- The Washington Times - Sunday, May 10, 2009



If I told you that the job loss so far in this recession has been 3.7 million - about the number President Obama plans to save or create - you would say I was delusional. As headlines and commentators across the country have already told us, the loss is 5.7 million, a number that comes straight from the U.S. Bureau of Labor Statistics (BLS).

If I told you BLS estimates the job loss at both 5.7 million and 3.7 million, you would be puzzled if not skeptical. But it’s true.

With the exception of a few die-hard economists, not many people know about the lower number because it wasn’t reported in the media or in the monthly BLS press release on the employment situation, and it’s pretty hard to find on the Bureau’s Web site. To locate the information, you have to search the Bureau’s Web pages by plugging in the title of the following report: Employment from the BLS household and payroll surveys: summary of recent trends.

On the first page of the report is a table showing both the 3.7 million and the 5.7 million total nonfarm job loss since the business cycle peak in December 2007. The larger job loss number comes from the monthly survey of establishment payrolls. The smaller number comes from the government’s monthly household survey with employment adjusted to the definition of payroll jobs for comparability.

Household-measured civilian employment is adjusted to a jobs measure by subtracting out agricultural employment, the self-employed, unpaid family and private household workers, and workers absent from their jobs without pay, and adding in multiple jobholders. Population control revisions are also smoothed.

The April BLS report comparing the two employment series is 11 pages long, but nowhere will you find an explanation or a discussion of the sizeable discrepancy between the two job loss figures in the current recession.

The question has important policy implications and needs addressing. If the reason for governmental silence is that the BLS can’t explain the difference, it should say so and at least put forth some plausible hypotheses.

Even though the monthly sampling error of household survey employment is larger than the error for payroll employment - a good reason to prefer the payroll data in the evaluation of month-to-month changes in employment - this can’t be used to justify ignoring valid trends in the household series over several months or longer.

In 2006, BLS wrote: “Divergences in the two surveys’ employment measures may occur again. At such times, data users and analysts would be well served not to think of one survey as ‘right’ and the other ‘wrong.’ ”

In the 2001 recession, household employment after adjustment to a payroll definition also fell by less than payroll jobs measured by establishment survey, though, like now, the discrepancy received little attention. (For earlier recessions, BLS doesn’t have the necessary data to adjust household employment to a payroll concept.) In the subsequent employment recovery, though gradual, the household adjusted job count continued to ring a more positive note, rising by 2.1 million more than the traditional establishment-based series.

The sharpest drop in employer-reported payrolls during the current recession occurred in the past half year, a loss of 3.9 million jobs. By comparison, household-measured employment adjusted to a payroll definition fell in the same period by 3.1 million, or by 800,000 less. That difference averages out to more than 130,000 fewer jobs lost a month, not a small number. In the first 10 months of the recession an average of 180,000 payroll jobs were lost a month compared to 55,000 a month for the household adjusted series. In both these recession subperiods, the adjusted household job count did not fall as fast as employer-reported payroll jobs.

All in all, at this juncture there’s no good reason to reject the smaller 3.7 million recessionary decline in jobs estimated by BLS in favor of the 5.7 million number.

One possible explanation for the discrepancy: Some workers who lose their regular jobs go off-the-books in their next job, a tendency that’s likely to increase in a recession. These workers would not be counted in the payroll survey, but could be picked up in household interviews. The BLS in 2006 wrote that although off-the-books employment “obviously is not counted by the establishment survey, because the workers would not be shown on employer payrolls, the household survey will capture some or all of such activity.”

We are in a serious job recession, but it may not be as bad as we thought.

Alfred Tella is former Georgetown University research professor of economics.



Click to Read More

Click to Hide