Will the September non-farm payrolls see another "significant downward revision," opening the door for a "50 basis point rate cut"?

U.S. employment data may face a significant downward revision, paving the way for a 50 basis point rate cut in September.

On September 9, the U.S. Department of Labor (BLS) will release its annual benchmark revision of non-farm payroll data. According to projections from Goldman Sachs, Standard Chartered Bank, and others, this could result in a substantial "employment reality check" of up to 550,000 to 800,000 jobs, which would directly impact market confidence in the U.S. labor market and could force the Federal Reserve to implement a significant 50 basis point rate cut, similar to last September.

There are two main reasons for the significant downward revision of the data. First, the birth-death model has become distorted, overestimating the jobs created by new businesses. Second, the substantial decrease in illegal immigration has led to a systematic overestimation of the labor force population. Estimates suggest that these biases may result in actual employment being overestimated by 40,000 to 70,000 jobs per month, equivalent to an annual cumulative inflation of 550,000 to 800,000 jobs.

The implications of this are very significant. A senior trader at Goldman Sachs stated that the key factor determining Powell's pace is not inflation, but employment. If this revision is similar to last September (when the BLS also revised down by 800,000 jobs, leading to a subsequent 50 basis point rate cut by the Fed), Powell may once again face the choice of whether to cut rates by 50 basis points in one go, even if just to "prove his innocence"—that last year's rate cut was not a political compromise, but based on real economic slowdown.

Standard Chartered estimates that the NFP published by the BLS is overestimated by 70,000 jobs per month

Goldman Sachs pointed out that the biggest source of distortion in employment data is the "birth-death model" used by the BLS for a long time. This model is used to estimate the number of jobs created by new businesses but is not based on actual business registrations or tax data; rather, it relies on model estimates, which can systematically overestimate job growth. In contrast, the QCEW (Quarterly Census of Employment and Wages) and BDM (Business Dynamics Statistics) are based on actual unemployment insurance records submitted by businesses and are considered a more reliable "gold standard."

Goldman Sachs used its own model, combined with BED data and more frequent business dynamic information, to find that the BLS model indeed overestimated job growth in the second half of 2024, averaging an overestimate of 45,000 jobs per month. Although the BLS has slightly adjusted model parameters in recent months and reflected a stabilization in the number of new businesses, the bias remains significant.

Standard Chartered's Steven Englander more directly referred to the birth-death model as a "fig leaf for the data." He estimates that the NFP published by the BLS is overestimated by 70,000 jobs per month.

According to his analysis, from the beginning of 2024 to the present, established companies have only added 25,000 jobs per month, while the BLS estimates that "new companies" contribute over 100,000 jobs per month. However, BDM data shows that new companies actually account for only 20% of all new jobs, far below the BLS assumption. More seriously, the number of jobs created by new businesses in 2024 is less than 20% of that in 2022. If the model reflects this reality, the NFP will be at least 70,000 jobs lower each month.

Englander further pointed out that to maintain a basic balance in the labor market, the "reasonable level" of non-farm employment data should be 170,000 jobs per month, with 100,000 coming from real natural growth and 70,000 being the overestimated portion from the model. **

It is worth noting that although the BDM is lagging (the latest data only goes up to 2024), like the QCEW, it serves as the data foundation for mid-year benchmark revisions used by the U.S. Department of Labor, and its authority far exceeds that of the sampling-based non-farm employment data. The employment benchmark revision to be announced by the BLS on September 9 is based on this data. Once the non-farm employment figures are adjusted according to the real trends reflected by the BDM, they may be revised down by 550,000 to 800,000 in one go, which will have a huge impact on market confidence and policy outlook.

Five Major Signals: Signs of Overstated Employment Data

Goldman Sachs pointed out that in addition to the birth-death model causing inflation, there are at least five additional reasons that further indicate serious issues with the data.

1. Decrease in Illegal Immigration

Goldman Sachs estimates that the number of illegal immigrants has significantly decreased in recent months. Illegal immigration has a substantial impact on labor supply. The "immigration wave" between 2022 and 2024 led to a surge in employment demand, but now immigration has slowed, and the actual need for new positions has also decreased. If the BLS continues to estimate employment demand based on old immigration assumptions, it will clearly be overstated.

2. Seasonal Adjustment Models Misjudge Trends

Seasonal adjustment models often initially misinterpret changes in real trends as seasonal fluctuations. It is only after confirming that the trend has indeed worsened that the model will revise previous data downward.

3. Historically, Original Data is Always Revised Down During Economic Slowdowns

Historical experience shows that during periods of economic slowdown, the initial raw employment data is often revised downward later. This phenomenon has occurred in every economic recession since 1979 (except once).

4. ADP Data Questions BLS's Overstatement of the Healthcare Sector

As a major provider of wage data in the U.S., ADP's data shows that employment growth in the healthcare sector is far less robust than reported by the BLS. In the past three months, new positions in the healthcare sector accounted for more than the total non-farm employment growth. Both ADP and industry analysts believe that the situation is not as exaggerated as the BLS claims, and the real situation may lie somewhere in between.

5. Household Surveys Overestimate Immigration and Employment

The household survey may currently overestimate U.S. population growth and employment growth. This is because the immigration estimates used at the beginning of the year were reasonably accurate but have now become significantly overstated. The current model assumes that the U.S. population grows by an amount that may be overestimated by 1 million people per year. This could lead to the employment growth data in the "household survey" being overstated by about 50,000 positions each month