contract. For example, mental health and substance abuse benefits are often carved out, meaning that the regular health plan chosen by the enrollee is not responsible for providing mental health and substance abuse care. A separate managed care company specializing in this area of care receives a contract from the employer or other payer to provide this benefit. Carve outs can be done for reasons of controlling moral hazard or moderating selection incentives (Frank and McGuire, 1998).

Our analysis shows how this might happen. A profit-maximizing plan might distort the shadow price for service s’ to affect its enrollee profile. If any service is carved out, this strategy will change. In particular, if service s’ is carved out, rationing strategies for all other services will generally be affected. Carving out any one service will affect the efficiency of that service provision as well as the nature of insurance market equilibrium overall.

Our model allows us to examine these effects. In this section we report illustrative results of carving out three services from the main insurance contract. We recalculate the profit-maximizing q’s in the case of carving out each service. The most commonly carved out service is mental health and substance abuse (known together as behavioral health), and we start with that.

The new q’s are shown in Table 5 for the case in which individuals know 25% and 40% of the information contained in prior spending and plan payments are made using the HCC risk adjustment system. Note that by comparing Tables 4 and 5 there is some effect on the q’s for services other than the one that is carved out. However, the effect is generally to move the calculated q away from unity. For example, the calculated q for heart disease assuming 40% information and HCC risk adjustment is 0.33 (Table 4).

When mental health and substance abuse are carved out, the calculated q falls to 0.28. Similarly, the q for mental health assuming 40% information and HCC risk adjustment is 0.76 (Table 4). When musculoskeletal services are carved out the calculated q is 0.72. Thus, while the distortion incentives for the carve-out service are reduced, distortions may increase for the services remaining in the health plan.


Health plans paid by capitation have an incentive to distort the quality of services they offer to attract profitable and deter unprofitable enrollees. Characterizing plans’ rationing as imposing a “shadow price” on access to care, we show that the profit maximizing shadow price for each service depends on the dispersion in health costs, how well individuals forecast their health costs, the correlation among use in illness categories, and the risk adjustment system used for payment. We further show how these factors can be combined to form an empirically implementable index that can be used to identify the services that will be most distorted in competition among managed care plans. A simple welfare measure is developed that measures the distortion caused by selection incentives.