On Sun, 6 Jun 2004, Ken Johnson wrote: > > >Date: Fri, 4 Jun 2004 17:45:49 -0700 (PDT) > >From: Forest Simmons <[EMAIL PROTECTED]> > >... > >No runoff method (or any other sequential elimination method) has simply > >"honest ballot" as optimal strategy: > > > >Suppose that there are three candidates that you honestly rank as A>B>C. > > > >Suppose further, that you are reasonably sure that A could not beat C if A > >survived to the last round, but that B could beat C if B survived to the > >last round. Then it might be to your advantage to rank A in second (if > >not last) place, contrary to your honest preference order. > >
> [Ken acknowledges and asks ...] > This is true, but what about zero-info strategy? Could "honest ballot" > be optimal for zero info? Forest replies: With zero information, honest ballot is optimal for just about every sequential elimination method that I know of (except frivolous ones where candidates are eliminated capriciously). > Also, in the more general case, could the > optimal strategy be to at least give honest relative rankings/ratings > among those candidataes who you think could possibly win? There would have to be a statistical indifference among them. What bothers me about this is that rich interest groups can manipulate voter perceptions, so that instead of zero information you get strongly touted disinformation, which results in voter incentives to vote insincere rankings that turn out to be contrary to voters' own interests. This is why some of us have proposed "Runoff Without Elimination:" The runoff stage is just a way of establishing relative candidate strengths more reliably than the polls. These candidate strengths can be used (for example) to determine reasonable approval cutoffs on the ballots. All candidates are included in the final approval tally. Suppose we use IRV as the runoff method. Then on each ballot we could (for example) put the approval cutoff adjacent to the IRV winner on the side of the IRV runnerup. Example: 40 ABC 25 BCA 35 CBA C is the IRV winner, and A is the IRV runnerup, so the approval cutoffs are 40 AB|C 25 BC|A 35 C|BA and the approval scores are 40, 65, and 55, respectively, for A, B, and C. This IRV without elimination method is a crude example of a "runoff without elimination" method. However crude, it seems to improve on plain IRV. I deem it to be inferior to plain approval, though there might be other "runoff without elimination" methods that could compete with Approval (in areas other than simplicity of ballot). Forest > > ---- > Election-methods mailing list - see http://electorama.com/em for list info > ---- Election-methods mailing list - see http://electorama.com/em for list info
