The Welfare Implications of Risk Adjustment in Imperfectly Competitive Markets

Risk adjustment is a common policy for mitigating the effects of adverse selection when government regulation limits insurer ability to rate consumers according to their expected risks. Under the ACA, risk adjustment requires that insurers with lower-than-average risk consumers make transfer payments to insurers with higher-than-average risk consumers, equalizing the expected risk borne by each insurer.

In this paper, I study the social welfare implications of risk adjustment. I specify a differentiated products model and show theoretically that risk adjustment can be welfare-reducing even if it perfectly equalizes risk across firms. The reason is that risk adjustment incentivizes firms to set premiums that may increase the expected risk of those who select into the insurance pool. Expected consumer risk may increase if (1) firm premiums and expected risk are positively correlated and (2) firm cost and adverse selection are negatively correlated.

I study the impact of risk adjustment in the ACA exchanges using consumer-level data from the California state exchange. The data contain about 2.5 million records across the 2014 and 2015 plan years. I combine these data with financial data from the ACA medical loss ratio (MLR) reports to estimate demand and cost. I estimate demand for health insurance using a nested logit discrete choice model. I obtain non-parametric estimates of plan marginal costs by inverting the firm's first-order conditions for profit maximization. I relate these estimates to premiums to measure how marginal costs vary with premiums. Adverse selection is present if higher premiums have a positive and statistically significant effect on marginal costs.

My estimates of demand and cost are consistent with theory. I find that low-income individuals, young adults, and single individuals have more premium-elastic demand. I estimate that a $100 annual premium increase would reduce a plan’s demand by 20 percent. If the premiums of all exchange plans were to increase by $100 per year, demand for exchange coverage would fall by 2 percent. My estimates of cost provide statistically significant evidence of adverse selection. Controlling for plan metal tier, I find that an increase in premiums results in higher marginal cost.

After estimating demand and cost, I simulate the impact of risk adjustment in the ACA exchanges. I find that risk adjustment compresses equilibrium premiums such that more expensive gold and platinum plans become cheaper and cheaper bronze and silver plans become more expensive. Consumer welfare increases by approximately $200 because premiums for more expensive plans decline and the ACA's price-linked subsidies offset the higher premiums for cheaper plans. Total social welfare is about the same under risk adjustment because the increase in consumer welfare is offset by an increase in subsidy spending. I conduct policy simulations using the estimated model and find the impact of risk adjustment is sensitive to the subsidy design. If ACA price-linked subsidies were converted to fixed subsidies or vouchers as proposed in the American Health Care Act of 2017, risk adjustment would decrease annual per-capita consumer surplus by $200 and social welfare by $400.