A while back I was pretty skeptical regarding the conspiracy theory that the 2004 U.S. Presidential election was rigged. The balance of evidence in favor of this conclusion was statistical, and, I felt, this was not a sufficient basis considering how difficult engineering a cover-up of that magnitude would be.

Then someone sent me to this article, a response to the Edison-Mitofsky Jan. 19th report on what the latter did wrong. Edison Media Research and Mitofsky International were the exit pollers for the U.S. elections. Their exit polls showed Kerry winning by 3.0%, although according to the official vote count, Bush won by 2.5%. This is a huge statistical deviation, one that cannot be attributed to mere chance.

Edison and Mitofsky initially (shortly after the election) advanced the hypothesis that their data was biased because Kerry voters might have been more eager to respond to polls than Bush voters. That was a plausible enough explanation for me, at least enough that I could write the issue off in my head.

But there's plenty of raw data, of course, and it can be tested. The article above sketches the details, but a bunch of statisticians did a much more thorough demolition of that hypothesis in a report released yesterday. The short, short version: reporting rates were actually slightly higher in districts that tilted towards Bush. Meanwhile, Edison-Mitofsky have no support for their hypothesis whatsoever.

So, it seems that the balance of evidence now leans in the direction of electoral error, or worse, fraud. Jesus Christ. Now what?