July 2010
MY REPLY ON VOTING
In response to some of the reactions to my piece on approval voting, I want to comment on some of the important issues that were raised. But first I need to reply to the serious charge that I did not do my research in writing the piece, which I consider quite unfair.
Last October when I met in London with a political advisor in Nick Clegg's parliamentary office, it was made quite clear to me that the Liberal Democrats were in favor of the voting reforms put forward by the Jenkins report. That proposal suggested using the alternative vote in single member districts and the single transferable vote in multi-member districts. I focused on the former because that is what the proposed referendum on voting reform will focus on. And I also did so in my January 12, 2010 memo to Nick Clegg which my posted comment briefly summarized.
Here in the States those associated with the Center for Voting and Democracy, or FairVote; which seems to have close ties with the UK's Electoral Reform Society; maintain that the nonmonotonicity problem outlined in my piece is a very rare occurrence. One response even suggested that it might be a once in a century phenomenon. But the fact is, as my earlier piece clearly showed, that it is quite feasible under the alternative vote, because of the nonmonotonic behavior of the system, for a candidate slated to win an election to lose by garnering more first place votes in the course of a campaign.
Aside from situations in which under the alternative vote one candidate has a clear first place majority and an instant runoff is not necessary, occurrences of nonmonotonicity do not appear to be so unlikely. Recent work by Joe Ornstein suggests that such occurrences in truly competitive elections may be disturbingly large involving from 13% of such elections up to a little over 40% in certain close elections. This means that under the alternative vote when a voter casts a vote for his or her favorite candidate he or she cannot know for sure whether that vote will help or hurt that candidate. Even under first-past-the-post, or plurality voting, such a perverse result of voting cannot occur.
The situation with respect to the single transferable vote (STV), which like the alternative vote is based on the Hare system, is hardly any better. As Sir Michael Dummett pointed out in a 2005 interview in the academic journal Social Choice and Welfare, the STV may have an advantage in that "it guarantees that the minority gets represented.
Otherwise is has no advantage at all. It's an almost chaotic system in the sense that a very small change in preferences would have results that change completely the outcome because it would affect the order in which candidates got eliminated, and so which votes were distributed at each stage. So I think that it really is a very bad system..."
In the same interview, Sir Michael made the point that "it is essential to a voting system that the count should be simple to carry out and for the voters to understand." Approval voting meets this requirement; a mixed voting system using the alternative vote and the single transferable vote does not.
The argument that approval voting leads to the election of a bland candidate is something of a red herring. There is a marked tendency under approval voting for the candidate favored by a majority over each of the other candidates to get elected. If not, the candidate elected will be the candidate supported by the largest number of voters. In terms of representing voters' preferences in the outcome, this is surely preferable to what can happen under first-past-the-post or plurality voting; or any variant of the Hare system (AV or STV).
Had approval voting been in place in the 1992 US presidential election, it is likely that Ross Perot, hardly a bland candidate, could have been elected. This and empirical research by Brams would strongly suggest that blandness is hardly likely to be a serious attribute of the results of an approval voting elections; unless that is what voters want.
If the use of approval voting does not give a satisfactorily less skewed result in terns of the proportional representation in a parliament or national legislature, it would give the information that could be used to allocate additional seats to party lists to give better proportional representation at the national level. Mexico, for instance, uses such a system. One could elect a 500 member parliament by a combination of 400 single seat constituencies and 100 seats allocated by a suitable algorithm to under-represented parties based on the national vote. Such a system may not be perfect, but then no voting system is likely to be; though some are better than others.