Recently someone asked about the best way to collect ordinal information. Jobst and Ted have recently suggested methods that use the basic information theoretic principle of encoding the most likely messages with the smallest code words, and getting approval information as a bonus. [The most likely messages are party and candidate preferences.] I would like to supplement their suggestions with one inspired by Joe Weinstein, a statistician who contributed to this EM list before his wife passed away a few years ago. Joe's "election jury duty" idea is based on the idea that in a large public election, a large enough sample of the voters is sufficient to determine the winner, and that, once singled out, a random sample of, say, ten thousand voters charged with deciding the election, would take this duty as seriously as a jury on a criminal case (since politicians often turn out to be criminals, anyway), and knowing that the outcome depended on them, they would study the candidates in depth, etc. and would be willing to rank the candidates on ballots more complex than mere plurality ballots, after receiving training. My idea is that in a large enough election, the individual pairwise contests could be farmed out at random to the voters. Here in Oregon everbody votes by mail. We get our ballots a month before the election, so we have weeks to study the issues and candidates, and fill in the ballots as we make our decisions. Even so when the ballots are long, it is hard to learn enough to make a wise decision on every contest. In an election with twenty single winner races, and with several candidates per race, it is hard to really get to know all of the candidates, not to mention all of the alternatives on the various "ballot measures." What if all of these races were broken down into pairwise contests, which in turn, were farmed out randomly so that nobody had to vote on all of them? To be specific, suppose that you had twenty single winner races with ten candidates each, and no ballot measures. Each of the ten candidate races could be broken down into 45 pairwise contests, so the total number of pairwise contests would be 20*45=900. If there were nine hundred thousand voters, and each of them received a random selection of ten pairwise contests to weigh in on, then each pairwise defeat would be based on ten thousand ballots, well above the statistical sample size requirement for 99% confidence. Forest
<<winmail.dat>>
---- Election-methods mailing list - see http://electorama.com/em for list info
