Early Evidence on Employment Responses to the Affordable Care Act
Using the 2010-2014 Medical Expenditure Panel Survey-Insurance Component (MEPS-IC) List Sample, this study investigates how labor market outcomes are changing over time as the ACA is implemented. The MEPS-IC is a nationally-representative, establishment-level survey of U.S. employers that collects detailed information on the provision of health insurance, including dimensions of offers, eligibility, and enrollment as well as the percentage of the establishment’s workers that are part-time. As an employer-level dataset, it also has the advantage that, unlike most individual-level survey data, it collects information about characteristics of the workforce at the establishment, such as the wage distribution and age distribution of its workers. These workforce characteristics are likely to be important predictors of employer decisions about offering health insurance and the use of part-time versus full-time employees. We augment the MEPS-IC with time-varying, geographic-specific information about local labor market conditions (e.g., county-level unemployment rate and establishment size distribution) and state-level insurance market attributes (e.g., state-based regulations for the employer group market, Medicaid eligibility for childless adults, Exchange type, and Exchange funding obtained from the federal government).
We consider two econometric modeling approaches. The first uses an instrumental variables econometric methodology to estimate the extent to which offer rates are changing due to changes in part-time work and vice-versa. The second approach takes advantage of a new question added to the MEPS-IC instrument in 2014, which asks each establishment whether it was offering insurance or not in 2014 and 2013. From this, we are able to model whether an establishment that was offering insurance in 2013 dropped it in 2014, and if that decision to drop coverage can be explained by variation in establishment, workforce, and policy-related variables described above.
Because we are using the MEPS-IC, the analyses are being conducted within a Census Research Data Center and require a rigorous disclosure process. We have not yet disclosed output for these analyses so cannot report preliminary results, but will go through the disclosure process in spring 2016. We have full confidence that we will have a completed in paper well in advance of the conference.