Why doesn't it make sense to check voter registration records against other databases? Because unless you're very sure that both databases are extremely accurate and high-quality, this will cause more problems than it solves. One problem is clerical error. Voter registration is not uniform across the year; instead, there are huge spikes just before the registration deadlines for major elections. This situation--with low levels of work most of the year, and a short-term spike--is tailor-made for temp work. Secretaries of State will bring in additional staff as needed to handle the extra workload. Unfortunately, in the rush of new applications, there will be additional errors as new workers get used to the system and existing workers rush to keep up. Remember--at some point, all data is entered by a human. There are doubtless many people listed in the Ohio file as "Jhon". These people, meanwhile, are probably listed correctly in other databases. So if you're "Jhon Smith" in the voter file, "John Smith" in the Social Security database and "Johnny Smith" on the property tax rolls, you could be dropped from the voter rolls. Even when data is entered correctly, there might still be problems. For example, you might live on "Sixth Avenue". Or "6th Avenue". Or "6 Ave". Any one of these is a valid way to record streetname, but will not be seen as the same by a computer. There is a solution to this problem; address standardization software will put most addresses into a standardized format with a relatively low error rate (see this for an example). But there is an error rate--and in a state with 10 million voters, for instance, even an error rate of one-tenth of a percent would lead to 100,000 potentially disenfranchised voters. Part of the problem is the onerous and antiquated system we have for monitoring voter registration; Matt Yglesias has some good points in that regard here. But part of it is simply planning for failure in the voter registration system. This goes back to the discussion that Dirty D, Student Redux and I had the other day about data mining. No database will ever be perfect, and even if it is perfectly accurate there's still the possibility of human error in interpretation or implementation. Failure is inevitable. What this means, of course, is that we should make failure graceful. Provisional ballots are a good first step in this regard. And we should absolutely not simply drop people from the rolls or force them to show restrictive forms of identification if we know that the problem is likely to be database error. Kudos to the Court (for now; this was only a preliminary step to determine if the GOP had standing to sue, not a full suit). |
