BLS response rates plunged after COVID and never recovered, especially job market surveys. 
BLS Response Rate Notes
- CPI C&S: Consumer Price Index Commodities & Services
- JOLTS: Job Openings and Labor Turnover Survey
- CES: Current Employment Statistics
CES is the establishment survey, monthly nonfarm payrolls jobs report.
Sample Sizes
- CES: 70,000 Establishments
- CPI: 10,500 Establishments
- JOLTS: 7,200 Establishment
- QCEW: 12.2 Million Establishments Quarterly
- BED: 9.2 Million Establishments
QCEW stands for Quarterly Census of Employment and Wages. It represents 95 percent of the data.
BED stands for Business Employment Dynamics. It is a large subset of QCEW.
Sample Size Data
QCEW, BED, CES Characteristics
CES Response Size
- The above image says the CES Survey Size was 629,000 establishments. However, the monthly survey size is 70,000.
- Of that 70,000 only 43.0 percent respond. That’s about 30,100 out of 12.2 million.
- Thus, monthly jobs reports are based on a 0.25 percent sample of the data.
JOLTS Response Size
- The JOLTS sample size is 7,200 establishments.
- Of that 7,200 only 33.4 percent respond. That’s about 2,405 out of 12.2 million.
- Thus, monthly JOLTS reports are based on a 0.02 percent sample of the data.
Who Responded and Who Didn’t?
The surveys are voluntary. So who responded and who didn’t?
Smaller businesses are much less likely to respond to such surveys because they don’t have a dedicated employment office that deals with government BS.
In particular, any struggling or tiny business likely does not have time for this BS. The moment they see the word “voluntary”, the survey goes in the ashcan.
Every Month, All Eyes Are on BS
Nonetheless, every month, mainstream media dissects the monthly jobs report as if it means something.
I too produce a monthly report, but I also point out the flaws, the discrepancies between the surveys, and the poor quality of the data relative to QCEW.
Employment Drops by 355,000 But Jobs Rise by 227,000 in November
On December 6, I commented Employment Drops by 355,000 But Jobs Rise by 227,000 in November
The strange jobs reports continue as the divergence between jobs and employment widens again.
Job Stats vs One Year Ago
- Nonfarm Payrolls: +2,274,000
- Employment: -725,000
- Full Time Employment: -1,342,000
In the last year, nonfarm payrolls are up by 2.3 million while employment is down by 725,000 and full-time employment is down by over 1.3 million.
A Breakdown, by Sector, of the Negative 818,000 BLS Job Revisions
On August 22, 2024 I gave A Breakdown, by Sector, of the Negative 818,000 BLS Job Revisions
Those negative revisions are a direct result of the BLS Birth-Death model gone haywire.
Quarterly QCEW Data Provides More Evidence of BLS Jobs Overstatement
On November 20, I commented Quarterly QCEW Data Provides More Evidence of BLS Jobs Overstatement
Hard evidence from QCEW report suggests more negative revisions coming for BLS nonfarm payroll report.
The 2024 Destruction of Small Business Employment in Pictures
The BLS payroll reports smack of oversampling large employers and undersampling small employers where jobs have been trending lower.
For more discussion, please see The 2024 Destruction of Small Business Employment in Pictures
So they may as well take the L out of BLS.
If these are the surveys the government emails me, I always bin them too. If they want me to even consider filling something out they can send it to me US mail. The only survey I ever filled out they kept sending increasingly hostile US Mail about penalties. But I only have one employee and probably never will have more so I never saw the point.
That is great info and even better analysis
And I love this quote
“Smaller businesses are much less likely to respond to such surveys because they don’t have a dedicated employment office that deals with government BS.” We have known that for years just did not know how to quantify the meaning of that
Is there any evidence that bias (pro or anti one political party) exists in the way the data is reported? Is lawfare supplemented by datafare?
Just the evidence that this is all enabled by the cashfare operated by the professional bureaucrats that populate the system, irrespective of political parties.
This is a job for DOGE. If it is useless then get rid of it altogether.
Yes. Doge is useless, and will achieve nothing.
But if would be good if DOGE is successful. The government leviathan is inefficient, loaded with waste, and keeps growing.
We are all paying for this waste. At some point we will feel it, if we already haven’t.
There is no point in responding until the Trump purge has had time to work.
Didn’t work last time…
The government gets payroll data on every employee, each pay period – Federal withholding, social security etc. Why does the government rely on survey data when they get real data at least every two weeks. D.O.G.E. where you at?
Very few employers submit pay records or SS payments to the federal government every two weeks (talk about an administrative burden). They usually do so on a quarterly basis (if they are on time) – thus the better estimates from the QCEW.
So what happens to the 7-14% of every paycheck that’s withheld by employers every pay period? That doesn’t sit idle u til end of quarter, it goes directly to IRS.
The IRS has tons of data not being used by other agencies. Maybe for good reason, but it Is an inefficiency.
How hard is it to push a button on a computer screen that sends info from your accounting software to the government over the internet?
“For taxes reported on Forms 941, 943, 944, or 945, there are two deposit schedules: monthly and semi-weekly. Before the beginning of each calendar year, you must determine which of the two deposit schedules you are required to use. The deposit schedule you must use is based on the total tax liability you reported on forms previously during the specified lookback period. The lookback period is different based on the form type.”
https://www.irs.gov/businesses/small-businesses-self-employed/employment-tax-due-dates#:~:text=Federal%20income%20tax%20withholding%2C%20Social,you%20are%20required%20to%20use.
Yes, bank deposits for taxes collected are due more often. But the linkage to specific workers and their Social Security numbers occurs quarterly or even annually with federal tax returns – thus the difficulty tracking employed workers at a moment’s notice like everyone on this thread wants
https://www.irs.gov/businesses/small-businesses-self-employed/depositing-and-reporting-employment-taxes#:~:text=Depositing%20employment%20taxes&text=There%20are%20two%20deposit%20schedules,Forms%20941%2C%20944%20and%20945.
Every employee? So how many millions are not on the books, paid cash in hand. I’m sure there are plenty of small operations with a lot of cash turnover employing illegals to pick crops or students to spin pizzas. The shadow economy is how big? Is it 5, 10 or 15% of GDP and only very rough guesses might be made. Does it follow the rest of the economy in rises and falls or not?
And if 5, 10 or 20 million illegals are flushed into the open by crackdowns on checking SS numbers and other methods then will those illegals that go onto the books give an artificial boost to the official economy or will it be offset by X million illegals either being pressured to go home or being officially deported cause the official employment level to fall?
Thanks, Mish, I appreciate the statistical analyses.
How is this so different from the Presidential polls? You and so many others analyzed those weekly/monthly statistics and did not they usually poll hundreds of people to figure out how many of the 250+ million US adults would vote for either candidate? So why in your POV were they more accurate?
Atlas Intel did an amazing job parsing and interpreting responses and non-responses.
Revisions show BLS is making huge errors somewhere. I suspect multiple places, sampling and nonsampling errors.
Thanks, and fair enough. I would disagree with the general characterization of “huge errors” in the BLS data though.
October payroll estimates were revised upwards by 24,000. That looks big when compared to the original change point estimate of +12,000 (3x). But that’s a revision for a change of a change. As the BLS table within your article shows, that CES survey is supposed to be used to calculate estimates for total employment. So the +24,000 revision was a .015% ‘error’ rate when compared to 159,288,000 employed workers in the US.
Of course, the estimate could still be off. But when looking at the data that way – as the survey intended, a 43% response rate (for a total of 30,000+ establishments) is probably statistically significant – just like a country-wide Presidential poll of a few hundred people may be too.
This important aspect will not stop the hype that wall street uses to convince investors to trade.
… until the next 50%-90% market crash teaches another generation how dangerous it is to believe Wall Street hype.
Whoa, overvalued markets, yes.
But it sounds like you’re investing in prepper bomb shelters
Stock Markets in USA are at the kind of historically unprecedented valuations that always and inevitably unwind with 50-90% real-value losses. The only question now is when & how. 1929? 2002? 1970s? 2008? Something totally different?
Well it depends if one believes that the stock market has reached “what looks like a permanently high plateau” – yet again.
(With a nod to Irving.)
50% is within the realm of reason when you consider it will overshoot to the low side.
I found this on poking around; “If you remove the Magnificent 7 stocks from the S&P 500, the P/E ratio would be around 15.4x.”
That’s not crash bait. Nvidia certainly is though. Take profits and head for the bunker on that one.
I’ve noticed in some other articles that Profits* are well above the normal range as a share of GDP. Mean-reversion of Profits/GDP will bring down share prices as much or more than mean-reversion of Mag7 P/E ratios. Earnings stagnation also tends to pull down P/E ratios since expectations of growth get reduced.
* Profits, not Earnings: I don’t think we can call them all “earnings”, since given the current monopolistic market structure, a lot of corporate revenues aren’t genuinely “earned” anymore. GOOG, for instance, would be paying a lot more to Mish if there were genuine competition in the online advertising market…
Very true. I stopped responding
How big is your company – # of employees?
I was kidding.
Lying you mean … as usual.