Please sign up for meetings at the link below:

docs.google.com/spreadsheets/d/1jFiVdlk-wKT7ONTIY0tLjZpSoRXTvqPwc-V5sDAkHAA/edit?usp=sharing

Abstract

I examine the preventive care decisions of diabetics in the U.S.before and after becoming eligible for Medicare coverage at age 65. I find that there is a discontinuous drop in the proportion of diabetics who report usinginsulin to manage their condition at age 65 of between 11.5% and 21.6% which is not explained by retirement, substitution for alternatives such as diet or medication, or an increased frequency of diagnosis. I show that the regression-discontinuity estimate of the effect of coverage on prevention is biased away from zero. This is due to its ignoring the compensatory effect of a longer lifteime horizon due to the increased access to emergency services granted by Medicare. I derive bounds for the Effect of Treatment on the Treated that result from this assumption, exploiting exogenous variation in the probability of survival due to unanticipated diagnoses of cancer. When combined with an additional assumption – that the probability of insulin usage is higher for those in worse health – I obtain tight bounds on this treatment effect that differ from the RDD results. The bounds indicate that among diabetic Medicare beneficiaries as a group, moral hazard decreased the proportion of insulin users net of the effect of improved access to treatment by between 0.6 and 2.2%.