The headline numbers for changes in non-farm payroll employment are significantly overstated according to our examination of the data in the Bureau of Labor Statistics (BLS) website. This occurs as a result of the distorting effects of two factors—- seasonal adjustments over a year’s time and the annual benchmark revisions that correct earlier incomplete data.
Let’s deal with the seasonal adjustment distortion first. The recently released employment number for October indicates that seasonally adjusted jobs increased by 829,000 over October 2009. However, BLS data show that non-seasonally adjusted employment increased only 626,000 over the same period. Since normal seasonal fluctuations are not a factor when comparing any number to a year-earlier period, monthly seasonal adjustments in this case are overstating the change in employment by 203,000. In other words, the actual increase in employment for the 12 months ending October is 203,000 less than the total you would get by adding up the last 12 months of the seasonally-adjusted data that everyone uses.
The second distorting factor is the annual benchmark revision, which the BLS estimates at this time each year. In line with this routine practice the Bureau recently announced an estimated downward adjustment to March 2010 nonfarm employment of 366,000. This essentially wipes out more than the entire birth/death adjustment for the period of 336,000. This will be the second consecutive year in which the benchmark revision more than wiped out the additional jobs added by the birth/death adjustment. The final benchmark revision will be issued with the employment report to be released on February 4, 2011.
We further note that for the 12 months ended October 31st, 2010, the birth/death estimate added 496,000 jobs to the non-seasonally adjusted monthly numbers. Since it is highly unlikely that new business startups have actually added any jobs in that period, it is probable that these estimates will also be wiped out by subsequent benchmark revisions. Thus, when we eliminate this birth/death adjustment of 496,000 from the 626,000 added for the 12 months ended October 31st, we are left with only 130,000 jobs added for the period rather than the 829,000 that everyone uses without actually analyzing the data that is readily available on the BLS site.
To sum up: Start with the 829,000 jobs added on a seasonally adjusted basis. Subtract the 203,000 jobs added through the distortion of seasonal adjusting. That leaves 626,000 non-seasonally adjusted jobs added. Then subtract the 496,000 jobs added by the birth/death adjustment that will probably be wiped out by a subsequent benchmark revision. That leaves an estimated 130,000 jobs added for the entire 12 months.
Note: All of the statistics cited above can readily be found in the BLS website at bls.gov. For monthly changes in payroll employment, both seasonal and non-seasonal as well as the birth/death adjustment, see “CES Net Birth/Death Model”. For the benchmark revision, see “CES Preliminary Benchmark Announcement”.