Estimating Heterogeneous Risk Preferences from a Panel of Real-World Betting Choices

Abstract

We develop a rich structural model that accounts for individual heterogeneity in risk preferences, both within and across utility types, and estimate it using a unique panel dataset of individual betting activity. We estimate the population distribution of utility curvature, loss aversion, and probability weighting, and evaluate their relative importance in explaining individual behavior. We find that, on average, utility curvature and loss aversion are mild, and there is light (heavier) probability weighting in the domain of gains (losses). We also find that substantial heterogeneity across individuals is present in all model parameters. Estimating a mixture model of utility types, we find that half of the individuals exhibit loss aversion, while almost all exhibit probability weighting, highlighting the importance of probability weighting in explaining observed behavior. These findings have substantial implications for microfounded models in economics and finance.