[EM] serious strategy problem in Condorcet, but not in IRV?

[James Green-Armytage]

Dave, et al,

You write:
>Shame - fine - and it can be worded to apply only to those who are guilty.
>Illegal - tempting, BUT
>       B's backers can claim a legitimate goal - A is the worst enemy, so
>they should do their best to help A lose.
>       A's backers can claim a similar goal - B and C are equally
>good/bad,
>so they should be neutral as to B vs C.
>How do you prove legal guilt without reading minds?  Unless you eavesdrop
>on their strategy session and hear an admission?

I reply:
Certainly you can't punish individual voters for voting insincerely. What
you could do is punish anyone, let's say a party official or employee, who
openly reaches out to the supporters of that party and instructs them to
vote insincerely. For something like this to have an impact in a public
election, it would be necessary to reach out to *a lot* of people, so it
wouldn't be too hard to find out about it.

Perhaps it is analogous to the laws against price fixing in capitalism. It
is seriously illegal for someone from one corporation to call up someone
from a rival corporation and bargain on setting a higher price to increase
both of their profits. Even doing so in a private conversation is illegal.
This is not considered an abridgement of free speech, but a necessary law
to keep the capitalist system from falling apart. Perhaps a law against
this sort of voting strategy conspiracy could be justified on similar
grounds.

However, there is still a vague possibility that such politicians will be
able to get the same effect through sheer negative campaigning, candidates
setting up a real enmity against their biggest rival. Organized strategy
aside, it is possible that in a situation where there are two close
opponents, voters will instinctively rank the other candidate dead last,
even if there are other candidates who they actually find more
objectionable. This kind of risky strategy definitely pays off a lot of
the time if the Borda count is being used. Monroe's argument is that it
also pays off sometimes when using Condorcet, too.

So the worry is that it won't take an organized conspiracy to lead to the
"turkey raising" phenomenon Monroe talks about. Indeed, in his paper there
is no mention of conspiracies; actually his logic is based on the same
kind of analysis as the Duverger law, that is, the probability that a
certain kind of vote will affect the outcome of the election, multiplied
by the marginal utility of that effect for the voter.

It is possible to argue that if two rival candidates smash each other to
bits with negative campaigns and the result is that some third guy gets
elected, then the candidates have gotten what they deserve. Hopefully the
voters will pick the most attractive option out of the remaining
candidates.

What is not so easy to argue is that it is okay for one group of voters to
snatch the election away from a sincere Condorcet winner using strategic
voting. Whether it would take an organized conspiracy for this to happen,
or whether it could happen due to one group of voters being just generally
more devious than the other, I don't know.

You write:
>Ok, but if they do this in real voting they are into dangerous gambling,
>for they could succeed in what they claim they want.

I reply:
Yes. Definitely. That is generally what we hope will hold sway if
Condorcet is being used. Surely, if you strategically reverse the order of
your preferences, you are risking that you will help elect a candidate who
you like even less than the one you are trying to thwart.

However, it is not guaranteed that voters will unwilling to take that
risk. 

When the A voters truncate in my example, then they are attempting to make
that risk real, the risk of electing C if the B voters don't shape up and
vote sincerely, or at least just truncate themselves. This creates a game
of chicken though, as I said, where risk is balanced against reward, and
the outcome depends not so much on voter preference as it does on how they
perform in the game of chicken, in terms of their temperament, their
strategy etc.

By the way, as far as the idea of throwing the steering wheel out the
window to win a game of chicken, perhaps the best strategy by members of
one party would be to all cast absentee ballots a certain way (as early as
possible), inform the members of the other party you have done so, and
force them to choose between submission and electing the car crash
candidate.

James