APPENDIX 2
PUTTING A VOTING SYSTEM IN CONTEXT
After Thomas Friedman's column appeared in the New York Times titled "A Tea Party Without Nuts" on March 24, 2010, the Times published letters dealing with voting reform. One was from Professor Eric Maskin at the Institute for Advanced Studies in Princeton, N.J. In response to that letter, I sent an e-mail on some of my observations to Professor Maskin to which he replied with some comments and some interesting material which he generously shared with me. Maskin is a proponent of Condorcet elections as outlined in my exchange with Dave Ketchum in the chapter on the Condorcet candidate. I wrote the following after reading the piece Maskin published in Scientific American along with Partha Dasgupts in March 2004. The main point of my response is that one has to look at voting systems in their broader context in deciding what would be the most suitable system to use.
April 16, 2010
Professor Maskin, Yesterday evening I read with great interest and pleasure the Scientific American 2004 article which you wrote with Partha Dasgupta. The piece is exceedingly well written and argued and I think a real public service.
I did notice, however one serious error in the article that you may be aware of. In the piece you state correctly that: "On the other hand, the American Constitution stipulates that, absent a majority of the electoral votes, the House of Representatives should determine the winner." But, the statement that "With a Republican majority in 2000, the House would presumably have gone for Bush" misrepresents the situation. If 26 of the smallest states all had Democratic representatives or a majority of Democrats each and the largest 24 states all had Republican representatives or a majority of Republicans each, the Republicans would surely represent a majority in the House, but the likely winner would have been Gore. This is because Article II Section 1 of the Constitution stipulates that in the event the election goes to the House, the members vote by state in which each state has only one vote. States like Alaska, Vermont, North Dakota, South Dakota and Delaware all have one representative each which in the voting is equivalent to the votes of states like California, Texas, New York, Florida and Illinois with 53, 32, 29, 25, and 19 representatives respectively in the current Congress.
I would not necessarily disagree with your statement that Condorcet voting is superior to all others. In the book I put together last year on "Third Parties and Voting Reform: The American Dilemma", I took the position that like the idea of the Pareto optimum in microeconomics, using the idea of the Condorcet candidate in judging the effectiveness of voting systems makes a lot of sense. This is a lot different from the criteria which the Range Vote people, like Warren Smith, advocate in judging voting systems which I maintain implicitly implies making interpersonal utility comparisons. But even if one accepts the Condorcet criteria, it surely is not the only factor that one needs to consider in judging the desirability of alternative voting systems since such systems exist in differing institutional and social contexts. Let me give some examples to explain more clearly why I might still consider approval voting to likely be the best choice currently available.
In your reply to me you stated that "Condorcet requires technology that is no more complicated that IRV or single transferable voting, which are already widely used in large elections." This may be true in places like Australia or Ireland. However, I would note that here in Michigan the city of Ferndale voted in a 2004 referendum to institute IRV in its elections. But it is my understanding that the Michigan Secretary of State told the city that it could not do this because the state lacked the voting infrastructure to institute such a system. I know that we have had one candidate for the office who has advocated getting such equipment, but doing so will clearly take time and money. From my perspective we need a voting reform now that will open up our political processes to third parties and this may well mandate looking at approval voting as a better way to go currently.
There are other considerations as well when looking at voting systems.
In 1991 when I gave a talk to a small group from the Academy of Sciences in Moscow, I advocated using a variant of approval voting to elect a constituent assembly to sort out the country's problems.
What I advocated involved ranking elected candidates by their levels of approval in percentage and using that to restrict the number of members of the KPSS in the assembly to the historical percentage (24%) in the Constituent Assembly that Lenin closed down in 1918. Thus beyond the 24%, the candidates in all the other districts with the highest approval rates that were not members of the KPSS would be those elected. Given that our legislatures are the bodies that provide oversight over our legal profession, it seems to me that a similar restriction might be very useful for lawyers in the American context for which an approval vote election could be more appropriate than a Condorcet election.
Another area where I might have some reservations in favor of approval voting involves the issue of transparency in two senses. First, in much of the literature I have seen on voting systems a sharp distinction is often made between voting sincerely and strategically.
It is my contention that it is important for voters in revealing their preferences to be able to do both. But to do so, one needs to have a simple voting system and feedback mechanisms, e.g., pre-election polls, to help voters on this. It is not so clear to me that a Condorcet election would provide such clarity better than an approval vote election. Second, a voting system itself needs to be easily understood by voters in terms of how it arrives at results. This is an area where I think that IRV really fails. But it is also not so clear to me that a Condorcet election would be superior to an approval vote election on this score as well.
In looking at voting systems, there is a crucial piece of information which I believe neither IRV nor Condorcet contain which approval voting may contain. And that is: Is an elected candidate really approved of by a majority, or at least a plurality of the voters, or in ranking candidates are voters simply ranking according to the lesser of bad choices? In the latter case, I would maintain that we have a problem. If under approval voting in a five person election no candidate could garner more than 40% of the vote, would not that be rather prima facie evidence that the slate of candidates was unsatisfactory and should be replaced by another slate in a new election? Once approval voting in the traditional sense were instituted one could consider adding a single vote in which people voted none of the above and include that in calculating approval rates with a mandatory second election with a totally different slate if no single candidate could garner at least a 50% approval rate. It is not so clear to me that such a voting system would be inferior to a Condorcet election.
In judging voting systems, I do not mean to imply that the considerations you set out are not important. On this score, I do think your work is interesting, relevant and clearly has merit. Given this, I think you and Dasgupta have done a real public service, which I wish more economists would do, in publishing the type of article you did in Scientific American. But I still advocate trying to push for approval voting which I think both you and Dasgupta would probably concede is superior to both plurality voting and IRV. In the event you may find it of interest, I attach below a copy of an unpublished op-ed I did on this which, with permission, plagerized a good deal from an earlier piece written by Professor Ben Zuckerman of UCLA. Regards.
John Howard Wilhelm
nytoped-17