What kind of preference profiles are you using for your simulation? Are you defining voter preferences by ordinal ranking? Are the rankings entirely random, or are they built around certain patterns, e.g. issue space, or some other type of 'closeness' of one candidate to another? I'd
What I did was very simple. For each voter I select a first choice at random, with equal probability of getting any of the three candidates. Then I select the second choice the same way from the remaining two candidates.
encourage you to increase the number of candidates when possible. I
I could generalize it, but the main idea now is simply to gain insight into the critical three-candidate case.
imagine that this will decrease the probability of convergence. However, it's possible that some non-convergent situations could be more troubling than others. For example, in a ten candidate situation, if it ends up oscillating between two candidates who are relatively similar to one another, there may be some amount of stability in this even though the outcome is non-convergent. I agree with you, though, that we would like to see convergent outcomes given stable preference profiles, and I think that it is interesting to try to figure out which methods do the best job of
It just bothered me that nobody seemed to be regarding this as the dynamic feedback problem that it is. Everyone seemed to think that static analysis based on "strategy" is sufficient to understand the problem. I'm now relieved to know that not everyone thought that. Perhaps nobody did; I'm just saying that that was the impression I was getting.
providing this. I'd ask you for the code, but I'm afraid I don't know anything about computer programming.
It's not anything to write home about yet. Maybe after I clean it up a bit and add some comments I'll put it out for anyone who might want it.
--Russ
---- Election-methods mailing list - see http://electorama.com/em for list info
