Using data from the Survey of Income and Program Participation (SIPP), this report examines issues related to the measurement of health insurance coverage and Medicaid participation and the simulation of Medicaid eligibility and draws some observations about the sources of health insurance coverage among children, the frequency with which children lack health insurance coverage, and the frequency with which these same children are eligible but apparently not enrolled in Medicaid. Subsequent reports will investigate the dynamics of health insurance coverage and Medicaid eligibility and participation among children and the characteristics of children by their health insurance coverage.
This report is organized as follows. Section A provides an overview of the SIPP, including its strengths and limitations, its general representativeness, and its representation of the population of children over time. Section B discusses our use of the SIPP to measure children’s health insurance coverage. Section C presents estimates of health insurance coverage among children. Section D details our methodology for simulating Medicaid eligibility and presents estimates of Medicaid- eligible children by their basis of eligibility. Section E presents estimates of Medicaid participation and other insurance coverage among Medicaid-eligible children, and Section F discusses strategies for evaluating the Medicaid eligibility simulation and presents some comparisons with program administrative statistics as well as estimates of the impact of one feature of our simulation--the definition of the family unit.
A. An Overview of the Sipp
The data on which this report is based are from the 1992 panel of the SIPP. The SIPP is a longitudinal survey whose respondents are interviewed every four months about their activity during the preceding four months. The questions include a lengthy series of “core” items, included in every interview, and periodic “topical modules” that collect data more infrequently on specialized areas. One quarter of the sample, constituting a “rotation group,” is interviewed in every month, so that the data for a given calendar month are based on a roughly equal distribution of respondents answering questions about activities one month ago, two months ago, three months ago, and four months ago. The staggered interviewing is intended to ensure that no calendar month of data is affected unduly by recall bias or other error associated with distance from the interview.
The Census Bureau collected nine waves of data--that is, nine interviews--from the entire 1992 SIPP panel sample and a tenth wave of data from three of the four rotation groups (that is, three- quarters of the sample). These data provide a common reference period covering three full calendar years--1992, 1993, and 1994--although, as we will explain, the Census Bureau is not releasing all of the data collected for the final three months of 1994.
B. Measuring Children’s Health Insurance Coverage
The SIPP ascertains health insurance coverage by means of a series of questions that ask about specific types of coverage or about coverage in general. These questions are placed relatively early in the SIPP interview, when respondents are likely to be more attentive. By contrast, the health insurance questions in the March supplement to the Current Population Survey (CPS) are placed near the end of the interview.
The applicable questions are reproduced below.
20a.(If ... is 65 years of age or older or ... has a work disability) Medicare is a health insurance program for disabled persons and persons 65 or older. People covered by Medicare have a card that looks like this (SHOW FLASHCARD L). Was ... covered by Medicare?
23a.(If ... is 18 or older or the designated parent or guardian of children under 18 years old who live in the household) During the 4-month period, was ... covered by (use local name for Medicaid) or another public assistance program that pays for medical care?
b.May I see ...’s (use local name for Medicaid) card to record the claim number?
23c.(If ... is the designated parent or guardian of children under 18 years old who live in this household) Were any of ...’s children (under 18) covered by (use local name for Medicaid)?
d.(If yes) Which children were covered?
23e.(If 23a or 23c is marked yes) Was (.../(and)...’s children) covered during the entire 4- month period?
f.(If no) In which months was (.../(and)...’s children) covered?
24a.Was ... covered by a health insurance plan at any time during the past 4 months? (Include CHAMPUS, CHAMPVA, and military coverage. Exclude Medicaid, Medicare, and plans paying benefits only for accidents or specific diseases.) (If no, skip to 24k.)
b.Was ... covered by a health insurance plan during the entire 4-month period?
c.(If no) In which months was ... covered?
d.Was ...’s health insurance coverage from a plan in ...’s own name (primary policy holder) or was ... covered as a family member on someone else’s plan? (If own name, skip to 24f.)
24e.Whose plan covered ...? (Skip to 24k.)
f.Was ...’s policy obtained through ...’s current employer or union, through a former employer, through the CHAMPUS or CHAMPVA programs, or in some other way?
g.(If current employer or union or former employer) Did ...’s employer or union (former employer) pay all, part or none of the premium (cost) of this plan?
h.Was ...’s plan an individual plan or a family plan? (If individual, skip to 24k.)
i.Other than ..., which persons in this household were covered by ...’s plan? (Include children as well as adults.)
j.Did ...’s plan cover anyone who did not live in this household during the past 4 months? (If yes) Who did the plan cover?
24k.(If ... is the designated parent or guardian of children under 15 years old who live in the household) Were all of ...’s children under 15 years old covered by a health insurance plan? (Include CHAMPUS, CHAMPVA, and military plans.) (Exclude Medicare, Medicaid, and plans paying benefits only for accidents or specific diseases.)
l.(If no) Which children were covered by a health insurance plan?
24m.(If 24k is yes or one or more children is listed in response to 24l) Were any of these children covered by the plan of someone who did not live in the household during the past 4 months? (If yes) Which children?
These questions were asked of all household members 15 and older, with some additional qualifying restrictions as noted. Coverage of children under 15 was ascertained from those questions that asked explicitly about the coverage of respondents’ children or other household members--that is, question 23d for Medicaid and questions 24i, 24k and 24l for other types of coverage. Note that question 24i could be the source of reported coverage for adults as well as children.
From the fields provided on the SIPP files we assigned each sample person to one of 13 categories of insurance coverage in each calendar month from January 1992 through September 1994. In making assignments to children, and presumably many spouses as well, we often had to refer to the record of a parent or other adult in the household to determine the source of coverage. This was not true for Medicaid or Medicare, but it was true for every other source. The SIPP data file identifies the household member providing the coverage when the plan is in another member’s name, and we used this information to link to that person’s record and access variables describing the source of coverage. When the coverage was provided by someone outside of the household, no source could be identified. We relegated to a separate category those children whose only coverage was provided by someone outside the household. Similarly, if a child’s coverage was reported under question 24k or 24l, then no information on the source was available.(4) We assigned such children to a residual coverage category indicating that we had no information on the type of coverage.(5)
Finally, a person could have been reported as covered by more than one type of plan during a given month. Clearly, Medicaid could be reported in combination with any other type of plan. Married persons could be reported as having their own employer-sponsored or other plan and being covered by their spouse’s plan. Similarly, children could be reported as covered by both parents’ plans, although the Census Bureau’s coding of the responses does not seem to allow for recording coverage under more than one parent’s plan (there is only one variable pointing to another household member as the source of coverage). Where two or more sources of coverage are reported, they may have been overlapping, or one source may have terminated while another began. Rather than trying to capture and display multiple sources of coverage, we elected to assign a single source of coverage to each person, following a priority scheme. In view of the focus of our research, we assigned Medicaid the highest priority. That is, any child who was reported to be covered by Medicaid, regardless of whatever other coverage may have been reported in that month, was coded as a Medicaid enrollee.(6) Priority was accorded to other coverages in the following order:
- Other coverage held in one’s one name
- Other coverage held in another household member’s name
- Coverage provided by someone outside the household
- Generic other coverage not associated with another household member
Beyond Medicaid, then, this scheme gives priority to assigning the type of coverage for which we have the most information.
C. Estimates of Health Insurance Coverage
Table 2 reports estimates of the health insurance coverage of children under 19 in each of the first three waves of the 1992 SIPP panel. These results are based on the full panel sample--that is, sample members who were present for the duration of the panel or until such time as they left the SIPP universe. Results are presented for the one month that was common to the reference periods of all four rotation groups in each interview wave--that is, January 1992 for wave 1, May 1992 for wave 2, and September 1992 for wave 3. Presenting Table 2 allows us to introduce the 13 categories of health insurance coverage that we identified with the SIPP variables and to show what change in the distribution of coverage may have occurred between the beginning of the 1992 and the eve of FY93, which along with FY94 is the focal period for the rest of this report.
The upper panel of Table 2 presents estimates of the number of children in each of the 13 coverage categories while the lower panel gives the percentage distribution for all children in each year. Given that the number of children represented by the panel declines over time while the total population of children rises, the percentage distributions provide greater comparability over time and may apply nearly as well to the full population as to the population that the panel actually represents in each month.
The first two sources of coverage, Medicaid and Medicare, require no explanation. Categories 3 through 6 refer to the current or former employer of a child’s parent or guardian, generally, although an older teen could have reported coverage by his or her own employer. Category 3, which accounts for 55 percent of children in each wave, refers to an employer- or union-sponsored plan with the employer or union paying all or part of the cost of the premiums. Category 4, which accounts for 2 percent or less of children, includes plans that the employer or union offers with no subsidization. Category 5 involves coverage that a former employer continues to subsidize while category 6, representing no more than half a percent of all children, is coverage associated with a former employer but with no (further) employer subsidy. Coverage obtained under COBRA would appear to fall in category 6, although the questions on which this category is based could easily lead COBRA participants to report their coverage elsewhere. Categories 7 through 9 refer to coverage provided by CHAMPUS, CHAMPVA, or the military. Category 10 refers to a source of coverage other than 1 through 9. This may include a state-only plan or coverage purchased in the private insurance market. Category 11 is coverage provided by someone living outside the child’s household--typically a divorced or absent parent. SIPP tells us nothing about the source of the coverage because the person in whose name the coverage is held is not interviewed, but we can infer that this coverage would be assigned to one of categories 3 through 10.
Category 12 consists of coverage that could not be classified under one of the preceding 11 categories. This category accounts for only .3 percent of all children in the first wave but then grows to 1 percent in the second wave and remains close to that level in wave 3. We suspected that most of the children classified under category 12 were placed in that category by their parents’ responses to question 24k. That is the children were reported as covered but were not identified with the plan of any adult in the household. If all of the children assigned to category 12 were allotted to the category for that reason, then we would expect to find that all of them were under 15 years of age, based on the final screen for question 24k. While disproportionate numbers of the children in category 12 were indeed under age 15, there remained enough who were over that limit to persuade us that some other explanation was operative. Because the health insurance variables on the SIPP file are not identified with specific questions, it is difficult to determine how a child could end up in category 12 other than by question 24k. We are left to infer that some of the information on children’s coverage must have been incomplete. For example, the Census Bureau may have coded the child as covered under another household member’s plan but either failed to identify or misidentified the household member responsible for the coverage. With further review of individual records it may be possible to resolve why these code 12 assignments were made, but it may not be possible to determine the correct coverage in each instance. Given the low frequency of this category, further investigation may have little merit.
The final category consists of children with no reported coverage. The relative frequency of this category remained at 13.7 percent between waves 1 and 3. During this same period, the proportion of children reporting Medicaid coverage rose from 16.1 percent to 17.3 percent. This rise was offset by a decline in other types of coverage sufficient to leave unchanged the proportion of children reported as uninsured.
Table 3 presents estimates of children’s health insurance coverage for selected months of FY93 and FY94. The first page provides estimates of numbers of children while the second page gives percentages of all children. Reported Medicaid coverage continues the rise observed in Table 2., growing to 19.0 percent of all children by September 1994. The percentage of children who are reported to be without health insurance declines by a full percentage point from September 1992 (in Table 2). The only other notable changes are a decline in the proportion reporting coverage under an employer plan to which the employer makes no contribution and a rise in the proportion of children who are reported to be insured but with no details provided. The former drops from 1.7 percent of all children to 1.2 percent while the latter increases from .8 percent to 1.1 percent.
The upper panel of Table 4 compares enrollment in Medicaid at the end of each fiscal year to enrollment ever during the year. The lower panel makes the same comparison for uninsurance. For Medicaid the number ever enrolled in each fiscal year is one-third higher than the number enrolled at the end of the year. The number of children ever enrolled in Medicaid during the full two-year period is 58 percent higher than the number enrolled at the end of either year.(7) When compared to the one-year ever-enrollment, the two-year figure indicates very modest turnover from one year to the next, as the number of children ever enrolled over two years is less than one-fifth higher than the number ever enrolled in the first or second year.
Turnover among the uninsured is considerably higher than it is among Medicaid enrollees. The number of children ever uninsured during a fiscal year is about 75 percent higher than the number who are uninsured at the end of the year. The number ever uninsured during a two-year period is about 30 percent higher than the number ever uninsured in either year alone and about 130 percent higher than the number uninsured at the end of either year (or at any one time). For the two-year period, the percentage of children ever uninsured matches the fraction who were ever on Medicaid: 27.1 percent.
Table 5 presents estimates of the number and percentage of children who were without health insurance coverage by month for the calendar years 1992 through 1994. The percentage uninsured declines over the course of the first year but shows no trend after that. There is no evidence of seasonality in these numbers. We wanted to select two or three months to present statistics in this and the other technical appendices, and this table suggests that the choice is not particularly important. We elected to focus on the final two months of FY93 and FY94 in most of our cross- sectional tables, supplementing these with the first month of FY93 when appropriate. September 1993 happens to have a relatively high estimate of the uninsured, at 13.1 percent, compared to other months whereas September 1994, at 12.7 percent, is close to the average. It is important to keep in mind that the month-by-month results suggest that the difference between these two estimates does not reflect any trend.
Table 6 presents similar figures for reported Medicaid coverage. Here it is notable that the reported coverage rises from 16.1 percent in January 1992 to 18.7 percent in June 1993, then levels off (with modest swings). It is notable that the percentage point rise in reported Medicaid enrollment is nearly three times the decline in the percentage of children reported to be without health insurance coverage. This suggests that two-thirds of the enrollment increase is due to net movement from other sources of coverage rather than net movement out of the uninsured. We address this issue more directly in Technical Appendix B.
Table 7 provides information on sources of health insurance coverage in FY93, FY94, and the combined, two-year period. Figures for FY93 refer to coverage at any time during that year, and likewise for the FY94 figures. Figures for “FY93 and FY94" refer to the two year period. These estimates are based on our monthly measure of insurance coverage. While a child may in fact have been covered by more than one source in a month (or even at the same time), and SIPP can tell us about multiple sources of coverage during a month, we measured only one source per month, as explained above. Persons with multiple sources in this table, therefore, were covered by those sources at different times during the year (or two-year period for the last two columns of the table).
We note that 6.7 percent of children reported no coverage of any kind in FY93, 6.4 percent reported no coverage in FY94, and 4.6 percent reported no coverage in either year. The remainder of the children in each of these periods reported coverage for at least part of the period. Between 84 and 85 percent reported only one source of coverage in each of the two years while 80.5 percent reported only one source over the two-year period. Within each of the two years, 63 percent reported only employer-provided coverage while about 19 percent reported only Medicaid coverage, with about 2.5 percent reporting coverage from another source. For the two-year period, these figures changed little: 60 percent reported only employer-provided insurance, 18 percent reported only Medicaid, and 2 percent reported only another source of coverage. We know from Table 3 that the proportion of children ever covered by Medicaid in the two-year period was higher than the proportion covered in either year alone. Table 7 shows that some of this increase is due to more children reporting Medicaid along with another source over the two-year period relative to one year. Specifically, 9 percent of children report Medicaid in combination with one or two other sources over the two-year period compared to 5 percent for either year alone. (The rest of the increase in Medicaid enrollment must be among children who were otherwise uninsured.)
In Table 8 we look at the incidence of uninsurance among children who reported any coverage during either or both fiscal years. Among children with any coverage, about 16 percent reported one or months of uninsurance in either fiscal year, and 24 percent reported a spell of uninsurance over the two-year period. The probability that a child with any coverage during a period was ever uninsured during that period varies substantially by type of coverage. Among children who were ever covered by Medicaid in either year, 28 percent were uninsured for part of the year. Among those who were ever covered by Medicaid over the two year period, the percentage who were ever uninsured during that period was 41 percent. For those with employer-provided insurance, the percentages uninsured were less than half these figures. Among those with other coverage, the proportions with any months of uninsurance were about midway between those for children ever covered by Medicaid versus employer-provided insurance.
If we look at just those children who report one source of coverage, so that we can isolate the “impact” of the source of coverage, we find even more pronounced differences by type of coverage. While the incidence of uninsurance is generally lower among children with only one source versus two or more, children whose only source was Medicaid during a year had a 27 percent incidence of uninsurance while those whose only source was employer-provided insurance had only an 11 percent incidence of uninsurance. Among those with coverage from another source during a year, between 21 and 23 percent were uninsured for at least one month in the year.
These figures are particularly interesting in light of the policy focus on the CPS, which counts as uninsured only those persons who reported having had no coverage during the year and provides no estimate of persons who were uninsured for only part of the year. Estimates of the kind reported in Table 8 cannot be constructed with CPS data