- The Washington Times - Monday, November 22, 2004

As the Roman philosopher Seneca the younger said, “Veritatem dies aperit” — time discovers truth. Perhaps for the subject at hand, Thomas Mann said it better: “Time clarifies.”

So it is with economic time series data that are subject to repeated revisions. Take employment, one of our most important economic indicators.

The total of U.S. nonfarm payroll jobs — the employment measure preferred by most economists and data users — is reported monthly by the U.S. Bureau of Labor Statistics (BLS). The data for a given month are reported on the first or second Friday of the following month. The initial payroll number is revised the next month and then revised again the following month as more returns come in from employers. The job count is further revised by continual updating of monthly seasonal adjustment factors.

Then once a year the sample survey data are benchmarked to a more comprehensive count of payroll jobs from state unemployment insurance tax records. The benchmarking revisions mainly correct for nonsampling errors of response, coverage, and processing.

Payroll jobs data revisions are also necessitated by sample redesign, updates to the model that estimates business births and deaths, changes in industry definitions and classifications, and changes in estimation methods.

How much do the data revisions change the initially reported job count, and do the changes have economic significance?

The short-term revisions in total payrolls, i.e., the revisions in the two months following the first preliminary estimate, vary considerably in size and sometimes tell a different story from initial estimates. The initially reported increases in payroll jobs for July and August were revised upward after two months by more than 50,000 each month, and the first September revision raised that month’s initially reported gain by 43,000.

Each of the first four months of this year show a similar understatement of job growth. Comparing the initial estimates of job change with the estimates reported two months later for January through April shows initially reported monthly increases were underestimated on average by about 50,000 jobs. Looking at the current job recovery, which started in September 2003, both the preliminary and revised data show the same turning point, and the short-term revisions to the jobs data do not alter the initially reported direction of any month-to-month change. However, the data show the initial estimates understated job growth in most months of the recovery. Only October 2003 and May and June of this year were overestimated, though the average monthly understatement during the recovery exceeded the average overstatement. Underestimates occurred both when the initial increases were large and when they were small. As more employer payroll reports came in, the job picture brightened.

Recently the BLS reported a preliminary estimate of the annual benchmark revision that will be released in February 2005. Nonfarm payroll jobs for March 2004 and subsequent months are expected to be revised upward by about 236,000, with the prior 11 months also adjusted upward on a graduated scale. Considering there was an underestimate in the initially reported level of jobs in the first few months of this year based on short-term revisions, this implies a significant overall initial understatement.

In 2006, the BLS will publish benchmark revisions for 2005, which will result in further revisions in the jobs data back to April 2004.

Continuing jobs data revisions are inevitable and are evidence of the BLS’s heroic efforts to constantly improve. But it would help if the lead-off statement about the latest job number in the monthly BLS press release used the word preliminary. Every month the media highlight the latest job change without proper caveat, and later revisions, which are sometimes much later, are often back-paged or ignored. The press and TV commentators would do well to hang a sign on their computers quoting the wisdom of Seneca.

Alfred Tella is former Georgetown University research professor of economics.

Sign up for Daily Newsletters

Manage Newsletters

Copyright © 2020 The Washington Times, LLC. Click here for reprint permission.

Please read our comment policy before commenting.


Click to Read More and View Comments

Click to Hide