- by New Deal democrat
Before I get to the main point at hand, let me make a quick note about this morning’s existing home sales report for January: it was more of the same. Sales remained within the sideways range they have been in for nearly the past three years; prices were nearly flat YoY, up only 0.3%; and inventory was above its post-pandemic levels but well below pre-pandemic levels.
But on to the main course. I have seen a surprising amount of commentary on the Seeking Alpha investment site that yesterday’s jobs gain of 131,000 for January means that employment is on the upswing, completely neglecting that one month does not make a trend, that revisions have been relentlessly downward, and that January is perhaps the most difficult month for the BLS to accomplish seasonal adjustments — in January there were 2,649,000 layoffs! It’s just that the adjustment mechanism expected even more.
By far more important for the trend, and in particular the trend for the leading indicators within the jobs report, were the revisions to the past 12+ months of data. And as we saw from the headline adjustment, it was very large and very bad. So let’s start there, and then go through the most important leading sectors and metrics.
The total adjustment over the relevant data was nearly a -1,000,000 jobs. For the year 2025, the adjustment was just over -400,000, causing a previous 584,000 gain to turn into only a 181,000 gain, for an average of only 15,000 jobs added per month:
Even worse, for the eight months from April through the end of the year, a grand total of only *12,000* jobs were added, and no that’s not a typo. That’s 1,500 jobs per month! That’s about as close to recessionary as you can get without actually being in one (that we know of, at this point).
So let’s turn to the leading sectors, starting with manufacturing. There a -166,000 decline through December since Mary 2024 turned into a -251,000 decline, before its very slight 5,000 increase in January:
Next, here is construction. There, a slightly increasing trend throughout the year that added 14,000 jobs turned into a declining trend through October that ended up with a net -1,000 decline for the year:
But even the rebound since October disappear when we look at the even more significant residential construction sector. There, an increase through March followed by a slightly declining trend thereafter, resulting in a -1,400 decline for the year turned into a nearly relentless decline since March 2024 that ended with a -12,900 decline during 2025:
The entire rebound in construction was because of nonresidential building construction (and asociated specialty trades, not shown below):
Through October of last year revisions added 3,700 jobs, and then 12,000 more since.
For the goods producing sector as a whole, the -90,000 decline from its April peak through December turned into a nearly relentless-184,000 decline from a peak in July 2024 through last October, before increasing 49,000 since (again, all due to nonresidential building construction and associated specialty trades):
In short, *all* of the leading employment sectors of the economy declined during 2025. The only significant leading indicator in employment that remained postive was the average workweek in manufacturing, but even that did not improve:
Finally, let’s turn to aggregate nonsupervisory payrolls. We won’t know what the “real” number was for January until we get tomorrow’s CPI report, but since there was a nominal 0.8% gain last month, it is likely the “real” number will be positive as well. Here the revisions subtracted roughly -1% from the previous trend, but retained an almost identical positive trajectory:
Decomposing the metric, revisions indicated a -0.6% decline compared with the previous index for aggregate hours worked:
But the previous vintage showed a 0.7% gain for the year, which was reduced to 0.4%, but still a gain.
This further compensated for by a 0.2% increase in average hourly earnings over the year ending in December:
In other words, while the absolute number for aggregate payrolls was revised downward, the upward *trend* remained intact. That, along with the intract trend of increased average weekly hours in manufacturing, were the sole leading positives that came out of the benchmark revisions. All of the others were negative.
This feeds into the dominant “K-shaped economy” narrative which I believe is correct: the AI data center boom has led to a stock market boom, which - aside from being a likely source of the increase in nonresidential construction employment - has been feeding a “wealth effect” increase in spending by the top 10%-20% of consumers.
