Thursday, June 03, 2010

Employment Report Prediction: Bad Reporting!

by Calculated Risk on 6/03/2010 12:51:00 PM

This is probably a safe prediction: There will be some really bad reporting tomorrow.

Note: The May headline payroll number will be large (consensus is for 540,000 payroll jobs added in May), but this will include around 400,000 temporary Census jobs.

I think this poor reporting will fall into two major categories:

1) Reporters who use the headline payroll number and write that this shows the recovery is picking up steam. We will probably see someone use the headline number and write something like "This is the largest monthly gain since September 1983" - well, were there 400,000+ temporary Census hires in September 1983? If not, why does this comparison matter?

2) Anyone who calls the number of temporary census workers "fake" or subtracts the birth/death adjustment from the seasonally adjusted headline number.

I've spoken with actual Census workers, and I guarantee their jobs are not "fake". Their jobs are real but temporary, and it is appropriate to subtract the Census 2010 workers from the headline payroll number to determine the underlying trend (even though it is obvious, I verified this calculation with the Census Bureau).

And no matter what anyone thinks of the birth / death model, the adjustment is added to the NSA numbers - and can't be subtracted from the SA numbers. Note: I think the birth / death model is useful although we have to aware of the weaknesses (it misses turning points - something I wrote about in 2006).

The employment report has many useful numbers. But the key payroll number is U-3 minus the number of Census 2010 temporary workers. The Census Bureau is tasked with reporting an estimate of the number of payroll jobs each month - so they include the Census 2010 jobs - but they also provide the information on the number of Census hires (the report will be released here).

It is the responsibility of anyone reporting on the numbers to do the heavy math and subtract the Census workers from the headline number. I will report both numbers.

Starting in June, the impact of the Census 2010 on the payroll report will be negative - see: Impact of Census 2010 on Payroll Report. That will require adding the temporary Census 2010 workers back to determine the underlying trend.

For some reason, print reports tend to contain the first mistake, and online reports the second error. Both are poor reporting, although the second is far more egregious (because the statements are blatantly false).

Oh well ... sorry for the rant, but I'm already seeing some really bad "analysis".