We have presented evidence that in the one-year period from July 1993 through June 1994 there were 23 million transitions among major categories of health insurance coverage among children--including transitions into and out of the uninsured. In frequency, these changes in coverage amount to one for every three children. And while the 23 million transitions probably affected no more than half that many children, this is still a lot of children. Moreover, the transitions out of (and into) uninsurance were nearly as large as the number of children who were uninsured at any one time, and the same could be said of transitions out of (and into) other insurance. For Medicaid, the number of transitions approached half the number who reported being enrolled in Medicaid at any one time during the year. And while the transitions out of ESI were proportionately much less common than the transitions out of these other statuses, there were still 7 million of them.
The purpose of this research was to investigate the contribution of trigger events to the occurrence of these transitions. To this end, we examined a large set of primarily economic and demographic changes in children's families as potential trigger events. We found evidence that many of these events occurred disproportionately among children who experienced these transitions versus children who did not, and a regression analysis of the effects of trigger events on subsequent transitions provided evidence of statistically significant net effects of particular events on the likelihood of a child experiencing specific kinds of transitions.
What does this analysis of trigger events tell us about why there are so many changes in health insurance coverage among children in as short a time as a year? While we did not address this macro level question explicitly, trigger events provide a mechanism that is capable of accounting for such changes--and for their fluctuation over time. The events that we examined occurred with varying frequency in the different coverage groups, and when they occurred some fraction of the children who experienced them reported changes in their health insurance coverage soon after. For children with ESI, 15 to 30 percent left ESI in the next four months. For uninsured children, 35 to 45 percent became insured in the next four months. Many of the events that we examined are potentially sensitive to changes in the economy. If particular events become more frequent or less frequent, will the transitions with which they are associated be affected as well? The question is important, but to answer it we need to observe changes in the frequency of events and then assess their impact on transitions. Comparison of the late 1990s with the earlier years included in this study may provide the material with which to answer this question.
Our research was not designed to explain why trigger events affect health insurance coverage. Clearly, the coverage offered by employers and the terms of its availability are important in mediating the impact of changes in the parents' employment on children's coverage. Data of this kind were not collected in the earlier SIPP panels, and the latest (1996) SIPP panel will provide only somewhat more information. The most promising national data source is the Medical Expenditure Panel Survey (MEPS), which collects data from employers as well as household members. Unfortunately, however, there are no nationally representative data that would allow us to look at change in the coverage offered by parents' employers--or its costs to employees--as a factor in the gain or loss of employer-sponsored coverage for children or adults .
Despite the high transition rates that seem to follow certain events, it was surprising that some rates were not even higher. In particular, when fathers of children with ESI lost employment, two-thirds of the children--and their fathers--retained coverage through at least the next four months. Obviously, some portion of this can be attributed to the source of ESI being separate from the job that was lost, but how often can this be true, and what else can explain our findings? Better data on the actual source of coverage would be helpful here as well.
Finally, there are research issues involving some of the transitions themselves. Given the modest year-to-year change in the national distribution of children's coverage, we knew that movements between sources of coverage must be canceling each other, largely, but we were surprised to find how often this was true at the micro level--that is, how often children reversed their own transitions. While there is independent evidence of churning in the Medicaid program (see Ellwood and Lewis 1999, for example), its pervasiveness across types of transitions was striking. Of particular note are the children who moved from ESI to Medicaid and back within a four month period--and often without their parents losing ESI. Additional research involving data from a source other than the SIPP would be enlightening.