Kudos to Floyd Norris over at the New York Times for characterizing total job losses to date as 8 million jobs, not “just” 7.2 million. As I discussed on Friday, the Bureau of Labor Statistics estimates that the number of jobs in March 2009 was 824,000 lower than it previously thought. But BLS won’t include this adjustment in its official data until early February.
The official, as-yet-unadjusted data indicate that 7.2 million jobs have been lost since the recession started in December 2007. The future revision to March figures, however, implies that a better estimate would be 8 million.
We can now expect several months in which commentators use different figures for total job losses. Those steeped in the details, like Norris, will use the 8 million figure. Those less-attuned to the details, like the authors of the NYT’s lead editorial (just four pages after Norris’s article), will use the 7.2 million figure.
Norris also addresses the obvious question: Why did BLS miss the March level of jobs by such a large amount? The answer is that BLS has to estimate jobs gained and lost at certain employers, and their model is not doing as well as we (or it) would hope:
The official job numbers are based on a monthly survey of employers, augmented by something called the “birth-death model,” which factors in jobs assumed to have been created by employers who are too new to have been included in the survey, and subtracts jobs from employers assumed to have failed and therefore not responded to the latest survey.
Victoria Battista, an economist at the Bureau of Labor Statistics, said the bureau was looking at whether that model needed to be changed, as well as at other possible issues, such as changing response rates to the questionnaire sent out to employers each month.
The newest revision is called a “benchmark revision.” Such revisions are disclosed each October, and led to reductions in job totals in both 2007 and 2008. But the changes those years were tiny when compared with the changes this year.
For the 12 months through last March, the birth-death model added 717,000 jobs to what the bureau would have reported had it relied solely on its survey.
While the government uses the survey of employers to estimate the number of jobs, the benchmark revisions are based on reports from states on the number of employees for whom unemployment insurance premiums are paid. Those numbers take longer to be available, but are considered to be more reliable.