Sunday, December 6, 2009

Questioning the November Employment Report

Many economists and analysts are finding the November BLS Employment Report released this last week hard to swallow.  I went over it in some detail on the Radio Show, but the questions on the data merit posting for reader's review.

The Report showed a job loss of 11,000 jobs when the month before was 195,000 and it showed the unemployment rate going down to 10% from 10.2 despite the loss.  However, the BLS figures are usually close to the ADP private survey and the ADP showed a loss of 169,000 jobs.  Calculated Risk asked "If the Economy lost Jobs, why did the Unemployment Rate decline?"  The article used scatter graphs of one month and two month rolling averages and a 3rd order polynomial and concluded it was at best statistical noise: "The bottom line is the decline in the unemployment rate this month was noise, and the unemployment rate will probably increase further. If the economy adds about 2 million payroll jobs next year, we'd expect the unemployment rate to still be at about 10% at the end of the year."  For every point of unemployment which lasts two years, the cost is $400 billion.

GDP growth below 2.5% will not substantially change the unemployment rate.  Slower growth could mean 10% unemployment through 2012 into 2013.  Krugman has commented on this as has Calculated Risk in "Employment and Real GDP".

If you look at page five (5) of the Report, you will see that there was a 98,000 decline in the civilian labor force with a 325,000 decline in unemployment but there was an increase of 291,000 in people no longer in the labor force.  You can also see this analysis on Shedlock's blog.  U6, discouraged workers no longer in labor force is shown at 17.2% in the BLS Report, an alternative model using the same statistical definition in use in the middle 1990's indicates 22%, unchanged from last month.

Some people have tried to reconcile the BLS Report with the ADP and have found they cannot get the BLS Report to add up, but that may be because different statistical methodologie3s are used for different components of the BLS Report.  It is well acknowledged that the BLS Birth/Death Model is flawed and will be corrected in its annual review in February for an overestimate of at least 824,000 jobs filled error, although some say the error may be an overestimate of about 80,000 per month, which would be 960,000.  One attempt at the ADP to BLS figures summed up with "We are looking at a three-month ADP-BLS net jobs divergence of roughly 230%, with ISM non-manufacturing employment still deep in contraction. Again, whom are we supposed to believe?"

Trim Tabs, which is a professional fee analytical service, calculated a November job loss of 255,000 and The Money Game provided a month by month Trim Tabs to BLS comparison estimates and revisions.  It is not pretty.

Print Page

No comments:

Post a Comment

Share This