Friday, May 12, 2006

Number bogosity depends on how you add 

Steve Verdon is accusing Powerline's Paul Mirengoff of using bogus employment numbers. Mirengoff says the "jobless recovery" has produced roughly 200,000 new jobs. Verdon says that's "anemic" compared to previous recoveries, and moreover
I�m not sure where Mirengoff is getting that 200,000 number. Taking the last 18 months of data I get about 165,000 jobs/month (this is for Nov. 2004 through April 2006�and taking Oct 2004 through March 2006 gives us 176,000). That is quite a rounding error.
Verdon uses a monthly employment change chart, and as I often say you need to smooth out those changes. The trailing 12-month change I use in that link shows about 2 million jobs per year consistently since late 2004. That does in fact come out to about the 165k/mo rate that Verdon shows.

What explains this difference? Part of it is a data revision in February that pulled down data from March 2005 by 158,000 jobs. That would account for about a third of the difference. (I know this one well, because for the QBR I had to re-write all my datafiles with the revision, as I do every spring issue. I know I pulled the St. Cloud numbers down this time.)

Part of it is Katrina, no doubt. In each of the Employment Situation reports from BLS since August 2005 you find an extra table or two explaining the effects of Katrina. The latest has a table for April 2006 looking at the 911,000 evacuees of the hurricane, of which about over half have returned to their homes. Suppose the rest were able to return, and they participated in the labor force and were employed at the same rate as those who have returned according to BLS. If so, there would be 282 thousand employed. Actually, though, there are 189 thousand employed. That 93,000 might be an overstatement -- maybe those who can't get back home have systematically different job experiences than those that can -- but then those workers put out by Katrina may also have a multiplier effect on those firms who survived Katrina. Ask your local bar owner in New Orleans about that.

If you add that on to the data revision you would get an average of 180,000 jobs per month over the eighteen months, and over 190k if you go back to October 2004. Had Verdon checked out the link he includes from his quote of Mirengoff, he'd also see that Paul was referring to the 2005 data through November, pulling out the Katrina months of Sept. and Oct. (You can play with graphs of the data at Economagic.) In short, different guys are adding data in different ways to get different results. There's much more going on underneath the averages being batted about.

We still don't know why employment growth has been slower in this recovery than the last (which itself was slower than the previous post-WW2 recoveries.) If it was a function of cheap money and heavy investment we'd not see increases in multifactor productivity like we have in this recovery. It's a puzzle.