Expected Utility Violations

Abstract

Our paper is an experimental inquiry into individual behavior under risk; we examine individual arbitrage between probabilities and payments in choosing among elementary lotteries and compare it to arbitrage predicted by expected utility theory. We draw from this empirical evidence some consequences concerning the form of the preference functional. The inquiry was conducted in France and in Morocco, which permitted to obtain samples of datas pertaining to two different cultural systems. The analysis of the datas obtained supports the assertion that there exists a certainty effect, whether we consider gains or losses. Our results are relatively more precise than earlier results [Kahneman and Tversky, 1979] and do not support the so-called “reflection effect”. Contrary to Cohen and Jaffray [1988], we do not find empirical evidence supporting the “linearity in probability” hypothesis prior to reaching the “certainty effect zone”. Our results are thus in line with those of De Neufville and Delquie [1988]. Finally, cultural differences do not bear any fundamental change concerning the violation of the expected utility axioms.