Experimental Economics: Reply
回应Heiner（1985），强调实验经济学中科学怀疑态度的重要性，指出检验理论时需区分实验设计前提与理论假设，并批评经济学界普遍存在的验证主义倾向。
More than in any particular method of inquiry, I think the hallmark of science is to be found in a constructively skeptical attitude toward knowledge.' The more fundamental are the concepts and assumptions of a science, the easier it is to take them for granted and to abandon this skepticism. In this spirit, Ronald Heiner (1985) is correct in emphasizing that the obtained from the study of the performance of experimental markets is only as secure as the classical preference model used to induce prespecified value structures on the agents in such markets. If the purpose of an experiment is to test a theory (for example, and demand), and the theory is not falsified by the test, this in no way supports any premise of the theory which was also a premise of the experimental design. When we falsify a theory, the implication is that one or more of its assumptions about the behavior of economic agents (maximization of expected utility, commonly shared (homogeneous) expectations, risk aversion, zero subjective costs of transacting, etc.) is in question, and the immediate task is to modify the suspected behavioral assumptions of the original theory. Other assumptions-such as that agents have well-defined preferences, or know the probability distribution from which other agent values were drawn-are not brought into question by the experiment because the experimental design reproduced (or should have) the environment posited by the theory being tested. When testing formal market theories in this way, we should always be aware of the fact that we are studying behavior within the context of our representations of the economic environment. If any of these representations is wrong, then our studies have only increased our self-knowledge, not our knowledge of things (natural economic processes). If we are to increase our knowledge of things, then our ultimate aim should aspire to more than discovering that the behavioral *Department of Economics, University of Arizona, Tucson, AZ 85721. 1 The principal contribution of Popper's falsificationist methodology is, I believe, the influential attempt to develop a formal logic of skeptical inquiry. That the attempt has failed, in the sense that it has produced no defensible codified set of procedures that yield a science of scientific method (happily it would appear that all such attempts will fail), should not detract from the disciplinary value of the falsificationist perspective in approaching scientific questions. Its value to the experimentalist is to force him to ask How can I design an experiment with the property that the set of potentially observable outcomes can be partitioned into those that are consistent with one (or a given) theory and those that are consistent with other theory(ies) (or inconsistent with the given theory)? That experimental life is such that his effort is about as likely to fail as to succeed by no means detracts from the value of the exercise. Its value to the theorist (if he will just forgo the career-advancing primeval incentive to publish yet another technically tractable extension of the existing theory literature) is to force him to ask How can I model this question so as to suggest (as Martin Shubik would say) a do-able experiment, and so as to yield observable implications that do not exhaust the set of possible outcomes? That this effort will often fail does not detract from the value of the exercise. Having said this I would not want to leave the impression that experiments that are fishing expeditions in the laboratory to see what will happen are of no value; seeing what happens can be essential in defining an analyticalempirical research program. Similarly, when a theorist builds (as Buz Brock would say) castles in the air, this is not necessarily useless, for it may lead to more operational forms of theory. We should impute some nonzero probability to the proposition that Feyerabend's anything goes posture is right. But at this stage I think it has become pretty obvious where our professional weaknesses are concentrated. Economists, while spouting the rhetoric (Donald McClosky, 1983) of the falsificationist, are in fact verificationist to the core. We all do it. We take a proposition, conjecture, or theory, then search for supportive historical or empirical examples. As everyone ought to know, seek and ye are likely to find, whether one is a Keynesian or a supply sider. What is not sufficiently appreciated is that this verificationist grubbing is a prescientific exercise in which one asks whether there is any supporting evidence, and how difficult it is to find; if there is none or if it is pretty hard to uncover, it suggests abandonment in the prescientific womb.