For an excellent overview of the issue of voting machines and error-rates, see the Caltech-MIT Voting Technology Project.

In terms of a straightforward summary of some of the informaiton in that study, here’s an excerpt of an article on Ballot Reform that I wrote for David Schultz, Ed. The Encyclopedia of Public Administration and Public Policy, New York: Facts on File, 2024. (forthcoming).

Currently there are five different types of voting methods in the United States: paper ballots, lever machines, punch cards, optical scan and electronic voting. Each of these reflects differing levels of technology and represents differing actual methods of casting and counting votes. The paper ballot system is, as the name implies, simply a piece of paper that a voter marks preferences upon. These ballots are counted by hand. Level machines directly record votes that are entered by flipping levers on the machines. Such machines have been in use since the late nineteenth century. Punch card ballots are read by computers, and require the voter to punch out a small hole in the card by knocking out a pre-perforated chad. Optical scan ballots are also read by computer, but voters instead use a pen to mark the ballot to indicate their preferences. Electronic voting entails the use of computers which record votes likely entered by keyboard or touchscreen.

The main issue is the question of “residual votes” which are defined by the Caltech-MIT Voting Technology Project as the combination of uncounted ballots, unmarked ballots and overvoted ballots. Residual vote rates for the 1988-2000 period for presidential elections are as follows: paper (1.8%), level machine (1.5%), punch card (2.5%), optical scan (1.5%), electronic (2.3%). The rates are quite higher for Governor and Senator during this same period: paper (3.3%), level machine (7.6%), punch card (4.7%), optical scan (3.5%), electronic (5.9%). Such numbers demonstrate that differing technologies do indeed have important effects on the balloting process.

The numbers on the electronic machines are partially based on older machines than the current touch-screen variety. There is no large pool of data to judge the touch-screen systems at this point, and the Caltech-MIT study deems that technoogy as “unproven.”