 1955. Journal of Economic Literature 32:30 – 59.

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 290 Philip Mirowski and D. Wade Hands

Cohen, Joshua. 1995. Samuelson’s Operationalist-Descriptivist Thesis. Journal of

Economic Methodology 2:53 – 78.
Epstein, Roy. 1987. A History of Econometrics. Amsterdam: North-Holland.
Friedman, Milton. 1941. Review of Monopolistic Competition and General Equi-
librium Theory, by R. Triffin. Journal of Farm Economics 23:389 – 90.
———. 1946. Lange on Price Flexibility and Employment. American Economic
Review 36:613 – 31.
———. 1949. The Marshallian Demand Curve. Journal of Political Economy
57:463 – 95.
———. 1951. Comment on Christ. In NBER Conference on Business Cycles. New
York: NBER.
———. 1966. Price Theory: A Provisional Text. Rev. ed. Chicago: Aldine.
Friedman, Milton and Simon Kuznets. 1945. Income from Independent Profes-
sional Practice. New York: National Bureau for Economic Research.
Friedman, Milton, and W. A. Wallis. 1942. The Empirical Derivation of Indiffer-
ence Functions. In Lange, McIntire, and Yntema 1942.
Hammond, J. Daniel. 1993. An Interview with Milton Friedman. In Philosophy and
Methodology of Economics. Vol. 1. Edited by B. J. Caldwell. Aldershot: Edward
Elgar.
Hands, D. Wade, and Philip Mirowski. 1998. Harold Hotelling and the Neoclassi-
cal Dream. In Economics and Methodology: Crossing Boundaries. Edited by
R. Backhouse, D. Hausman, U. Mäki, and A. Salanti. London: Macmillan.
Hendry, David, and Mary Morgan, eds. 1995. The Foundations of Econometric
Analysis. Cambridge: Cambridge University Press.
Hicks, John. 1939. Value and Capital. Oxford: Oxford University Press.
Hildreth, Clifford. 1986. The Cowles Commission in Chicago. Berlin: Springer-
Verlag.
Hirsch, Abraham, and Neil De Marchi. 1990. Milton Friedman. Ann Arbor: Uni-
versity of Michigan Press.
Hood, W., and T. Koopmans, eds. 1953. Studies in Econometric Method. Cowles
Commission Monograph 14. New Haven, Conn.: Yale University Press.
Hotelling, Harold. 1932a. Edgeworth’s Taxation Paradox and the Nature of Demand
and Supply Functions. Journal of Political Economy 40:577 – 616.
———. 1932b. Review of Ragnar Frisch’s New Methods of Measuring Marginal
Utility. Journal of the American Statistical Association 27:451 – 52.
———. 1935. Demand Functions with Limited Budgets. Econometrica 3:66 – 78.
———. 1954. Hotelling to Lloyd Reynolds, 24 September. Box 5, Miscellaneous
Correspondence “R,” Harold Hotelling Papers, Columbia University, New York.
———. N.d. On the Nature of Demand and Supply Functions. Box 26, File “Math-
ematical Economics,” Harold Hotelling Papers, Columbia University, New York.
Houthakker, Hendrik. 1950. Revealed Preference and the Utility Function. Eco-
nomica 17:159 – 74.
———. 1983. On Consumption Theory. In Paul Samuelson and Modern Economic
Theory. Edited by E. C. Brown and R. M. Solow. New York: McGraw-Hill.

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 Postwar Stabilization 291

Ingrao, Bruna, and Giorgio Israel. 1990. The Invisible Hand: Economic Theory in
the History of Science. Cambridge, Mass.: MIT Press.
Knight, Frank. 1944. Realism and Relevance in the Theory of Demand. Journal of
Political Economy 52:289 – 318.
Koopmans, Tjalling, ed. 1950. Statistical Inference in Dynamic Economic Models.
Cowles Commission Monograph 10. New York: Wiley.
Lange, Oskar, Francis McIntire, and Theodore Yntema, eds. 1942. Studies in Math-
ematical Economics and Econometrics. Chicago: University of Chicago Press.
Leonard, Robert. 1991. War as a Simple Economic Problem. In Economics and
National Security. Edited by C. Goodwin. Durham, N.C.: Duke University Press.
———. 1995. From Parlor Games to Social Science: Von Neumann, Morgenstern,
and the Creation of Game Theory, 1928 – 1944. Journal of Economic Literature
33:730 – 61.
Lewin, Shira. 1996. Economics and Psychology. Journal of Economic Literature
34:1293 – 1323.
Marschak, Jacob. 1939a. Personal and Collective Budget Functions. Review of Eco-
nomics and Statistics 21:161 – 70.
———. 1939b. Review of Henry Schultz’s Theory and Measurement of Demand.
Economic Journal 49:486 – 89.
———. 1943. Demand Elasticities Reviewed. Econometrica 11:25 – 34.
———. 1944. Marschak to Robert Redfield, 15 February. Rockefeller Foundation
Archives, Sleepy Hollow, N.Y.
Mirowski, Philip. 1989a. The Measurement without Theory Controversy. Econ-
omies et sociétés 11:65 – 87.
———. 1989b. More Heat than Light. New York: Cambridge University Press.
———. 1989c. The Probabilistic Counter-revolution. Oxford Economic Papers
41:217 – 35.
———. 1990. Problems in the Paternity of Econometrics: Henry Ludwell Moore.
HOPE 22:587 – 609.
———. Forthcoming a. Machine Dreams. Cambridge: Harvard University Press.
———. Forthcoming b. RAND/OR. Social Studies of Science.
Morgan, Mary. 1990. The History of Econometric Ideas. Cambridge: Cambridge
University Press.
Mosak, Jacob. 1938. Interrelations of Production, Price, and Derived Demand.
Journal of Political Economy 46:761 – 84.
Patinkin, Donald. 1995. The Training of an Economist. Banca Nazionale del
Lavoro Quarterly Review 48:359 – 95.
Reder, Melvin. 1982. Chicago Economics: Permanence and Change. Journal of
Economic Literature 20:1 – 38.
Samuelson, Paul. 1938. A Note on the Pure Theory of Consumer’s Behaviour. Eco-
nomica 5:61 – 71.
———. [1948] 1968. Consumption Theory in Terms of Revealed Preference. Eco-
nomica 15:243 – 53. Reprinted in Utility Theory: A Book of Readings. Edited by
A. N. Page. New York: Wiley.

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user
 292 Philip Mirowski and D. Wade Hands

———. 1950. The Problem of Integrability in Utility Theory. Economica

17:355 – 85.
———. 1965. Professor Samuelson on Theory and Realism: Reply. American Eco-
nomic Review 55:1164 – 72.
———. 1972. Maximum Principles in Analytical Economics. American Economic
Review 62:249 – 62.
———. 1983. Economics in a Golden Age: A Personal Memoir. In Paul Samuelson
and Modern Economic Theory. Edited by E. C. Brown and R. M. Solow. New
York: McGraw-Hill.
———. 1992. My Life Philosophy: Policy Credos and Working Ways. In Eminent
Economists: Their Life Philosophies. Edited by M. Szenberg. New York: Cam-
bridge University Press.
Schultz, Henry. 1938. The Theory and Measurement of Demand. Chicago: Univer-
sity of Chicago Press.
Sonnenschein, Hugo. 1972. Market Excess Demand Functions. Econometrica
40:549 – 63.
Stigler, George. 1952. The Theory of Price. Rev. ed. New York: Macmillan.
———. 1965. Essays in the History of Economics. Chicago: University of Chicago
Press.
———. 1988. Memoirs of an Unregulated Economist. New York: Basic Books.
Weintraub, E. Roy. 1991. Stabilizing Dynamics: Constructing Economic Knowl-
edge. New York: Cambridge University Press.
Weintraub, E. Roy, and Philip Mirowski. 1994. The Pure and the Applied: Bour-
bakism Comes to Mathematical Economics. Science in Context 7:245 – 72.
Wong, Stanley. 1978. The Foundations of Paul Samuelson’s Revealed Preference
Theory. Boston: Routledge and Kegan Paul.

Downloaded from https://read.dukeupress.edu/hope/article-pdf/30/Supplement/260/426779/ddhope_30_Supplement_260.pdf

by THE NEW SCHOOL user