Hysteresis and path dependence
Hysteresis and path dependence
The notion of path dependence, which I will define precisely in the third part of the thesis, has a predecessor in economics. This is the notion of hysteresis, coming from physics and chemistry, and adopted by new-Keynesian and post-Keynesian economists to model unemployment dynamics or, more recently, financial markets and currency fluctuations.
But the concept has been fading away in the mainstream literature and now the notion of path dependence has since then superseded it in many scopes of research, and much beyond the economics of innovation. Even though, I will treat hysteresis briefly for interest it gives to taking account of history.
Path dependence is a broader concept that can be traced back to economic history and the economics of technological change (David 1975, Flichy 1995, David 2001). And both refer to a more general class of phenomena: historical processes that are not often possible to be couched in mathematical terms, at least in traditional calculus.
Here, like evolutionary concepts, path dependence has also become a fad, but much more than a fad, it is a change in perspective for some economists and economic historians. To be fair, much of the historical concept is coming from innovation studies and the history of techniques. It is implicit in the works of Maurice Daumas (1996/1962) and Bertrand Gille (1978) as I have introduce them in the introduction. What is new is the stochastic and mathematical treatment of path dependence that gives “history matter” a more specific and rigorous meaning. The notion is not free of controversies, as I will show it in the third part (see the introduction and chapter IX and X).
Before, and also, later, simultaneously with the emergence of the concept of path dependence in the 1980s (David 1985, Arthur 1989, 1994), hysteresis consisted of a formal recognition that past events had a bearing on current phenomena. One aspect must be stressed here, while stating that history matters, “economists often subscribe to an approach that does not exploit hysteresis in a manner that reveals the full extent of its potential significance and the full impact of history on the present” (Katzner 1999: 171).
Hysteresis is very often represented by a periodic equation such as:
y t = B y t – 1 + a (1)
and by setting y t = y t – 1 = … = ¢ , the stationary state of equilibrium is:
¢ = a/1 – B, (2)
which does not depend on time t.
And this is not very appealing because the variable yt converges to a history free value ¢ (Katzner 1999: 173).
A more interesting case is one that makes the second element from the right hand side depends on t like that:
yt = B y t – 1 + et (3)
But this equation does not yield, in the general case (et¹0), a stationary state:
¢ = et/ 1 – a (4)
where et brings irreversibility for the present and future value of yt (Katzner 1999: 173).
There is, however, still one more general case that can be formulated in the following way:
yt = f t (y t – 1 ; et), (5)
where f t is determined at time t by interactions that change with t. This third conception of hysteresis is one that is compatible with (1) radical uncertainty, as defined in the first part of the thesis; (2) path dependence, in the sense given by Paul David; and (3) historical economics or the explanations of historical processes, with one caution: historical time is a combinatorial intertwining of various variables that have their own partial autonomy and sphere of influence with no precise borders.
Accordingly, it is clear that the case of multiple causal factors; the feedback mechanisms from the very precise process itself (y t and its partial determination by past values); the causal interaction with and from other variables, etc.; turn the analytical and quantitative handling difficult with very few cases of data available for the relevant factors behind an historical process under study. This is one reason for the success of the simulation methods in evolutionary economics, but this does not diminish the relevance of careful empirical studies.
I consider this distinction between historical processes and mere mechanical reproduction mechanisms (equation 1) as fundamental which argues against the use simple methods that are based on a simple reproduction (equation 1) or a sophisticated model that “only allows parameters to vary exogenously through time” (Katzner 1993:332-33, 1999: 176).
Historical processes are tackled through different classes of tools: one is hysteresis, another, philosophically related to this first, is path dependence, and, finally, narratives are the third.
Depending on how we construe hysteresis, we can consider that it integrates path dependence, but even here the inclusion is not perfect. Path dependence, in the versions of either Paul David or Brian Arthur), stresses the particularity of the stochastic processes (Arthur 1994, David 2001). Nonetheless, differences are also at other such as the origin: path dependence comes from historical research and is applied to technological change while hysteresis is coming from physics and chemistry and has been applied above all on unemployment dynamics.
The third class of tools to analyze historical processes consists of narratives that include path dependence and hysteresis, not just as a previous discussion but also as techniques of study. Usually, narratives are more “appreciative”, to use an expression of Dick Nelson, than theoretically grounded. Most of my research will follow the path of narrative exposition and by this I means not only the literary device of telling (hi)stories but also quantitative and qualitative techniques that fit into the overall explanation.
Historical time and the quest for abstraction in economics (continued from methodos XIII and XX)
Let me define here the notion of theoretical history. If a lot of ink has been dedicated to the philosophy of history, a necessary question with an arguable answer, much less has been dedicated to theoretical work about historical processes. I adopt here a position –a philosophy of history or a theory of history, as it is usually referred to by historians– in line with the works of Fontana (1992, 1999a, 1999b) and Rodriguez de las Heras (1976) and the method and the theory of complexity. In other words, I integrate a more general conception of history that allows for greater richness in understanding reality (see the introduction and the first chapter for more details about my stance xxx).
From a rather methodological point of view, I affirm that (1) a theoretical history is possible (and even necessary, see introduction); it is not a contradiction with the critique of sources in historiography and the high value scholarship of some historians; and (3) it is a “grounded” theory purported to explain what the historical processes are and how things change or remain the same. By grounded, I mean, without adopting any empiricist dogma, recognition of the complexity of reality and the necessity of empirical work for the very theoretical work. Theories do or do not gain legitimacy, not just on their own, but in relation to the reality they try to explain. At some stage of the inquiry, the theory has to be confronted to empirical cases.
 Most of the literature about unemployment dynamics does not interest us here. See the next section on hysteresis.