Here are 8 charts and tables to discuss.
The lead chart compares the Quarterly Census of Employment and Wages (QCEW) to the Current Employment Statistics (CES) nonfarm payroll report.
Year-Over-Year QCEW and CES Nonfarm Payrolls Through March
- CES: +211,000 jobs
- QCEW+ 86,000 jobs
These numbers are down from over 2 million jobs in January of 2024, and down from over a million jobs in January of 2025.
Two Key Points
- The recent uptrend in nonfarm payrolls may be a mirage of sampling and non-sampling errors.
- Everything depends on whether the chart reflects a genuine trend change, is an arbitrary bump, or worse yet a mirage that will be revised away.
QCEW vs Nonfarm Payrolls
The above image from QCEW Concepts.
The CES survey is 629,000 worksites (with optional reporting) compared to mandatory responses by 12.2 million establishments.
CES Response Rates
The above image is from BLS Response Rates.
Comparison: QCEW consists of 12.2 million establishments with about 95 percent coverage.
In contrast, the Nonfarm payroll report surveys ~629,000 establishments with a response rate of 43.3 percent as of January 2026.
Discrepancies are inevitable. So, the BLS adjusts CES with annual benchmark revisions.
In July, the BLS released the preliminary revision for March of 2026 compared to March of 2025.
Annual Benchmark Revisions
After a series of huge BLS benchmark revisions based on the Quarterly Census of Employment and Wages (QCEW), the latest revision was small.
The BLS reports the Preliminary Benchmark Revision for March of 2026 (from March 2025) is -79,000 total jobs.
- The preliminary estimate of the Current Employment Statistics (CES) national benchmark revision to total nonfarm employment for March 2026 was -79,000 (-0.1 percent)
- The preliminary benchmark revision for total private employment was -178,000 (-0.1 percent).
Over time, the slope of the QCEW and the slope of CES will match, and that’s what the lead chart shows.
Month-Over-Month Comparisons
To make month-over-month comparisons, one needs seasonally adjusted numbers.
The QCEW should also list seasonal adjustments but it doesn’t.
Lack of month-over-month analysis makes QCEW even less timely.
I do not have a financial package to seasonally adjust QCEW, but I do have permission from John “Jake” Bush at Piper Sandler & Co. to post his seasonally adjusted QCEW data.
However, that data is only through 2024 for a post I did in 2025.
QCEW vs Nonfarm Payrolls Seasonally Adjusted Detail
QCEW vs Nonfarm Payrolls Seasonal Adjustment Notes
- Blue Line: QCEW SA from “Jake” Bush at Piper Sandler & Co.
- Red Line: BLS reported Nonfarm Seasonal Adjustments
- Green Line: I used BLS Seasonal Adjustment factors applied to QCEW
I am confident of using the BLS adjustment applied to QCEW, at least as a ballpark estimate. The following chart shows why.
QCEW vs Nonfarm Payroll Seasonal Adjustments
The blue line shows derived factors from Piper data vs derived factors from BLS numbers.
Curiously, QCEW seasonal adjustment factors are very close to BLS seasonal adjustment factors for October through March.
However, there are consistent directional flip-flops April through September.
Since March is inside the relatively safe window, we can make some comparisons.
| CES SA | QCEW SA | Gap | |
|---|---|---|---|
| Last Piper (Dec 2024) | 158,316 | 155,386 | 2,930 |
| Mish (green, Mar 2026) | 158,436 | 155,654 | 2,782 |
Understanding the Gap
- Nonfarm payrolls include some railroads and religious organizations that QCEW doesn’t.
- QCEW includes some agricultural households that CES drops.
- The net result is a constant ~3 million CES to QCEW discrepancy.
The gap itself is not important, it’s the slope of the charts. However, the BLS should modify nonfarm payrolls to eliminate this continual discrepancy.
Recommendations to the BLS
- Normalize QCEW to Nonfarm Payroll datasets
- Provide seasonally adjusted QCEW numbers
- Use actual tax data to augment CES numbers
Are Deportations and Surveys Impacting CES Numbers?
- CES is a weighted sample of establishments.
- If a firm doesn’t report, the BLS imputes from similar reporters (industry, size, region).
- If nonrespondents are weaker than respondents, imputations overstate jobs.
- The smaller, newer, and more heavily immigrant dependent firms are more likely to fear responding to any government form.
- To the extent points 2 and 3 are happening, the sample over-represents large, old, native-owned reporters, which were more stable in 2025–26.
- Also, birth-death adjustments would also miss new immigrant firms and be slow to kill disappearing ones, with the latter much more likely.
- Birth-death refers to the birth or death of corporations, not individuals.
That is nonresponse bias. Collection rates did collapse after Covid. And this remains a real CES issue.
Given poor response rates, the key question is who is it that isn’t responding? If it is firms leaning more heavily on immigrants, then the BLS is overweighting the rest.
| Channel | Hits CES? | Hits QCEW? | Verdict |
|---|---|---|---|
| Fewer immigrant workers | Lower payrolls | Yes, Same | Real |
| Fear, no-show, slower sites | Lower payrolls | Yes, Same | Real |
| Lower CES response at immigrant-heavy firms | Imputation from healthier reporters makes CES too high until benchmark | No | Plausible. Unmeasured with a lag |
| Birth-death missing deaths / extra births | CES too high until benchmark | No | Measured with a lag |
Deportations and outflows reduce actual payrolls and should show in both CES and QCEW.
QCEW lags by four to seven months, four immediately after a QCEW release (5 to 8 actually because nonfarm payrolls lag by a month).
It’s September, and the latest QCEW data is March.
The risk is that the CES sample leans on firms that still file the survey, which may over-represent the stable part of the universe.
I suspect this is happening, but there is no way to prove it, now. Proof comes later.
The Business Employment Dynamics (BED) report will catch birth-death sampling and non-sampling errors but with an even bigger lag.
| Report | Reference Period | Scheduled Release Date & Time | Official Source Tracking |
|---|---|---|---|
| BED (Business Employment Dynamics) | First Quarter 2026 | October 28, 2026 at 10:00 AM ET | BLS BED News Release Page |
| QCEW (Quarterly Census of Employment and Wages) | Second Quarter 2026 | December 2, 2026 at 10:00 AM ET | BLS QCEW Schedule Calendar |
I am skeptical there is a sustained trend change in nonfarm payrolls.
Unfortunately, BLS methods are such that we will not know until December, what really happened in June.
The Business Employment Dynamics (BED) report on which the infamous BLS Birth-Death model is based lags even further.
On October 28, we will find out how accurate the Birth-Death model was in March. And that lag is as good as it gets for BED.
Meanwhile, the BLS is using stale Birth-Death models for every monthly nonfarm payroll report.
Lather-Rinse-Repeat with Annual Adjustments. Lovely.
But hey, look on the bright side. We now know with good accuracy what actually happened in March.
And in December we will understand June.
Related Posts
September 4, 2026: Unexpectedly Strong Jobs Report, Payrolls Rise by 162,000 in August
Jobs Rebounded in August, Unemployment Rate Steady at 4.1 Percent.
September 7, 2026: Hello Farmers, Diesel New Record Near $6, Labor Day Gasoline Record Too
A week ago we were looking for $5.80 Diesel, now $6.00 is in sight.
September 8, 2026: Seven Forces Support High and Rising Bond Yields
There are national and international forces pressuring yields.
Eh i would not be surprised if the numbers were juiced up.
Is it really growth. With a high number of boomers retiring i would lean more to replacement. Maybe some immigrant replacement thrown in for good measure.
rubbish in, rubbish out. owned a bunch of different small companies in my first 66 years. i always just winged it with no correlation to reality. my favorite gov mandated form to mess with is the long form census form some are targeted with. i checked off every box and received a call from DC about it. that phone call itself was hysterical for me. i would have payed money to have explained how i really did not know if my father was the black garbage man or the chinese launderer or the mexican gardener and on and on and on. it’s incumbent on real men to fuck with an evil empire. even in they smallest and most idiotic of ways. imho.
AI private credit investment summer’s last hurrah: restaurant service sector, government, and medical and education service sector.
Diesel to hit $6.00 dollars/gallon today or tomorrow, and 7 – 8 dollars (maybe more) by midterms … B4NI.DE up 2.3% today equivalent to a 12 cent per gallon price increase.
After reading this employment reporting analysis, my takeaway is there are far too many opportunities to manufacture the desired outcome and obfuscate a true reflection of the employment situation.
A light upturn in jobs seems reasonable. It shows up in both Revelio’s number and in ADP’s. Also the data that’s available suggests the highest payoff to job switching in a couple of years, and it’s stronger at the hourly level. That’s something you might expect from a slightly tighter labor market. The bulk of the evidence suggests an inflection point, but durability of the pickup is a question mark.
Mish, your analysis on Payrolls and employment numbers keeps nudging me to ask: Since the numbers are AGED (Not new) or HARD TO ASSESS AT FACE VALUE and require YOUR work to “interpret’ the data – – nearly as strange as my point here – – it is like we are reading a FOREIGN LANGUAGE (Japanese or Chinese) and thus they WANT it murky, not comprehensible easily – – and thus their goal is achieved:
MURKY, STRANGE, IRRITATING and INCOMPREHENSIBLE. THIS is the goal of these numbers.
And, surely the next question arises (Esp going into an Erection in November – – oh, sorry, ELECTION): THEY ARE HIDING the terrible numbers that would lead the MSM to HAVE to not comply with the Propaganda Dept of the US Government (THE WH)….and continue to the bullshit that we are contending with as employers or even as retired folks.
I owned a company up until a few years ago where a similar survey would land on my desk to fill out. I would ignore it sometimes when busy, because it was a pain.
Eventually I would get a threat in follow-up letter informing me that a response was mandated by law. I could genuinely never figure out what the penalty would be.
I sold the company, but worked for the new owners for some months to help them transition to full autonomy.
I continued to receive the surveys and the associated threats although I no longer had employees. No matter how many times I responded with an explanation of why I didn’t have any need to respond, I continued to get threats and surveys. That persisted for months.
QCEW is mandatory but BLS BS is Not
Millions of people who could not get jobs because of the millions of illegals who were laying on their couches playing video games while living on welfare and voting for Democrats are now getting jobs as roofers, concrete workers, bricklayers, sheetrock workers, carpenters, plumbers, painters, etc., and they are grateful and will be voting for their savior in the midterms.
Funny! Long ago we used to have a poster who liked to twist everyone’s tail with similar satire.
They were pretty funny, too. They would catch the unwary, although responding with faux rage would be some next-level trolling.
You mean the Schrondinger’s Immigrant?
This guy is being truthful, why the downvotes?
Some immigrants are getting jobs in congress and doing nothing and getting paid lots of money and getting free healthcare.
The blue-collar scene is worse. Millions of immigrants would spend all day long doing drugs and robbing people, and also loitering at Home Depot, working for relative pennies doing day labor and taking our jobs.
Immigrants are bad.
Immigrants pulled up all the shrubs around my house, and got the cat pregnant.
“Millions of immigrants would spend all day long doing drugs and robbing people, and also loitering at Home Depot, working for relative pennies doing day labor and taking our jobs.”
You’re supposed to say “those pesky immigrants are stealing our jobs”, as if a job were like a wallet that is liable to disappear any time you leave it out unattended.
Come on, get with the program, sheesh.
Be alert!
Reverse Yen carry trade is coming?
Another possible reason for the falling CES response rate is that nonresponse itself may be correlated with economic stress. Firms cutting staff or restructuring are more likely to have reduced accounting/admin capacity, payroll disruption, or simply more urgent priorities than completing a voluntary survey. I think you point at this but not sure.
If so, the remaining respondents could systematically be healthier firms, creating exactly the upward bias you describe. It would be interesting to see whether BLS has compared subsequent QCEW outcomes for CES responders versus nonresponders. They could add this bias to their models.
correct