Phantom jobs and job losses

Public Interest, Wntr, 2005 by Tim Kane

The household survey offers a marked contrast. It is based on personal contact with 60,000 individuals, and provides demographic details by gender, education, and race. By interviewing respondents, it seeks to differentiate between people who are "in" and "out" of the labor force, and calculates an unemployment statistic representing the percentage of people in the labor force who are unable to find work.

Household data have their own set of problems, and controversies have flared up periodically over the mechanics of the survey--when phone calls are made, and to whom, for example--as well as over more fundamental issues. But it seems fair to say that these biases have been largely static over much of the life of the survey. The methodology of the payroll survey, by contrast, is unable to compensate for a new economy in which people have a plethora of employment options.

A tale of two (and more) surveys

The evidence that something is amiss in the payroll data lies in a phenomenon that astute economists have noticed in recent years: The payroll and household surveys have been diverging widely in the views they present of the American labor market.

The payroll survey indicated that 800,000 jobs were lost between President Bush's inauguration and the 2004 election. At its low point, 2.7 million jobs had been lost relative to the peak level of March 2001. After August 2003, payrolls expanded every month leading up to the election, adding 1.9 million jobs in just over one year--a strong showing, but not enough to create net job gains. The sluggish recovery of payrolls fueled the media's "job-less recovery" story line in 2003, and allowed candidate Kerry to call Bush's the worst presidential record on jobs since Herbert Hoover.

And yet, this labor market crisis only seemed to be appearing in the payroll survey data. Other official government measures signaled net job growth during the first Bush term, especially the household survey, which indicated that the number of American workers grew by 1.6 million under Bush. Other labor data confirm the positive view: Initial jobless claims were 10 percent below their historical average during 2004, and private-sector hiring indexes were setting records. As Andrew Sum, Paul Harrington, and Ishwar Khatiwada put it in a November 2003 paper for the Center of Labor Market Studies at Northeastern University,

  Total employment in the U.S. has increased much more markedly since
  the end of the recession than indicated by findings of the [Current
  Employment Statistics] payroll survey .... Since most of these sources
  of employment gains have occurred outside the formal employment
  system, total hours of labor input in the U.S. economy are likely
  being underestimated, thus exaggerating part of the rise in labor
  productivity since the end of the recession.

If all the other indicators suggest that the labor market is keeping up with other signs of economic improvement, where did all those "lost jobs" go? The most obvious answer is that they didn't go anywhere. Or to be more precise, more workers stayed put in their current jobs, thus depriving the payroll survey an opportunity to double count them.