Grading Medicaid: Evidence from Transitions to Medicare

Medicaid spending in the highest-spending states is more than three times Medicaid spending in the lowest spending states for all four defined categories of Medicaid enrollees (children, adults, individuals with disabilities, the elderly). Despite this wide variation in spending, we know very little about the extent to which it is driven by differences in the pools of Medicaid enrollees across states (selection) vs. differences in state Medicaid programs or local market factors (treatment effects). To the extent that some portion of the variation comes from differences in state programs (treatment effects), we also know little about what program characteristics drive spending differences and whether the marginal Medicaid dollars improve the health of Medicaid enrollees.

In this paper, we seek to unpack the variation in Medicaid spending across states using a novel empirical strategy. We leverage data describing demographics, fiscal spending, and mortality for the universe of Medicaid enrollees linked to similar data for the universe of Medicare enrollees. We use this data to allow us to compare fiscal spending and health effects of each state’s Medicaid program relative to a single, homogeneous alternative program: Medicare. By comparing each state’s Medicaid program to Medicare, we can effectively compare each state’s Medicaid program to each other state’s Medicaid program, allowing us to assess the extent to which program factors influence the variation in observed Medicaid spending across states.

Specifically, our empirical strategy consists of first identifying individuals enrolled in the Supplemental Security Income (SSI) program but not enrolled in the Social Security Disability Insurance (SSDI) program at age 63. These individuals are enrolled in Medicaid at age 63 but not in Medicare. We follow these individuals into Medicare at age 65, at which time they become dually enrolled in Medicaid and Medicare. We observe how the fiscal spending and health of these individuals change discontinuously at age 65 in a difference-in-differences design where individuals who are enrolled in both the SSI and SSDI programs (who are also enrolled in both Medicaid and Medicare) at age 63 act as a control group (for them there is no change in program enrollment at age 65).

We use this design to first compare Medicaid and Medicare coverage overall in terms of fiscal cost and health impact. We then divide the population according to their state of residence and estimate separate Medicaid vs. Medicare effects for each state. We then rank state Medicaid programs according to their Medicaid vs. Medicare spending effects and according to their Medicaid vs. Medicare mortality effects. From these rankings, we make conclusions about whether states with low spending effects relative to Medicare achieve lower levels of spending without hurting health or instead whether there appears to be a clear trade-off between spending and health.