March 21, 2012

The Demographics of the Jobs Recovery

Appendix A: Revisions of the Current Population Survey

Each January, the U.S. Census Bureau makes adjustments to the population controls in the Current Population Survey. That means the sample weights are revised so that estimates from the CPS agree with pre-specified national population totals by age, sex, race and Hispanic origin and with state level totals by age, sex and race. The effect of the revision in January 2011 was to reduce the estimate of the working-age population by 347,000, the labor force by 504,000 and the number of employed workers by 472,000 (“Adjustments to Household Survey Population Estimates in January 2011,” Bureau of Labor Statistics, February 2011). Rates—for employment, labor force participation and unemployment—are either not affected or only slightly affected by the January CPS revisions.

The adjustments to CPS weights are typically based on revised estimates of net international migration and updated vital statistics. Methodological changes also play a role. In the 2007 and 2008 population estimates, introduced into the CPS in January 2008 and January 2009, respectively, the Census Bureau made significant changes in the methodology used to measure international migration from 2000 onward. The impacts of those changes are concentrated in groups where a high percentage of the population is foreign born, notably working-age Hispanics and Asians. As such, the new population controls have the potential for affecting the measured size of the foreign-born population and labor force.

Labor market statistics published by the Bureau of Labor Statistics are not revised historically to account for the effect of annual revisions to CPS weights. However, for each new “vintage” of population estimates, the Census Bureau releases the entire time series of monthly population estimates from April 2000 through the year when the latest estimates are used for CPS weights. These revised population estimates can be used to produce a consistent series of CPS data from 2000 onward by reweighting the CPS.

The estimates in this report are adjusted in two ways to account for the revisions of the CPS. First, the estimates for 2007 to 2008 are produced using reweighted data that incorporate the Vintage 2008 population estimates for the civilian non-institutional population (Vintage 2008 population controls were introduced into the CPS in January 2009). The new weights are derived using procedures that follow the weighting procedures of the U.S. Census Bureau (2006) to the extent possible with public-use data applied to Vintage 2008 population estimates—both published and unpublished data supplied by the U.S. Census Bureau to the Pew Hispanic Center (see Passel and Cohn, 2010, for additional details). With this consistent series of CPS data, it is possible to more accurately measure changes over time in working-age population, labor force and employment of various racial, ethnic and nativity groups.

Second, estimates for 2007 to 2011 are revised to account for the effects of CPS revisions in January 2010 and January 2011 using a BLS-published methodology (see “Creating Comparability in CPS Employment Series,” by Marisa L. Di Natale). This methodology first produces revised estimates for the overall Hispanic and non-Hispanic populations. The overall revisions to the Hispanic and non-Hispanic estimates are then further allocated to subpopulations by gender, race, nativity, year of arrival, place of birth and industry. The allocations are based on the distributions of the Hispanic and non-Hispanic working-age population by these characteristics in the first quarter of 2011.