JEL classification: C15, C23, C25, D81 Keywords: Panel Data, Unobserved heterogeneity, Choice under risk
ABSTRACTExperimental economics focuses on eliciting preferences, studying individuals one at a time to take into account their heterogeneity. Experiments have the appealing property of collecting enough observations to perform such an analysis. In real word, and in natural experiments, individuals cannot be observed according to experimenters' needs.We propose a method that aggregates over individuals taking into account their heterogeneity. Using data from a natural experiment, we estimate three models of decision making under risk: Expected Utility, Rank-Dependent Expected Utility and Regret-Rejoice. Our results show that individual-wise analyses can be substituted by pooled approaches without losing information about individual heterogeneity.
RISK ATTITUDE IN REAL DECISION PROBLEMSThis paper aims at providing new evidence on risk aversion, focussing on aggregation over individuals eliciting their heterogeneity. The relevance of this topic is twofold: from a theoretical point of view, by showing that differences among people significantly affect their decisions, we highlight the need for theoretical developments which can better account for diversity. From an applied viewpoint, the statistical significance of such individual factors allows for better estimates than the ones one obtains when disregarding this issue.In lab experiments, individuals are observed as many times as the experimenter needs.This has provided ground for a flourishing literature focussing on the analysis of choice rules estimated individual by individual. Such a kind of approach is nearly unfeasible when data stemming from natural experiments are used. However, natural experiments' datasets are very attractive for they have the benefit of salient incentives. Therefore, in order to study choice rules when the number of observations on each individual is exogenously determined, researchers need to pool individual data.Participants both in lab and natural experiments differ in crucial characteristics. In pooling their data, we cannot discard this fact. Using data from a natural experiment, namely a TV show, which involves players taking decisions between risky prospects with outcomes up to half-a-million euros, we estimate three models of decision making under risk: Expected Utility, Rank-Dependent Expected Utility and Regret-Rejoice. The characteristics of the game, which involves lotteries composed of small and very large prizes, allow us to investigate the performance of the Rank-Dependent Expected Utility 2 (RD) and Regret-Rejoice (RR) utility functionals 1 in addition to the standard Expected Utility (EU) functional.Our research is designed to capture nuances in players' behaviour neglected by the EU formulation. Specifically, with the introduction of a RD functional, we can take into account the fact that some players appear to overweight the extreme prizes of a lottery with respect to the others; while the RR functional allows us ...