Friday, January 8, 2010

The Seasonal Adjustment

From Bonddad: SilverOz also wrote on this here. Also, the Census Bureau has a great explanation of seasonal adjustments here.

Bonddad asked me to write a little bit on the seasonal adjustment for December, as the adjustment used this month was significantly larger (to the downside) than in previous years for this month. This large downward adjustment to the seasonal adjustment (I am discussing the Establishment Survey here) is a bit inexplicable, as the typical seasonal factor that is attributed to the downward adjustment is Christmas hiring, which is fairly stable over time and if anything was weaker this year. Thus, it may be that after last year's huge revisions (coming to an employment report in February) the BLS decided to err on the side of conservatism and overestimate the seasonal adjustments instead of underestimating them (because presumably no one will publish conspiracy theories about the BLS "cooking the books" if the numbers are revised up at the end of the year).

To quickly look at the data behind my theory, I will look at the December adjustments, as these were the most "off" from their averages (or even from 2008). In December, the difference can be seen in comparison to both the average and the 2008 adjustment, as the historical adjustment is -.56 versus the -.69% used by the BLS in 2009, which would yield a jobs difference of 171,000 (netting a positive jobs number for December), however the 2008 adjustment was -.62%, which would take us to a positive 7,000 jobs for the month. The interesting thing about this large downward change in the seasonal adjustment for December is that the large seasonal factor for the month (ie Christmas Jobs) was consistent in its adjustment with previous years (subtracting 545,000 jobs from the report), meaning that the additional adjustment came from a sector outside of retail.

So, what we will really have to watch closely is the January job numbers to see what changes to the seasonal adjustment are made then, as it would seem highly suspect to subtract extra seasonal jobs in December and then not add back to account for the seasonal firing in January/February. It may very well be that the BLS has tweaked their adjustments to a more conservative number (ie less jobs) in order to avoid any future embarrassing revisions like will take place next month.