返回Google圖書搜尋

Modeling Employee Choice of Health Plan

出版	SSRN, 2007
URL	http://books.google.com.hk/books?id=cpLezwEACAAJ&hl=&source=gbs_api

註釋Much of the health care insurance coverage available in the United States is offered in the context of plan choice. An understanding of how to predict individuals' choice of health plans is instrumental in understanding and managing these markets. In addition, choice of plans is a key component in many proposals for national health coverage in the US, so a good understanding of the mechanisms of choice will assist in evaluating these policy options. This has lead to a broad class of literature, primarily in the context of choice in employer-sponsored plans. The majority of this literature relies on a logit model, a utility-based choice framework. This model presumes the errors in the utilities are independent and homoskedastic, an assumption often described as the independence of irrelevant alternatives. This assumption is violated when one alternative is a better substitute for a plan than another, as we presume would hold when an indemnity plan is marketed with two HMO plans. In order to accommodate this violation, many researchers group the closer substitutes in a nested logit model. This introduces heteroskedasticity across nests, but still presumes independence of error terms across choices. This work adds to the literature on health plan choice and the presence of asymmetric information in health insurance markets by studying the health plan choice behavior of employees of a large US employer. The work contributes to the literature by providing a unified, logically consistent, utility-maximizing framework to analyze health plan choice in the presence of asymmetric information and risk aversion. Bayesian inference provides the tools needed to allow unobserved preference shocks to be correlated across choices in a flexible way, and for those shocks to be correlated over time, within a multiperiod, multinomial probit model. Various papers in the health plan choice literature incorporate some of these features, but this is the first effort that combines them all in a logically consistent way. In this more flexible model, dimensions of health risk typically unoberservable are found to influence choice, with sicker individuals less likely to enroll in the tightly managed plan and the high-deductible consumer-driven health plan. This is a significant finding, as empirical evidence of adverse selection, beyond that measured by observable factors, is rare. Understanding the interactions between premiums and health risk in determining health plan choice is important for assessing the functioning of broadly defined health insurance markets. Non-discrimination regulations in the United States require that the employer not adjust individual contribution levels for individual risk characteristics. Absent adjustment for differences in the average health risk mix attracted to the plan, plans that are expensive because their benefit design attracts sicker patients are penalized in price equally with plans that are expensive due to poor management. A greater understanding of how asymmetric information plays out in the employer-based health plan market will allow better discrimination between the premium differences due to adverse selection, and those due to plan inefficiencies. That information, in turn, can lead to more cost-effective health plan benefit design.