Florida. November 2002. Optical scan vs. Touch screen.
Undervote rate is much higher on touch-screen voting machines.
Eighteen months ago, the Florida Division of Elections analyzed the November 2002 gubernatorial election. Their report showed that the undervote rate (races for which no vote was recorded) was 2 to 3 times higher on Florida's touch-screen machines than on its optical scanners. Since touch-screen machines warn voters of undervotes and optical scanner don't, these results are contrary to what logic would suggest.
The undervote rate varied from 0.31% to 0.58% on optical scan machines. The rate varied from 0.92% to 0.93% on touch screens.
The overvote rate (votes for too many candidates in a single race) on optical scan machines varied from 0.01% to 0.31%. Overvotes are impossible on touch-screen machines.
So, while touch-screens warn of undervotes and disallow overvotes, and precinct-based optical scanners only warn of overvotes and blank ballots, the total rate for both types of errors was higher on touch screens (0.93% max) than optical scanners (0.89% max).