JAMA Health Forum | 2021
Fewer People May Have Become Uninsured in 2020 Than Feared
Abstract
The effect of the COVID-19 pandemic on health insurance coverage is still not known with certainty. Yet health economists and policy makers are eager to understand how many people became uninsured in 2020 so that we can assess how well the safety net established by the Affordable Care Act (ACA) is working. In this issue of JAMA Health Forum, Bundorf et al1 make an important contribution to our understanding of what happened in 2020 with their study that draws on data from the US Census Bureau’s Household Pulse Survey (HPS). The study finds that rates of employersponsored insurance (ESI) steadily declined by 0.2 percentage points each week from April through December 2020. However, rates of other sources of coverage, especially public sources, increased. The authors1 divide their data into 2 distinct time periods in 2020: the spring and summer and the fall and winter. During the first period, they find that losses of ESI were only partially offset by increases in other sources of coverage. This leads to their main result that 2.7 million people aged 18 to 64 years became newly uninsured during the spring and summer of 2020. However, in the later period, the authors find that weekly gains in rates of coverage from alternative sources, primarily Medicaid, mostly offset weekly declines in rates of ESI coverage, with no statistically significant change in the number of uninsured people during the fall and winter of 2020. This study is important because the authors1 have overcome various data limitations with the HPS to provide some of the most robust and credible estimates available to date of coverage changes during the pandemic. Administrative data has revealed part of the story by documenting a surge in Medicaid and CHIP enrollment2 and a somewhat smaller increase in Marketplace plan selections,3 but household survey data are the only source of information on the number of people who have ESI and who are uninsured. The emerging story suggests that nearly 3 million people lost ESI and became uninsured for several months. This estimate is in line with a July 2020 report4 that projected 2.9 million people younger than 65 years (including children) would become uninsured in 2020 due to the COVID-19– related recession, based on detailed analysis of data on employment losses by industry, state, and demographic characteristics. A key result from Bundorf et al1 is that ESI coverage rates declined steadily over 9 months. We also know from the administrative data that Medicaid enrollment has steadily increased each month since April 2020. This suggests that some of the people who lost ESI and became uninsured during the spring and summer may have eventually found an alternative source of coverage, such as Medicaid, later in 2020 and, perhaps, in early 2021. We also have more robust evidence that the ACA safety net is making a difference in outcomes by state expansion status. The authors1 found that in states that had expanded Medicaid as permitted under the ACA, declines in ESI coverage were offset to a greater degree by gains in other sources of coverage compared with states that had not expanded Medicaid. The robustness of the study by Bundorf et al1 rests on the authors’ use of data from the US Census Bureau’s HPS. The experimental internet survey started in April 2020 and missed the first month of the pandemic’s influence. Response rates were very low, ranging from about 1% to 10% each week. Most concerning to observers of the HPS estimates, there was a discontinuous jump in insurance coverage estimates in September 2020 that strongly suggested measurement error rather than a real change in coverage levels. The discontinuity was likely the result of a change in the survey instrument first fielded in September. Despite these limitations, the HPS has some important strengths. The survey covered a large sample of the population drawn from the US Census Bureau’s Master Address File and included tens of thousands of respondents each week, more than tenfold + Multimedia + Related article