[EM] IRV's "majority winner". What if we let the people choose?

[James Gilmour]

> James Gilmour wrote:
> > > >49  A<C<B
> > > >48  B<C<A
> > > >  3  C<B<A
> >
> >James Green-Armytage replied:
> > >       Well, if the votes were sincere to begin with, then it is 
> > > axiomatic that C will win a runoff election against B.
> >
> >But if you did decide this by a separate run-off election, I should not be
> >surprised to find large numbers of voters changing their  preferences in 
> >that run-off election, and in so doing, reject the CW.

Adam Tarr wrote:
> You realize the sophistry in that argument, don't you?

No, there is no sophistry in my argument.

> You're essentially 
> saying that those preferences shouldn't be respected, since 
> the people 
> wouldn't actually express them if they had thought they mattered.

No. that is not what I am saying.  I am suggesting they would vote sincerely but then reject the
outcome of their own actions when they saw the consequences and all the evidence.  I am also
suggesting that if they fully understood that such outcomes were possible, they would reject a
voting system that could produce such an outcome.  IRV has a similar problem the other way up, but I
don't think IRV's problem would lead to its rejection (see below).


> >Imagine a "real-life" scenario: Bush, Gore, Nader.  Would we really 
> >have had four years of President Nader?
> 
> OK, explain to me how, in your "real life" scenario, every 
> Bush and Gore 
> supporter liked Nader more than the other guy.  Every one of them!
> 
> If you're going to transpose this into a "real-life" 
> scenario, then pick 
> someone who would actually be perceived as a moderate, in 
> between Bush and  Gore, such as John McCain.

Thanks for the corrections.  Over here all we heard about when it came to the actual election were
Bush, Gore and Nader - with Nader as the spoiler, having a very small percentage of the votes in a
very close election.  (We heard about hanging chads too, but that's another story!)


> >This is about more than voting arithmetic and measures for identifying
> >"the most representative candidate".  It brings in systems of values which 
> >are expressed in different dimensions from those used to measure representativity.
> 
> I don't understand this.

That is exactly the problem.  NB nothing personal, just a general observation about much of the
discussion on this list.

> Are you trying to say that people might not vote 
> in a way that pairwise counting measures meaningfully?  If 
> so, could you  justify that?

I'm not sure how people might respond in such a Condorcet election, but my main point is that once
they appreciate such an outcome is possible they would never accept the Condorcet voting system.
There may be good intellectual arguments that the 3/49/48 CW is the most representative candidate of
the voters, but all my political experience leads me to believe that such a result would provoke a
massive public outcry and demands for immediate changes to the voting system.

As others have appreciated, such a response operates on a different dimension from that used to
assess "representativity" as defined by the comparisons of Condorcet, IRV, etc.  You may think this
contrary, but we know from political surveys that real people are quite capable of holding contrary
views and even mutually incompatible views.  So while they might accept the CW on one dimension,
they will reject the outcome of the election on another dimension.  I don't have a neat definition
for that dimension, but it is characterised by the reactions I suggested in the two different
scenarios, ie accept CW if 32/35/33, but reject CW if 3/49/48.

IRV has a similar problem, but the other way up.  If the votes are cast
49  A<C<B
48  B<C<A
  3  C<B<A
I believe most voters would accept the IRV winner, B, as a fair outcome.  This is conditioned by the
very low level of first preference support for C.  But in the 35/33/32 scenario, I would expect
quite a hullabaloo about the unrepresentative and unfair result of the IRV election.  Again, this
response is not operating on the same dimension as that used to assess representativity.  Other
values in the total system of values come into play and determine the response.

James Gilmour