[EM] Approval Runoff

[Forest Simmons]

I would like to make a couple of suggestions relating to Approval Runoff
that address your questions.

1. The natural ballots for Instant Approval Runoff are Dyadic Approval
Ballots. At each stage of the runoff the coarsest remaining inequality is
the current approval cutoff.

Remember that on Dyadic Approval Ballots every category of approval or
disapproval may be further divided by a finer inequality than the one
delineating the category itself, thus providing the approval cutoff that
would apply if all were eliminated from the race except the members of
that category. example: A > B >> C >>> D > E >> F > G >>>> H >> I > J .
The coarsest inequality is the one found between G and H. If H, I , and J
are all eliminated but candidates on both sides of the >>> inequality
between C and D still remain, then that inequality is the current approval
cutoff.

2. However, for most practical purposes all that is needed is a way to
choose pairwise between the two candidates with the best approval scores,
as in the experience you described below. Dyadic Approval ballots are not
needed for this. This is easy to do when either rankings or ratings are
available. 

On either style ballot include MAC (Minimal Acceptable Candidate) as one
of the candidates. The rank or rating received by this virtual candidate
marks the approval cutoff for that ballot. After the approval scores have
been calculated, a head-to-head comparison between the two candidates with
the highest approval is made using the rankings or ratings. 

3. Another use of these ranked or rated ballots with MAC is (my version
of) Demorep's ACLA (Approval Condorcet Least Approval) method. First
eliminate all of the candidates with below 50% approval (unless this
leaves none, in which case proclaim the one with greatest approval to be
the winner). Then among the candidates with more than 50% approval find
the Condorcet winner if there is one. If not, eliminate candidates one by
one on the basis of least approval until there is a Condorcet winner among
the remaining candidates. (In the almost impossible event that it comes
down to two candidates with both head-to-head and Approval ties, then use
random ballot or some other traditional tie breaker.) 

I hope that you can use the ideas behind these methods to customize a
method for your own purposes. I am interested to hear about what you come
up with. 

Forest


On Mon, 30 Apr 2001, Moe St. EverGreen wrote:

> At the Green Party of Florida meeting last weekend,
> the issue of how to handle our officer elections came up.
> 
> The first suggestion was, predictably, to use IRV.
> 
> In fact, we also received several calls to endorse IRV.
> 
> I was encouraged that I was not the only one there aware
> of the problems with IRV.
> 
> We ended up endorsing Preference Voting in general, taking 
> that at its most generic meaning.. that of ranking vote references,
> with the final counting method not specified. The purpose of
> the endorsement being to make sure any new voting machines,
> (as most US people now know Florida needs) would support ranking 
> of ballots.
> 
> Anyway, back to the point, IRV was not accepted by the group.
> 
> Instead, I made a counter proposal that we use Approval Voting.
> 
> Approval voting was accepted, and worked wonderfully for most of 
> our elections (with a large majority of voter satisfaction, it seemed)
> until we came to a tie for spokeperson.
> 
> 2 people tied, with the third candidate having a lesser number of 
> supporters.
> 
> So, I suggested a runoff between the top two.
> 
> That was accepted (I was the third person losing, by the way, so
> perhaps my suggestion was given more credence in that I was
> accepting my defeat).
> 
> And although some people might have predicted otherwise,
> the removal of a person (and possibly further reflection on
> the fact that there was a tie), resulted in a different amount 
> of votes for the two people after the runoff, again with a high
> majority of voter satisfaction.
> 
> In this case, I think that a random determination had a high
> chance of being perceived as unsatisfactory, and similarly
> a chair breaking the tie would have not been acceptable.
> 
> However, I'm not really sure the runoff needed to have eliminated
> a person. Since we don't use secret ballot, and actually count 
> the people voting, and allow people to caucus and discuss for
> two minutes before voting, it seemed that rather than the elimination
> changing the dynamic, politicking changed the dynamic. 
> Some people were likely convinced by others to absent, 
> because the winner's votes only increased by 2, while the other 
> person's votes decreased by 10.
> 
> Anyway, I though this story might interest those searching for
> hybrid systems. 
> 
> It seems Approval makes the use of a serious runoffs slightly more 
> acceptable.
> 
> Runoffs are usually a serious of pluralities, having all the problems of IRV.
> 
> A series of Approval Runoffs which might be seen as an extension of
> what we used for a tie breaker, seems to have more merit for a meeting
> than a series of Plurality Runoffs as described in Roberts.
> 
> Especially when the vote is not secret and here are not too many
> people (or there is some good voting mechanism) to easily count 
> them quicky.
> 
> I'm can not see how, though, a ballot could be designed
> for any kind of useful Instant Approval Runoff.
> 
> In that circumstance, I hope to find a way to describe
> something like Cloneproof SSD that would be acceptable 
> for an instant balloting for use with expensive one-time mail ballots.
> 
> IRV has a main disadvantage (and one big advantage
> for its opponents) in that it would be much more expensive to 
> implement than most Condorcet methods.
> 
> However, all of the definitions of cloneproof SSD that I have seen
> are very difficult to describe, and often don't have very distinguishable
> examples that show various aspects of the process.
> 
> - Moe.
> 
>