Kevin Drum has a good rundown of the problems with a UC Berkeley sociologist's analysis of voting in Florida (PDF). It's the same issue as Steven Freedman's paper and the supposed optical scan anomalies -- a tendentious result produced by people who don't have a good background in the field. Kieran Healy shows in a working paper (PDF) that if you control for the two anomalous counties Drum flags, the supposed e-voting effect goes away. And Alex Strashny at UC-Irvine has an excellent methodological and substantive critique up as well, which shows (among other things) that if you account for changes in party registration, the e-voting effect also goes away.
The larger problem here is that lefty academics with access to voting data can come up with a million supposedly unexplainable anomalies and put them up on the Internet, fueling partisan suspicions that Bush stole the election. It's the same thing as bogus accusations of voter fraud from Republicans - part of the never-ending assault on the legitimacy of electoral results. This pattern is not good for our democracy...
[Disclaimer: It's possible that someone could use statistical methods to discover fraud, and we want to pay attention to them if and when they do. But the burden of proof cannot be entirely borne by the system - elections are messy and complex, and it's easy to raise questions that cannot be disproven. Going forward, people making claims about "anomalies" should have to meet a much higher standard of proof than these papers have.]
Comments