Some data analysis
Nov. 6th, 2004 07:11 pmData from here and here and eyeballing voting technologies used from here.
The source for that graph that freaked me out earlier.
Another nerd doing the same kind of stuff. His superior use of gnuplot and actual statistical techniques (confidence intervals and the like) is nice. He concludes that there is enough evidence to justify a thorough investigation, but not enough to accuse.
The error rate is attained by subtracting each candidates exit poll numbers from their actual percentage of votes tabulated.
State | Kerry Error | Bush Error | Methods |
---|---|---|---|
Alaska | -12.0 | +8.9 | Optical |
North Carolina | -4.5 | +4.1 | DRE and mixed |
Florida | -3.9 | +3.1 | Optical and DRE |
Minnesota | -2.9 | +3.6 | Mixed, optical, and DRE |
Ohio | -2.5 | +2.0 | Punch card, some DRE |
Pennsylvania | -2.2 | +2.6 | Some of everything - lever, optical, DRE |
Wisconsin | -2.2 | +2.4 | Optical, some mixed |
Colorado | -1.7 | +1.5 | Optical, some paper |
New Mexico | -1.1 | +1.0 | DRE mostly |
Louisiana | -0.8 | +0.8 | Lever and some DRE |
Arizona | -0.5 | -0.1 | Optical |
Missouri | +0.1 | -0.6 | Mixed, punch card, and paper |
Iowa | +0.2 | +1.1 | Optical, some DRE |
Michigan | +0.2 | +0.8 | Mixed, some DRE |
Systemic errors in Kerry's favor in the polling data? Or systemic errors for Bush in the vote count? Something is messed up here. Also, I found out that the historic tendency has been for Republicans to vote early, while Democrats vote late. So the idea that this data is biased towards Dems because Repubs haven't voted yet would also go against historical trends. I would like to find more confirmation of that idea, however.
The source for that graph that freaked me out earlier.
Another nerd doing the same kind of stuff. His superior use of gnuplot and actual statistical techniques (confidence intervals and the like) is nice. He concludes that there is enough evidence to justify a thorough investigation, but not enough to accuse.