- The Washington Times - Friday, February 9, 2007

Earlier this month the Bureau of Labor Statistics (BLS) reported nonfarm payroll jobs grew by a modest 111,000 in January, or about 75,000 a month less than the average monthly gain in 2006. But the bigger news was in the revisions to past jobs numbers.

The revisions included:

(1) Routine changes to the November and December monthly job count as more reports came in from employers in the payroll survey.

(2) Revisions resulting from annual benchmarking.

(3) Adjustments to the data from updating of the model used to estimate business births and deaths.

(4) Revised seasonal adjustment factors.

Total civilian employment collected by household survey was also revised for December 2006 as the result of updated population controls reflecting new estimates of international migration and vital statistics. Helped by more complete employer reporting for the last two months of 2006, the increase in payroll jobs in the final quarter of last year was revised upward by 104,000.

But by far the biggest revision to the jobs data resulted from the latest annual benchmark adjustment based on a full job count from state unemployment insurance tax records. This month BLS reported that nonfarm jobs for March 2006 were revised upward by 754,000. As a result, job estimates were raised by that amount in subsequent months and by declining amounts back to April 2005.

The 2006 benchmark revision was the largest on record, more than 3 times bigger than the average revision for the prior 10 years. More employment is always welcome, but an error of that magnitude is disconcerting.

BLS is investigating the sources of the problem. Some of the error, the Bureau says, has been traced to “the model-based estimation process for birth employment. The largest portion of the revision likely results from an accumulation of differences from several other error sources.” One hopes the ongoing research will bear fruit by reducing the size of benchmark revisions for 2007 and later years.

BLS plans include examining the benefits of more frequent benchmarking, bringing businesses into the sample sooner, further revising its business birth-death model, and establishing a survey to study reporting discrepancies.

In contrast to the large 2006 benchmark revision to the payroll jobs data, the annual update to the population controls used in the calculation of employment collected by household survey increased total civilian employment in December 2006 by 153,000.

Together, all the latest revisions to the payroll data boosted the level of jobs for December 2006 by 933,000 over the pre-revised job count for that month — a tidy addition.

Since the bottom of the last recession in November 2001, payroll jobs before the latest revisions were reported to have risen by about 5.4 million. The revisions boosted that to nearly 6.4 million. This is still short of the growth of total employment, which was about 9.5 million in the same period.

Still, the large upward revisions in the jobs data narrowed the gap in growth between the two series by nearly a million. If anything, the revisions add credibility to the household employment series, which deserves more attention than it’s been getting, certainly as an indicator of trend.

When the media report on the latest labor market data every month, they focus on payroll jobs and rarely mention total employment, primarily because the payroll sample size is larger. Although month-to-month changes in the payroll data are comparatively less volatile, over periods of several months and longer the household data may be the better indicator of trend. (Some economists take the average change in the two series as the preferred measure of growth.)

It’s good news that the expansion in jobs has been stronger than we thought. But large revisions to the jobs data are unsettling to those who depend on these data and have broad implications for the economy. Faster job growth may mean productivity growth is slower, the labor market tighter, and the threat of inflation greater than we thought.

Alfred Tella is former Georgetown University research professor of economics.

Sign up for Daily Newsletters

Manage Newsletters

Copyright © 2020 The Washington Times, LLC. Click here for reprint permission.

Please read our comment policy before commenting.


Click to Read More and View Comments

Click to Hide