Models of Investor Forecasting Behavior - Experimental Evidence

Investors' forecasting behavior affects their trading decisions and the resulting asset prices. It has been shown in the literature how different, apparently reasonable investor forecasting behaviors can lead to qualitatively different asset price trajectories. For example, investors' confidence or panic in relying on observed asset prices to forecast future prices can lead to price bubbles, price crashes, and unpredictable price cycles. In this paper, we report the results of behavioral experiments to study investors' forecast formation. We present the design of an experiment consisting of multiple investors that participate in a market for a virtual asset. We conducted three experimental sessions with different participants in each session. Unlike most other market experiments with known finite durations and predictable decreasing prices toward the end of the experiment, our experimental sessions had random durations chosen to emulate an infinite horizon market with discounting. Asset price bubbles and cycles occurred in all three sessions. We fit different models of forecast formation to the observed data. There is strong evidence that the investors forecast future prices by extrapolating past prices, even when they know the fundamental value exactly and the extrapolated forecasts differ significantly from the fundamental value. The rational expectations hypothesis seems inconsistent with the observed forecasts. There is some evidence for adjustment of forecasts towards the fundamental value as the extrapolated forecasts deviate from the fundamental value. The calibrated forecasting models of all participants were consistent with dynamical systems that exhibit price bubbles and cycles