'Tis The Season For Lou Costello Labor Math

The January Employment Situation Summary Is Simply Garbage

As is apocryphally attributed variously to Mark Twain and Benjamin Disraeli, “There are three kinds of lies: lies, damned lies, and statistics.”

To that one must add the January Employment Situation Summary from the Bureau of Labor Statistics. It is in January where the BLS goes all out on the Lou Costello Labor Math to obfuscate, with its “seasonal corrections”, the true state of employment in this country.

Thus we can say with certainty that the lead paragraph of the summary is factually false.

Total nonfarm payroll employment rose by 517,000 in January, and the unemployment rate changed little at 3.4 percent, the U.S. Bureau of Labor Statistics reported today. Job growth was widespread, led by gains in leisure and hospitality, professional and business services, and health care. Employment also increased in government, partially reflecting the return of workers from a strike.

No, total nonfarm payroll employment did not rise by an unbelievable 517,000 in January. That is simply not true, and close inspection of the detail of the summary shows us that it is not true.

Of course, detailed inspections of narrative promoting government propaganda is anathema to the corporate media, which is how we get delusional headlines like this celebrating these nonexistent employment gains.

After all, why rain on President Asterisk’s parade as his handlers take an undeserved victory lap on Twitter celebrating job gains that never happened?

President Biden @POTUS

This morning, we learned the economy created 517,000 jobs in January. We also learned we did better last year than previously thought. Add that up, and we’ve created 12 million jobs since I took office. That's the two strongest years of job growth in history – by a long shot.

Since the corporate media will never do any actual thinking or analysis, and merely parrot the BLS propaganda report, let us examine all the ways that jobs number is 100% pure horse hockey.

As a starting point, we must understand that it is no coincidence that January is such a wildly flawed month for BLS jobs reporting. Those who have subscribed to my Substack for a while will recall that last January I had a similar condescension about the BLS Employment Situation Summary.

The Employment Situation Report: Lou Costello Math

This year we will begin delving into some of the reasons why this happens (whether those reasons constitute malice or stupidity will be a judgment I leave to the reader’s discretion).

Next, a reality check: The ADP National Employment Report missed some 411,000 of those jobs when it released its numbers a few days ago.

Private sector employment increased by 106,000 jobs in January and annual pay was up 7.3 percent year-over-year, according to the January ADP® National Employment Report™ produced by the ADP Research Institute® in collaboration with the Stanford Digital Economy Lab (“Stanford Lab”).

While any large-scale measurement has the capacity for error, are we really prepared to believe that ADP undercounted overall employment in this country by a multiple of four?

We should not be so cynical of outside sources.

A primary reason we know the lead paragraph of the Employment Situation Summary is wrong is that January has long been the month of statistical corrections by the BLS.

Straight away we can see that it is impossible for 517,000 jobs to have been created in just the one month of January when the BLS is revising and re-benchmarking all twelve months of the Establishment survey. That 517,000 figure is the number after the revisions have been implemented, which means in the best case scenario, the BLS jobs report is capturing all manner of employment not previously counted in the monthly reports.

Moreover, when all of the revisions are to the upside, that alone is cause for extreme skepticism.

The process behind these revisions is the BLS’ routine annual benchmarking protocol, which, in conjunction with its application of seasonal variances to the raw data, provides a measure of curve-fitting and smoothing to that data, ideally in order to filter out regular seasonal fluctuations in employment. Historically, that process more or less has worked over the longer term.

The raw employment totals show a clear seasonal pattern of fluctuations. The seasonally adjusted figures seek to filter out that pattern so that the underlying macroeconomic trends can be more plainly seen.

As a result, each January the unadjusted total employment numbers take a substantial downward revision, and this January has been no exception.

If we ignore the backdrop of the benchmarking process, the 517,000 jobs claim becomes even more absurd because it comes as the unadjusted data shows a whopping 2,505,000 jobs lost in January alone.

Perversely, that still makes January 2023 the best January in the post-pandemic era.

It is also important to note that the benchmarking is done to the Establishment Survey data only. The BLS does not revise the Household Survey numbers.

In accordance with usual practice, BLS will not revise the official household survey estimates for December 2022 and earlier months. However, to show the impact of the population adjustments, table B displays differences in selected December labor force series based on the old and new population estimates.

What the BLS does is provide separately the data which would make up the revisions were they to be so applied.

Thus of the supposed 894,000 persons added to the Employment Level in the Household Survey, 810,000 of that is a statistical “correction”, and the actual increase is 84,000.

It is important to note at this point that, late last year, the Philadelphia Fed did an early rendition of the benchmarking process, and determined then that the BLS stats for the second quarter were wrong, with the total jobs created during that quarter was approximately 10,000.

Philadelphia Fed Confirms BLS Is BS

The BLS uses similar adjustment methods to massage the unemployment numbers, making a seeming 1,028,000 increase among the unemployed become a 28,000 decrease.

In terms of presenting data accurately and faithfully, the BLS has been abysmal. While there is nothing intrinsically mathematically invalid about applying these seasonal corrections, the failure to mention them when discussing the 517,000 increase in the seasonally adjusted total payrolls figure on the Establishment Survey is tantamount to deception. The lack of such disclosures in that lead paragraph, of which one has to parse the rest of the jobs report to ascertain even their existence, is sufficient to render that paragraph factually false.

Yet we need not confine ourselves to critiques of the BLS statistical methods to realize the jobs number is beyond sketchy. If we include the ADP National Employment Report alongside the Establishment Survey and the Household Survey, we have three notionally independent surveys of employment. If the surveys are being honestly and appropriately done, the three surveys should show broadly similar results. While they did so at the beginning of 2022, in the last half of 2022, that changed, and the Establishment Survey became the clear outlier.

With few exceptions, the Household Survey and the ADP National Employment Report remain reasonably close together throughout 2022. After about July, however, the Establishment Survey began reporting numbers consistently higher than either the Household Survey or the ADP report. By December, the Establishment Survey had become a clear outlier—thus again demonstrating that the 517,000 number is a complete fabrication.

What was the actual number of jobs created in this country during the month of January? Based on the BLS Employment Situation Summary, we cannot plausibly say. We may be certain that it is not 517,000. It might be somewhere around the ADP figure of 106,000.

In truth, we do not know. And we are not going to be able to divine a reasonable figure from the data contained with in the Employment Situation Summary. By mixing actual changes with statistical revisions, the BLS has irreparably tainted the data, making it unusable.

What the BLS should do with their benchmarking process is release two summary reports in January—one with the benchmarking applied and one without. They should fully disclose in the very first paragraph the net effects of the benchmarking. That would be the proper way to present the data, which would allow for consistent interpretation and analysis across the benchmarking date.

Yet these are things the BLS does not do and likely will never do. They seem to prefer to release an Employment Situation Summary that is simply garbage.