Tuesday, August 16, 2011

The Weekly Standard: "PolitiFact’s Problem with Long Division"

Jeffrey H. Anderson may have a Ph.D., but it's not in mathematics. So when he's faced with the daunting task of taking one number and dividing by another number, he should just leave it to the rocket surgeons over at PolitiFact. This is especially important if old math doesn't produce PolitiFact's desired result.

Anderson sums up the numerical details while answering a PolitiFact analysis:
Last month, I wrote that President Obama’s own handpicked Council of Economic Advisors had released an estimate that the president’s economic “stimulus” had added or saved just one job for every $278,000 of taxpayer money spent. Obama’s economists said the “stimulus” had cost $666 billion to date and had added or saved 2.4 million jobs. $666 billion divided by 2.4 million is $278,000. Yet when Speaker John Boehner tweeted, “POTUS’ economists: ‘Stimulus’ Has Cost $278,000 per job,” PolitiFact Ohio rated [*] his tweet as “False.” PolitiFact Texas and PolitiFact Wisconsin have chimed in with identical scoring of similar statements.

So, what does PolitiFact have against long division?

Had Anderson been a reader of this blog, he would know that when numbers act in defiance of predetermined talking points, PolitiFact simply invents new standards to measure them against. And when it comes to inventing new standards, PolitiFact Ohio gets its cue straight from the top:

After Republicans began to circulate the blog item, White House spokesman Jay Carney said its conclusions were "based on partial information and simply false analysis." White House spokeswoman Liz Oxhorn issued a statement that noted the Recovery Act bolstered infrastructure, education, and industries "that are critical to America’s long-term success and an investment in the economic future of America’s working families."

The White House points out that Recovery Act dollars didn’t just fund salaries - as the blog item implies - it also funded numerous capital improvements and infrastructure projects.

Lumping all costs together and classifying it as salaries produces an inflated figure.

Of course, PolitiFact fails to offer evidence that Anderson did classify the entire stimulus spending as salaries.

Here's where PolitiFact Ohio tags out, and PolitiFact Texas brings some new moves to the ring:

The White House points out that Recovery Act dollars didn’t just fund salaries — as the blog item implies. Lumping all stimulus costs together and classifying the total as salaries produces an inflated figure.

Oops! PF Ohio already said that. Let's try again:
We checked the White House report, and of the $666 billion stimulus total, 43 percent was spent on tax cuts for individuals and businesses; 19 percent went to state governments, primarily for education and Medicaid; and 13 percent paid for government benefits to individuals such as unemployment and food stamps.

The remainder, about 24 percent, was spent on projects such as infrastructure improvement, health information technology and research on renewable energy.

How would Anderson respond to this arithmetical assault?

There are a number of problems with these claims.

First, I never said that the $278,000 per job was all spent on salaries or wages. I would never attribute anything close to that degree of efficiency to the federal government.

I'm really starting to like this Anderson guy.

He continues:

As I wrote in my response to the White House, “This much is clear: Based on an estimate by Obama’s own economists, for every $278,000 in taxpayer-funded “stimulus” money that the Obama administration has spent — whatever it may have spent it on — the “stimulus” has added or saved just one job.” That remains an undeniable fact.

Anderson's article takes up another issue with the stimulus, which is that not only is it incredibly expensive per job created, he also contends that the stimulus is actually causing jobs to be lost. PolitiFact, unsurprisingly, took issue with that claim as well. But Anderson effortlessly debunks PolitiFact's debunkery:

The entire response on this point from PolitiFact (both the Ohio and Texas versions) is to cite Moody’s chief economist Mark Zandi, who told the left-leaning website TPMDC that “the Weekly Standard misinterpreted that data.” That was good enough for PolitiFact. Never mind that Zandi is a Keynesian economist whose estimates of the stimulus’s likely effects were cited (see table 4) by Christina Romer, the first head of Obama’s Council of Economic Advisors, before the “stimulus” was even passed. In other words, Zandi said it would work, and now he says it worked.

In the end we're left with yet another (and multiple) examples of PolitiFact taking an objective, verifiable statement, constructing a straw man, and quickly demolishing their creation. Anderson never implied that the $278,000 figure represented salaries per job. He was making a point regarding the expense of the stimulus overall and putting it into a context that was easily digested by readers. It's impressive (if not disturbing) the lengths PolitiFact went to in order to distort and discard Anderson's valid premise.

Our goal at PolitiFact Bias is to consolidate and condense the best critiques of PolitiFact and provide a collection point of those criticisms. It is not our prerogative nor desire to reprint full articles.This brief review doesn't do justice to Anderson's excellent and thorough work. As always, we encourage you to go to the source and read the whole thing.

On a side note, I'd like to add something I found amusing, repeated verbatim in both the PF Ohio and Texas editions:
Furthermore, the publication created its statistic with the report's low-end jobs estimate. Had it gone with the 3.6 million job figure at the top end of the range, it would have come up with a smaller $185,000 per job figure.
Are we to assume $185,000 per job created would bump the stimulus into the "successful" category?

Bryan adds:

I find it amusing that PolitiFact accepts Mark Zandi's opinions without comment yet spends much of another recent fact check attacking Florida governor Rick Scott's source because of its supposed partiality.

PolitiFact's work generally leaves the impression that it favors liberal sources in terms of both numbers and reputation.  One might say PolitiFact represents Groseclose liberal media syndrome on steroids.

Pending a rigorous evaluation, of course.

*In Anderson's article the original "PolitiFact Ohio rated" hyperlink linked back to Anderson's own piece. I changed it to link to the PF Ohio rating of Boehner's tweet that was described.-Jeff 

No comments:

Post a Comment

Thanks to commenters who refuse to honor various requests from the blog administrators, all comments are now moderated. Pseudonymous commenters who do not choose distinctive pseudonyms will not be published, period. No "Anonymous." No "Unknown." Etc.