[EM] Summability

[Markus Schulze]

Dear Adam,

you wrote (24 July 2003):
> Well, I don't think anybody here would deny that.  But the simple fact
> that one can more easily get a sense of how the polls will translate into
> election results, does not mean that the method is more manipulable in
> reality.  A well-designed Condorcet system may be transparent to the
> voters, but they will have a hard time manipulating it, and will
> generally find an honest ballot to be a good choice.
>
> IRV, on the other hand, will produce fairly chaotic results in close
> races between more than two candidates.  This means voters will be unsure
> what their best strategy is, any may well end up regretting their vote.
> If all the voters had good information, then they would be more likely to
> know when to compromise and when to vote honestly, and the average voter
> would probably be more satisfied with their vote.
>
> So the intractability that comes with IRV does not, in my mind, count as
> a significant advantage.  Denying people their optimal strategies due to
> lack of information does not make better results especially likely when
> the results are chaotic.
>
> Note that I'm not arguing that summability is a crucial criterion - I'm
> just saying that I don't see non-summability as an advantage.  All things
> being equal, I'd choose a summable method, but all things are rarely
> equal.

The task of an election method is to find the best candidate according
to a given heuristic. The task of an election method is not to help
voters vote strategically.

Markus Schulze