At 05:53 PM 4/23/2007, Juho wrote: >Political elections are typically competitive. Polls are typically >less competitive. Voting on which family size Pizza (of several good >ones) to buy for the family today may well be a quite non-competitive >election.
That's true. And one might ask why. Certainly it's understandable in a family. But it is also understandable in any functional neighborhood or community organization. Why does this "non-competitiveness" break down, and under what conditions? My point in bringing up the pizza example is that I find it odd to expect that an election method will bring good results in a competitive environment when it miserably fails in a non-competitive one. That is, a group of people want to quickly find which of a number of solutions is maximally satisfactory. Isn't this the problem of elections? Somehow it is believed that this doesn't apply to politics. In politics, the thinking goes, people only want to win, to see "their side" be the victor. Could it be that the *system* -- which includes plurality elections -- actually encourages this? What we see about Range is that it works quite well under competitive conditions, it is certainly not *worse* than FPTP, and, many of us would argue, than Condorcet methods. It usually finds the Condorcet winner, anyway, in examples which stick close to real-world conditions. And when it doesn't, it finds a *better* winner. The argument against Range typically goes: in competitive elections, where people really want their favorite to win, they will bullet vote, thus reducing the election to Approval. It's probably false. Sure, *partisans* will vote that way, we can expect, but many, many voters consider themselves independents. There is a cost to bullet-voting. It is essentially an abstention from every pairwise election other than those involving the favorite. In a strong two-party contest, where third parties are essentially irrelevant, we can expect partisans not to care about that. And, indeed, many of them may *not* bullet-vote, but will assign some value to one or more third party candidates. It is a way for them to express their leanings without risk of "losing." It's important to realize that Plurality *often* chooses the correct winner, i.e., the candidate who will win under either Range or a Condorcet method. It probably does this more often than not, though strong two-party systems tend to shift positions and match each other so that they remain close to parity. The problem is in the few percent of nonpartisan voters and third-party voters, what can *they* do? Range and Approval (and IRV, though not as safely) bring them in out of the cold, eliminating the immediate spoiler effect. That alone is well worth, with Approval, simply dropping the no-overvoting rules, which in one fell swoop and with minimum fuss and no cost turns elections into the most basic Range method. The IRV promoters here have Proportional Representation as a long-term goal. They don't really talk about their strategy, as far as anything we have seen, they are *not* people who believe in democracy in reform movements, they are one-pointed and, as far as we have seen, quite inflexible. But I suspect that their strategy is to bring IRV, which is, of course, single-winner STV, and then, they might think, the path to PR is open. But STV and IRV are actually worlds apart. The offensive effects possible in IRV, particularly center-squeeze, which can easily defeat a strong Condorcet winner (who will also win under Approval), is far less damaging in the multiwinner context. STV is quite a reasonable method, multiwinner, though we think there may be better. The strategy could backfire, if there is a prominent election failure. That is, the bad taste generated as a result of IRV defects could attach to STV and thus to PR. But the IRV people (by which I mean the political activists, there are some theoretical analysts who favor IRV who are exceptions) won't actually discuss the issue. It's a shame, actually. ---- election-methods mailing list - see http://electorama.com/em for list info
