“Did Medicare Decrease Diabetics’ Insulin Usage Until the Advent of Part D?”
I document evidence that diabetics previously reliant on insulin stopped using insulin to a significant extent once they became eligible for Medicare in 1998-2006. This pattern only ends in 2006 when Medicare Part D is established, which subsidises Medicare recipients’ purchases of diabetic medication and insulin. I interpret these results as evidence for the relative strength of moral hazard on preventive medicine usage before and after Part D. When individuals don’t bear the costs of hospitalisation, but have to pay for preventive care, they will take fewer precautions than is optimal for holding down the cost of publicly provided health care. By contrast, a subsidy for preventive care can reduce this effect by making it less expensive to reduce the probability of a future hospitalisation from a given individual’s point of view. I conclude that Medicare Part D significantly increased insulin usage among elderly diabetics.The first piece of evidence is from a regression-discontinuity design using the 1998 wave of the Health and Retirement Study (HRS), which finds a large and statistically significant effect of Medicare eligibility on insulin usage – 10.77% fewer individuals above 65 manage their condition using insulin than those who are under 65. Additional evidence to support the regression-discontinuity design comes from following those who were under 65 in 1998 as they cross the eligibility threshold in subsequent survey years in 2000, 2002, 2004 and 2006. In 2000-2004, significantly more of those who were ineligible for Medicare in 1998 stop using insulin than start using insulin once they become eligible; in 2006, this pattern is reversed.
“Extrapolating Away from the Cutoff in Regression Discontinuity Designs by Identifying the Marginal Treatment Effect”
Would more or less preventive medicine be used in the United States if the age of eligibility for Medicare were lowered? I provide a new way to answer questions such as this one by joining together two literatures that have been developed in parallel. The recent literature on extrapolating away from the cutoff in RDDs has used observable characteristics of those away from the cutoff to evaluate the effect of changing the value of the cutoff. The literature on the Marginal Treatment Effect (MTE) has focused on how to identify the effect of a policy on those at the margin of being treated when treatment effects vary with unobservable characteristics. I show that previous claims of extrapolation away from the cutoff rely on strong assumptions regarding unobservable characteristics. However, combining the two approaches yields a strategy for identifying treatment effects away from the cutoff that avoids these assumptions and allows for selection on the unobservables.
