We’ve detailed how the Bureau of Labor Statistics has been grossly overestimating the number of jobs in the economy, and now the Federal Reserve Bank of Philadelphia says the trend is continuing—at least it did continue in the third quarter of last year.
Each month, the BLS estimates the number of payrolls in the economy and then revises that figure over the next two months. These figures come from a survey of a relatively small number of businesses, but the data are later revised again once a year.
This annual revision process is known as the benchmark and covers a period from March of one year to March of the following year. In the last benchmark, about 600,000 jobs were wiped from the government’s statistical spreadsheets.
That means we never had the number of jobs we thought we did, at least not from March 2023 to March 2024. This is a pattern that has existed for a couple of years now, where initial monthly job reports consistently overestimate the number of jobs.
For the period included in the last annual benchmark, about one in four jobs that was originally thought to have been created never even existed.
But these benchmark revisions don’t simply come out of the blue. The BLS also publishes a quarterly census of employment and wages based on much more comprehensive unemployment insurance data. While this is not nearly as timely as the monthly job reports, it’s much more accurate.
This dataset is slightly different from that used in the monthly job reports, with each excluding certain categories of workers or employment, like government or railroad employees.
Still, the two sets of numbers trend very closely together with relatively small differences between them in most circumstances. That means we can impute from the quarterly census data what kinds of revisions are expected during the benchmark.
And you don’t have to do the math yourself if you don’t want to (we do it because we’re geeks) — you can let the Philly Fed do it for you. Once a quarter, they publish an “early benchmark” which is like a sneak peek on the upcoming BLS data.
The Philly Fed takes state-level data and adjusts the national statistics to develop surprisingly accurate estimates of the upcoming benchmark form BLS, on a quarter-by-quarter basis. Summing the four consecutive quarters (starting with Q2 of the year) in question will estimate the revised change far in advance of its official release.
Of course, the Philly Fed doesn’t release a payroll figure in thousands of jobs the way BLS does. Instead, it presents its findings as a percentage change, at an annualized rate.
That means it takes one more step to convert this seasonally adjusted percent change into the raw change in payrolls. Additionally, because the Philly Fed’s estimated change is an annualized rate only taken to the tenth of a percent, there’s a significant band within the minimum and maximum possible change because of the size of the labor force.
Nevertheless, the numbers from the Philly Fed have been so bad that even a best-case scenario represents a big drop in the next benchmark.
The numbers from the second quarter of last year were terrible, with the Philly Fed estimating payrolls fell 0.1 percent at an annualized rate, compared to the monthly job reports showing an increase of 1.6 percent. So, not only was job growth wildly overestimated for those three months, but net job growth was negative.
This past week, we got the Philly Fed’s estimate for the third quarter, and it was another ugly report. Although not as bad as the second quarter, it again showed job growth being overestimated, this time for the period from July through September.
The monthly job reports for the third quarter showed a 1.0 percent annualized rate of increase, and now Philly Fed’s early benchmark shows just a 0.7 percent increase.
Over the 12-month period ending with the third quarter, the initial monthly job estimates amounted to a 1.3 percent increase, while the sum of the state data showed a 1.5 percent gain. Meanwhile, the Philly Fed’s early benchmark showed only a 0.9 percent increase.
Fractions of a percent may not sound like much, but we’re talking an economy with hundreds of millions of people, which is why you end up with these revisions causing changes of tens of thousands or hundreds of thousands of payrolls.
So, the next benchmark revision is already shaping up to be a real doozy. Between the second and third quarters of last year (the first half of the benchmark’s covered period), the annualized rate of job growth was overestimated by 2.0 percentage points.
Even if the fourth quarter of last year and the first quarter of this year see no significant downward revisions to the Philly Fed’s early benchmark, we’re still looking at a massive drop to payrolls once the BLS releases the official benchmark.
How massive? Almost 1,000,000 jobs.
Once again, the payroll figures are going to prove wildly inaccurate, and we’ll only find out way after the fact. The next benchmark will cover the period from March 2024 to March 2025 (right now) and won’t be released until early 2026.
As we’ve pointed out plenty of times before, problems have been evident in the labor market statistics since the Spring of 2022, and nothing has been done to correct these errors. That’s not to say revisions are abnormal—they’re perfectly normal and standard operating procedure. What’s not normal is very large revisions that all go the same way, in this case, down.
Here’s hoping that the new administration cleans up the mess at BLS and makes some heads roll. Major decision makers from Wall Street to Washington rely on this data and the lack of confidence in the numbers will prove paralyzing if nothing changes.
For now, we just have to take all the BLS numbers with a grain of salt, remembering what the BS in BLS stands for…