October 29, 2010

The Undead Undead Voter Claim

[Note: Justin here, with thanks for another generous offer to guest post.]

Halloween isn't just a time for scary political ads. It's also the favorite season for reports that the dead are walking--straight to the voting booth.

Back in 2006, I wrote about the flaws in many of these undead-voter claims, which are often based on careless attempts to match voter rolls to Social Security Administration lists. In 2007, I wrote about them again. In 2008, I wrote about them again.

It turns out that unwarranted claims of fraud by dead voters are ghoulishly difficult to kill. This year, "Saw" isn't the only retread that's been resurrected for late October.

Amidst the general buzz about voter fraud, the dead-voter stories are back. Yesterday, in Kansas, a candidate for Secretary of State held a press conference bemoaning (again) the state of the voter rolls and calling (again) for photo ID rules at the polls. He said, "When you look at the voter rolls, you will find many people on the voter rolls who, if they were still alive today, they would be 104, 106, 110 years old . . . . Let me just give you one example."

The claim then got specific, which is unusual ... in part because specificity creates the opportunity for verification. The candidate cited Alfred K. Brewer as having died ... and then voted. To be fair, he then hedged his claim, stating that though he had not wholly confirmed the fraud, "it certainly seems like a very real possibility."

A little research quickly undermined the certainty. The voting Alfred K. Brewer is very much alive (when called to inquire about his journey back from the grave, he was raking leaves). A death entry for his father, who passed away some time ago -- and was also named Alfred K. Brewer -- had triggered the false alert.

Such "false positives" -- improper assumptions that two people with the same listed information are actually the same person -- are far more statistically common than most people think, even when the matching exercise isn't sloppy. But conclusions from a "match" aren't the only issue. Sometimes the "dead voter" isn't dead. Sometimes she didn't vote. Follow-through on wild claims, it appears, is an even more effective way to whittle down the zombie voter hordes than a shotgun.

(Indeed, the research underlying the Kansas press conference claims was substantially more thorough in its follow-through -- and the number of undead suspects dropped accordingly. Unsupported claims are bigger, and much more exciting.)

To this end, a few years ago, I put together a short "investigator's guide" for claims of voter fraud, to help people -- before hosting the press conference -- sort the unsupported claims from those with merit. The guide is based on my experience that behind most reports of dead voters, scary as they may seem, there's a whole lot more truth left to unearth.

Posted by Justin Levitt at October 29, 2010 06:05 PM