Persistence in Medicare Reimbursements and Personal Medical Accounts

Abstract

This study uses the Continuous Medicare History Sample (CMHS), a longitudinal sample of Medicare beneficiaries from 1974 to 1997, to identify persistence in Medicare reimbursements. Consistent estimates of Medicare reimbursement persistence are obtained using a dynamic panel Tobit model. Lagged reimbursements are statistically significant in explaining a beneficiary's current year reimbursements, yet their effect is moderate in magnitude. An additional dollar of Medicare spending in the previous year gives rise to $0.19 in spending in the current year for those who have positive expenditures, and the effect ranges between $0.16 and $0.25. Past spending patterns for up to five years are also used to evaluate the out-of-sample prediction capabilities of a multi-equation cross-sectional model. Simulated spending patterns for several cohorts are derived and compared to actual spending. The simulated spending distributions are used to study the feasibility of establishing retirement health accounts for Medicare patients.

Keywords: dynamic panel, Tobit, Medicare, persistence

JEL Classification: I1, C23, C24