Thine own petard lieth in wait for such as thee, Andrew Sullivan

It's simply amazing. 

Sullivan shows us how to hoist one's self with his own petard like a father patiently demonstrating to his young son how to tie a tie.

No, really.  Sullivan shares with his readers a chart derived from PolitiFact's "Truth-O-Meter" readings.

Sullivan comments:

Of course, there's a big bold disclaimer: Politifact picks and chooses what topics it covers; it itself is not unblemished in its impartiality; none of this pretends to be a comprehensive, independent analysis of large swirling, now uncoordinated campaigns. Nonetheless, you can separate out party leaders on both sides, as selected by Politifact, and gauge their truthfulness, as measured by Politifact.

Exactly!  The chart tells you nothing dependably about the candidates, but it does potentially tell you something about PolitiFact.

Compare:



That's a nicely tied tie! 

Unfortunately, Sullivan doesn't seem aware that he has undercut the point of his post.

Again, what you see is a GOP advantage in truthiness in general but a huge discrepancy when it comes to total, massive, pants-on-fire whoppers.

Sullivan's claims especially tickled me since I'm completing the first of a series of studies examining PolitiFact for signs of bias.  The first such study looks specifically at "Pants on Fire" ratings compared to other ratings.  The difference between "False" and "Pants on Fire" is important because PolitiFact defines the difference between the two (so far as I can tell) on entirely subjective terms.

FALSE – The statement is not accurate.

PANTS ON FIRE – The statement is not accurate and makes a ridiculous claim.

Until PolitiFact reveals its objective means for determining that a statement is worthy of ridicule, it's reasonable to take the definitions literally.  Far from showing that Republicans lie outrageously, Sullivan's stats simply show PolitiFact's tendency to think Republican untruths worthy of ridicule.

Of note, PolitiFact Wisconsin is often criticized as leaning to the right.  This measure bears that suspicion out to some extent, though I suspect the reasons behind the numbers differ significantly.