| 简介： | Expected utility theory (EUT) is a partimonious theory that explains behavior under risk and uncertainty. Rabin and Thaler (2001) showed that EUT of wealth is not a satisfactory explanation of risk aversion. We use empirical data from a controlled laboratory experiment to show that EUT of income does not explain risk aversion either and loss aversion is a more convincing explanation. More importantly, we demonstrate that an empirical model with loss aversion and rank-dependent probability weighting can best fit the experimental data. Lacking knowledge on reference points is the usual obstacle in explaining risk averse behavior with loss aversion. When certainty equivalents are directly elicited, it is often unclear what the decision maker's reference point is. We instead elicit valuations-Willingness to Pay (WTP) and Willingness to Accept (WTA) -for risky prospects. This experimental design combined with an empirical model based on Koszegi and Rabin (2006) is robust/insensitive to changes in reference points. Our empirical model features a single loss aversion parameter and a non-parametric probability weighting function and also takes EUT as its special case. Through a structural estimation, we show that both loss aversion and probability weighting are needed to explain risk aversion. The obtained estimates are consistent with literature. |