In two months, the total "downward revision" reached 260,000! Can we still trust the U.S. non-farm payroll data that keeps "slapping us in the face"?

Wallstreetcn
2025.08.02 02:18
portai
I'm PortAI, I can summarize articles.

The downward revision of 260,000 jobs in the United States is mainly due to adjustments in institutional data, with 40% stemming from sample corrections by state education departments. However, a deeper structural issue lies in the low response rate of surveys and the potential impact of the Trump effect. Both factors not only undermine the immediate accuracy of the non-farm payroll report but also make an already complex economic picture even more elusive

The U.S. Bureau of Labor Statistics has significantly revised down the non-farm employment reports for the previous two months, raising profound doubts in the market about the credibility of this key economic indicator.

On August 1, according to Wall Street News, the U.S. added only 73,000 non-farm jobs in July, far below expectations, with the data for the previous two months revised down by 258,000, marking the largest downward revision since the COVID-19 pandemic. Among them:

  • The May data was revised down by 125,000, from +144,000 to +19,000.
  • The June data was adjusted down by 133,000, from +147,000 to +14,000.

This massive "statistical reversal" has exceeded the scope of normal seasonal adjustments. Behind this: the continuously declining response rate to surveys, the resource challenges faced by statistical agencies, and potential policy changes are collectively eroding the foundation of non-farm data, making this highly anticipated report increasingly resemble a "draft" awaiting validation.

Divergence Between Job Openings and Employment Numbers

To understand the non-farm report, it is essential to recognize that it consists of two independent surveys:

  • One is the "establishment survey" aimed at businesses, used to generate the number of non-farm job openings we typically focus on. If a person holds two jobs, they will be counted as two job openings in this survey.
  • The other is the "household survey" aimed at families, used to calculate the unemployment rate. If a person holds two jobs, they will only be counted as one employed person in this survey.

The controversial downward revision of 260,000 job openings stems from adjustments in the establishment survey data. The U.S. Bureau of Labor Statistics stated that about 40% of the revisions came from the education departments of state and local governments, which "mainly resulted from the routine inclusion of additional or revised samples after the initial release."

However, a deeper contradiction lies in the fact that since 2022, the number of job openings in the U.S. has been continuously increasing, while the number of people employed has hardly changed.

(Since 2022, the number of job openings in the U.S. (red line) has been continuously soaring, while the number of employed people (green line) has hardly increased.)

Data Distortion: Survey Response Rate Sounds Alarm

The accuracy of the employment report is being challenged by a fundamental issue: fewer and fewer businesses and households are willing to participate in government surveys.

The Bureau of Labor Statistics conducts wage surveys of businesses three times a month, and as more samples are collected, the data becomes more complete However, in recent months, the sample response rate for initial surveys has repeatedly fallen below 60%, far lower than the pre-pandemic norm of about 70% or even higher. Omair Sharif, president of Inflation Insights, pointed out:

The more missing data there is, and the later it is obtained, the greater the likelihood of significant revisions. A 50% response rate is far from sufficient.

Underlying Causes: Budget Tightening and Erosion of Trust

Reports indicate that the decline in survey response rates is not a recent phenomenon; it is rooted in long-standing social and institutional factors.

On one hand, the public has long grown weary of cumbersome surveys, while trust in government and other institutions continues to erode.

On the other hand, statistical agencies themselves are facing increasingly tight budgets and staffing constraints, a problem that became particularly pronounced during the Trump administration. Economist Gregory Daco stated:

We see that cuts to government agency funding are affecting their ability to collect and analyze economic data. All reports conducted by the Bureau of Labor Statistics may experience greater volatility in the future.

This resource strain is not an isolated case.

Earlier this week, the Bureau of Labor Statistics stated that, on average, about 15% of the samples used to compile key inflation data (CPI) have been suspended from collection. As early as June, the agency announced the suspension of data collection in three metropolitan areas, citing that existing resources could not support it.

Potential Impact of the "Trump Effect"

In addition to the aforementioned reasons, some economists have proposed another explanation that may exacerbate data quality issues.

Derek Holt, head of Capital Markets Economics at Scotiabank, suggested in a report that the rapid changes in trade, immigration, fiscal, and other policies under the Trump administration may also be a contributing factor. He believes:

As businesses struggle to cope with the impacts of various policy upheavals, the already low survey response rates may further compromise data quality.

In addition to monthly rolling revisions, the Bureau of Labor Statistics also conducts an annual benchmark revision every February, calibrating against more accurate but less timely data sources. Last year, the preliminary estimate of the annual revision released by the agency recorded the largest decline since 2009.

This further confirms a common concern in the market: the initially released economic data may increasingly resemble a "draft" that requires repeated revisions.