I. INTRODUCTION
The effort to make economics mathematical is set in different historical moments, each according a transformed place for mathematics in economics. Among them one may cite the first introduction of calculus in the 1860s, the increased use of statistics in the late nineteenth century, or the development of econometrics in the 1920s and 1930s.
The one that served as a critical turn, though, where economics, conceived as a science akin to physics, but especially to mechanics, was to be comprehended as mathematical in its language, techniques, and formal relations or structures, occurred with the rise to prominence of neoclassical economics at the turn of the twentieth century. But, as is the case for any historical moment, one ought to inquire as well about paths not pursued. What alternative ways of thinking about, or constructing, economics might have been available in that mathematical turn?
Two essential and noteworthy developments in the period of the ascendancy of neoclassical economics mark the mathematical turn within it. First of all, the marginalist rethinking of the field tended to encourage the greater use of, and reliance upon, geometric representation, culminating in Alfred Marshall’s early synthesis of neoclassical economics, which, despite Marshall’s own reluctance and ambiguity about the role of mathematics in economics, produced what still remains a recognizable microeconomics replete with graphical techniques.
Second, for all the broad differences on the subject among the marginalist schools, with Carl Menger the most notable exception, there were those like Stanley Jevons, Francis Edgeworth, Leon Walras, Vilfredo Pareto, and Irving Fisher who championed the mathematizing of economics. As distinct from those who sought to create statistical measures in economics (Porter Reference Porter, Klein and Morgan2001, pp. 17–18), they envisioned linking the quantification of economic phenomena to establishing algebraic or other mathematical relationships across economics variables.
Among such economists, Walras was perhaps the most ardent, as he saw his relentless and seemingly unyielding advocacy of a suitably mathematical economics met with disdain in France, resulting in his being regarded as an outsider by its established academic world of political economy. His main economic work, Elements of Pure Political Economy, through its revisions the cynosure of his writing over a lifetime, was intended to demonstrate the possibilities of an algebraic construction of, and solution to, the interplay of resources, output, and demand in any national economy. His extensive correspondence, organized and made available to a wider public audience by William Jaffe (Walras Reference Walras and Jaffe1965), reveals Walras’ dogged and sometimes beleaguered campaign to win support for his idée fixe.
Yet, there is more than one irony in this period’s turn toward mathematizing economics, especially as one looks more closely at Walras and his intellectual milieu. For one, as Philip Mirowski noted, the case for mathematizing economics was typically advanced as a way of making economics more scientific, through an alignment of economics with classical mechanics, at a time when its presumptive universality was already being delimited by contrasts with thermodynamics and when the core ideas of classical physics would shortly undergo the challenges of relativity and quantum theory (Mirowski Reference Mirowski1989, pp. 59–89). Moreover, both relativity and quantum theory drew upon, and entailed a recasting of, the nature and understanding of physical measurement, with the introduction of the idea of space–time through the former and the principle of uncertainty through the latter.
It was also a period in which mathematicians found themselves engaged in strenuous debate, lasting several decades, as efforts to achieve greater rigor—typically with greater formality—and to counter inconsistencies in mathematics led to a crisis about the nature of the foundations of mathematics. As E. Roy Weintraub has pointed out, there was a palpable shift over the course of the nineteenth century from an alignment of mathematical rigor with the capacity of mathematics to capture the truthfulness of the facts of the external world to an emphasis upon the need for internal consistency and the avoidance of contradiction (Weintraub 2001; also Weintraub 2002, pp. 48–49). Yet, the explicit grounding of all mathematics in the rules of logic or through a limited number of fundamental principles or axioms could not eliminate all such inconsistencies.
But perhaps most striking, if little noticed, in Walras’ case was the fact that France’s foremost mathematicians, including those whose views on the prospect of mathematizing economics he had sought out, raised questions and concerns that might easily have guided this project in another direction, away from the neoclassical tack it tended to take. Of the French mathematicians active and prominent in academic and professional circles around 1900, one might readily include, among others, Gaston Darboux, by then permanent secretary of the French Academy (Grattan-Guinness Reference Grattan-Guinness1997, p. 487); Henri Poincare, the most outstanding figure of the period who bridged the world of mathematics and physics; Emile Picard, who provided a similar bridge; Jacques Hadamard, Picard’s foremost student; and Emile Borel.
If one considers carefully the specific exchanges between Walras, on the one hand, and Henri Poincare and Emile Picard, on the other, one can discern within the writings of these mathematicians a raft of major elements for a modified economics program. Moreover, a similar set of such elements can be found as well in the writings of a third French mathematician with whom Walras corresponded, Emile Borel.
This modified economics program would include: the necessity of identifying with forthright clarity the assumptions underlying and limiting any economic hypothesis put forward; a recognition of the central role of uncertainty in assessing economic behavior; the complications posed by the discontinuities in economic measurement and the possibility of using qualitative measures that might require a different set of mathematical techniques; and an awareness of the place of memory and history in economic systems, or what has commonly become known as “hysteresis,” a term used by both Poincare and Picard.
Yet, for all intents and purposes, what these mathematicians proffered was a road not taken at that juncture. In part, one may read this as a desire on Walras’ part to accept the apparent approbation of Poincare and Picard, and cite the apparent approbation of Emile Borel, Gaston Darboux, Maurice d’Ocagne, and Jules Tannery (Walras Reference Walras and Jaffe1965, pp. 314, 352), rather than absorb broadly the points they raised, in either their responses to him or in their own writings.Footnote 1 To the end, in 1908 and 1909, Walras identified Poincare, Picard, and Borel as supporters of his approach to economics, whom he hoped, without apparent success, might assist in promoting his last work, “Economique et mecanique,” intended to make plain the alignment of a mathematical economics with mechanics, to the mathematical scientific community (Letters #1677, 1722, Walras Reference Walras and Jaffe1965, pp. 347, 388).
It was also the case that there were others, like the Italian mathematician Vito Volterra, who accepted the basic premise that economics ought to become more mathematical, along the lines of the physical sciences, though, as Pareto discovered, even Volterra could find difficulties in the details, notably in the integratibility of the utility function (Mirowski Reference Mirowski1989, pp. 248–249; Weintraub 2002, pp. 32–35, 48–51).
The inclination not to modify the program of economics may reflect as well a seemingly commonsensical understanding about the nature of science and the role of mathematics in it, where mathematics, whether in physics or economics, was grasped intuitively as serving as a medium, whereby number, and hence quantification, in words enshrined by Tobias Dantzig, became “the language of science” (Dantzig Reference Dantzig1930). And, to the extent that a framework for mathematics might be applied to economics, Walras’ own comprehension of the uses of mathematics in mechanics inclined him to view it as offering up an axiomatic structure, even though that structure and its primitives were actually dependent upon an analogy with classical mechanics (Turk Reference Turk2006). This was quite distant and distinct from the conventionalism that characterized Poincare’s mathematical thought, where experience and imagination shaped mathematical concepts rather than being shaped by idealized principles and primitives, and influenced the choice of hypotheses that might be fitted to a given set of facts (Heinzmann and Nabonnand 2008).
II. LANGUAGE, TECHNIQUE, OR FORMAL RELATION?
To appreciate what matters hung in the balance in this century-old mathematical turn in economics, it is necessary at the outset to step back and recognize the ambiguities and critical shadings entailed by the notion of ‘mathematizing.’ In part, the program of mathematization rested upon an understanding of mathematics as it was re-formed and shaped in the course of the nineteenth century into analysis, with a new emphasis upon rigor and a close bond with what came to be known as mathematical physics (Kneebone [1963] 2001, pp. 139–140; Grattan-Guinness Reference Grattan-Guinness1997, pp. 312–317, 405–408).
This kind of mathematical thought was identified along a chronological and intellectual pathway through such figures as Joseph-Louis Lagrange, Augustin-Louis Cauchy, and Karl Weierstrass, where Lagrange sought to algebraize mechanics, while Cauchy replaced more loosely conceived notions about limits and the infinite with more formal and precise constructs of them, formalizing proofs in the process (Grattan-Guinness Reference Grattan-Guinness1997, pp. 373–376), and Weierstrass led the effort to set all mathematics as analysis, free of geometric representation and intuition (Grattan-Guinness Reference Grattan-Guinness1997, pp. 481–486). In this transformation it would prove difficult to distinguish the mathematics that economics would seek to appropriate, from the science, classical physics, that found mathematics essential to its expression. As Weintraub put it, by the late nineteenth century, the effort to make economics as solid as a hard science required that it “emulate” mathematics itself (Weintraub 2002, p. 37).
The development of analysis represented a major trajectory in the broader advance of mathematical ideas, but by the same token, this meant that other understandings of mathematics might exist and inform a mathematization of economics. In the case of the early twentieth-century turn in economics toward greater mathematization, one might think of topology, or, as it was then labeled, analysis situs, as one such alternative, a mathematical subfield pioneered by Poincare and explored by Picard.
There are three different possible understandings of the nature of mathematics and the role it might play in economics that are routinely melded into each other, although they may be viewed as quite distinct in crucial ways.
The first of these is the notion that, as Paul Samuelson later put it in the motto for the Foundations of Economic Analysis: “mathematics is the language of science” (Samuelson Reference Samuelson1965, p. 1751). Walras himself had taken the position that mathematics would afford economists a clarifying and more precise language with which they might more readily engage in a scientific inquiry. In a letter to the Nobel Peace Prize Committee on January 22, 1906, Walras praised the work of Vito Volterra, citing the latter’s article entitled “Les mathematiques dans les sciences biologiques et sociales,” which had just been translated into French and published in La Revue du Mois twelve days before:
The author, as he himself states, is more a mathematician than an economists; but, just the same, no one has better recognized in our ‘analytical economics’ the proceedings by the same means as which Maxwell and Helmholtz have transformed physics and physiology, and makes comprehensible how economic questions such as free exchange, for example, which are and remained so tangled and confused in everyday language, can be brought to a solution so clear and sure through the use of the mathematical method (Letter #1618, Walras Reference Walras and Jaffe1965, p. 296).
It may be the case that Walras had not read—or read too closely—Volterra’s article. In his notations to Walras’ letter, Jaffe remarked that Volterra had made no reference to “free exchange” (Walras Reference Walras and Jaffe1965, p. 296). It is also worthwhile to note that in the article, Volterra had approvingly cited Walras, along with Whewell, Cournot, and Gossen, for their “assimilation of mathematical methods” (Volterra Reference Volterra1901, p. 447). Of these four, three belonged to an earlier generation: Whewell died in 1866, Cournot in 1877, and Gossen in 1858. Thus, Walras stands out as the only listed living contemporary of Volterra, suggesting perhaps that Walras’ overwhelming reaction served as more than a bit of an act of mutual admiration.
The identification of the mathematical method in economics with the resort to a language affording scientific precision also can be found a generation later in the writings of Charles Roos, an early American advocate of the mathematical method in economics. In Dynamic Economics, published in 1934, Roos stated: “In order to state hypotheses so clearly that they will not be misunderstood, it is necessary to choose language carefully” (Roos 1934, p. 10). He goes on to describe the need for, and virtue of, a “symbolic language” in any scientific inquiry: “In fact, the more variables there are the more necessary it becomes to have a symbolic language to keep track of them. No one could advise an astronomer to develop a science of the stars without using mathematics” (Roos 1934, p. 11).
But mathematics could also be regarded as a toolbox, from which the best technique, or, at the least, the one most likely to succeed, would be chosen. In the process the mathematical economist became more of a professional craftsman, on the order of a jeweler whose skill in devising the best approach to cutting drew upon his/her experience and knowledge in sorting through the array of tools and techniques at his/her disposal. The range in technique has proven to be quite broad: as Walras made use of linear equations, so Roos made use of integral equations.
For his part, Edgeworth highlighted the possible gains to economic analysis through the use of the calculus of variations. He began by noting that “[i]t is the first principle of the calculus of variations that a varying quantity attains a maximum when the first term of variation vanishes, while the second term is negative (mutatis mutandis, for a minimum)” (Edgeworth [1881] 1932, p. 91). He then applied this principle to economics: “In the simple cases which in the infancy of Mathematical Psychics are along presented in these pages, we know by observation not what the second term is, but that it is continually negative” (Edgeworth [1881] 1932, p. 91).
As Edgeworth’s text suggests, mathematical methods and techniques might, on occasion, also be regarded as fundamentally formal and structural: as mathematical statements expressing universal relations that hold regardless of any specific content. This notion is akin to Ivor Grattan-Guinness’ idea of “structure-similarity” in mathematical forms and patterns that precede any attribution of meaning to them (Grattan-Guinness 1992, p. 93).
It may provide, as well, a mathematical resonance between economics and physics, like the one identified by Samuelson in his later reflections upon the genesis of the Foundations of Economic Analysis in the 1930s:
I was vaccinated early on to understand that economics and physics could share the same formal theorems (Euler’s theorem on homogeneous functions, Weierstrass’s theorems on constrained maxima, Jacobi determinant identities underlying LeChatelier reactions, etc.), while still not resting on the same empirical foundations and certainties (Samuelson Reference Samuelson1998, p. 1376).
This, then, serves as the backdrop to the alternatives, whether expressed fully or only in inchoate ways, that economists early on in the twentieth century might divine as pertinent in mathematics to the nature and workings of economics. For Walras, whom Samuelson had praised as the Newton of economics (Samuelson Reference Samuelson1965, p. 1756), both the power associated with the more scientific discourse of quantification through the use of algebra, and the possibility of analogizing formal elements and structures from one discipline to another, had great appeal.
III. ON THE IMPORTANCE OF MEASURE
The exchange of letters between Leon Walras and Henri Poincare took place largely in 1901; the exchanges between Walras and Emile Picard in 1906.Footnote 2 Walras’ letters to Emile Borel followed shortly thereafter. This was the last decade of Walras’ life, and so his effort to confirm the legitimacy of his life’s work and influence the future course of economics as a discipline was, if anything, reinforced in this period.
The exchange of letters between Walras and Poincare has been largely overlooked, with the critical exception of an essay by William Jaffe dating from 1977 (Jaffe Reference Jaffe and Walker1983, pp. 213–220) and a discussion of the import of Poincare’s response for notions of attaining a general equilibrium by Bruna Ingrao and Giorgio Israel in The Invisible Hand, published in 1990 (Ingrao and Israel 1990, pp. 154–161). Yet, the discussion of the nature of scientific measurement and, hence, the role of mathematics in the construction of science is central to the question that prompted the exchange: what is the proper role and scope of mathematics in economics?
That is so at least in part because of the timing. For Walras, the issue had been joined with greater urgency when Hermann Laurent, a mathematician with whom Walras had a long-standing relationship and had engaged in an especially extensive correspondence in December 1898 and again in May 1900, threw down the gauntlet at an annual conference of mathematical actuaries in 1900, and challenged the validity of applying mathematics to notions like satisfaction (Walras Reference Walras and Jaffe1965, p. 113), a core element in Walras’ “pure political economy,” and, like Stanley Jevons’ idea of utility, the essential concept in the reconstruction of economics in marginalism and ultimately in neoclassical economics.
Philip Mirowski has excoriated the feebleness of Walras’ response to Laurent, emphasizing the weakness of Walras’ mathematical intuition and skills and his tendency to cite the approbation of others, especially those with strong scientific credentials, as proof of the rightness of his theory. For Mirowski, Walras’ failing here to grasp the challenge and address it adequately—and a similar, contemporaneous such failing by Pareto in response to an otherwise highly supportive Vito Volterra—represented the hollowness of the mechanical analogy upon which they based their mathematization of economics (Mirowski Reference Mirowski1989, pp. 241–250).
As to Laurent’s challenge to Walras, Mirowski paid particular attention to Laurent’s ‘compression’ of Walras’ case into mathematical language, readily amenable to problem solving through the technique of integration, and revealing thereby an unresolved factor of integration necessary for measuring rarete that called into question—and, in Mirowski’s view—nullified the physical metaphor underlying Walras’ economic model (Mirowski Reference Mirowski1988, pp. 32–24). Of course, the reference to metaphor also suggests that what was being sought by Walras, and contested by Mirowski, was a kinship of form or structure established through mathematics. As Mirowski saw it, the “program of mathematical economics” required both “a shared language and a shared metaphor” (Mirowski Reference Mirowski1991, pp. 147–148).
Mirowski, though, did not pursue the full extent of Walras’ efforts to salvage and ultimately sustain his mission to mathematize economics.Footnote 3 Walras turned to Henri Poincare, the pre-eminent French mathematician of his day, and asked if Laurent was right or not (Letter #1492). Initially, Poincare responded by affirming the possibility that economics could be mathematized, within certain limits (Letter #1494). Walras pressed harder. Had he exceeded those limits (Letter #1495)? Poincare then responded once more, again rather swiftly, within a week of receiving Walras’ query (Letter #1496), in terms that Walras, not unreasonably, interpreted as positive, but which came with reservations and a difference in perspective that rendered the response quite ambiguous.
One may surmise, if only on textual grounds, that the rapidity of Poincare’s response was due in part to the attention he had been giving to the matter of measurement in science, most especially with regard to physics. That same year, 1901, saw the publication of the first of Poincare’s three popular works about the nature of science, Science and Hypothesis. One can identify phrases in Poincare’s letter to Walras that are drawn directly from Science and Hypothesis, specifically the subsection on “measurable magnitude” and his discussion of magnitudes that are not ‘measurable.’ In particular: “we can indeed say whether a given one of these magnitudes is greater than another, but not whether it is twice or thrice as great” (Poincare Reference Poincare1905, p. 23).
One may compare this wording with Poincare’s response to Walras:
Your definition of rarete impresses me as legitimate. And this is how I should justify it. Can satisfaction be measured? I can say that one satisfaction is greater than another, since I prefer one to the other, but I cannot say that the first satisfaction is two or three times greater than the other. That makes no sense by itself and only some arbitrary convention can give it meaning. Satisfaction is therefore a magnitude but not a measurable magnitude (Letter #1496, translated by and in Jaffe Reference Jaffe and Walker1983, p. 217).
Thus, from Poincare’s perspective, the matter raised by Walras had little to do with any questioning of the status of economics as a field of intellectual inquiry—should it be regarded as suitably and sufficiently scientific, or not?—but instead turned on the efficacy and appropriateness of scientific measurement, thereby challenging established notions about how science itself was constituted.
Poincare’s reflections on the nature of measurement drew upon a half-century of examination and reimagination among leading mathematicians of the “problem of space” (Friedman Reference Friedman, Greffe, Heinzmann and Lorenz1996, pp. 335–338), initiated by the recognition that the intuitive understanding of space thought to have been captured by Euclidean geometry represented only one of several different possible geometries calling forth different standards of measurements and different shapes to, and curvatures of, space, and conditions for the stability of objects in such spaces, including celestial bodies, needed to be re-examined.
For Poincare, this led to seeing many of the physical concepts through which measurements were made, whether cast in terms of Newtonian physics or from the vantage point of energetics, as merely conventions (Poincare Reference Poincare1905, p. 67ff). He tried as well to introduce the pertinence of a more qualitative form of measurement, paradox notwithstanding, found in unmeasurable magnitude (Poincare Reference Poincare1905, pp. 23–24).
In effect, in responding to Walras, Poincare set Walras’ query about the possibility of giving ‘satisfaction’ a suitable mathematical measure in the context of his own work. Thus, while Walras was keen to establish the efficacy of mathematics in the discourse of economics so as to buttress the scientific claims of the discipline, Poincare was concerned with the limitations of a mechanistic view of science in general and, hence, the scope and limits of mathematics in any science.
Consider the following juxtaposition: in his letter to Poincare, Walras relies upon an understanding of the notion of ‘mass’ and its capacity to be measured through an imagined counting that Poincare would in no way accept. Walras cites the definition of the mass of any physical body in Poinsot’s Statics, the work from which Walras built his analogy of economics to classical physics, as “the number of molecules that compose it” (Letter #1495). Such a number was not actually countable, yet formed the basis of physical construction and informed mathematical formulations of physical laws. Why should the satisfaction of an individual be regarded as any less suitable for a similar quantification and mathematical construction (Letter #1495)? By contrast, in the section on measurement in Science and Hypothesis, Poincare challenged the concept of ‘mass’ as found in, and derived from, Newton’s laws as fundamentally tautological (Poincare Reference Poincare1905, pp. 73–76).
Moreover, from Poincare’s perspective, setting foundations for any scientific inquiry was questionable, as it was only “relations between objects” that mattered in the construction of any science (Poincare Reference Poincare1905, p. 18). Hence, if satisfaction—or, presumably, any other economic category or variable—could be comprehended as elucidating a set of pertinent relations, then it passed the threshold of science. In his letter to Walras, Poincare finds that relation in “defin[ing] satisfaction by any arbitrary function providing the function always increases with an increase in the satisfaction it represents” (Letter #1496, Jaffe translation).
At first blush, this may appear to meet the terms of the third way that economics could be rendered mathematical, by establishing a structural correspondence between a formal mathematical relation and an economic relation, but the strictures of Poincare’s conventionalism rapidly call that into doubt. First, there would have to be a basis in experience that “guides” the construction of that formal relation (Poincare Reference Poincare1905, p. 53). Otherwise, a ‘formal relation’ would remain an abstraction, a product merely of a choice of words. Second, the arbitrary functions used to construct the correspondence constitute a scaffolding that must, in the end, be removed for the relation exhibited to be meaningful. “If the arbitrary functions still appear in the conclusions, the conclusions are not false, but they are totally without interest, because they depend upon the arbitrary conventions made at the start” (Letter #1496, translation in Jaffe Reference Jaffe and Walker1983, p. 217).
Intriguingly, Poincare gave Walras a pass on this objective. Yet, Jaffe correctly noted that that was not the case. “In [Walras’] own proof of the fundamental theorem of proportionality of marginal utilities to parametric prices as a condition of equilibrium . . . the arbitrary assumption of diminishing utility remains embedded in the conclusion and entails the retention throughout of the further arbitrary assumption of cardinally measurable utility” (Jaffe Reference Jaffe and Walker1983, p. 218). Ingrao and Israel also found that Walras did not meet Poincare’s strictures, but from an altogether different vantage point. They take as suspect and arbitrary the assumption retained by Walras that different individuals respond uniformly in their satisfaction to various combinations of goods (Ingrao and Israel 1990, p. 158). One suspects that this challenge, rooted in the subjectivity of human response, was at the heart of Laurent’s initial rebuke to Walras. If Ingrao and Israel are right, then Poincare also completely missed the mark. Poincare’s focus was upon whether satisfaction itself was measurable, rather than how satisfaction that is experienced by different individuals would meet a standard of sufficient uniformity and consistency as to be measurable.
IV. POINCARE’S RESERVATIONS
Overall, Poincare appeared supportive of Walras’ efforts. Nonetheless, it may be fair to ask if Walrasian ‘satisfaction’ did meet the further test laid out by Poincare. Poincare had offered up two reservations. First, he took the positing of unbridled self-interest as a plausible approximation of individual behavior. Second, and with greater hesitation, he questioned the reliance upon full clairvoyance, the notion that individuals have a complete and certain grasp of the future.
Walras did not explore the implications of either of these reservations. Instead, he marginalized their gist, taking note of them essentially as subsumable deviations of an idealized treatment of consumer behavior, akin to the frictionless world of classical physics. What follows is his response to Poincare:
As to hypotheses, it is quite certain that one must be on guard when one passes from abstractions to reality. In reality there are frictions in the economic mechanism; and as to the other concerns [raised by Poincare] men are not utterly self-interested nor perfectly clairvoyant. From this it follows that the applied theory of the production of wealth should indicate these frictions with care and bring about their suppression as completely as possible in view of approximating the maximum of utility as closely as possible (Letter #1498, Walras Reference Walras and Jaffe1965, p. 167).
Nevertheless, one could envision the potential impact of probabilities or asymmetries if such reservations were taken fully into account. More generally, Poincare’s reservations could be regarded as the basis for weighing and incorporating the consequences of a lack of sufficient information in the analysis and interpretation of consumer behavior. In Walras’ schema, as it was for Irving Fisher and Vilfredo Pareto, the individual in economics was aligned with the molecule in mechanics. Yet, nowhere in statistical mechanics did molecules have to contend with clairvoyance, or the lack thereof, as they collided with one another. Even if at only an initial stage, Poincare had opened the way to an appreciation of the problem of integrating intention and consciousness into economic matters.
In his paper examining the Walras–Poincare correspondence, William Jaffe essentially cast the matters discussed into a narrative of realization whereby the recognition that utility ought to be understood as an ordinal rather than cardinal measure is seen as a “remarkable premonition [by Poincare] of subsequent trends in microeconomic theory” (Jaffe Reference Jaffe and Walker1983, p. 218). This, even on its face, seems unduly positive in its reach. There is no evidence that Walras took up Poincare’s insight or his conventionalist critique in his own work, or shared it with other economists. Nor is there any evidence that economists of the succeeding generation had knowledge of, or had somehow absorbed into their own thinking, his caution about the need to contemplate economic behavior beset by a lack of clairvoyance.
Moreover, that Poincare’s comments cannot simply be read as an endorsement of ordinal over cardinal measures of utility, as Jaffe envisioned them to be, is apparent from Poincare’s discussion of the physical continuum in Science and Hypothesis, where full transitivity, a critical element in the ranking system of ordinal measure, cannot be maintained. Relying upon Ludwig Fechner’s standards for measurement, Poincare saw contradictions in measurement emerge that led him to discern a disjuncture between the physical and the mathematical continuum (Poincare Reference Poincare1905, pp. 25–28). At the same time, it would be too facile to see Poincare’s critique of classical physics as effectively deconstructing the kind of scientific framework that Walras sought for economics, as Mirowski would have it, simply because Poincare did give explicit support to the quest undertaken by Walras to mathematize economics, as long as it was done with suitably framed hypotheses. This means that the deconstruction one might reasonably see unfolding through Poincare’s insights and strictures must be somewhat more nuanced, establishing where the ‘relations between objects,’ built out of experience, would hold.
To what extent did Poincare’s introduction of the notion of ‘non-measurable’ magnitudes into his exchange with Walras open up the possibility of making use of qualitative yet mathematical measurement in economics? Although Jaffe interpreted this as the case for ordinal as opposed to cardinal measurement, one might demur on a number of counts. For one, as Nicholas Georgescu-Roegen noted, it might be possible to devise instead a measure of weak cardinality, partaking of certain elements of quantitative measure (Georgescu-Roegen Reference Georgescu-Roegen1966, p. 51). For his part, Poincare had identified a variability in the nature of measurement, using the example of temperature. He cited the introduction of the notion of absolute temperature as laying the groundwork for a new mathematical construct of temperature. While it had been limited formerly to unmeasurable magnitudes, based upon correspondences with changing intervals of columns of mercury, it could now be conceived as a mathematical function that was ‘constantly increasing’ (Letter #1496, Walras Reference Walras and Jaffe1965, p. 163).
One might extend Poincare’s critique of measurement further, to encompass qualitative changes for which a different kind of mathematics would be required, employing different sorts of techniques and drawing upon a different model for mathematical analysis; namely, topology. It raises the possibility that mathematization, including that of economics, need not be viewed as synonymous with the quantification of economic phenomena into mathematical variables. As the French mathematician Rene Thom, one of the major developers of ‘chaos theory’ and recipient of the 1958 Fields Medal, would later put it: “A topological analysis of a situation has a qualitative content which is not quantitative” (Thom Reference Thom1991, p. 93). This stands as a refutation of Ernest Rutherford’s apothegm that “Qualitative is nothing but poor quantitative” (Thom Reference Thom1991, p. 93), a ‘formula’ that recalls the contention by Walras, Roos, and Samuelson, among others, about the precision gained by mathematizing economics through a more scientifically established vocabulary.
In this context it is noteworthy that Poincare is often viewed as having laid the groundwork for chaos theory in an 1890 paper in Acta Mathematica entitled “Sur le problem des trois corps et les equations de la dynamique” (Grattan-Guinness Reference Grattan-Guinness1997, pp. 614–618). And, as Mirowski has noted, Poincare’s exploration of the implications of chaos theory led him to recognize the centrality of “stochastic processes” in classical mechanics itself (Mirowski Reference Mirowski1989, pp. 74–76).
V. OF ‘FROTTEMENTS,’ IRREVERSIBLE PHENOMENA, AND HYSTERESIS
In 1905 Emile Picard, one of France’s leading mathematicians, published a popular work on the current state of science, entitled La science moderne et son état actuel. In the tradition of such popular writing in France, Picard’s exploration of the subject was high-minded and intellectually engaging; not dissimilar, therefore, from Poincare’s popular writings about the nature of, and prospects for, science at the turn of the twentieth century.
Picard, while less well known than Poincare, also had a distinguished career as a mathematician, making particularly noteworthy contributions in algebraic geometry and analysis, for the latter of which he produced a standard multi-volume treatise (copies of which he sent to Walras as a courtesy) (Letters #1634, 1636, and 1637). He first taught at the Ecole Normale Superieure in Paris, where his most famous student was Jacques Hadamard. From 1898 on, he was ensconced at the University of Paris. Like Poincare, Picard had an abiding intellectual interest in physics, especially mechanics, and in 1922 wrote one of the first textbooks on relativity. Picard’s Treatise, first published in the mid-1890s, helped spread to a wider academic audience both Poincare’s and Lyapunov’s theorizing about the conditions for establishing equilibrium in dynamical systems.
In La science moderne et son état actuel, Picard produced a short treatment, no more than a couple of paragraphs, of the attempt made to apply mathematical analysis to economics, and attached it essentially as an addendum to a section entitled “Analysis in its Relations with Chemistry and Biology.” It was this passage to which Walras was drawn and responded so favorably, causing him to seek out Picard in October 1906 in further support of the effort to mathematize economics (Letter #1633). Walras’ reaction appears somewhat outsized. The next month, as if offering blandishments, he inquired of Poincare as to whether the rumors were true that Poincare and Picard had made the case for a mathematized economics at an international conference at the 1904 World’s Fair in St. Louis, Missouri (Letter #1638). The conference had been organized around the theme of the “unity of science,” and had drawn, in addition to Poincare and Picard, Gaston Darboux as head of the French delegation (Darboux Reference Darboux1912, pp. 371–372). Poincare, though, denied that the matter of a mathematized economics had ever been taken up at the conference (Letter #1639).
In this passage, presented below in its entirety in translation, one is struck by the cautionary notes sounded by Picard in his overview of the field.
Picard begins by stating: “[B]ut, despite legitimate hopes, it is clear that, overall, biology is still far from entering a truly mathematical period.” He then goes on to say: “This is not the case, according to certain economists, for political economy” (Picard 1905, p. 45).
That case then follows:
After Cournot, the Lausanne school made an extremely interesting effort to introduce mathematical analysis into political economy. Under certain hypotheses, which hold for at least limited cases, one finds in the scholarly treatises an equation between the quantities of goods and their prices, which corresponds to the equation for virtual velocities in mechanics: it is the equation of economic equilibrium. A function of quantities plays a role in this theory corresponding essentially to the potential function. Moreover, the most authoritative representatives of the school insist upon the analogy of economic phenomena with mechanical phenomena; ‘as rational mechanics, one of them says, considers material points, pure economics considers homo oeconomicus.’ Naturally, one also discovers the analogues of the equations of Lagrange, the obligatory model for all mechanics (Picard 1905, p. 46).
However, Picard’s assessment concludes in less than glowing terms. “All the while admiring these hardy efforts, one is led to fear that its authors have neglected certain hidden masses, as Helmholtz and Hertz might have put it. But, in the event, there are in these doctrines a keen application of mathematics, which, at least in cases that are well circumscribed, have already rendered service” (Picard 1905, p. 46).
What are those limits to which Picard refers? Or, more precisely, what are their implications for those seeking to mathematize economics? There are a host of masses or movements in the discourse of physics at the end of the nineteenth century that were regarded in the terminology of Helmholtz and Hertz as hidden; these included: the ether, vibrating materials, potential energy, and the dissipative motion association with entropy. Critically, the last of these involved actions or movements that were irreversible, though Picard’s overall concern, like that of Poincare, was with respect to all such actions or movements that could not be comprehended in a conservative force field.
Elsewhere in La science moderne et son état actuel, Picard distinguishes between physical systems that acted upon the principle or postulate of ‘non heredite,’ in which “the future of the system depended solely upon its current state” (Picard 1905, p. 123), and those which were subject to the principle or postulate of heredity, in which the past or history did matter. The former were typically regarded as conservative, while the latter were taken not to be so. Picard saw these latter systems marked by certain ‘frottements,’ or frictions, which he felt were characteristic of systems where the past influenced the future. And it was to these ‘frottements’ that Picard specifically drew Walras’ attention, throwing out a clear-cut caution, in his response to him, restricting “the analogies of the problems of pure political economy” to “the problems of the rational mechanics of systems sans frottements” (Letter #1634, Walras Reference Walras and Jaffe1965, p. 313).
In La science moderne et son état actuel, Picard expatiated as well on the relation between ‘frottements’ and the notion of irreversibility:
The great majority of physicists are of the opinion that irreversible systems cannot be conservative; to discuss this assertion it is necessary to comprehend that the notion of irreversibility is not independent of the number and the nature of the variables envisaged in the system. A mathematical physicist of great penetration, M. Brillouin, thinks that rational mechanics is not essentially reversible, and irreversibility can introduce instability into mechanics (Picard 1905, p. 124).
This carries with it the implication that when irreversible systems are involved, any reliance upon a natural movement toward an economic equilibrium may be mistaken. Attaining such an equilibrium would come to require the use of Lagrange’s equations, through a direct borrowing from mechanics. But, as Poincare had noted in Science and Hypothesis, the existence of ‘irreversible phenomena’ called into question the scope of the applicability of Lagrange’s equations: “It also has always been verified, at least in so far as concerns reversible phenomena which thus obey the equations of Lagrange, that is to say, the most general laws of mechanics” (Poincare Reference Poincare1905, p. 125). The underlying principles here are those of the conservation of energy and least action. But Poincare goes on to say: “Irreversible phenomena are much more rebellious” (Poincare Reference Poincare1905, p. 125). These too can be comprehended by universal principles or laws, but of a somewhat different sort: Carnot’s principle and that of entropy.
For his part, Picard links systems in which history or heredity matters to the phenomena of hysteresis:
The word ‘heredity’ ought not cause one to think necessarily of living beings; it signifies simply here the anterior history of the system under study. The type of these is found in the deformations said to be permanent and in the phenomena called hysteresis. Here again one ought to remember the notice taken immediately of systems with ‘frottements’ (Picard 1905, p. 125).
As Emile Picard’s extended discussion of the significance of irreversible phenomena in mechanics reveals, consideration of the scope, nature, and implications of such phenomena had clearly entered into the discourse of France’s leading mathematical physicists at the turn of the twentieth century. Picard had specifically cited the insights of Marcel Brillouin. The latter was known for his contributions to the development of quantum mechanics, and was among the attendees at the first Solvay conference, which would be held in 1911, an event that is widely regarded as the first international physics conference, and that brought together the most illustrious physicists of that generation.
Henri Poincare, who also attended the first Solvay conference, had challenged the model of energetics, which had gained popularity among physicists in the second half of the nineteenth century, as failing to comprehend the import of irreversibility: “Another objections seems to me still more grave: the principle of least action is applicable to reversible phenomena; but it is not at all satisfactory in so far as irreversible phenomena are concerned; the attempt by Helmholtz to extend it to this kind of phenomena did not succeed and could not succeed” (Poincare Reference Poincare1905, p. 93).
VI. ECONOMIC HYSTERESIS REVISTED
Hence, Rod Cross’ version of the basis for the great dominance of reversibility in the conceptualizing of economic systems must be called into question. Writing in the Journal of Economic Methodology about “Metaphors and Time Reversibility and Irreversibility in Economic Systems,” Cross described a two-step process. First, he cited the “reluctance of physicists . . . to admit that time irreversible phenomena were other than aberrations” (Cross Reference Cross1995, p. 127). Then he noted the “propensity of neoclassical economists to draw their metaphors from what they perceived to be the most prestigious scientific discipline” (Cross Reference Cross1995, p. 128).
Yet, it is clear that by the first decade of the twentieth century, many leading physicists were not all that reluctant to take up the subject of irreversibility as a central matter in mechanics. That being the case, Cross’ second proposition would also need to be modified: neoclassical economists tended to “draw their metaphors from what they perceived to be” physics, which they regarded as “the most prestigious scientific discipline.” Or, in the alternative, it was a case of the tail wagging the dog, as nineteenth-century mathematical physics had produced an abstracted system and a set of mathematical techniques that appeared remarkably amenable to establishing analogous structures in other disciplines.
However, it cannot be denied that the existence and significance of irreversible phenomena had entered the domain of public discourse. Both Poincare’s and Picard’s accounts of this matter were found in their popular writings, as was their use of the term “hysteresis.” The introduction of it in their popular works is noteworthy on two counts: first, it meant that what had been a new technical coinage, as “magnetic hysteresis,” in 1891 (Ewing 1891), had found its way into a more public discourse only a decade later; and second, especially in Picard’s hands, its use had been broadened to any system in which history or memory mattered. For his part, Borel also emphasized the likelihood of irreversibility in physical phenomena. While he did not use the term “hysteresis,” his challenge to the scope of Ludwig Boltzmann’s hypothesis on ergodicity introduced a way of thinking about physical phenomena that was closely allied to it. Jacques Hadamard was yet another contemporary French mathematician who had explored statistical mechanics (Hadamard Reference Hadamard1906).
As for Walras, it is well known that he clung to Poinsot’s Statics as his vade mecum to physics throughout his entire adult life (Turk Reference Turk2006, pp. 200–201). This is remarkable, as Poinsot first wrote his text on statics at the beginning of the nineteenth century, and even the edition that Walras used, Poinsot’s eighth, was published in the early 1840s (Letter #1495), which predated Clausius’ work on the conservation of energy and Helmholtz’s inquiry into the principle of least action. Thus, Walras’ mental construction of physics also predated the exploration of energetics, whose flourishing as a school of thought itself had come to an end by the time of Walras’ exchanges with Poincare.
Walras was aware of both Poincare’s and Picard’s popular writing, and had, in fact, written to Picard precisely because of what he learned Picard had stated about the effort to mathematize economics in La science moderne et son état actuel. Moreover, in a letter to Georges Renard, Walras also claimed, shortly after his exchange of letters with Picard, that he was reading La science moderne along with Poincare’s La valeur de la science (Letter #1641). All courtesies aside, Picard had been explicit in his response to Walras that one might analogize from the rational mechanics of systems ‘sans frottements.’ The text of La science moderne et son état actuel makes plain that such ‘frottements’ or frictions are associated with memory, history, irreversibility, and, thus, hysteresis. While briefer in its depiction, Poincare’s Science and Hypothesis links the irreversible phenomenon of the friction of solids to hysteresis (Poincare Reference Poincare1905, p. 125). Thus, even at the turn of the twentieth century, for anyone who wished to take into account current thinking in physics and attempt to address the role of hysteresis in analogized systems, the requisite materials and direction were fully in evidence. This appears to be especially so in the world of French mathematics and physics.
Moreover, Walras’ exchanges with Poincare and Picard were direct, and both pointed him to the problematic nature of friction for conservative mechanical systems. Picard in particular had judged it to be pertinent to any application of mechanics to economics. This was plainly a moment when the idea of hysteresis might have been taken up in the discourse and thinking of economics, and thus might have come to occupy a formative role in economic thought, but it was not to be so.
Furthermore, the historical moment is all the more striking when one takes into account the fact that many of the same issues were taken up in Emile Borel’s rethinking of the philosophical basis of economic measurement in his 1907 article, “Un paradoxe economique.” Borel was a younger French mathematician with whom Walras had also corresponded, but, it appears, only in his capacity as editor of La Revue du Mois.Footnote 4
In this article Borel turned his attention to the practical reality of the discontinuity of number inherent in economic matters like setting prices or applying taxes and other governmental rates. This extended his initial exposition about the general difficulty involved in quantifying sets of things ontologically, starting with what makes a ‘bundle’ of wheat. (The article’s subtitle was “The Sophism of a Bundle of Wheat.”)
At the same time, Borel focused upon what he called “unstable equilibria” and the potential for seemingly minor and likely unnoticed events to introduce more profound destabilizing effects. As Borel put it, the “breach” of any equilibrium may be set in motion by a “very small action” that nonetheless leads to “important consequences.” He saw the affinity of his notion of the instability of equilibrium with Poincare’s definition of chance (‘hasard’), whereby “small causes produce large effects. The insignificant cause is not perceived, but the effect strikes us and appears to us as fortuitous” (Borel [1907] 1972, p. 2206f).
These mathematicians thereby presaged, by nearly a century, the efforts by Brian Arthur and Paul David to incorporate similar ideas into economic thought (Turk Reference Turk2010, pp. 481–486). Arthur would champion a form of hysteresis propelled forward by the accident of ‘small events’ (Arthur 1989, p. 122). Applying terminology that was first introduced by Boltzmann, both Arthur and David would seek to distinguish ergodic from non-ergodic pathways in economics, the latter of which linked past events to future developments, so that, in David’s words, “history matters” (David Reference David1997, p. 25).
Arthur has traced the scientific origins of his heterodox views in economics to, variously, Jacques Monod on the interplay of chance and necessity, and Ilya Prigogine on the irreversibility of physical systems, but also Henri Poincare on a science of ‘complexity’ drawn from an anti-Cartesian perspective on mechanics: “But there is another way to do science—a growing movement and that is to look from the bottom up at how interactions form structures and patterns. Certainly there are many precedents for that, Poincare’s dynamics for example” (Delorme and Hodgson Reference Delorme, Hodgson, Finch and Orillard2005, p. 20).
VII. HYSTERESIS, EXACT MEASUREMENT, AND DISCONTINUITY
What emerges from the critiques of classical mechanics by both Poincare and Picard is an inclination to give greater weight to the role of history or memory in science, extending, especially as Picard conceived of it, to all forms of scientific inquiry and including, therefore, economics among the disciplines so affected.
In their work there also appears to be a recognition of the discontinuities associated with the irreversible nature of many physical phenomena. This was also true for Borel, who saw a disjuncture between the continuity found in abstract mathematical measurement, with its assumption of unbounded enumeration or division, and the discontinuity and boundedness inherent in any physical measurement. To what extent, then, is hysteresis linked to the discontinuity inherent in such a lack of exactitude?
Those who have limned the history of hysteresis across disciplines have emphasized its roots in the notion of continuity. Writing in the 1970s, Jonathan Elster, who is often cited in this regard (Cross and Allan 1988, pp. 26– 27), traced a broadly conceived version of the idea of hysteresis back to Leibniz (Elster Reference Elster1976). It was Leibniz who asserted that nature does not make leaps or jumps; that is, among natural phenomena, changes assume an evolutionary form or follow an evolutionary path requiring a set of links between past events and experiences and the present state of affairs. The weight of the past then becomes the basis for the continuity implicit in historical experience. Though not cited by Elster, Alfred Marshall invoked the same expression as Leibniz in his support of continuity in economics: “the motto of my Principles . . .is:—Natura non facit saltum: i.e., economic evolution is gradual and continuous on each of its numberless routes” (Marshall Reference Marshall1919, p. v).
But there is a countervailing theme among the critiques of the philosophical foundations of classical mechanics provided by the turn-of-the-century mathematicians.
Poincare, Borel, and Volterra all appear to recognize that discontinuity must be reckoned with, however differently each pursued the matter. Poincare pointed to the abstracted construct of the mathematical continuum, in contrast to the lack of smooth, continuous gradations evident in measures of the physical world.
In a similar vein, Borel saw a substantial and substantive gap or breach between the exactitude of abstract quantification as a mathematical operation and the blurred and less distinctive measurement of the physical world. Accordingly, he turned to measures of probability instead as the sole means whereby the reality of discontinuity in the external world could be addressed and reconciled: “[T]he calculus of probabilities alone permits us to exit from the logical impasse where the reasoning by continuity thrusts us” (Borel [1907] 1972, p. 2203).
By contrast, Volterra accepted the existence of discontinuity in the physical world, but allowed for the need of theorists of physics to create essentially a fiction of continuous change, extending this practice to economists as well: “Now, by analogy to that which the creators of the theory of elasticity have done, and Fourier with regard to heat, the economists suppose that the quantity of goods for which each may dispose, the nature of such would be known as discontinuous to them, vary by continuous degrees” (Volterra Reference Volterra1901, pp. 446–447).
The nomothetic view of economics, so essential to the neoclassical program of economics, appears to embrace continuity as a means of establishing universal patterns transcending time and place. Discontinuity, though, introduces additional nuances. On the one hand, one may envision the possibility that discontinuity alone is insufficient to bar identifying such universal patterns, as long as more disruptive patterns can be comprehended as recurring. On the other, discontinuity may help set in motion the irreversibility and hysteresis that mark many physical and economic processes.
Thus, the turn-of-the-century French mathematicians offered up the basis for embracing alternatives to the neoclassical program in both approach and conceptualization, by calling forth, instead, variously, a larger role for memory, hysteresis, and the inexact in concept and measure, as well as a recognition of the complexities accompanying discontinuity.
