New nonfarm payrolls rose by only 29,000, well below expectations, while the unemployment rate climbed to 4.175% and average hourly earnings growth slowed to 3.025% year over year.
On the establishment survey side, September added just 29,000 positions (construction, manufacturing, education and health services, and leisure and hospitality performed relatively well, while information, financial activities, professional and business services, and government were weaker), far short of the roughly 90,000 consensus estimate. August payrolls were revised down from 162,000 to 133,000, and July was revised from 21,000 to -10,000, for a cumulative downward revision of 60,000. The three-month average fell from 51,300 to 50,700. Average hourly earnings rose just 0.132% month over month, with the prior reading revised up to 0.32%, while the annual pace slowed to 3.025%, also below expectations.
On the household survey side, the unemployment rate came in at 4.175% versus 4.141% previously, slightly above the 4.1% forecast, though the labor force rebounded modestly. The participation rate edged up to 61.8% but remains at a low level. In September, total full-time employment rose by 88,000 month over month and part-time employment rose by 205,000, yet full-time employment is still down a cumulative 839,000 for the year. As a comparison to nonfarm payrolls, ADP private employment excluding government rose by 90,000 in August, with a three-month average of 57,000, highlighting that nonfarm payrolls are more volatile across months.
Overall, the softer June and July nonfarm readings rebounded above expectations in August, but September fell short again. On one hand, this suggests the labor market has not been meaningfully boosted by AI investment and is not accelerating. On the other, because of cross-month adjustment factors, nonfarm payrolls are oscillating between beating and missing expectations, which is a statistical issue. After the Fed began raising rates in September, equities rebounded, but the Treasury market rapidly priced in a hiking cycle and ramped up expectations for an October move. However, following this week's Fed official comments and the jobs data, the urgency of an October hike looks limited, though the Fed still has reason to hike in December given elevated energy prices and relatively strong nominal growth.
Cross-month payroll swings do not disrupt the view that the labor market is near full employment
The weaker-than-expected September labor data reduces the urgency of an October hike but is unlikely to stop the Fed from raising rates again in December. From the current employment picture, slowing wage growth and a low participation rate continue to constrain aggregate consumption, but the Fed's framework focuses mainly on the unemployment rate, and persistently weak labor supply means the jobless rate will struggle to rise even without an improvement in labor demand. By employment structure, both goods-producing and service-providing job creation declined in September, and government employment weakened notably. In goods, construction added 11,000 jobs on the back of AI investment and manufacturing added 9,000. In services, traditional sectors such as education and health services (up 20,000) and leisure and hospitality (up 10,000) remained resilient, but payroll gains in information, financial activities, and professional and business services kept shrinking. Government shed 17,000 positions, mostly nearly 11,000 local government non-education jobs, though the decline in government employment is unlikely to persist.
AI lifts investment and GDP growth but does not appear to add many jobs
If manufacturing and construction are treated as AI-driven sectors and information, financial activities, and professional and business services as AI-disrupted sectors, it is clear that AI's net contribution to nonfarm payrolls has stayed clearly negative since 2025. In September, AI-driven sectors added 20,000 jobs, but AI-disrupted sectors shed 26,000. Since the third quarter of 2026, AI-driven sectors have contributed 89,000 jobs while AI-disrupted sectors have lost 70,000, roughly offsetting each other. From the Fed's perspective, although AI has a positive overall effect on the economy, it does not significantly accelerate labor demand. Therefore, after the September hike, it makes sense to watch core inflation and economic data, and there is no need to rush another hike in October.
Nonfarm payrolls are swinging across months due to statistical issues
The August upside surprise does not point to a marked acceleration in the labor market, and the September miss does not point to a sharp deterioration in employment; the overall picture remains stable with a slight softening bias. After the BLS adjusted its model early in the year, monthly nonfarm payroll volatility has been extreme, with subsequent revisions also large. The August and September seasonal adjustment magnitudes similarly reveal cross-month adjustment problems. Because adjacent survey months can span four or five weeks, the BLS seasonal adjustment aims to correct this incomparability, but data volatility and methodological changes in recent years have made 2026 nonfarm payrolls swing sharply. We therefore still believe it is more reasonable to read the overall employment trend using the three-month average of nonfarm payrolls alongside ADP data. Currently, ADP employment, the ISM PMI employment subindex, and high-frequency demand indicators suggest no major improvement in labor demand. The three-month averages of ADP and nonfarm payrolls roughly indicate monthly job gains of 50,000 to 60,000 since the start of 2026.
The small rise in unemployment does not support an October hike but will not stop the Fed from considering one in December
The unemployment rate rose to 4.17% in September from 4.14% previously, slightly beating expectations, but it has not risen to a range the Fed would consider "dangerous" and is unlikely to prevent another hike this year while inflation risks (crude oil and food) persist. This month's rise in the jobless rate came alongside an increase in the labor force, but a higher share of those added were unemployed, pushing the rate slightly higher. Breaking down the structure, the main contributor to the higher unemployment rate was an increase in unemployed people entering the labor market from the supply side (re-entrants and new entrants). Judging by the Fed chair's "reaction function," a stable unemployment rate still means the labor market is near full employment. Even with weak September payrolls and a clear year-to-date decline in full-time employment, this is still interpreted as reduced supply caused by tighter immigration policy rather than a risk of economic weakening. On that basis, the softer September labor market only confirms that an October hike is not necessary, but it will not push the Fed off a path of moderate tightening.
Wage growth remains on a slowing track, limiting the risk of a major core inflation de-anchoring
AI's dependence on imported supply chains also means the risk of broad price increases driven by stronger domestic demand is limited. Former Fed official Clarida once noted that core inflation converges toward labor costs over the medium to long term, and current labor cost growth (such as the ECI) is roughly consistent with 2.0%-2.5% inflation, although the Fed chair views wages as a lagging indicator that may not reflect inflation risks in real time. We believe that, judging by the structure of new employment, AI has not broadly and substantially boosted manufacturing, which is tied to the fact that US AI investment relies on imports from supply chains in South Korea, Japan, China, and elsewhere, making net exports a drag on real GDP growth. In addition, relatively stable wage growth means that, if energy prices ease and food prices do not significantly exceed expectations, inflation should slow markedly in 2027, at which point the Fed may not need to raise rates further.
In sum, we believe the below-expected September nonfarm payrolls and slightly above-expected unemployment data show that no rate hike is needed in October, though this view awaits confirmation from September CPI. Based on the labor market and other demand indicators, the Fed will still not enter a sustained, aggressive hiking cycle like 2022; two to three hikes, including the one in September 2026, remain a reasonable decision. Our estimates of AI's boost to productivity and the natural rate of interest show that roughly two to three hikes are enough to push the nominal rate into restrictive territory. Considering that the five-year average TFP growth estimated by Fernald in 2024 was less than 0.8%, even if TFP growth rebounds significantly above 2% from 2025 to 2027, the corresponding five-year average should be slightly above 1.2%, roughly matching the two-sided HLW estimate of the natural rate. Under a very optimistic AI assumption, the neutral federal funds rate in 2027 would be around 3.75%-4.0%; if the inflation center is also assumed to run above expectations in the short term, the neutral rate in 2027 would be around 4.25%-4.5%, corresponding to about one to three hikes since September 2026. If AI's boost to TFP is less optimistic than above, then even with a 2.5% inflation center, the two hikes penciled into the September FOMC dot plot would be enough to push the federal funds rate above neutral, which is also our baseline assumption.
Below-expectation nonfarm payrolls helped equities and pushed the dollar index lower, but the bond market stayed cautious and yields edged up. After the September labor data, the market cut the implied probability of an October hike to about 22.7%. Two Fed officials had also been actively guiding the market to reduce pricing for an October move, so there is currently little reason for a hike in October, and watching the data trend is more appropriate. In markets, after the labor data came in clearly weaker than expected, Treasury yields and the dollar index fell while equities rebounded. The dollar index ultimately dropped to 101.9303, closing below 102. The three major US indexes rebounded, with the Dow up 0.49%, the Nasdaq up 1.19% and hitting a new intraday high, and the S&P 500 up 0.73%. However, after an initial decline, the 10-year Treasury yield rose 3.2 basis points to 5.273%, and the 2-year yield rose 3.33 basis points to 4.825%, showing the bond market remains cautious about a single month of weaker economic data. In our analysis of the September FOMC meeting, we argued that although long-end Treasury yields are unlikely to stay high for long, short-term oil prices, food prices, and better-than-expected economic data could all trigger a short-term spike in yields. While nominal and real yields on long-end Treasuries are starting to look attractive for long-term allocators, the turning point for yields may not have arrived yet. In the short term, the front end of the curve, where hike expectations are more fully priced, still offers a higher probability of gains.
Risks
(1) The risk that tighter immigration policy leads to a further decline in labor supply. (2) The risk that US inflation rebounds as geopolitical issues flare up again. (3) The risk of errors across different statistical series and methods for the labor market.