Chris Vogt brought to my attention an article from today's Wall Street Journal (available online to subscribers, page A18 in hardcopy) that is related to this post and this post. The article argues that the employment numbers from the household survey ought to be given more weight than they are by economists and policymakers.
The Bureau of Labor Statistics finally confirmed what we and others believed all along, that its "establishment survey" underestimated the number of jobs created from March 2003 to April 2004. The government bean-counters offered a preliminary upward revision of 236,000, and some believe the final correction early next year will be between 300,000 and 400,000 jobs.
It turns out that this economic expansion is different from those in the past, but not in the way that many thought. New jobs are being created as usual, but they are different kinds of jobs. The U.S. economy is undergoing a structural change as more people become self-employed or form partnerships, rather than working for large corporations.
This transformation confounds the government's employment surveyors because they rely on the payroll data of about 400,000 existing companies. These figures are then compared to the number of workers who were paying into state unemployment insurance programs in the previous year. This provides a fairly accurate picture of employment as long as the same proportion of employed people continue to work for existing companies.
But when a higher ratio of people make their livelihood as independent consultants to their old company, or as power sellers on eBay, they don't show up in the establishment survey that provides the most widely used employment figures -- at least not until the benchmark figures are revised long after the fact.
These jobs do, however, show up right away in the "household survey."
I'm not yet convinced that the payroll survey is giving us the wrong idea about the broad contours of the employment story. In particular, I still believe that, when all is said and done, job creation in this recovery will be viewed as an anomaly, if not a mystery. But I'd love to find out I'm wrong.