As the Economy Improves, the Number of Uninsured Is Falling But Not Because of a Rebound in Employer Sponsored Insurance

The data for this report is based on Urban Institute analysis of the Census Bureau’s March Supplement to the Current Population Survey (the CPS Annual Social and Economic Supplement or ASEC). The CPS supplement is the primary source of annual health insurance coverage information in the United States.

There is debate over whether the CPS is measuring the number of uninsured for an entire year (as intended) or whether responses more closely reflect the number of uninsured at a point-in-time. In this paper, we assume that the CPS is essentially a measurement of point-in-time coverage, primarily because the number of uninsured in the CPS has historically been significantly closer to point-in-time estimates and well above the full year estimates of other surveys. While there is also a concern that the CPS understates Medicaid/CHIP enrollment and thus, possibly overstates the number of uninsured,*none of the estimates presented here have been adjusted to take into account possible underreporting of Medicaid/CHIP coverage. However, it is unlikely that the size of the Medicaid undercount varies substantially over time.

We use the health insurance unit (HIU) as the unit of analysis for determining family-level income. A HIU includes members of the nuclear family who can be covered under one health insurance policy (i.e., policyholder, spouse, children who are under age 19 and full-time students under age 23). Use of HIUs in determining family-level income leads to results that differ from those obtained when household income is used because the latter includes the income of all relatives and unrelated individuals living together. The income of the HIU more accurately reflects the income available to individuals when purchasing private insurance or determining eligibility for public programs. We look at changes in coverage dividing the population into three income groups based on percent of the federal poverty level (FPL). The FPL’s are useful because they adjust for both inflation and family size.

In 2011, the Census Bureau revised its health coverage imputation methodology for those who did not respond to health insurance questions. The revisions address the differences between the way that health insurance coverage is collected in the CPS ASEC and the way it is imputed. Previously, dependent coverage assignments were limited only to the policyholder’s spouse and/or children. The revisions now allow all members in the household to be assigned dependent coverage, and the increase in the imputed number of dependents with coverage more accurately reflects individual reporting. These revisions were reflected in the calendar year 2010 CPS ASEC data, and revised extracts were released for 1999 to 2009 data years allowing a methodologically consistent trend to be examined from 1999 to 2010. Overall, the new editing process led to a 0.6 percentage point decrease in the number of uninsured in 2009. The release of the 2010 Census has impacted our use of the 2010 CPS dataset and our ability to create time trends spanning the last decade. Every year, the CPS survey is weighted according to the most recent Census so that the results of the survey sample may be generalized to reflect the composition of the entire population. Since 2000, the CPS datasets have been created using weights based on the demographic information from the 2000 Decennial Census. With the release of the 2010 Decennial Census, the Census Bureau updated the previously published 2010 CPS data using weights based on the newly gathered information from the 2010 Census. While this update enables the 2010 CPS dataset to more accurately reflect the current demographics of the population, it leads to two different sets of estimates for 2010: those based on the 2000 weights and those based on the 2010 weights. It is important to take note that CPS data for previous years in the decade continues to use weights based on the 2000 Census. Through rigorous testing, we found that the changes resulting from the updates were too small to be considered statistically significant, with a few minor exceptions. The most important is in race/ethnicity, where the change in weights resulted in a statistically significant decrease in the total number of whites and a statistically significant increase in the total number of Hispanics and people from other races. There were no changes in the rates of ESI, Medicaid, or other forms of insurance. However, the changes in the numbers of whites, Hispanics, and “other race/ethnicity” meant that some of the reported decline in the number of whites without insurance and increase in the number of Hispanics and “other” were due to the change in weights. For example, 600,000 of the 1.4 million person decline in the number of white uninsured that we report in this paper was due to the change in weights (reflecting a greater decline in the white population). Similarly, there were 300,000 more Hispanic uninsured and 400,000 more “other” uninsured because of the larger estimated size of these populations.

We use NHIS data to assign coverage to young adults, ages 15-26 years old, who report private coverage from outside the household but don’t report what that coverage is. We analyzed the corresponding year’s NHIS data to obtain the total proportion who have such coverage and the share of ESI among those with coverage outside the household. We applied this proportion to the CPS, assigning these individuals to either ESI or private non group coverage so that the rates match those seen among this population in the NHIS. This method results in the overall share among this group with ESI and private non group coverage matching that of the NHIS.

* Davern M, Klerman JA, Ziegenfuss J, Lynch V, Baugh D, Greenberg G. A partially corrected estimate of Medicaid enrollment and uninsurance: results from an imputational model developed off linked survey and administrative data. J Econ Soc Meas. 2009; 34(4):219-40; Call KT, Davidson G, Sommers AS, Feldman R, Farseth P, Rockwood T. Uncovering the missing Medicaid cases and assessing their bias for estimates of the uninsured. Inquiry Winter 2001/2002;38(4): 396-408

Conclusion Tables

KFF Headquarters: 185 Berry St., Suite 2000, San Francisco, CA 94107 | Phone 650-854-9400
Washington Offices and Barbara Jordan Conference Center: 1330 G Street, NW, Washington, DC 20005 | Phone 202-347-5270

www.kff.org | Email Alerts: kff.org/email | facebook.com/KFF | twitter.com/kff

The independent source for health policy research, polling, and news, KFF is a nonprofit organization based in San Francisco, California.