Negative recency, randomization device choice, and reduction of compound lotteries

We report an experiment in which subjects are not indifferent between real-money lotteries implemented with randomization devices that are equivalent under the Reduction Axiom. Instead, choice behavior is consistent with subjective distortion of conditional probability, and this persists in treatment conditions that control for (i) computational limitations and (ii) possible confounding by ratio bias.

► Across groups, 80% choose the reduced form, violating the Reduction Axiom. ► Reduced form is chosen over compound lottery, consistent with negative recency. ► Compound structures can introduce biases into experimental data. ► Negative recency: an alternative explanation for the myopic loss aversion result. ► Negative recency: an alternative explanation for the St. Petersburg Paradox.