How do people reduce compound lotteries?

This paper provides experimental evidence for a rather important question: How do people reduce compound lotteries? As an alternative to the reduction of compound lotteries axiom of expected utility, I also test the compound independence axiom that can be employed by several decision theories. While the non-parametric test does not reject the compound independence axiom, I do not find support for evaluation of compound lotteries by the compound independence axiom through rank dependent utility that was used to motivate the axiom. The reduction of compound lotteries axiom is tested by two methods used in the literature. The validity of the axiom depends on the particular method used. While binary choices support the validity of the reduction axiom, there is no evidence of evaluation of compound lotteries through the axiom. Furthermore, out-of-sample predictions indicate that expected value is the best predictor of elicited certainty equivalents of compound lotteries. Interestingly, expected utility is the best predictor of elicited certainty equivalents for simple lotteries. The results suggest that subjects follow different mechanisms when evaluating compound lotteries as compared to simple ones.