We leverage the assumption that preferences are stable across contexts to partially identify and conduct inference on the parameters of a structural model of risky choice. Working with data on households' deductible choices across three lines of insurance coverage and a model that nests expected utility theory plus a range of non‐expected utility models, we perform a revealed preference analysis that yields household‐specific bounds on the model parameters. We then impose stability and other structural assumptions to tighten the bounds, and we explore what we can learn about households' risk preferences from the intervals defined by the bounds. We further utilize the intervals to (i) classify households into preference types and (ii) recover the single parameterization of the model that best fits the data. Our approach does not entail making distributional assumptions about unobserved heterogeneity in preferences.

Inference insurance partial identification revealed preference risk preferences stability D01 D12 D81 G22