Examining Substitution: State Strategies to Limit "Crowd Out" in the Era of Children's Health Insurance Expansions. B. There is a range of substitution estimates reported in the literature.


Previous research has examined substitution in the context of Medicaid expansions. Although most researchers agree that substitution occurs at some level in all social insurance programs, it may be difficult to accurately predict future levels of substitution for new state children's health insurance programs based on the current, disparate research findings. The variation of substitution estimates based on current literature has yielded values ranging from 0 to 50%. Thus, the range itself indicates the complexity in estimating substitution with any level of precision.

There have been debates in the research community regarding the appropriate methodology for measuring substitution. Although two of the research teams (David Cutler and Jonathan Gruber and Lisa Dubay and Genevieve Kenney) used Current Population Survey (CPS) data, their approaches differed considerably. Cutler-Gruber focused on the question: what is the reduction in private insurance coverage as a share of the persons who enrolled in Medicaid directly as a result of the expansions? Dubay-Kenney, on the other hand, posed the question: What is the reduction in private insurance coverage as a share of the total increase in Medicaid coverage over this period?

The Cutler-Gruber study examined data from 1987 through 1992 of women of childbearing age (15-44) and children made newly eligible via Medicaid expansions. Cutler-Gruber looked across states for differences in Medicaid eligibility and public and private insurance, controlling for economic and demographic factors of the population in each state. They attempted to separate the substitution effect from secular trends in employer coverage, but admitted that there is no definitive way to isolate one from the other. Based on regression analysis, Cutler-Gruber found that the decline in private insurance was approximately 50 % of the increase in Medicaid coverage induced by expansions. This study identified that states with a greater increase in Medicaid eligibility had larger declines in private coverage when compared to states with smaller increases in Medicaid eligibility. The authors considered this to be strong evidence exemplifying substitution. The authors additionally noted in their analysis that opt out occurred largely due to the fact that employees enrolled in employer-sponsored insurance less frequently; however, employers may have encouraged this behavior among their employees by contributing less for their insurance. Therefore, Cutler-Gruber make the important point that the issue of substitution is not merely opt out or push out, but perhaps a complex interaction between both forms of substitution.

Another commonly cited study was conducted by Dubay and Kenney. This study examined the occurrence of substitution as Medicaid expanded into higher income ranges by focusing on poor (incomes below 100% FPL) and near-poor (incomes of 100-133% FPL) populations. Their research identified virtually no substitution of employer-sponsored insurance among poor pregnant women, but approximately 8.5% of poor children under age 11 were substituting coverage. Among near-poor pregnant women and children, the rate of substitution was more pronounced. Dubay and Kenney estimated a substitution rate of 45% for pregnant women and 21% for children under age 11 in the near-poor category. Though these findings suggest the effects of substitution, Dubay and Kenney cited secular changes in the insurance market as additional contributors that may have caused a portion of this effect. Overall, Dubay and Kenney estimated that 14% of the increase in Medicaid coverage for pregnant women and 12% of the increase in enrollment for children under age 11 could be attributed to substitution of private insurance. Both Dubay and Kenney and Cutler and Gruber found substitution to be less of an issue among children than among pregnant women. A study conducted by Lara Shore-Sheppard at the University of Pittsburgh additionally supported the work of Cutler-Gruber and Dubay-Kenney. Shore-Sheppard also utilized CPS data to conclude that declines in private coverage were unlikely to have occurred as a result of recent Medicaid expansions.

A July 1997 study conducted by The Lewin Group developed estimates of substitution consistent with those found by Dubay and Kenney. This analysis used two separate databases, the CPS and Survey of Income and Program Participation (SIPP), and identified that between 20.3% and 24.2% of children enrolled in Medicaid under eligibility expansions could be substituting Medicaid for available employer-sponsored coverage. This study also indicated that rates of substitution are highest among children with incomes above 100% of poverty. SIPP data (1990) were also used by Linda Blumberg, Lisa Dubay and Stephen Norton of The Urban Institute to examine the movement from private coverage to Medicaid under Medicaid expansions. This analysis supports the previous finding, that with respect to children, the displacement of private insurance coverage for Medicaid as a result of recent expansions accounted for only a small portion of enrollment into the Medicaid program.

Researchers have identified the concern of using population-based surveys to determine accurate levels of substitution in state children's health insurance programs. For example, the difficulty in interpreting CPS data to estimate the extent of substitution has been acknowledged by many researchers. Therefore, estimates of the uninsured may reflect only the number of individuals without insurance for the entire year, rather than portraying the true transitory nature of health insurance experienced by many low-income families. Of additional concern is that respondents may report health insurance coverage at the time of interview, rather than their experience throughout the previous calendar year. Studies which have utilized other population-based surveys such as SIPP have yielded comparable conclusions, yet are concerned with similar interpretation and data accuracy issues.

While research identifies a wide range of substitution estimates, it is important to note the vast majority of researchers have identified that there is some existence of substitution within state programs. However, the main weakness of current literature is its limited practical relevance to state programs and their experiences. For example, the definition of substitution used by Cutler and Gruber, Dubay and Kenney, and Shore-Sheppard does not include any concept of affordability, an issue that is of major concern to state policymakers. The nature of the research data required researchers to define substitution as any reduction in private coverage consequent with an increase in Medicaid coverage while controlling for secular trends. The data used did not allow for examination of the burden of private insurance premiums and out-of-pocket payments as a share of family income on the newly eligible, who are by definition low income. Therefore, by definition, any family that refuses an offer of health insurance from an employer, regardless of the expense, has "opted out" of the private health insurance market. State administrators have specifically operationalized the notion of individually-based substitution as the refusal of an offer of "affordable" private insurance. Some states have defined this as an employer subsidy of at least 50% of the total premium (e.g., Minnesota).