Underlying these revisions is that the birth-death model used by the BLS is badly flawed. My calculations are roughly in line with the estimates of Bloomberg’s chief economist.
This is a complex post detailing some of the more arcane steps the BLS uses to compile its monthly employment report.
But keep reading, I’ll put it all together after I explain the data I used in this post.
Additionally, we have confirmation from Bloomberg’s chief economist that he has given us similar numbers.
BED Note
- What is BED? Business Employment TrendsThis is about 9.1 million private companies It will be very accurate.
- The BED report accurately shows how many jobs are being created through new businesses and how many jobs are being lost through business closures.
- BED is Quarterly Employment and Wages Census QCEW covers 11.6 million businesses
- The issue with BED is timeliness. On July 24, the BLS BED Data for Q4 2023 (October, November, December).
- There is a nine-month wait from October to July, and an even larger lag for QCEW data.
Notes on births and deaths
- To avoid the large time lag of BED, BLS Birth and death model Estimate the net job creation from new and closure businesses.
- The BLS explains that “net birth and death projections are not seasonally adjusted and are applied to the seasonally unadjusted monthly employment estimates to derive the final CES employment estimates.”
- CES stands for Current Employment Statistics and is part of the monthly employment statistics report.
- The BLS does not calculate and does not know how many jobs the birth-death model adds or subtracts from the monthly employment report.
CES Notes
- The CES monthly employment report is about 670,000 companies 9.1 million in BED and 11.6 million in QCEW.
- Because the CES does not have up-to-date data on business openings and closures, the BLS uses a birth-death model to estimate them.
- The monthly employment report summary says, “Do not subtract reported births and deaths from the reported headline figures. That method is statistically invalid and highly inaccurate,” yet I see people doing that calculation every month.
overview!
In normal times, when we are not in or out of a recession, birth-death models are primarily subject to random fluctuations.
Monthly employment reports are not as accurate, but there are errors. Typically It will balance out over time.
Approaching a recession is a different story, but the last recession was one for the record books.
It is important to note that both the BED and birth-death model figures are produced by the BLS and are supposed to measure the same thing: the creation or loss of net new jobs by businesses.
BED Net Employment vs. Birth and Death Model
The COVID recession caused a huge number of businesses to close, and then six months later a huge number of businesses opened.
Let’s throw all that away and look at what has happened from 2022 to now.
For convenience, we repeat the lead chart.
BED net employment vs. birth and death model net employment
Chart Details (Both figures are taken from the BLS)
- Total for Q2-Q4 2023: 484,000 jobs
- Total births and deaths Q2-Q4 2023: 1,263,000
- The difference between the two reports is 779,000 cases.
Here’s a second opinion.
Statement by Anna Wong, Chief Economist at Bloomberg (on mail)
Labor-market observers have split into two major camps over the past year: those who see a recovery that justifies the Federal Reserve’s high-interest-rate stance for longer, and those who see momentum fading and that interest rates may need to be cut sooner.
Bloomberg Economics takes the latter position, and has increasingly taken that stance lately: We believe the labor market weakness has been evident for some time, especially in our real-time data on business creation and closure.
The Bureau of Labor Statistics’ fourth-quarter employment trends report confirmed that nonfarm payroll data had been overstating the strength of the labor market for nearly a year.
Thus, we believe the labor market is not just approaching a turning point, but has likely passed it. If the Fed waits until September to cut rates (which is still our base case), it will probably be too late.
The Nonfarm Payrolls report is one of the most timely labor market indicators and one of the most closely watched economic releases. The data is collected through the BLS’ Current Employment Statistics (CES) program. However, the tradeoff for timeliness is that it relies on estimates of net employment changes due to business openings and closures, and its true value is only known over very long time lags.
This creates the possibility of revisions to the CES at key points in the economic cycle. For example, The 2023 CES-based payroll data will not fully reflect the underlying net business closures until early 2025.
For more timely information on business openings and closures, we turn to two other BLS reports: the Quarterly Employment and Wage Survey (QCEW), which provides the most comprehensive employment data of any official survey, and the Business Employment Trends Series (BED), a longitudinal subsample of the QCEW. Both of these provide a much broader coverage of businesses than the CES survey, which covers only 670,000 establishments.
Based on this data, we estimate that nonfarm payrolls reported in the establishment survey are likely to be revised downward by 730,000 in the final three quarters of 2023.
QWEC establishment counts have declined sharply since mid-2023. BED data for Q4 2023 supports that assessment.
The reason the CES birth and death drivers are overstated is because corporate mortality data are available with a delay and the predictive equations for those drivers are retrospective. This is heavily influenced by the increased rate of entrepreneurship during the pandemic..
Looking beyond 2023, we used Chapter 11 bankruptcy filings and business formation filings to estimate what net job creation from business openings and closures would have been in the first quarter of 2024. The results suggest that the downward trend continues. While bankruptcy data is not available for the second quarter of 2024, business formation data for the second quarter of 2024 is even cooler, down 5.3% from the fourth quarter of 2023.
Bottom line: The labor market has been weak for some time, and the downturn is not sudden. Given its dual mandate, the Fed will likely be slow to cut rates, which is why it expects the unemployment rate to hit 4.5% by the end of 2024..
Pending Negative Amendments
Let’s compare Anna Wong’s overvalued estimate of 730,000 with my calculation of 779,000.
The 730,000 indicates that the BLS employment report overstated jobs by 81,111 per month over a nine-month period.
Additionally, does anyone believe that the total number of births and deaths in the fourth quarter of 2024 will be 653,000?
That’s the highest quarterly total since at least the first quarter of 2017. And we won’t know how wrong this number is until the end of 2025.
Hiring is much weaker than expected, unemployment rate rises
Counting downward revisions, June saw unexpected weakness across the board, particularly in private sector and manufacturing payrolls.
Jobs are taken from the CES monthly survey described above, and employment and unemployment are taken from a separate BLS household survey.
The discrepancy is large, but BED and QCEW confirm which is accurate.
But every month, economists get obsessed with bullshit jobs statistics.
On July 5th, I Hiring is much weaker than expected, unemployment rate rises
Comparison with job vacancies from one year ago
- Non-farm employment (blue): +2,611,000
- Employment (red): +195,000
- Full-time employment (yellow): -1,551,000
BLS Nonfarm Payrolls Previous Revision
On July 7, I asked How much trust do you have in the BLS Jobs Report?
Let’s take a look at the revised BLS employment report for January 2023 through June 2024 in four charts.
There have been only three second upward revisions in 18 months.
The post above looks at the QCEW data in a bit more detail than the BED subset and estimates that there are around 900,000 pending revisions.
These point-by-point revisions may or may not reduce the numbers, but a staggering 653,000 births and deaths in the second quarter are not included in the QCEW data, and we will have to wait until late 2025 to know what actually happened.
The recession has already begun
We reiterate our July 8, 2024 recession prediction. Weak data suggests the recession has already begun. Now let’s discuss when the recession will start.
Enough already. The recession is here. Let’s start the discussion with a very good indicator that has few false positives and no false negatives.
“Hell will break loose” as recession deepens in coming months
For more information on the Second Opinion on the Recession and other charts, please see “Hell will break loose” as recession deepens in coming months
