Consumer Demand for Health Insurance in the ACA Marketplaces: Evidence from Covered California

Creation of the Affordable Care Act (ACA) marketplaces, where eligible nonelderly individuals can receive subsidies for purchasing insurance, offers researchers a unique opportunity to study how consumers choose health plans.  In particular, researchers can exploit variation in consumer demographics, plan premiums, plan benefit design, and geographic markets.  Previous efforts to estimate nonelderly consumer demand for health insurance often consider a single firm, where plan choice is typically limited and the covered population may be relatively homogeneous, or the individual market, where detailed data are very limited. 

In this study, we use 2014 and 2015 individual-level enrollment data from Covered California, the state-based marketplace in California, to estimate a discrete-choice structural model of demand for health insurance.   Obtained through the California Public Records Act, the data provide detailed enrollee and plan information for the 1.2 million and 1.3 million people who enrolled in Covered California in 2014 and 2015, respectively.  For each of the 2.5 million enrollee-year observations, the data indicate the insurer, metal tier, network type, and premium of the plan selected by the enrollee.  In addition, we observe key enrollee characteristics, such as age, income, gender, county of residence, and household composition.

Our analysis indicates considerable responsiveness to price and heterogeneity in consumer preferences.  To obtain unbiased estimates of demand price elasticities, we are able to exploit an exogenous shock across enrollment years.  After pricing well above its competitors in 2014, Kaiser Permanente reduced its premiums by close to 10 percent in 2015, while other insurers either increased or held premiums steady.  We find that Kaiser significantly increased its market share from 17 percent in 2014 to 24 percent in 2015 across the state, suggesting that consumers are responsive to price.  Furthermore, we observe considerable heterogeneity in consumer preferences.  Older and higher-income individuals are more likely to avoid HMO plans and purchase a plan from one of the Blues (i.e., Anthem Blue Cross and Blue Shield of California), while younger and lower income individuals are more price sensitive and tend to select relatively cheaper HMO plans.  These results provide suggestive evidence of adverse selection in plan choice.  We also find evidence of inertia, as 90 percent of returning enrollees choose the same plan in 2015.  In addition, we observe over 20,000 enrollees who are eligible for cost sharing subsidies choosing a dominated plan (e.g., a more expensive gold plan with an 80 percent actuarial value instead of a cheaper affordable cost sharing plan with a 94 percent actuarial value), suggesting the need for improved consumer outreach and education efforts.         

Once we recover the primitives of the demand model, we estimate the market equilibrium in order to run a set of counterfactual policy simulations.  In particular, we estimate the social welfare change associated with the ACA’s premium tax credits and individual mandate penalties.  Furthermore, we derive the welfare-maximizing tax credit and penalty policy, and compare this policy to the one implemented under the ACA.  This counterfactual analysis could have important implications for policymakers seeking to improve the ACA.