Goldman Sachs: Significant downward revision in non-farm payrolls, what is the true situation of US employment?

Recently, a series of employment data released by the U.S. Bureau of Labor Statistics has raised market suspicions, with Goldman Sachs believing that the Quarterly Census of Employment and Wages (QCEW) data implies misleading information.

The data shows a significant downward revision of 818,000 non-farm jobs, while over the past three months, employment has grown by 170,000 jobs per month. On August 26, Goldman Sachs analysts Jan Hatzius, Alec Phillips, David Mericle, and others released a report stating that revising down by 300,000 or 25,000 per month would be accurate, with an overall employment growth rate of about 160,000 jobs per month.

Goldman Sachs pointed out that the birth-death adjustment model may be too optimistic, leading to an overestimation of non-farm employment data. In addition, continuous downward revisions due to seasonal factors can cause data deviations, so the unemployment rate may be a better observation indicator.

The Federal Reserve's policy decisions will consider the actual situation of the job market, and poor non-farm employment reports due to fluctuations or survey noise may prompt the Fed to cut interest rates faster.

Birth-Death Adjustment Model May Be Too Optimistic

The data shows that employment has grown by 170,000 jobs per month over the past three months. According to Goldman Sachs analysis, the non-farm employment level in 2023 was significantly exaggerated, with the birth-death model being a possible culprit, overestimating by about 25,000 jobs per month, with the overestimation possibly accounting for about 300,000 of the 818,000 job cuts in this revision.

However, Goldman Sachs also pointed out that by 2024, this adjustment has significantly decreased, with the overestimation error narrowed down to about 10,000 jobs per month, currently only slightly higher by 5,000-10,000 jobs per month compared to pre-pandemic levels.

Furthermore, the continuous downward revisions of seasonal factors that grew between 2022 and 2023 have largely ceased as the employment growth trend stabilizes. Goldman Sachs noted that with the confirmation of subsequent data, it is now recognized that the employment growth trend has indeed slowed down. So far this year, seasonal factor adjustments have added an average of 2,000 jobs per month, while in 2022 and 2023, there was an average decrease of 30,000 jobs per month.

Discrepancies Due to Illegal Immigration and Survey Method Biases

Regarding the 818,000 baseline adjustments based on QCEW, Goldman Sachs pointed out that the scale of the adjustment is as expected. This adjustment may have deviated due to many companies not reporting illegal immigrants, affecting employment data. After capturing the correction to the birth-death adjustment model, revising down by 300,000 or 25,000 per month would be accurate.

Goldman Sachs noted that part of the gap between household surveys and non-farm employment growth is due to the inability of household surveys to capture the surge in immigration and differences in definitions, with illegal immigrants potentially causing a greater discrepancy between the two surveys than the preliminary baseline revision earlier this week However, despite considering these factors, there is still a gap of 1.2 million people between household employment and non-farm employment growth. Goldman Sachs research found that the survey methods used by the BLS may lead to one or both indicators becoming less reliable, thus introducing more uncertainty.

Therefore, Goldman Sachs has adopted a statistically optimal rule to balance the difference signals between household surveys and non-farm employment surveys, placing approximately three-quarters of the weight on the more reliable and less noisy non-farm employment growth, and one-quarter of the weight on household employment growth. Taking into account adjustments for immigration and birth-death models, Goldman Sachs estimates the true pace of employment growth to be around 160,000 per month.

Data Fluctuations or Rate Cuts for Acceleration Unemployment Rate Data More Accurate

However, given the uncertainty of the true pace of employment growth, the report points out that it is recommended to pay more attention to the unemployment rate as it is less affected by data challenges.

There is also a risk that labor demand may be too weak, and if this occurs, or if poor employment reports result from fluctuations or survey noise, the FOMC may cut rates more quickly. However, even if labor demand is sufficiently healthy, the volatility of monthly employment growth and noise in the data mean that data can sometimes significantly undershoot the average level.

Furthermore, predicting seasonally adjusted employment growth is also challenging, as even a small deviation in any adjustment can have a significant impact on estimating seasonally adjusted employment growth.

Therefore, due to the difficulty of confidently distinguishing signals from fluctuations in real-time, soft data resulting from normal fluctuations may also lead the FOMC to cut rates more aggressively