|
Medicaid For Children: Federal Mandates, Welfare Reform, And Policy Backsliding
The 1996 federal welfare reform law delinked Medicaid enrollment from welfare participation. This paper estimates the impact of welfare reform on children’s Medicaid enrollment using a methodology that both adjusts for income and other demographic differences over time and across states, and provides income-specific estimates of enrollment. The results indicate large enrollment declines: Between 1995 and 1998, enrollment probabilities for children in families with no income declined from 81 percent to 68 percent, while at half the poverty line, the decline was from 61 percent to 53 percent. This implies that 926,000 to 1.37 million fewer children were enrolled after welfare reform. At the state level, Medicaid declines and welfare reform were strongly associated, with only a few states succeeding in preserving children’s Medicaid coverage.
During the 1980s the federal government created options and imposed mandates on the states to expand Medicaid eligibility for children and pregnant women in low-income families. For children, the most important mandates were enacted in the Omnibus Budget Reconciliation Acts (OBRAs) of 1989 and 1990, which required that states extend Medicaid eligibility to all children under age six with family incomes below 133 percent of the federal poverty level, as well as children born after 30 September 1983 with family incomes below 100 percent of the poverty line. In 1996 Congress enacted and President Bill Clinton signed the welfare reform law that repealed Aid to Families with Dependent Children (AFDC) and replaced it with Temporary Assistance for Needy Families (TANF) block grants to the states. While early Republican-sponsored bills had included Medicaid as part of these welfare block grants, this alternative was rejected. Instead, Sen. John Chafee (R-RI) and Sen. John Breaux (D-LA) modified the welfare reform legislation to delink Medicaid from the welfare programs funded by the new block grants. Section 1931 of the welfare reform law not only preserved the Medicaid program as a federal/state entitlement but explicitly maintained for the Medicaid program both the pre-welfare reform, AFDC-related eligibility rules for adults in families with dependent children and the implementation schedule for the child-related expansions established in 1989–1990. This paper uses data from the Current Population Survey (CPS) for 1979–1998, including two years of data collected for the years after enactment of welfare reform. With these post-welfare reform data we can examine whether Medicaid coverage for low-income children has been preserved as intended. The methods used here, applied to the large sample sizes of the CPS compiled over two decades, permit a longitudinal analysis of enrollment trends at different income levels and for individual states. This level of disaggregation is important because recent policy trends have had very different impacts on families with no or extremely low incomes compared with those at or above the poverty level. Moreover, these enrollment patterns vary across states, with only a few states successfully preserving Medicaid coverage.
The goal of this analysis is to understand the enrollment impacts of Medicaid policy changes that take place over time and may differ across states. However, observable enrollment levels are influenced by factors other than policy, especially changes in potential recipients’ needs for safety-net support—needs that are related to a family’s income and that fluctuate over time as economic conditions change. When analyzing enrollment across states, we want to correct for the economic situations of potential recipients so that the results are comparable. For example, in some states the poverty rate is higher than the national average, and so it would not be surprising to see higher-than-average welfare participation rates in those states. Similarly, we want to correct for the changing economic situation over time so that the results from the 1990s are comparable to those from the 1980s. To do this, I have analyzed Medicaid (and AFDC/TANF) enrollment using a two-step estimation strategy that models individual and family decisions to enroll in programs and calculates a standardized measure of state-level policy outcomes.1 In the first step the data are divided by state and time period (to have a sufficient number of observations for small states, three annual surveys are pooled together: 1980–82, 1983–85, and so on). Using these data I estimate logit models for each state that control for income, income squared, race, age, sex, and eligibility for the mandated coverage. The income variable used here is "pre-welfare" family income, which is transformed by dividing by the poverty level for each family’s size. This corrects for varying sizes of families, and because the poverty level is annually adjusted for inflation, this also provides a consistent measure of income over time. Race is coded as a dichotomous variable distinguishing whites from nonwhites, and the model includes the age and sex of the potential recipient child. Because the mandates were phased in over time, the model includes an age/income interaction term, which is a dichotomous variable set equal to one when a child in a particular year falls within the age and income cutoffs established by the federal mandates. The estimated models include separate intercepts for the different years and year-specific slope coefficients for the age/income interaction, which permits the computation of year-specific predicted probabilities. The data used here are drawn from the March CPS supplement, which includes questions on receipt of benefits from a number of government programs, including Medicaid and AFDC/TANF. The sample was limited to children age fourteen and younger in families with income at or below 300 percent of the federal poverty level. For the March 1999 CPS, in which the questions refer to enrollment in 1998, all of the children in the sample are age-eligible for the OBRA mandates. The CPS provides relatively large sample sizes, permitting estimates for the United States as a whole and for most states.2 To compare enrollment across the states and over time, I calculated a "standardized enrollment probability" in a second step. This involved taking the national mean values for the variables in the model above and multiplying them by the coefficients relevant to each state and year. This provides an estimate of low-income persons’ prospects of enrolling in each state. Because this method eliminates variations in need associated with differences in income across states and over time, it provides a clearer comparison of the impact of policy design and administration on program enrollment. The results are expressed as enrollment probabilities at given income levels, which can be interpreted as the percentage of the population at that particular income level who are enrolled in the program. In the discussion below, standardized probabilities are calculated at various percentages of the federal poverty level ($13,003 for a family of three in 1998).
National trends. Exhibits 1  and 2 portray national trends in the standardized enrollment probabilities for both children’s Medicaid enrollment and enrollment of their families in AFDC/TANF for 1979–1998. The Medicaid expansion created by OBRA 1989–90 is depicted in Exhibit 1 by the rising enrollment probabilities across the range of income levels after 1989. The rise in these enrollment probabilities accelerated in the years between 1992 and 1995, reaching a peak in 1995 or 1996, depending on income level. These changes were largest among children in families with incomes at 50,100, and 150 percent of poverty—gains of twenty-five to thirty-two percentage points from 1989 to 1995–96. These are substantial changes, especially when compared with the Medicaid enrollment probabilities that prevailed prior to 1989. In the early 1990s there were even important gains affecting children in families with no income, whose Medicaid eligibility had been nearly constant over the 1980s. The contrast with welfare is striking, since the probability that a child’s family received AFDC/TANF has been declining since 1992–93 (Exhibit 2).
The two trends have created a substantial difference in enrollment probabilities betwen these two programs at each income level. Prior to the OBRA mandates, children’s Medicaid coverage was largely related to their parents’ eligibility for welfare. The mandates altered this, establishing an independent path for children’s eligibility. By 1995–96, which represents the peak of the expansions, enrollment probabilities were sixteen to thirty percentage points higher for Medicaid than for AFDC/TANF among children with family incomes below the poverty line, and ten to twenty-six percentage points higher for those from 100 to 200 percent of poverty. Focusing only on the trend in Medicaid enrollment, the change from 1989 to 1995 equaled eight to twenty-six percentage points for those with incomes at or below poverty. The increases at these low income levels are particularly notable because during this period AFDC enrollment had already begun to fall, even before enactment of the national welfare reform law in 1996. Since 1995–96, however, children’s Medicaid enrollment levels have fallen. The enrollment probability for children with no income was 81 percent in 1995. By 1998 it had fallen to 68 percent. Similarly, for those with incomes at half the poverty line, the drop was from 61 percent to 53 percent. It is important to note that the drop-off was largest for those with the lowest incomes. These declines coincide with the decline in welfare enrollment, and, as discussed below, there is a strong relationship between the magnitudes of the Medicaid and welfare declines at the state level. In 1998,18.9 percent of U.S. children under age eighteen (about 13.5 million children) lived in families with incomes below the poverty line.3 Among these children in poverty, about 5.8 million were living in the very poorest families, those with incomes below 50 percent of poverty, while the remaining 7.7 million fell between 50 percent and 100 percent of poverty. The enrollment probabilities reported here are estimated at particular income levels, and so applying these results to all those falling into a range of incomes will necessarily be an approximation. If the change for children from families with no income of thirteen percentage points between 1995 and 1998 is applied to the entire group of 5.8 million children with incomes below 50 percent of the poverty line, this would imply a coverage loss for 754,000 children. If the change of eight percentage points estimated for those at 50 percent of the poverty line is applied to this group, then 464,000 children would be estimated to have lost coverage. The impact on the children with incomes between 50 and 100 percent of poverty can be similarly estimated. Applying the estimated six to eight percentage point decline to the 7.7 million children in this income range implies that an additional 462,000 to 616,000 children lost coverage. Combining the two income ranges together, the results imply that in 1998 Medicaid enrolled between 926,000 and 1.37 million fewer children with incomes below the poverty line than would have been enrolled if the policy that prevailed prior to welfare reform had simply been continued.
It is particularly noteworthy that the Medicaid enrollment decline for those with no income leaves them at an enrollment probability that is lower than at any year since 1979, the first year of the data series used here. Since AFDC eligibility in the 1980s was more restrictive than it was in the 1970s, today’s Medicaid enrollment is lower than it was in the 1970s as well. While the 1981 Reagan budget cuts had reduced AFDC eligibility for working parents, and thus lowered Medicaid enrollment levels for children in families with incomes between 50 and 100 percent of poverty, children in families with no income were not affected by these changes. This is illustrated in Exhibits 1
State-level trends.
Because the states are responsible for set ting Medicaid policy and administering the program, it is important to examine state-to-state variations in these enrollment trends (Exhibit 3
Impact of welfare reform. The last three years of Medicaid enrollment declines are strongly associated with recent declines in welfare enrollment. This is illustrated in Exhibit 4 , which plots AFDC/TANF enrollment changes on the horizontal axis against Medicaid enrollment changes on the vertical axis. For most of the states there is a clear association: The size of a state’s Medicaid enrollment change between 1995 and 1998 is directly related to its welfare enrollment change. Overall, there has been a noticeable re treat in Medicaid enrollment from the levels that had been achieved by!995.4
The exact relationship between the changes in these two programs depends on the nature of state responses, as well as the relative proportions of the Medicaid recipient pool who receive their eligibility through AFDC/TANF versus those with non-welfare-related eligibility. For example, for the nation as a whole, the welfare decline equaled twenty-one percentage points for families with no income. For these families, about 81 percent of children enrolled in Medicaid were also enrolled in welfare in 1995, which implies that the twenty-one-percentage-point decline in welfare would lead, in the absence of measures to protect Medicaid enrollment, to a Medicaid decline of seventeen percentage points. The actual Medicaid decline of thirteen percentage points for the country as a whole indicates that there has been only limited success in preserving poor children’s Medicaid coverage. The exceptions to this are states along the top of the figure: Alabama has had a Medicaid enrollment increase, while several other states have experienced relatively small declines in Medicaid enrollment (five percentage points or less)—Connecticut, Mississippi, Indiana, Tennessee, and Massachusetts—even though these states have had substantial declines in welfare enrollment. (Missouri has also had an increase in Medicaid enrollment, but this has occurred along with an increase in TANF enrollment.)
Many of the remaining states fall between two dashed lines in Exhibit 4
There are two major limitations to this analysis. The first is that it relies on the accuracy of the CPS, which is conducted monthly by the Census Bureau using a large sample and consistent methods across the country. The March supplement is a standard source of information on population, income, and program participation. Nonetheless, it has been criticized because it relies on self-reported participation in government programs. The number of total program recipients implied by these self-reports appears to be lower than the number of recipients estimated from administrative data. To address one type of underreporting, the Census Bureau imputes Medicaid enrollment for those who report receipt of public assistance payments. Since welfare enrollment has declined, fewer respondents are subject to this imputation; this may create a downward trend in Medicaid enrollment.6 In addition, the CPS survey instrument and sampling frame have been revised over time, introducing other potential changes, although these changes should be improving the survey’s accuracy. Finally, the questions of the March CPS refer respondents to income and program participation in the prior year. Recall bias may lead some respondents to actually report current income and participation instead.7 For this analysis the assumption is that this potential underreporting has been constant over time and across the states. Nonetheless, despite the potential limitations of the CPS, administrative data also show Medicaid declines that are associated with welfare caseload declines. A second limitation is that the analysis relies on a two-step methodology to develop standardized enrollment probabilities for each state. The first step uses a relatively simple model to control for individual and family characteristics. Important factors may have been left out of this model, and to the extent that these factors both explain program enrollment and differ across the states, the results presented here may be inaccurate. Analysis using more complex models, however, shows that the trends in enrollment identified above are robust to these issues of model specification.8
Research findings. These results show that preserving children’s Medicaid in the face of welfare reform has turned out to be much harder than anticipated. Despite the mandates, Medicaid does not yet cover all potentially eligible poor children, as indicated in Exhibit 1  by enrollment probabilities that are less than 1.00. This is true even for children whose family incomes are well below poverty: One-third to half of children in families with extremely low incomes are still not enrolled in Medicaid a decade after enactment of the OBRA mandates. While Medicaid enrollment for children had expanded during the first half of the 1990s, this progress appears to have been reversed. Despite the explicit language of the 1996 welfare reform law ordering the states to maintain the implementation schedule, in practice this guarantee of coverage has failed. Previous analyses have revealed important limits in the extent of the "take-up" of Medicaid benefits by the potentially eligible, suggesting that 4.7 million uninsured children in 1996 were not enrolled in Medicaid even though they were eligible.9 Using data from the states for the years just prior to the enactment of federal welfare reform, Marilyn Ellwood and Leighton Ku predicted that welfare reform might lead to Medicaid declines.10 Administrative data on Medicaid caseloads indicate a decline in caseloads in the first two years after enactment of welfare reform.11 Prior research on the impact of welfare reform on Medicaid has often been limited to examining the overall trends in administrative caseloads or has used raw tabulations of CPS data. An important difficulty for both approaches has been distinguishing the effects of welfare reform from the economic expansion that also occurred during the 1990s. The two-step analysis used here corrects for changes in income over time and variations in income levels across states. This removes the effects of the economic expansion, leaving only the effects that result from state policy and implementation. While previous researchers may have been hesitant to link Medicaid declines to welfare reform because of this potential confounding, the analysis above controls for the level of income for each family, and so the enrollment declines reported above are not the result of rising incomes associated with the economic expansion. Similarly, because the analysis standardizes for income, the state-to-state differences in enrollment probabilities reflect differences in policy and implementation, not variations in poverty rates across states. The analysis also examines the trends at different income levels, allowing for a disaggregation of policy effects. Previous analyses have only approximated this by comparing those falling above and below the poverty line, or distinguishing those eligible through welfare from those who receive Medicaid without also receiving cash benefits. So, for example, Ku and Brian Bruen report a 20.4 percent decline in cash assistance-related Medicaid enrollment for children, using administrative data, and a decline of 4.4 percentage points for children with incomes below 200 percent of the poverty line, using data from the CPS.12 The estimates presented here show that these declines are actually concentrated among those with the very lowest incomes: a thirteen percentage-point-decline for those with no income, in contrast to declines of only three to five percentage points for those with incomes at 150 and 200 percent of poverty.
Statistical modeling to correct for economic and demographic changes has been employed in two recent analyses of Medicaid enrollment changes.13 However, these studies give only national results and do not provide state-specific standardized enrollment estimates such as those presented in Exhibits 3 Policy implications. Policymakers and administrators face a number of difficulties in trying to provide a guarantee of Medicaid enrollment: Individuals and families have to explicitly sign up and, in fact, generally must initiate the process themselves. Potential enrollees must surmount a number of barriers including misunderstandings and lack of information about programs and their rules, burdensome applications and documentation requirements, complicated redetermination requirements, language difficulties, intrusive questions, demeaning experiences in the enrollment process, as well as the historic connections to welfare, which create the possibility that negative feelings and perceptions about welfare receipt will inhibit Medicaid participation as well. Policy recommendations naturally flow from this: simplify the application process, reduce documentation requirements, expand office hours, provide assistance in Spanish and other languages, accept applications through the mail and at sites separate from welfare offices, and provide more information about programs to those who are potentially eligible.14 These problems have been compounded by welfare reform initiatives, which are largely being evaluated by policymakers in terms of their impact on reducing welfare caseloads. Administrators face the difficulty of both encouraging the full use of programs that federal law has explicitly maintained (Medicaid and food stamps) at the same time that they are discouraging recipients from enrolling in the block-granted TANF programs. Program staff need to deal with a complicated set of Medicaid eligibility rules that are now independent of welfare eligibility, and computer systems must be modified to reflect these changes.15 Outreach efforts have been developed, but they may be better serving the relatively higher income families eligible for the Medicaid expansions and the new State Children’s Health Insurance Program (SCHIP). The results presented above show substantial backsliding for children’s Medicaid enrollment in most states. At this point, it is not clear what the few relatively successful states, identified in the results above, have done differently to preserve Medicaid enrollment. It is probable that both policy choices, such as the overall income eligibility limits, and administrative activities, such as the nature of the application process, will affect program use. Tennessee quite prominently established TennCare with an eligibility standard of 400 percent of poverty—higher than any other state. Massachusetts and Connecticut have set relatively high income limits for their SCHIP implementation; these programs have served to recruit families into public insurance programs.16 In addition, Connecticut provides two years of transitional Medicaid care for families leaving welfare. Alabama, Connecticut, Indiana, and Mississippi are among the fourteen states that provide twelve months of continuous eligibility for children’s Medicaid. All six of the relatively successful states use mail-in applications, although thirty-four other states have done so as well.17 Future research should examine how and why these states were relatively successful in preserving Medicaid enrollment for children, while others were not. In April 2000 the Health Care Financing Administration (HCFA) ordered that the states reinstate the eligibility of recipients losing Medicaid because they were leaving welfare, pending redeterminations of their Medicaid eligibility based on non-welfare-related coverage. This includes coverage related to transitional Medicaid benefits, Section 1931 preservation of Medicaid coverage under pre-welfare reform rules, Supplemental Security Income (SSI)-related eligibility, and, for children in particular, the mandated coverage of children in low-income families.18 This is a valuable first step that directly addresses one part of the problem: the loss of welfare-related Medicaid coverage among those leaving welfare. As discussed above, those who had received Medicaid through welfare generally will fall into the lowest income ranges which have had the greatest declines in Medicaid enrollment. The policy thus is targeted at an important group of families, many of whom are indeed eligible for continued Medicaid coverage. It should be noted, however, that many others who are eligible have never been enrolled, perhaps because they never sought enrollment or because they became discouraged at some point in the application process. The redeterminations of eligibility will do little to address the needs of those who have never obtained welfare or Medicaid benefits. The fear that welfare reform would harm access to health insurance is no longer a hypothetical risk. The evidence presented above indicates that the states have largely failed in the task of protecting Medicaid coverage for low-income children. This should lead state and federal policymakers to give greater attention to the policies and administrative practices of their welfare, Medicaid, and SCHIP programs that impede enrollment, and should direct new attention to the more general problem of the nonuse of program benefits by those who are eligible.
Karl Kroncbusch is assistant professor in the Department of Epidemiology and Public Health at Yale University. Special thanks arc due to Laura Tiehen, who aided in developing the original program used for this analysis. The author also thanks Mark Schlesinger, Jennifer Stuber, and two anonymous reviewers for their helpful comments on an earlier draft.
This article has been cited by other articles:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Data And Methods
Relating Welfare And Medicaid...
