Health Plan Choice and the Valuation of Plan Attributes in the ACA Marketplaces
Discussant: Benjamin Sommers
To explore these issues, detailed person-level plan enrollment data from the California, Colorado, and Washington state-based marketplaces for 2014-2017 are linked to plan characteristics from the Robert Wood Johnson Foundation HIX Compare data and provider network data from Vericred. Conditional and mixed logit regressions are used to examine enrollee plan choice as a function of various financial and non-financial plan attributes, and logit regressions are used to explore plan switching as a function of changes in both chosen-plan attributes and overall choice-set characteristics. These estimates are also used to calculate changes in expected consumer surplus due to simulated changes in the choice sets.
I find that enrollee plan choices are strongly related to premium, deductible, provider network size, and quality star ratings in the expected directions. In particular, enrollees have a significant willingness-to-pay for larger provider network size, in the range of $76 to $266 for a 10 percentage point increase in network size across these three states. Moreover, enrollees are sensitive to star rating measures of plan quality, with, for example, a willingness to pay of $427 for four vs. three stars in California and $361 for four vs. three stars in Colorado (as star ratings were not used in Washington during this time) Examining changes in plan attributes and choice set characteristics on switching, there is some evidence that enrollees are more likely to switch plans when a plan’s provider network size decreases and when a plan’s star rating decreases, but there is no strong evidence that individuals switch due to increases in premiums.