[EM] The Allure of IRV

[Dave Ketchum]

What I read here is that you stumbled into the wrong argument.

IRV and Condorcet equally capably attend to spoilers with few votes:
      IRV sees them as low votes when and if it sees them in first place 
on a ballot.
      C does the numbers on all pairs, and then discards these as losers.

Usually they will pick the same winner from what is left:
      IRV, if and when the proper winner gets a lead, will discard the 
losers, one by one, and zero in on that winner.
      C gets there by doing the numbers on all pairs.

THEREFORE, you ONLY look at cases where they pick different winners to 
decide which you choose to call better.

Even here, if the method you are arguing for calls it close to a tie, you 
go look for a stronger case.

France did not do either IRV or C, so we can only guess at what the 
numbers might have been on ranked ballots, but I will construct a simple 
case to promote C from the following voters:
      20% dedicated ultra conservative (Le Pen) - they do not need ranked 
ballots for they will ONLY vote for their candidate.
      21% dedicated conservative (Chirac) - ditto).  In the real life 
runoff Chirac did much better, but most of these were moderates and 
liberals looking for the least obnoxious, given an unhappy choice.
      About 16% each for 3 moderates and liberals.
      11% only for Joe, a trusted moderate-liberal.  While the above 48% 
vote their current stars in first place, they all vote for Joe in second 
place.

 From this IRV discards Joe first and then the 16%ers.  While Joe turns up 
behind them on each of their ballots, he gets discarded, for IRV never 
sees more than 16% for him at a time.  Thus IRV finishes with Chirac over 
Le Pen.

C, looking at whole ballots, will see 59% for Joe, declare Joe winner, and 
get LOUD CHEERS.

Dave Ketchum

On Wed, 24 Apr 2002 16:55:40 -0700 (PDT) Alex Small wrote:

> I concur with Adam.  Once you support ranking, unless you've heard of
> Condorcet the runoff idea makes intuitive sense, since plenty of places in
> the US use 2-stage runoff.  The question is how to sell Condorcet over IRV.
> 
> My original message was prompted by an argument with a very intelligent
> person who heard of an election in Ireland where IRV happened to find the
> centrist.  She concluded (justifiably) that if France had used IRV instead
> of two-stage runoff the final round would likely have had at least one
> liberal or moderate rather than two conservatives.  A liberal vs. a
> conservative may or may not be as good as a centrist vs. one of those, but
> it offers more freedom of choice than conservative vs. ultra-conservative.
> 
> I tried Hitler-Stalin-Washington on her (not those names, but that idea)
> but she pointed to the Irish example.  She is very intelligent.  However,
> take the intuitive notion of runoffs and combine it with an anecdote of a
> very good result under IRV, and even a very intelligent person like my
> fellow student will be difficult to persuade.
> 
> (It probably doesn't help that I'm a Libertarian and hence I frequently
> clash with her on politics.  If people don't like the messenger the message
> will fall on deaf ears.  Hence it's important for people from different
> third parties to collaborate when selling Approval or Condorcet.)
> 
> I tried summability, thinking that an engineer would appreciate the
> difference between exponential scaling and n^2 scaling.  She didn't care.
> I said that the Condorcet candidate is by definition the one whom the
> electorate prefers.  She said "Well, it seemed to work pretty well in
> Ireland."
> 
> Any thoughts on how to overcome the tag-team combo of IRV's seemingly
> intuitive nature and IRV anecdotes?  I'm sure CVD is collecting such
> stories.
> 
> Alex

-- 
  davek at clarityconnect.com    http://www.clarityconnect.com/webpages3/davek
   Dave Ketchum    108 Halstead Ave, Owego, NY  13827-1708    607-687-5026
              Do to no one what you would not want done to you.
                    If you want peace, work for justice.

----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em