Every year, the Bureau of Labor Statistics (BLS) revises its employment estimates based on surveys, which is the source of the widely reported monthly job gain and loss statistics, to sync it to a more comprehensive dataset.
This time around, the preliminary results suggest that the revision is likely to cause a reduction of more than half a million jobs. The final changes will be implemented early next year, just in time for election madness.
To dive a bit more into the weeds, each year, the survey-based Current Employment Statistics (CES) estimates are benchmarked to comprehensive counts of employment.
These comprehensive counts are derived from state unemployment insurance tax records that nearly all employers are required to file. Over the past decade, the changes from this benchmarking have averaged plus or minus 0.2% of total nonfarm employment.
This time, the preliminary estimate from the benchmarking indicates a downward adjustment to March 2019 total nonfarm employment of -501,000 (-0.3%).
At the industry level, the largest revision is in mining and logging, which will change by -2.2%, with transportation and warehousing up 1.4%.
The largest numeric changes are leisure and hospitality (down 175,000), professional and business services (down 163,000), and retail trade (down 146,400). Several industries will see upward revisions.
The CES sample covers about one third of all workers and is a starting point for the overall change in employment. Then, a statistical model is used to deal with the inability of the sample to capture, on a timely basis, employment growth generated by new business formations. The issue of business “deaths,” or nonresponses to the survey, also leads revisions.
For most series, there are also seasonal adjustments to eliminate the part of the change in employment which can be attributed to normal seasonal patterns.
For example, employment in retail naturally drops in January due to the temporary workers hired to help with Christmas crowds; thus, it makes sense not to count this expected change as a negative signal about the economy. Similarly, construction jobs tend to peak in the summer.
Is there a nefarious plot at work with BLS trying to make the economy look better or worse than it is?
No! The estimating is a complicated process, and as anyone who has built a large economic or statistical model (including me) will tell you, perfection is simply not possible.
While 501,000 is a large number, it’s a small percentage (just 0.3%).
While no incumbent likes to see the numbers go down for any reason (especially in an election year), the bottom line is that estimation is difficult at best, particularly when underlying economic conditions, policies, and global phenomena are rapidly evolving.
M. Ray Perryman, PH.D., is president and chief executive officer of The Perryman Group (www.perrymangroup.com). He also serves as Institute Distinguished Professor of Economic Theory and Method at the International Institute for Advanced Studies.