Jobless Claims Data Gyrate Due To Seasonal Adjustment Difficulties

Jobless Claims: -22,000 to 353,000 in 12/13/03 Week

I took yesterday off, saw the headline above, and was pleasantly surprised. This data point was well below the 365,000 consensus. Then today I came across this interesting take from Joshua Shapiro, Chief U.S. Economist for MFR, an independent research house:

“We continue to maintain that the claims data are too dependent on questionable seasonal adjustment at this time of the year to be of much if any use in detecting underlying labor market trends. While other sources of information (survey evidence, the monthly employment report, etc.) point to an incipient labor market recovery, we will need to wait until after the New Year to see if the claims data corroborate this.

Over the past 6 weeks, we have seen NSA changes of +51,453, -49,282, +50,143, -40,404, +129,108, and now -75,048. SA changes have been +16,000, -7,000, -8,000, +11,000, +10,000, and now -22,000. The NSA increase 5 weeks ago was due to normal layoffs in construction as weather turned colder, the NSA decline 4 weeks ago was due to the Veteran’s Day Holiday, the NSA increase 3 weeks ago was due to a bounce-back effect from the holiday-shortened week, the NSA decline 2 weeks ago was due to the Thanksgiving Day holiday, the very large NSA increase 1 week ago is the NORMAL bounce-back effect from the Thanksgiving Day holiday and the NSA decline this week is the NORMAL drop-off after that very large increase. There is NOTHING unusual in the underlying NSA data, but the seasonals are once again unable to account adequately for the weekly changes.”

Note that last week’s claims had to be adjusted up. Does this decline of 22,000 (seasonally adjusted) represent a positive data point or not? Its hard to tell, but Joshua Shapiro notes that “the CUMULATIVE effect of poor seasonal adjustment has likely left SA claims at a lower level than the trend in NSA [non seasonal adjustment] data would suggest.”

Wouldn’t it be nice if we just got clean, unspun data?

Maria Fiorini Ramirez, Inc. (MFR).

Print Friendly, PDF & Email

Read this next.

Posted Under