At 10:58 AM 10/15/2005, Warren Smith wrote: >I did not say voters always want full rankings. > >What I said was, I think there is a substantial SUBCLASS of elections, >in which, all voters (except perhaps for a few who are insane or writing >illegible ballots or something random like that) will want to >provide full rankings.
Even if there are ten candidates? "Substantial?" The basis for this claim? Yes, if there are two candidates, all voters will want full rankings.... And there are voters who will feel confined with three ranks. Fewer with four, I submit, and drastically fewer with five. At some point there is no substantial confinement. Frankly, I think two or three explicit ranks, plus the default last rank, is probably the balance point, beyond that the benefits of increased ranking come to be outweighed by ballot complexity considerations. However, I see no harm in additional ranks being available, provided that somehow the technical problems are addressed and are not significant. Note that if the method allows equal ranking, adding clones does not require additional ranks. >If your voting method misbehaves badly commonly when confronted >with elections from within >that subclass, then it is not a good voting method. Only if the subclass is reasonably possible. Further, if there is a marginal possibility, and the failure is minor under those conditions (such as a narrow failure to elect the Condorcet winner but instead electing one who is *almost* the Condorcet winner), we could consider the defect not serious. After all, we saw a rather large majority criterion failure in the 2000 U.S. Presidential election, and practically no consequences ensued: those who could have complained, and were in a position to do something about it (the members of the House and Senate) did nothing: a few House members complained and it would have only taken one Senator signing off on the complaint for it to proceed through formal process. And not one Senator agreed. At a point when the Democrats had a slim majority in the Senate.... I'd say that the failure then was serious. If the only problem with the 2000 election had been a narrow loss for Bush in the popular vote, and a narrow win in the electoral vote, then, no, it would not have been serious. But there were more substantial problems than that, starting with the fact that the popular vote margin was not all that small for Gore. And it gets worse.... Nor do I mean that there were no serious consequences from that election. However, it is not my purpose here to engage in partisan debate. Just to note that election methods don't have to be perfect; when the electorate is divided in two, and you have a single-winner election, no matter how you slice it, 50% are going to be dissatisfied. Does it really matter if it is 49.9% or 50.1%? Only in "principle," not in reality. In both cases the society is divided and weakened as a result. >That was all I claimed. Now since I am asked for supporting >evidence, I respond >that many books and many authors, simply refuse to consider truncated ballots. In other words, the evidence for the view that truncation is bad is that many consider it bad. Oh. Okay, never mind.... If you fall for that one, I can sell you on a whole series of myths. Let's start with the myth that saturated fat is bad for you, regardless of the rest of your diet.... Practically everyone believes that now. Yet it was never demonstrated, and is probably an error. Indeed, I'm betting my life, literally, that it is an error. And, so far, so good, lab work looks great, my doctor was impressed. But I won't *really* know until I pop for another cardiac CAT scan, and even then, it's only anecdotal. >Also some countries make them illegal. Some countries make democracy illegal. Others imagine that democracy is enhanced by coercing the population. > I suggest to you, that much as you may >dislike this, there is a reason they did that, Undoubtedly. Perhaps they have leaders with a habit of forcing their brilliant ideas on others. > and it is not a good idea to >ignore it and pretend it does not exist when we design voting methods. Pretend that *what* does not exist? I don't think anyone pretended that nobody desires full ranking. After all, some people here do just that. But what is unclear is how important this would be to an actual electorate. I have yet to hear, so far, any complaints about the IRV elections in San Francisco, which did not allow full ranking. And cogent arguments have been advanced that people who actually do rank a large number of candidates would not generally be able to remember later what exact ranking they expressed. With a large number of candidates and with the vast majority of voters, most candidates would fall into clumps that could be equally ranked within the clump; what seems to be most important is that voters be able to express a favorite and that this help the favorite to win and not help someone else to win instead, maybe that voters be similarly able to express a most-disfavored, with maximum effect away from the election of that candidate, and that there be perhaps a few ranks in between. Full unique ranking with a large number of candidates just does not seem to be a highly desirable characteristic to me. And I think that actual experience with elections would bear this out. But, of course, the proof is in the pudding. ---- election-methods mailing list - see http://electorama.com/em for list info
