Experimental asset markets are multiple-period laboratory double-auction markets utilizing human subjects who trade asset units with fundamental values determined by well-defined (perhaps stochastic) dividend streams. Traders’ monetary payoffs are typically tied to individual performance, e.g. traders attempt to maximize earnings in the form of per share dividend payments and capital gains. The seminal work on double auction design was due to Smith (1962). Methodological precepts, which govern virtually all current experimental asset market designs, originated with Smith (1982).
Essentially, experimental asset markets were developed to investigate and test various hypotheses which followed from the theory of efficient markets. In particular, attention has centered on various predictions concerning market efficiency in the presence of rational expectations. These include tests for the existence of both weak-form and strong-form efficiency (i.e. asset prices reflect all public and private information respectively), along with the ability of the market to both disseminate and aggregate diverse private information, as well as the study of individual expectation formation. As such, a fundamental cornerstone of the research investigates the diffusion of information in the market in the presence of trader uncertainty.
In general, the multiple period setting of ass et markets presents several sources of trader uncertainty. Uncertainty may derive from diverse expectations among traders concerning the movement of future prices conditional on a distribution governing states of the world such as dividend payout. Uncertainty may also present itself in the form of private information concerning trader type (e.g. assets have differing valuations depending on trader-type endowments). Otherwise, uncertainty emanate s from individual differences in home-grown expectations governing the expectation formation process and uncertainty regarding the future movements of prices. It is the assumptions regarding the formation of expectations that discriminate between competing models of asset valuation. As such, the object of investigation is to observe individual decision making in an environment in which uncertainty generated by the diversity of states and trader types is the experimental control.
Early laboratory studies examined whether ass et markets were informationally efficient, and presented results which indicated that market efficiency is generally robust to information asymmetries. Controlling for trader type uncertainty, Forsythe et al. (1982) present evidence indicating convergence to strong-form efficiency. Essentially, the design consisted of repeated two-period asset markets with trader type uncertainty, i.e. share value was private information and differed in each period for each trader type. Two types of equilibria are possible: a naive equilibrium which results when traders value assets based solely on their private information regarding dividend values and a full information, and rational equilibrium which results from the dissemination of otherwise private information into the market. It is the rational equilibrium that the market converged to in repeated sessions and, hence, generated support for strong-form market efficiency. Plott and Sunder (1982) examined essentially the same issue but in the presence of state uncertainty in which dividend payout followed a probability distribution and certain traders were given more information than others regarding payout states. The experimental evidence showed that the market reveals insider information; market prices converge quickly to a fully revealing rational expectations equilibrium.
Extensions followed. Forsythe and Lundholm ( 1990) examine information aggregation rather than information dissemination in a series of experiments looking at whether markets are capable of efficiently aggregating a highly diverse, but sufficient, body of information. The issue here, as above, is whether traders can form inferences about market fundamentals through an examination of publicly available information on bids, asks, and contracts. They conclude that, in the presence of diverse private information, trading experience and complete information are jointly sufficient to generate a rational equilibrium. Copeland and Friedman (1987, 1991) extend the analysis in an examination of the sequential revelation of private information to uninformed traders.
Other asset market experiments differ markedly in their designs and investigative intent. Smith et al. (1988) utilize a multiperiod finite horizon model with state uncertainty regarding dividend payout, but with no private information beyond individual endowments, to examine bubble behavior (e.g. market prices that deviate from fundamentals). Traders’ one-period ahead forecasts of market prices were simultaneously solicited to test various theories of expectation formation. Market prices were observed repeatedly to exhibit bubble-crash behavior and expectation formation was best characterized as adaptive (not rational) in character. What caused these bubbles is not clear. Speculative behavior is theoretically impossible in finite horizon experiments. Other explanations, however, suggest these bubbles occur due to incomplete learning. Subsequent experiments did indicate that expectations do converge to rational expectations as traders gain experience, and prices tended to vary little relative to fundamentals.
Other experiments have examined the efficiency of dividend signals by measuring the noise content of the signal but still leave the dividend puzzle an unresolved issue. Still others have altered the finite dimensionality of the design to test for the presence of speculative behavior. These designs essentially involve a probabilistic horizon in which subsequent trading periods occur conditional on the outcome of a random draw. In such cases, the (known) probability of continuing serves the same function as the discount rate in conventional asset valuation mathematics. These experiments show definite evidence of speculative bubbles which occur despite the absence of private information. As such, the concept of market efficiency is not yet a resolved issue. In addition, investigator s have designed asset market experiments to examine the issue of form versus substance, that is, whether traders prefer one asset over another because form matters even though both assets are substantively identical in terms of fundamental value. The evidence at this point is inconclusive but the research is important nevertheless, because should form matter, then asset values may be more than merely functions of discounted dividend streams.
Experimental asset markets are an invaluable research resource since they permit the investigator to extract a sufficient level of institutional detail necessary to an examination of the research objective and abstract away unnecessary sources of noise. At the same time, the experimental designs allow one to exercise the necessary level of control through a judicious choice of structural designs and parameterizations in order to collect data necessary for strong statistical tests of underlying hypotheses.
