Does Medicaid Coverage Reduce Short-term Mortality Among Childless Adults?

Monday, June 13, 2016: 1:35 PM
Robertson Hall (Huntsman Hall)

Author(s): Thomas DeLeire

Discussant: Andrew Goodman-Bacon

The 2010 Affordable Care Act (ACA) enabled states to expand their Medicaid programs to all adults and children with incomes up to 133 percent of FPL. As of November 2014, 27 states had expanded their programs and there had been an increase of over 10 million individuals enrolled since implementation.  As the ACA expanded eligibility primarily for adults and in particular, childless adults, most of this increase was likely among those populations.  

As the lack of access to regular sources of care is associated with increased risk of mortality, and because low-income childless adults have been found to be at increased risk of chronic disease, there is reason to believe that increased coverage of low-income childless adult populations would lead to reductions in avoidable death even in the short-term.  In fact, two studies have found large reductions in mortality following coverage expansions to childless adults (Baicker and Sommers 2012 and Baicker, Long, and Sommers 2014).  The Oregon Health Insurance Experiment, however, found no sizeable or statistically meaningful reductions in mortality in the year following enrollment into Medicaid (Finkelstein et al. 2011).  Given the importance of knowing how Medicaid affects mortality, and given the divergent estimates in the recent literature, there is a need for additional credibly designed studies. 

In this study, we exploit a policy reversal in Wisconsin, during which a major public insurance expansion for childless adults was implemented and, several months later, abruptly frozen.  Individuals who applied after the program was frozen were placed on a waitlist. We obtain estimates of the effect of Medicaid on mortality by comparing those who applied prior to the program freeze and received benefits to those who applied after the program freeze and did not receive benefits.

We use a sharp regression discontinuity design that employs the timing of the enrollment suspension and waitlist introduction to estimate the size of any mortality effect. A particular strength of our study is that we rely on the state’s own administrative records rather than on self-reported enrollment. The data for our study are Medicaid enrollment files matched to a census of mortality records from the National Center of Health Statistic’s National Death Index (NDI).  The NDI indicates both the year and cause of death. The Medicaid records allow us to observe all enrolled and waitlisted applicants, including their exact date of application. We merge the two administrative datasets using names and Social Security numbers.

We find no evidence that public insurance reduces mortality among low-income childless adults in the two years following enrollment into public insurance in Wisconsin.  Our point estimate is both small (0.003 percentage points from a baseline rate of 2 percent) and is also not statistically meaningful.  We also find no reductions among any specific cause of death.  This effect size similar in magnitudes to what was found in the OHIE, but is much smaller that estimates based on differences-in-differences from state-level insurance expansions.  Our results suggest that not every expansion in public coverage will lead to short-run reductions in mortality.