What Precedes Spells without Insurance?
Between October 1992 and September 1994, children started nearly 19 million new spells without insurance. Each month, on average, about 800,000 children became uninsured. In the month preceding each new spell, 52 percent of these children were reported to have been covered by employer-sponsored insurance, 39 percent by Medicaid, and 6 percent by another source of insurance. The remaining 3 percent were born uninsured.

Altogether, more than 550,000 children left Medicaid each month, on average. More than half of this group--56 percent--started new spells without insurance. Just over one-third, or 34 percent, gained employer-sponsored coverage, 2 percent obtained coverage from another source, and 7 percent turned 19 or became ineligible for the SIPP.¹

These findings beg the question: why do so many children leave Medicaid only to become uninsured? Loss of Medicaid eligibility appears to explain fewer than half of these transitions from Medicaid to uninsured. That is, more than half of the children who left Medicaid and became uninsured still appeared to be eligible for Medicaid, according to a state by state simulation of the Medicaid eligibility rules. It is possible that reporting error may account for some of these apparent departures from Medicaid by eligible children, contributing to the recognized undercount of Medicaid enrollment.² Research in progress will shed light on the potential role of reporting error in particular, and on the circumstances surrounding these changes in Medicaid enrollment and insurance coverage more generally. Even if the number of children moving from Medicaid to uninsured proves to be fewer than the estimate reported here, policymakers need to shift some attention from the question of why children are not enrolling in Medicaid to the question of why they are losing coverage, especially when many of them may still be eligible.

What Follows Spells without Insurance?
Of the more than 9 million children who were uninsured in any one month, about 8 percent, or 750,000 children on average, became insured in the next month, and another 60,000 simply aged out of the population of children (turned 19) or left the SIPP universe. Medicaid had a disproportionate impact on these transitions. Of the children who became insured each month, 41 percent enrolled in Medicaid, 52 percent obtained coverage from their parents' employers, and 7 percent gained coverage from another type of plan. Thus, while Medicaid was reported to insure only 19 percent of all children under 19 at any one time in 1993 or 1994, it accounted for 41 percent of the uninsured children who gained coverage. This suggests that the number of uninsured children will be particularly sensitive to changes in Medicaid enrollment patterns. Indeed, the decline in the Medicaid caseload that followed the introduction of welfare reform appeared to produce a full percentage-point increase in the proportion of children who were without health insurance, as measured in the CPS (Fronstin, 1997b).

How Much Churning Is There among Uninsured Children and Medicaid Participants?
"Churning" in a population consists of exits and re-entries by the same group of people over a relatively short period, whereas "turnover" more literally involves exits and entries by different people. The frequency of churning and turnover among uninsured children and Medicaid participants influences the magnitude of the task of insuring the uninsured and the efficiency with which it may be possible to do so. For example, churning in the Medicaid population has come to be viewed as a serious problem by Medicaid managed care organizations.

As we noted, 12 million children started nearly 19 million new spells without insurance between October 1992 and September 1994. In other words, more than one-third of the new spells that began during this two-year period were the second or third spells of children who had become uninsured earlier in the period. In each year alone, 20 to 23 percent of the new spells were additional spells of children becoming uninsured for at least the second time in the year. The magnitude of this churning among the uninsured raises questions that must be addressed if we are to make significant progress in permanently reducing the number of uninsured children. Why do so many children regain coverage only to lose it again within a year or two? What are the roles of public and private sources of insurance in accounting for this phenomenon?

If we look at who is uninsured from one year to the next, we find less turnover than we might imagine, given the sizable numbers of children entering and leaving the uninsured population each month. While 800,000 children regained insurance coverage each month over the study period, only to be replaced by another 800,000 who lost coverage, and while 12 million different children became uninsured over the two-year period, we find that 62 percent of the 9.5 million children who were uninsured in October 1992 were still uninsured in September 1993, and 52 percent were still uninsured a year later. In other words, after two years, only about half of the uninsured children or between four and five million had been replaced by children who became uninsured during the intervening period. Some of the children who were still uninsured had undoubtedly remained uninsured continuously over the two-year period, but the estimates of churning suggest that children who gained insurance coverage and then lost it again may account for a significant proportion of the children who were uninsured at both the beginning and end of the period.

While the turnover in the population of uninsured children over one or two years may appear small, we would underscore two aspects of these findings:

• The turnover is not small in terms of actual numbers of children. If the federal government and the states could immediately and permanently insure the estimated 11 million children who are currently uninsured, these findings suggest that a year later, an additional 4 to 5 million children would be uninsured.
• The figures on year-to-year turnover hide the many children who go without insurance for some period during the intervening 12 months--losing but regaining coverage before the end of the year. Again, in the scenario that we described, with current estimates of the uninsured, more than 4 million additional children would go without insurance for some period during the year.

Both in absolute number and as a proportion of the total caseload, however, new spells on Medicaid were less common than new spells without insurance. The 13.4 million new spells on Medicaid that started between October 1992 and September 1994 contrast with the 19 million new spells without insurance initiated over the same period. The month-to-month turnover in the Medicaid child population, at 4 percent of the caseload, was only half the monthly turnover in the uninsured. Further, 80 percent of the children who were enrolled in Medicaid in October 1992 were still enrolled in Medicaid in September 1993 and, a year later, 76 percent. Again, some portion of the children who were enrolled in Medicaid at the beginning and end of the period actually left the program and returned, but given the lower exit rates from Medicaid than from the uninsured, we would expect a larger share to have been continuously enrolled in Medicaid than the fraction of uninsured children who were continuously uninsured over the same period.

In sum, while churning has attracted interest in studies of the Medicaid population, we find more churning and substantially more turnover among uninsured children than among children enrolled in Medicaid. However, we must recognize that changes in the Medicaid caseload that are developing post-welfare reform could begin to alter the Medicaid dynamics as well.

2. Surveys report fewer children enrolled in Medicaid than the statistics compiled by the Health Care Financing Administration (HCFA(now known as CMS)) indicate. This finding is typical of comparisons between survey data and program administrative statistics for means-tested entitlement programs. Some researchers adjust their estimates of uninsured children to compensate for Medicaid underreporting (Ullman et al. 1998). Others caution that we know too little about the sources of the Medicaid undercount to reliably attribute a large portion of it to children who are misreported as uninsured. In our case, adjusting reported insurance coverage to compensate for a Medicaid undercount was not practical, for the most part, because there is no evidence on how Medicaid underreporting might be correlated with the complex dynamics and the many characteristics we examined. Nevertheless, it is important to recognize when we and other researchers examine Medicaid eligibility among uninsured children that the survey data on which we base our estimates fail to account for all children who are counted in the Medicaid program statistics.