Today, the Census Bureau reports that there are 44.9 million uninsured. But this Census Bureau number is inadequate and, in several important respects, unreliable.

How best to help the uninsured has been a prominent feature of domestic policy for decades. But before there is a solution to this pressing national problem, it must be accurately defined. And therein lies the inadequacy of the Census Bureau data.

In fact, there are several different estimates of the number of uninsured individuals in America and their characteristics. Discrepancies between these various estimates are not simply a matter of who was counted, but rather how they were counted and when they were counted. As the Census Bureau releases its numbers on the uninsured, federal and state policymakers should keep in mind that how the Census counts the uninsured matters as much or more than the aggregate number.[1]

Getting it Right

The problem of the uninsured is complex. For policy makers crafting a solution, the problem first must be clearly defined. But a solution is impossible unless policymakers can determine exactly who are the uninsured, how many are in their ranks, and why they lack insurance. We do know that, overwhelming, the uninsured are in working families and work for small businesses.[2]

Today, the job of counting the uninsured falls on the U.S. Census Bureau. The Bureau's chosen instrument for this task is a database called the Current Population Survey (CPS). But as Census officials themselves admit, the Bureau is ill-equipped to complete this task with accuracy. In Appendix C of its report on "Income, Poverty, and Health Insurance Coverage in the United States," the Census Bureau admits that "Health insurance coverage is likely to be underreported in the CPS"[3] and that underreporting "appears to be a larger problem than in other national surveys that ask about health insurance."[4] The Bureau speculates that "Some reasons for the disparity may include the fact that income, not health insurance, is the main focus of the ASEC (Annual Social and Economic Survey) questionnaire."[5]

Census Bureau officials also acknowledge underreporting of the number of Medicaid enrollees. They say that the problem arises because their survey "is largely a labor force survey, interviewers receive less training on health insurance concepts. Additionally, many people may not be aware that they or their children are covered by a health insurance program if they have not used covered services recently, and therefore, they would fail to report coverage."[6] Whether this explanation would close the gap between the Bureau's numbers and other reports on the uninsured is unclear.

The Medicaid Mystery

It is clear, though, that policymakers should carefully scrutinize the Census Bureau's estimates on Medicaid enrollment.

Medicaid is the joint federal and state program that provides health coverage to the poor and indigent.

Just as among the uninsured, there is churning within the Medicaid population. When a Medicaid recipient leaves the program-whether by receiving coverage from an employer or by surpassing Medicaid income eligibility limit-from that point on, he or she is no longer counted as enrolled in Medicaid in "point in time" surveys taken after that point. But this former recipient is counted in surveys that look at Medicaid enrollment "at any point" during that year.

Part of the apparent over-counting of the uninsured in the Census data is likely due to a serious undercounting of Medicaid enrollees. While the Centers for Medicare and Medicaid Services (CMS) reported Medicaid enrollment of 51 million in 2002, the Census reported only 33 million, a difference of 18 million people. This trend continues in 2003 with a .7 percentage point increase in Medicaid enrollment by the Census Bureau, putting that number at 35 million, but CMS reports 53 million enrollees. This discrepancy is, to say the least, problematic.

Conclusion

At the very least, the undercounting of Medicaid recipients and the undercounting of insurance coverage, as admitted and described by Census officials, demonstrate that the Census Bureau's figures on the uninsured do not accurately reflect reality and may lead policymakers and the public to incorrect impressions about the uninsured. Policymakers and policy experts have no excuse for not owning up to this fact and should supply it as a major caveat whenever making use of the Census data on the uninsured.

The Census Bureau's statement that its survey "is not designed primarily to collect health insurance data"[7] should be accepted at face value: it is an internal recognition that the Bureau's methodology for this task is flawed.

Public officials should require the Census Bureau to make appropriate methodological adjustments and thus better calculate the number of the uninsured and their durations of uninsurance. Congress and the Administration should take the necessary steps to see that the Census produces this sort of data in a clear, comprehensive, and reliable fashion.

The issue of uninsurance is simply too important for its public face to come from an indifferent and inaccurate survey. Only when the Census Bureau significantly revises its methodology should policy makers and the public trust this dataset as the basis for public policy.

Derek Hunter is a Research Assistant in the Center for Health Policy Studies at The Heritage Foundation.