SHADAC Staff

p 612.624.4802
e shadac@umn.edu

Understanding the New CPS Processing System and New 2018 Health Insurance Coverage Estimates

On Tuesday, September 10^th, the U.S. Census Bureau released estimates of income, poverty, and health insurance coverage for 2018 from two large-scale federal data surveys conducted annually and administered by the Bureau: the American Community Survey (ACS) and the Current Population Survey Annual Social and Economic Supplement (CPS ASEC). The CPS ASEC is a key source of data on poverty, health insurance coverage, health status, and medical out-of-pocket spending, in addition to a range of other economic and demographic characteristics for the United States population. For information on the ACS estimates see the recently published blog, “National Uninsurance Increased alongside Eight States; Three States Saw Decreases.”

The 2019 CPS ASEC (2018 calendar year estimates) is the first release of a “Production File” to feature a new data processing system that takes into account changes made to the health insurance questions in the 2014 CPS ASEC redesign. Changes include the addition of new variables related to types of coverage that a person held at the time of interview, Marketplace coverage, and sub-annual coverage.^[1] For a detailed discussion of these changes, please see several papers produced by the Census Bureau that are available here. This blog briefly discusses the implication of the new processing system on the ability to trend estimates of health insurance coverage.  

According to Census, health insurance estimates created using the new processing system should not be compared with estimates produced using the old processing system.¹ Currently, the Census Bureau has released three files that that were produced using the new processing system: the 2017 CPS ASEC Research File, the 2018 CPS ASEC Bridge File, and the 2019 CPS ASEC Production File. The Census Bureau publication released this past week—“Health Insurance Coverage in the United States, 2018”—compares the calendar year 2018 health insurance estimates from the 2019 CPS ASEC Production File with the calendar year 2017 estimates from the 2018 Bridge File because both of these files use the new processing system.

The table below illustrates the importance of distinguishing between estimates of health insurance coverage produced by the old versus the new processing system.

Sources: The uninsurance estimates are from “Health Insurance Coverage in the United States, 2018” and “Health Insurance Coverage in the United States, 2017”.

For example, when comparing estimates of the 2017 national uninsured rate, the old processing system (2018 Production File) produces an estimate of 8.8 percent, while the new processing system (2018 Bridge File) produces an estimate of 7.9 percent, nearly a whole percentage point lower. When examining changes in the uninsured rate between 2017 and 2018, the Census Bureau uses the 2017 estimate from the 2018 Bridge File (7.9 percent), created using the new processing system, to compare against the 2018 estimate from the 2019 Production File (8.5 percent), also created using the new processing system.

Using only estimates from data produced by the new processing system thus allows for a true apples-to-apples comparison.