As things shake out in the aftermath of the 2004 election, it seems less and less likely that the huge fraud theory will pan out. I tend to go along with David Corn's piece. Read the whole thing for a pretty good overview of some of the major scandals in the counting and voting irregularities. Here are a few bits:

In pieces for The Nation magazine, I've noted that there is good cause to worry about the integrity of a voting system that is overseen by partisan players and that relies in part upon paperless electronic voting machines that are manufactured by companies that are led by pro-GOP executives and that refuse to reveal the computer codes they use. But I've also cautioned against declaring that the potential for abuse means the system was abused to flip the results. Exit polls that differ from reported vote counts are not necessarily proof of foul play, and statistical analyses that seem to raise questions need thorough vetting before they are waved about as signs of chicanery.

Take one of the early arguments for the "stolen election." Shortly after E Day, a former high school math teacher named Kathy Dopp sent out a chart that showed George W. Bush faring unusually well in Florida counties that used optical scan voting machines. A-ha, some folks exclaimed, this chart demonstrated the vote had been fixed. A team of political scientists led by Walter Mebane, a professor of government at Cornell, then examined the votes in these counties and found they were consistent with a years-long trend of registered Democrats in rural counties voting for Republican presidential candidates. Their findings were disputed by some "stolen election" advocates. But the Caltech/MIT Voter Technology Project released a study that reached the same conclusion as the Mebane paper....

For instance, in Union County, more than 75 percent of registered voters were listed as Democrat. The official vote count was 3,396 for Bush and 1,251 for Kerry. The Herald found 3,393 votes for Bush and 1,272 for Kerry – practically no difference. The results were the same in the other two counties. This hands-on exercise demonstrated that statistical analyses that rely on predicted outcomes based on voter registration figures can only prove so much. There's no substitute for inspecting actual evidence, such as ballots.