A Non-Parametric Test of Risk Aversion

Abstract: In economics, risk aversion is modeled via a concave Bernoulli utility within the expected-utility paradigm. We propose a simple test of expected utility and concavity. We find little support for either: only 30 percent of the choices are consistent with a concave utility, only two out of 72 subjects are consistent with expected utility, and only one of them fits the economic model of risk aversion. Our findings contrast with the preponderance of seemingly "risk-averse" choices that have been elicited using the popular multiple-price list methodology, a result we replicate in this paper. We demonstrate that this methodology is unfit to measure risk aversion, and that the high prevalence of risk aversion it produces is due to parametric misspecification.