Initial Reply to Banki, Simonsohn, Walatka and Wu (2025)

Abstract

Banki et al. (2025) ("BSWW") show that by simultaneously (i) dropping subjects who made even one error in Oprea (2024)'s training questions (75% of the sample) and (ii) focusing attention on the median, Oprea (2024)’s finding of prospect theoretic behavior in "deterministic mirror" tasks seemingly disappears.  I show that it is BSWW's focus on the median within this subsample that drives this result: Oprea (2024)’s findings continue to hold in this subsample at the mean, and non-parametric tests continue to affirm prospect theoretic behavior in mirrors.  I present the full distribution of behavior and show that the median is a poor summary statistic for these data: the distributions in BSWW's subsample look quite similar to the full sample, with a clear and significant skew toward prospect-theoretic errors.  I also present evidence that BSWW's analysis substantially overestimates the degree of confusion among subjects in Oprea (2024)'s experiment, and show that most of BSWW's further lines of criticism falsify claims that Oprea (2024) does not make and that are not part of his empirical argument that lottery valuations are influenced by complexity.

Keywords: complexity, prospect theory JEL codes: C91, D91, G0