Friday, September 14, 2012

Path Dependency and Hysteresis

I promised to discuss the difference between these two concepts a while ago. The idea of path dependency is related to Joan Robinson’s famous objection according to which equilibrium is not an actual outcome of real economic processes, and it is for that reason an inadequate tool for analyzing accumulation.  Her view would suggest that path dependency should be seen as a property of models that break with conventional methodological stances, and, in particular, with the dominant neoclassical school.

It is important to note that mainstream defenders of the idea that ‘history matters’, like Paul David (of QWERTY fame), tend to disagree with the view that path dependency implies a rupture with neoclassical economics. David (2001, p. 22) says, in this regard: “imagine … my utter surprise to find this approach being attacked as a rival paradigm of economic analysis, whose only relevance consisted in the degree to which it could be held to represent a direct rejection of the normative, laissez-faire message of neoclassical economics!”  For David, path dependency is a property of dynamic and stochastic processes and cannot be used to assert anything about models and propositions derived in static and deterministic setting [which is, apparently, what he thinks neoclassical economics is all about].

Mark Setterfield's research might hold the key to this issue, by differentiating hysteresis [a concept from physics, that shows that mainstream economists do have physics envy!] and path dependency, and suggesting that the former, more typical of mainstream models, is a special case of the latter, more general and the concept often linked to heterodox models. He suggests that hysteresis is a variation of traditional equilibrium analysis, which implies that some displacements from equilibrium would be self-correcting while others would not. Hysteresis results from the non-uniqueness of equilibrium and under certain conditions the economic system would adjust to a new equilibrium. On the other hand, Setterfield argues that the typical path dependent model is based on cumulative causation, a concept that harks back to Gunnar Myrdal and Nicholas Kaldor’s contributions to economics. In this case, transitory shocks always have permanent effects.

A simple example might illustrate the difference between the more restricted notion of hysteresis and cumulative causation. In the conventional mainstream description of labor markets, an increase in unemployment insurance that allows workers to hold out longer for better paid jobs increases the natural rate of unemployment. After a fall in demand (an external shock), if structural changes to the labor market like higher benefits take place, the level of unemployment will increase and eventually fall, as real wages fall, but to the new and higher natural rate [think of Gordon's Time Varying NAIRU]. Hysteresis implies that history matters, but the system is still self-adjusting.

The quintessential example of cumulative causation is associated to the Kaldor-Verdoorn Law, which says that output growth leads to rising labor productivity. Thus, higher demand leads to higher output growth, which implies higher productivity, lower costs, and higher income in a virtuous circle of expansion. There are several possible expansion paths, depending on the strength of the multiplier-accelerator forces and the Kaldor-Verdoorn coefficient, rather than a single equilibrium to which the system adjusts. There is no adjustment to an optimal equilibrium level, no natural rate fixed or varying. The heterodox notion demands the rejection of the natural rate.

2 comments:

  1. This is interesting.

    It seems to me that in practice, hysteresis is used almost exclusively to refer to labor supply, while path dependence is used much more broadly.

    It also seems to me that the basic idea behind hysteresis is that, in general, each set of fundamentals produces a unique equilibrium but that in certain conditions a period of disequilibrium can change some of the fundamentals. The paradigmatic case is the decay of human capital in a period of excess unemployment.

    Whereas path dependence is often used to suggest that it is not possible to specify a set of exogenous fundamentals such that there is a unique stable equilibrium at all. (I guess that's more or less the same as what you said.)

    ReplyDelete
  2. I disagree that path dependence is a non-orthodox method or view of the world, and so I think parsing path dependence and hysteresis is a red herring in the debate about the usefulness path dependence. There is nothing unorthodox in arguing that your economic explanation (be it preferences, technology, geography) for the origin of some institution or idea explains why it's still around today. Even if it admits of institutional inefficiency, there is still the implication of some "efficiency". And by the way, who cares about typewriters anyway? There are bigger fish to fry for heterodox economists and I doubt those fish can be fried with a simple application of path dependence.

    Path dependence might come with the saying "history matters", but it's a very crude understanding of history that path dependence reinforces. Historians such as Marc Bloch have warned historians to "beware of false idols", i.e., attributing modern phenomena to something from the very distant past. I've been reading Boldizzoni's "The Poverty of Clio", a critique of modern economic history, and he mentions a lot of weaknesses in this approach.

    I think the relevance of path dependence to the "orthodoxy or heterodoxy" debates, from David's article onward, is the deeper issue of the reliance on mathematical models/methods/metaphors to talk about economic (or historical-economic) phenomena. That is to say, why do we still wish to frame history in terms of mathematical formulae? History is all about unearthing detail and telling a rich narrative while emphasizing important themes like contingency and power and social relations. Path dependence ignores all that in favor of equilibrium concepts and determinism (at least on some level).

    ReplyDelete