From swillett@igc.apc.org Thu Oct 12 10:26:34 1995 To: loeb@psyche.mit.edu Subject: Re: Cambridge PR Voting Demo Cc: jlindsay@netcom.com ---------------------------------------------------------------------------- > From loeb@psyche.mit.edu Tue Oct 3 13:23:18 1995 > Date: Tue, 3 Oct 95 16:23:31 EDT > From: Eric Loeb > To: swillett@igc.apc.org > Subject: Re: Cambridge PR Voting Demo > > I mailed you the code. Let me know if you have any questions > or comments. It was hacked together rather quickly... > > eri ---------------------------------------------------------------------------- Thanks, Eric, for sending me your code for this demonstration. I appreciate the effort you have put in on this demonstration, since I know from my own experiences that people can get really confused trying to figure out how PR systems work! Anything which gives them an example helps make the process more real. My concern is that I don't think your demo correctly implements voting the way the Cambridge rules require. I don't claim to be an expert on the Cambridge election code, but I talked to Jim Lindsay, who authored PRMaster, a PR vote processing program (which successfully duplicated the count of Cambridge's last election) and we are concerned with the way you have handled resolving a tie. One of the key features of STV voting (the single transferrable vote method which Cambridge uses) is that additional (lower) preferences a voter expresses can in no way adversely affect a candidate they prefer more. If, however, you tally second choice ballots while the fate of first choice candidates is still undetermined then someone's second choice could knock out their first choice candidate! This implies that you MUST resolve ties if there are no more surplus votes to be transferred and the election is not yet resolved before you can proceed to count second (or third, etc.) preferences. I think Cambridge does this randomly (select one of the tied lowest vote-getters and eliminate them). Then the votes of that candidate are transferred to the next candidate ON THOSE BALLOTS. You do not count ALL the second (or third, etc.) choice votes, just the next choice on the ballots which have been "released" by eliminating their current top choice. I hope this makes some sense. I would be happy to continue to discuss these rules and, if I am incorrect, I hope you will let me know. Once again, I appreciate the time you have spent trying to work out this demo! Steve Willett