[EM] FYI: Tacoma park IRV vote data

[Jonathan Lundell]

On Apr 15, 2009, at 8:20 AM, Kristofer Munsterhjelm wrote:

> Jonathan Lundell wrote:
>
>>> C wins by 70/93 = 75.3% of the votes. What a landslide!
>>>
>>> (Schulze and MAM gives A > D > C > B, and IRV gives A > B > C = D.)
>> That's really a mischaracterization of IRV. IRV (and STV in  
>> general) does not produce a candidate ordering. It simply finds a  
>> winner by interpreting the ballots as a list of contingent choices.  
>> In particular, no ranking is implied by order of elimination.
>
> Alright. The point of the ordering was to show that the elimination  
> method of "eliminate whoever got the most Plurality votes" is rather  
> bad, and also that it produces a different result than IRV. Schulze  
> and MAM agrees that C is the next worst candidate, and IRV (like  
> Schulze and MAM) elects A as the winner.
>
> Let's call the method of my previous post "Worst Runoff". Worst  
> Runoff, like all single-elimination methods, passes Condorcet Loser  
> - so there's no way to get it to elect B.
>
> Whether or not IRV produces an ordering is not really important.  
> What is important is that you can stretch the "elected by a  
> majority" definition to apply to any elimination system - at least  
> unless you provide the caveat of "assuming the voters know what the  
> method does, so that when they submit ranked ballots, they  
> implicitly consent to the method's treatment of those ballots". If  
> one uses this qualification, then IRV's property of "elects what a  
> majority of the voters want" is either incorrect (because the voters  
> don't "know what the method does" to the intuitive level required)  
> or is a near-tautology (as any method that takes a majority of the  
> ballots into account would satisfy this criterion).

I agree to this extent: "majority" is a fuzzy concept, and in any  
particular application it needs to be precisely defined.

I think it's easy enough to phrase IRV voting instructions in terms of  
contingency, though I don't think it's particularly harmful for voters  
to think of their ballot as a simple preference ordering, because the  
likelihood that a voter will have sufficient information to make it  
worthwhile to vote strategically is vanishingly small, at least in  
real public elections.