Employment was stagnant last month as measured by two different government surveys, one of business payrolls and the other of U.S. households. Both surveys, however, showed a rising trend in employment in the second half of last year. Between July and December, payroll jobs rose by a tepid 278,000, whereas employment as measured by household survey rose by more than 3 times as much, by a healthy 875,000. Which number are we to believe?
Most economists and the media rely on the payroll data as the preferred measure of job market conditions. But ignoring the household employment data can sometimes result in misleading conclusions.
There are pluses and minuses to both surveys. The household survey, which is based on monthly interviews with about 60,000 households around the country, counts people, not jobs. It is more comprehensive in that it includes the farm sector, the self-employed, and unpaid family workers. Total civilian employment as measured by the household survey is currently about 6 percent greater than the number of nonfarm jobs counted in the payroll survey.
The payroll data are drawn from a larger sample of businesses and government agencies covering about 400,000 worksites nationwide. But by measuring jobs, not people, the payroll survey double-counts workers with more than one job. Unlike the household survey, it counts workers under age 16, civilian jobs held by members of the Armed Forces, and workers who commute to the U.S. from Canada and Mexico. As economists have noted, in periods of economic recovery, as now, the payroll survey may lag in recording jobs created by new small businesses.
As a recent Federal Reserve Board analysis pointed out, employment as measured by the household survey depends on population estimates that in the last few years may have been too high. Estimating population requires assumptions about the numbers of illegal immigrants entering the country.
The estimates have probably been too high because of the unforeseen slowing effect on immigration of a weak job market and more stringent entrance controls since the 2001 terrorist attack. On the other hand, the household survey probably has the advantage of picking up more off-the-books employment than does the payroll survey.
After looking carefully at the differences between the payroll and household employment data, Federal Reserve Gov. Ben S. Bernanke concluded in a Nov. 6, 2003, talk at Carnegie Mellon University that “the truth probably lies in between the two series.” But noting the larger sample size of the payroll survey and the questionable population estimates underlying the household employment data, he went on to say “somewhat greater reliance should probably be placed on the payroll survey.”
Based on this conclusion, if, say, twice as much weight is given to changes in business payroll data as to changes in household survey employment data, a recalculation of the numbers shows “true” employment since last July increased about twice what was recorded by the payroll survey. But putting this conjectural estimate aside, it’s fair to say the job market is doing better than just the payroll data indicate.
In analyzing employment, more information is better than less. The media should resist the temptation to oversimplify by reporting only the payroll jobs data. As the economic recovery unfolds in the months and quarters ahead, the household as well as the payroll employment numbers will need to be weighed in the balance if we are to better know the condition of the job market and where the economy is headed. If that sometimes means having to contend with a little more uncertainty, and even contradiction, so be it. Trying to grasp the truth is sometimes a messy business.
Alfred Tella is a former Georgetown University research professor of economics.