The sensitivity of expected utility violations to the experimental design : how context affects risky choice

Abstract

Expected Utility Theory is tested under different question contexts. It is hypothesized that previously cited independence violations may result from experimental biases rather than a shortcoming of the Theory. An experimental survey presents risky choice questions as lotteries and as "real life" scenarios to test the relative frequency of independence violations under different test conditions. Simple proportion-difference test statistics show that some choice pairs elicit significantly different choices under the scenario contexts. A more sophisticated analysis, using logit regression models, finds that the scenario contexts reduce choice biases caused by the similarity of the alternatives. Choices over scenario-contexts are found to be consistent with Expected Utility Theory. Violations of Expected Utility Theory over lottery contexts are attributed to the similarity of the alternatives.