More numbers fudging...
Beware the Jobs Report of July
By BINYAMIN APPELBAUM
The American economy almost certainly shed more than a million jobs last month, just as it does every July.
But that’s not the “jobs number” that will get most of the attention when the Bureau of Labor Statistics speaks Friday morning.
No, the number we care about will be much smaller. Indeed, it is very likely to be modestly positive. And that is because of something called “seasonal adjustment.”
Every month, the Bureau of Labor Statistics estimates the number of jobs based on a survey of businesses. Then it adjusts the number to eliminate recurring fluctuations — such as from teachers going on summer break — so long-term trends are more visible.
In June, for example, the bureau estimated that the economy added 391,000 jobs. But the headline number, after seasonal adjustment, was just 80,000 jobs.
This is a good and important idea. But right now we have a problem: Some economists believe economic turbulence has disrupted the calibration of those adjustments, undermining the accuracy of the bureau’s estimates.
Federal Reserve officials have cited doubts about the accuracy of the monthly jobs number as one reason for their uncertainty about the health of the economy.
As my colleague Floyd Norris explained last month, one possible distortion that has arisen in recent years, thanks to the weakness of the economy, is that “seasonal adjustments make things look better than they are in the winter, when fewer workers are being let go than the government expects, and worse in the spring and summer, when the workers who were not let go cannot be rehired.”
The issue will loom particularly large on Friday, because the report for July is annually adjusted by a larger amount than for any other month save January, when holiday workers lose their jobs.
In the last 10 July jobs reports, dating back to 2002, the agency has added an average of 1.33 million jobs to its original estimate. Last year, for example, the agency estimated that payrolls declined by 1.3 million jobs in July, but it reported a seasonally adjusted increase of 96,000 jobs.
That places a huge premium on the accuracy of the adjustment: A 5 percent error in the adjustment would have shifted the reported total last July by two-thirds.
And even in the best of times, the bureau’s estimates are rarely that accurate.
The government has estimated an average change of 149,700 jobs in the last 10 July jobs reports, but it has since revised those estimates by an average of 92,900 jobs per year. In other words, the initial estimate is generally off by about 62 percent.
In three of those 10 years — 2002, 2003 and 2007 — the agency wasn’t even correct about whether the economy gained or lost jobs.
So take Friday’s report with a measure of caution.
And one more thing: The other headline number, the unemployment rate, is derived from a separate household survey. It’s also adjusted seasonally. And it has its own problems. But that’s a blog post for another month.