Risk Attitudes, Sample Selection and Attrition in a Longitudinal Field Experiment

Longitudinal experiments allow one to evaluate the temporal stability of latent preferences, but raise concerns about sample selection and attrition that may confound inferences about temporal stability. We evaluate the hypothesis of temporal stability in risk preferences using a remarkable data set that combines socio-demographic information from the Danish Civil Registry with information on risk attitudes from a longitudinal field experiment. Our experimental design builds in explicit randomization on the incentives for participation. The results show that the use of different participation incentives can affect sample response rates and help identify the effects of selection. Correcting for endogenous sample selection and panel attrition changes inferences about risk preferences in an economically and statistically significant manner. Estimates of risk preferences change with these corrections. In general we find evidence consistent with temporal stability of risk preferences when one corrects for selection and attrition.