Preference for Randomization: Empirical and Experimental Evidence

We investigate violations of consequentialism in the form of the stochastic dominance property. The property is shared by many theories of choice and implies that the decision-maker prefers receiving the best outcome for sure over all lotteries that involve multiple outcomes. We run experiments to demonstrate that dominated randomization can be attractive. In treatments where decision-makers are asked to submit multiple decisions without knowing which one is relevant, many participants submit contradictory sets of decisions and thereby induce a dominated lottery between outcomes. Explicit choice of non-consequentialist randomization is observed in a separate treatment. A possible reason for the effect is the desire to avoid having to make the decision. A large data set on (high-stake) university applications in Germany shows patterns that are consistent with a preference for randomization.