Annual Revisions to Metro-Level Jobs Data Sheds New Light on Job Growth
On March 21, the Bureau of Labor Statistics (BLS) released benchmarked State and Metro-Area Employment, Hours, and Earnings (SAE) data. The benchmarked data, which come out once a year, are the most accurate employment statistics available for metropolitan areas. Previous versions of the metro-level jobs data, like the initial SAE estimates released five weeks after the close of each month, can often be revised substantially when the benchmarked data come out.
So the release of the benchmarked SAE is a moment of truth for the BLS and for metro areas. For the BLS, the size of the revisions to the initial SAE shows how accurate the earlier estimates were. For metro areas, the benchmarked SAE provides a reliable and complete estimate of their employment through September 2013.
Every month, the initial SAE estimates how many jobs states and metro areas had two months earlier. These initial estimates are based on a combination of responses to the Current Employment Survey and imputation models for areas where the sample size is too small to generate reliable estimates. Even if done perfectly, estimates from survey data or models are subject to errors. Once a year, the SAE data are benchmarked to the Quarterly Census of Employment and Wages (QCEW), which provides counts of jobs based on administrative data covering 98 percent of employment. These data come from a census of jobs, so they are more accurate than the initial SAE. (For more detail on the benchmarking process, the reliability of the initial estimates, and the advantages of using the QCEW to measure employment growth, see “Which Estimates of Metropolitan-Area Jobs Growth Should We Trust?”
We assess the magnitude of the latest revisions by comparing final and benchmarked jobs estimates for September 2013. We focus on this month because the most recent benchmarking revised the SAE data from October 2012 to December 2013, but the QCEW is only available through September 2013. The SAE estimates for October 2013 to December 2013 are subject to additional revisions with the next round of benchmarking in March 2015. We define revisions as the difference between the benchmarked SAE and the initial SAE.
For all metro areas in the US, the average revision in the benchmarked data is plus or minus 1.2 percent of employment. So if a metro area with 500,000 jobs had the average revision, benchmarking would add or subtract 6,000 jobs to or from the initial estimate. Revisions are typically larger for small metro areas than large ones, in percentage terms. The average revision for metro areas with populations below 250,000 is plus or minus 1.6 percent, double that of metro areas with populations above one million.
We now turn our attention to a dozen metropolitan areas in the Fourth Federal Reserve District. Their revisions ranged from −2.0 percent in Lexington to +2.7 percent in Lima. Only Akron, Dayton, and Youngstown had smaller revisions than the national average.
The revisions can have a big impact on a metro area’s year-over-year change in jobs. If the initial SAE shows that the number of jobs in a metro area grew 1.0 percent, then the average revision would change that to either a loss of 0.2 percent or a gain of 2.2 percent.
The chart below compares the initial and revised percent change in jobs from September 2012 to September 2013 for these Fourth District metro areas. The horizontal axis is the initial year-over-year percent change in jobs, and the vertical axis is the revised year-over-year percent change in jobs. The initial estimates underestimated the jobs growth of the metro areas above the diagonal line and overestimated the growth for the metro areas below the line. The further away the point is from the line, the more impact the revision had on the year-over-year change in jobs. To see the exact figures for each metro area, look at the last two columns in the table below (or see this spreadsheet for all metro areas in the US).
What do these revisions in the year-over-year changes mean in terms of jobs? The initial SAE said that the Cleveland metro area lost 7,200 jobs from September 2012 to September 2013, but it actually gained 7,400 jobs. Columbus gained almost twice as many jobs as originally estimated, 25,900 instead of 12,900. Revisions giveth, revisions taketh away: Pittsburgh gained 2,400 jobs, not the 20,000 jobs suggested by the initial estimate. (See the table below for the numbers for the other Fourth District metro areas or see this spreadsheet for all metro areas in the US.)
In the table below, the metro areas are ordered by the percent change in employment from September 2012 to September 2013 based on the revised data. Overall, metro areas in the US had jobs growth of 1.9 percent over this time. Columbus and Lexington both grew faster than metro areas overall. With gains between 1.7 and 1.8 percent, jobs growth in Wheeling, Cincinnati, and Toledo was just under that of US metro areas as a whole. In Lima, Akron, Cleveland, and Pittsburgh, jobs grew but less than half as fast as metro areas overall. Dayton and Youngstown had no change in the number of jobs, and Erie lost 0.3 percent of its jobs over the 12 months.
|Metro areas||Jobs in September 2013||Year-over-year change in jobs||Year-over-year percent change|
Note: See this spreadsheet for all metro areas in the US.
Source: Authors’ calculations from Current Employment Statistics State and Metro-Area Employment, Hours, and Earnings
accessed through Haver Analytics.
People are naturally interested in the first set of jobs estimates that are made available, but they should be equally interested in the revisions to those estimates. The initial estimates are like a rough sketch, giving the general outline. The benchmarked estimates are like a good photograph, showing exactly what is.