The annual revision of the U.S. non-farm payrolls is worse than expected, revised down by 910,000, increasing pressure on the Federal Reserve to cut interest rates

U.S. job growth in the year ending in March was far less than previously reported, increasing pressure on the Federal Reserve to cut interest rates.

According to preliminary benchmark revision data released by the U.S. government on Tuesday, non-farm employment for the year ending in March was revised down by 910,000, equivalent to an average monthly increase of nearly 76,000. This revision accounts for about 0.6% of the total labor force of 171 million. This is the largest downward revision since 2000.

Market expectations for this data revision were 682,000. Before the release of this data, several economists indicated that U.S. employment numbers for the year ending in March could be significantly revised down by nearly 800,000. U.S. Treasury Secretary Janet Yellen also warned that annual employment could be revised down by as much as 800,000. However, the final revision was even worse than the market generally expected.

By industry, employment numbers were revised down in almost all sectors. The sectors with the largest downward revisions included: leisure and hospitality – 176,000, professional and business services – 158,000, retail trade – 126,200, and manufacturing employment was also significantly revised down. Only transportation and warehousing, as well as utilities, saw slight upward revisions. Almost all revisions were concentrated in the private sector; government employment data was revised down by 31,000.

Before the release of this report, U.S. government data indicated that employers added nearly 1.8 million jobs in the year ending in March, unadjusted for seasonal variations, averaging 149,000 per month.

The non-farm annual revision report covers most of the time period before Donald Trump took office as U.S. President, indicating that the employment situation in the country had already deteriorated before he began imposing tariffs on U.S. trading partners. The data released on Tuesday will also be further adjusted when the BLS publishes the final benchmark revision in February 2026.

The previous year's non-farm benchmark revision covered the 12 months ending in March 2024, with an initial total downward revision of 818,000, later corrected to 598,000 in February 2025, still the largest downward revision since 2009.

Non-Farm Data Revision

Non-farm annual benchmark revisions differ from monthly adjustments in that their scope is much broader. Monthly adjustments primarily come from new survey data added by the BLS; annual benchmark revisions are based on quarterly employment and wage surveys, as well as tax data, which provide a nearly comprehensive recalculation rather than just incremental corrections to monthly reports:

The BLS's monthly employment report is based on two surveys. Benchmark revisions involve payroll employment numbers, which are obtained through business surveys; revisions do not affect the unemployment rate, as the unemployment rate is derived from household surveys.

Once a year, the BLS benchmarks March employment numbers against a more accurate but lagging data source—the Quarterly Census of Employment and Wages (QCEW). This data is based on state unemployment insurance tax records and covers nearly all jobs in the U.S.

Although non-farm employment benchmark revisions occur annually, this year has received more attention as investors and Federal Reserve observers are looking for signs that the labor market is slowing faster than previously expected. This adjustment has also impacted the political arena, as President Trump has previously criticized the revisions to employment data Last year, after a similar significant downward revision of the preliminary benchmark in early 2024, Trump also fiercely criticized former U.S. President Biden, questioning the integrity and economic performance of his administration.

The political and economic impact of this data revision could be quite significant. More signs of a weak job market will provide Trump with new reasons to advocate for interest rate cuts by the Federal Reserve. This annual revision of non-farm employment not only exacerbates concerns about the health of the economy but also raises questions about the credibility of labor statistics.

After the weak non-farm employment report in July this year, accompanied by a significant downward revision, Trump dismissed then-BLS Director Erika McEntarfer and nominated Heritage Foundation economist E.J. Antoni to take over. However, the non-farm employment number in August was lower than in July, and it was also revised down to a reduction of 13,000 jobs in June, marking the first negative growth since December 2020.

It should be noted that although Trump criticized the revisions, both monthly revisions and benchmark revisions are routine processes that update estimates based on more data. In recent years, the magnitude of revisions has been greater than usual, which some economists attribute to the unique economic dynamics post-pandemic. Several economists have stated that the initial employment data may be influenced by various factors, including adjustments for business openings and closures, as well as how undocumented immigrant workers are counted.

In recent years, most monthly initial employment data has been stronger than QCEW data. Some economists believe this is partly due to the so-called "birth-death model"—adjustments made by BLS to reflect the net effect of business openings and closures, which may have been skewed in the post-pandemic era.

Others believe there is another reason: immigration factors. Since QCEW is based on unemployment insurance records, and undocumented immigrants cannot apply for unemployment insurance, this data likely excludes thousands of undocumented workers included in the initial estimates.

Market Interpretation and Reaction

The adjustment by the U.S. Bureau of Labor Statistics indicates that the slowdown in the U.S. labor market in recent months has actually occurred after a period of cooling employment growth, which may lay the groundwork for a series of interest rate cuts by the Federal Reserve as early as next week. Federal Reserve Chairman Powell recently acknowledged that the risks facing the job market have increased, and two of his colleagues leaned towards rate cuts in July.

The revised data itself does not reflect the current situation, as this data can be traced back as far as a year and a half ago. However, data from recent months also shows a softening labor market. The average monthly increase in non-farm employment for June, July, and August was only 29,000, below the "break-even level" needed to maintain a stable unemployment rate.

Traders widely expect the Federal Reserve to announce interest rate cuts at the end of its two-day meeting on September 17.

In terms of market reaction:

• Spot gold briefly plunged, approaching $3,640, with the daily increase narrowing to about 0.3%. It had risen to $3,674.27 when the U.S. annual non-farm employment revision data was released at 22:00 Beijing time
• The yield on the 10-year U.S. Treasury bond rebounded from a daily low of 4.0359%, approaching 4.07% again, with an overall increase of about 2.5 basis points during the day; the yield on the two-year U.S. Treasury bond rebounded from a daily low below 3.48%, rising above 3.51%.
• The S&P 500 index rebounded from a daily low of 6494.64 points, rising about 0.2% and refreshing the daily high to 6507.10 points