Testing Preference Voting on a Computer Program
Rob Richie and Howard Fain
For over fifty years, the City of
Cambridge has counted ballots in its municipal elections by hand. The preference voting
form of proportional representation system used for city council and school board
elections requires counting and transferring voters' ballots in a manner for which there
was no known machine technology count when the system was adopted in Cambridge in 1941.
A study by The Center for Voting and Democracy indicates that Cambridge and other cities that adopt preference voting will never again be forced to count ballots by hand. A city may choose to conduct the count by hand, but with computer technology, the study conclusively demonstrates that it is entirely feasible and affordable to conduct a count by computer.
The Center's program -- called PRMaster 1 -- produced final results using Cambridge's counting rules with 22,962 ballots from the 1991 city council election in two minutes on a personal home computer. The accuracy of the results are well within the margin of error created by the manual data entry process used in the study.
Cambridge will need to weigh the value of maintaining a community tradition of hand-counting that is important to many people versus the delay and citizen misunderstanding of the process that comes with the hand-count. The following were judged by the Center as important considerations in the city's decision to convert to a computer count:
Cost: The current hand-count involves costs to the taxpayer. Computerizing the count also would cost taxpayers, but after one-time acquisition of software and hardware, computerization undoubtably would save money.
New voting machines: Many localities around the country are considering or have already implemented upgrades of their voting machines to such systems as electronic voting booths and scanning systems. These mechanisms for voting would make computerization of the count particularly sensible because they would permit direct recording of data to be used by a computer program. Without further research, it is unclear whether other voting machines would allow the City to avoid having to input ballot data manually in order to use a computer program.
Manual data entry: To save money, the Center did not seek to validate the data entry with a system of "reverse checking." There are errors in the manual data entry used in this trial, although we calculate total errors as likely less than 1%. Even though the data entry process could be improved, it still might cause some voter distrust that could be avoided by voting machines that would record data directly and thus eliminate the need to use manual entry.
Recounts and filling of vacancies: Whatever savings that would come with a computerized count would be magnified by any situation that requires a re-count of ballots. For a contested result calling for a re-count or to fill a vacancy, the computer program could conduct a re-count within minutes, with no additional costs.
More precise ballot transfers: An important finding of this report is that Cambridge voters should be confident that the current "random draw" rules of transferring surplus ballots from elected candidates are unlikely to affect the results. Nevertheless, given civic traditions emphasizing the value of every vote, an argument can be made in favor of a more precise method of ballot transfer that would ensure that every voter's ballot is treated equally and that would remove any uncertainty about "randomness." Computer programs provide great flexibility for such refinements of the method of counting.
Invalid ballots: New voting machines that directly record data from the voter could lessen the number of invalid ballots cast. Although the current percentage of invalid ballots is well under 2%, new voting machines could reduce this percentage to near zero. Even without new machines, a computer program could reduce invalid ballots by slightly modifying the rules to allow "x" voting, as described in the report -- a computer program can deal with the counting complexities that come with "x" votes (which are treated as tie votes) much more easily than is possible in a hand-count.
Voter participation: It is not clear whether voter participation in Cambridge elections is affected by misunderstanding of the preference voting system -- particularly because turnout is very high compared to most localities -- but it is possible that any such misunderstanding is magnified by the length of the count. Computerizing the count would produce results in a timely manner more in keeping with other elections in which Cambridge residents vote.
Other localities' consideration of the preference voting system: As other localities often are wary of the prospect of organizing and running a hand-count, the successful example of a computerized count in Cambridge would increase prospects of other localities adopting preference voting. Such a trend could be good for Cambridge, both for strengthening its voters' confidence in the system and in removing whatever burden the city might have in using an "exotic" voting system.
Community traditions: Although the count can be conducted in under two minutes using computers, the program could be adapted to maintain the Cambridge tradition of studying and discussing each transfer of ballots in a public setting. The program could be written to stop after each count until activated to do the next count.
This article is from the introduction of a report prepared by the Center for Voting and Democracy for the Cambridge Board of Elections in June 1994. For more information, contact the Center.
Table of Contents