In a public speech yesterday, U.S. Attorney General Eric Holder called for universal, automatic voter registration, and stated that current technology can accomplish that, despite the fact that the current system is complex and error-prone. As Reuters reported on Holder's remarks:

By coordinating existing databases, the government could register "every eligible voter in America" and ensure that registration did not lapse during a move.

That's easy to say, but it requires some careful thought to make it easy to do. After some discussion with election officials recently, I've concluded that it is in fact easy to do in tech terms, but not in a way that you might think. To explain, let me first say that one thing that's not going to happen anytime soon is a "federal government takeover" of voter registration. VR will remain a state responsibility for the medium term, I predict.

Second, something that might happen, but would be a bad idea, is the combination of inter-state record matching and automatic registration. Why? Because we've already seen that in some states, recent practice includes automatic de-registration: if a computer's matching algorithm says that you moved from one state to another, you get un-registered in the first state. (Though not registered in the second!) Of course that's a problem if the match is incorrect -- and we've seen plenty examples of dodgey databases yielding false positive matches -- but it also can be a problem even if it is correct.

Ironically, the most recent instance of that story I've heard personally was from a Yale political science professor who specializes in election observation in other countries, and is keenly aware of voter registration issues as a bar to voting. While retaining her residence in CT, the prof did something that looked to some computer like taking up residence at another address -- result: CT's VR system de-registered her. Not the right way of doing universal, automatic, permanent.

One state election official explained the higher-level issue to me recently with two main points.

1. The current system places responsibility on the citizen to apprise the appropriate government. So when it appears that there has been a change of address, the state VR operators should reach to the voter in question to get the real story from them. That includes making it easier for voters to quickly find out their VR status and get help on what they can do next. (Which is what we're doing with online VR technology this year.)

2. When deciding what to do about a reported VR change, the responsibility is the election official's not some computer's. Technology can help suggest to an election official that a voter's record may be out of date, but that should not mean that the voter record should invalidated, either automatically or with a pro-forma confirmation by a person who has no more information than the computer did. What should the election official do instead? See point #1 above!

In other words, a simple interpretation of Holder's words about database co-ordination can lead to data-mining and matching that is error prone not just because the databases have imperfect information, but also because some of the most important information -- voter's intent -- is not in the database, for example "I did a postal address forwarding from my CT home to a DC address not because I moved but because I'm visiting for several weeks and don't want to miss my mail."

So that got me thinking about functional requirements - surprise, techie thinks about requirements not policies! - and we came up with a way to use those two principles to deliver many of the benefits of universal automatic permanent registration, without actually changing election laws and overhauling existing voter database systems. What's required is an inter-government information sharing system:

  • that can notify state VR system operators about events that are possibly relevant to VR, without having to be authoritative about the event or even the person involved;
  • that can enable state VR system operators to take further steps to determine whether there's been an change in voter eligibility;
  • is sufficiently flexible for a wide variety and number of government organizations to participate with ease.

In addition, not required, but darned useful to residents of the 21st century, this system would be complemented by online assistance to members of the public to help them quickly and accurately respond inquiries from election officials.

The latter we are, as I have said, already working on, and well into it. But that inter-government information sharing system, what is that? It would clearly have to be not complicated, not expensive, and not requiring changes in election law or policy. Is that possible?

I think so. Stay tuned, we may be on to something.