An employee on the production line at the General Motors assembly plant in Arlington, Texas. The overall US payroll employment was ratcheted down by more than 400,000 jobs because of an annual “benchmark revision” Image Credit: Bloomberg

The generally positive news from Friday’s US employment report was tempered a bit by the fact that overall payroll employment was ratcheted down by more than 400,000 jobs because of an annual “benchmark revision” — the biggest such downward revision since the last recession.

It wasn’t tempered that much: The 422,000-job revision for December reduced employment by just 0.3 per cent. It also wasn’t exactly news: The Bureau of Labor Statistics announced the size and broad contours of the revision in August. The revised data do allow us, though, to see more clearly what has been happening in a few industries that have been going through changes lately, such as oil and gas, retail, and restaurants. But first, a brief explanation of why these revisions are being made in the first place.

The monthly payroll employment numbers are derived from a survey of 142,000 businesses and government agencies. (The unemployment rate, which is announced at the same time, is derived from a different survey of 60,000 households.)

This “establishment survey,” as it is known, tends to miss newly formed establishments, which the BLS attempts to correct for using something known as the “birth-death model.” But the BLS doesn’t always get it right.

Happily, the agency has a fallback data source to check its numbers against: The administrative filings that employers have to make with the state agencies that administer unemployment insurance, which cover about 97 per cent of all wage and salary civilian employment in the country. These data are released with about a four-month lag, and for various reasons can’t be relied upon to track month-to-month changes in employment. So the BLS makes an annual revision based on the March unemployment-insurance data, adjusting all the other months’ data using a “linear ‘wedge-back’ procedure,” which you should definitely not try at home. Sometimes these revisions are upward, but for last March it was negative 514,000 jobs.

And now, on to those industry numbers I promised. The industry supersector for which the BLS reported the biggest downward revision in percentage terms was mining and logging. Now that we have the detailed data, it’s clear that the revision was mainly about oil and gas. Everybody already knew that 2019 was kind of a tough year for the frackers, but their retrenchment is clearer now.

Coal mining saw a downward revision, too, but not a large one. The big coal-related news from the benchmark revisions was that the BLS is going to discontinue several earnings and hours-worked series for the industry because there just aren’t enough coal miners anymore (that is, they “no longer have sufficient sample to be estimated and published separately”).

For all their current political salience, mining industries just don’t employ all that many people, so the biggest downward revisions in numerical terms were in the much larger supersectors of retail, leisure and hospitality, and professional and business services. In retail, the downward changes were pretty widely distributed — yes, department stores saw a downward revision, as one would expect given all the dire headlines about them, but so did the non-store retailers supposedly to blame for their travails.

In leisure and hospitality, the downward revisions do appear to have been concentrated in one industry: full-service restaurants, which have been booming over the course of this economic expansion, but booming less than previously thought in 2018 and 2019.

In professional and business services, legal employment was actually revised slightly upward, and management and technical consulting employment barely changed. But temporary help services, often seen as an economic bellwether, saw big downward revisions for late 2018 and early 2019, although it appears to have been recovering a bit lately.

[In the publishing sector] software publishing, which includes games, turns out to be the only industry in this category that has been seeing jobs gains in recent years. So no, I do not think this revision indicates a great newspapers/magazines/book revival.

What the revisions overall seem to indicate is that the US economy really did hit something of a speed bump in late 2018 and early 2019, one that may become more apparent in gross domestic product data as those go through their own benchmark revisions this summer. But it seems to be cruising along again now.