[EM] Why I think IRV isn't a serious alternative 2

[Juho Laatu]

--- On Mon, 22/12/08, James Gilmour <jgilmour at globalnet.co.uk> wrote:

> The two situations I had in mind were:
> Democrat candidate D;  Republican candidate R; 
> "centrist" candidate M
> 
> Election 1
> 35% D>M;  33% R>M;  32% M
> 
> Election 2
> 48% D>M;  47% R>M;  5% M
> 
> M is the Condorcet winner in both elections, but the
> political consequences of the two results would be very
> different.  My own view
> is that the result of the first election would be
> acceptable, but the result of the second election would be
> unacceptable to the
> electorate as well as to the partisan politicians (who
> cannot be ignored completely!).  If such an outcome is
> possible with a
> particular voting system (as it is with Condorcet), that
> voting system will not be adopted for public elections.

Example 1.

In this country there are 10 small parties.
Each party has one candidate. In addition
there is one neutral candidate that has
worked to make the fighting parties agree
with each others. All voters vote
MyPartyCandidate > TheNeutralCandidate > ...
The neutral candidate gets 0% first
preferences but he might be the best
president anyway.

Example 2.

Voters of the large parties think that one
should not elect a president from a 5%
party. The votes are 48% D>R; 47% R>D; 5% M.
Candidate D will win. If the voters think
that it is better to elect a candidate from
a 5% party than to elect the candidate of
the large competing party then they could
vote as in your example. Maybe they gave M
a mandate although he is from a small 5%
party.

Example 3.

IRV philosophy is to emphasize the
importance of first place preferences
(among the remaining candidates during the
elimination process). First place
preferences could be considered to be a
sincere target in elections where the
elected person will need support of his
first place supporters when in office
(well, the voters could also simply vote
for candidates that have sufficient first
place support among the citizens). One
could do a similar trick also in Condorcet.
Let's say that a pairwise preference has
strength 1.5 if it includes the most
preferred candidate of the ballot. In that
case with preferences 43% D>M; 42% R>M;
15% M the pairwise comparison results would
be D-R 64.5-63; D-M 64.5-64.5; R-M 63-65.5.
That is a tie between D and M. Would this
be a good balance between the "35% M" and
"5% M" examples?

Juho