ADVERTISEMENT

U.S. Payrolls Revisions Are More Reliable, Fed Study Shows

U.S. Payrolls Revisions Are More Reliable, Fed Study Shows

(Bloomberg) -- Explore what’s moving the global economy in the new season of the Stephanomics podcast. Subscribe via Pocket Cast or iTunes.

Revisions to monthly U.S. payrolls figures have become more reliable over the short term while those on industrial production have become less so, according to a Federal Reserve study that finds that policy makers could build such evolutions into their real-time assessments of the economy.

The volatility of the Labor Department’s revisions to its payrolls count is down by almost by half since the early 1980s -- to about 50,000 currently from 100,000, Cleveland Fed research economist Mark Bognanni wrote in a report released Thursday. Thus, the data provide a more reliable read on real-time changes in employment than in the past. Meanwhile, the typical size of revisions to industrial output figures has trended up over time.

“Revisions to many monthly economic indicators display systematic behaviors that policy makers could build into their real-time assessments,” Bognanni wrote. “Some indicators’ revision series have varied substantially over time, suggesting that these indicators may now be less useful in real time than they once were.”

The analysis coincides with Fed Chairman Jerome Powell’s recent comment that the central bank is using big data to improve its grasp of the job market after some major revisions, citing Labor’s announcement in August that payroll gains over the year through March were likely a half-million smaller than previously reported.

“Thus, the currently reported job gains of 157,000 per month on average over the past three months may well be revised somewhat lower,” Powell said Tuesday in a speech in Denver.

Multiple Revisions

Each jobs report includes revisions for the prior two months. For example, last week’s report showed nonfarm employment rose by 136,000 in September while July’s tally was revised up by 7,000 and August’s by 38,000.

Bognanni examined the real-time reliability of six important monthly economic indicators over time by studying short- and long-term revisions. The analysis included industrial production and manufacturing data from the Fed, personal consumption and housing starts from the Commerce Department, and Labor Department data on payroll employment and aggregate weekly hours worked. Each of those has a first, second, and third release, with the versions coming at one-month intervals that Bognanni considered short-term revisions.

“Substantial revisions tend to occur indefinitely after the initial data release, a result which suggests a certain degree of caution is in order when using even thrice-revised monthly data in policy making,” said Bognanni, whose research focuses on the macroeconomic effects of monetary and fiscal policies.

ADP Collaboration

In his remarks to business economists this week, Powell cited the central bank’s years of collaboration with the payroll processing firm Automatic Data Processing Inc. to build a payrolls gauge covering about 20% of the private U.S. workforce. Those figures are available with a roughly one-week delay -- faster than the government’s main jobs report and ADP’s own monthly count of private-sector employment.

“While experience is still limited with the new measure, we find promising evidence that it can refine our real-time picture of job gains,” Powell said.

He said the new gauge would have shown policy makers at the start of the Great Recession that the job market was much worse than official data showed. The Labor Department’s payrolls figures showed job losses of about 750,000 in the first eight months of 2008, while later benchmark revisions later put the loss at about 1.5 million.

To contact the reporter on this story: Jeff Kearns in Washington at jkearns3@bloomberg.net

To contact the editors responsible for this story: Scott Lanman at slanman@bloomberg.net, Vince Golle, Alister Bull

©2019 Bloomberg L.P.