Wednesday, February 04, 2009
The highlight of the week's economic reports comes on Friday when the Labor Department releases the employment data for January which, in itself, may be quite a shock. But, this report also includes the annual revisions to prior data where the BLS (Bureau of Labor Statistics)updates its "birth-death model" numbers, among other things, to try to bring the revised numbers a bit closer to reality, from which they strayed quite noticeably over the last few years. Here's the official announcement from last month:
A large portion of those estimated 125,000 construction jobs that were created by the birth death model in 2007 are going to disappear and the construction job growth of 100,000 in 2008 may turn negative as many more home-building businesses were dying than being born whilst the housing bubble was rapidly deflating.
One of the great deficiencies in the employment data coming from the BLS is that, due to the amount of estimation and modeling that is required to provide timely data, much of the estimation and modeling go horribly wrong when the economy turns from expansion to contraction or vice-versa.
The BLS is quite up front about this in their documentation, however, this doesn't stop huge revisions from largely invalidating the data that was reported prior to the revisions. Recall that a couple years ago, over 800,000 jobs showed up in one of these annual revisions and suddenly, all the talk of Bush's "jobless recovery" stopped.
As it turns out, many of those jobs involved nailing drywall and peddling mortgages to people who either didn't understand what they were doing or didn't care because they thought they'd soon be rich beyond their wildest dreams, but that's a discussion for another time.
Look for many hundreds of thousands of 2007-2008 jobs to "vanish" in a similar fashion on Friday as another turning point was reached and the BLS attempts to correct all of those embarrassingly large positive numbers with more reasonable figures.
As I understand it, the revisions will only cover the period up to March of 2008, primarily for the twelve months prior to March 2008, meaning that, even this revised data will quite old.