Adapting the American Community Survey for the Affordable Care Act

To measure the accuracy of questions on health insurance premiums and subsidies added to the American Community Survey (ACS) and their utility in categorizing coverage type following the Affordable Care Act (ACA).

A reverse record check study where households in Minnesota with individuals enrolled in five different types of coverage-employer-sponsored insurance (ESI), non-group (outside the marketplace), marketplace, Medicaid and MinnesotaCare (a public plan requiring premium contributions from the enrollee)-were administered a telephone survey that included the ACS health insurance module appended with experimental questions on premiums and subsidies.

Enrollment records from a private insurer were used as the sample for primary survey data collection in the spring of 2015 using the ACS health insurance module. Survey data were matched back to enrollment records, which indicated coverage status at the time of the survey. The analytic sample includes matched data on about 600 individuals.

In total, 100%, 95.3%, and 86.9% of marketplace, non-group, and ESI enrollees, respectively, were correctly reported to have a premium. 74.6% of Medicaid enrollees were correctly reported NOT to have a premium and 77.4% of MinnesotaCare enrollees were correctly reported to HAVE a premium. For the subsidy item, correct reports of no subsidy were 99.1%, 93.8%, and 80.9% for ESI, non-group, and unsubsidized marketplace enrollees, respectively. A total of 72.4% of subsidized marketplace enrollees were correctly reported to have a subsidy. Analysis also indicates that an algorithm leveraging these two new data points can be used to separate the overall "direct purchase" category into two sub-groups: subsidized marketplace and unsubsidized marketplace combined with individual non-group.

Results indicate high levels of reporting accuracy for questions about premiums and subsidies. Thus, this post-ACA module of the ACS is capable of rendering more detailed coverage types than previously possible.