Monday, February 16, 2015

Weekly jobless claims as forecasts of job growth and unemployment

 - by New Deal democrat

I've found that the weekly new jobless claims reports are useful in forecasting that month's employment number, and also the unemployment rate several months later.  Since I haven't done that in a while, I thought this would be a good time to update each of those metrics.

First, here is a spreadsheet that was prepared for me by Craig Eyermann at Political Calculations, showing the relationship between the monthly average of initial jobless claims at that month's subsequently reported jobs number during this expansion:

Roughly, this shows that for every 20,000 decrease in initial jobless claims, there has been a similar 20,000 increase in the monthly jobs report =/-80,000. This graph was published last October.  So how well has the past trend forecast the last four months of jobs?  Here's the predicted value (left), and the actual report (right), both in thousands:

Oct 260. 221
Nov 248.  423
Dec 252.  329
Jan 248.  257

Three of the four months have been within that range, with November obviously much higher. Still, the results show that the average actual jobs numbers in the last four months has been higher than predicted by the trend.  Or, put another way, the trend in new job creation is accelerating.

Next, here's a graph I haven't updated in awhile.  This shows the leading relationship of jobless claims to the unemployment rate a month or two thereafter:

This relationship completely blew out in 2009, but after that returned:

 The question recently has been, with initial jobless claims at near record lows as a percent of the population, would the unemployment rate stall, or make up the ground it lost in 2009? So far the answer is the latter.

Finally, the longer term rate of new jobless claims also leads the longer term unemployment rate.
First is the graph covering the last 50 years:

Based on both of the above trends, I am expected further declines in the unemployment rate, although perhaps at a slower rate.