Controls for Risk Aversion in the Laboratory

Abstract

Experiments are being used to generate measures of characteristics of individuals that need to be controlled for economic analyses. Examples include the measurement of risk aversion, inequality aversion, and trust. Imprecision in measuring those characteristics may lead to misleading conclusions in tests of the underlying theory of choice. We consider the need to control for risk aversion in tests of expected utility theory. We show that the conclusions drawn about the choice theory may be sensitive to statistical imprecision in measuring risk aversion. We show that the use of instruments to estimate risk attitudes does not allow sufficiently precise estimates of risk aversion for these purposes. Given existing laboratory technology and statistical models, we conclude that controls for risk aversion should be implemented using within-subjects, revealed preference designs that utilize the direct, raw responses of the subject. These statistical issues are generally applicable to a wide variety of experimental situations.

Keywords: Expected utility theory, risk aversion, experiments

JEL Classification: D81, C91