The Effects of Medicaid Expansions on Insurance Coverage of Children

Transcription

1 152 REVISITING THE ISSUES The Effects of Medicaid Expansions on Insurance Coverage of Children Lisa C. Dubay Genevieve M. Kenney Lisa C. Dubay, Sc.M., is senior research associate at the Health Policy Center at the Urban Institute, Washington, DC. Genevieve M. Kenney, Ph.D., is senior research associate at the Health Policy Center at the Urban Institute, Washington, DC. Editors Note In the Winter 1995 issue of The Future of Children, the Child Indicators article examined recent trends in health insurance coverage for children. A major conclusion of that article was that the proportion of children covered by private, employment-based health insurance has been declining in recent years while changes in the rules for eligibility for Medicaid have made that program increasingly important for children. Nearly eight million more children were covered by Medicaid in 1993 than in 1988, largely as a result of the eligibility expansions. However, the number of uninsured children grew by about one million during that same period. In this Revisiting the Issues article, Lisa Dubay and Genevieve Kenney of the Urban Institute look more closely at the effects of the changes in the eligibility rules for Medicaid on children s insurance coverage between 1988 and They examine the extent to which Medicaid is covering the expansion-targeted population, many of whom would have been otherwise uninsured, and also the extent to which increased levels of Medicaid coverage for children merely substituted for employer-sponsored health insurance. They find that, although the Medicaid expansion led to a substantial increase in Medicaid enrollment for children, many uninsured children who are eligible for Medicaid are not being enrolled. With regard to the voluntary substitution of Medicaid coverage for employer-sponsored coverage, the analysis suggests that only 17% of the increase in Medicaid coverage for children between 1988 and 1993 is attributable to this substitution. Estimated rates of substitution are much higher, however, for near-poor children than for poor children. This suggests that further raising Medicaid income eligibility thresholds could substantially reduce the number of uninsured children but only at the price of also reducing the number with employersponsored insurance. EML and LSB In response to declining health status indicators for low-income children and growing disparities in access to health care between the insured and the uninsured, 1 4 Medicaid coverage for young children was expanded in the late 1980s. Whereas Medicaid eligibility for children had been restricted to categorical eligibility based on eligibility for Aid to The Future of Children SPECIAL EDUCATION FOR STUDENTS WITH DISABILITIES Vol. 6 No. 1 Spring 1996

2 153 Families with Dependent Children (AFDC), new legislation expanded eligibility for children based on income alone, with income cutoffs higher than those specified for AFDC. Congress permitted and eventually mandated states to provide Medicaid coverage for children up to age six in families with incomes up to 133% of the federal level and to children born after September , with family incomes up to 100% of. It also gave states the option (starting in 1988) to cover infants with family incomes up to 185% of. 5 The hope was that these expansions would reduce the number of uninsured children and, thus, improve children s health. Between 1988 and 1993, the number of children receiving Medicaidcovered services grew by 53%. 6 Over the same period, however, employersponsored insurance coverage was declining, 7,8 and the number of uninsured children remained high (8.7 million in 1993). 9 This article examines several important questions about the impact of the Medicaid expansion. First, for which groups did Medicaid coverage actually expand? Second, did this expansion merely substitute for privatesector coverage rather than covering children who were previously uninsured? These questions are important because, to the extent that Medicaid fails to enroll the population targeted by the expansions or simply substitutes public for what had been private coverage, the expected, desired health effects may not occur. 10 Two other studies have recently examined the changing insurance status of children using data from the Current Population Survey (CPS). While these studies find similar trends, they differ substantially in their interpretation of the trends. One study, by Newacheck and colleagues, 11 attributes all of the change in Medicaid coverage to secular declines in employer-sponsored coverage; the other study, by Cutler and Gruber, 12 attributes a substantial portion of the change in Medicaid coverage to crowd-out but fails to control adequately for secular declines in employersponsored coverage. 13 This Revisiting the Issues article sheds more light on the issue of crowdout (the voluntary substitution of free health insurance coverage for children under Medicaid for employer-sponsored insurance when the terms of the employer-sponsored coverage have not changed substantially) by explicitly controlling for the secular declines in the offering, financing, and takeup of employer-sponsored insurance using data from the Current Population Survey to track changes between 1988 and Because these data are cross-sectional and do not follow specific families and children over

3 154 THE FUTURE OF CHILDREN SPRING 1996 time, it is not possible to be definitive about what caused the changes that were observed over this period. But it is possible to trace the extent to which trends in employment-based coverage for Medicaid-eligible children did or did not mirror trends for the ineligible population, an analysis which provides a preliminary assessment of the likely extent of any crowding out that might have occurred. Data Sources The Urban Institute s Transfer Income Model, Version 2 (TRIM2), a well-accepted microsimulation model of tax and transfer programs affecting individuals and families, is used for this analysis. 14 The database underlying TRIM2 is the March Current Population Survey, a nationally representative sample of the U.S. population. Used for this analysis were the 1989 and 1994 versions of the March CPS, which include questions regarding income levels and health insurance coverage for each member of the household in the previous year. In the Winter 1995 issue of The Future of Children, Lewit and Baker used TRIM2 data to analyze health insurance coverage of children under age 18 over the same period. 15 Three weaknesses in the CPS health insurance information deserve note. The first is that the CPS does not ask directly if a person is uninsured. Rather, estimates of the uninsured from the CPS reflect the number of persons who failed to say yes to any of the questions about specific types of health insurance. The second weakness is that, although the survey asks for coverage during the previous year, 16 comparisons with other surveys indicate that some respondents are, in fact, providing information about their current insurance status. 17 Third, the CPS probably underreports program participation, including participation in Medicaid, because CPS participation estimates are considerably lower than the number of participants reflected in program data. Using the TRIM2 edited version of the CPS rather than the CPS itself is important because it mitigates some of these problems by using the CPS health insurance questions to create individual indicators and by correcting for underreporting. In addition, TRIM2 identifies each Medicaid-eligible person and the person s eligibility category, even for persons who do not report Medicaid coverage. 18 Health Insurance Coverage of Children: 1993 In 1993, some 24% of all children 18 years and under had Medicaid coverage only. Fifty-two percent had employer-sponsored insurance only. An additional 6% had both Medicaid and employer-sponsored coverage, 19 and 13% had no health insurance. Medicaid coverage of children varies greatly by family income, age, and race. In 1993, Medicaid provided coverage at some point during the year for almost 80% of all poor children and more than 50% of children with incomes from 100% to 133% of. In 1993, Medicaid provided coverage for 45% of all infants and 40% of all children ages one through five. Medicaid played a smaller role for older children, covering 25% of all 6- to 10-year-olds and 19% of all 11- to 18-year-olds in Finally, 23% of all white children, 54% of all black children, and 31% of all children of other races were covered by the Medicaid program in To What Extent Is Medicaid Covering the Expansion Target Population? Medicaid participation rates (the proportion of eligibles who were enrolled in the program) in 1993 were much higher for the traditionally eligible population (whose eligibility is tied to welfare participation) than for the expansion-eligible population (those eligible based on the higher incomeeligibility cutoffs). The Medicaid participation rate among children whose Medicaid eligibility was tied to welfare participation was 90%. In comparison, the proportion of expansion-eligible children without employer-sponsored coverage who enrolled in Medicaid was 69%. The proportion of expansion-eligible children who report coverage by both employer-sponsored insurance and Medicaid in 1993 was 51%. 19 Whether the lower participation rates for

4 Revisiting the Issues: The Effects of Medicaid Expansions on Insurance Coverage of Children 155 Table 1 Changes in Insurance Coverage for Children 18 Years and Under Between 1988 and 1993 Year Percentage of Children According to Health Insurance Status Employer-Sponsored Medicaid Employer-Sponsored Only Only and Medicaid Uninsured Change -9.3* 7.3* 2.1* 0.4* * The difference between 1988 and 1993 is statistically significant at the 0.05 level. the expansion population are the result of lack of knowledge about the new eligibility rules, an unwillingness to enroll in Medicaid, or persisting problems with the Medicaid eligibility determination process is unclear. 20 More than a quarter of the children who remained uninsured all year were eligible for Medicaid, a proportion that varied considerably by age. Forty-five percent of uninsured children five years old and under were eligible for Medicaid, compared with 30% and 15% of uninsured 6- to 10-year-olds and 11- to 18-year-olds, respectively. The lower Medicaid eligibility rates for uninsured older children are not particularly surprising, given the very low income-cutoff levels for AFDC-based Medicaid eligibility and the exclusion of most older children from the higher income-eligibility provisions over the analysis period. 12 Did the Medicaid Expansions Crowd Out Employer-Sponsored Insurance? Between 1988 and 1993, the proportion of all children with Medicaid coverage increased by only seven percentage points, and the percentage with both Medicaid coverage and employer-sponsored coverage increased by two percentage points. The proportion with employer-sponsored coverage declined by only nine percentage points. The proportion of uninsured children increased by almost half a percentage point nearly one million children (see Table 1). That these increases in Medicaid coverage occurred along with large declines in employer-sponsored coverage makes it plausible that Medicaid did crowd out some employer-sponsored coverage as has been recently argued by David Cutler and Jonathan Gruber. 12 However, at the same time the Medicaid expansions were being implemented, changes in the employer-sponsored insurance market were occurring. Specifically, the share of premiums being borne by employees was increasing, and employers were reducing their offers for and financing of insurance coverage, especially for dependents. 8 In addition, the period was marked by a recession with an accompanying increase in unemployment and, hence, possible loss of employer-sponsored health insurance coverage. These factors were likely to result in declines in employer-sponsored insurance for children separate from the expansions. Children who lost their employer-sponsored coverage for these reasons either enrolled in Medicaid or became uninsured. In this analysis, the decline in employer-sponsored coverage of children that is attributable to the secular declines in coverage is separated from the decline that is attributable to the expansions, and only the latter is considered to be crowd-out. Once that part of the decline in employersponsored coverage that is the result of factors beyond the secular declines had been identified, total crowd-out effect is estimated. Total crowd-out effect is defined as the percent of the total increase in Medicaid coverage of children, which is the result of

5 156 THE FUTURE OF CHILDREN SPRING 1996 Table 2 Insurance Coverage for Children Under Age 11 by Income Group Between 1988 and 1993 Year Percentage of Children According to Health Insurance Status Employer-Sponsored Medicaid Employer-Sponsored Only Only and Medicaid Uninsured 1988 Under 100% of % to 133% of Under 100% of % to 133% of Percentage Change Under 100% of -2.3* 10.3* % to 133% of -21.3* 20.4* -14.4* -9.6* * The difference between 1988 and 1993 is statistically significant at the 0.05 level. the decline in employer-sponsored coverage related to the expansions (in other words, not related to secular declines). To assess the extent of crowding-out, changes in insurance coverage for poor children (those with family incomes up to 100% of ) and near-poor children (those with family incomes from 100% to 133% of ), the two groups most affected by the expansions, are examined separately (see Table 2). The crowd-out analysis is limited to children under age 11 because these are the children who were affected by the expansions. To control for secular trends in health insurance coverage, the experience of these children is compared with the experience of a group of individuals not affected by the expansions men ages 18 to For poor children (those in families with incomes below the federal line), the proportion reporting only Medicaid coverage increased by 10 percentage points and the proportion reporting only employersponsored coverage decreased by more than two percentage points. Thus, the share of poor children who were uninsured fell by almost eight percentage points. This pattern suggests that the expansions resulted in a decline in the number of uninsured children in and that there was little crowding out of employer-sponsored insurance for this group. Given the dramatic drop in the share of these children who were uninsured, the expansions in Medicaid appear to have provided a safety net for children living in. For near-poor children (those with family incomes from 100% to 133% of the federal line), the trends are different. There was a large (21-percentage-point) decline in children with only employersponsored coverage and a commensurate increase in those with only Medicaid coverage. The share of near-poor children with both Medicaid and employer-sponsored coverage also increased substantially. The share of near-poor children who were uninsured declined by 10 percentage points. These

6 Revisiting the Issues: The Effects of Medicaid Expansions on Insurance Coverage of Children 157 Figure 1 Percentage Point Change in Insurance Coverage Experienced by Those with Incomes Between 0% and 99% of Poverty Between 1988 and Percentage Point Change Employer-Sponsored Medicaid Only Employer-Sponsored Uninsured Only and Medicaid Children under age 11 Males ages 18 to 44 trends suggest not only that Medicaid may, indeed, have crowded out employersponsored coverage for near-poor children, but also that the expansions reduced the share of children in these income and age groups lacking insurance coverage. Although plausible, the crowding-out hypothesis is not the only possible explanation for the observed trend in private insurance coverage for near-poor children. As mentioned earlier, changes in the private insurance market over the same period rising premiums and increased cost sharing by employees could have accounted for reduced private coverage, which would not be considered to be crowding out. To further identify the reasons for reduced employer-sponsored insurance among poor and near-poor children, the trend in insurance coverage for men ages 18 to a group unaffected by the Medicaid expansions was compared with the trends for children under age 11 in the same income groups (see Figures 1 and 2). Between 1988 and 1993, the proportion of men ages 18 to 44 with family incomes under 100% of with employersponsored coverage declined by two percentage points, and the share who were uninsured declined by three percentage points. For men ages 18 to 44 with family incomes between 100% and 133% of, employer-sponsored coverage declined by eight percentage points. The proportion without insurance increased by five percentage points. The larger declines in employersponsored coverage for children relative to men in the same income group suggests that some crowding out of employer-sponsored insurance did occur. But the fact that employer-sponsored coverage for men declined at all suggests that some of the switching from employer-sponsored coverage to Medicaid for children was unrelated to the expansions. How much of the decline in employersponsored coverage can be attributed to crowding out? To account for changes that

7 158 THE FUTURE OF CHILDREN SPRING 1996 Figure 2 Percentage Point Change in Insurance Coverage Experienced by Those with Incomes Between 100% and 133% of Poverty Between 1988 and 1993 Percentage Point Change Employer-Sponsored Medicaid Only Employer-Sponsored Uninsured Only and Medicaid Children under age 11 Males ages 18 to 44 were occurring in the provision and takeup of employer-sponsored coverage irrespective of Medicaid changes, controlling for income, it is assumed that the trends for men are predictive of what would have occurred for children in the absence of the expansions. To account for the possibility that some portion of the increase in nearpoor children reporting both Medicaid and employer-sponsored coverage is the result of crowd-out, it is assumed that half of this increase is attributable to the substitution of Medicaid for employer-sponsored coverage (crowd-out) and that half is attributable to other factors. The total crowd-out effect is estimated in four steps. First, the percentage point decline in employer-sponsored coverage attributable to crowd-out is estimated. To do this the percentage point decline in employersponsored coverage for men is subtracted from the percentage point decline in employer-sponsored coverage of children for each income group (poor and near poor) to control for the secular declines. 22 Second, half of the percentage point increase in children reporting both Medicaid and employersponsored coverage is subtracted to account for crowding out in this group. 23 These calculations produce an estimate of the percentage point decline in employer-sponsored coverage that is attributable to crowd-out. It is estimated that, of the decline in employer-sponsored coverage for the near poor and poor, six percentage points for the near poor and one percentage point for the poor are attributable to crowd-out. Third, the percentage point decline in employer-sponsored coverage attributable to crowd-out obtained in the first two steps is divided by the percentage point increase in Medicaid, again separately by income group. 24 The third step produces estimates of the proportion of the increase in Medicaid coverage of children attributable to the crowding out of employer-sponsored coverage for the poor and near-poor groups. Based on these calculations, it is estimated

8 Revisiting the Issues: The Effects of Medicaid Expansions on Insurance Coverage of Children 159 that 22% and 15% of the increase in Medicaid coverage for the near poor and poor, respectively, is attributable to crowdout. Finally, an average across the two income groups is estimated using weights equal to the proportion of the increase in Medicaid coverage accounted for by each income group to obtain the total crowd-out estimate. 25 Thus, the total crowd-out effect is defined as the proportion of the increase in Medicaid coverage of children that is attributable to the crowding out of employer-sponsored coverage. Using this method, it is estimated that 17% of the total increase in Medicaid enrollment, which occurred between 1988 and 1993 for young children with incomes below 133% of, is attributable to the crowding out of employer-sponsored coverage. When it is assumed that all of the increase in those reporting both Medicaid and employer-sponsored coverage is attributable to crowd-out, then an upper bound on the total crowd-out effect is estimated to be 26%. 26 Conclusion Overall, the take-up rates among children eligible for Medicaid are fairly high, and there is evidence that the Medicaid expansions for children resulted in only a small rate of substitution of Medicaid for private employer-sponsored insurance coverage. The participation rates among children eligible under the expansions who lack employer-sponsored coverage were around 70%. This is significantly lower than participation rates for children eligible for AFDC, however, indicating that more intensive outreach may be needed to further reduce the number of young children lacking insurance coverage. Only 17% of the increase in the Medicaid coverage of children between 1988 and 1993 is attributable to crowding out. 27 Crowding out is more of an issue for near-poor children than for poor children. This suggests that raising eligibility thresholds beyond the federal level could significantly reduce the number of children who lack insurance coverage but only at the price of also reducing the number with employer-sponsored coverage. Together these findings have mixed implications for the impacts of the Medicaid eligibility expansions on children s health. First, while the expansions led to substantial increases in Medicaid enrollment for children, the program is still not reaching a significant portion of the target population. Second, the apparent shift from employersponsored to Medicaid coverage, resulting from both secular declines in and crowding out of employer-sponsored coverage for the near poor, means that Medicaid is substituting to some extent for what would have been private insurance coverage in prior years. Because other research has demonstrated access problems under the Medicaid program, 28 this could mean less access to care for some children. At the same time, lowincome children who switch from private to Medicaid coverage may experience greater access to preventive health care under Medicaid. Despite these issues, the expansions likely protected large numbers of lowincome children who would have lost their employer-sponsored insurance coverage and become uninsured during this period, and may have staved off the resultant declines in child health. Finally, and most disturbingly, despite the dramatic Medicaid expansions for children, the number of uninsured children grew by almost one million between 1988 and This suggests that many more children would have lacked insurance coverage without the expansions, reflecting a transformation of the health insurance market that leaves more and more children with state Medicaid programs as their primary potential source of health care coverage. 1. Kasper, J.D. The importance of type of usual source of care for children s physician access and expenditures. Medical Care (1987) 25,5: Rosenbach, M.L. Insurance coverage and ambulatory medical care of low-income children: United States, Reports from the National Medical Care Utilization and Expenditure Survey. Series C, Analytical Report No. 1. Washington, DC: U.S. Department of Health and Human Services, Public Health Service, National Center for Health Statistics, 1985.

9 160 THE FUTURE OF CHILDREN SPRING Short, P.F., and Leftkowitz, D.C. Encouraging preventive services for low-income children: The effect of expanding Medicaid. Paper presented at the 1991 Annual Meeting of the Association for Health Services Research. San Diego, CA, Leftkowitz, D.C., and Short, P.F. Medicaid eligibility and the use of preventive services by low-income children. Paper presented at the 1989 Annual Meeting of the American Public Health Association. Chicago, IL, For a more detailed discussion of state initiatives, see Hill, I.T., Bartlett, L., and Bostrom, M.B. State initiatives to cover uninsured children. The Future of Children (Summer/Fall 1993) 3,2:142 63; DeGraw, C., Park, E.J., and Hudman, J.A. Revisiting the issues: State initiatives to provide medical coverage for uninsured children. The Future of Children (Spring 1995) 5,1:223 31; or Lewit, E.M., and Baker, L.S. Child indicators: Health insurance coverage. The Future of Children (Winter 1995) 5,3: Unpublished tabulations of the Health Care Financing Administration, Form 2082 data. 7. Holahan, J., Winterbottom, C., and Rajan, S. The changing composition of health insurance coverage in the United States. Health Affairs (Winter 1995) 14,4: KPMG Peat Marwick. Health Benefits in Newark, NJ:KPMG Peat Marwick, See note no. 5, Lewit and Baker. 10. This statement is not meant to imply that Medicaid and private coverage can be equated. On the one hand, Medicaid, in contrast to many employer-sponsored plans, covers all preventive care visits and does not require the deductibles, which may inhibit some lowincome families from obtaining necessary preventive or illness-related care. Therefore, some children may be better off with Medicaid than with employer-sponsored coverage. On the other hand, the well-documented access and quality problems within the Medicaid program may present significant barriers to obtaining appropriate care for some children. Therefore, substituting Medicaid for private insurance coverage could lead to better or worse access, depending on the characteristics of both the employer-sponsored coverage and the Medicaid program. 11. Newacheck, P., Hughes, D., and Cisternas, M. Children and health insurance: An overview of recent trends. Health Affairs (Spring 1995) 14,1: Cutler, D., and Gruber, J. Does public insurance crowd out private insurance? Working Paper No Washington, DC: National Bureau of Economic Research, For a more complete critique of this paper, see Dubay, L. and Kenney, G. Did the Medicaid expansions crowd out private coverage? Working paper no Washington, DC: Urban Institute, For a complete discussion of TRIM2, see Giannarelli, L. An analyst s guide to TRIM2. Washington, DC: Urban Institute Press, Lewit and Baker (see note no. 5) examine changes in insurance coverage for children under 18. Eighteen-year-olds are included in this analysis because, at this age, they still may be covered under Medicaid as dependent children and under private insurance policies as dependents. 16. This practice allows individuals to report more than one type of coverage. 17. Swartz, K. Interpreting the estimates from four national surveys of the number of people without health insurance. Journal of Economic and Social Measurement (1986) 14:233 42; U.S. Department of Commerce, Bureau of the Census. Health insurance coverage: Current Population Reports, Series. P 70, No. 17. Washington, DC: U.S. Government Printing Office, 1990; Winterbottom, C., Liska D., and Obermaier, K. Statelevel data book on health care access and financing. Washington, DC: Urban Institute Press, The term Medicaid-eligible is used in this paper to identify children who met all the eligibility requirements for Medicaid regardless of whether they actually received any Medicaid services. This is in contrast to the use of the term Medicaid-eligible by the Health Care Financing Administration to signify individuals who are actually enrolled in the program. 19. The Current Population Survey does not distinguish between children who have both Medicaid and private insurance coverage simultaneously in a year and those who have these sources of coverage at different times during a year. See the Child Indicators article by Lewit and Baker in the Winter 1995 issue of The Future of Children for a description of how children s insurance status is determined in the CPS and TRIM2.

10 Revisiting the Issues: The Effects of Medicaid Expansions on Insurance Coverage of Children Children appear to enroll in Medicaid at much higher rates than pregnant women despite the effort to streamline the eligibility process for pregnant women. (See note no. 13, Dubay and Kenney.) 21. Consideration was given to using women of childbearing years and older children as other comparison groups. Women of childbearing years were rejected because their experience over this period could have been influenced by the expansions for pregnant women. Older children were rejected because their experience might have been affected by younger siblings eligibility for Medicaid. 22. For the near poor, this calculation is ( ), or 13.18; for the poor, it is ( ), or For the near poor, this calculation is (13.18-(0.5 *14.42)), or 5.97; for the poor, it is ( (0.5 * 0.1)), or For the near poor, this calculation is (5.97/( (0.5 * 14.42))), or 0.216; for the poor, it is (1.06/( (0.05 * -0.1))), or The weights are for the near poor and for the poor. The calculation is then ((0.216 * 0.269) + (1.54 *0.731)), or This upper bound probably represents an overestimate of the shift in financing because some of these children are moving from Medicaid to employer coverage and some are maintaining both types of coverage for some portion of the year. Clearly, more research is needed to understand the insurance coverage patterns of these children. 27. This crowd-out estimate (and the upper-bound estimate) are less than that reported by Cutler and Gruber (see note no. 12), who estimate the total effect of crowding out for children to be between 31% and 49%. These estimates are not readily comparable because the estimates in this article are based on children under age 11 while those reported by Cutler and Gruber are based on all children. Furthermore, this article focuses on children with incomes below 133% of while Cutler and Gruber examine all income groups. Although there are many differences between the work of Dubay and Kenney and that of Cutler and Gruber, the key difference is that Cutler and Gruber fail to account for the secular decline in employer-sponsored coverage, which causes their estimates to overstate crowd-out. When controls for secular trends are omitted, the resulting estimates are similar to those of Cutler and Gruber. For a more complete discussion of these issues, see note no. 13, Dubay and Kenney. 28. Gavin, N. I. Review and synthesis of the literature on financial barriers to health care services for children. Report prepared under the Evaluation of the Medicare Catastrophic Coverage Act. HCFA Contract No Abt Associates, Inc. Washington, DC: Systemetrics, 1991.