Using National Survey Data to Analyze Children’s Health Insurance Coverage: An Assessment of Issues. 1. Forms of Underreporting


It is important to differentiate among different sources or forms of underreporting, as they carry quite different implications for estimates of the uninsured.

a. Misreporting of Medicaid as Private or Other Public Coverage

In an effort to increase and maintain enrollment, a number of states have taken steps to give their Medicaid programs the appearance of private insurance plans. Such tactics may be successful to the point of confusing participants as to the source of their health insurance coverage. As a result, some Medicaid enrollees may report that they are covered by some type of private insurance plan or a public plan other than Medicaid and thus not get counted as Medicaid participants. Such people will not be counted among the uninsured, but their actions will contribute to Medicaid enrollment being understated relative to administrative estimates.

b. Incomplete Reporting of People Included in Family Coverage

The fact that multiple family members are often but not always included under the same coverage creates a measurement problem that different surveys approach in different ways--for example, going through the household person by person to ascertain each member’s source(s) of coverage versus identifying the person in whose name a particular coverage is held and listing all household members included in that coverage. However insurance coverage is measured, there exists the potential for individual household members to be omitted. The likelihood of such omissions may be greatest for children, whose coverage is often collected somewhat differently than that of adults.(30) A finding that is consistent across multiple surveys is that roughly one-fifth of uninsured children appear to have at least one insured parent in the household. Czajka (1999) finds that 7 percent of uninsured children as measured in the SIPP report having a parent covered by Medicaid while about 13 percent have a parent with employer-sponsored coverage. The 7 percent figure strikes us as high. While there are circumstances under which a parent could be Medicaid-eligible without the children also being eligible, the parent would still have to meet a means test, which in most cases would imply a family income low enough to qualify the children under the poverty-related criteria.(31) Furthermore, while there is nothing implausible about uninsured children having parents with employer-sponsored coverage, we suspect that if parents can report their own Medicaid coverage but omit that of one or more children, they can do likewise with employer-sponsored coverage. In short, much of the 7 percent of uninsured children with Medicaid-covered parents and at least some of the 13 percent with employer-insured parents may be misreported as uninsured.

c. Not Reporting Medicaid at All

Rather than misreporting Medicaid as another type of insurance or reporting Medicaid for only some of the family members who are actually covered, respondents may fail to report any Medicaid coverage at all. If such people report no other coverage during the reference period, they will be classified, incorrectly, as uninsured. If they do report other coverage, then they will be identified as insured, but they will still contribute to the underreporting of Medicaid. When the reference period is short, we suspect that nearly all of these cases will be recorded as uninsured. With a reference period as long as a year, however, a nontrivial share of those who fail to report their Medicaid coverage may have had--and reported--other coverage and, therefore, not been classified as uninsured.

d. Population Undercoverage

We have discussed how population undercoverage is endemic to surveys and how this type of error affects the low-income population disproportionately. While this almost certainly accounts for some of the underreporting of Medicaid, we suggest that population undercoverage may have a relatively greater impact on estimates of Medicaid-eligible nonparticipants. The circumstances that contribute to people being left off of household rosters--basically, their transiency or weak attachment to the household--may also make it unlikely that they would be covered by Medicaid if eligible. In general, population coverage may affect the uninsured more than the insured. Thus it may tend to push estimated Medicaid participation rates up rather than down and lower rather than inflate estimates of the proportion of children who are uninsured.