UCLA

In this dissertation, I build a novel collection of income thresholds used by all states to determine Medicaid eligibility for non-elderly, non-disabled adults, for all years between 1996 and 2010. I use this collection of income thresholds to explore same-year impacts of access to Medicaid, using a simulated instrumental variables approach. First, I explore effects of Medicaid access on opioid use, finding heterogeneous effects. For small states, opioid sales and opioid use disorder treatment episodes increased. However, for larger states, and for the country on average, effects on these outcomes are muted. I find no evidence that opioid overdose deaths increased, and I interpret this heterogeneity as evidence that prescription regulations in larger states were effective in preventing increasing access to Medicaid from leading to increased use of opioids. I also explore effects of Medicaid access on substance-use disorder (SUD) treatments. I find that while SUD treatment episodes increased substantially after increases in income thresholds for Medicaid eligibility, daily client loads and overall facility counts did not increase. This implies that SUD treatment episodes must have become faster following these Medicaid expansions. This change may be explained by another set of findings: hospital-based SUD treatment facilities increased their client counts, with a particularly large increase within relatively quick detoxification treatments. Comparing my results to prior research, I note that the 2010 Affordable Care Act (ACA) significantly broadened Medicaid’s coverage of SUD treatments, which may explain past findings that when focusing on the period following 2010, Medicaid expansions led to an outsize increase in outpatient rehabilitation services, rather than hospital-based detoxification services. In the final chapter of this dissertation, I provide a detailed discussion of my sources and methods for building the novel collection of income thresholds for Medicaid eligibility, including how I decide what values to use when existing data are missing or conflicting.