Robust Risk Adjustment

With Aaron Banks, Pietro Tebaldi, and Matthew Zhan

Managing risk selection is one of the central challenges of regulated insurance markets. To offset the unraveling forces of adverse selection, regulators implement risk adjustment policies that attempt to compensate insurers for cost differentials in the population. Despite their popularity, it is well known that these policies create new incentives for insurers to select enrollees to maximize regulatory payoffs. Using extensive variation in risk adjustment regulation in Medicare, we show that private insurers systematically attract consumers more rewarded by future risk payments, conditional on true spending risk. Intuitively, insurers have an informational advantage relative to the regulator, having more knowledge about patients' medical histories and more advanced technologies to predict future spending. This allows insurers to selectively attract patients who the regulator might consider ex-ante expensive to insure (e.g., diabetics) but who are ex-post cheap (e.g., diabetics with good eating habits). Using this variation, we estimate a model of demand for insurance that accounts for heterogeneous correlations between enrollees' preferences, their observable risk factors, and their true underlying risk. The same variation allows the identification of insurers' private information. We study the optimal design of risk adjustment policies that are robust to the presence of this information asymmetry. We contrast robust policies' performance to those with access to insurers' private information, highlighting the gains and losses from adopting robust design perspectives in insurance markets.