Experimental Economists Economics

Experimental Economists and Simulation, Experimental Economics

Recently, a group of economists has begun to undertake a different approach to empirical work in economics. Using animals or people to act as buyers and sellers of an unnamed commodity, and knowing the underlying supply and demand conditions, they determine whether the theory correctly predicts the results that occur in an "experiment. These experimental economists claim to have proved various economic propositions through their experiments.

Let us consider a test they did using a procedure called a "double oral auction market," in which buyers and sellers publicly announce bid and offering prices. Vernon Smith, a leader and developer of much of this work, conducted a laboratory experiment in 1956 to test whether equilibrium would be achieved in a double oral auction market. Students took roles as suppliers and demanders and called out their prices. Within fifteen minutes, with a market of fourteen students on each side, the price came very close to the equilibrium price; once it arrived there, it tended to stay there. When demand shifted (when students were given sheets of paper telling them different demand conditions), the price adjusted relatively quickly to the new equilibrium price. This experiment has been replicated by a number of other economists.

Such an approach has several possible uses. By using the experimental method, economists can see how markets react under different institutional conditions. In a recent experiment, researchers tested a posted-price market and compared it to a double-oral auction market. In a posted-price market, firms and buyers post a price for a period of time and stick to it. Researchers found that prices tended to be higher in posted-price markets than in double-oral auction markets, a finding that led the U.S. Department of Transportation to ask the help of experimental economists in solving a problem concerning the pricing of railroads and barges. The railroads had asked the Department of Transportation to switch from privately negotiated freight rates to publicly posted rates, arguing that public posting would protect both themselves and small barge owners from unannounced price-cutting by large barge owners. When experimenters simu­lated the two types of markets, however, they found the opposite to be the case: price posting tended to yield higher prices than private negotiation and hurt small barge operators. The railroads dropped their request.

Another test done by experimental economists was of the Coase theorem, which states that parties who are capable of harming one another but who can negotiate will bargain to an efficient outcome, regardless of which side has the legal right to inflict damage. The experimental results confirmed this prediction. However, the experiment found that when individuals were endowed with the legal right by means of a coin flip, they almost inevitably did not extract the full individual rational share of the bargaining surplus that is predicted by game theory. Instead, the bargainers almost inevitably shared the surplus equally. This suggests that a fairness ethic, not pure rational individual maximization, governs distribution. This in turn suggests that individuals do not perceive asymmetric property rights as legitimate if they are awarded randomly. However, when property rights were awarded to the individual who won a game of skill before the experiment, the experimenters noted that two-thirds of the individuals with the property right obtained most of the joint surplus, whereas under the random assignment treatment none did.

Given the problems of empirically testing theories, it is not surprising that this work has gained in importance. Its acceptance by the profession would have wide-ranging implications and would require significant changes not only in the training of economists but also in their role in society and their entire approach to economic problems.

A related development is analysis through simulations. In this work models are designed that have multiple agents who follow simple, locally based rules. Then simulations are run, and it is determined which rules survive and which do not. This allows modelers to choose assumptions by their survival rather than by introspection.