From a comment on BlogForAmerica.com (let me know if there's a more primary source) comes an analysis of the California gubernatorial election, breaking down total number of votes vs. votes in counties where Diebold equipment was used. The figures look like this (note that not all candidates were counted):
| Candidate | Cast in Diebold counties | State total | % of the total votes cast. |
|---|---|---|---|
| All | 1,403,375 | 7,842,630 | 17.89% |
| Schwarzenegger | 581,145 | 3,552,787 | 16.36% |
| Bustamante | 447,008 | 2,379,740 | 18.78% |
| McClintock | 186,923 | 979,234 | 19.08% |
| Camejo | 39,199 | 207,270 | 18.9% |
| Huffington | 7,498 | 42,131 | 17.79% |
| Ueberoth | 3365 | 21378 | 15.74% |
| Flynt | 2384 | 15010 | 15.88% |
| Coleman | 1869 | 12443 | 15.02% |
| Simon | 1351 | 7648 | 17.66% |
| Palmieri | 2542 | 3717 | 68.3% |
| Louie | 598 | 3198 | 18.7% |
| Kunzman | 1957 | 2133 | 91.75% |
| Roscoe | 325 | 1941 | 16.7% |
| Sprague | 1026 | 1576 | 65.10% |
| Macaluso | 592 | 1504 | 39.36% |
| Price | 477 | 1011 | 47.18% |
| Quinn | 220 | 433 | 50.8% |
| Martorana | 165 | 420 | 39.28% |
| Gosse | 60 | 419 | 14.3% |
Now, the Diebold equipment is new, and there are bound to be a few glitches in a new product, especially when it's being used for a purpose it wasn't designed for - an election featuring 130-odd candidates when you expect at most 10. So that's why you have logs and a paper trail, so you can work out what went wrong. Right?
Incidentally, if there are faults in the Diebold system that can result in pathologically wrong results for small candidates, for edge cases, what's to say that there aren't also wrong results for some parts of the electorate of major candidates?
