U.S. Job Loss Should Get Policy Fix After Downward Revision
Parsing a harsh downward revision in the latest jobs data.
You've reached your free article limit
You've read 0 of 1 free Pro articles.
We just had downward revisions of 911,000 to the jobs data from Q2 2024 to Q1 2025.
My understanding is that a large portion, even the biggest contributor, to the revisions came from overstatement of the birth/death model. That is the model that uses EIN applications and other inputs to attempt to estimate how many jobs were created by new business formation (birth) and how many jobs were lost by businesses that closed (death).
Over time, with additional data (primarily state unemployment information), they can better analyze what actually happened and "true up" the data to reflect information that has come in since the model estimates.
For Perspective
The Q2 2024 to Q1 2025 period had an NFP figure of 1.8 million and birth/death figure of 1.2 million (I don't think the entirety of the birth/death counts in the NFP data as there are adjustments made, but it is not a horrible starting point).
On this simplified metric (again not all of the unadjusted birth/death model jobs make it into the final number), it would indicate that 67% of jobs came from the net of new business formation.
We have argued time and time again that in a gig economy, EIN applications don’t mean what they once did.
EIN applications once represented people who were starting a business, often as sole proprietors, with several employees.
With the gig economy, we believe that a lot of individuals set up EIN(s) for their gig economy job. Some, maybe just one, but some, possibly more, so they can track their various gig jobs separately.
There is even reason to believe that many of the EIN applicants, in a gig economy, continue to work and are just nervous about their prospects. This shows up in the JOLTs QUIT rate (low) and Law School Applications (high).
From April to August 2025, cumulative NFP is 265,000. The birth/death model says 963,000 (seems a bit extreme, but there are seasonals). Knowing that there are adjustments to how many of the birth/death model jobs show up in the final data can make a big difference. For instance, if all of the birth/death model findings were determined the be inaccurate and therefore discounted, the economy since the start of Q2 could have total jobs as low as -700,000 for the period. Of course, that is an overstatement, but it puts things in perspective.
That is a bit extreme, but given the size of the write-off of jobs for the prior period, which came close to the total amount of birth/death jobs (I don’t think that is completely coincidental), we should be worried how overstated current jobs are.
To me — and this is simplistic, but seems reasonable — new business jobs should be in line with existing business jobs (directionally). If the economy is adding jobs, then new businesses are likely being created because times are good. If no one is hiring, how realistic is that people are starting new businesses and hiring? Or are more likely doing something in the gig economy?
For the same April to August period in 2024, NFP was 560 and birth/death was 1,000 (a bit less skewed than this year’s data, but maybe a clear indication of how many issues there are with the model).
Also, our other “favorite subject” has been seasonality. Overstating jobs in the start of the year, based on old seasonal patterns, that no longer exist.
Given the magnitude of last year’s revisions, the linkage to birth/death model overstatement and the disproportionate amount of birth/death versus actual jobs since April, if I was a policy maker, I’d assume the U.S. is currently losing jobs and set policy accordingly.
They probably won’t do that, but they should seriously consider it.
Also, I toss out OER and use Zillow or Cleveland Fed real time rent figures, and take down the inflation fears a notch or two!
I'm bullish bonds and flatteners here.
