Payroll Revisions: More than Meets the Eye

The recent July employment report sparked an opportunity to revisit a common misunderstanding about payroll revisions and their implications for the economy. The Bureau of Labor Statistics reported a nonfarm payroll employment loss of 23,000 jobs in July, with the unemployment rate decreasing to 4.1 percent.

However, these numbers are estimates based on a survey of approximately 119,000 businesses and government agencies representing about 622,000 worksites. The sample size is significant, covering about one-quarter of American payroll employment, but it's still subject to revisions as additional information arrives and seasonal factors are recalculated.

The BLS notes that the two preceding months are routinely revised after nearly all reports have been received. In this case, May's payroll growth was revised from 129,000 to 63,000 jobs, and June's from 57,000 to 20,000 jobs, resulting in a combined downward revision of 103,000 jobs.

A closer look at the data reveals that revisions are not necessarily corrections but rather updates as more information becomes available. The 90% confidence interval surrounding a monthly payroll change is roughly plus or minus 122,000 jobs. Applying this to July's estimate means the underlying change could be anywhere between -145,000 and +99,000 jobs.

The analysis of historical data suggests that downward revisions are common during good economic times, while upward revisions are more frequent during stable contractions. A single negative revision is not a strong indicator of recession, but persistent deterioration across successive estimates does signal increased risk. Once an ongoing string of negative revisions accumulates to approximately 30,000 to 40,000 jobs, subsequent recession risk rises noticeably.