Decisions Under Risk are Decisions Under Complexity: Comment

A re-analysis of Oprea (2024)’s data suggests that that measurement error produced by a confusing experimental design underlies the provocative claim that prospect theory’s risk attitudes reflect mistakes arising from “complexity” rather than underlying preferences. In the reported studies, participants valued risky lotteries (e.g., a 10%chance of $25) and riskless “mirrors” of those lotteries (e.g., 10% of $25 for sure) with similar means, exhibiting the fourfold pattern and loss aversion for both. This equivalence, however, was driven by the 75% of subjects who erred on comprehension questions. These subjects produced excessively noisy data, with first-order stochastic dominance violation rates 5 to 10 times higher than in previous studies. The remaining 25% of subjects largely valued mirrors at their expected value and lotteries in line with prospect theory. Participants with a higher likelihood of understanding experimental instructions (e.g., students over online subjects, STEM majors, better-compensated, higher CRT scorers) behaved more in line with prospect theory for lotteries than mirrors.