[EM] Real IRV Election, Disputable Result

[Jan Kok]

Holy sacred cow, Batman!!

According to Brian's analysis, Miller was the Condorcet winner, but
Kiss won the actual IRV election.  Miller was preferred over Kiss 3991
to 3455.

Brian, the number of first-choice votes according to your histograms
doesn't exactly match the numbers on the Burlington election site, for
the first round summary.  For example, you show Kiss got 3808
first-choice votes, but the election site shows 3809.  Do you know
what caused the discrepancies?  Did you account for spoiled and
truncated ballots correctly?

Thanks for doing this analysis!  This is BIG news in the small world
of voting methods! :-)

- Jan


On 3/10/06, bql at bolson.org <bql at bolson.org> wrote:
> Here it is folks, the one we've been waiting for:
>
> http://www.ci.burlington.vt.us/ct/elections/
>
> A real live IRV election, run last Tuesday, and different methods give
> different results.
>
> The raw data can be downloaded here:
>
> http://www.burlingtonvotes.org/20060307/
>
> I've adapted the data and run it through my counting software to get this:
>
> http://two.bolson.org/a.html
>
> I should run it through more of the systems I've implemented. Maybe I'll
> do that tonight.
>
> Anyway, I've collapsed the voting data to one file here:
>
> http://two.bolson.org/20060308BurlingtonVT.nameeq.bz2
>
> It's in a url-encoded format "name one=3&name two=1" etc. That's been
> modified to be ratings in a sorta borda fashion; in this case 6 - ranking.
>
> Brian Olson
> http://bolson.org/
> ----
> election-methods mailing list - see http://electorama.com/em for list info
>