Medicare Part D is an experiment in the public-private provision of health care. In Part D, insurers and pharmaceutical manufacturers negotiate over what drugs will be covered (eligible for plan reimbursement), the price paid by the insurers for the drugs, and the beneficiaries'drug copayments. Medicare then pays the insurance plans a subsidy worth, in expectation, 75% of drug costs. Any beneficiary can enroll in a given Part D plan at the same premium, regardless of health status;insurers are willing to cover both the sick and the healthy because the subsidies vary by a beneficiary's diagnoses. Insurers receive a predetermined amount for each of 84 conditions, where the amount is meant to cover the expected costs of treating a beneficiary with that condition. The subsidy amounts were set using data from federal retirees and Medicaid beneficiaries in 2000 and 2002. I argue that some diagnoses are over-reimbursed, meaning that the subsidy amount exceeded the average treatment cost in Part D;others are under-reimbursed. Such inaccuracies in the subsidy scheme could be caused by differences in the drug needs between the retiree/Medicaid sample and the true Part D market, or by the higher out-of- pocket costs in Part D. Additionally, industry dynamics since 2000 (entrance of new branded drugs or onset of generic competition) could raise or lower the costs of treating certain diagnoses. I use data compiled by Center for Medicare and Medicaid services on the medical claims and Part D spending of a random 5% subsample of beneficiaries. I estimate by linear regression the portion of plan liability attributale to each diagnosis (here I follow the methods used to set the original subsidy amounts). I find that, indeed, some subsidies greatly exceed or are exceeded by the mean treatment costs in Part D. I then develop a theoretical model of contracting for branded drugs in the presence of inaccurate subsidies. I argue that drugs that treat generously reimbursed conditions will be more likely to be covered by Part D plans and that their out-of-pocket costs will be lower, since insurers want to attract beneficiaries with these conditions. But because insurers cannot credibly threaten to exclude such drugs, the negotiated price paid for these drugs will be higher. I find evidence for my hypotheses in Part D plans'pricing and coverage in 2009 and 2010. Given that higher reimbursements were shared among beneficiaries (who paid less for drugs), pharmaceutical firms (who received higher prices), and insurers (who presumably benefitted from the remainder), I now estimate the fraction of an additional dollar in reimbursement that would accrue to each of these groups. This step utilizes a structural model of joint drug-insurance plan demand joined to a bargaining model of moment inequalities.

Public Health Relevance

For this proposed project, I will investigate how the system of subsidies used to support Medicare Part D affects the incentives of insurers and pharmaceutical firms negotiating over the price and out-of-pocket costs of branded drugs. I first demonstrate that drugs that treat conditions that were generously subsidized were more likely to be covered, were offered to beneficiaries at lower out-of-pocket costs, and were purchased by insurers at higher prices. I then estimate how much of each extra dollar of subsidy would have accrued to insurers (in the form of higher profits), pharmaceutical firms (in the form of higher negotiated prices), and beneficiaries (in the form of lower out-of-pocket costs).

National Institute of Health (NIH)
Agency for Healthcare Research and Quality (AHRQ)
Dissertation Award (R36)
Project #
Application #
Study Section
HSR Health Care Research Training SS (HCRT)
Program Officer
Willis, Tamara
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Johns Hopkins University
Schools of Arts and Sciences
United States
Zip Code