Rupture and continuity in the original divide between microdynamics and macrodynamics

Abstract

In 1933, Ragnar Frisch introduced a distinction between microdynamics and macrodynamics. His claim that he proposed the first macrodynamic analysis and that microdynamic schemes were limited to single markets or individual behaviours influenced the subsequent development of economic models. But the division was above all between tools: maximisation at the individual level, and dynamic analysis at the aggregate level. What was at stake was the question of economic rationality and whether it could be extended to an aggregate system. This paper reconstructs the different research programs to give a new picture of the development of early mathematical models in economics.

Keywords:

• Microdynamic
• macrodynamic
• Ragnar Frisch
• Jan Tinbergen
• Charles Roos
• cobweb
• intertemporal optimisation

JEL CODES:

• B21
• B22
• B23
• C50

Acknowledgements

I thank Michaël Assous, Erwin Dekker, Kevin Hoover, Francesco Sergi, Hans-Michael Trautwein and two anonymous referees for their helpful comments on earlier versions of this paper, as well as the participants to the 18th STOREP conference, to the Young Scholar Seminar at the 2021 ESHET conference and to the H2M Seminar at Paris I University, where different versions of this paper were presented.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Archives

Ragnar Frisch’s Archives (RFA)

Notes

1 Hoover (2012, 23) argued that “the issue of the relationship of macroeconomics to microeconomics is simultaneous with the introduction of the terminology,” and this position can be furthered by saying that the issue appeared because of the introduction of the terminology, as it was an important element in its creation. See also Hoover (2012, 22 ff.) on the development of the terms microeconomics and macroeconomics, and the rest of his article for the ways in which different microfounded macroeconomic agenda developed. I am concerned here with what Hoover called the “prehistory” of the microfoundation literature.

2 Andvig (1981) has argued that this “pragmatic” program of Frisch opened the door to “shallow ad hoc model constructions” and contrasted it with the hayekian program of basing business cycle research on economic theory and individual behaviors (Andvig, 1981, 713–714).

3 Andvig (1981, 713) has noted Evans and Roos as precursors to Frisch but did not make the link between their work and microdynamics.

4 Earlier studies that have looked at these different modeling strategies include Dimand (1988), Weintraub (2002, Chapter 2), Dimand and Veloce (2007), Duarte (2009), Pomini and Tusset (2009), Duarte (2016) and Pomini (2018). While I use in the following the results of these works, my main contribution is to emphasize that the different modeling strategies of intertemporal optimization that they defended formed a body of knowledge against which Frisch reacted as a whole in the early 1930s.

5 Dimand (1988, 157) argued similarly that “[t]he strongest reason for adopting this modeling strategy appears to have been mathematical convenience.”

6 While Evans does not cite his sources, there is little doubt that he was referring to Fisher’s analysis that had been presented in 1923 and 1925 and sought to explain business cycles with variations in prices (Fisher 1923, 1925). In those papers, Fisher drew a clear distinction between the price level and its rate of change, and emphasized the importance of the latter. Roos explicitly referred to Fisher (1923) in his 1927 article (Roos 1927a, 634).

7 Roos had a postdoctoral fellowship in Chicago in the late 1920s (Dimand and Veloce 2007, 522). While the economics of his analysis was published in the Journal of Political Economy (Roos 1927a), his more mathematical developments were published in the Proceedings of the National Academy of Sciences (Roos 1927b).

8 See Duarte (2016) on the different approaches to time-discounting in interwar economic models and how this assumption became necessary in postwar models.

9 That utility maximization at the aggregate level was not accepted in the interwar has been underlined elsewhere: “Ramsey's model combines mathematical formalism, utility-maximizing agents, and aggregate-level control with clear normative content. … the latter two aspects were considered to be mutually exclusive prior to the 1950s, and such incompatibility is also behind the delayed recognition of Ramsey's (1928) analysis among economists” (Duarte 2009, 164). According to Pomini (2018, 68), the Italian school of economists working on intertemporal optimization that survived until the late 1930s was also concerned with the maximization of utility.

10 “In spite of these researches [on cyclical fluctuations], no successful attempt to relate the theory of business cycles with economic theory has been made” (Roos 1930, 501).

11 I thank Erwin Dekker for pointing out to me that cobweb studies were in-between the micro analysis of individual behavior and the macro analysis of the economy as a whole, in the meso level of market structures. Dimand (1988) has noted the connection between approaches based on empirical cobwebs or the calculus of variation, a connection which is also discussed by Morgan (1990, Chapter 5) who linked the work of Roos and Evans with that of Moore although they were based on different formalisms (1990, 152–155).

12 Ezekiel (1938) discussed the development of the cobweb, and his paper became the primary source of many subsequent studies. Lendjel (1998) based his account of early cobweb developments on a comprehensive view of the original texts. Moore’s difficulty in building a dynamic approach and his poor reception by his contemporaries, except his most devoted student, Henry Schultz, was described in Epstein (1987, Chapter 1), who also gave an account of Tinbergen (1930).

13 He added Moore’s (1929) approach as a third one, which he called “comparative statics.”

14 Several studies have already established Tinbergen as one of the main economists of the interwar period, in particular Morgan’s previously cited study (1990, Chapter 4), which centered on his macroeconometric program of the second part of the 1930s. Boumans (1993) examined his doctoral work and some of his early models in (Boumans and Boumans 2013). Dekker (2021) recently published a comprehensive biography of his intellectual trajectory from the point of view of cultural history.

15 On Frisch’s use of mechanical analogies and physical concepts, see Dupont-Kieffer (2003), in particular the last part of her doctoral thesis (Chapter 5, 283 ff.). See also Andvig (1981) on Frisch’s work during the interwar, and the models he developed before 1933, among which was a 38 equations disaggregated model that Andvig sees as important to steer Frisch away from microdynamics. Andvig (1988) makes the case that in the second part of the 1930s, Frisch moved away from macrodynamics, toward macroeconomic decision models.

16 This paper was partly translated in English in Frisch (1992), but the microdynamic model was left out, and only the first four sections which developed a general view of dynamics were present; this attests the relative ignorance in which microdynamic models, which were dominant in 1929, had fallen after Frisch’s macrodynamic approach was developed and spread.

17 Wade Hands (2012) argued that Walrasian models and certain Keynesian models (principally IS-LM) were characterized by an absence of path-dependencies and irreversibilities, while “such path-dependencies and irreversibilities were common features of many of the demand theories the Walrasian program was competing against during the 1930s and 1940s” (2012, 109). It is precisely this program of path-dependencies that characterized the macrodynamic research program, as shown below with the work of Tinbergen. A more detailed examination of the protagonists of this program, and of the importance of the stability and instability issues that shaped it, can be found in Assous and Carret (2022a).

18 On the role of this debate see Dupont-Kieffer (2012) and Assous and Carret (2022a, Chapter 4, section 2).

19 In the report, Marschak added that “[t]his point had been emphasized by Professor Fréchet in a recent discussion at the Institut Henri Poincaré” (Marschak 1934, 192), a clear reference to the Poincaré lectures, where Fréchet’s presence is attested by his correspondence with Frisch (see in particular Fréchet to Frisch, 16 February 1933, and Frisch to Fréchet, 13 March 1933, RFA, where they discussed the organization of the conferences).

20 The problem of stochastic optimization became a topic of intense research after the war but from the point of view of the 1930s economists’ methodology based on the classical calculus of variations this was an intractable problem.

21 Louçã (2007) has given a thorough history of the Econometric Society centered around Frisch’s role, although he did not relate it to what had been done previously in dynamic analysis as is done here.

22 “The grouping of the elements, which has its statistical counterpart in the calculation of index numbers of all sorts, is characteristic of what Frisch calls macrodynamic economics in contrast with microdynamic economics” (Tinbergen 1935, 243, original emphasis).

23 Hoover (2004) argued that “causal language flourishes when empirical economists start with data, but withers when they start with mathematical theory” (2004, 161), and that the rise of Walrasian general equilibrium models after the war were the dominant factor in the decline of causal analysis in the postwar period, which is consistent with the story developed here, where the causal analysis characteristic of the macrodynamic program was developed in rejection of a teleological intertemporal optimisation.

24 See Assous and Carret (2020, 2022b) on these models and others developed by Tinbergen, which were elaborated in particular to discuss the influence of economic policies and to interpret the long depression of the early 1930s.