Problems come when using databases to disqualify voters

Guest Opinion, Iowa City Press Citizen, Aug. 23, 2012

Part of the Voting and Elections web pages, http://www.cs.uiowa.edu/~jones/voting/
by Douglas W. Jones
THE UNIVERSITY OF IOWA Department of Computer Science

During their news conference Friday, Iowa’s Republican secretary of state, Matt Schultz, and Democratic attorney general, Tom Miller, presented evidence suggesting there are non-citizens who have registered to vote illegally and that some of these illegal registrants have voted.

Clearly, further investigation is called for, and if indeed these people have voted, they should be prosecuted.

I am worried, however, about the effort to run a database matching effort to ferret out and remove non-citizens from the voting rolls.

The central problem here is that we have no requirement of registering to vote under the same name as we use for other purposes.

For a driver’s license, you present a birth certificate, so your name on the driver’s license will match your birth certificate.

To register to vote, you can use your employer ID card and a phone bill.

As it turns out, my voter registration is in the same name as my driver’s license. That’s because I used my license to register about 32 years ago.

On the other hand, my employer’s ID card lists my name differently (just a middle initial). I could have registered to vote with that card, had I wanted to. There is no legal requirement that I use the same name everywhere, and in fact, I use a variety of names and nicknames:

I’m not trying to confuse people. It’s just that, at various times, I’ve used different and obvious variations on my full name.

That’s why cross checking voter lists with driver’s license databases is very problematic. If you demand exact matches, you’ll miss many people; and if you accept partial matches, you’ll start to confuse people.

The exact rules used to determine whether you’ll be more likely to err by disenfranchising people who were legally entitled to vote, or to err by allowing people to vote who shouldn’t.

With more than 3 million residents and fewer than 2,000 alleged improper voters, it is extremely easy to disenfranchise people accidentally.

You’re trying to build a filter fine enough to catch the 1 in 2,000 voters who might have cheated, but if your filter makes a mistake in just 1 percent of the cases it examines, you’d disenfranchise tens of thousands of voters.This risk is high enough that such a filtering effort is extremely dangerous, particularly right before an election.

There are two people with my name at the University of Iowa. There have been as many as five: including one other with the same middle initial.

In 2010, someone with my name (same middle initial) pleaded guilty to beating a man to death in Mississippi.

Google finds a number of people who share my full name — including a financial adviser in California and a game warden in North Carolina.

If someone tried to match a database of convicted felons in the U.S. with Iowa voters, I could easily be disenfranchised.

We don’t have a national database in this country that uniquely identifies each citizen and permanent resident. There are good solid reasons we don’t have one — expense being one, and threats to civil liberties being another.

With a national ID system, we could do reliable cross-checks. Without such a system, any cross-checking of databases must be undertaken with extraordinary care.

Vigorous prosecution of those who’ve cheated, on the other hand, seems to be entirely appropriate — and such prosecution would have a very strong deterrent value.

Douglas W. Jones is an associate professor of computer science at the University of Iowa and the co-author of the book, Broken Ballots: Will Your Vote Count?