Federal job fakery hits peak balderdash: the January 2022 payrolls report

On the morning of 4 February 2022 the White House warned that the January payrolls report would be weak because of the Omicron wave. Bank economists had forecasts ranging from low to negative. The number came out at 467,000 jobs added, four times the average forecast and double the most optimistic one. Wages were up 0.7 per cent on the month and 5.7 per cent on the year. Bond yields broke through a ceiling that had held for two years, on the reasoning that a booming labour market meant the Federal Reserve could tighten as hard as it liked. This page recorded how the number was made, because the making was more informative than the number.

The raw count fell

The unadjusted payroll count, the actual number of people on payrolls, fell from December to January by 2.8 million. That is not unusual; it falls every January as holiday hiring unwinds, which is what seasonal adjustment exists to remove. What was unusual was the size of the adjustment. December’s total was adjusted down and January’s adjusted up by enough to turn a fall of 2.8 million into a rise of 467,000.

At the same time the Bureau of Labor Statistics published its annual benchmark revisions. November’s gain, originally 249,000, became 647,000. December’s, originally 199,000, became 510,000. The last five months of 2021 were revised up by more than 800,000 jobs in total, while months earlier in the year, which nobody was watching any more, were revised down by more than that. The whole apparent strength of the labour market going into 2022, on which the Fed’s tightening was being justified, came from adjustments and revisions to numbers that had told a different story when first reported.

The contradiction

In the same report the number of unemployed rose slightly, to 6.5 million, and the unemployment rate ticked up a tenth of a point. A hiring boom and a rise in unemployment in the same month are possible, since they come from two different surveys, but the combination did not support the headline’s story of an economy running hot, and 6.5 million unemployed remained 800,000 more than before the pandemic. The number of people outside the labour force entirely, who are counted as neither employed nor unemployed, was the larger fact the headline left out.

Why it mattered that week

The report landed in a market that had just watched Meta lose 26 per cent of its value in a day, about 230 billion dollars, the largest one-day loss for any company in history, on an earnings report that showed its user growth had stopped, and had then watched Amazon rise 14 per cent overnight on its own. This page took both as evidence of a market pricing fantasy rather than earnings, which is the kind of market a recession corrects. Into that, the payrolls print told the Fed that it could raise rates without risk, and told the bond market to price for it. The Fed did, six weeks later.

The standing objection

The question asked here in earlier years, with end-of-year revisions of this size, was why anyone trusted the adjusted numbers at all, when up could be turned into down by revisions to the previous revisions, and the final figures typically arrived after March when nobody cared. The seasonal model, it was observed, seemed to have one setting for adjusting reports up and none for adjusting them down until the revision months later. The presidents who oversee the Bureau have an interest in the direction of the first print; the revisions come when the interest has moved on.

Later note. The January 2022 figure was itself later revised, to 504,000, then in the following year’s benchmark down again; the pattern of large upward first prints followed by downward benchmark revisions became the subject of a formal debate in 2024, when the annual benchmark cut the prior year’s job gains by 818,000, the largest downward revision since 2009. The concern recorded here about the method was, by then, mainstream.