If adjustments are made to the household survey to make it count jobs in a manner comparable to the establishment survey, then instead of an increase of 842,000 jobs since August 2002, the household survey indicates a loss of 425,000 jobs — almost as large as the 560,000 jobs lost according to the establishment survey.
Krueger does not go into detail about what "adjustments" are made. One of the controversies concerns an increase in self-employment, which would show up in the household survey but not in the payroll survey. If one of the "adjustments" is to take self-employment out of the household survey, then I would see that as a substantive adjustment, not simply a technical one.
One concern I have is that if Jane Doe separates from the payroll of BigCorp to become a self-employed consultant for BigCorp, her job disappears from the statistics but her output does not. If that is the case, then productivity growth would be artificially boosted. In private correspondence, Brad DeLong suggested to me that this is not the case--he says that BigCorp's value added would be net of Jane's consulting services. I agree that this would happen in principle, but I am not sure that the productivity statistics would use the correct net value added measure in practice.
For Discussion. What is the likelihood that some of the weak job growth and strong productivity growth of the past two years is in part a statistical mirage?