Evaluating consumers’ choices of Medicare Part D plans: A study in behavioral welfare economics

We propose new methods to model behavior and conduct welfare analysis in complex environments where some choices are unlikely to reveal preferences. We develop a mixture-of-experts model that incorporates heterogeneity in consumers’ preferences and in their choice processes. We also develop a method to decompose logit errors into latent preferences versus optimization errors. Applying these methods to Medicare beneficiaries’ prescription drug insurance choices suggests that: (1) average welfare losses from suboptimal choices are small, (2) beneficiaries with dementia and depression have larger losses, and (3) policies that simplify choice sets offer small average benefits, helping some people but harming others.