Medicaid and Medicare together constitute the two largest government programs in America, accounting for almost a trillion dollars of annual spending (or over 5% of GDP). Both programs were introduced 50 years ago, as part of the Social Security Amendments Act. Medicaid was created as a health care program for the poor, while Medicare was primarily meant to serve the elderly, although both now serve a sizeable number of disabled individuals.
While Medicaid and Medicare were not originally designed to work together, there are presently over 9 million individuals enrolled in both. Those qualifying for both programs are typically either disabled individuals (who simultaneously receive SSI and SSDI), or the poor elderly (who receive SSI only). These individuals, who are referred to as ‘dual-eligibles’, account for a disproportionate 34% of Medicaid and Medicare spending, while amounting to only 13% of enrollees (CBO, 2013).
Given the substantial amount of federal as well as state funds spent on dual- eligibles (almost 2% of GDP), this population constitutes an important topic for public …nance and policy research. Moreover, this population’s experiences could be generalizable to Medicare recipients with other types of secondary insurance, since such insurance is carried by almost 90% of all Medicare recipients. As such, this re- search may relate to existing Medicare work on employer sponsored coverage (Chandra et al, 2010) and Medigap (Cabral et al, 2014).
In addition, dual-eligible care relates to questions of more general economic and theoretical interest. For example, cost-sharing is an important theme in dual-eligible care, given that this population is subject to minimal amounts of it. As such, iden- tifying cost sharing’s impact on this population’s spending constitutes an important and unanswered empirical question, especially since the impact of cost-sharing might
be di¤erent for this population than for others; to this end, the impact of cost sharing could be heterogeneous across health and economic status.
Moreover, this research ties into existing theory on cost-sharing, particularly on the theoretical mechanisms driving cost-sharing’s e¤ects. Cost sharing could reduce utilization through reduced moral hazard (Manning et al 1987), but also increase it through reduced preventative care and resulting o¤sets in preventable utilization (Chandra et al, 2010), and it is valuable to understand the relative importance of each mechanism. Moreover, the magnitude of these o¤sets could vary by health status, and be greatest for sicker individuals (Chandra et al, 2010).
To start, I examine the e¤ect of supplemental Medicaid coverage on those already in Medicare. Supplemental Medicaid coverage would a¤ect treatment composition and spending primarily through the e¤ective elimination of cost-sharing, which would otherwise be at around 20% under Medicare-only coverage (MedPAC, 2004).
Past work has been hindered by a number of empirical issues, which I attempt to overcome in this paper. First, given compositional di¤erences between those simulta- neously in Medicare and Medicaid, and those in Medicare-only, it has been di¢ cult to separate out the e¤ect of supplemental Medicaid coverage from that of underlying enrollee characteristics. As part of a novel identi…cation strategy, I leverage involun- tary disenrollments from Medicaid among those who were previously dually-enrolled in Medicaid and Medicare. Critically, disenrollment is not concurrent with individual- level changes to health or economic status. Rather, the disenrollments resulted from a 2009 Tennessee court decision, which allowed Tennessee to check the eligibility of existing Medicaid enrollees, and to disenroll anyone who no longer met the eligibility requirements.Prior to this ruling, Tennessee could not check the eligibility of a subset of Medicaid recipients (those who initially quali…ed via SSI), nor could it disenroll those no longer eligible, as a result of a 20-year long court prohibition (Wadhwani,
2010).1
In my analyses, I focus on Tennessee residents who were simultaneously in Med- icaid and Medicare as of 2008. Using a di¤erence-in-di¤erence approach, I examine health utilization among those who were and those who weren’t exogenously dis- enrolled, before and after disenrollment,. As part of my analysis, I focus on those disenrollees who no longer met the economic requirements for Medicaid, but who otherwise still met the disability quali…cations. By construction, this group is better o¤ than typical Medicaid or typical dual-enrollees, but probably only marginally so, meaning that my …ndings could still have external validity. After all, these individ- uals maintained their Medicaid coverage for an extended period of time, when they could have voluntarily dropped coverage. Throughout these analyses, I make use of Medicare administrative data, which comprehensively tracks health utilization for those in Medicare-only as well as for those dually-enrolled in Medicare and Medicaid (seeing as Medicare functions as the primary payer for both groups).
Altogether, I …nd that Medicaid disenrollment is associated with a 25-30% de- crease in overall utilization and spending, which is highly signi…cant. My results also suggest that moral hazard dominates over prevention-driven o¤sets in this setting, although both are present to some extent. To this end, I …nd that Medicaid dis- enrollment is associated with a substantial, 30% decrease in outpatient care, driven disproportionately by a reduction in elective care; at the same time, I …nd evidence of o¤sets-albeit less pronounced-in the form of increased inpatient utilization, dispro- portionately driven by increases to potentially avoidable care. Unfortunately, I am not able to e¤ectively identify the accompanying e¤ect on health outcomes or overall patient welfare.
1Incidentally, the 2009 Medicaid disenrollment is distinct from the one used in Garthwaite et al
(2014), given that the 2009 episode is judicially rather than legislatively driven (although coinciden- tally, both take place in the same state).
These results have substantial relevance for policy, suggesting that part of the high spending on dual-enrollees in Medicaid and Medicare is attributable to a lack of cost-sharing. The absence of cost sharing also results in increases to potentially low-value care. While my results imply that the imposition of cost-sharing could reduce spending and improve care e¢ ciency, even for low-income populations, this justi…ably raises concerns about a¤ordability (and also about health outcomes and patient welfare, which I do not address here). One worthwhile approach could be to o¤set increased cost-sharing with cash voucher payments (which could also be used for non-health expenses), and thereby …nancially incentivize e¢ cient care while ensuring that it remains a¤ordable. A similar approach has already been adopted in Indiana, and proposed in several other states (Saloner et al 2014).
In Section 2, I go over relevant institutional features of Medicaid, Medicare, as well as overlaps between them. I also review the dynamics of Medicaid disenrollment in Tennessee. In Section 3, I go over the data used in these analyses. In Section 4, I review my empirical design and implementation. In Section 5, I go over the results. In Section 6, I conclude.