Why This Was a Weak US Jobs Report

• Payrolls beat expectations due to one single sector—health

• Ex-health, payrolls continue to grind to a halt

• Ambulatory and counselling services drove the job gain

• Revisions wiped out 1.029 million jobs by year-end 2025…

• ...due to combination of annual March 2025 revisions, SA factor revisions since, and birth-death model changes

• The UR dipped, but with very low confidence in the household survey data…

• ...for multiple reasons, not least of which the use of outdated pre-Inauguration Day population counts before immigration policy tightened

• Weather impaired data collection and the quality of household survey estimates

• Hours worked continue to grow softly

• Wage growth was solid

• US nonfarm payrolls m/m 000s / UR %, SA, January:
• Actual: 130 / 4.3
• Scotia: 0.0 / 4.3
• Consensus: 65 / 4.4
• Prior: 48 / 4.4 (revised from 50 / 4.4)
• Two-month payroll revision: -17k
• Annual payroll revision to March 2025: -862k
• Total payroll revision to December 2025: -1.029 million

This is a weak overall report in my opinion and so I disagree with how markets interpreted the readings for a slew of reasons.

Markets only paid heed to the dip in the unemployment rate and the headline beat without fussing over details. The Fed should fuss over details that paint a continued picture of a very weak job market under the hood and fade the unemployment rate.

NOTHING UNDER THE HOOD

For starters, the nonfarm payroll gain of 130k drops to a gain of only 6.5k after excluding the 123.5k jobs created in the health and social services sector. If not for that one sector, I would’ve been very close. This extends the trend of very weak nonfarm payrolls ex-health (chart 1) though private payrolls ex-health picked up somewhat against a weak trend (chart 2).

Health care added 82k jobs driven by ambulatory health care services (+50k), hospitals (18k), and nursing and residential care facilities (13k). Ambulatory health care services include professionals offering out-patient services. Social assistance hiring was up 42k mainly in individual and family services (38k) which is defined as “primarily engaged in providing non-residential social assistance services for individual and families.”

So, more ambulatory service jobs and more counselling services employment buoyed job growth and that’s supposed to be a good sign for the economy and to markets. Righto, everything makes perfect sense.

In short, I blew my headline nonfarm call entirely because of this one sector as health and social assistance hiring was the strongest since August 2020 during the pandemic (chart 3). That’s curious in that the expiration of some ACA subsidies in December is likely to impair health’s role in hiring over the year even if not in this first report of 2026. I have no good explanation for why health hiring was higher than since the dark days of the pandemic. It even conflicts with the drop in Indeed job postings for health sector positions.

Private nonfarm payrolls gained 172k but only 48.5k after excluding health and social assistance and only 35k after excluding education.

Chart 4 shows the sector breakdown of the change in payrolls. Government fell 42k driven by the federal government (-34k). Most sectors were flat to lower.

REVISIONS WIPED OUT 1.029 MILLION PAYROLL JOBS

There were three types of revisions here that combined to remove 1.029 million nonfarm payroll positions by year-end 2025.

One was to March 2025 in the annual final benchmarking process that removed 862k jobs compared to the initially estimated –911k revision last September. Chart 5 shows the pattern of massive recent revisions to March payroll levels through successive annual benchmarking exercises. Recent years have tipped the scales on the size of the preliminary and final revisions which should have everyone seriously doubting the quality of the jobs numbers being produced by the BLS.

Second was that the BLS revised seasonal adjustment factors and this had the effect shown in chart 6 which portrays more downward than downward revisions from this effect. The BLS notes “Seasonally adjusted data from January 2021 forward are subject to revision. In addition, data for some series prior to 2021, both seasonally adjusted and unadjusted, incorporate other revisions.” 

Third was the customary two-month rolling revisions that in this case combined regular revisions with SA factor revision effects to December and November.

There were also birth-death model adjustments that removed 61k seasonally unadjusted jobs in January compared to no adjustments, -99k in December, and –31k in November with wild ups and downs before that.

PAYROLLS HIT THE SKIDS AFTER ‘LIBERATION DAY’

Where this leaves us is chart 7. The light blue line shows payroll levels up to December before this morning’s December data and all revisions. The dark blue line shows payrolls after this morning’s figures. The red line then takes the blue line and adds Chair Powell’s estimated 60k/month overcount of payrolls since April of last year. He offered that estimate on a lark in my opinion and we won’t know until September when the preliminary annual benchmarking revisions to March 2026 payrolls arrives.

THE DIP IN THE UNEMPLOYMENT RATE IS NOT TO BE TRUSTED

The unemployment rate dipped to 4.3% from 4.4%. At least I got that one right, but for wonky reasons that are tough to take credit for. The household survey generated a job gain of 528k in January versus a 387k rise in the size of the labour force. I just don’t trust that data.

For one thing, the BLS said that severe weather impaired collection of household survey data and pushed that survey’s response rate to 64.3% which was below average (chart 8).

For another, the 90% confidence band around the household survey’s estimate of the change in the number of unemployed is +/- 425,000. That means that the estimated -141k drop in the number of unemployed Americans last month was somewhere between -566k and +424k at the 10% level of significance. Right. The statistical confidence in the change in unemployment basically covers a decent sized US city of about 1 million people.

Third, as the BLS noted in the previous release for December, the annual incorporation of revised population counts that would have normally been incorporated in this month’s household survey won’t be incorporated until next month’s report. This adds to data quality question marks until we see that effort. For now, the household survey uses short-term projections of monthly population estimates derived from population adjustments “introduced in January 2025 (based on Vintage 2024 population estimates.” They go on to say “As soon as practicable, BLS plans to revise January 2026 estimates to incorporate the updated population controls.”

This could be key to qualifying the reported change in the UR and its drivers. A whole year’s worth of tighter immigration policy that was introduced after the January 2025 population counts before inauguration day may not be adequately reflected in today’s numbers.

OTHER DETAILS

The US had better keep up productivity growth. Hours worked were up only 0.8% q/q SAAR in Q4. They are very tentatively tracking 1.5% in Q1 based only on Jan and the Q4 average. GDP is hours worked times labour productivity (chart 9).

Wage growth was 0.4% m/m SA last month but revised down from 0.3% to 0.1% m/m in December. The result left the y/y rate unchanged at 3.7%. Chart 10 shows annualized m/m seasonally adjusted wage growth.