Any trend in the collection, analysis, interpretation, publication or review of data that can lead to conclusions that are systematically different from the truth.
--J. A. Last, "A dictionary of epidemiology"
The fact checkers:
Aaron Sharockman: writer, researcher
John Bartosek: editor
Sharockman and Bartosek have collaborated on a sensational example of selection bias. Let's observe their method as they quote a series of claims by Florida state representative Eric Eisnaugle:
"We have seen allegations of fraud, (Florida Department of Law Enforcement) investigations. (We have seen) falsifying of hundreds of registrations, including the registration of an actor who was already deceased at the time. In another case, Mickey Mouse was registered to vote.The claims are:
"In yet another case, hundreds or thousands ... of students were registered to vote without their knowledge after they simply signed a petition, having no idea that their information was then going to be turned around and used to register their names on the voter rolls here in Florida."
- a deceased actor had a voter registration application falsified
- a cartoon rodent was registered to vote
- signers of a petition were registered to vote without their knowledge
Eisnaugle made three different fraud claims, and we plan to touch on each in this fact check to see if they're true. But for the sake of the Truth-O-Meter, we're going to home in on the line that stuck out because of the absurdity of it, if true: "Mickey Mouse was registered to vote."PolitiFact chose the most potentially absurd of the three claims to position the needle of the "Truth-O-Meter."
What about the claims that don't count on the "Truth-O-Meter"?
The group workers turned in 1,400 cards, of which 888 were found to be fraudulent. Included among the fraudulent voter applications -- deceased actor Paul Newman.That one sounds true.
The workers would have the students sign a petition to legalize marijuana or a petition urging stiffer sentences for child abusers. Then, workers would have students unknowingly fill out a second form registering them to vote as Republicans.The claim about a third-party group registering students surreptitiously also sounds true. Eisnaugle's not bad on the Meatloaf-O-Meter, even if the Mickey Mouse claim doesn't pan out.
And what of the Mickey Mouse claim? Was Walt Disney's signature creation registered to vote in Florida?
It comes down to the meaning of "registered to vote."
When I read the PolitiFact story, I took "registered to vote" to mean that Mickey was eligible to walk up on election day, had he desired, and cast his vote. Was my view prejudiced by the accompanying story? Could be. I can't say for sure. What I do know is that when I listed to Eisnaugle speak, I took him to mean that Mickey had been registered in the same sense that Paul Newman had been registered--via the creation of a fraudulent voter registration application.
I encourage readers to conduct their own tests. Download the video by clicking here and watch Eisnaugle speak at about the 2:28 mark (props to PolitiFact for providing the link even if the alternate interpretation was never considered in the story).
The context supports the interpretation scorned by PolitiFact. All of it surrounding the Mickey Mouse example concerns the actions of third-party registration groups.
PolitiFact did not address the comments of Lyndsey Cruley, spokeswoman for Florida's House Republicans:
"In sum, although Rep. Eisnaugle’s remarks could be interpreted, if taken out of context, in two different ways, he clearly intended to demonstrate that some third-party registration organizations have committed fraudulent actions here in Florida."Cruley's explanation fits the facts. Why wouldn't PolitiFact acknowledge the point?
And if PolitiFact is narrowly focused on the truth of the potentially absurd registration of Mickey Mouse, then why does PolitiFact editorialize regarding the bill Eisnaugle was extolling?
What's worth noting is that, in all three stories Eisnaugle cited, the fraud was exposed. The only real impact was that students in 2004 were registered as Republicans when they wanted to be registered as something else. It's also worth noting that it's unclear how HB 1355 -- had it been part of election law -- would have altered any of the above three scenarios.I suppose Eisnaugle should have used examples of unexposed fraud. Though how he would know about unexposed fraud ought to cause its own set of suspicions. That, in fact, is one of the reasons why bills such as the one in question may prove helpful. Our system makes it pretty easy to hide fraud. For example, felons may have illegally voted in Florida's elections. How does one detect it other than by catching them in the act?
In the end, PolitiFact's editorializing misses the mark. Eisnaugle did not claim that the examples he cited would have been altered with the bill in effect. He cited the claim to silence the complaint that no problem with election fraud exists in Florida.
Aaron Sharockman: F
John Bartosek: F
Flunk factors: Ignoring the obvious charitable interpretation of Eisnaugle's statement, editorializing about Eisnaugle's point while disclaiming interest in Eisnaugle's point (not to mention missing the latter). The selection bias problem is not specific to Sharockman and Bartosek. That is a systemic problem.
The insidious effects of the type of selection bias found in this PolitiFact story have their full effect when PolitiFact encourages readers to look at the collected statements about a person in order to develop an impression about the person. Using Eisnaugle as an example, the reader of Eisnaugle's PolitiFact "report card" (they actually call it that!) would see a lone "False" rating. Yet if PolitiFact had chosen either of the other two statements to rate, Eisnaugle would have a lone "True" rating. Or if PolitiFact had rated all three, Eisnaugle could expect to have two "True" ratings and one "False" rating.
What good is a grade point average created under these types of conditions?