Are the BLS Employment Numbers Right?

On the first Friday of each month, the U.S. Bureau of Labor Statistics (BLS) publishes a payroll employment estimate. It’s the broadest and most accurate measure of the number of people working in this country and one of the most widely watched data releases. The Texas Real Estate Research Center (TRERC) closely follows this report because payroll jobs represent demand for real estate. The more workers there are in the area, the more offices, factories, and stores are needed. Further, the income from those jobs creates the demand for still more goods and services, including local housing.

Given its importance, can we trust the estimate to show the real situation?

On the plus side, the number comes from a large sample drawn from organizations that account for 97 percent of all jobs. The sample for this estimate is drawn in such a way that a monthly change of only 130,000 jobs is statistically significant. In other words, we can be pretty sure that just a tenth of a percentage point change in one month is real, and not statistical “noise.” If this were the end of the story, we could turn to the monthly BLS releases in complete confidence.

Not everyone appreciates that the BLS revises these numbers after publication. Two revisions are released in the months following the first report. For instance, the January 2024 headline number was 155.6 million jobs. It has since been lowered by 202,000 jobs, reflecting more recent data. In any one month, the revision may not be significant. However, the BLS tends to make consistently higher revisions or consistently lower revisions for many months in a row. For example, all the changes since April 2023 have lowered the original job number (see figure) The result is that job estimates today are substantially below where they would have been without those changes. BLS also makes annual benchmark revisions to the employment numbers when it updates its master list of businesses.

Beyond revisions, another complication can cause uncertainty over the job numbers.

At any time, new businesses open while some close. This means that the BLS’ master list of businesses is always a little out of date. To account for this, the BLS adjusts the sample job number by some amount using its business “birth-death” model. The adjustment is based on historic patterns in net business openings and job growth. The model can overcompensate when there have been major changes in the business cycle. Indeed, recently the model seems to be influenced by the dramatic restart of the economy after the COVID shutdown. According to TRERC’s estimates, the model has overstated job growth since January 2023 by a net 300,000 jobs.

In time, we expect that BLS will correct for errors such as this with its benchmark revisions. That gives little solace, however, to those of us trying to understand where the economy is today and where it is going.

While the payroll estimate is the most comprehensive and most precise employment measure we have, we should be aware that the headline number almost always changes. At best, the latest estimate is still a look in the rearview mirror.