[Election-Methods] RELEASE: Instant Runoff Voting

[Aaron Armitage]

--- On Tue, 7/29/08, Terry Bouricius <terryb at burlingtontelecom.net> wrote:

> From: Terry Bouricius <terryb at burlingtontelecom.net>
> Subject: Re: [Election-Methods] RELEASE: Instant Runoff Voting
> To: eutychus_slept at yahoo.com, election-methods at lists.electorama.com
> Date: Tuesday, July 29, 2008, 1:19 PM
> Aaron,
> 
> Just four little points to what Aaron Armitage wrote...
> 
> 1. <snip>
> "You claim, in short, that using the same inputs
> differently makes them 
> different inputs, and that producing the same kind of
> outcome differently 
> makes it a different outcome."
> <snip>
> 
> I believe James was arguing that while a voter's
> preferences in her mind 
> might be the same, but knowing whether the ballot would be
> counted using 
> one method or another set of vote processing rules (with or
> without 
> later-no-harm protection, for example) will change how the
> voter will mark 
> the ballot. Or put another way, two ballots with identical
> rankings on 
> them may in fact reflect very different actual preferences
> by these two 
> voters depending on which vote processing rule is going to
> be used. Thus, 
> one can't simply say inputs (if one means actual voter
> preferences) are 
> identical by looking at the rankings without regard to the
> vote processing 
> rule in place.
> 

But voters' sincere preferences aren't the input for any voting system
because we don't have access to them. Every ranked ballot system works with
ordered lists of candidates provided by voters. I know that strategic
voters will adjust how they use the input based on their understanding of
the exact mechanism by which the input is turned into an outcome (that's
the definition of a strategic voter). But susceptibility to these kinds of
manipulations is simply one more way of comparing what remain different
ways of converting the same kinds of inputs into the same kinds of
results; in fact, several of the social choice criteria directly relate to
strategic vulnerability. We're talking about how well Condorcet and IRV
perform the same function, not two different functions.

> 2.  <snip>
> "Under any definition of "one person, on
> vote" that Bucklin fails, IRV 
> also
> fails. But that wouldn't be a proper definition
> anyway."
>  <snip>
> 
> Not so. A single transferable vote is very different than a
> Bucklin 
> additive vote. Under IRV each voter has one vote for one
> candidate counted 
> in the final tally. Under Bucklin, voter A may have one
> vote in the final 
> tally, but voter B has two votes for two candidates in
> opposition. One 
> court ruled that Bucklin violated the one vote-one person
> concept, while 
> another court ruled that IRV upheld it. Since these were
> different courts, 
> it certainly isn't conclusive, but the difference is
> significant. I 
> personally think that methods like Bucklin and Approval
> might be seen as 
> satisfying one-person one-vote (nearly as well as IRV)
> because a "vote" is 
> an expression of the voters choice on the matter at hand,
> and all voters 
> have equal rights to mark the ballot with no class of
> voters getting an 
> automatic advantage.
> 

But voter A is the strategic voter and has an advantage over voter B, who
has voluntarily diluted his voting strength by voting sincerely. The one
person one vote objection was to allowing voters any fallback position at
all, which IRV also does.

> 
> 3. <snip>
> "Take an example. Louisiana uses the same election
> system that France 
> does,
> and it malfunctioned the same way in both places; a
> crypto-fascist got
> enough votes to make it to the runoff, produced a fair
> amount of panic, 
> and
> duly lost to an opponent whose only real selling point was
> being the only
> alternative."
> <snip>
> 
> But the method is not IRV. In France with sequential
> elimination, all 
> experts agree that le Pen would not have made it into the
> final runoff, 
> and that Jospin would have been the finalist with Chirac.
> Louisiana is a 
> better example, though still weak...since we can't know
> for sure how IRV 
> in a single November election rather than the lower turnout
> October 
> primary Duke passed through, would have changed voter
> turnout and 
> outcomes.
> 

Yes, of course I know it wasn't IRV. That's why I tried to make reasonable
guesses about what might have happened, rather than just repeating what
did. Even though the particular contingent facts of the election might
have gone differently, my overall point stands: there are very reasonable
scenarios where the majority preference for the CW over the IRV winner is
a serious and substantively-based preference, rather than an artifact of
strategic voters' using the CW as a placeholder. Many if not most voters,
if allowed to truncate, will omit unknown candidates altogether.