The Effects of Trigger Events on Changes in Children's Health Insurance Coverage

C. FREQUENCY OF TRANSITIONS

Table 1 reports the average number of children who reported each major source of coverage between July 1993 and June 1994, including the number who were uninsured and the number who were reported as insured but with missing information on the actual source of coverage. The number of transitions out of each category of coverage (except the one indicating source not reported) and the number of transitions into each category are presented as well.

Altogether there were 23 million transitions among the types of coverage listed in Table 1, excluding any that began or ended with an unknown source. With an average population size of 70.7 million children over this period this amounts to one transition for every three children. Transitions out of uninsurance and transitions out of ESI totaled 7.8 million and 7.2 million, respectively. In addition to these there were 5.9 million transitions out of Medicaid and 2.4 million transitions out of other insurance. The 7.8 million transitions out of uninsurance compare to an average monthly uninsured child population of about 9 million. While the nearly 8 million transitions may reflect some children becoming reinsured twice during the year, we have shown elsewhere that turnover among uninsured children is very high, with about half the children who were uninsured at the end of a year being a different group of children than those who were without insurance at the beginning of the year (Czajka 1999). Transitions out of other insurance represented 87 percent of the number covered by other insurance at any one time, whereas the exits from Medicaid were about 45 percent of the average monthly reported enrollment. Transitions out of ESI were only 17 percent of the total ESI coverage, but this is hardly an insignificant fraction.

Transitions into each category of coverage were very nearly equal to the transitions out of each category--consistent with the findings of cross-sectional surveys that the distribution of children among types of coverage (or lack of coverage) changes very little from year to year. Comparison of the transitions out from and into each category of coverage suggests minimal net change in the number of children with ESI, a decline of about 400,000 in the number with Medicaid, and increases of 200,000 to 300,000 in the numbers with other insurance or no insurance.

1. Transitions Between Types of Coverage

Sample counts and population estimates of the transitions that are the focus of this study are reported in Table 2A along with the percentage distribution of the total transitions. The transitions are grouped by the status of origin. In Table 2B the transitions are grouped by the destination status. For the 12-month period from July 1993 through June 1994 the SIPP provides 3,753 sample observations of the 12 types of transitions.

Just over half of the children who left ESI became uninsured--more than 3.6 million. Those who obtained coverage from another source instead were about equally likely to enroll in Medicaid or to obtain other insurance. The 7.8 million children who left uninsurance obtained ESI or enrolled in Medicaid with about the same frequency whereas the 5.9 million who left Medicaid ended up uninsured more than two times out of three. Of those who obtained insurance from another source after leaving Medicaid, 90 percent or 1.8 million out of 2 million obtained ESI. Children who left other insurance tended to report ESI as their next source of coverage. Nearly three out of four or 1.8 million children who left other insurance did so while about half a million became uninsured and 100,000 enrolled in Medicaid.

Something else that is evident in Table 2A is that forward and backward transitions between a given pair of coverage statuses tended to occur with similar frequency. For example, 3.6 million children moved from ESI to uninsured during the year while 3.5 million moved from uninsured to ESI. Similarly, 3.6 million moved from uninsured to Medicaid while 3.9 million moved from Medicaid to uninsured; 1.7 million moved from ESI to Medicaid while 1.8 million moved from Medicaid to ESI; and 1.8 million moved from ESI to other insurance while 1.8 million also moved from other insurance to ESI.

These patterns of reciprocal movement help to explain why we see relatively little change in the distribution of children's health insurance coverage between one year and the next despite observing such a high volume of transitions. The near equivalence of forward and reverse transitions between almost any given pair of insurance types prompted us to ask whether a large part of this phenomenon could be attributed to behavior at the micro-level--that is, individual children moving from one type of coverage to another and then returning to their original coverage after a relatively short amount of time. ⁽⁸⁾ We found that this was indeed the case but to differing degrees for different types of transitions.

TABLE 2B
SAMPLE COUNTS AND POPULATION ESTIMATES OF SELECTED TRANSITIONS IN CHILDREN'S HEALTH INSURANCE COVERAGE: JULY 1993 TO JUNE 1994
Type of Transition	Sample Count	Population Estimate	Percent of All Transitions
All Children Under 19	11,666	70,719,000
Total Number of Transitions	3,753	23,325,000	100.0
Transitions to ESI from:	1,175	7,121,000	30.5
Uninsured	584	3,546,000	15.2
Medicaid	285	1,798,000	7.7
Other Insurance	306	1,777,000	7.6
Transitions to Uninsured from:	1,274	8,038,000	34.5
ESI	605	3,619,000	15.5
Medicaid	578	3,872,000	16.6
Other Insurance	91	547,000	2.3
Transitions to Medicaid from:	846	5,472,000	23.5
Uninsured	549	3,640,000	15.6
ESI	278	1,720,000	7.4
Other Insurance	19	113,000	0.5
Transitions to Other Insurance from:	458	2,694,000	11.5
ESI	320	1,840,000	7.9
Uninsured	104	645,000	2.8
Medicaid	34	209,000	0.9
SOURCE: Survey of Income and Program Participation, 1992 Panel. NOTE: The number of children reported in the first line refers to the average number under 19 at any one time, whereas the counts of transitions represent annual estimates.

2. Secondary Transitions

Table 3 reports how often each of the 23 million transitions was followed by a second transition within the next four months--that is, by the next SIPP interview--and the frequency with which the initial transitions were reversed by these second transitions. Altogether, 40 percent of the 23 million transitions were followed by a second transition within the next four months, and 33 percent were reversed. ^{(9), (10)} Thus more than four-fifths of the second transitions (32.6 divided by 40.3) involved a reversal of the initial transition.

The frequency with which transitions were followed by second transitions varied with the type of transition. Children leaving ESI were the most likely to have a second transition--54 percent of the time--and the most likely to reverse the initial transition--about 46 percent. Children leaving any of the other three coverage statuses were much less likely to have second transitions--30 to 38 percent--and much less likely to reverse their initial transitions--25 to 28 percent. The frequency of second transitions showed marked variation depending on the destination of the initial transition, however. Children who moved into ESI were the least likely to have a second transition, with probabilities ranging from 23 to 24 percent. Children moving into other insurance, on the other hand, were as likely to have a second transition as children leaving ESI, with probabilities ranging from 49 to 59 percent. Children moving to Medicaid from other insurance or no insurance had intermediate probabilities of experiencing second transitions, but it is noteworthy that the single transition with the highest probability of being followed by a second transition and the highest probability of being reversed was the transition from ESI to Medicaid. For these transitions, 64 percent were followed by a second transition within four months, and 55 percent were reversed. In short, transitions from ESI to Medicaid were exceedingly temporary. At the same time, transitions in the reverse direction--from Medicaid to ESI--were the most long-lived. Only 23 percent of the children who moved from Medicaid to ESI had a second transition in the next four months, and less than 16 percent returned to Medicaid within that time frame.

Only two types of initial transitions were more likely to be followed by a transition to a third coverage status than a return to the original status. Children who left Medicaid for other insurance had a 59 percent chance of having a second transition within four months, but less than 18 percent returned to Medicaid. Instead, nearly 19 percent obtained ESI, and 23 percent became uninsured. Children who made the reverse transition initially--that is, they moved from other insurance to Medicaid--had a 40 percent chance of making a second transition within four months, but less than 16 percent returned to other insurance. All of the remainder, or about 25 percent, became uninsured. Both types of transitions were quite rare, however, and the sample sizes on which these estimates are based are quite small (see Table 2A).

It is of particular interest how often children who became uninsured had a second transition and became covered again within four months. Depending on the original coverage (ESI, Medicaid, or other insurance), between 44 and 50 percent of the children who lost coverage regained some form of coverage within four months. ⁽¹¹⁾ At least three-quarters of these children returned to the same coverage that they had prior to becoming uninsured.

If all initial transitions were ultimately reversed, we would expect to observe that about one-half of our sample transitions were reversed. The other half would be secondary transitions and, therefore, not followed by reversals. Looking only four months ahead, we find that nearly half of all transitions out of ESI were reversed but only somewhat more than a quarter of other transitions. Because of the seam bias, we would have to look an additional four months ahead to see appreciably more reversals--if indeed there are many more--and for much of our sample of transitions this would take us beyond the end of the 1992 panel. Nevertheless, from what we observe we can conclude that children's reversing their transitions does indeed help to explain why we see nearly equal flows backwards and forwards between any given pair of coverage statuses. Yet the differences in the rates at which particular transitions were reversed indicate that reversals alone cannot explain the near equivalence of most forward and backward transition rates. The dynamics of transitions between types of health insurance coverage are too complex to be summarized so simply.

3. Response Error

We also considered the possibility that the frequent reversals of transitions between one SIPP interview and the next could reflect error in the reporting of health insurance coverage. Specifically, if coverage during a reference period were misreported and then corrected in the next, this would give the appearance of an initial transition followed by a reversal. One way in which such error could occur is through changes in the respondent. While the intent in the SIPP is that each adult respondent answer his or her own questions, nearly a quarter of all SIPP interviews are conducted with proxies--that is, someone in the household other than the intended respondent answers the questions for that respondent. ⁽¹²⁾ A change in respondent from self to proxy occurring between one wave and the next could result in the respondent's health insurance coverage being misreported for that reference period. If the respondent returned to answer the questions in the next wave, the earlier, correct coverage could be reported again, giving the appearance of a reversal of the "transition" recorded during the previous wave.

To assess whether changes in respondent may account for a disproportionate number of reported transitions and, in particular, transitions that were reversed by the next interview, we examined the frequency with which transitions coincided with changes in respondent for the father, mother, or child 15 and older. For all children, even those 15 and older, we identified changes in the proxy status of each parent. For children 15 and older we also identified changes in their own respondent status. Any of these respondents--or their proxies--could have been responsible for reporting a child's health insurance coverage.

Our findings, presented in Table 4, indicate very clearly that we can reject the possibility that the frequent reversals of transitions between one SIPP interview and the next can be explained by changes between self- and proxy respondent. The frequency with which transitions were reported and the frequency with which they were reversed were no more common among children with changes in respondent than among children with the same respondent over the interviews in question. In all, 28.5 percent of the estimated 23 million transitions coincided with changes in respondent, and 16.5 percent coincided with changes in respondent that were reversed at the next interview. Of the estimated 9.4 million transitions that were followed within four months by second transitions, 28.0 percent were accompanied by changes in respondent and 14.6 percent by changes in respondent that were reversed. Clearly, then, respondent changes were no more common among transitions that were followed by second transitions than they were among all transitions. Likewise, of the 7.6 million transitions that were reversed within four months, 26.1 percent coincided with changes in respondent and 13.2 percent coincided with respondent changes that were themselves reversed at the next interview. Again, these respondent changes were no more frequent among transitions that were reversed in four months than they were among all transitions.

We will review additional but more indirect evidence relating to response error in Section E.

Footnotes:

8. If the time period over which such reversals occurred were much more than a few months, we would not observe many such reversals within a one-year window.

9. These second transitions are already counted among the 23 million when they occurred within the July 1993 to June 1994 window.

10. Because of the aforementioned seam bias in the reporting of transitions, "four months" undoubtedly includes a number of transitions that happened in five or six months. If we had looked ahead six months instead of four, however, we would have found relatively few additional transitions because so many five and six month spells are reported as only four months in length.

11. The frequency with which spells of uninsurance end in four months or less has been noted in other analyses of SIPP data (see, for example, Bennefield 1998, Copeland 1998, and Czajka 1999). Again, the seam bias in the reporting of transitions overstates very substantially the extent to which these durations were exactly four months instead of two, three, five, or six. By the same token, however, spells of only a month or two in length are almost certainly underreported.

12. Children under 15 are not interviewed. Information on their health insurance coverage and other characteristics is always provided by another household member, who can change from one interview to the next.

Where to?

Top of Page
Table of Contents

Home Pages:
Office of Health Policy (HP)
Assistant Secretary for Planning and Evaluation (ASPE)
U.S. Department of Health and Human Services (HHS)

Last updated September 5, 2000