Here are Three Things to Remember about The Government's Monthly Employment Reports:
First, the initial report for the current month by the Bureaus of Labor Statistics (BLS) has usually contained significant upward revisions to previous months, as shown here:
For the past seven months, the number reported for jobs added in the current month has been, on average, less than 2/3 of the total reported increase in jobs, because of significant revisions to prior months.
Second, as you would expect because of the first point, the current month's initially reported total has usually been revised upward quite a bit in subsequent months:
For the last five months of 2006 (the most recent months for which all revisions have been made), the number the BLS reported for jobs added in the current month has been, on average, only about 60% of what was ultimately reported for that month after revisions.
Third, BLS does an annual comprehensive revision to reported employment numbers that is released every February, as described more fully here. The latest comprehensive revision added over 900,000 jobs that were spread over the previous two years of revised results.
The problem, simply put is simply this:
- The initial report gets the news coverage.
- The routine revisions in the following two months, if they get coverage at all, almost never make the headlines and usually don't get into the first couple of paragraphs of any media report.
- The annual comprehensive revision is almost completely ignored.
The net effect is that even people who are paying reasonably close attention to news about the economy almost by definition have to be underestimating just how many jobs are being added. Because of the nearly invisible revisions and annual adjustments, almost no one would believe you if you told them that over 7.5 million jobs have been added in the past 3-1/2 years, but that is indeed the case:
You would hope that initial reports of increases in employment would be treated with at least a little bit of caution. But that is not the case, as this typical report from Reuters on the February jobs numbers from last week demonstrates:
The U.S. economy added a slightly weaker-than-expected 97,000 jobs in February, the smallest gain in more than two years, as increases in service-sector employment offset declines in construction and manufacturing, a government report showed on Friday.
Dan Clifton at the American Shareholders Association blog made a reasonable prediction last week that February's numbers will go up, as they did in almost every month noted above, and, according to Dan, as they have in almost every month for the past two years:
These numbers will be revised higher and bring the level over 100,000. Yet, the naysayers will still be saying job growth is slowing. But this wrong.
If the general form of the recent past noted above holds, the final jobs increase number for February, which we won't know until May, will go up by 40,000 or so and end up in the neighborhood of 140,000.
I therefore submit that reporting an initial figure as Reuters did with the "scare description" of "lowest in 2 years" will probably turn out to be wrong, and that reporting the initial number as if it's a done deal, or that it won't change much (which is what Reuters' report at least implied), should therefore be seen as more than a little irresponsible. Why? Well, ask yourself this: What are the chances that Reuters will correct the initial report's "lowest in two years" hysterics if the job numbers do go up quite a bit? (Similar headlines appeared with September's initial reports, and were followed by whopping September revisions that were mostly ignored -- Or have you seen any corrections I haven't seen?) Even in the remote chance that a correction occurs, how many million readers have been influenced (probably incorrectly) to believe that job growth was mediocre and on a downward trend in February?
Cross-posted at BizzyBlog.com.