After Census Bureau field representatives collect Current Population Survey Annual Social and Economic Supplement (CPS ASEC) survey responses, the Census Bureau processes the information into a data file that can provide a nationally-representative snapshot of the U.S. population. Specifically, the Census Bureau uses a series of statistical programs to resolve logically inconsistent answers (clean), fill in missing or incomplete information (impute), and account for survey sampling (weight). Collectively, these efforts are known as data processing.

Revisions to the data processing system were informed by over a decade of research and evaluation. Prior research identified limitations related to health insurance content in the CPS ASEC and suggested that improvements to the questionnaire and processing would produce higher quality estimates of health insurance coverage. After a decade of research showing that the CPS ASEC captured less health insurance coverage in comparison with other federal surveys (Hess et al., 2001; Kenney & Lynch, 2010; Pascale, 1999), the Census Bureau introduced a redesigned questionnaire in the 2014 survey. Data based on the new questionnaire became available during the traditional fall release (e.g., Smith & Medalia, 2014, 2015; Barnett & Vornovitsky, 2016; Barnett & Berchick, 2017; Berchick, Hood, & Barnett, 2018). However, to release these data in this timely manner, the Census Bureau used the existing processing system, which could not fully incorporate the richer information contained in the redesigned questionnaire.

The 2017 CPS ASEC Research File builds on the redesigned questionnaire and debuts the revised data processing system. That is, the Census Bureau has produced two CPS ASEC data files for this survey year. The first file is the 2017 CPS ASEC Production File which used existing programs of cleaning, imputing, and weighting the data. The second file, the 2017 CPS ASEC Research File, introduces a new method of estimating coverage which builds from subannual estimates to determine whether a person was covered at any point in the previous calendar year, and refines the methods by which respondents’ data are cleaned, imputed and weighted to better reflect the rich economic and demographic data collected by the CPS ASEC. As most health insurance coverage questions concern the previous calendar year, coverage estimates in both 2017 CPS ASEC files corresponds to coverage in 2016.

In this working paper, we describe key changes implemented in the new processing system and present health insurance coverage estimates from the 2017 CPS ASEC Research File and 2017 CPS ASEC Production File.