Puzzling “BLS Wonderland” report does not signal a strong labour market: A deep dive

The recently released BLS jobs data has not only surprised to the upside but smashed through expectations.

The Nonfarm payrolls (NFP) rose 339K in May 2023, against expectations of 190K.

‘Tis the time for much cheer. Or is it?

The report has received plenty of positive coverage, with Jim Lebenthal, Chief Equity Strategist, Cerity Partners, stating,

This was pretty darn good. This was pretty darn good.

It wouldn’t be the first time that an employment situation report outperformed industry expectations.

Far from it, in fact.

According to a May article published by ZeroHedge, the BLS report outperformed industry expectations for a record 13^th consecutive month.

Of course, this now stands at 14.

Despite the unending flurry of layoffs in a year which saw three of the four largest financial meltdowns in US history, trepidation amid the possibility of widespread banking contagion, and of course, the Fed’s unprecedented tightening, the jobs numbers don’t seem the least bit rattled.

So, the jobs market has been doing far better than anyone expected, but is it so straightforward?

Employment situation fundamentals

The BLS publishes the Employment Situation report on the first Friday of every month.

Quoting from a January 2023 article published on Invezz,

The BLS’s employment situation report publishes two different indicators – employment level and nonfarm payrolls (jobs), which each come from separate surveys – the household survey and the establishment survey, respectively. The two measures do not imply the same thing.

Breaking down the difference, we have employment measures,

Roughly, employment refers to persons (16 years or above) who have worked 1 hour or more over the week in which the survey was conducted. This could be in a professional organization, their own company, or a farm. In the case of a family enterprise, unpaid workers are also included if they work more than 15 hours during the week. Employment also includes workers who were temporarily absent from work under a variety of scenarios including vacation, illness, personal issues, and a host of other reasons. The key point is that while tallying up the employment number, each worker is only counted once, even if they are working in multiple jobs.

Regarding the NFP,

On the other hand, nonfarm payrolls drawn from the establishment survey theoretically count all US workers leaving out farm employees and some other categories such as proprietors, private household employees and unpaid volunteers. However, this number is vulnerable to double counting, especially if workers are employed in multiple jobs at once.

Definitional distortions

It should be noted that even the employment level may be hiding the true picture, given that in many cases the respondents need only have worked one hour to qualify.

Those who do not work and have given up on looking for jobs are thought to have dropped out of the labour force, reducing the unemployment rate.

Looking for work online via ads or portals does not count which means such people would not be included in the labour force either.

Regarding the NFP data, the BLS notes,

…multiple jobholders are counted for each nonfarm payroll job.

Mike Shedlock, a well-known economic blogger, and registered investment advisor, who referred to the latest numbers as ‘The BLS Wonderland’, noted,

These distortions artificially lower the unemployment rate, artificially boost full-time employment, and artificially increase the payroll jobs report every month.

Grave divergences

During May, the rosy NFP data saw a rise of 339K in May 2023, while employment fell by 310K.

That’s right.  

The employment level in the household survey came down by pretty much the same amount as the improvement in the establishment survey.

Shedlock has written about the divergences between jobs and employment for nearly a year, pointing out that the headline number is not the full story.

The following graph shows the month-on-month changes in employment levels and total NFPs.

The NFP which is susceptible to double counting has continued to rise despite the difficult economic conditions.

The change in employment, on the other hand, has been much more variable and in some cases, has shifted in the opposite direction to the change in NFPs, complicating the labour picture.

Relying solely on the NFP figures especially may be masking these divergences.

Unemployment rate and labour participation

Despite the surge in NFP this month, the unemployment rate has increased rather sharply from its historic low of 3.4% to 3.7%, even though labour participation has remained flat at 62.6%.

The historic lows in the unemployment rate and strong NFP results have not been in alignment with the news we have been hearing about job losses across the tech sector.

Although tech layoffs have moderated in recent months, May numbers were still above the 2022 average of 13,725 at 14,555.

In addition, manufacturing jobs appear to have peaked registering a decline of 2,000 in May 2023.

This was backed up by last week’s JOLTs report according to which manufacturing quits fell to the lowest number since October of 2020, signalling the unwillingness of workers to exit their jobs.

Regarding the health of manufacturing and what it tells us about the labour market, Ronald-Peter Stöferle and Mark J. Valek wrote in their In Gold We Trust report (2023),

…the number of overtime hours worked has declined sharply year-on-year, which has been a reliable recession indicator in the past. In times of restrictive monetary policy and economic slowdown, companies try to cut costs by reducing staff or cutting working hours. Weakening orders also favors this managerial behavior.

Interested readers can read detailed commentary on the report here.

Moreover, the JOLTs report also shows that in aggregate, workers are relatively unwilling to quit across total nonfarm jobs, implying worker anxiety and a weaker labour market.

The total nonfarm quits rate declined to 2.4% in April 2023 compared to peaking at 3.0% in April 2022.

Revisions

The most troubling aspect of the employment numbers has been the magnitude of revisions.

In the case of the NFP, at the end of 2022, revisions due to the benchmarking process and seasonal adjustments were colossal.

Even more striking was the employment data which improved by 894,000 in January 2023, of which a mammoth 810,000 turned out to be revisions from 2022.

To add to the complexity, unlike the NFP data, the US government does not report which months were modified and by how much, preferring an aggregate number for the year without giving much context.

2023

In 2023, revisions (two per report) have lowered NFP estimates by 45K, 63K and 19K in January, February, and March, respectively.

Thus, the first three months had a job creation that was 127K lower than initially reported.

However, looking at the second March revision in the May report, the NFP was raised by 52K while the April data increased by a strong 41K, for a total of 93K more jobs than first reported.

Other than the size of the revisions, there are concerns that the sample sizes are too small with the employment survey interviewing 60,000 households and the establishment survey compiling data from 122,000 businesses and government agencies.  

Hours worked and wages data

In May, average hourly earnings were up by 0.3% MoM while the average workweek fell to 34.3 hours from 34.4 in April.

Prima facie, this reduction isn’t in line with the optimism around strong NFP figures either.

Further, the all-important wage data is often updated in subsequent reports too, with Investing.com noting 9 such revisions since January 2022.

U-6

Although the unemployment rate (U-3) has risen to 3.7%, the most comprehensive measure, U-6, was recorded at 6.7%.

U-6 includes unemployed persons, as well as those ‘marginally attached’ to the labour force and part-time workers who would prefer a full-time job.

Similar to the U-3, a definitional underestimation of the labour force may be resulting in under-reported U-6 levels.

Birth-death model

The birth-death model reported by the BLS shows the birth and death of corporations, as a source of the change in the labour market.

A SchiffGold report notes,

Given the economic uncertainty, it’s hard to believe that enough new businesses formed in May (to support such a strong report)

In regards to the model in general, Shedlock noted,

The model is wrong at economic turning points and is also heavily revised and thus essentially useless.

Outlook

The key question is whether the employment situation report is reflecting the reality of the labour market.

The degree of revisions, as well as the limited sample size, may raise questions about the representativeness of the data.

This is especially important for the Fed, which has been placing a great deal of emphasis on employment data to determine the interest rate trajectory.

The latest data suggest that the manufacturing sector may be slowing down, while tech layoffs have come thick and fast over the past year.

In December 2022, I wrote a piece regarding Danielle DiMartino Booth’s concern of a ‘compressed lag effect.’

With the rapid tightening over the past year, could we still see this lag effect play out, and if so, is the labour market up to the challenge? Can we be certain that the labour market is as strong as reported?

Authors at Schiffgold are not optimistic, noting,

That’s when the Fed will rush to step in with liquidity. They already did it on a small scale with SVB. Expect the next one to be much bigger.