Virtually no economist would maintain that learning processes are unimportant for economics – especially if economics is understood as a genuinely social science that studies human behavior more generally. Indeed, it is easy to see that some sort of learning lies at the heart of the economic approach.
As discussed more extensively next, traditional economics, in its most general methodological characteristic, is the analysis of changes in behavior as a response to changing restrictions or relative prices. This basic methodological concept – in conjunction with the rationality assumption – provides for the predictive power and the relative success of the rational choice approach as compared to other disciplines or approaches in social sciences. Hence, the need for some notion of the adjustment of behavior inevitably follows from this basic concept.
Standard Approach
Traditionally, economics has assumed a perfect adjustment mechanism that allows the analysis to concentrate on the steady states of some unspecified underlying process:
“Economics has tended to focus on situations in which the agent can be expected to ‘know’ or to have learned the consequences of different actions so that his observed choices reveal stable features of his underlying preferences. We use economic theory to calculate how certain variations in the situation are predicted to affect behavior, but theses calculations obviously do not reflect or usefully model the adaptive process by which subjects have themselves arrived at the decision rules they use. Technically, I think of economics as studying decision rules that are steady states of some adaptive process, decision rules that are found to work over a range of situations and hence are no longer revised appreciably as more experience accumulates.” [LUCAS (1987, 218)].
As SIMON (1978, 10) has pointed out, the interest was – and mainly still is – not in the adjustment process itself, but in its outcome, i.e., in which decisions are made by individuals and not how decisions are made.[6] Hence, in most standard equilibrium models learning is reduced to some sort of passive “adjustment” or “adaptation” in that individuals are thought to automatically react to changing circumstances (i.e., to changes in restrictions or relative prices) in otherwise stable environments.[7]
This reduction (in the spirit of Lucas’ above statement) implies strong assumptions about learning – in that a perfect and complete learning process is implicitly presumed – that could only be defended in an environment that provides perfect conditions for learning. Despite the fact that this kind of reasoning not only prevails in microeconomic theory, but extends to macroeconomics (e.g., to models that involve the rational expectations hypothesis which implicitly assumes a perfect and complete learning process) in the standard literature virtually no attempt is made to justify these assumptions that are so crucial for the functioning of economic theory.
Rather, the strong learning assumptions are commonly defended in the same vein as other strong assumptions in economic theory (e.g., the rationality assumption) – that is, by the traditional “as if” argument à la Friedman, or by similar arguments like “the market leads to perfect learning”.[8] Nevertheless, the critique from various sides continues. Aside from other forceful counter arguments – concerning the impossibility to justify the rationality assumption as a result of learning,[9] and the inability of neoclassical theory to genuinely account for evolution or development[10] – an important point of criticism that has been put forward (e.g., by WINTER, 1987) is that a theory that gives no rationale for its underlying strong assumptions provides no internal criteria for its applicability.
That is, neoclassical theory assumes perfect and complete learning but “itself provides no indication as to how long it takes for adaptive processes to reach something like steady-state conditions, it provides no guidance regarding the quality of the predictions that...may be expected to provide in particular cases.”[11] Furthermore, although the limits of the underlying assumptions may generally be acknowledged, those who defend these assumptions along the lines mentioned above “do little to help define [these limits] and nothing to explore the important phenomena that lie beyond them.”[12]
Generally speaking, standard economics has treated human learning as a black box process of perfect adaptation and has not attempted to explore either the conditions under which this may be justified, or the limits that are implied by the learning assumptions.
6 Note that Lucas had admitted earlier that hypothesizing about the quantitative implications of economic policies “involves imagining how agents will behave in a situation which has never been observed”, and therefore “one must have some understanding of the way agents’ decisions have been made in the past and some method of determining how these decisions would be altered by the hypothetical change in policy.” [LUCAS, 1981, 180] This statement clearly calls for the study of learning processes.
7 WITT (1992, p. 4):"In the [neoclassical paradigm] change is always interpreted as reactive. Individual agents or economic aggregates are viewed as responding to events that affect the basis of decision making. Economic actors are portrayed as attempting to adapt optimally to new conditions imposed on them. They are not credited in any way with creating the new conditions themselves. The reason for this narrow interpretation is the very core of the neoclassical paradigm, the synthesis of optimization and the equilibrium concept. Used together, the two ingredients rule out any explanation of individual behavior other than that of adapting to changing circumstances."
8 A long list of such defensive arguments with respect to the rationality assumption can be found in CONLISK (1996, pp. 683) who presents learning as an argument in favor of rationality but criticizes it since it holds only under very perfect conditions. Hence, it is the weakness of this argument and theories that make strong assumptions about learning that these conditions are not identified and the limits imposed by them are not analyzed. See also THALER (1991, p. 158).
9 See ARROW (1987, 201).
10 See NORTH (1994).
11 WINTER (1987, 248).
12 WINTER op. cit.
Prof. Tilman Slembeck
Next: Approaches to Economic Learning
Summary: Index