Each month we report the results from our payroll employment model. This is not really a forecast. We take a number of different economic indicators from the middle of the preceding month and ask what change in payroll employment is consistent with the other data. The variables are not causally related, but are all different measures of the economy.
We get a pretty good fit from the University of Michigan sentiment reading (yet another new low this month), the four-week average of initial claims (using the period ending in mid-month), and the ISM manufacturing report.
The data are all consistent with continuing weak economic growth and a loss of about 90,000 jobs. The market is looking for a loss of 60,000. Since the 90% confidence interval (sampling error only) is about +/- 100,000, we think that an actual gain is pretty unlikely, while a big loss is possible.
It is interesting that whatever is reported is overly hyped and interpreted as an official count, rather than as a statistical estimate.
Readers interested in learning more can do three things:
- Take a look at last month's article, where we explored forecasting issues.
- Check out Little Known Facts about the Payroll Employment Report. They are still little known!
- Try playing our Payroll Employment Game. It shows you the range of different results the BLS might get even from a well-designed survey. It is cheaper than trading the actual report.