In the January jobs report, the Bureau of Labor Statistics (BLS) incorporated its benchmark revision and lowered the number of jobs estimated by 376,000. This revision is based on data from state unemployment insurance (UI) offices, which covers more than 99 percent of payroll employment. The downward revision implies that employment growth had initially been overestimated by approximately 30,000 jobs per month in the year from March 2006 to March 2007. While the size of this revision was known since September, when BLS released its preliminary estimate, the inclusion of the revision in this month's report should have opened a few eyes. The BLS had been imputing roughly the same number of jobs for new firms not counted in the survey in 2007 as in 2006. This imputation is presumably the source of the overestimate of job growth. If we were imputing too many jobs in 2006, then we must be over-estimating job growth in 2007 by even more, since the economy is growing more slowly. This means that when we get the benchmark revision based on the March 2008 UI data, it is likely to show an even sharper downward revision than the one we just saw. There is one other noteworthy item that the vigilant reporters writing on the jobs report managed to overlook. The percentage of the workforce employed in manufacturing fell below 10 percent for the first time since we've been keeping the data. The data show that we fell below the 10 percent threshold in October. We didn't know it at the time because BLS had not yet adjusted its employment numbers for its benchmark revision. Of course, those who get CEPR data bytes already knew all this.
--Dean Baker
