Updated below.

The Associated Press unleashed something of a firestorm earlier today, when it published an article critiquing the recent Recovery Act jobs data. Performing a "limited review" of "some" of the recipient reports of Recovery Act contracts on Recovery.gov, the Associated Press concluded that the "government has overstated by thousands the number of jobs it has created or saved with federal contracts" under the Recovery Act. A bold claim certainly and one not supported by the facts.

The AP's secondary claim that the jobs data discrepancies "[raise] questions about the reliability of a key benchmark the administration uses to gauge the success of the stimulus" is, however, supported by a more thorough analysis of the jobs numbers in the initial round of recipient reporting. But the unreliability goes in the other direction: It is more likely that federal contractors under-counted the number of jobs created or saved by the Recovery Act.

According to the piece, "[t]he AP's review identified nearly 600 contracts claiming stimulus money for more than 2,700 jobs that appear to have similar duplicated counts..." and cites five instances where jobs counts were apparently erroneous. Of the five, only one contract overstated job creation by more than 300 (a 3,000 overcount by Teletech Government Solutions).

Is this troubling? Yes. Does it call into question the reliability of the job count numbers? Yes. Does it support the claim that the aggregate job count of the some 9,000 recipient reports was over-counted by thousands? Not really. Here's why:

• 335 reports state that no jobs were created or saved despite the recipient receiving Recovery Act funds and fully completing the funded project
• 169 reports indicate zero job creation despite being more than 50 percent complete, but not yet fully completed
• 77 reports show no jobs created or saved for completed contracts valued over $100,000 (and additional 60 for contracts that are more than 50 percent complete, but less than fully complete).

For example, a report from Chrysler on a $53 million contract to build cars for the federal government, which has been fully completed, shows that no jobs were created or saved. It's hard to imagine that $53 million in added revenue didn't save at least one job.

All of these reports indicate likely under-counting by recipients. And if one reads the narrative descriptions associated with these reports, one finds curious statements like "58 Employees retained employment" next to a job count of 0. Also found are statements like "Kept company from laying off employees" accompanied with a job created or saved count of 0.

The Recovery.gov website shows that over 30,000 jobs were reported as being created or saved by $16 billion in awarded contracts, and the AP story underscores the unreliability of the data. However, the 30,000 figure is more likely a low-end estimate than a radical overcount.

For more reactions to the AP article, check out Ed O'Keefe from the Washington Post, as well as Ed Deseve's response on behalf of the administration on the White House blog.

Image by Flickr user stephenjohnbryde used under a Creative Commons license.

UPDATE: John Irons at the Economic Policy Institute has performed an actual analysis of the recipient reporting data and concludes:

The first wave of recipient reports posted on Recovery.gov provides a wealth of data for observers to sift through on a range of Recovery Act funding. However, it is now apparent that the jobs data are of only limited value because of a range of issues from inconsistent methodologies for estimating jobs to implausible reports of jobs created or retained.

"The jobs data are of only limited value" is, unfortunately, less attention-getting than "Stimulus jobs overstated by 1,000s." The former, however, seeks to educate while the latter seeks to titillate.