August 7, 2004


A Problem with Payrolls? (Tim Kane, Ph.D., and Rea Hederman, August 6, 2004,

Here. We. Go. Again. The Department of Labor's July Employment Situation Report is out this morning, and the news is good. And not good. The unemployment rate stands at 5.5 percent in July, compared with 6.3 percent a year ago and an average of 5.8 percent in the 1990s. Fundamentally, the employment situation in America is solid, but the sluggishness of payroll job growth is stunning and out of line with a host of positive indicators. The strength of today’s labor market is yet another sign that the President’s 2001 and 2003 tax cuts were effective policies and must be extended.

Employment growth is not a problem in America, but the establishment survey of payrolls has a clear problem. According to the official statement from the commissioner of the Bureau of Labor Statistics, which publishes the monthly figures, “As measured by the household survey, employment rose by 629,000 over the month, compared with a change of 32,000, as measured by the establishment survey.” Since 1948, there have been 62 instances where household data reported 500,000 new jobs or more, coinciding with a payroll average of 240,000. July’s net gain in payroll jobs marks the 11th straight month of rising payroll employment but still came in well below analysts' expectation of 247,000. What's going on?

The key, once again, is the difference between the two employment surveys. The payroll survey measures employment at established firms but does not count self-employed consultants, contractors, farmers, and the like. Worse, the payroll survey is highly susceptible to changing rates of turnover. When job-changing dips due to higher uncertainty, the payroll survey artificially deflates. So the gap between the two surveys, which had narrowed a bit in recent months, is now at 3 million (see Chart 1). Since March 2001, the payroll survey suggests there are 1.24 million fewer jobs, in contrast to the household survey’s measure of 1.81 million more working Americans—more than ever before.

And today of all days, the Labor Department published its very first, preliminary assessment of the impact of job-changing on the payroll survey (see "Effects of Job Changing on Payroll Survey Employment Trends"). It acknowledges the Heritage Foundation's analysis from March 2004 (see “Diverging Employment Data: A Critical View of the Payroll Survey” by Tim Kane, Ph.D.) that worker turnover has a significant effect on the payroll survey, inflating the total count by over a million jobs, but also by over-emphasizing job losses during business cycles when turnover declines. Since March 2001, BLS estimates that the "job-changing effect" has led to an overstatement of 251,000 job losses in currently published data.

When you have a stack of data and one datum contradicts the others, which do you suspect?

Posted by Orrin Judd at August 7, 2004 2:46 PM

I wonder if the market went down on the jobs numbers because of fear that W will now lose instead of because of the data itself.

Posted by: JAB at August 7, 2004 9:16 PM


Drug company & health care stocks have been going down for the same reason.

Posted by: at August 8, 2004 1:53 AM

"When you have a stack of data and one datum contradicts the others, which do you suspect? "


Posted by: Jeff Guinn at August 8, 2004 9:55 AM

"When you have a stack of data and one datum contradicts the others, which do you suspect? "

The one that confirms your bias.

Posted by: Robert Duquette at August 8, 2004 11:41 PM

Written like a Democrat Robert.

Posted by: genecis at August 8, 2004 11:45 PM