The origin of predictable behavior : Ronald A. Heiner, Department of Economics, Brigham Young University, Provo, Utah 84602, U.S.A.

Abstract

Despite vigorous counterargument by its proponents, optimization theory has been persistently attacked an acceptable explanation of behavior. In one form or another, these attacks repeat the oldest critique of economics; namely, the ability of agents to maximize successfully. Over the years, this critique has taken various forms which include information processing limitations in computing optima from known preference or utility information, unreliable probability information about complex environmental contingencies, and the absence of a well-defined set of alternatives or consequences, especially in an evolving world that may produce situations that never before existed. These complaints are not new to economics. Indeed, they have been present during the very intellectual sifting process that produced neoclassical optimization and general equilibrium theory. Thus, we are to further elaborate this critique of conventional theory, the basic issue is whether there is anything new that is worthy of attention by someone well versed in standard tools and concepts. Are we simply advancing more refined or cleverly argued versions of older critiques, or extensions of them to areas not previously emphasized? Such arguments would still represent an attack on the basic rationality postulate of economics (that agents are able to maximize), but without providing a clear alternative to traditional optimization theory. However plausible these arguments might be, ultimately they must be set aside by someone desiring a theoretical understanding of behavior, unless they lead to another modeling structure whose analytical ability can be explored and compared with existing optimization theory. Another argument focuses on the desire to understand the real dynamic processes that actually generate observed behavior. In contrast, optimization is thought of a surrogate theory based on false assumptions about agents' capacity to maximize. Thus, it can be defended only in terms of empirical testability, without really illuminating the underlying processes determining behavior. Nevertheless, even this view was fully accepted, it is unlikely by itself to cause a major shift away from conventional thinking. The reason is that evolutionary processes have long ago been interpreted one of the key mechanisms tending to produce optimizing behavior; or conversely, optimizing models will predict the behavior patterns that will survive in an evolutionary process tending to select relatively superior performance.' The latter interpretation is in fact one of the dominant justifications for standard models against the criticism of unrealistic assumptions (i.e., the surviving agents of a selection process will behave as if they are able to maximize).2 *Department of Economics, Brigham Young University, Provo, UT 84602. I am indebted to Axel Leijonhufvud for constant encouragement about applications to economics, and for numerous stylistic suggestions. Harold Miller helped familiarize me with a broad range of issues across the sociobiological, psychological, and behavioral science literatures. James Buchanan provided stimulating discussion about conceptual issues. I have also benefited from the advice and criticism of Armen Alchian, Ron Batchelder, Bruce Brown, Robert Clower, Daniel Friedman, Jack Hirshleifer, Kai Jeanski, Randy Johnson, Edward Leamer, Stephen Littlechild, John McCall, James McDonald, Richard Nelson, Gerald O'Driscoll, Dennis Packard, Clayne Pope, Lionello Punzo, Ezio Tarantelli, and Sidney Winter. Needless to say, these colleagues are not responsible for inadequacy in the conceptual framework or scope of ideas presented. 'See in particular Armen Alchian's well-known 1950 paper, and also Sidney Winter, 1964, 1971; Jack Hirshleifer, 1977; Richard Nelson and Winter, 1974. 2A still used reference on the as if point of view is Milton Friedman's 1953 paper. Some recent journal illustrations are Benjamin Klein and Keith Leffler, 1981, p. 634; Richard Posner, 1980, p. 5; Hirshleifer, 1977, p. 50; Nelson, 1981, p. 1059. The ultimate extension of this view is to claim not that agents are able to maximize (select most preferred actions), but rather that any ob-