Jobs Data Might Be A Statistical Mirage

Includes: DIA, QQQ, SPY
by: Paulo Santos

Some observers are somewhat surprised by the last non-farm payrolls job creation statistic. These statistics somehow deviate from historical relationships and not surprisingly that can easily make for a good story or conspiracy.

However, there's another possible explanation - the jobs number might be part of a statistical mirage now making its way through the official numbers. Unfortunately, we can't quantify this statistical mirage because the seasonal adjustment models are not public, but we can explain the logic that brings about the mirage.

The first thing one needs to understand is that the numbers reported by the BLS are seasonally adjusted. They are seasonally adjusted because, if they were not, there would be significant swings in the data that just correspond to seasonal cycles in employment. For instance, right before the holiday/Christmas season there's a lot of seasonal hiring, and right after it ends, a lot of seasonal firing.

Now, the problem here is that how much of this seasonal hiring/firing is derived from the data itself. And during the 2008 recession, there was an implosion in employment right around the end of the year, so the cycle was much deeper than usual. Now, part of that implosion will be picked up by the methods that seek to seasonally adjust the data. What these methods will think is "around the end of one year and the start of another it's normal for employment to fall hard" (and indeed it is, but the methods will pick a harder "hard").

Now, fast forward to 2011 or 2012 ... when employment regularly falls at the start of the year, it won't fall "as hard" as expected by the seasonal adjustment that's now also using data from 2008. The end result? The seasonal adjustment will over-compensate and turn something normal into something more positive (source for chart:

This effect has already been noticed by a few analysts. And there's not just upside from it, there's downside as well. Once we get to April/May, the effect turns into its head - then the seasonal adjustments will be expecting more improvement than they really should expect (as the seasonal adjustments over the entire cycle even themselves out), and once that improvement does not materialize, the seasonal adjustments turn what might otherwise be a normal performance, into a weak one. Unless, of course, the BLS revises the seasonal adjustment method to do away with the effect, as the ISM already did.


There's reason to believe that a part of the much-celebrated January non-farm payrolls number was simply a statistical mirage. Unless changes are done, this positive effect will happen again in the following two months or so, and turn negative afterwards. This might be relevant for the trading of S&P (NYSEARCA:SPY), Dow Jones (NYSEARCA:DIA), Nasdaq (NASDAQ:QQQ) around economic releases. The same effect might be overstating the improvement on the jobless claims numbers.

Disclosure: I have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.