Estimates of the number of children eligible to participate in Medicaid and the State Children's Health Insurance Program (SCHIP), but who remain uninsured, will play an important role in the evaluation of SCHIP's success. There is no single data source that directly tells us how many children are eligible for these programs, however. Rather, estimates of eligibility must be simulated by applying state eligibility rules, household by household, using a survey database that contains measures of the same characteristics a caseworker would record in an actual eligibility determination. To carry out such a simulation, we require knowledge of each state's program rules for determining eligibility, as well as a nationally representative database that provides data on the variables most influential in determining eligibility.
In this report, we present key findings on variations in state eligibility rules across 10 states and discuss the implications for simulations of income eligibility for Medicaid and SCHIP, based on national household survey data. We also assess five national household surveys and their ability to support simulation of Medicaid and SCHIP eligibility.
Mathematica Policy Research, Inc. (MPR) collected detailed information on the rules used in 10 states to determine eligibility for Medicaid and SCHIP--Alabama, Arkansas, California, Colorado, Connecticut, Florida, Massachusetts, Michigan, New Jersey, and New York.1 The data involved four broad topic areas: income standards, income and asset methodologies, family composition rules, and other eligibility rules such as presumptive eligibility and third-party insurance coverage. Information about eligibility rules was collected for the Temporary Assistance for Needy Families (TANF) program, for two Medicaid programs--Section 1931 Medicaid and Medicaid poverty expansions--and the state's SCHIP program. As appropriate, we collected information on the state's SCHIP Medicaid expansion (M-SCHIP) and its state-designed program (S-SCHIP).
Based on the 10 states in our sample, we find that:
- Income eligibility rules vary considerably across states and across programs within a state.
- States have used the greater flexibility of the TANF program to introduce a larger variety of earnings disregards. In addition, TANF policies regarding the treatment of unearned income and household expenses vary considerably across states.
- Some, but not all, 1931 Medicaid programs have adapted TANF income eligibility rules, that provide some degree of uniformity across programs within the state. Medicaid poverty expansion programs, however, continue to use the disregard standards of the old Aid to Families with Dependent Children (AFDC) program when they use a net income test. As a result, income eligibility rules within a state can vary across the TANF, 1931, and poverty expansion programs.2
- In a further departure from uniformity, the S-SCHIP programs often use less complex rules than the Medicaid programs. Many S-SCHIP programs, for example, use gross income tests, and their family composition rules have fewer exclusions. Because of these differences, important variations have arisen between the Medicaid and S-SCHIP programs within a state and make simulating eligibility for these programs more complex.
While variability in eligibility rules was expected, the level of variability was somewhat surprising. In particular, the relatively common practice of using net income tests in Medicaid programs and gross income tests in S-SCHIP programs has implications for understanding the degree to which SCHIP actually extends coverage for children. To better understand the implications of differences across states and programs, we derived effective income levels for each program in our sample. Using a hypothetical family of three, we determined the maximum amount of earnings after accounting for earnings, child care, and child support disregards the family could have and still qualify the children for benefits under each program. We find that the effective income levels used by TANF and Medicaid 1931 programs vary widely across states. Medicaid poverty expansion programs extend coverage for children in families with income substantially above the levels for the TANF and 1931 programs. The effective income levels for most Medicaid poverty expansion programs are above the mandated levels. Our example also demonstrates that many M-SCHIP programs extend coverage for adolescents, while covering few (if any) additional younger children. We also find that the difference in the effective income levels between the Medicaid poverty and SCHIP programs is less than that suggested by the income thresholds used in income eligibility tests. This is one of several factors that may affect the success of SCHIP programs in covering additional children.
These findings demonstrate that the task of simulating eligibility for SCHIP has been made more difficult by the variation in eligibility rules across and within states. Cross-state variability requires knowledge of 51 different sets of eligibility rules, whereas within-state variability requires multiple layers of algorithms. This additional complexity makes evaluating SCHIP's success in reaching its target population more difficult. The estimation task is particularly challenging if estimates are desired for separate programs within states.
No single survey captures the full range of information needed for a relatively complete simulation of Medicaid and SCHIP eligibility. For example, the widely used Current Population Survey (CPS) collects no data on assets or expenditures, and nearly all the income data refer to the previous calendar year, rather than the monthly reference period often used to determine Medicaid and SCHIP eligibility. Furthermore, the CPS cannot be used to simulate the waiting periods that states have implemented in SCHIP as an anti-crowd-out feature. The more comprehensive Survey of Income and Program Participation (SIPP) provides detailed income data on a monthly basis and can simulate waiting periods. Its principal limitation is the long interval between data collection and the release of data products--a problem SIPP shares with the Medical Expenditure Panel Survey (MEPS). The MEPS collects nearly as much relevant information as the SIPP, with some of it stronger than the corresponding SIPP data; public use files, however, do not identify individual states, thus making it impossible for users to simulate differences in state thus eligibility rules. The National Health Interview Survey (NHIS), from which the MEPS sample is drawn, also does not identify individual states, and it lacks certain key variables (such as family income by source and assets). The National Survey of America's Families (NSAF) has some of the same limitations as the CPS with respect to the reference period for which income is reported; it too collects fairly limited data on expenditures and resources. The chief strength of NSAF is the comparative data it provides for 13 states that were assigned large samples. We conclude that, of the five surveys we examined, SIPP appears to collect the largest share of the information needed to simulate Medicaid and SCHIP eligibility in the 10 states.3
Clearly, to be credible, the eligibility simulations will have to be detailed. At the same time, they must incorporate major simplifications in order to be feasible. An important task (which is the subject of another report) is determining how simulations of Medicaid and SCHIP eligibility can be adapted to the data limitations without seriously weakening their value as a policy tool (Schirm and Czajka 2000).
The next chapter summarizes our findings on eligibility rules in the 10 states, and Chapter III discusses the implications for comparing eligibility across programs. Chapter IV assesses the data needs for a simulation of eligibility rules, while Chapter V compares the data collected by five national household surveys. We conclude with a discussion of our findings (Chapter VI).