[EM] Simple illustration of center-squeeze effect in runoff voting

[Dave Ketchum]

Here it is noted that IRV has a black mark for failing to correctly award W 
as deserving winner.  They seem not to notice that IRV's failure is also 
describable as incorrectly discarding W as an undeserving loser.

As to escaping two party domination, think on:

Plurality:  If I prefer one of their candidates over the other, I must vote 
my preference between them, and wait til next time to think of voting for a 
minor party candidate.

Approval:  Here I can vote for both a major and a minor, but must vote as 
if equally desiring the barely tolerable major over the much better minor.

IRV:  See above.

Condorcet:  Can use IRV ballots and voting, but Condorcet promises to read 
all that I vote on them.  Further, its N*N array is a useful record as to 
relative strength of candidates/parties.

And "center-squeeze effect" or "Center-pull"?  Makes sense if there is only 
one issue of interest for an election.  Makes less sense when, as usually 
happens, there are multiple important issues with each major party doing 
better at satisfying each voter on part of the issues.

DWK

On Tue, 20 Jan 2009 05:10:50 -0000 baring001 wrote:
> Tom,
> 
> The problem is that the center-squeeze effect can easily keep a center
> candidate from reaching 20 percent of the vote.
> 
> Of course the more basic problem is that IRV uses plurality to decide
> which candidates to eliminate.  If plurality can't be trusted to pick
> the winner, why would we expect it to reliably pick losers so that we
> can somehow "back into" finding a winner?
> 
> In over 10 years of following election reform, I still haven't seen a
> good argument for anything more complicated than approval voting.
> 
> If you must insist on elimination rounds, the best method would be to
> conduct an approval vote and then during the vote tally iteratively
> eliminate candidates with the fewest approvals.
> 
> Short of that, plurality could be called a "good enough" system.  It's
> just that if you're not happy with the two parties' nominees, your
> best recourse is to join one party and attempt to change it from within.  
> 
> 
> --- In instantrunoff-freewheeling at yahoogroups.com, "Tom Ruen"
> <tomruen at ...> wrote:
> 
>>Baring01,
>>
>>I'd support using IRV as an "elimination process" rather than picking a 
>>winner. That is, eliminate one candidate at a time from the bottom
> 
> until all 
> 
>>remaining candidates have at least 20% of the vote. THEN see if a
> 
> Condorcet 
> 
>>winner exists. If none, then randomly pick a winner from the Smith
> 
> Set (The 
> 
>>smallest set of candidates who can defeat all candidates outside the
> 
> set 
> 
>>head-to-head.)
>>
>>I think this handles the "Center-squeeze" you describe AND handles an 
>>opposite problem of a perhaps called "Center-pull" of a pure Condorcet 
>>method where a center candidate gains power merely by being cautiously 
>>inoffense to all sides.
>>
>>Well, this IRV-Condorcet Hybrid system is my favorite, although in
> 
> practice, 
> 
>>I continue to believe Plurality is the real enemy and its silly to
> 
> argue 
> 
>>about minor differences in majority systems.
>>
>>IRV is "good enough" in practice, and perhaps best, encouraging
> 
> candidates 
> 
>>and voters to "agree to agree" before the election (like the
> 
> plurality) due 
> 
>>to spoiler fears, and that pre-election organization is what helps
> 
> keep the 
> 
>>number of candidates managable for voters!
>>
>>Plurality rewards a "two party system", and IRV/runoffs rewards "the
> 
> two 
> 
>>strongest candidates" who may not always be from the same two may
> 
> parties in 
> 
>>different elections.
>>
>>Tom Ruen
>>
>>----- Original Message ----- 
>>From: "baring001" <baring001 at ...>
>>To: <instantrunoff-freewheeling at yahoogroups.com>
>>Sent: Thursday, January 08, 2009 10:22 AM
>>Subject: [IRV-freewheeling] Simple illustration of center-squeeze
> 
> effect in 
> 
>>runoff voting
>>
>>
>>
>>>For a simple illustration of the center-squeeze effect, imagine a
>>>three-way race in which the electorate is uniformly distributed across
>>>some policy continuum. Candidates A, B, and C are positioned at the
>>>top edge of the 25th, 50th, and 75th percentiles, respectively.
>>>
>>>If the voters rank their choices in order of closeness on this
>>>continuum, then voters from 1st to 37th percentiles will vote for A,
>>>voters from the 38th to 62nd percentiles will vote for B, and 63rd to
>>>100th will vote for C. The first round totals (as percentages) will be
>>>37.5 - 25 - 37.5 and the centrist, B, will be eliminated first by a
>>>margin of 12.5 percent.
>>>
>>>If you assume a normal instead of uniform distribution of voters, with
>>>candidates again at the top of the 25th, 50th, and 75th percentiles,
>>>the results are nearly the same except the centrist is eliminated by a
>>>margin of approximately 10.7 percent instead of 12.5.
>>>
>>>It makes surprisingly little difference whether the voter distribution
>>>curve is flat or bell-shaped. Even though the voters move inward with
>>>a normal distribution, the location of the 25th and 75th percentiles
>>>move inward as well (with a z-score of plus or minus 0.67 s.d.)
>>>
>>>In either case, so long as the two wing candidates position themselves
>>>anywhere between the 25th and 76th percentiles, there is no way for a
>>>middle candidate to avoid being eliminated first. The best he can do
>>>is try to mimic one of the wing candidates and hope for a
>>
> first-round tie.
> 
>>>Of course, as the wing candidates move toward the center, the centrist
>>>loses by an even larger margin. Hence the term "center squeeze
>>
> effect".
> 
>>>Of course this changes if the candidates are skewed around more than
>>>one policy dimension.  Obviously where there is no centrist, there can
>>>be no center-squeeze effect.
>>>
-- 
  davek at clarityconnect.com    people.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.