“Decisions and Dynamics”: An Ad Hoc Exploration of Intellectual and Institutional History January 3, 2015Posted by Will Thomas in Commentary Track, History of Economic Thought.
Tags: Ben Bernanke, Charles Hitch, E. Roy Weintraub, Herbert Scarf, John von Neumann, Judy Klein, Kenneth Arrow, Leon Walras, Paul Samuelson, Perry Mehrling, Philip Mirowski, Robert Dorfman, Robert Solow, Roger Backhouse, Till Düppe, Verena Halsmayer, Wade Hands
This post offers some background information on my new paper, “Decisions and Dynamics: Postwar Theoretical Problems and the MIT Style of Economics.”
The 2013 “MIT and the Transformation of American Economics” conference was one of those conferences where the invitation arrives years before the event. When I agreed to attend, I thought I would just offer a complement to the conference’s focus on MIT economics with a discussion of the early history of operations research at MIT, a subject I already knew a lot about.
What I didn’t realize until the run-up to the conference was that it was part of the annual series, which results in the publication of the annual supplementary volume to the journal History of Political Economy. I had already published pretty much all my material on MIT in my 2009 Science in Context article on Jay Forrester’s industrial dynamics, and in my 2012 Business History Review article on OR at MIT and Arthur D. Little. So, that scuppered my plans for an easy recycling job.
The obvious direction in which to go was to discuss the intellectual relationship between economics and operations research. The problem with this plan was that, while there are many interesting things to say about that relationship, the relationship at MIT was pretty thin throughout the 1950s, the period I’ve studied carefully. At this point I didn’t have the time to try and suss out any subsequent relationship (if it was even substantial) through the publications record, nor, being based in London at the time, did I have much chance to do new archival research.
What did seem like a good opportunity was to engage with the thinking of the only historian to publish on the relationship between OR and economics, Phil Mirowski.
In Machine Dreams (2002), and building on work he had done with Wade Hands,* Mirowski posited that a growing orthodoxy in American postwar economics could be divided into three branches: the “Chicago doctrine, the Cowles approach, and the MIT style” (191). He further posited that the differences between these branches could be interpreted in terms of their respective relationships with OR as it emerged from World War II.
I give Mirowski a great deal of credit for being the only person to really notice that the OR-economics relationship is historically important. However, he seems to understand military OR not as something in its own right, but as a thing that acted upon economics. Thus he draws a dubious distinction between an empirical, modestly statistical “British” version of OR (which he identifies with the Chicago school of economics) and an imperiously theoretical “American” version of OR (which he identifies with the RAND Corporation and the Cowles Commission for Research in Economics).
I could discuss Mirowski’s argument on this point at some length, but, for our present purposes, it is unnecessary to go into detail. Mirowski makes clear that, in his view, of the progenitors of the three branches of postwar American economics he identifies, “[MIT’s Paul] Samuelson [1915–2009] appears least influenced by any of the signal innovations of OR, be it of the outdated British variety or the cyborg heritage of [John] von Neumann” (226).
The key trouble with this take on MIT is that the institute’s most influential economists were, in fact, not divorced from OR at all. As Mirowski himself relates, beginning in the late 1940s, Samuelson began working on a project suggested by Charles Hitch (1919–1995), the head of the new RAND Corporation’s economics department, to illuminate the relevance of the new technique of linear programming to economics. This project continued for most of the 1950s, enveloping his junior colleague Robert Solow (b. 1924) and the Harvard economist and mathematical statistician Robert Dorfman (1916–2002). At the same time, linear programming was becoming a central plank of a new body of OR theory. Linear Programming and Economic Analysis was finally published in 1958.
It is arguable that MIT was much more influenced by OR than Chicago, at least, ever was. The reason, then, why Mirowski does not suppose OR to have been influential at MIT is because it was not influential in the right way (226):
Incongruously, Samuelson managed to coauthor a commissioned text for RAND on the promise of linear programming with Robert Dorfman, an important figure in American OR, and his protégé Robert Solow, a text that strains to deny any linkages to game theory and the computer, and conjures an imaginary lineage where linear programming merely “resolv[es] the problems of general equilibrium left unsolved by [19th-century French economic philosopher Léon] Walras”… Unlike Chicago, which simply ignored von Neumann and the cyborgs, and Cowles, which struggled mightily to assimilate them to the neoclassical organon, it is hard to regard MIT as doing anything other than actively trying to purge economics of any residual trace of cyborgs, which it knew was already there.
Mirowski takes the absorption of the techniques of “American OR” (allegedly, the prototypical “cyborg science”) into economics to be a kind of natural, military-funded vector for the development of models of economic markets as Walrasian calculating engines—the target of Mirowski’s book. Because MIT (rather vocally) did not favor the development of such models, the interaction that did occur between linear programming and economics at MIT must be (rather grudgingly) explained away as a puzzling irony (231):
MIT tried to have it both ways: there was nothing really new under the sun, but its epigones were nonetheless busy trying to co-opt the latest mathematical techniques to demonstrate that they were the most au courant of scientists. Not the least of the ironies besetting their project was that they were housed in one of the biggest nests of cyborgs to be found on the planet—the MIT of Whirlwind, Lincoln Labs, Bolt Beranek & Newman, and the Media Lab.
As I argued at Duke, the situation can be much more readily understood if we assume that Mirowski’s conceptualization of OR, and of its relationship to economics, simply makes no sense.
Definitively, this is not to say that the relationship between economics and practical mathematical techniques—some of it associated with OR, some of it not, or only later—is unimportant. Quite the contrary. Rather, if we want to understand this relationship, we would do best just to listen to Judy Klein, who traces the key line of influence to… the Carnegie Institute of Technology (notably absent from Mirowski’s institutional troika). Unfortunately, most of her work on the subject remains unpublished, but here, in any event, is a video of her speaking about it with historian of economics Perry Mehrling:
There is still quite a lot to be worked out about which economists have gravitated to what sorts of models, and why. But, as I mentioned in my previous post on the “MIT” conference, there actually wasn’t all that much interest in the intermixed virtues and vices of Mirowski’s historical model.
(However, before moving on, one should point out that it is precisely because Mirowski minimizes the links between OR and MIT economics that, contra Roy Weintraub’s introduction to the new HOPE volume, the relatively weak relationship my published paper portrays between OR at MIT and economics at MIT can’t really have much to say about the validity of Mirowski’s overarching argument about OR and economics.)
In retooling this paper for publication, I actually ended up taking quite a bit away from the conference. First and foremost, there was the insight that it should be about the “MIT style” of economics. But I think it also benefited from a stronger understanding of the uses to which Samuelson and Solow put linear programming methods, which I gleaned from the papers of Roger Backhouse and Verena Halsmayer (as well as from a helpful conversation I had with Till Düppe in Athens in 2012). Backhouse also provided me with some unpublished items he had written on Samuelson’s history with linear programming, which was quite valuable, among other things, confirming an affinity in Samuelson’s and Hitch’s views of linear programming.
My final appreciation of my paper balances two points: 1) although the published paper is much more coherent than the talk it is based on, it is—full confession—still unquestionably an ad hoc pasting together of a number of things I happened to have had on hand; but 2) the “pertinent observations” I offer in the paper actually do open up some very specific points for research, which I haven’t really heard addressed elsewhere, and which I think are well worth doing more work on. Among these:
- The need to attend to how decision making is treated in decision theory and economic modeling, prior to (and, frankly, probably also after) the institution of any clearly defined intellectual program of “microfoundations.”
- How did modelers treat “dynamic” effects in economic models, e.g., how were adaptive behaviors regarded as rationally anticipatory, and how were delays and time limitations regarded as giving rise to unintended pathological phenomena, and how were or were not these (equally plausible) phenomena reconciled?
- The need to understand “rationality” assumptions as more closely associated with a preferred heuristics of modeling, rather than with presumptions of actors’ actual rationality, i.e. should, in any given case, model refinement be regarded as a process of making idealized models more realistic, or, just as kosher to my mind, should it be regarded as including additional rational behaviors in otherwise overly rudimentary models?
- More specifically, there appears to have been a widespread interest in inventory problems in the postwar period. Inventories were regarded as playing a crucial role in business cycle phenomena. They were also of strong interest to key economists, including Kenneth Arrow (b. 1921) and Herbert Scarf (b. 1930), developed decision theoretical models of optimal inventory policies. What sustained these figures’ interest in this problem?
Finally, I sketch out a model of how very distinct programs in OR and economics at MIT only gradually grew together, particularly through the mediation of MIT’s School of Industrial Management, and drawing on disciplinary alignments pioneered elsewhere, particularly the Carnegie Institute of Technology.
However, a lot of work could still be done here. The received wisdom I have heard indicates that OR and economics have traditionally not actually been very close to each other. Yet, only the other day, on a totally separate project, I was reading a 1980/1983 paper by MIT PhD Ben Bernanke—Ben Bernanke!—that treats investment behavior using a dynamic programming model. So, yes, it would be good to know the history of all this.
At the moment, though, I think we’re still at the stage of mere observation, where we can’t even formulate coherent historical theses concerning these issues, at MIT or elsewhere. Here’s a not-particularly-helpful video of Kenneth Arrow (who was never at MIT) recently discussing his own experiences in OR and economics:
Really, nobody is capable of articulating this stuff very well yet. Which, if you ask me, is pretty exciting.
*D. Wade Hands and Philip Mirowski, “Harold Hotelling and the Neoclassical Dream,” in Economics and Methodology: Crossing Boundaries, edited by Backhouse, Hausman, Maki, and Salanti (London: Macmillan, 1998).
D. Wade Hands and Philip Mirowski, “A Paradox of Budgets,” in The Transformation of American Economics, edited by Mary Morgan and Malcolm Rutherford (Durham: Duke University Press, 1999). (link)