Learning about Stability of Risk Preferences

Classical theories posit that risk preferences are stable across decision contexts. Two branches of empirical literature assess this claim. Structural approaches examine within-person consistency of model-based estimates of risk aversion, while “model-free” approaches examine within-person correlations of risky choices. We elucidate the latent structure underlying the model-free approach. Using this structure, we develop a new approach to assessing preference stability that has stronger testable implications, and we partially identify features of stability and heterogeneity of risk preferences. Our study illustrates a general principle: partial identification through minimal assumptions grounded in economic theory robustly bridges fully structural and model-free methods.