The Instability of a Heterogeneous Cobweb economy: a Strategy Experiment on Expectation Formation

Which strategies do agents use when forming expectations about future prices, and how often do combinations of these strategies lead to stable or unstable outcomes? To answer these questions we performed a four-round strategy experiment in a cobweb economy. Each market consisted of 20 periods and in each period strategies had to forecast next period's price. It was common knowledge that the realized market price was a function of all individual expectations, but subjects did not know the underlying market equilibrium equations. Subjects gained experience in a 'normal' experiment before submitting their first strategy. All strategies were programmed and after each round the subjects received feedback about the relative performance of their strategy, and were allowed to revise their strategy for the next round. Subjects use a wide variety of different strategies. Over the rounds quadratic forecasting errors decrease and realized market prices move to a neighborhood of the rational expectations (RE) steady state, but at the same time the complexity of the price fluctuations increases. Convergence to the unique RE steady state occurs in less than 10% of all cases. In the final round 60% of the price fluctuations appears to be chaotic. Strategy simulations with homogeneous agents typically show regular behaviour, with prices converging to a steady state or to a 'far from the steady state' stable cycle. Heterogeneous interaction of simple prediction strategies thus seems to be the main source of the endogenous price fluctuations, frequently leading to a boundedly rational equilibrium of 'close to the steady state chaos'.