Why Is The Government Overrepresenting Raw Continuing Claims Numbers?

Tyler Durden's picture

One of the most recent statistical aberrations to be noted by Investment Research firm Oscar Gruss and subsequently referenced by Bloomberg highlights the dramatic disconnect between raw and seasonally adjusted continuing claims numbers. The divergence has manifested itself in a discrepancy of nearly 900,000 people. It is represented below:

From Bloomberg:

The pre-adjustment total dropped for the last nine weeks, the longest streak since May 2008. The number of claims fell by
1.05 million, or 17 percent, during the period. Adjusted claims changed direction on a week-to-week basis throughout the period and declined by only 113,000, or 1.8 percent.

Another way of visualizing the discrepancy is presented by Oscar-Gruss:

Furthermore, the key issue according to Oscar-Gruss' economists is the odd behavior of the 2009 data series which is a visible outlier compared to prior historical periods: the delta between unadjusted and adjusted hit a recent record, and to the downside at that.

And here is the proposed explanation:

“Our assumption is that the sheer brutality of the current cycle has caused the statisticians to cease to trust the ‘raw’
data and therefore fall into the trap of abusing the process of seasonal adjustment.”

Is the discrepancy merely a function of improving fundamentals catching up with the labor pool:

We would expect to see another “catch up” reduction in the “headline” seasonally adjusted continuing claim data in the coming weeks which will take claims down by somewhere between 400 to 500K based on the current data. We would also conclude that the pace of employment repair is running significantly faster than anyone relying on the “headline” seasonally adjusted data would be led to believe.

Yet keep in mind, in late 2008 and early 2009, the opposite was true, when unadjusted data was over a million higher than adjusted. Is the raw-to-adjusted ratio skew merely a function of the two data-series catching up on average with each other?

It would be odd for the government to present a cautiously bearish representation of economic reality here, while in all other economic data series it has not hidden its bias to promoting an upside case. Compare historical data revisions: what better way to indicate sequential improvement than by making just prior data look worse upon further "information" - this has occurred on far too many occasions over the past year.

Or is this merely a way for the government to provide "improving" numbers and thus stoke the market higher as the adjusted continuing claims number catches up with the raw number?

Alas, the simplest explanation is that while the rate of attrition could be slowing, it is not manifesting itself into any real economic improvement. Recall our recent presentation of the exhaustion rate, which hit an unprecedented 52% in the most recent data. This means that a whopping 52% of those collecting claims exhaust their benefits before they can find a job.

Combining the raw continuing claims data with the exhaustion rate presents a much more troubling picture: the raw data is "improving" due to the lack of material new "entrants", yet the lagging end of the curve keeps getting worse and worse.

This is yet another matter on which the BLS would be best suited to present their reasons for this divergence.

Lastly, it would be prudent to recall the interview with Charles Biderman of TrimTabs, who has for years been promoting the idea that the BLS continues to misrepresent the reality of the labor picture: be it raw or adjusted. One thing is for certain: with the U-6 unemployment rate breathing on 17% and soon likely to hit 20%, much of this speculation is merely semantics. The economy is still a long way away from not only stopping the labor bleeding but from actually reporting an uptick in employment. Until that happens, we expect that all bets are off when it comes to the government presenting the most optimistic possible case, even as more aberrations such as this become apparent.