Panel Paper: Uninsured Spells and Transitions in the Current Population Survey

Thursday, November 8, 2018
Wilson B - Mezz Level (Marriott Wardman Park)

*Names in bold indicate Presenter

Edward Berchick and Brett O'Hara, U.S. Census Bureau

Objective. Earlier research has demonstrated that collecting health insurance coverage information at the monthly level in the CPS Current Population Survey Annual Social and Economic Supplement (CPS ASEC). helps to improve the quality of annual estimates of health insurance coverage (e.g., Medalia, O’Hara, & Smith, 2015; Pascale, Boudreaux, & King, 2015). However, it is possible that data collection at the monthly level may not yield accurate monthly-level estimates of the percent of the population covered by insurance or transitions on or off insurance. In this paper, we present CPS ASEC monthly estimates of health insurance transitions (from insured to uninsured and vice versa), as well as estimates of uninsured spell durations. We benchmark the CPS ASEC against estimates from the Medical Expenditure Panel Study (MEPS). We also highlight the importance of considering left and right censoring in such research.

Data Sources/Study Setting. We use the internal preliminary research file for the 2016 CPS ASEC and the public-use 2015 MEPS both of which include monthly health insurance information for the civilian non-institutionalized population for calendar year 2015. We restrict both samples to the population under age 65 (when many individuals become Medicare-eligible) and to the months of coverage that fall during the 2015 calendar year.

Study Design. We employ a number of analytic strategies. We first calculate transition probabilities across months; that is, we compare the month-specific probability of moving from insured to uninsured and from moving from uninsured to insured, primarily using nonparametric Kaplan-Meier methods. We compare the duration of uninsured spells in CPS ASEC and MEPS by calculating quantiles (25th, 50th, and 75th percentiles) of months spent uninsured. We also examine the sensitivity of each model to issues of right and left censoring, as well as weighting. For all analyses, we analyze two age categories (0-18 years and 19-64 years) separately.

Results. TBA

Conclusion Most existing research on monthly health insurance dynamics uses the Medical Expenditure Panel Study (MEPS) or the Survey of Income and Program Participation (SIPP). However, compared with these other surveys, the CPS ASEC has a much larger sample size and is released in a more timely manner, respectively. This allows the CPS ASEC to be used to answer more research and policy-relevant questions than either MEPS or SIPP. Our analysis will highlight some strengths and weaknesses of potential CPS ASEC monthly data.