Journal of Economic and Social Dynamics

Abstract

We replicate the two classic choices of the Allais paradox in a randomized survey that manipulates only the framing of the probabilities: one group saw them as percentages, the other as outcomes of a two-digit lottery draw (00–99). The paradoxical pattern appears in full force under the percentage framing but is markedly attenuated when the same probabilities are presented as identifiable numeric events. The shift in preference for the safe option between Problem 1 and Problem 2 is −39 percentage points under percentages versus −12 under numbers; the difference between the two shifts is significant (difference-in-differences z = 2.34, p = 0.019; logistic interaction p = 0.032). We read the result through regret theory and through the certainty-vs-ze ro-aversion debate: making the states of the world identifiable changes both the anticipated regret attached to each option and the salience of the zero outcome, and thereby the strength of the certainty effect that sustains the paradox. We discuss the direction of the effect, competing accounts (probability weighting, event-splitting, and aversion to zero), and a design capable of isolating the mechanism.