Strategy experiment in dynamic asset pricing

This study presents a strategy experiment in a simple dynamic asset pricing model, where the current market equilibrium asset price is determined by aggregated individual expectations of next period's price. After participating in an introductory laboratory experiment on expectation formation participants formulate a complete forecasting strategy. These strategies are programmed and markets are simulated. Participants receive feedback from the results of these simulations and can adapt their strategy. Four rounds are played and the entire experiment lasts eight weeks. A final laboratory experiment compares predictions of participants with the predictions of the submitted strategy. We find that most of the participants submit complicated strategies and that strategies become more complicated over the rounds. Most markets converge to a steady state price only after many periods, if at all. The number of converging price sequences increases over the rounds. These results suggest in general slow convergence and learning of the subjects over the rounds. Even in a stationary environment learning the correct fundamental price level turns out to be difficult. An important part of the non-convergence seems to be caused not by individual strategies but by the interaction of several strategies. From the final experiment we conclude that the strategies are a good representation of what participants do in a laboratory experiment. (c) 2004 Elsevier B.V. All rights reserved.