A preview to tomorrow’s employment.

David Taylor’s site often provides useful commentary and economic data and indicators.  Take a  look at  his comments  on tomorrow’s employment release, and then return for our take.  (While we often read this site, we appreciate the tip from reader LC.)

Link: A preview to tomorrow’s employment..

If you were to take a statistics class, you’ll learn this motto: There are lies. There are damn lies. And then there are statistics.

The suggestion of a link between weekly jobless claims and the employment report is a logical one, which we discussed last month when the same topic came up at Calculated Risk.

Regular readers know that we warn about relying too much on what our eyes tell us about correlation.  It is worth taking the time to collect the data and use statistical methods.

Our own model gets a very good fit from three variables (out of the four provided on our Payroll Employment Game site).  Last month we challenged readers to guess which of the four variables added no predictive value to the model.  The weekly jobless claims is one of those that we use, and does have a strong correlation by itself.  It is much better to add the additional variables.

Our forecast for this month is for a gain of 126,000 jobs, slightly less than market expects.  As David Taylor points out, we will not know the real answer until after all of the government revisions, including next year’s "benchmark" changes.  This uncertainty will not stop the media from reporting the results as fact, nor traders from making big bets.

You may also like


  • Jim April 6, 2007  

    Jeff- Have you considered including a weather variable in your payrolls model? That probably would have captured some of the upside in today’s number..

  • Jeff April 9, 2007  

    Hi Jim,
    Your point about the weather is a good one. There are certainly times when weather helps a particular employment-week survey as compared to the normal seasonally-adjusted data.
    The problems in any additional variable are determining how much it adds overall and how to measure it.
    Even if something is not a large explanatory factor, it may be important to a particular month — your point here, I think.
    We’ll think more about the problem.