May 2025
·
3 Reads
We present a comprehensive meta-analysis of prospect theory (PT) parameter estimates, synthesizing data from 166 papers that report 812 estimates based on the decisions of approximately 52,000 subjects across 69 countries. We develop and implement a novel approach that enables the joint meta-analysis of multiple PT parameters, explicitly modeling potential covariation among them. Our analysis reveals that the average utility curvature (measured via the coefficient of constant relative risk aversion) is 0.33 for gains (95% credible interval [0.31, 0.36]) and 0.29 for losses (CrI [0.25, 0.32]). The average likelihood sensitivity parameter is 0.68 (CrI [0.66, 0.70]), and the average elevation parameter is 0.98 (CrI [0.95, 1.02]). These values align with the stylized facts of PT: diminishing sensitivity to changes in wealth relative to a reference point, and to changes in probabilities moving away from the endpoints of the unit interval. Beyond these central tendencies, we uncover substantial heterogeneity in parameter estimates, limiting the precision of predictions about future parameters. Among the study-level characteristics examined, the measurement method emerges as the most significant predictor of parameter variation, highlighting potential violations of procedural invariance. Nonetheless, a large portion of the heterogeneity remains unexplained even after adjusting for a wide range of study features, suggesting that subtle design differences meaningfully influence behavior. Our results thus offer not only a synthesis of existing findings but also a roadmap for future research.