Sunday, September 15, 2013

PolitiMath at PolitiFact Florida

For some time we've been interested in how PolitiFact handles the magnitude of error for numbers claims.  We refer to ratings that apparently rely on some degree of numerical error as "PolitiMath," and it serves as one tool helping us assess PolitiFact's inconsistency.  It can also give us another sign that PolitiFact allows ideological bias to affect its rulings.

PolitiFact Florida provides us an excellent test case this year with a pair of fact checks that appear to betray inconsistency.

Charlie Crist and the Ineligible Voter List 


On Sept 12, PolitiFact did a fact check on former Republican and former Governor Charlie Crist of Florida.  Crist claimed that out of over 100,000 people the state of Florida thought were ineligible to vote, less than 10 were eventually found ineligible.

PolitiFact ignored Crist's deception in framing the state's list, which was a list of people it suspected might be ineligible to vote.  The lists, as PolitiFact noted, were pared down to 2,600 names and then turned over to local elections supervisors for investigation, precisely because those on the list were merely suspected of ineligibility.  And from that list of 2,600 names PolitiFact reported that "about 85 noncitizens" were removed from Florida's voter rolls.

Crist's "under 10" figure, charitably interpreted as nine, represents a 87.1 percent error compared to the PolitiFact estimate.  The calculation takes the difference between the correct number and the incorrect number and divides the difference by the correct number ((85-9)/85).  If that's not clear, go here or here.

PolitiFact ruled Crist's claim "Mostly False."

Will Weatherford and the College-ready Florida High School Graduate

On Sept. 11, PolitiFact did a fact check on Florida House Speaker Will Weatherford.  Weatherford, during a speech, said only one out of three Florida high school graduates qualify as "college-ready."  Weatherford repeated the claim twice during the speech.

PolitiFact said it heard from Weatherford spokesman Ryan Duffy, who said Weatherford misspoke.  Duffy also tweeted:
Weatherford's "one in three" number, read strictly, represents a 48.6 percent error compared to the number PolitiFact presented as accurate ((65.7-33.3)/65.7).

PolitiFact ruled Weatherford's claim "False."

Explaining the discrepancy?


We can think of one mitigating factor that could help justify giving Crist a better "Truth-O-Meter" rating than Weatherford:  Out of 180,000 names (or 2,600), the difference between "less than 10" and 85 doesn't make a great deal of difference.

On the other hand, where Crist received some credit for his underlying point, Weatherford received no credit at all for his underlying point, that the number of Florida high school graduates unprepared for college is way too high.  Add to that a very plausible case, despite Weatherford repeating himself, that he simply misspoke.

We don't see a plausible justification for giving Weatherford a lower rating than Crist.


Is it just me or do the Miami Herald's Marc Caputo and PolitiFact's Katie Sanders seem a tad snarky in their Twitter communications with Weatherford's spokesperson?

No comments:

Post a Comment