Monday, February 6, 2012

To Rush Limbaugh: A Lesson in Seasonal Adjustments

This article is by Silver OZ.

While perusing various articles on the Employment Situation Summary that was released several days ago (2/3/2012), I came across this transcript from the Rush Limbaugh radio show where Rush decides essentially that we shouldn't bother with the seasonally adjusted numbers and only look at the raw (NSA) job creation number from the Establishment Survey (likely because this number is negative and thus "bad" for Obama). What Rush says is:

"That's part of this two and half million fewer jobs. Are you hearing me on this, folks? It is corrupt as it can be. Well, now, the wait a minute, though. There's nothing new here in the seasonal adjustment. Normally we never talk raw numbers. Nobody ever reports or talks about raw numbers. I happened to see today the raw numbers, and little red flags are going up, my curiosity is being piqued here. And then I see that the labor force participation rate, 1.2 million people dropped out of the labor force in one month, and it happens to be December to January."

What exactly are are these "red flags" Rush? These is a very good reason we have seasonal adjustments and more importantly, there is very good data on just how big the January adjustments typically are. What follows is is the "raw data" for January job reports going back to 1983 (the first year of the recovery from the 81 recession to pick an arbitrary date):

1983-01-01 -1667
1984-01-01 -1490
1985-01-01 -1744
1986-01-01 -1960
1987-01-01 -1966
1988-01-01 -2092
1989-01-01 -1979
1990-01-01 -2094
1991-01-01 -2550
1992-01-01 -2388
1993-01-01 -2167
1994-01-01 -2250
1995-01-01 -2309
1996-01-01 -2700
1997-01-01 -2546
1998-01-01 -2559
1999-01-01 -2755
2000-01-01 -2639
2001-01-01 -2876
2002-01-01 -2889
2003-01-01 -2685
2004-01-01 -2661
2005-01-01 -2706
2006-01-01 -2653
2007-01-01 -2794
2008-01-01 -3035
2009-01-01 -3698
2010-01-01 -2869
2011-01-01 -2858
2012-01-01 -2689

Notice a pattern here Rush? You see, every January many people get laid off regardless of underlying economic conditions. It is a very predictable and very consistent pattern (not one year with fewer than 1.49 million job losses), hence the seasonal adjustment. The BLS adjusts most months (although some adjustments are larger than others). For instance, last year, the "raw numbers" for February were +821,000 jobs, for March were +913,000 jobs, and April were +1,179,000 jobs, but obviously the seasonal adjustments took those numbers down considerably. So Rush, I am going to issue a challenge to you. I am going to challenge you to quote only the "raw numbers" for job creation every month this year, right up to the election, since you don't trust the seasonal adjustments. I am going to take a guess and say that you won't take me up on the challenge and instead will likely just pick and choose which number to use based on what is worse for Obama, but maybe there is a hint of journalistic integrity somewhere in your body.

Also Rush, since you decided to quote Zero Hedge regarding the "1.2 million people that dropped out of the labor force in January" (you do realize that Tyler Derden isn't a real person right?), I have decided to correct you on that point as well since I am already writing to you. That 1.2 million increase in those not in the labor force is an adjustment that the BLS applied based on actual data from the 2010 census. Essentially, since the census provides more exact numbers on the population statistics every 10 years, the BLS adjusts their estimates that were made in the intervening period (ie 2001-2010), but since the census doesn't break down monthly changes, the BLS simply applies the adjustment in one month (ie this January), which they clearly pointed out in the report. Conveniently, the BLS also provided a very nice table that showed what the December 2011-January 2012 change in the adjusted categories would have been without the once a decade adjustment. The actual change (without the revisions) from December to January in those "not in the labor force" was only -75,000. I wrote a post on this yesterday to correct both Zero Hedge and Rick Santelli, who made the same reading comprehension mistake that you succumbed to. And in case you just don't want my word on it, you can go to Calculated Risk for another take on it.

I hope this brief summation helped you understand a little bit more about the seasonal adjustment and the once a decade census adjustment, but if not, feel free to contact me, I'd be happy to have more discourse with you on the subject.