[EM] RE: [Condorcet] A "Condorcet" by any other name still smells as sweet?

[Abd ulRahman Lomax]

At 06:46 PM 9/30/2005, Simmons, Forest wrote:

>Approval is unacceptable to voter psychology because once you 
>approve Compromise, you erase the ballot distinction between 
>Favorite and Compromise.  That's why ordinary Approval is not a 
>viable ballot proposal.

This is the main objection I have seen to Approval, other than the 
phoney one that Approval allegedly gives one voter more than one vote.

>What does that leave?
>
>1. Creative use of simple ballots to get voter rankings, as (for 
>example) in Jobst's DFC proposal.
>
>2.  Ranked or not, ballots have to allow distinction of Favorite 
>without strong incentive to rate or rank Compromise with the 
>distinctive mark along with (or in place of) Favorite.  This means 
>that the distinctive mark has to be mostly expressive or symbolic, 
>with very weak instrumentality, such as having a role in breaking ties.
>
>After Asset Voting, Approval Plus (A+) is the simplest such a proposal.

Depending on how it is implemented, A+ is actually a Condorcet 
method.... But Forest does not go there with it. The Condorcet 
variation I've called A+PW, PW standing for pairwise.

>Approval Plus  is Approval with the ability to distinguish one 
>candidate with a plus.

It is Approval "plus" specification of a Favorite. The word 
"Approval" does cause some problems. Some people think that it is 
insincere to "Approve" a candidate who you can barely stand, merely 
because he or she is a frontrunner and is not as bad as the other. So 
I've switched the names to

Favorite, Preferred, [blank = Not Preferred]

There is no reason to prohibit the designation of more than one as 
Favorite, but most voters won't do it. The reason *not* to prohibit 
it is simply that it does no harm.

>Along with the approval tally, there is a tally of each candidate's plusses.
>
>When the max approval candidate is unique, that candidate is elected.

The max approval candidate will so rarely not be unique that we could 
practically neglect the possibility and leave the resolution to 
existing law, which might be a coin toss. However, one could use the 
Favorite information to resolve a tie, and I see no reason to resort to:

>When two or more candidates are tied for max approval, then the plus 
>tally for each candidate is used to determine how many marbles will 
>correspond to that candidate in an urn full of different colored 
>marbles (one color for each candidate).
>
>A marble is drawn.  The candidate owning the drawn marble picks the 
>winner from among the tied candidates.

However, the Favorite (Plus) information could be actually used, in 
more ways than one. First of all, I've never seen Approval advocates 
(other than myself) mention it, but Approval has a problem when mated 
with public campaign finance laws. What happens when you approve two 
candidates? If you are a Nader voter, is the public funding resulting 
from your Green vote split with the Democrats?

However, Plus solves this problem. Campaign financing would be 
determined by the Plus votes. Plus also solves the psychological 
problem of being unable to specify a favorite.

However, in basic A+, one problem remains. When there is a two-party 
system and the chances of a third party winning are vanishingly 
small, basic Approval solves the spoiler problem (as does any 
Condorcet method). However, as a third party approaches parity, the 
spoiler problem will loom again. If there are three parties, about 
equal in voting strength as seen in polls going into the election, a 
voter who has a strong preference will have a difficult choice: vote 
for only the Favorite and thus abstain from the pairwise election 
between the two other candidates, or vote for the Favorite and a 
Preferred, and abstain from *that* pairwise election.

Plus suggests a possible path: Use the Favorite information in the 
pairwise contest between any Favorite and Preferred candidate. This 
must be done ballot-wise, it can't just be totals, I think. 
Essentially, this is Condorcet, with only three ranks allowed.

Because it is a Condorcet method, it should be vulnerable to cycles. 
But, by its nature, it has clear Approval information, which can be 
used to resolve cycles. I think that the DMC method could be used, 
but I prefer simply awarding the election to the Approval winner 
among the members of a cycle.

I have *not* examined all the implications of this proposal. There 
have been other attempts to suggest methods using a similar ballot, 
but I have not yet found an existing proposal that analyzes the 
ballots in the same way. It seems simple to me, and obvious, but I 
didn't think of it until a few days ago, and it *may* be that somehow 
this procedure was overlooked.

It is an Approval method. And it is a Condorcet method, but with 
limited ranking.

I think it is worthy of examination, and this is happening to some 
degree on the EM list. As I have written many times, I don't consider 
myself an expert.....

I have come to realize this, though: much of the concern over 
strategic voting may be misplaced. Strategic voting, by definition, 
risks a failure to elect a preferred candidate just as much as it may 
result in a supposedly better outcome. Strategic voting can be 
employed by more than one group. I think that it would actually be 
rare that an attempt would even be made. So simply noting that a 
method is vulnerable to some strategy may not be fatal to the method. 
I'm much more concerned about the possibility that, with sincere 
voting, Condorcet failure may occur *without* the compensating fact 
that the Approval winner was elected.