[Election-Methods] RELEASE: Instant Runoff Voting

[Aaron Armitage]

--- On Mon, 7/28/08, Chris Benham <cbenhamau at yahoo.com.au> wrote:

> From: Chris Benham <cbenhamau at yahoo.com.au>
> Subject: Re:RELEASE: Instant Runoff Voting
> To: "EM" <election-methods at lists.electorama.com>
> Cc: eutychus_slept at yahoo.com
> Date: Monday, July 28, 2008, 9:57 AM
> Aaron,
> "In an important respect, Condorcet is more natural
> than IRV: if a majority prefers Brad over Carter, this
> preference exists whether the voting system does anything
> with it, or even elicits enough information to determine
> that it exists. "
> Yes, except that "Condorcet" is a criterion and 
> IRV is a method, and  "more natural" doesn't
> have a precise meaning.
> 

I think the meaning of "more natural" was sufficiently clarified by my
explanation, namely that Condorcet methods are based on a property that
exists in the preferences themselves, rather than being an artifact of the
counting rules. Which is why there can be such a thing as a Condorcet
criterion; it refers to a real correspondence between the rankings
provided and the result produced.

> "Condorcet simply discovers and applies this
> preference. IRV, on the other hand, elicits enough
> information to discover it exists, but may decide to ignore
> it based purely on procedural grounds. There are no good
> reasons for this, ever."
> IRV meets Later-no-Harm and  Later-no-Help and  is immune
> to Burial strategy, and these properties are incompatible
> with the Condorcet criterion.
> Some people think these "reasons" are
> "good". 

I know some people claim that there are good reasons to use non-Condorcet
methods. I claim that when it gets down to installing a candidate over a
candidate who beat him, which all non-Condorcet methods do sometimes (or
else they'd be Condorcet methods), those reasons aren't good enough.

Suppose my preference is Andrea > Brad > Carter, and Brad > Carter is a
majority preference, but Andrea > Brad and Andrea > Carter isn't. Under
IRV, I am sometimes (but not always) forced to hold on to Andrea
until I no longer have any chance of helping Brad beat Carter, even if my
preferences are more like Andrea > Brad >>>>> Carter. This quality strikes 
me as perverse. Nevertheless, Condorcet's failure of LNH is connected to
a serious weakness, vulnerability to burial. But this means that when my
later preferences do "hurt" my favorite, things are as they should be 
because it means I didn't create an artificial cycle.

> BTW, which of the many methods that meet the Condorcet
> criterion is your favourite? 
> Chris Benham
> 

You're really asking which completion method is my favorite, and I don't
have any strong preferences. I believe an actual runoff for the Smith set
would be the most appropriate: it would give voters a chance to reconsider
with their attention focused on only the serious candidates, and make
strategy harder to use. If, for example, the cycle existed because a large
number of ballots were Republican > Green > Democrat, the Democrats will
have an opportunity to use counter-strategy, or even make the apparent
manipulation a campaign issue. It's not really reasonable to have a third
or later vote, so some reasonable completion method will need to be used.