Risk aversion and expected-utility theory: A calibration exercise

Rabin (Econometrica 68(5):1281–1292, <CitationRef CitationID="CR19">2000</CitationRef>) argues that, under expected-utility, observed risk aversion over modest stakes implies extremely high risk aversion over large stakes. Cox and Sadiraj (Games Econom. Behav. 56(1):45–60, <CitationRef CitationID="CR5">2006</CitationRef>) have replied that this is a problem of expected-utility of wealth, but that expected-utility of income does not share that problem. We combine experimental data on moderate-scale risky choices with survey data on income to estimate coefficients of relative risk aversion using expected-utility of consumption. Assuming individuals cannot save implies an average coefficient of relative risk aversion of 1.92. Assuming they can decide between consuming today and saving for the future, a realistic assumption, implies quadruple-digit coefficients. This gives empirical evidence for narrow bracketing. Copyright Springer Science+Business Media, LLC 2007