Panel Paper: Health Insurance and Early Retirement: Evidence from State Exchanges and Medicaid Expansion

Saturday, November 10, 2018
Madison B - Mezz Level (Marriott Wardman Park)

*Names in bold indicate Presenter

Kevin Feeney, University of Southern California


Policies implemented under the Affordable Care Act (ACA) to increase the number of insured persons may affect labor supply through “job lock” mechanisms. These include the expansion of state Medicaid programs and federal subsidies on the exchange marketplace. I study whether these policies encourage early retirement, examining persons aged 62-64 who are eligible for (early) Social Security benefits. I use a quasi-experimental differences-in-differences framework to compare individuals eligible for subsidies/Medicaid to those not eligible for either within a given state.

Prior work using policy variation from the ACA has focused largely on Medicaid expansions, ignoring the exchange subsidies offered to individuals in non-expansion states, or, under differences-in-differences designs, limiting the analysis to those below the poverty line who don’t qualify for marketplace cost reductions/tax credits. The marketplace subsidies may also encourage early retirement, and individuals at higher incomes may be more likely to exhibit behavior consistent with job lock since employer-sponsored insurance is more common. Past research has also focused on a wide age range, pooling across individuals with different demand for health insurance and labor supply.

I use a variation of the differences-in-differences model, comparing individuals eligible for Medicaid or subsidies (“treated”) to those at higher incomes eligible for neither (“control”). I bin individuals into thresholds of the poverty line based on eligibility for Medicaid (100-138%), cost sharing reductions on the exchanges (100/138-250%), or tax credits (250-400%), and those ineligible for these benefits (401-600%) to identify the effect of health insurance on early retirement. I estimate the model separately among Medicaid-expansion and non expansion states. This specification is robust to variety of fixed effects, such as state-specific year fixed effects. I take care to address potential limitations of this design, such as parallel trends and endogeneity of income, labor supply, and health insurance. In a separate differences-in-differences analysis of Medicaid expansion only, I examine those below 100% to reconcile results extant studies.

Data comes from the American Community Survey (2010-2016), which has information on employment and health insurance for over 200,000 individuals aged 62-64. By focusing on this age group, I hope to better measure the impact of health insurance on retirement since these individuals are eligible for early Social Security benefits, which may encourage withdrawal from the labor force if individuals substitute between earning and benefits.

Results show both exchanges subsidies and Medicaid increased health insurance coverage. Those that experienced increases in coverage were more likely to drop out of the labor force, retire, and take up Social Security benefits. Roughly half of this effect is driven by unemployed individuals ending their job search. The increase in retirement also coincides with reductions in full-time employment and employer-sponsored health insurance, consistent with traditional models of job lock. I find no change in self employment, and insignificant, small increases in part-time work. As a placebo check, I examine individuals ages 65-67 who have Medicare and find no change in their employment or Social Security benefits.