I have some new charts today on job revisions. Let’s discuss.
BLS Survey Background
- Every month, the BLS monthly payroll reports produce a preliminary estimate of nonfarm payrolls.
- The BLS does two monthly revisions of nonfarm payroll as well as annual revisions.
- The monthly report is also called Current Employment Statistics (CES). Nonfarm payrolls are Establishment Survey (a survey of business employment).
- The unemployment rate comes from the Household Survey, a survey of individuals.
BLS Nonfarm Revision Chart Notes
- The above chart reflects the initial nonfarm payroll report and the two following monthly revisions.
- The chart runs through June because the July data has not yet been revised. June reflects the difference between the first report and the first revision in July.
- The other months reflect the difference between the initial report and the second revisions.
For fourteen of the last 20 months the revisions have been negative.
BLS Nonfarm Payrolls as Released and Revised
There is no revision yet to July so data runs through June.
The June second revision will come out this next monthly report this Friday as will the first revision to July.
Employment Revisions Six Month and 12-Month Rolling Sum
Negative revisions are not always associated with recessions, nor positive revisions with non-recessions.
Dating to November 1964 (that’s how far back the Philadelphia Fed data goes), there were 306 months of negative initial to third (or current for most recent months).
That’s 306 out of 717 months of negative revisions, 42.7 percent of the time. Positive revisions happen 57.3 percent of the time. But ….
BLS Revises Jobs Down by 818,000, About 68,000 Per Month
On August 21, I reported BLS Revises Jobs Down by 818,000, About 68,000 Per Month
Do I get to say I told you so? My advance estimate a month ago was 779,000 lower. Bloomberg estimated 730,000.
A Breakdown, by Sector, of the Negative 818,000 BLS Job Revisions
Here’s a A Breakdown, by Sector, of the Negative 818,000 BLS Job Revisions
Yesterday morning, I had difficulty finding the BLS job revisions. So did everyone else. Eventually the BLS posted -818,000 but my unadjusted calc of -915,000 is accurate too.
Q: How do those numbers factor into the monthly revisions?
A: They don’t.
CES Preliminary Benchmark Announcement
Please consider the CES Preliminary Benchmark Announcement
In accordance with usual practice [but not in the usual way], the Bureau of Labor Statistics (BLS) is announcing the preliminary estimate of the upcoming annual benchmark revision to the establishment survey employment series. The final benchmark revision will be issued in February 2025 with the publication of the January 2025 Employment Situation news release.
Each year, the Current Employment Statistics (CES) survey employment estimates are benchmarked to comprehensive counts of employment for the month of March. These counts are derived from state unemployment insurance (UI) tax records that nearly all employers are required to file. For National CES employment series, the annual benchmark revisions over the last 10 years have averaged plus or minus one-tenth of one percent of total nonfarm employment. The preliminary estimate of the benchmark revision indicates an adjustment to March 2024 total nonfarm employment of -818,000 (-0.5 percent).
Preliminary benchmark revisions are calculated only for the month of March 2024 for the major industry sectors. The existing employment series are not updated with the release of the preliminary benchmark estimate. The data for all CES series will be updated when the final benchmark revision is issued.
Big Bang Revision
Long after everyone cares about what happened in the first quarter of 2024, the BLS will post an annual revision in one big bang.
Every year, this makes all of the monthly data reports historical nonsense “in accordance with usual practice.“
Thought of the Day
68k downward revision per month on average. This means that productivity will be revised up while unit labor costs will be revised down.
Personal income and Gross Domestic Income will be revised downward and the personal saving rate will be reduced from its already very depressed level of 3.5 percent.
We should remember that the markets traded on this faulty data. The downward revision is confirmation that the birth/death model, which imputes the gain in jobs for small firms, is not working.
Happy Labor Day
Thank you Mish for your analysis.
Full time jobs are down and part time jobs are up with the net difference being what is reported. Excusable obfuscation?
Wouldnt total hours worked be a more accurate way to portray?
BLS Survey Background:
– Every month, the BLS monthly payroll reports produce a preliminary estimate of nonfarm payrolls. Then the BLS does two monthly revisions (middle & end of month.of nonfarm payroll as well as annual revisions. (Actually “Corrections”)!
> Every month, I wonder why on earth, WHY do we feel the need to estimate something that already is estimated? So we can change it around to better meet the original estimate? That would be called Altering a report to make the estimate look better, more accurate, and to make you look like you were doing your job right in the first place imo! We then allow for Two Corrections, which is Truly What They Are!
– The monthly report is also called Current Employment Statistics (CES). Nonfarm payrolls are Establishment Survey (a survey of business employment).
> Should be called the “Monthly “Made Up” Report, for what it actually is. A damn Estimate, with multiple estimate corrections along the way, to better reflect what they want it to (reality), and to make it appear that they knew what they were doing upon the estimation. Unfortunately that appears to be very far from the truth.
– BLS Nonfarm Revision Chart Note
> For fourteen of the last 20 months the revisions have been negative. Showing you just how accurate they truly are… NOT! Just a glossy report, massaged and manipulated (there’s a surprise) a few times along the way. Everything sort of works out better, when I do ALL MY Numbers at the end of every month, rather than the beginning, where the accuracy truly matters, but I digress…
– BLS Nonfarm Payrolls as Released and Revised.
Negative revisions are not always associated with recessions, nor positive revisions with non-recessions.
> No Kidding, and did anybody think otherwise, JS…
– Dating to November 1964 (that’s how far back the Philadelphia Fed data goes), there were 306 months of negative initial to third (or current for most recent months).
> So Since 1964, they have all been just making numbers up along the way. Then they change them (re-estimate) Two More Times before the month is over, and they STILL GET IT WRONG? These should be some of the most accurate reported numbers in all of our Government, with this sort of attention to detail!
– That’s 306 out of 717 months of negative revisions, 42.7 percent of the time. Positive revisions happen 57.3 percent of the time.
> so a literal crap shoot / roll of the dice if you will. Nearly a 50% accuracy rate, OR FLIP A COIN!!!we should be so proud of these mathematicians… NOT!
– We should remember that the markets traded on this faulty data.
> Pretty Scary right there!
Happy Labor Day!
To you as well Mish, and “Thank You” for. All your time in dealing with such complex issues. People like You make a difference, and it’s a pleasure to see, and I for one, truly enjoy your efforts!! I question numbers all the time, and always will, because I see too many “Revisions” and that just tells me over and over again, that we are dealing with “Faulty Data” in many/most cases in my experiences., albeit limited to “Intl. Supply Chain Management”
And yet with all of these negative revisions, the NBER hasn’t declared we’re in a recession yet.
Imagine that?
How much extra demand is 12.5M illegals created when coupled with $1.92T in deficit spending.
Like seriously? Who in their right mind thinks we’re teetering towards a recession with 3% Q2 GDP?
At some point, the labor market was going to show signs of slowing down. Okay, so it took 12 months longer than expected. The 4W moving unemployment claims average peaked a month ago @ 241K and has fallen to 231.5K. Could they turn north again? Absolutely, but given reasonable consumer spending and all the other factors, will not be surprised by a slow turn north in manufacturing.
As expected, it’s month-to-month these days, but I for one will not be surprised to see sub 6% 30YFM by October and 5.5-5.75% headed into next spring. Housing sales are already turning positive and will continue pushing prices higher in most cities, so get ready for an early 1980’s inflation head fake in the next 3-5 months.
It’s unfortunate we don’t have both timely and accurate labor market data.
I want to commend you btw for not veering into Conspiracy Land. Most everyone on the Right it seems thinks the BLS is being purposely inaccurate.
If you manage a system that everyone relies upon, but which you know to be hopelessly inaccurate at economic turning points … and you fail to fix it … that’s a form of purposeful inaccuracy.
The nature of the game isn’t so much to lie outright, as to mislead through “excusable” obfuscation. So long as the marks are misled you can make a profit, but by not lying you can sleep at night and stay out of jail…
You’re assuming there is a fix available. Some problems are just difficult.
So much of the country’s payroll is controlled by a few companies…couple that with known tax payments and data from services like Venmo, and the government absolutely SHOULD be able to generate more timely and accurate data by now. They enjoy being able to use the fudge factor for their benefit.
whatever jobs there are in the USA they aint in chip making
https://asia.nikkei.com/Business/Tech/Semiconductors/China-buys-more-chip-tools-than-South-Korea-Taiwan-U.S.-combined2