Data Revisions Change Region’s Employment Picture
The Bureau of Labor Statistics (BLS) has revised its labor force estimates for the period 2005 through 2010. This was done to reflect new Census Bureau population controls, updated input data, new statewide controls, and re-estimation. While these revisions occur regularly, this latest round of revisions produced fairly large changes in Pittsburgh area data for the number of unemployed, employed and labor force. These changes show the area’s unemployment rate to have been significantly lower than first reported during much of 2010.
As we noted in January 2009 (Policy Brief Volume 9, Number 2), the labor market reports indicated the area’s labor force had exhibited inexplicably rapid growth from 2007 to 2008, averaging nearly 21,000 more labor force participants every month compared to the year earlier level. In that Brief we speculated that the questionably strong numbers would eventually be revised and questioned whether the number of employed and/or unemployed might also actually have been greater than those being reported at the time. In April 2011, we have the answer. All three labor market measures have undergone major revisions.
As mentioned earlier the largest jump in the labor force occurred with the 2008 monthly data and as expected, the BLS has revised this data downward for ten of the twelve months (excluding January and February), but the reduction was modest, a couple hundred people in every month except August (1,031), November (1,166) and December (1,160). However, in 2009 every month was revised downward ranging from a drop of just over 1,000 in February to nearly 17,500 fewer in November. This pattern of downward revisions continued through the entirety of 2010. January 2010’s labor force count was lowered by more than 24,300 compared to the number being reported just a couple of months ago. February’s cut was similarly large at nearly 28,400. In all, the average monthly labor force reduction for 2010 was just over 16,000. Thus, as we had expected would eventually happen, it appears as if the heavy increases tabulated in the 2007-2008 period were largely reversed with the 2009 and 2010 revisions.
In the 2009 Brief, it was noted that the labor force count in the region has been running at roughly one half the total population in the region. Therefore a jump in the labor force, as was evidenced in the data from 2007 to 2008, would most likely be the result of substantial population gains in the seven counties that comprise the Pittsburgh region. To be sure, the climbing labor force could have resulted from a large and unexplained increase in the number of people in the working age group while population remained flat or perhaps a major upward move in the labor force participation rate of women. Neither of these possible explanations seems very likely as they would require sizable and rapid shifts in demographics or in behavior governing labor force attachment.
In Allegheny County, the region’s most populous county, population numbers show that, with the exception of a small increase between 2008 and 2010, most of the years in the decade were marked by population decreases. In fact, over the decade only two counties, Butler and Washington, had population increases. But they were not enough to offset the losses from the remaining five counties, and certainly not enough to explain such large increases to the labor force. Thus, there was a high probability that downward revisions would take place. The problem in these revisions coming in later years rather than in the inexplicable jumps in 2008 is they appear to be compensatory rather than correcting the real mistake made earlier.
Moreover, as could be readily foreseen, with large decreases to the labor force, there were almost certainly going to be accompanying reductions in the number of employed and unemployed. In nearly every month during 2008 through 2010, the only exception being January and February of 2008, the employment numbers were revised downward. The average monthly downward revision in 2008 was 500 while in 2009 it was nearly 5,800 and in 2010 it was just over 11,700. Unless the unemployment rate was to fall to low levels no one would believe, the labor force had to be revised downward. From 2008 through 2010 there were steady real declines in the area’s employment count as the region and country sank into the recession.
The number of the unemployed was also revised during this period. In 2007 and 2008 (pre-recession) the majority of monthly numbers were revised up slightly. The average monthly increase for 2007 was just less than 100 and for 2008 it was 134. However for 2009 and 2010 the number of unemployed was revised downward in 23 of 24 months (February 2009 had an increase) because the reductions in labor force estimates exceeded the reductions in employment. To illustrate; if 1000 people leave the work force and employment drops by only 800, then the number out of work falls by 200. The average monthly downward revision in unemployment in 2009 was just under 1,500 while in 2010 the average monthly decrease was about 4,350.
Of necessity, the unemployment rate (the ratio of the number unemployed to the labor force) was affected by the revisions to the labor force and the number of unemployed. Because of the relatively small revisions to labor force in 2007 and 2008, the rate was virtually unchanged during the months of 2007 and 2008. However, there were more significant changes in late 2009 and 2010. In three of the final four months of 2009, the unemployment rate was revised downward by two tenths of a percent (September, November, and December). In May, June, and July in 2010, the unemployment rate was revised downward by more than one-half percent-being revised from 8.5 to 8.0 percent in both May and June and from 8.6 to 8.0 percent in July.
As we anticipated back in January 2009, there were major revisions to the Pittsburgh region’s household survey data. The unexplained jump in the labor force from 2007 to 2008 was largely negated by downward revisions to the data in 2009 and 2010. But as noted above, the corrections should logically have focused on 2008.
In sum, the labor market data so eagerly awaited by the media and analysts, with its large periodical revisions and its oftentimes truly head scratching anomalous numbers, must be viewed with a degree of skepticism and taken to be a first approximation and not precise scientific truth.