Let’s start with a discussion of poor employment data, but the problems don’t stop there, or even in the US. Poor data is international. 
I got the above chart without the annotations from Michael Green, Chief Strategist Simplify Asset Management, a few days ago.
As you can see, the third revisions are negative for all of 2023.
I was interested in how this data looked over time so I created my own spreadsheets as follows.
Employment Revisions Thousands of Persons
That looks like a random, unusable mess. However, revisions are clustered as a 6-month sum of revisions shows.
Employment Revisions Six Month Rolling Sum
Employment Revisions Six Month Rolling Sum Detail
The Fed is making decisions on baseline data that is increasingly unreliable with wild swings in revisions.
It’s much worse than it looks because the chart does not include annual revisions.
National Current Employment Statistics March 2023 Revisions
The BLS announced those revisions in the CES Preliminary Benchmark Announcement on August 23, 2023.
In accordance with usual practice, 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 2024 with the publication of the January 2024 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 a downward adjustment to March 2023 total nonfarm employment of −306,000 (−0.2 percent).
Revision Double Normal Size
This revision was double the normal size. But it will not reflect in the data until February 2024. Then what?
Let’s ponder the prior annual benchmark revision.
It was so huge I comment on it every month in my monthly jobs report.
Jobs Rise by 187,000 But 110,000 Negative Revisions and Unemployment Soars by 514,000
Payrolls vs Employment Gains Since May 2022
- Nonfarm Payrolls: 4,377,000
- Employment Level: +3,185,000
- Full Time Employment: +1,446,000
- Only 45.4 percent of the employment gains for the last 15 months was full time employment.
But who knows how accurate any of that is?
On September 1, I commented Jobs Rise by 187,000 But 110,000 Negative Revisions and Unemployment Soars by 514,000
Accounting for negative revisions, jobs effectively increased by 77,000 while unemployment surges as people looking for work can’t find it. Bloomberg labeled this “Goldilocks”.
Full time employment is down by 150,000 since January of 2023.
Here is a statement I have made in my jobs report every month since February.
Poof, Here You Go
Of the 894,000 rise in employment in January, 810,000 was due to annual benchmark revisions. And the BLS does not say what months were revised, just poof, here you go.
Again, we cannot, with strong confidence, suggest these reports portray an accurate picture of either jobs or employment.
The BLS does not say what months were revised. It just dumps huge revisions into the current bucket, without adjusting the prior data.
The annual revision for 2022 was a massive 810,000 but God only knows how.
I have been working on this idea for several days. The Wall Street Journal beat me to a similar headline today by hours.
Bad Data Everywhere, Not Just the US
Please consider the WSJ article The Problem With Economic Data Is Getting Worse emphasis mine.
Data revisions are occasionally so big that they upend our shared understanding of what’s going on.
The latest example comes from the U.K., where it turns out the economy grew much more than previously estimated. Rather than being the sick man of Europe with GDP still smaller than before the pandemic, and the weakest recovery in the Group of Seven industrialized nations, it has beaten Germany and grown in line with France (at least until their figures are revised).
The problem isn’t new, but it’s getting worse. Timely data is typically based on surveys, and even companies can’t be bothered to fill in surveys any more.
This isn’t only a problem for investors in the U.K., where the $3 trillion economy is almost 2% bigger than previously thought.
In the summer of 2011, the labor market appeared to have stalled. The Bureau of Labor Statistics reported zero job creation that August, and the Obama White House withdrew proposed new clean-air rules, citing the need to support the economy. The Wall Street Journal’s report on the figures was headlined “Job Growth Grinds to a Halt” and quoted an economist warning that “clearly, the economy is unwinding.” Stocks fell more than 2% on the day.
The timely data that drives markets on a day-to-day basis has become less reliable. The response rate for the survey of 651,000 businesses behind the payrolls data was steady at about 60% before the pandemic, but has since dropped to just 42%.
It’s even worse for the important Job Openings and Labor Turnover Survey used to work out how many vacancies there are and how many people are quitting their jobs. Investors have followed it fanatically over the past two years as a gauge of the labor market. The response rate has more than halved to just 32%.
WSJ Nonfarm Revision Chart
That chart matches the one I had been working on, shown above. I have detail dating back to 1964.
Let’s now discuss massive discrepancies in the data we do have.
Negative Revision to 2nd Quarter GDP, Huge Discrepancy with GDI Continues
On August 30, I commented Negative Revision to 2nd Quarter GDP, Huge Discrepancy with GDI Continues
GDP vs GDI Chart Notes
- Real means inflation adjusted
- GDP is Gross Domestic Product
- GDI is Gross Domestic Income
- Real Final Sales is the bottom line assessment of GDP. It excludes inventories which net to zero over time.
GDI was negative for two consecutive quarters and has been weaker than GDP for four quarters. GDI is now positive, but it is subject to greater revisions than GDP.
Two Measures of the Same Thing
Bear in mind that GDP and GDI are two measures of the same thing. Income should match products sold.
The last three quarters of GDP are +2.6%, +2.0%, and +2.1%.
The last three quarters of GDI are -3.3%, -1.8%, and +0.5%.
Philadelphia Fed GDPplus Measure Sure Looks Like Recession Started in 2022 Q4
GDPplus is a measure of the quarter-over-quarter rate of growth of real output in continuously compounded annualized percentage points.
It’s a blend, but not an average, of Gross Domestic Product (GDP) and Gross Domestic Income (GDI). It is much smoother than either GDP or GDI as the above chart show.
In 100 percent of the cases, with no false signals, no misses, and no lead times more than two quarters, every time GDPplus had two consecutive quarters of negative growth, the economy was in recession.
And except for one negative print of a mere -0.1 percent, the economy was in or would soon go into recession as soon as the first negative GDPplus number surfaced, and stuck.
Of course, the fly in this ointment is revisions. The key words above are “and stuck.” But we have already gone through two revisions of GDI and thus GDPplus and those revisions were negative,
GDPplus Recession Signals
GDPplus Recession Signals Synopsis
- GDPplus signaled every recession
- GDPplus was on time 4 times, early by a quarter 3 times, and early by 2 quarters twice.
For discussion of the advantages of GDPplus, please see Philadelphia Fed GDPplus Measure Sure Looks Like Recession Started in 2022 Q4
People believe what they want and certainly Biden along with mainstream media is touting GDP.
But assuming we have enough negative revisions to employment under our economic belts, and those revisions are in the correct direction, the GDI/GDPplus version of the story is far more likely the accurate portrayal vs GDP.
Months or years after the facts are revised, if GDPplus is accurate, the NBER may determine the US went into recession in the second quarter of 2022.
And all that lording over the UK for Brexit by the “remoaners” now appears to be laughably false. Germany is worse off.
Central banks are making decisions off this terrible data.
Bernanke censored the Ph.Ds. that worked for the FED. You used to get an honest response from communicating with the FED’s technical staff.
It’s time for the states to hold a constitutional convention and eliminate the fed and jerk a knot in the federal government’s tail.
The problem would be worse, for the fact that since 1980 hourly wages haven’t kept up with inflation, the Fed’s been the band-aid to cut rates to free disposable incomes.
Across the board we need laws on bribery (“money is free speech”), as long as the private sector continues buying the rule of law, we’re screwed, with or without the Fed.
The highest bidder owns the law, we’re the best Democracy money can buy.
.
The methodology The Fed employs for determining what two plus two is; is to poll a bunch of dilettante “investors,” and then fitting the responses to some form of curve.
That’s what “Data Driven Economics” mean.
Whether their so called “data” is “timely” or not, is pretty far down the list of their issues.
I hate conspiracy theories, that said, I highly doubt the Fed, with the immense resources at their disposal is “accidentally” missing a thing.
My distrust of all things government came to the front after 2011, when Bernie Sanders leaked 2008 CFTC records showing it was domestic (TBTF) banks (Goldman Sachs, Citi, Morgan Stanley, Koch Bros, etc) that had manipulated oil futures to the extent that it restricted disposable incomes spawned the onslaught of sub-prime defaults.
They knew what they were doing.
The CFMA (of 1999) allows complete secrecy in futures/derivatives, these banks were short sub-primes at the same time they were selling them fraudulently as “AAA” assets, they manipulated commodities (mainly oil) and – voila, Goldman Sachs was paying all-time record bonuses to it’s prop desks in 2009.
Full circle 2023- My inner foil-hat wearing self wonders if lumber prices were partially manipulated in 2021, in turn drove up home prices, in turn the Fed increased rates dramatically to counter “inflation”.
Bottom line, C-suits despise higher wages, since the inception of the CFMA and removal of Glass-Steagall in ’99, we are sitting ducks, left scratching our heads at bizarre price fluctuations, but more importantly as with myself, left frustratedly pondering who paid who to manipulate what, growing more and more distrustful of my government and its wealthy bribers.
So, while politicians have currently resorted to “scandalous” accusations about the individuals across the aisle, and we bicker, infight and name-call, NOTHING has been done to stop the real problem in politics, money buying rule of law.
Bribery is NOT free speech.
.
All that can be said for sure is that there is definitely all manner of skullduggery afoot, and by random luck, some of us have guessed what some of it is. We don’t know which ones, or which guess is correct though.
In the absence of evidence we will likely never see, its all conjecture.
Not to worry. They say Aslan is on the move- perhaps has already landed.
As long as they put enough money into the right confidential Super-PAC’s, the war is won before it even starts.
It absolutely blows my mind that Super-PAC’s aren’t required to disclose their funders.
.
The Grease Must Flow…
Well written commentary it speaks of thoughtfulness.
lobbying has been a part of democracy since the Roman Bread company 2000 plus years ago. my god some of you people have no grip on reality of civilizations.
You work with the data you got and not the data you want but can’t get.
A callback to the OG Donald… my favorite mass murderer.
The Fed might stay put, facing UAW strike and gov shut down. Higher
unemployment rates a lower gov spending.
I’ve brought this up before. To be refuted because 1Q GDP was revised up. So, one data point is revised up while all the others are down and that’s supposed to even things out. Even mobsters give police good leads from time to time to make them look on the up and up.
My guess is those who initially reported 1Q GDP screwed up. They didn’t realize they could have told the truth.
The inflation stats are probably even worse for accuracy. The are lies, there are damn lies and then there are statistics. I do believe that the economic politburo does deliberatly understate true inflation to keep interest payments on their debt lower and decrease COLA adjustments.
Are you pleased?
Anyone that has a legal job has to have a social security number. Every social security number is associated with a zip code on file. There are social security deductions that happen every week, bi-week or month for every single employed person depending how they get paid. If the Fed wanted to know what real-time (or as close to it) employment was they would just work out some sort of data reporting from SSA down to the zip code level.
Why this has never happened when it’s the easiest and most viable way of knowing employment numbers as real-time as possible is beyond me. I don’t know why the BLS doesn’t just use that data set, what is the point of surveys and other nonsense when there is digital data?
The only problem with your idea is that not all jobs pay into Social Security. Many government jobs (which is now a HUGE percentage of jobs overall) do not because they have their own pensions so they’d have to look elsewhere for some data.
https://www.ssa.gov/pubs/EN-05-10051.pdf
Also many other jobs (selling stuff on EBay, buying/selling stocks for your own portfolio and living on that money etc) don’t have social security number requirements either though they definitely pay taxes and are part of GDP.
It’s definitely not perfect but would be better than the nonsense we get now. And the whole point is to get trends not exact numbers. If last month there were 400k social security numbers with activity in zip 10001 and this month there is 300k we can calculate the reduction in employment for that zip. Do that for every zip and you get a pretty good analysis in near real time.
It’s no different than Nielsen ratings or voting samples. As for ebay and uber drivers, how do you think those people are counted now? How do you call an ebay seller and ask them if they are employed?
My preferred method now is just to watch what companies are doing. Walmart is lowering wages for new hires, that’s a big indicator the labor market for those roles isnt that tight anymore. I’ll keep an eye on Apple and if they announce layoffs then we know the economy and job market is definitely softening.
Ebay has to report any income greater than $600 in a year to the IRS and requires your social security number to sell on ebay.
You are correct that it does.
But that doesn’t help figure out what amount of money that person is making since they still don’t pay into social security etc. All that number does is let them file and pay taxes at the end of the year.
If they add up all the other 1099 and W2’s they can get gross over-the-table income. Benjamin Franklyn does a surprising amount of the business of America though.
They can just look at tax withholding instead. But, it wouldn’t matter because they would adjust things for the headline. Tax withholding numbers have been trending bad by the way,
Because they do not MEAN it to be honest and concise. That is not their goal.
IT IS CONCOCTED to “please.”
How much of bad data sampling methods is by design? National layoff data was suspended when numbers got really bad.
It was? Can you post a quick reference/link? I never heard anything about that, I’m curious to read about it.
He might be confusing the suspension of data lockups pre-release with the release of the actual data. The early release/embargo of data was suspended in June of 2020 or so. The actual data wasn’t suspended (at least in the case of the JOLTS separations data).
Personally do not believe any numbers that come out from the government, Numbers will always meet the needs of who it benefits in the White house, not the true reality of what is happening.
Thank you. NOTHING is real. Look at how they ‘Concoct’ CPI data, removing components that are PRICE inflationary.
The Green Movement. The Wars. The Promises ALL broken.
I used to think: “Why do they not try to hide this shit?”
NOW we know that it is AS IT IS MEANT TO SEEM: THEY DO NOT CARE what we think in reality and only purvey to FASCIST interests, those that serve lobbyists.
There is NO people’s LOBBY, is there? If you find one, let me know when they are taken seriously.
It’s hard to create a solid case for intentional bad data. Depending on your own political leanings, you can think of motivations for more favorable data or more pessimistic data, but it’s tougher to think of who might consistently benefit from *poor* data. More to the point here, the quality of nearly *all* data that’s necessarily dependent on survey or sample information has declined over the last couple of decades, mostly due to simple distrust of institutions and decline in the level of respect that’s accorded to subject matter experts. You see a lot of the symptoms of that on this blog (which I enjoy, but damn!)
Having known some of the people who put this stuff together, they’re mostly career statisticians, analysts, and economists just trying to put in an honest day’s work for a salary amount they could probably beat if they took their skills to the private sector. Most of them take their jobs extremely seriously and would consider the question moderately off-putting.
Have you ever worked for the federal government? Everything is political. The only thing people care about is keeping their job. The heads of BLS and BEA are no different. If they don’t do what they’re told, they might lose their positions and they have final say of every number.
correct. worked in gov in my youth putting together demographic and cartographic andscientific data for NASA and NYS. the plebes doing the grunt work, are just sheep. the decisions and final say are political hacks.
Here’s an example of intentional bad data in the CPI: Owner’s equivalent rent. It undercounts housing costs BY DESIGN.
Who benefits if CPI is understated? Who loses?
OH, WOW, really: HOW can you attach that WONDERFUL WORD, “HONEST” to any of these outcomes?
Never ascribe to malice that which can be explained by incompetence.