Introduction The ex post efficiency of risk segmentation

Ž . Journal of Health Economics 19 2000 499–512 www.elsevier.nlrlocatereconbase Risk segmentation: goal or problem? Roger Feldman , Bryan Dowd DiÕision of Health SerÕices Research and Policy, UniÕersity of Minnesota, 420 Delaware St., SE, Box 729, Minneapolis, MN 55455, USA Abstract This paper traces the evolution of economists’ views about risk segmentation in health insurance markets. Originally seen as a desirable goal, risk segmentation has come to be viewed as leading to abnormal profits, wasted resources, and inefficient limitations on coverage and services. We suggest that risk segmentation may be efficient if one takes an Ž . ex post view i.e., after consumers’ risks are known . From this perspective, managed care may be a much better method for achieving risk segmentation than limitations on coverage. The most serious objection to risk segmentation is the ex ante concern that it undermines long-term insurance contracts that would protect consumers against changes in lifetime risk. q 2000 Elsevier Science B.V. All rights reserved.

1. Introduction

Whenever consumers are offered multiple health plans, it is unlikely that an identical distribution of health risks will select every plan. The degree to which the distribution of risks varies among health plans, and the impact of risk variation on the efficiency and fairness of the health care system, are topics of considerable interest to health economists. There appears to be a growing consensus that risk segmentation is undesirable, that ‘‘risk-adjusted’’ payments to health plans are necessary either to forestall risk segmentation or to offset its effects, and even that competitive approaches to allocating health care resources may not be viable without such adjustments. Corresponding author. Tel.: q1-612-624-5669, fax: q1-612-624-2196. Ž . E-mail address: feldm002tc.umn.edu R. Feldman . 0167-6296r00r - see front matter q 2000 Elsevier Science B.V. All rights reserved. Ž . PII: S 0 1 6 7 - 6 2 9 6 0 0 0 0 0 5 1 - 5 Our goal is to discuss the optimal pricing of health insurance plans. A synthesis of the literature is needed because different authors have compared different contracts and have used different perspectives on the problem. We begin by reviewing how the analysis of risk segmentation has changed over the last 20 years. Specifically, in the 1970s health economists thought risk segmentation was efficient because their perspective was after people know their health-risk status Ž . ex post . Current concerns over the efficiency of risk segmentation are based on a Ž model in which people make choices before they know their health-risk status ex . ante . The paper is organized as follows. In Section 2, we discuss the efficiency of ex post risk segmentation, and suggest that managed care may address some long- standing problems that impeded efficient ex post segmentation of risk. In Section 3, we discuss how the goal of efficient ex post risk segmentation became a problem when risk segmentation was viewed from an ex ante perspective. In Section 4, we suggest ways to improve efficiency from both an ex post and ex ante perspective. Finally, we compare concerns about efficiency vs. fairness in an ex ante, multi-period world where people do not know their future risk status.

2. The ex post efficiency of risk segmentation

The current interest in risk segmentation is preceded by a rich history in the Ž . health economics literature. Arrow 1963 noted that ‘‘hypothetically, insurance requires for its full social benefit, a maximum possible discrimination of risks’’. However, Arrow believed that insurance markets display ‘‘a tendency to equalize, rather than to differentiate premiums,’’ thus redistributing income from low-risk to high-risk consumers. Arrow interpreted this tendency as evidence that the market was not ‘‘genuinely competitive,’’ because in a competitive market any plan that failed to charge lower premiums to low-risk consumers would lose those con- sumers to competing insurers who charged lower premiums. Arrow’s crucial point was that competition leads to increased efficiency in health insurance markets by forcing the segmentation of risks. As explained in Ž . greater detail by Pauly 1970 , Arrow’s reasoning was simple: when consumers in different risk classes face an average or ‘‘community-rated’’ premium, low risks pay too much for coverage and demand an inefficiently low level of coverage, while high risks pay too low a price and demand too much coverage. Ž . Ž . Rothschild and Stiglitz 1976 RS provided a different rationale for the efficiency of risk segmentation. They compared markets in which insurers have perfect information about the risk of the insured population, to markets with imperfect information. RS show that experience-rated policies will be offered to each risk class of consumers. When insurers have perfect information, the resulting ‘‘separating’’ equilibrium is efficient for both groups. However, if insurers cannot distinguish between the two types of consumers, their only strategy is to offer policies with characteristics that induce consumers to reveal their risk ex Ž . post by choosing one plan vs. another . RS refer to these policy characteristics as ‘‘self-selection mechanisms.’’ The specific self-selection mechanism they discuss Ž . is the extent of coverage e.g., the coinsurance rate . The policy intended for low risks must have such a low level of coverage that it is not chosen by high risks. Because low risks cannot buy the amount of coverage they would like at premiums that are actuarially fair for their group, RS conclude that the separating equilibrium under imperfect information is not efficient. Thus, Arrow, Pauly, and RS reached the same conclusion for very different reasons. Arrow and Pauly argued that the decision to pool risks through commu- nity-rated premiums induced market failure by distorting prices. RS argued that failure to segment risks was the result of market failure in the form of poor insurer information. Imperfect information led to inefficiency in the RS model because coverage is an imperfect self-selection mechanism. If a better self-selection mechanism could be discovered, then a Pareto-superior equilibrium could be attained. With better technology, insurers could create policies designed specifically to appeal to low and high risks. Thus, throughout the 1960s and 1970s, economists’ primary concern with risk segmentation was how to achieve it. The discovery of a self-selection mechanism more effective than coverage would have been hailed as the solution to inefficiency in the market for health insurance. The market did not have to wait long. By the early 1980s, a new type of self-selection mechanism had arrived in the form of managed care. Because managed care — particularly the type of managed care that uses restricted provider networks — appears to be more attractive to low-risk consumers than Ž . high-risk consumers Brown et al., 1993; Hellinger, 1995 , managed care might Ž . allow low risks to purchase more generous coverage i.e., reduced cost-sharing without fear that ‘‘their’’ plan will attract high risks. Managed care is by no means a perfect or costless self-selection mechanism, either to consumers or insurers Ž . Newhouse, 1996 . Low risks might reject managed care if risks could be segmented some other way, but they may find managed care less onerous than subsiding premiums for high risks, if it lets them buy coverage at prices that are actuarially fair for their risk group. Thus, managed care may be an efficiency-im- proving self-selection mechanism.

3. The goal becomes a problem