Why you shouldn't believe everything you read in the jobs report
The Bureau of Labor Statistics reports an unexpectedly sharp rise in job-creation for the month of October, and also an unexpectedly sharp rise in job losses—in the very same monthly data release.
Economists today were going through the confusing and conflicting numbers, to sort out effects of the government shutdown in early October from the underlying forces at work in the economy and labor market.
Non-farm payrolls increased by 204,000 according to the report, approximately 80,000 higher than the consensus estimate of economists. Job growth occurred across a broad range of industries, from retail and leisure (bars and restaurants), to manufacturing, professional services and health care.
Meanwhile, the unemployment rate rose 0.1 percent to 7.3 percent. The number of Americans reporting that they had work during the survey period in October declined by 735,000, and the total civilian labor force declined by 720,000—both very large, and statistically significant, declines.
The October government shutdown accounts for at least some of this contradictory data. Economists used words like "distorted," "noisy," "difficult," and "misleading" to describe the report.
"The data’s mushy, you could call it that," quipped Nariman Behravesh, chief economist at IHS. He said there’s a more a technical term: "The ‘noise-to-signal ratio’ of the recent data is fairly high, there’s a lot of volatility in the data, and special factors to sort through."
A key factor is that there are two different surveys conducted by government researchers. In the household survey, a sampling of approximately 60,000 American households are asked about the employment status of adults (age 16 or older) living there.
In the Establishment Survey, a much larger sampling of businesses and government offices (approximately 500,000) are asked how many people they have on the payroll.
In October, as many as 800,000 federal workers were furloughed at some point, though they were later given back-pay.
"In the (establishment) survey of firm payrolls, you’ll be counted as employed," if you’re paid for the time you were on furlough, explains economist Justin Wolfers of the University of Michigan and the Brookings Institution. “And in the household survey they call ‘temporary layoff a form of unemployment. So the same person is either employed or unemployed, depending on which survey you look at."
Also, because government workers didn’t return to their desks until mid-October, they conducted the household survey one week later than usual.
"My memory of what I was doing two weeks ago just isn’t as good as my memory of what I was doing a week ago," says Wolfers. "Small changes in how we conduct surveys can have very large unintended consequences."
Tom Nardone at the Bureau of Labor Statistics -- whose staff meticulously explained these and other anomalies in the October jobs data as part of the Employment Situation Summary release -- says there are often statistical anomalies in the data. In addition, he says, the normal margin of error on the surveys ranges from more than 90,000 in the establishment survey, to approximately 400,000 in the household survey.
"The numbers this month are as reliable as they are most months, which may sound like a surprising statement," says Nardone.
One particularly troubling trend indicated by the October report is the precipitous decline in the labor force, and a concomitant fall in labor-force participation, says finance professor Dan Seiver at San Diego State University.
"A big departure of people -- three-quarters-of-a-million -- from the labor force: that’s a puzzle the BLS doesn’t have an explanation for," says Seiver. "If it’s retirements or other voluntary leaving of the labor force, we need to understand that better." If it’s due to long-term unemployed people giving up the job search and leaving the workforce altogether, that would be more alarming, Seiver says.
Justin Wolfers says, unless you’re an economist or on the Federal Reserve, you might as well wait till next month, when some of the data distortion and mushiness has diminished, to figure out how the job market’s doing.