Do Preference Reversals Disappear When We Allow for Probabilistic Choice?

The "preference reversal phenomenon," a systematic disparity between people's valuations and choices, poses challenges for theory and policy. Using a very general formulation of probabilistic preferences, we show that the phenomenon is not mainly due to intransitive choice. We find a high degree of regularity within choice tasks and also within valuation tasks, but the two types of tasks appear to evoke very different cognitive processes, even when the experimental environment tries to minimise differences. We discuss possible implications for modelling and eliciting preferences.