Tuesday, October 16, 2012

Retroactive record setting unemployment rates

Little noticed in the media, but there was a big change in labor force stats routinely reported on each month for the Pittsburgh region. 

A few may remember this post from April when I mentioned that the Bureau of Labor Statistics was reporting seasonally adjusted unemployment rates for the Pittsburgh region that were different from what the state's own numbers were showing.  Not the biggest of discrepancies so no big deal.  The difference between the two data points was not an error, but an artifact of two different methodologies for adjusting raw labor force data for seasonal variation.  

It turns out that with the data just released last week (beginning with the August MSA data) the state has basically given up on using their own models and are now reporting the BLS version of the same regional labor force data for MSAs within the state.  OK, not a problem there.   They also are using the BLS data going back in time to 2000.   Basically all the historical unemployment rate data as reported contemporaneously has been changed.  In some months the differences between the old and new unemployment rates for Pittsburgh can be quite substantial up to as much as 4/10ths of a percent. 

But one theme here in recent years is that we have been generally bouncing around or in a few months tying a month in the 1970s which was the last period in which the regional unemployment rate was so far below the nation's unemployment rate.  The new data actually works out to be a new record (in the past).  In October of 2009 the national unemployment rate was finally reported at 10.0%.  For the Pittsburgh MSA the originally reported final unemployment rate was reported at 8.0% which gives a difference of 2.0 percentage points.   The new unemployment rate being reported for the region that month is 7.7% which gives a difference of 2.3 percentage points below the national unemployment rate.  That would be the largest gap by which Pittsburgh's unemployment rate below the national unemployment rate in any data since 1970 and likely much further back.

There are actually a slew of contemporanous news stories, punditification, and headlines that all would have to be qualitatively rewritten if the revised data were known at the time. It all gets again to how much we overinterpret these monthly labor force data dumps. Hold that thought because there are some bigger issues in that I may get back to. 

With the revision of data back to 2000, the entire time series has been changed.  Here is the updated version of my chart showing the difference between local and national unemployment rates. If you really want to discount what the green means, I don't have time to update my calculation of the cumulative difference in this chart, but basically these are unprecedented times in some ways for the region's labor force..  Think all that may have something to do with the record size of the regional labor market and recent net migration flows into the region?  You bet. 


Blogger Josh Clark said...

From where did you get pre-1990 unemployment data for Pgh - from the census bureau? chartistfriend@gmail.com

Thursday, October 25, 2012 8:09:00 AM  
Blogger Josh Clark said...

Found it - great site, bro.

Thursday, October 25, 2012 8:19:00 AM  
Anonymous Jennifer Graham said...

Did you have updates for this post? There must have been something serious that the government did for these numbers. I mean, we no longer know who should do the move in decreasing the unemployment rate in the US. Much as I know, a lot of people do their part in exerting effort and trying to look for a living. However, we still get these number of slaps on our face due to the undeniable numbers showing.

Monday, December 17, 2012 10:38:00 AM  

Post a Comment

<< Home