سیاست عمومی و بیمه سلامت انتخاب ها از سالمندان: شواهد از خرید برنامه درمانی

This study provides evidence on health insurance decisions of senior citizens. Nearly all senior citizens have health insurance coverage through Medicare, but poor seniors may also qualify for Medicaid that fills many gaps in Medicare coverage. Since 1987, the Medicaid program has expanded eligibility. Using the SIPP, I find that Medicaid eligibility increased from 8.7% in 1987 to 12.4% in 1995. For every 100 elderly who became eligible, approximately 50 took up Medicaid, but more than 30 dropped private coverage, resulting in crowd-out of 60%. Crowd-out came from individuals dropping privately purchased health insurance rather than dropping employer-provided retiree health insurance. The roles of asset tests, health status, and the panel structure of the SIPP are also explored. I find that a major strength of the SIPP is in its point-in-time asset information for determining Medicaid eligibility, while the changing income and demographic information over the course of the 2-year panel adds little insight beyond cross-sectional data.

The US federal government spent more than $352 billion on entitlements for the elderly in 1990.1 Although 80% of the money went to two programs — Social Security and Medicare — a significant amount was also spent on means-tested welfare programs, such as health insurance through Medicaid, cash assistance through Supplemental Security Income (SSI), food stamps, public housing, and energy assistance. Welfare programs for the elderly do not receive as much attention as those for the young, but combined federal spending on elderly SSI and Medicaid recipients amounted to $11.7 billion in 1990, approximately 54% of the amount spent on cash assistance and health insurance for younger households on AFDC.2 An aim of this study is to provide evidence on the economic behavior of senior citizens with respect to the largest of these means-tested programs, Medicaid. Nearly all senior citizens have health insurance coverage through Medicare, but poor seniors may also be eligible for Medicaid, which fills many gaps in Medicare coverage and offers first-dollar coverage. During the past decade, the Medicaid program has undergone a series of changes relating to eligibility. In particular, two new categories of elderly Medicaid recipients, known as Qualified Medicare Beneficiaries (QMBs) and Specified Low-Income Medicare Beneficiaries (SLMBs), were created. The income and asset limits to qualify under these programs were less strict than the limits under existing Medicaid categories, and 1.9 million senior citizens were enrolled in the QMB program in 1993.3 My particular focus will be on two issues relating to the QMB (and SLMB) expansions. First, how much did the QMB expansions increase Medicaid eligibility? The QMB expansions added to an existing and confusing patchwork of Medicaid rules that varied across states — in states where Medicaid was already generous the QMB expansions may not have made many individuals newly eligible. Second, how did increases in Medicaid eligibility affect supplemental insurance coverage? To address this, I estimate the effects of Medicaid eligibility on Medicaid coverage, private insurance coverage, and total supplemental insurance coverage. In addressing these questions, this study makes two primary contributions. The first contribution is adding evidence to the growing literature on government provision of health insurance and crowd-out of private insurance through a conceptually clean example. Although a recent and controversial literature has addressed the magnitude of the effect of Medicaid expansions for pregnant women and children on private insurance coverage, there are two real-world problems associated with the structure of employer-provided health insurance and the timing of Medicaid participation that make crowd-out estimates among the young inherently difficult.4 First, employer-provided health insurance is usually sold only to individuals or to entire families without gradations among types of dependents. Thus, a family that wants to cover both parents, but not the children (because the children qualify for Medicaid) may find it impossible to do so with only one employer health insurance policy. As Cutler and Gruber (1996a) explain, this lack of distinction may increase or decrease crowd-out compared with the case where a policy covered only individuals, and could conceivably result in crowd-out estimates of more than 100%. Second, the Medicaid expansions for younger groups provide no immediate benefit unless the family uses health care services — thus many families may wait until their child gets sick to enroll.5 Both Cutler and Gruber (1996a) and Currie and Gruber (1996a) find very low Medicaid take-up rates — for every 100 children made eligible for Medicaid, approximately 25 took it up. If Medicaid take-up is reduced because it provides little immediate benefit, then crowd-out estimates will be overestimated. Both complications mean that the appropriate benchmark for judging the economic importance of crowd-out is not obvious. These two problems are avoided in the context of the elderly. First, the QMB expansions provided immediate benefits because they paid for Medicare premiums (worth $550 per year). Second, the distinction between individuals and families is irrelevant because both elderly family members would be covered by QMB. The second contribution is data-oriented. To estimate the effects of QMB, I use longitudinal data from the Survey of Income and Program Participation (SIPP) covering the years 1987–1995. The SIPP provides detailed questions on health insurance coverage and the ingredients of Medicaid eligibility including liquid assets, life insurance, the value of the car, and out-of-pocket medical expenses. I use these questions to impute Medicaid eligibility incorporating all the major categories of elderly Medicaid recipients. These additional variables, which are largely unavailable in the Current Population Survey (CPS), make a major difference to the estimates. The main findings can be summarized as follows. First, Medicaid eligibility increased dramatically, from 8.7% in 1987 to 12.4% in 1995. Second, the expansions for the elderly resulted in dramatically higher Medicaid take-up rates than similar expansions for children. For every 100 elderly who became eligible, approximately 50 took it up. More than 30 elderly dropped private coverage, however, resulting in crowd-out of 60%. Third, crowd-out appears to come from individuals dropping privately purchased health insurance rather than dropping employer-provided retiree health insurance. Finally, the wide array of variables in the SIPP topical modules results in much less measurement error for computing Medicaid eligibility than if these variables were ignored. For example, Medicaid take-up results that ignore the Medicaid asset tests are less than half as large. I also find that the short length of the SIPP does not allow for many insights from its panel structure. The remainder of the paper is arranged into four sections. Section 2 describes the supplemental health care choices facing the elderly. Particular attention is paid to key features of the Medigap and Medicaid policies. The section also presents some basic numbers and magnitudes of Medicaid participation. Section 3 describes the data construction and identification issues. Section 4 presents the results and robustness checks. Section 5 concludes.

This study has examined the consequences of Medicaid expansions for the elderly. The primary results show that take-up rates for the expansions were around 50%, but more than half of those who took up Medicaid coverage also dropped private supplemental coverage. These results, then, provide a confirmation in a different setting of Cutler and Gruber’s (1996a) findings on crowd-out for pregnant women and children. Crowd-out was concentrated among the youngest of senior citizens, who are likely to find Medicaid coverage a better substitute for private supplemental coverage, and may be most responsive to different supplemental options at the time they become eligible for Medicare. Moreover, the wide array of variables in the SIPP data leads to more precise measures of eligibility than one could obtain in the CPS, and better estimates of take-up rates and crowd-out.