Thursday, September 27, 2012

BLS adds 386,000 jobs to the economy

The economy may be doing better than we initially thought.

This morning, the Bureau of Labor Statistics (BLS) released its preliminary revision of the March 2012 jobs tally. The BLS expects that when the revised estimates are released next year, the figures will show that the U.S. had 386,000 more jobs in March 2012 than were initially reported.

Revisions to the jobs figures are necessary because the initial estimates are based on a survey with some sampling noise and with an imperfect sampling frame. Some months after the initial estimates are made, the BLS is able to compare and benchmark them to nearly complete data gathered from payroll tax records. Estimates throughout the year are then readjusted to this benchmark.

When the economy was declining in 2008 and 2009, the BLS initially under-estimated the job losses by 902,000 positions, and we didn't find out how truly awful the employment situation was until a year or so after the actual decline. When the Obama administration famously forecast that its stimulus package would hold unemployment to 8 percent, no one realized that the U.S. was actually nearly a million jobs further in the hole. Everyone knew things were bad--we just didn't know that things were that bad.

Now that the economy is improving, the BLS appears to have under-estimated job gains by roughly a third of a million.

Why does the BLS keep missing the mark? Part of the reason has to do with the entry and exit of businesses during economic swings. Each month the BLS chooses businesses to survey to get its employment numbers. The sampled businesses are selected from a roster of businesses that provided payroll tax records in a recent quarter. Businesses that were created since the last set of payroll tax records were processed aren't on that roster and can't be part of the frame. Similarly, the BLS misses some of the businesses that stopped operating. The BLS makes statistical adjustments for these issues, but these adjustments are based on historical trends, not the sort of huge swings that we saw during the recession and are now seeing during the recovery. The recent pattern has been for the BLS initially to undercount losses during declines and to undercount gains during upswings.

The revision for 2009 was unusually large--a change of 0.7 percent in the employment figures. The current revision, which works out to 0.3 percent, is right at the edge of the historical 0.1 to 0.3 percent average.

In February when the final numbers are released, we will probably discover that the net number of jobs in the U.S. increased over the course of President Obama's first term. For time being, however, the official estimates indicate that there are fewer jobs in the U.S. than there were in January 2009.