For the second week in a row, actual (i.e., not seasonally adjusted) unemployment claims as reported by the Department of Labor came in greater than the analogous week in 2012.
At the same time, and also for the second week in a row, the department's seasonally adjusted claims number -- the only one the business wire services ever specifically identify in their reports -- came in lower. In today's instance, raw year-over-year claims were almost 5 percent higher than the same week a year ago, but the year-over-year seasonally adjusted figure came in 11 percent lower. That's bad enough, but then the wires compounded the problem by running with indefensible conclusions based on DOL's contradictory data.
The following graphics demonstrate the contradiction. First, here's last week (issued on January 17 for the week ended January 12):
Both weeks had five buisness days, so it's hard to understand why DOL used very different seasonal adjustment factors in the two years. If DOL had used the same factor in 2013 as it did in 2012 (1.442 instead of the 1.660 it used), seasonally adjusted claims would have come in at a radically higher 385,000 instead of the reported 335,000 (555,708 divided by 1.442, rounded).
Now here's today's related graphic for the week ended Janary 19:
In this case, if DOL had used the same factor in 2013 as it did in 2012 (1.120 instead of the 1.323 it used), seasonally adjusted claims would have come in at 393,000 instead of the reported 330,000 (555,708 divided by 1.442, rounded), or 63,000 claims (19 percent) higher.
There is some justification for treating the most recent week differently than the comparable week a year ago, because the same week last year included the Martin Luther King holiday. But I don't believe that comes anywhere near justifying the entire 63,000-claim difference.
Let's see how the three major business wires covered the results (bolds are mine):
Associated Press, Christopher Rugaber ("US JOBLESS CLAIMS DROP TO 5-YEAR LOW OF 330,000")
The number of Americans seeking unemployment aid fell last week to the lowest level in five years, evidence that employers are cutting fewer jobs and may step up hiring.
The Labor Department said Thursday that weekly unemployment benefit applications dropped 5,000 to a seasonally adjusted 330,000. That's the fewest since January 2008.
The four-week average, a less volatile measure, fell to 351,750. That's also the lowest in nearly five years.
Bloomberg, Lorraine Woellert ("Jobless Claims in U.S. Decrease, Prolonging Seasonal Swings")
Claims for jobless benefits in the U.S. unexpectedly dropped last week to a five-year low, highlighting the challenges in adjusting the data for swings at the start of a year.
... Initial jobless claims reflect weekly firings and tend to fall as job growth accelerates.
Reuters, unbylined ("Jobless claims drop to five-year low")
The number of Americans filing new claims for unemployment benefits unexpectedly fell to its lowest since the early days of the 2007-09 recession, a hopeful sign for the sluggish labor market.
The fact is that year-over-year claims came in higher for the second week in a row -- the first time that has happened since October 2009. Because claims have really been higher, it seems fair to turn each of the three reported cited above around:
All three wire reports paid lip service to the difficulties involved in seasonal adjustments, yet none of them reported the fundamental problem the seasonal adjustments have been masking, namely that year-over-year claims are up.
The thing that's so maddening about the press's fixation on seasonal adjustment is that we're at the 4-1/2 year mark (since roughly June of 2008) where underlying economic activity has been far more influenced by government- and recession-induced distortions than by normal seasonal patterns. Yet no reporter wants to look past government calculations which are almost by definition inherently flawed to see what the real numbers might be telling us.
Cross-posted at BizzyBlog.com.