[EM] Real IRV Election, Disputable Result

[Jonathan Lundell]

At 11:08 PM -0500 3/10/06, Abd ul-Rahman Lomax wrote:
>At 09:39 PM 3/10/2006, Jonathan Lundell wrote:
>>  >Thanks for doing this analysis!  This is BIG news in the small world
>>>of voting methods! :-)
>>
>>How so? It's well known that IRV/AV/STV doesn't necessarily find the
>>Condorcet winner. It shouldn't be too surprising that there are
>>real-world examples.
>
>Actually, IRV proponents routinely claim that, yes, theoretically, 
>IRV can fail to select the Condorcet winner, but the circumstances 
>are so rare that they don't happen. So this is significant.
>
>I've also seen the claim that the IRV winner deserves to win in a 
>case like this, for some BS reason that I don't remember. It is hard 
>for me to remember nonsense.

One reason is that IRV honor later-no-harm. Methods that violate 
later-no-harm (like any Condorcet method) encourage insincere voting. 
That's a problem. You may prefer the tradeoff of Condorcet methods, 
but they have drawbacks as well.

That said, I'm a little confused by Brian's summary of the election. 
As I read it, Miller is preferred to Kiss 3991 to 3455 in the 
pairwise table (at least that's what I take the "virtual round robin" 
to mean).

But the Burlington results page give Kiss 3809 first-round first 
choices. So how can only 3455 voters prefer Kiss to Miller? Or am I 
misreading Brian's tables?

Likewise, in the final round, Kiss beats Miller 4761 to 3986. Kiss 
must be preferred to Miller on 4761 ballots, no?

Help me out here.
-- 
/Jonathan Lundell.