The February payroll estimate for the Establishment Survey was 242,000 with upward revisions to December and January, so we are told the labor market is surging once more. Last month when the Establishment Survey suggested “only” 151,000 (before revisions) it was taken as disappointing even though there is statistically no difference between 151k and 242k (a 90% confidence interval leaves out a lot of, well, confidence in the exact number). The saving grace last month despite the “weak” headline number was the purported surge in wages, which was immediately seized upon as the long awaited signal for “oveheating.” In February, though, where job gains have been taken as inordinately better, wages collapsed. Statistics.

In fact, average weekly earnings (combining wages and hours) fell by the most in the relatively short record of that series dating back to 2006. That suggests, strongly, the wages gains from January weren’t and the wages disappearing in February still more confirmation of the dangers of statistical over-interpretation. There are seasonal artifacts (and bias defects, really) that are on full display in these numbers, but that only makes it into the mainstream on the one side.

Last month for the January payroll report, one Reuters reporter wrote:

U.S. employment gains slowed more than expected in January as the boost to hiring from unseasonably mild weather faded, but rising wages and an unemployment rate at an eight-year low suggested the labor market recovery remains firm.

The same reporter this morning suddenly notes “calendar quirks” in the wage data even though the same calendar quirks were present in January’s estimates on the way up:

The only blemish in the report was a three cent drop in average hourly earnings, but that was mostly because of a calendar quirk.

That $0.03 drop in wages translated weekly earnings from January’s $878.15 to just $872.04 in February, again the biggest reduction on record (and even worse as if wages were really growing the relative change is not compared to last month but what February growth “should” have been). In short, “calendar quirks” were seasonal adjustments that in January boosted the wage rate though that wasn’t part of the story, but statistics being what they are only means that those adjustments have to be “given back” the next month unless there is actual growth; making them the story on the back half when it is that latter summation that truly counts. The seasonal adjustments for January across a large range of economic accounts were similarly robust (retail sales most conspicuously), meaning that the behavior of wage estimates here are a likely preview of what those February seasonals will be.