A Structural Model of Health Insurance Choice and Health Care Demand in the Medicare Managed Care Program

Brand, Keith Joseph, Department of Economics, University of Virginia
Stern, Steven, Department of Economics, University of Virginia
Friedberg, Leora, Department of Economics, University of Virginia
Pepper, John, Department of Economics, University of Virginia
Merwin, Elizabeth, School of Nursing, University of Virginia

I examine the joint determination of health insurance choice and subsequent health care utilization in the Medicare managed care and supplemental insurance markets in the United States. The objectives are to evaluate the welfare impact of the Medicare Advantage, the Medicare managed care program, taking into account the effect of risk selection, and to advance the applied literature on health insurance choice and health care demand. I model health care demand as simultaneously determined in five dimensions: inpatient care, outpatient care, doctor visits, prescription drugs, and dental care. The model incorporates uncertainty about efficacy when treatment decisions are made, limits on the efficacy of treatment, and diminishing marginal product both intensively and extensively. Demand functions are implicitly defined by the first order conditions of the consumer. Likelihood contributions are generated by numerically solving for the vector of unobservables that is consistent with characterizing the observed utilization data as utility-maximizing. I employ a mixed multinomial logit approach to health plan choice. Risk averse consumers choose health plans by taking expectations of indirect utilities over a known distribution of health states. The model uses a unified framework in that the health insurance and utilization decisions are based on the same underlying preferences. I use individual level data on health insurance choice and subsequent utilization, and data on aggregate enrollment in Medicare Advantage. The aggregate data is observed at the county/age/gender level. The decomposition to the age/gender level allows me to solve for a vector of unobservables that is specific to every managed care plan/county combination. These unobservables are used to construct moment conditions that supplement the likelihood function in estimation. The results suggest that while each age/gender group benefits from Medicare Advantage on average, younger age groups benefit more. The results also indicate that a small percentage of each age/gender group is adversely affected by Medicare Advantage and this share is larger in older age groups. In a second policy experiment, the results indicate that the extra cost to Medicare through supplemental insurance because of increased ex post moral hazard is greater than the extra consumer surplus generated by supplemental insurance.

Note: Abstract extracted from PDF text

PHD (Doctor of Philosophy)
All rights reserved (no additional license for public reuse)
Issued Date: