Rather than restate one of the key points, here is an excerpt from http://www.foxnews.com/story/0,2933,108290,00.html
"The Department of Labor provides two different sets of labor market statistics: the Establishment Survey and the Household Survey. The Establishment Survey polls 400,000 companies on how many employees they have. The Household Survey questions 60,000 households each month on whether they have jobs and whether someone is looking for one.
There are a number of technical reasons that the two surveys can yield different numbers. For example, people with more than one job will be counted multiple times in the establishment survey. On the other hand, the self-employed are only counted by the household survey.
The so-called ?jobless recovery? (search) picture is painted using the Household Survey for calculating the unemployment rate but using the Establishment Survey for the number of jobs created. The Household Survey can be used for both measures as it, too, provides estimates on the total number of people employed. But the two surveys have implied dramatically different changes in employment over the last few years. Over the last year, the Household Survey shows that almost two million new jobs have been created, while the Establishment Survey indicated a job loss of 62,000 jobs. Over the entire Bush administration, the Household Survey found that about 2.4 million new jobs have been created. By contrast, the Establishment Survey shows a net addition of only 522,000.
Why the difference? The number of companies does not remain fixed. Old firms die and new ones are born. The Establishment Survey finds out about the company deaths quickly, but it takes longer to learn about births. The current list of firms surveyed excludes firms started over almost the entire last two years. What the Establishment Survey shows is that total employment in older firms has changed little over the last three years. It completely missed the growth in new jobs among new startups and self-employment.
Not surprisingly, the choice of numbers is central to the political debate. Using the Establishment Survey, Democratic presidential candidates charge that over two million jobs have been lost under the Bush administration from January 2001 to January 2003...Part of the problem may simply be the complicated nature of these numbers. The Establishment Survey is much larger, surveying 400,000 businesses, compared to the 60,000 households used by the other survey. The Establishment Survey makes the claim of being more comprehensive, but the Household Survey is still quite large and it has always been the ?official? measure for calculating the unemployment rate. The systematic bias in the Establishment Survey is what is important: it completely ignores new startup businesses, and we have had an usually large expansion in this category over the last couple of years.
By only referring to the Establishment Survey numbers, the media has implicitly taken sides in the debate, albeit perhaps unknowingly. A simple Nexis computer search of the news media from Nov. 1, 2003, to Jan. 10 finds 975 stories using the term ?jobless recovery? to describe the U.S. economy... Eventually the Establishment Survey numbers will be adjusted for all the new startups that have sprung up over the last couple of years. Unfortunately, much of this won?t be reported until after the 2004 elections when all but a handful of historians and economists will pay attention."