Thursday, June 4, 2009

Revised Preliminary Updated Prediction!

by the Sandwichman

Sandwichman previously predicted a job loss of between 775,000 and 835,000 jobs in the May employment report. This morning, however, when parsing the CNN Money happy talk about signs of improvement, I realized a wrinkle I had overlooked. Ken Houghton also called attention to the wrinkle in a comment. Am I comparing a preliminary report for May to a revised one for April? That hadn't been my intention. But I did make a mistake. I was carelessly thinking of the April job report as if it was static and the May one as if it was the "final edition".

From last September to February, as the employment situation has hemorrhaged, the average adjustment from first preliminary to final revision has been 136,800 per month. Comparing this month's preliminary with last month's first revision could thus handicap the reported job loss by an average of 205,250 jobs. So I'm going to make a revised prediction of a NOMINAL job loss of between 570,000 and 635,000 jobs in May. Because the April job losses will likely be revised upward by, say, 60,000 to 70,000 (to 599,000 to 609,000), CNN Money can even report the preliminary May figure as an "improvement."

It may not be obvious that the 570,000 figure represents the SAME estimate of jobs lost as the original 775,000, the difference being that an estimated 205,000 of them are split between one upward revision of the April report and two future revisions of the May report. Phew!

2 comments:

Anonymous said...

Ah ha! So now we have to wait until August to nail you. If you weren't so contrary, I'd say you're "hedge' ready for CNBC. I'd say you're going to overshoot the mark by about 100k. that is revised (semi) final for May per the July empsit report will be about 670.

Joe K.

Sandwichman said...

Here's another way of putting it, Joe: the preliminary April non-farm employment count was 132,414,000. My May estimate would have been for a 131,639,000 to 131,579,000 final May count. If the April number gets revised downward to something like 132,300,000, the May final may well be "about 670" (or even 661).

Not to put too fine a point on it, my basic observation remains that the April number was not the "improvement" or breathing spell that it was made out to be. The appearance of improvement was a statistical artifact.

Based on the ADP numbers and the way they're being reported, May doesn't look any better than April while April is looking worse than it did a month ago. On the other hand, I'm not saying that April or May are any worse than February and March, just more of the same -- that is to say, really, really bad.