Let’s check out BLS job report revisions from January 2023 through June of 2024 in a series of four charts.
The trend is obvious. There were only three upward second revisions in 18 months.
BLS Nonfarm Payroll Revision Magnitude
Sum of BLS First and Second Revisions
Note: The sum of revisions is first to original and second to original, not second to first.
Average BLS First and Second Revision
Average monthly revisions are about -35,000 a month for the second revision.
This is barely a start of what’s really happening.
At tip of the hat to Gary Brode at Deep Knowledge Investing. He asked me if I had anything on revisions and I told him I would produce some charts.
He has a post coming up on this subject but I was a little late in getting my charts to him. You may wish to subscribe to his reports.
QCEW Revisions
These monthly revisions do not incorporate annual BLS benchmark revisions based off more complete QCEW data.
QCEW (Quarterly Census of Employment and Wages) is a count of Unemployment Insurance (UI) administrative records submitted by 11.9 million establishments. It represents about 92 percent of data.
The monthly nonfarm payrolls are based on about 6.5 percent of the data with inaccurate imputations on top of that.
Nonfarm Payrolls NSA Minus QCEW
QCEW data is not in a friendly format to download. I manually added the above quarterly numbers from a QCEW download into a better formatted CES quarterly download from the St. Louis Fed.
In normal times, neither heading into or out of recessions, the differences seem to vary randomly in a tight range of about 2.5 million.
QCEW vs BLS
The base discrepancy is ~2.5 million. At the end of 2023 the discrepancy was ~3.4 million.
That implies an overcount in the monthly jobs reports of 0.9 million in six months.
That’s 150,000 a month on top of the monthly revisions!
For discussion of the above chart with more details and charts, please see How Much Does the BLS Overstate Monthly Jobs?
So, how much faith do you have in the monthly reports?
