Alternation bias and reduction in St. Petersburg gambles

Reduction of compound lotteries is implicit both in the statement of the St. Petersburg Paradox and in its resolution by Expected Utility (EU).We report three real-money choice experiments between truncated compound-form St. Petersburg gambles and their reduced-form equivalents. The first tests for differences in elicited Certainty Equivalents. The second develops the distinction between ‘weak-form’ and ‘strong-form’ rejection of Reduction, as well as a novel experimental task that verifiably implements Vernon Smith’s dominance precept. The third experiment checks for robustness against range and increment manipulation. In all three experiments the null hypothesis of Reduction is rejected, with systematic deprecation of the compound form in favor of the reduced form. This is consistent with the predictions of alternation bias. Together these experiments offer evidence that the Reduction assumption may have limited descriptive validity in modelling St. Petersburg gambles, whether by EU or non-EU theories.