اثر برنامه بیمه سلامتی کودکان دولت در پوشش بیمه سلامتی

This paper presents the first national estimates of the effects of the SCHIP expansions on insurance coverage. Using CPS data on insurance coverage during the years 1996–2000, we estimate instrumental variables regressions of insurance coverage. Our regression results imply that 9% of children meeting income eligibility standards for SCHIP gained public insurance. While low, our estimates indicate that states were more successful in enrolling children in SCHIP than they were with prior Medicaid expansions that were focused on children just above the poverty line. Crowd-out of private health insurance was estimated to be nearly 50%, which is in line with estimates for the Medicaid expansions of the early 1990s. In addition, state anti-crowd-out provisions in the form of waiting periods were found to significantly affect both take-up and crowd-out.

The State Children’s Health Insurance Program (SCHIP) was signed into law as part of Title XXI of the 1997 Balanced Budget Act. The goal of the legislation was to increase the insurance coverage of children in the United States by extending eligibility for public insurance to children in families earning too much to qualify for Medicaid yet earning too little to afford private health insurance. Touted as the largest expansion in health insurance since the enactment of Medicaid in 1965, the SCHIP legislation apportioned more than $40 billion in federal matching funds over 10 years beginning in fiscal year 1998. States are allowed to use these funds to expand Medicaid eligibility, to develop new insurance programs, and increase outreach for children already eligible for public coverage. The efficacy and cost of public insurance expansions depend critically on two factors: program take-up and the possible substitution from private to public coverage, i.e., “crowd-out”. A number of recent studies find low take-up rates for individuals who became eligible for public insurance through the Medicaid expansions of the late 1980s and early 1990s (Cutler and Gruber, 1996, Dubay and Kenney, 1997, Shore-Sheppard, 1997, Yazici and Kaestner, 2000, Blumberg et al., 2000, Ham and Shore-Sheppard, 2001 and Card and Shore-Sheppard, 2003).1 Previous studies also suggest that the Medicaid expansions contributed to a decline in private insurance, though the estimated magnitude of this effect varies considerably. Low take-up might be an even greater problem for SCHIP as many of the families made eligible for the program had little prior experience with public insurance programs and, therefore, may lack good information about the program. Negative perceptions of the quality of public health insurance and stigma related to public assistance also represent potentially serious barriers to enrollment. Evidence from the Medicaid expansions suggests that take-up falls as coverage is extended to relatively higher income families. For example, Currie and Gruber (1996) find that take-up was higher among women who gained eligibility from expansions that were targeted at families in poverty than for those who gained coverage from broader expansions. Similarly, Card and Shore-Sheppard (2003) find that expansions targeting children under the federal poverty line (FPL) led to a 10–15% increase in Medicaid coverage, whereas legislation extending eligibility to children with incomes up to 133% of the FPL had essentially no effect. Since private insurance coverage increases with income, crowd-out is potentially a greater problem for SCHIP than for prior expansions of public insurance. However, in response to research showing large crowd-out effects for Medicaid, SCHIP programs were designed with explicit mechanisms intended to keep newly eligible families from dropping private coverage in favor of public insurance. Thus, whether crowd-out is more or less of a problem in the case of SCHIP as compared to the earlier Medicaid expansions is an empirical question.2 Because the program is relatively new, research on the effects of SCHIP is quite limited. Two recent studies present evidence on trends in insurance coverage for children in the early years of SCHIP implementation (Zuckerman et al., 2001 and Rosenbach et al., 2001). They show that the percent insured fell slightly for children in families with incomes above 200% of FPL, while remaining relatively constant for lower income children who were more likely to be affected by SCHIP. Since coverage had been falling previously for children in the SCHIP target population (incomes between 100 and 200% of the FPL), this suggests a positive, if small, effect of the program. Neither study attempts to calculate rates of take-up or crowd-out. Kronick and Gilmer (2001) study the effect of four state-level programs implemented prior to SCHIP on the insurance coverage of adults. They estimate that these programs were successful in enrolling between 6 and 24% of the target population. As expected, Kronick and Gilmer find greater evidence of crowd-out for adults with incomes above the FPL as compared to those below. Our study is the first to estimate the effect of SCHIP eligibility on public and private insurance coverage of children. We use 1997–2001 data from Current Population Survey (CPS) and econometric techniques similar to those used in previous research on the Medicaid expansions (e.g., Currie and Gruber, 1996, Cutler and Gruber, 1996 and Ham and Shore-Sheppard, 2001). Specifically, we use an instrumental variable approach that relies on variation across and within states in the timing and extent of coverage expansions to identify the effect of SCHIP on insurance coverage. In addition to estimating the overall effect of SCHIP eligibility on coverage by public insurance (take-up), private insurance (crowd-out) and uninsurance, we test for differences between states that implemented SCHIP by expanding Medicaid eligibility and those that established new stand-alone programs. We also test for the impact of the most common anti-crowd-out provisions implemented by states: waiting periods requiring that children be without private insurance for a length of time before enrolling in SCHIP. We find that the enactment of SCHIP led to a small but statistically significant reduction in the rate of uninsurance for children. The take-up rate implied by our results is lower than comparable estimates for the Medicaid expansions of the 1980s and 1990s as a whole, though larger than estimates pertaining to Medicaid expansions that, like SCHIP, were targeted at children above the poverty line. We find some differences related to the way states implemented the program. The requirement that children be uninsured for a period of time before being eligible for SCHIP appears to be effective at reducing crowd-out, though at the cost of lower take-up of public insurance. We also find weak evidence that take-up was higher in states that established new stand-alone programs compared to states that expanded their existing Medicaid program, though the differences between program types are not precisely estimated. Estimation of take-up and crowd-out effects in our data are complicated by the fact that some public health insurance plans, particularly newly created non-Medicaid programs, could be mistaken by CPS respondents for private health insurance plans. Indeed, our results provide some evidence that is suggestive of such errors in reporting. When we treat the CPS insurance data as accurate, we obtain small and statistically insignificant estimates of crowd-out. In contrast, models that account for potential misreporting suggest that crowd-out represents nearly one-half of the increase in public coverage.

This paper presents the first national estimates of the effects of the SCHIP expansions on insurance coverage. Several key findings emerge from our analysis. First, we find that SCHIP had a small, but statistically significant, positive effect on insurance coverage. Our regression results imply that approximately 9% of children meeting income eligibility standards for the new program enrolled in public insurance. Take-up rates calculated for children who were both income-eligible and previously uninsured are considerably higher and are in line with estimated take-up rates for the Medicaid expansions of the 1980s and early 1990s. The take-up rate based on the uninsured is probably too large, though, as is implicitly assumes that all new SCHIP enrollment came from the ranks of the uninsured, which our crowd-out results suggest was not the case. Nonetheless, our estimates based only on income eligibility rules indicate that states were more successful in enrolling children in SCHIP than they were with prior Medicaid expansions focused on children just above the poverty line. One possible explanation for SCHIP’s greater success relative to earlier Medicaid expansions is that newly established SCHIP expansion programs were viewed more favorably by low income families than Medicaid. However, when we distinguish between states that implemented SCHIP by expanding Medicaid and those that started new programs, we find limited support at best for this argument. Our point estimates indicate that eligibility for stand-alone SCHIP programs has a slightly more negative effect on the probability of being uninsured, but the differences between the two types of programs are not statistically significant. An alternative explanation is that as states enacted SCHIP they devoted considerable resources to outreach and marketing, and streamlined many enrollment procedures (US GAO, 2000). It is not possible with our data to assess the impact of these efforts or to compare the strategies taken by different states. A recent study by Aizer (2002) finds that outreach through community-based organizations and a television advertising campaign increased Medicaid take-up in California. More research in this area would be valuable. One factor limiting SCHIP enrollment is the various policies that were enacted to prevent crowd-out, such as waiting periods and premium contributions. We test for the impact of the most readily available and widely used means of limiting crowd-out: the waiting period. We find that waiting periods had a statistically significant effect on both take-up and crowd-out. Specifically, if no state enacted a waiting period, our results indicate that the take-up rate would have increased to 13%, implying more than 400,000 additional children enrolled, with a relatively small rise in the predicted rate of crowd-out to just over 50%. By contrast a waiting period of 5 months essentially eliminates crowd-out, but a waiting period of this length is predicted to reduce the take-up rate to 3.4%. Finally, our results point to a serious data problem facing researchers in this area. State efforts to make public insurance programs more attractive to eligible families and increased contracting with private health plans has blurred the line between public and private health insurance. The finding of a positive relationship between SCHIP eligibility and private non-group coverage suggests that some CPS respondents whose children are insured through SCHIP report that coverage as private insurance. This ambiguity in distinguishing private and public coverage makes it difficult to use these data to track changes in type of insurance and accurately estimate rates of take-up and crowd-out.