Title: A semiparametric approach to model Medicare insurance choice and expenditures
Authors: Chan Shen - University of Texas MD Anderson Cancer Center (United States) [presenting]
Abstract: Medicare insurance covers a large percentage of the US population. Within Medicare, there are several choices beneficiaries can make including whether to purchase supplementary insurance and whether to choose Health Management Organization (HMO) plans. There are debates about the impact of these choices on medical expenses. We use a semiparametric approach to model Medicare insurance choice and expenditures. For insurance choice, this semiparametric approach allows a very flexible form for choice probabilities, thereby avoiding the independence of irrelevance alternatives assumption in multinomial logit models. In addition to a number of important demographic variables, we find that that market penetration of HMO, healthcare knowledge level and number of chronic conditions have significant impacts on insurance choice. For health expenditures, we treat the insurance variables as endogenous choices whose impacts flexibly depend on exogenous covariates. Both equations are estimated using a recursive-differencing estimator for a nonparametric expectation, which has desirable asymptotic and finite sample properties. We used the Medicare Beneficiary Survey data to create the study cohort.