Modeling demand for Medicare Part D plans

The expansion of Medicare benefits to include coverage for prescription drug (Part D) is the largest since the program's inception. The benefit is delivered through private insurers competing for beneficiaries by offering a variety of options and features. Understanding preferences of beneficiaries for coverage features allows for a better assessment of the value of this private arrangement to Medicare beneficiaries. This research uses aggregate market outcomes to examine preferences of Medicare beneficiaries for the different features of Part D plans. Previous analyses of demand for Part D plans examined the early years of the program, when beneficiaries and insurers were still learning about the market. The main contribution of this research is the use of the most recently available public data, from 2009, to estimate separate demand systems for beneficiaries receiving low-income-subsidies (LIS) and for those who are not. I also examine the heterogeneity of consumer preferences and use the results to compute a lower-bound estimate of the potential welfare gains to LIS beneficiaries from an alternative assignment strategy that enrolls them in plans that best match their needs.;For non-LIS beneficiaries, the results show that they are significantly more sensitive to premiums than previously estimated, consistent with the evidence of consumer learning. Non-LIS beneficiaries value coverage in the gap less than the price of this coverage, which could be another piece of evidence for adverse selection in Part D markets. There is significant parameter heterogeneity, explained mainly by the level of medical expenditures. Beneficiaries with higher medical expenditures are less sensitive to premium and deductible, but more sensitive to average out-of-pocket spending.;LIS beneficiaries are found to be significantly more responsive to premium and measures of plan generosity. This heightened sensitivity to out-of-pocket spending implies that matching LIS beneficiaries to plans that best cover their medications could significantly improve their welfare, compared to the current policy of random assignment. Simulations of this strategic assignment policy show substantial welfare gains, with a lower-bound monthly estimate that is equivalent to 30% of average monthly spending on LIS beneficiaries by the government.