Shortcomings of Instant Runoff Voting
Dave Farber’s Interesting People list recently carried a brief article by the computer scientist Stephen Unger, warning about the shortcomings of Instant Runoff Voting.
As Unger points out:
… while IRV does work as advertised in some important situations, there are also situations in which it produces bizarre results. For example, there are cases where a vote FOR X can help DEFEAT X. Another problem with IRV is that counting of votes cannot be decentralized. All ballots for a statewide election, for example, must be processed together. This is in contrast to the distributed counting process (precinct by precinct) used conventionally for other methods. This increases costs and makes it more difficult to safeguard against error or fraud.
Instant Runoff Voting certainly has its virtues (as I aim to show), but I’ll always argue that anyone who wants to repair the US public election system should focus on other priorities, such as: 1) assuring vote integrity through auditable paper ballots; 2) weekend voting and other measures that would boost turnout, and; 3) better voter education. Move to IRV (or range voting, or Condorcet) without addressing those other problems, and peoples’ faith in the system is likely to drop even lower than it is now.
Filed under: Instant Runoff Voting
One Response to “Shortcomings of Instant Runoff Voting”
Leave a Reply
You must be logged in to post a comment.
Among those who know election methods, there is an emerging consensus that the first election method reform should be a very simple one: just Count All the Votes. I.e., stop discarding overvoted ballots, just count them. This effectively implements Approval Voting, which is the special case of Range Voting with only two ratings (Yes and No). Clearly, it is not the most sophisticated method, but it is *almost* as good as the very best at practically no cost. It does not suffer from the IRV problems.
As to vote integrity, there is also a very simple solution, an expansion on the use of “auditable paper ballots,” and that is ballot imaging. If when ballots are first removed from the deposit boxes they are serialized and imaged, both by election officials and independently by any observers who care to do it, and the images are made public, *anyone* can verify vote counts. Media, I’m sure, will rush to do it, since with good automatic recognition software they may be able to generate unofficial results, quite accurately, before the official counts. And the official counts can be done on copies of the ballots rather than on the original ballots, which can be sealed and sequestered for court review if it is ever necessary.
The technology necessary? Present in every office. Any fax machine can handle the imaging. So can most digital cameras, quickly and cheaply. Images can be counted by hand or by open-source recognition software. Reading marked ballots automatically is a far easier task than optical character recognition, and OCR software is common now.
The serialization makes possible distributed counting methods where people divide up the counting tasks, and counting is redundant, which can then be cross-checked since ballots are numbered.
The only argument of any weight I’ve seen against this proposal is that it could help with vote buying and vote coercion, schemes where the fraud involves the voter specially marking the ballot. Those are not currently big problems in the U.S., and can be addressed in other ways than by keeping ballot visibility confined to election officials and observers, especially given that most election fraud involves precisely those officials. Corruption thrives in secrecy.