part 2 of VA & insincere extension

Mike Ositoff ntk at
Sun Sep 27 18:12:02 PDT 1998

Additionally, David said that it's reasonable to assusme a
spatial ordering, and if there's a 1-dimensinal policy space,
then the only voters who'd be indifferent between any candidates
would be voters situated between them in policy space. The
extreme voters in our examples would never be indifferent between
any candidates. The A voters would never be indifferent between
B & C.

Though it isn't what VA is optimized for, it isn't quite true
that, when nothing is known about the other voters, you have
nothing to lose by randomly ranking all the candidates, though
you don't have preferences between all of them. That's because
when you do that, you might be one of the B voters in my
41,39,20 example, and you might be helping the A voters to
make A win by insincere extension, voting C over B when they're
indifferent between those 2. True, either way, the A voters
aren't penalized & B isn't protected from defeat, so it's in
a rather weak sense that you have something to lose from
insincere extension. 

But for one thing, the A voters are dis-rewarded in the sense
that they can't gain by the strategy, while they can lose by
it just as you can (because it can help someone use the strategy
against them).

For another thing, this situation isn't what VA is optimized for.
Earlier in this message, & in the previous one, I gave some
reasons why insincere extension isn't a problem in VA.


More information about the Election-Methods mailing list