Friday, May 02, 2008

U.S. Employment Report Weaker Than It Seems

Stocks and the dollar rallied, while bonds sold off, after the slightly stronger than expected U.S. employment numbers. Still, while the numbers were more bullish than I had expected in some aspects, the details do not look as relatively benign as the headline.

The headline talks of a decline of 20,000 payroll jobs, less than most analysts including me had expected. And unlike in previous months, government employment boosted the number only slightly, by 9,000 to be more precise. Meaning that private sector employment fell by only 29,000, much lower than in previous months. However, as in previous months many of those jobs were imputed jobs by the flawed so-called "birth/death" model. Indeed, they again raised the number of assumed jobs to 66,000 per month, up from 64,000 in March and 53,000 in February. Thus, even as reported employment continues to decline, the government implausibly assumes that new business start-ups are accelerating.

During the latest 12 months, private sector employment according to the government is up by 238,000. However that includes the 787,000 "birth/death" model-jobs, so actually reported private sector employment is in fact down by 549,000. That decline is of course something which occurred entirely during the latest 6 months, when total private sector employment is down 282,000 to which one can add the 394,000 imputed jobs, meaning that actually reported private sector employment is down 676,000, or on average 113,000 per month.

This was probably what was at work in the sectors that supposedly created so many jobs, and so supposedly to a large extent compensated for continued massive job losses in manufacturing and construction. Supposedly, professional services -meaning more specifically accountants, computer programmers and administrative services- and health care saw massive employment gains of a magnitude that looks suspicious. They did in fact also have massive number of jobs imputed by the "birth-death" model, although it is unclear to what extent that reflects seasonal factors and to what extent it reflects seasonally adjusted numbers, as the Bureau of Labor Statistics strangely only report non-seasonally adjusted numbers for jobs imputed by the "birth-death" model.

The one bullish indicator in the report was the decline in the unemployment rate and the rise in employment in the household survey. However, the household survey number tends to be extremely erratic and volatile on a month to month basis so that number can be ignored unless it is repeated during the coming months, which is highly unlikely. Meanwhile, the decline in hour's worked and hourly earnings in particularly manufacturing implies a decline in industrial production and real income excluding the temporary so-called tax rebates.


Anonymous Anonymous said...

Hi Stefan

The birth/death numbers are not seasonally adjusted, the headline numbers are seasonally adjusted. So you cant just subtract one number from the other.

10:56 AM  
Blogger stefankarlsson said...

I know they're not seasonally adjusted, and I wrote that explicitly in this post. And that's why I didn't subtract the particular monthly number of 267,000. Instead, I took the total for the latest 12 months, 787,000 and divided it by 12, which is approximately 66,000. That is the average monthly addition during a whole year and so it is the best proxy for the seasonally adjusted addition from birth/death.

6:23 PM  

Post a Comment

<< Home