[EM] Free riding

[Raph Frank]

On Sun, Aug 31, 2008 at 8:21 AM, Juho <juho4880 at yahoo.co.uk> wrote:
> Woodall free riding uses some irrelevant candidate that is ranked first.
>
> Hylland free riding does not rank the favourite candidate.
>
> A third approach to free riding is to rearrange the candidates to reflect
> the estimated probabilities.
>
> The true preference order of a voter is A>B>C>D>E>... The voter expects A to
> be elected quite certainly. Candidates B and C are less certain. The voter
> considers B and C to be almost as good as A. Candidates starting from D are
> considerably worse. As a result the voter decides to vote B>C>A>D>E>...

I think this is the strategy that most parties actually use for vote
management.  They never recommend to the voters not to rank a certain
party member.

They just tell them to rank a certain candidate first.  This means
that the high probability candidate ends up getting less first choice
votes.

In fact, it is often not even recommendations.  Each candidate is
often given a specific area of the constituency where he can canvas
in.  This pushes up his first choice total.

> It is possible that B or C gets elected while A will not,
> but the risks are not too big.

Yeah, that has happened in Ireland.

http://en.wikipedia.org/wiki/Mary_O'Rourke

She was a cabinet minister, but they over-estimated her popularity.

> In any case the three clearly best candidates
> (A, B, C) will get all possible power of this vote.

Not necessarily.  If the popular candidate doesn't get elected, then
some of his personal vote is lost.

E.g if there are 2 candidates and they get

A1:
0.5 (personal)
0.1 (party)

A2:
0.7 (party)

A1 is eliminated first.  A2 gets the 0.1 party vote transferred and
thus has 0.8 quotas and may not got a seat (depends on how much of the
personal vote of A1 stays with the party).

If A1 had been allowed to campaign normally, it might have gone

A1:
0.5 (personal)
0.3 (party)

A2
0.5 (party)

A2 is eliminated and A1 gets 1.3 quotas and thus takes the seat.

> There is also no risk of
> the vote going to some irrelevant candidate (as in Woodall free riding).

That is true.  Apparently, the election reform society ran a PR-STV
election and a candidate almost got elected by suggesting that people
rank him 1 in order to use Woodall free riding.

It is clearly better that if such a thing happens, at least you have
elected a fellow party member.

> This generalizes to any preference order, not only to the handling of the
> first favourite.

True, but it is probably not really worth the effort.  You would be
estimating the odds on the state of the count after many round.