A great deal of effort goes into forecasting future levels of economic activity, and many people understand the difficulties that are faced in making such forecasts.

There is significantly less understanding, however, of the difficulties faced in determining economic performance in the recent past. In fact, all kinds of economic data undergo revisions over time, meaning that our view of the past is constantly changing. Sometimes, the revised data tell a story that differs greatly from what was told by original data. This is especially true of employment data for metro areas.

On March 8, the Bureau of Labor Statistics (BLS) released its latest annual benchmark revisions to the payroll employment data for every metro area in the United States. Monthly employment estimates going back to April 2005 were affected by these revisions. In addition, new population controls resulted in small revisions to the data further back in time.

Compared with the initial estimates released in January, the latest revisions result in a significant upgrading in the perceived economic health of metro areas in the Eighth District.1 The table presents the pre- and post-revision estimates of employment growth in 2006 for all 18 metro areas in the Eighth District. Summing over the 18 metro areas, the benchmark revisions greatly increased estimated job growth: The initial estimate across all of the metro areas in the Eighth District was an increase of 34,900 jobs, whereas the new, post-benchmark estimate indicates 60,700 more jobs.

All four of the largest metro areas in the District saw upward revisions in the estimates of their employment growth. The St. Louis and Louisville metro areas saw dramatic improvements in their employment estimates, while the revisions in Memphis and Little Rock were more measured.

Before the revisions, the employment estimates suggested that St. Louis and Louisville both experienced relatively grim years in 2006. St. Louis was said to have seen a loss of 400 jobs (–0.03 percent), while Louisville was thought to have seen an increase of only 3,900 (0.6 percent). The new estimates, however, present completely different pictures. St. Louis and Louisville are now seen as having generated 13,600 jobs (1.0 percent) and 11,900 jobs (1.9 percent), respectively. The new story is that St. Louis saw positive but slow employment growth that was somewhat below the rate for the United States (1.7 percent), while employment in Louisville grew somewhat faster than it did for the country as a whole.

Before the revisions, employment in the Memphis metro area was thought to have increased last year by 6,100 (1.0 percent), while the corresponding employment growth in Little Rock was 5,900 (1.7 percent). The new estimates indicate that the Memphis metro area generated 9,300 more jobs (1.5 percent), which puts its performance much closer to the national rate. For Little Rock, the new estimate of the net number of jobs created in 2006 is 6,800 (2.0 percent), which now puts the area as an above-average job producer for the year.

Some very large revisions, both up and down, occurred for the employment growth estimates of the 14 small and medium metro areas in the Eighth District. Three large upward revisions, each of which was of greater than one percentage point, occurred for Bowling Green, Elizabethtown and Fort Smith. Bowling Green saw its employment picture change from slightly below the national average to well above it, while Elizabethtown and Fort Smith saw their employment pictures completely reverse. Each had initially been thought to have experienced mild job losses in 2006, but now are estimated to have experienced job gains well above the national rate.

At the other end, Evansville, Hot Springs and Pine Bluff all saw significant downward revisions (one percentage point or more) in their employment growth rates for last year. For Hot Springs, the downward revision meant that its 2006 performance was downgraded from startlingly good to only above average. Evansville and Pine Bluff, on the other hand, went from having positive to negative job growth for the year. The revision was especially dismal for Pine Bluff, which went from above-average job growth to significant job losses.

Jobs Data and Benchmarking

So how is it that the pictures of local economies can change so much? The payroll employment in a metro area—the number of jobs—is provided by the Current Employment Statistics (CES) program of the BLS. According to the BLS, each month it surveys “about 160,000 businesses and government agencies, representing approximately 400,000 individual work sites,” from around the United States. Although the survey covers hundreds of thousands of employers, these employers make up only a small percentage of all businesses and work sites in the country. (According to the BLS, there were more than 8.8 million such establishments in the United States in June 2006.)

To calculate a comprehensive measure of metro area employment, the BLS needs to estimate the number of establishments in the area. This is the primary reason for the sometimes-large revisions to the CES data: the difficulty in estimating the number of establishments. When the economy is in recovery, for example, new firms might be setting up and hiring workers very quickly. The BLS doesn’t find out about the new firms or jobs until the unemployment insurance records are updated, which can take several months or more. This lag is compounded by the fact that small firms, which provide the bulk of jobs, might need to provide unemployment insurance information only once a year rather than monthly or quarterly, as is required of larger firms.

To estimate the number of establishments, the BLS relies on the Quarterly Census of Employment and Wages (QCEW). The QCEW is a tabulation of employment information for workers covered by state and federal unemployment insurance programs. Because of its comprehensive nature, data from the QCEW cannot be produced as quickly as data from the CES: Initial data are released six to seven months after the end of a quarter and are subject to subsequent revision. To fill in the blanks, the BLS estimates the number of establishments using the QCEW as a benchmark. Each year, the BLS establishes new benchmarks using updated data from the QCEW. Because of the lags and revisions to the QCEW data, the yearly benchmarking affects employment data from the CES going back 21 months. This is why the estimates just released have affected the yearly employment changes for 2005 and 2006. Note also that the estimates for job growth in 2006 will change again in March 2008 because much of the data for 2006 will be affected by the benchmark revisions that will occur then.

Joshua A. Byrge and Kristie M. Engemann provided research assistance.

Endnotes

1. A detailed analysis of these revisions and the recent history of benchmark revisions is provided by Michael R. Pakko and Howard J. Wall in, "Post-Casting Employment in the Eighth Federal Reserve District: Revised Data for Metro Areas," CRE8 Occasional Report No. 2007-01, March 2007. See http://research.stlouisfed.org/regecon/op/CRE8OP-2007-001.pdf. [back to text]