The kerfuffle today over volatile jobless claims data is of a piece with the long history of how these numbers are reported and then revised.
Continue Reading Below
These kinds of problems have cropped up in the past, and the number involved falls within the statistical range of revisions to prior week numbers.
Initial jobless claims–a measure of layoffs–were down a whopping 30,000 to a seasonally adjusted 339,000 in the week ended Oct. 6, the Labor Department said. However, revisions to the government's numbers can be significant, ranging anywhere from 30,000 to 60,000, data show.
You can see the examples of anomalies below.
The broad focus, however, according to economists at Wall Street investment banks, is that the average annual GDP growth since 2009 when the recovery began has been around 1.5%, versus the post–World War II average of 3.3%. That 3.3% number counts nearly a dozen recessions and economic downturns.
In addition, 83 out of the last 84 weeks of jobless claims data have been revised higher, a curious trend economists on Wall Street have already taken note of.
For example, Nomura Global Economics earlier this year warned analysts to take into account that the revised numbers will likely be skewed higher, adding that the undistorted trend likely lies closer to 370,000 to 380,000 in weekly jobless claims.
When the Department of Labor publishes weekly jobless claims, it does so after compiling data it gets from all 50 states, all of which submit figures to the Labor Department, whose team of economists and analysts then crunch the numbers on tens of thousands of households and businesses.
You’ll see the Labor Dept. notes in its weekly releases that its present weekly number is an advance number.
Meaning, the fact that the prior week number is revised versus the following week is simply a matter of function, because the initial count is not yet completed. In its disclosures, the Labor Department specifically notes that "the advance number of actual initial claims under state programs” is “unadjusted.”
Below are some government explanations given for big drops. One example of a big drop after a May 2011 report cites lagging data as a reason, the other reasons are for “unprocessed claims” – be it due to the holidays or the weather shutting down government offices.
Other examples of why the numbers can be volatile include layoffs at automakers or temporary factory shutdowns.
“Claims data tend to be volatile around the summertime “due to temporary shutdowns in the automotive industry,” economists at Goldman Sachs have indicated in a research note, adding any improvement should likely be “partly discounted.”
Economists note they tend to look at the August data, when weekly claims data typically tend to be less volatile. Specifically, the August numbers tend to present a better indication of whether labor conditions are improving or worsening.
Here’s what the Department of Labor has said in its releases:
April 2009 (-54,000)
2009-04-04 657,000 initial number
2009-04-11 603,000 revised number
“Keep in mind that the movement in weekly claims is likely to have been skewed by the inability of seasonal adjustments to fully capture the effects of the Passover and Good Friday holidays.”
July 2009 (-30,000)
2009-07-04 576,000 initial number
2009-07-11 546,000 revised number
“A department analyst said the drop in new claims didn't point to improvements in economic conditions. The second straight weekly decline reflected problems adjusting layoffs for temporary shutdowns at General Motors and Chrysler plants to retool for new models.”
January-February 2011 (-30,000)
2011-01-29 433,000 initial number
2011-02-05 403,000 revised number
“Some analysts cautioned that severe winter weather that affected 30 states could have contributed to the sharp drop, closing some government offices and preventing people from filing applications.”
April-May 2011 (-44,000)
2011-04-30 478,000 initial number
2011-05-07 434,000 revised number
“On Thursday, the Labor Department said that the number of continuing unemployment benefit claims--those drawn by workers for more than a week--rose by 5,000 to 3,756,000 in the week ended April 30. Continuing claims are reported with a one-week lag.”
July 2012 (-31,000)
2012-07-14 388,000 initial number
2012-07-21 357,000 revised number
“Auto manufacturers used to schedule brief shutdowns of plants at the start of every July to retool for new models, laying off tens of thousands of workers and making them temporarily eligible for jobless benefits. Yet auto makers have closed plants less often and laid off fewer workers after the government bailout several years ago.
"The changing employment patterns in the auto industry have yet to be fully captured, however, in the government’s process of making seasonal adjustments to the claims report. As a result, the weekly data are especially erratic in July and less useful in gauging labor-market trends."
Other reports note that the Labor Department has been forthright about the volatility of its data.
“Our data has been quite volatile lately,” a Labor official said in one release, after the decrease in claims in one week wiped out an equally large increase in the prior week.