Monday, May 2, 2011

For Ohio voter rolls, a birthday surprise

By Associate Professor Justin Levitt

Recently, the Ohio press was abuzz with news of a study from Ohio Secretary of State Jon Husted on one aspect of the state's voter registration rolls. He'd attempted to match the registration lists with records of individuals listed by Ohio's department of health as deceased, and apparently found 18,460 matches.

But that's not what the public heard.

In one respect, press reports on the study showed remarkable restraint. It's quite common to see such reports heralded by shrill cries of voter fraud, with unfounded allegations of wrongdoing as overblown as the exclamation points at the end of an all-caps email. It's also unfortunately frequent to use such reports as fodder for heightened pollsite "security" measures that don't actually address the purported problems: the election-law equivalent of an airport frisk designed to stop only when it meets "resistance." Neither showed up in the Ohio press on the Secretary's report this week. And that silence is particularly notable given the highly controversial photo-ID bill working its way through the legislature at this very moment (which Husted, to his credit, recognizes as unnecessary). Huge credit to the press corps for reporting the news, and not the red meat.

Well, almost.

Unfortunately, not one Ohio headline I've seen about the Secretary's report actually got the news right.

The Cleveland Plain Dealer: "Thousands of dead on Ohio voter rolls, report shows."

No, it didn't.

The Toledo Blade: "Ohio elections chief says health records show thousands of 'active' voters are dead."

No, he didn't.

The Cincinnati Enquirer: "Hamilton County leads state -- in dead voters."

No, it doesn't. (Or at least, the study that is the basis for the article says no such thing.)

These are just a few examples from a few of Ohio's bigger papers. But around the state, the story was the same: the Secretary of State had found 18,460 dead people on the registration list. Except that's not what the Secretary found.

To his credit, Secretary Husted was far more measured in his own report's release. He explained that he'd matched the names and birthdates of more than 8 million voters on the registration rolls to the records of individuals listed by the health department as having passed away. He found 18,460 matches. And he asked county officials to investigate further before removing the voters in question from the list, which is exactly the right response.

Let's assume that the health department's records are accurate. Why not jump immediately to the "Delete" key? In a word: math.

Matching names and birthdays runs headlong into a problem straight from Statistics 101. Next time you're in a large group of friends, write everyone's birthday down on a piece of paper. With just 23 people at your party, chances are better than 50 percent that two of them share the same birthday (month and day). With 50 people, it's almost a lock.

Adding the year to the mix, you need more people in the room -- but still far fewer than most expect. With about 180 people at the party, chances are better than 50 percent that two of them were born on exactly the same date. 460 people makes it 99 percent certain.

What this means is that in any big pool, it's pretty likely that people will share the same date of birth. All you need are 460 Anne Millers, or 460 Thomas Smiths, or 460 Maria Johnsons. And there are more than enough Millers, Smiths, and Johnsons in the Buckeye State. You'd make money betting that there were plenty of different people with the same names and dates of birth at any given Ohio State game. Now expand the pool to the 8 million registered voters in the whole state.

There is a real probability of identifying the wrong person, many times over, with matches like these. Matching on names and birthdates isn't, on its own, enough information to act -- which is why a federal bill has tried to require the same caution seen in Secretary Husted's approach. Election officials who instead rush to judgment on the basis of a match alone may find that they've delivered their constituents a very unpleasant birthday surprise.

No comments: