Evolutionary selection of forecasting and quantity decision rules in experimental asset markets

Abstract

Bao et al. (2017) find that bubbles are less likely to emerge in experimental asset markets when subjects make price forecasts only (Learning to Forecast treatment, LtF) than when they make trading quantity decisions (Learning to Optimize treatment, LtO) or both price forecasts and quantity decisions (mixed treatment). This paper provides two explanations for this difference. First, the subjects in the LtO and mixed treatment usually have a high intensity of choice parameter, which leads them to switch faster between the decision rules and a greater fraction of the population to choose the destabilizing strong trend-following rule. Second, the actual quantity decision may deviate substantially and persistently from the conditionally optimal level given the price forecasts in the mixed treatment, which amplifies the price deviation from the fundamental value. Our findings are helpful for understanding the root of financial bubbles and financial crisis, and designing policies to stabilize the market.