اثر پوشش مدیکر غیر فعال در بازار بیمه خصوصی

We investigate whether the removal of high-cost individuals from private insurance markets leads to greater coverage for individuals who are similar but not as high cost. Using data on insurance coverage from the Panel Study of Income Dynamics, we estimate the effect of the extension of Medicare to the disabled on the private insurance coverage of non-disabled individuals. We find that the insurance coverage of individuals who had a health condition that limited their ability to work increased significantly in states with high versus low rates of disability.

Does the removal of high-cost individuals from private insurance markets lead to greater coverage for individuals who are similar but not as high cost? In theory, if removing high-cost individuals reduces the range of hidden information in insurance markets, then it will dampen insurers’ incentives to protect themselves against adverse selection. As incentives to protect against adverse selection decline, pooling increases, which benefits the high-cost individuals who remain (Newhouse, 1996). The answer to this question is central to current health policy debates. Subsidies for insurance for the chronically ill, for example, seek to provide high-cost individuals with coverage at something like a community rate, but without forcing low-cost individuals to finance the cost through their purchase of insurance (Swartz, 2003 and Holahan et al., 2003). The general equilibrium effect of these subsidies, however, depends on how they affect the form and extent of coverage in the broader insurance market. Yet, despite this, there is little empirical evidence how such policies might perform. In this paper, we assess an historical example of a policy intervention of this sort, the extension of Medicare to the disabled, on the private insurance coverage of non-disabled individuals. In 1973, Congress extended Medicare benefits to beneficiaries of the Social Security Disability Insurance (SSDI) program; prior to then, there was no uniform, comprehensive public insurance program for the disabled. More important for the purposes of our study, extending Medicare to the disabled also had the effect of removing high-cost individuals from the broader pool of the privately insured. No empirical evidence exists of the impact of this policy, or similar policies, on the private insurance coverage of non-disabled individuals. We use data on insurance coverage from the Panel Study of Income Dynamics (PSID) from before and after the extension of Medicare to the disabled to estimate the effect of the program on private insurance coverage rates in the broader population. We find that the insurance coverage of individuals who had a health condition that limited their ability to work increased significantly in states with high versus low rates of SSDI beneficiaries. These “work-limited” individuals included, but were not limited to, SSDI beneficiaries. The increase in the number of work-limited individuals with insurance was far greater than the number of Medicare eligibles. Thus, the expansion of Medicare not only increased coverage among the targeted population of the disabled, but also among people who were similarly situated but less seriously impaired, suggesting the potential usefulness of subsidies to high-cost individuals in promoting insurance coverage generally. Then, we use data from the Health Insurance Council2 from 1970 to 1980 to estimate the effect of the extension of Medicare on private insurance comprehensiveness. As we discuss below, the same model that predicts that the extension of Medicare could have spillover effects also predicts that it could lead to increases in the comprehensiveness of coverage. We find that the comprehensiveness of private health insurance increased significantly after versus before the extension of Medicare in states with high versus low rates of SSDI beneficiaries. Our analysis proceeds in the next five sections. Section 2 presents a theoretical framework that explains how targeted subsidies for health insurance can have effects in the broader population. In Section 3 we discuss the data we use for our analysis, describe our methodological approach, and present tabular results which show evidence of a large impact of the extension of Medicare on non-disabled coverage rates. We embed this analysis in a more general econometric model in Section 4 and present results. In Section 5, we estimate the effect of the extension of Medicare on the scope of the policy offerings of private insurers. Section 6 concludes.

Using an important policy natural experiment, we have estimated the extent to which subsidies to cover high-cost individuals affect insurance coverage of others. According to our point estimates, extension of Medicare to an additional 0.5 percentage point of the population through the Social Security Disability Insurance program led to an increase in private insurance coverage of between 7.8 and 13.3 percentage points among individuals who described themselves as limited in the kind or amount of work that they can do. In 1980, these work-limited individuals were 15.7% of the total population. Thus, extending Medicare to an additional 0.5 percentage points of the population increased total insurance coverage by 1.2 (=0.157 × 7.8) to 2.1 (=0.157 × 13.3) percentage points. Subtracting off the 0.5 points due to the direct effect of the program gives a range for the spillover effect of 0.7 (=1.2 − 0.5) to 1.6 (=2.1 − 0.5) percentage points. Our results can be used to calculate the marginal “target efficiency” of extending Medicare to the disabled in the 1970s. Gruber (2003), for example, suggests evaluating such programs in terms of a “bang for the buck” — the total government spending per dollar of insurance cost covered (that is, the cost per newly insured weighted by the cost of those who are gaining insurance). Medicaid expansions to low-income adults, according to Gruber, have a budget cost of $1.30 per dollar of previously uncovered health costs. The budget cost exceeds $1.00 because of crowding-out of private coverage; for every $1.00 the government spends to newly insure someone with Medicaid, it must also give insurance to some number of individuals who would have had private coverage. Using the midpoint of our range of estimates of the extent of crowding-in of the non-disabled population of 1.15 percentage points (1.15 = ((0.7 + 1.6)/2)) per 0.5 percentage points of Medicare expansion, the formula for the target efficiency of the program is: View the MathML source Turn MathJax on where λ is the extent of crowding-out of private insurance by Medicare; cd is the cost of public insurance for a newly covered disabled person; and cn is the cost of private insurance for a newly covered non-disabled person. As the formula shows, one cannot calculate the target efficiency of the policy without information on extent of crowding-out of the program, the health spending of the newly covered disabled individuals, and the health spending of the newly covered non-disabled individuals. However, it is possible to calculate the conditions under which it is more target-efficient than a typical Medicaid expansion,5 or under which it achieves a target efficiency of less than one dollar per dollar of previously uncovered health costs — that is, no target efficiency cost. For example, assuming complete crowding-out of private insurance coverage (λ = 1) for the disabled, the target efficiency of the program reduces to 0.5cd/1.15cn. Thus, even with 100% crowd-out, the program is still more target-efficient than a typical Medicaid expansion, as long as 0.5cd/1.15cn < 1.3, or (cn/cd) > 0.33; the program has no target efficiency cost at all as long as 0.5cd/1.15cn < 1, or (cn/cd) > 0.43. In 2003, Medicare reimbursed $6471 per disabled enrollee in the program, while the annual premium for a conventional, single-person employer-sponsored insurance policy was $3576 (Kaiser Family Foundation, Employer Health Benefits 2003 Survey), yielding an estimate of (cn/cd) of 0.553. Hence, at recent values of cn and cd, the program is highly target-efficient. The extent to which our estimates of the effects of the extension of Medicare can be extrapolated to future targeted subsidies remains an open question. Although our point estimates are quite large, several factors suggest caution in applying them out-of-sample. First, our estimates are measured with considerable error: the lower 95% confidence bound in most specifications is approximately half the size of the point estimate. In addition, the extension of Medicare to the disabled arguably removed the most seriously chronically ill from the private insurance pool, and removed them completely; the people who would be covered by most proposed policies are likely to be less costly over the long run than SSDI recipients, and are likely to be subsidized less than fully. Future work might seek to account for these differences and use these estimates to simulate the effects of programs that are similar but not identical to the extension of Social Security to the disabled.