There was a lot of discussion surrounding the BLS’s benchmark revision. Dave Altig at Macroblog has an excellent review, while Ritholz at Big Picture is perturbed about the magnitude of the data revision. Without claiming to have greater insights into why the revision is so large, I do think a look at the context would be useful.
First, note that while the 2006 preliminary benchmark revision, at 0.6 percent, is unprecedented in size, 0.4 percent revisions have occurred twice before in the 11 years revisions have occurred.
Table 1. CES national total nonfarm benchmark revisions,
Second, as has been remarked elsewhere, the benchmark revision moves the establishment series closer to the “adjusted” household series, and so potentially solves one puzzle. But recall that this is a preliminary adjustment. We’ll have to wait and see if indeed the 810,000 figure sticks. (This powerpoint presentation by Steven Braun notes some of the reasons why the establishment and the household measures can go astray, and makes an observation on the firm life/death ratio issue.)
A graph of the data shows how the revision affects overall trends as measured by the two series. In Figure 1, the actual October payroll employment series (blue), the BLS’s “research” adjusted household series (made to mimic the establishment series, see here) in green are shown. In addition, I have pro-rated the April 2005-March 2006 establishment figures to incorporate the 810,000 additional jobs implied in the preliminary benchmark. I assume the payroll employment series remains 0.6 percent higher than in the official series. This “adjusted” series is red.
Figure 1.Payroll employment (blue), adjusted payroll employment (red) and “adjusted” Household employment series (green). Sources: FRED II, BLS, NBER, and author’s calculations.
Third, we come to perspective. I thought it would be useful to see whether the revision, against a backdrop of 136 million employees, changes one’s view regarding the trends in employment growth during this expansion vis a vis the previous one.
Figure 2.Official payroll employment-current expansion (blue), adjusted payroll employment-current expansion (red) and official payroll employment series-previous expansion (green). Sources: FRED II, BLS, NBER, and author’s calculations.
The answer is no. While large in absolute number, the revision does not change the fact that employment growth has been remarkably low in this expansion.
Finally, if my assumptions are not far off the mark, then the upward revision in the employment growth up to March 2006 merely accentuates the deceleration in employment growth in recent months.
Figure 3.Month-on-month annualized (log) percentage change in official payroll employment-current expansion (blue), adjusted payroll employment-current expansion (red) and official payroll employment series-previous expansion (green). Sources: FRED II, BLS, and author’s calculations.
The “adjusted” household series is still exhibiting strong, and accelerating, growth. Which one we should pay attention to has been a source of vigorous debate here and here. While Brookings’ George Perry (Brookings Papers on Economic Activity 2005(2)[abstract]) argues for equal weighting on the household and establishment series, while in his comments William Wascher (Deputy Associate Director, Division of Research and Statistics at the Fed), demurs, and argues for continued reliance on the establishment series.