In lieu of an abstract, here is a brief excerpt of the content:

  • Comment and Discussion
  • William Wascher

William Wascher:1

George Perry has written a very interesting and provocative paper aimed at anyone engaged in analyzing and forecasting the U.S. economy. As Perry notes, analysts place considerable emphasis on the Bureau of Labor Statistics' release of employment data on the first Friday of each month, essentially because these data provide one of the earliest readings on the pace of economic activity in the preceding month. One difficulty in interpreting the data, however, is that the release includes results from two independent surveys, of households and of establishments, which often give different signals about the strength of the labor market. For example, in the last month for which data were available as this volume went to press (November 2005), the household measure of employment fell by more than 50,000, whereas the payroll measure increased by more than 200,000.

Perry argues that analysts place too much weight on the monthly employment changes in the payroll survey and consequently too little on those in the household survey. In his view, analysts have misinterpreted the smoothness in the employment changes from the payroll survey as evidence of greater accuracy, because they have not taken into consideration some other shortcomings of that survey. Instead, according to Perry, forecasters would be better advised to average the monthly changes from the two surveys.

The paper's first section presents evidence on whether market participants and policymakers focus more heavily on the monthly employment changes [End Page 312] from the payroll survey than on the corresponding data from the household survey. Perry's results show that federal government bond yields respond significantly to surprises in the monthly change in payroll employment immediately following the release of the data, and hardly at all to surprises in household employment. An unavoidable limitation of this analysis is that the "surprises" for both the household and the payroll employment changes are based on a single mean employment forecast gathered in a survey conducted by Money Market Services, which explicitly refers to the payroll survey. I suspect that if market participants were asked instead about the household survey, they would take into account various technical factors and time-series properties of those data, which would lead their forecasts for that employment measure to differ from that for the payroll survey. This suggests that the household employment surprise is likely measured with considerable error, which would seem to diminish the information content of Perry's regressions. Even so, the simple fact that market participants forecast only the monthly change in payroll employment, along with the existence of a derivatives auction for that portion of the release, indicates that these analysts do pay considerably greater attention to the payroll measure.

At first glance the same appears to be true of policymakers. The results in Perry's table 2 show that the announced federal funds rate target after each FOMC meeting responds more strongly to changes in payroll employment than to changes in household employment. Taken literally, the coefficients in column 2-3 of that table suggest that policymakers place about twice as much weight on the payroll survey as on the household survey. However, such a result is not necessarily evidence of a strong preference for the payroll survey. The FOMC meetings often take place well after the employment data are released, and during that period a whole host of other indicators of economic activity become available, all of which are in the FOMC's information set. In this context an alternative interpretation might be that the payroll data line up better than the household survey data do with the broad array of information used by the FOMC in making monetary policy decisions.

Nonetheless, there does seem to be a general preference among forecasters for the payroll survey, and this preference presumably reflects the view that the payroll survey's measure of employment growth has a higher signal-to-noise ratio than does the household survey measure. Perry's Figure 2 illustrates the rationale for this prevailing wisdom. With the monthly change in [End Page 313] household employment regularly exceeding plus or minus 500,000, one often would seem hard pressed to make heads...

pdf

Share