August 23, 2024

This past week the Bureau of Labor Statistics released its preliminary estimate of the annual revision to the payroll employment data.  It caused quite a stir.   The BLS reported that in March 2024 there were 818 thousand fewer workers on the job than are being reported currently.  Some people argued that the BLS had covered up the data to make the economy appear stronger than it actually is for political reasons.  Nonsense.  But also there was a hint that some Wall Street firms obtained the data ahead of the official release time.  Maybe.  Maybe not.  Given that the Fed is now paying at least as much attention to the labor market as to inflation, the revised data created a sense that the Fed may be far behind the curve and should have begun its rate-cutting initiative months ago.  We do not buy that either.    The bottom line is that there is little reason to be concerned about the fact that there are 818 thousand fewer workers on company payrolls than the current data indicate.  Here’s why.

Regarding the BLS cooking the books to make the economy appear stronger than it is during this important election period, that is absolute rubbish.  If there were some sort of conspiracy, employment data from a variety of other sources would have to be in cahoots.  For example, in addition to the payroll employment data, there is civilian employment which is collected from a separate data stream and used to calculate the unemployment rate.  There is the monthly ADP report that estimates changes in employment.  There are weekly data on layoffs (initial unemployment claims) and the number of people receiving benefits.  If there is some sort of a conspiracy then all these other sources of employment would need to be involved.  This type of shot at the integrity of the BLS is unwarranted. The BLS is doing the best job in can in an imperfect world and it is totally transparent in what it is doing.  Every economic indicator released by every government agency and a variety of private sector firms is at least partially estimated and, therefore, wrong and subject to revision.  To figure out what is really happening to employment every month we should take data from all these sources and combine them to get an overall sense of the big picture for employment (or sales, or inflation, or GDP).  View every economic indicator with a healthy dose of skepticism.

Apart from the magnitude of the revision there was a concern that some Wall Street firms may have obtained a copy of the report prior to its official release.  That may – or may not – have happened.  Every month the BLS releases the employment report for the previous month on the first Friday of the following month at 8:30 a.m. (ET).   People anxiously await its release.  Regarding the release of the so-called benchmark revision, the BLS  indicated quite clearly that it would release its preliminary estimate at 10:00 a.m. (ET) on August 21, 2024.  The revisions are typically small and for that reason many economists do not pay much attention.  Over the past 10 years the average revision has been plus or minus 0.1%.  This one was a downward adjustment of 0.5%.  It got so much attention because it was much larger than normal.  It is possible that some Wall Street firms got the information early and the Fed needs to investigate whether that was, or was not, the case.  More likely, in our opinion, those Wall Street firms were simply paying more attention than everybody else.   But the real issue is whether 818 fewer jobs really matters much.  The short answer is, not really.

When we see the payroll employment data each month it is viewed as gospel.  It is not.  Some of the employment data each month are estimated.  The BLS collects data from most large firms, but it does not have the ability to survey millions of small firms.  Your local automobile dealership, restaurant, lawn care firm, and dry cleaner do not report employment data every month.  The BLS collects some information from small firms and then estimates the change in employment for the universe of small firms from that sample which may or may not be an accurate reflection of what is actually happening.  Hence, the monthly  payroll employment data are never exact.

But every year the BLS receives data from the survey of state unemployment insurance filings which contain data from virtually every employer in the country.  If those data are stronger or weaker than the existing payroll employment data the BLS adjusts the payroll employment data accordingly.  Hence, the annual benchmark revision.

So how important is the loss of 819 thousand jobs?  Not a lot.  Think about it.  Between March 2023 and March 2024 the current data tell us that the economy generated 2.9 million jobs or 242 thousand per month.  Now we learn that in actuality, the economy generated 2.1 million jobs in that period of time or 172 thousand per month. That is still a very healthy pace of job creation.  By comparison, in the one-year period leading up to the recession the economy generated 187 jobs a month.  In our opinion, the revision is largely much ado about nothing.

Keep in mind that what was just released are preliminary data.  The final revised data will be released in early February.  The revision could be larger, or smaller.

The other thing that will happen when the revised data are actually published in February is that the BLS will change the so-called seasonal adjustment factors.  The data that we see every month are adjusted to account for the typical change in employment for that month.  For example employment soars in November and December as firms hire thousands of temporary workers to help out in the business Christmas season.  Those workers are then laid off in January.  Rather than report that employment soared in November and December and then plunged in January the BLS adjusts for the “normal” change in employment for those months.  But the economy is constantly changing and those normal monthly changes shift over time.  So at the same time that the BLS will incorporate the 819 thousand fewer workers, it will also change the seasonal factors for each month.  That means that the monthly pattern of employment we see will also change as a result of this revision.  It happens that in the first three months of this year payroll employment soared by a whopping 307 thousand per month.  Without knowing anything, it is likely that when we see the new data in February 2025 much of the 819 loss of workers for the past year will have occurred very early in 2024.  That is ancient history.

In the end, in the most recent 3-month period payroll employment has risen 170 thousand per month.  When the final data are released in February we may find that employment in that time period was, perhaps 140 thousand.  It matters  a bit, but not enough to get super-excited.

Stephen Slifer

NumberNomics

Charleston, S.C.