QUICK LINKS Beige Book (Federal Reserve Board) Texas Index of Leading Indicators

Regional Economy
March Madness: A Major Upset with Job Numbers

Julia Kedrova looks at the effect of the BLS's recently revised employment data.

© 2002 The New Yorker Collection from cartoonbank.com. All Rights Reserved.

Data revisions arise because two sources supply information about employment. The first is a monthly employment survey that covers about 350,000 firms nationwide and gathers initial job numbers. In Texas, the survey covers over 24,000 firms. The second source is the unemployment insurance-based job count that covers nearly all U.S. employers. The latter is more accurate, but it is released six to nine months after the survey. Therefore, every year the initial job numbers are realigned with the unemployment insurance-based numbers. When the two sources diverge, significant revisions result.

We once thought that Texas employment grew 1 percent in 2001; we now know it shrank 1 percent—a difference of 215,000 jobs. The Texas revisions were proportionately most extreme in temporary employment industries (16 percent) and wholesale trade (5 percent).

For metropolitan areas, the most significant downward revisions in job growth were 3.2 percent for Dallas, 2.74 percent for Fort Worth and 2.63 percent for Austin (Chart 1). Before the revision, Dallas, with its 33,700 new jobs, was second only to Tampa, Florida, in job gains nationwide. After the revision, Dallas lost 31,500 jobs—an error of over 65,000 jobs.

Chart 1

The main explanation for such a large but not unprecedented revision is the economic downturn. Data gathering near economic turning points is always problematic for both public and private forecasters. Analysts believe the sample overestimates job growth during an economic slowdown because it fails to account for business failures. The BLS does a bias adjustment at the national level. However, at the state level, bias adjustments are either not done or are applied conservatively. As a result, the employment numbers for the states undergo greater revisions when the employment trends are changing rapidly.

The revisions are meaningful to anyone who relies on economic forecasts. For example, construction companies use job data to plan their projects. City governments use job data to forecast tax revenues and budget appropriately. After being caught off guard by the magnitude of this revision, many decisionmakers may have to change budget and profit forecasts.

One economic implication of the revision is the timing of the recession as it incorporates the national estimate of employment. The national and sum-of-states estimates of job numbers differ. [1] The national estimate is routinely bias adjusted, but state estimates are bias adjusted at the states’ discretion. Thus, generally the national estimate is more accurate.

However, the sum-of-states estimate was revised with the unemployment insurance-based count in March. After the revisions, the sum of the states is much closer to the unrevised national estimate (Chart 2). Arguably, until the national estimate is revised, the sum of the states offers a better estimate of job numbers because it draws on more accurate information. The national estimate is revised with the unemployment insurance-based count in June. Historically, after both estimates are revised, the numbers are nearly identical.

Chart 2

The National Bureau of Economic Research (NBER) officially declares the timing of national recessions. The NBER’s dating a national recession to March 2001 appears to coincide with the peak in nonfarm employment. If the national estimate is revised into harmony with the revised sum-of-states estimate, nonfarm employment peaked in January. This may imply that recession began in January, not March.

About EYI | Global Economy | U.S. Economy | Regional Economy | Free Enterprise | Money & Banking | Technology
Federal Reserve Bank of Dallas | FRB Dallas Publications