|
They ran an analysis on Ohio and Florida and compared e-vote and non-e-vote precincts using a statistical method taking into account about 6 or 7 variables, including, for example income, race, population, change in turnout between 2000 and 2004, and voting history (how they voted in the 2000 and 1996 presidential races). The model was an extremely accurate predictor of voting behavior everywhere except the e-vote precincts in Florida. (There was no variation attributable to electronic voting in Ohio -- so if there was fraud in Ohio, it was statewide, evenly distributed and is historically consistent.)
The three worst counties in Florida were Broward, Palm Beach and Miami-Dade.
Interestingly, the biggest predictor of variation was the 2000 vote for Gore. The better Gore did in 2000 in a county, the bigger the variation between the results the model predicted and the results that were reported.
The scope of the research was merely to judge the accuracy of electronic voting.
The team commented on the Dixiecrat Florida counties which were heavily Democratic in terms of registration but which went for Bush. They say that these were NOT an animally. Those counties went heavily for Dole, and have traditionally voted for Republicans for president despite overwhelming Democratic registration. (The Berkeley team's analytical method does NOT take into account registration, which is what those claims of fraud are based upon, and which seems to correlate very poorly with how those counties vote for President.)
Another thing, and this is just my 2 cents: people talk about whether there was some intentionality in this. The professor addressed this by saying that there could be a software "glitch" -- there were stories about some machines counting down to 0 once they reached their capacity -- or there could be a mechanical explanation like a build up of "smudge" on a touch screen, which makes it less sensitive to a selection the more often that selection is made. The professor said, however, that was for someone else to judge -- they're only statisticians and not software or computer experts.
The discussion of intentionality might be illuminated by considering that punch card machines also have a defect. If you don't clean them, they become harder to punch through. Punch card machines might not have been designed with fraud in mind, but the people who knew they should have cleaned out those machines and didn't probably committed fraud in 2000.
|