[EM] IRV in action

[Markus Schulze]

Dear Dave,

I wrote (7 April 2003):
> Example:
>
>    40 voters vote X > Y > Z.
>    35 voters vote Y > Z > X.
>    25 voters vote Z > Y > X.
>
>    Candidate Y is the IRV winner.
>
>    Suppose that candidate Z asks his supporters to bullet vote.
>    Then this example looks as follows:
>
>    40 voters vote X > Y > Z.
>    35 voters vote Y > Z > X.
>    25 voters vote Z.
>
>    Now candidate X is the IRV winner. Now candidate Z can hope
>    that some of the supporters of candidate Y will give their
>    first preference to candidate Z to keep candidate X from
>    winning. Then this example looks as follows:
>
>    40 voters vote X > Y > Z.
>    35 voters vote Z > Y > X.
>    25 voters vote Z.
>
>    Now candidate Z is the IRV winner.

You wrote (7 April 2003):
> Given the candidate's positions being equidistant:
>       IRV - I see not how bullet voting matters, for nothing else is seen
> until/unless your candidate loses.
>
> Given two candidates being siblings, far from the third:
>       IRV - voters for a sibling candidate almost certainly want to give
> the other sibling second preference, to decrease the chance of the third
> candidate getting elected.  I would expect these two candidates to promote
> this voting.  I do not see the third candidate caring, as I said above.

Suppose that there are three potential winners: X, Y, and Z.
Suppose that Y pairwise beats X when every voter votes sincerely.
Suppose that Z pairwise beats X when every voter votes sincerely.
Suppose that Y cannot pairwise beat X without the votes of the Z supporters.

Then it makes sense for the Z supporters to bullet vote so that the
Y supporters see that Y cannot win. At least those Y supporters who
have Z as their sincere second preference will then give their first
preference to Z so that Z wins.

Markus Schulze