On Friday, September 4, the Bureau of Labor Statistics will publish the August 2026 Employment Situation. Our forecast is a gain of 41,000 nonfarm payrolls, which would lift the seasonally adjusted PAYEMS level from 158.858 million to approximately 158.899 million.
Why we do not simply add ADP’s 38,000
ADP reported a 38,000 increase in private payrolls. But ADP is not the BLS establishment survey, and it does not measure government employment. The two private-payroll series share a direction at times, but their monthly changes can diverge substantially. A one-for-one translation would be false precision.
Step 1: forecast private payroll growth
We first match the target to the predictor. ADP measures private payrolls, so the transfer function forecasts USPRIV, not total PAYEMS. Both series are converted to monthly log changes and the estimation starts in January 2021, avoiding the 2020 collapse that overwhelmed the full-sample relationship.
RainbowStats prewhitens the two growth series with ARIMA models and searches the residual relationship. The selected ADP residual enters at lag 11, but it is statistically weak: β = −0.080, t = −0.672, and adjusted R² = −0.031. That is an important result. The model is not pretending that one ADP release contains a reliable one-month payroll signal.
| Model diagnostic | Result |
|---|---|
| Estimation period | January 2021–July 2026 |
| Observations | 53 |
| Selected ADP residual lag | 11 months |
| ADP residual β / t-statistic | −0.080 / −0.672 |
| R² / adjusted R² | 0.0088 / −0.031 |
| August private log-growth forecast | 0.0004976 |
The model’s August private-payroll log-growth forecast is 0.0004976. With July private payrolls at 135.588 million, the conversion is:
The model’s 95% private-payroll interval is broad—approximately −102,000 to +237,000—so this is a published point estimate, not a claim of certainty.
Step 2: add government payrolls separately
Total nonfarm payrolls equal private payrolls plus government payrolls. Government employment has recently weakened, including a 53,000 decline in July. We use the latest 12-month average government change, −26,250, as the transparent baseline rather than asking ADP to explain a sector it does not cover.
How much depends on government payrolls?
More than we would like, and we show it plainly. Using the government ARIMA forecast produces +28K. Using the 12-month mean gives our official +41K call. Using the recent median gives about +57K—effectively the +58K consensus.
A rebound in government hiring or an unusually large positive BLS–ADP divergence would push the headline toward or above consensus. Another government contraction would pull it toward the lower end of our range.
The prediction
There is no hedging the headline: we predict that August nonfarm payrolls increased by 41,000. That corresponds to a PAYEMS level of approximately 158.899 million. We will compare the forecast with Friday’s release and examine the miss—direction, magnitude, revisions, and which component failed.
Run the RainbowStats replication
Publishing note: this is the localhost link supplied with the analysis. Replace it with the production share link before publishing publicly.
Sources and replication
- ADPMNUSNERSA — ADP total nonfarm private payroll employment.
- USPRIV — BLS total private employment.
- USGOVT — BLS government employment.
- PAYEMS — BLS total nonfarm payroll employment.
- BLS release calendar — August 2026 Employment Situation scheduled for September 4 at 8:30 a.m. Eastern.
Data accessed through RainbowStats on September 2, 2026. Payroll values and changes are seasonally adjusted where defined by their source. Forecasts are estimates and are not investment advice.