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

[Markus Schulze]

Dear Jonathan Lundell,

Greg wrote (25 Nov 2008):

> While complete ballot data is ideal, I think a convincing
> case as to how a voting method might perform in a particular
> election can sometimes be made from polling data. For example,
> there's good exit polling data for the Senate race in Minnesota
> that's being recounted, showing that supporters of the
> Independence party candidate would have preferred Al Franken
> over Norm Coleman by a 5% margin. That would have given Franken
> at least another 20,000 votes, way more than the 215 votes he
> trailed by pre-recount. I think that's a pretty good case that
> IRV would have selected Franken, regardless of the results of
> the plurality recount.

I wrote (25 Nov 2008):

> Then what do you say about the opinion polls that said that
> Bayrou was a clear Condorcet winner in the 2007 French
> presidential elections (although IRV would have chosen
> Sarkozy)?

You wrote (25 Nov 2008):

> We don't actually know who IRV would have chosen, since
> the polling (and the campaign) didn't happen in the
> context of an IRV election. It's not an unreasonable
> conjecture that Bayrou would have gotten a larger
> percentage of first choices (some from Sarkozy and
> Royal) under IRV. Nor do we know how the smaller
> party votes would have transferred.

So is it feasible to use polling data to show that
an election method would have violated some desirable
criteria? Or is complete ballot data needed?

Or are only IRV supporters allowed to use polling data
to show the greatness of IRV, while advocates of other
methods have to use complete ballot data?

Markus Schulze