[EM] Can we come to consensus? this way?

[Abd ulRahman Lomax]

At 12:24 AM 9/10/2005, Dave Ketchum wrote:
>Approval gets mentioned so often that I comment up front:
>      Approval as the method.  Simple, but a loser because I too 
> often come up with something like:  I WANT Nader, but I cannot 
> tolerate Bush - so far, so good - But, Nader is not a likely winner 
> so I WANT to show liking Kerry less than Nader but more than 
> Bush.  With this common desire I am ready to reject Approval as an 
> acceptable method.

Doesn't it depend on your options? If the option is the status quo or 
Approval, would you still reject Approval?

Approval has the singular advantage of requiring no ballot changes, 
only a tweak of the election rules: simply stop discarding overvoted ballots.

As to what you want, i.e., to indicate preference for Nader over 
Kerry, would you be consoled by the fact that Nader would, under 
Approval, likewise share in increased votes from those who preferred 
Kerry but thought Nader acceptable? Approval would be, for third 
parties, a *drastic* improvement in the situation. Would Condorcet be 
better? Maybe. Maybe not. Condorcet would indeed allow you to express 
your preference, but in simple Condorcet there is no indication of 
approval and no indication of degree of separation of preference. A 
small preference may have the same vote pattern as a drastic 
difference, between love and abhorrence. Besides, what is the purpose 
of elections?

It is not to express your preference, per se, it is to select an 
officeholder. Preference, approval, ratings, are *means* which may be 
used in this process. What qualifies an officeholder?

As I have written here and elsewhere, there are two philosophies 
about this: one is that the candidate with the broadest approval 
should be elected, the other is that the most popular candidate 
should be elected. Note that in a polarized multiparty election, the 
"most popular candidate," a Condorcet winner, might actually be 
preferred by much less than a majority of voters. It is easiest to 
see with truncated votes:

30: A
20: B
20: C
20: D
10: E

Under Condorcet rules, A wins, with 30 pairwise victories over all 
other candidates.

I'm not arguing against Condorcet, just pointing out how different 
the two philosophies are. Were the same election run Approval, A 
would also win; Approval voting is more likely to encourage multiple 
votes; but we really don't know, there is not very much experience on 
which to base predictions.

>      Approval to resolve Condorcet cycles.  Worth considering, but 
> I question the explaining, the doing, and the counting being worth the pain.

Condorcet is complicated to count. Once you are doing that, Approval 
cutoffs add very little complication, and greatly increase the 
ranking information. (It is equivalent in complexity to adding one 
candidate to the ballot.) A>>B>C>D has a meaning quite a bit 
different from A>B>C>>D. Elections are not only used to elect 
officials, but they also indicate the degree of support an official 
enjoys, which should theoretically encourage a candidate elected with 
narrow support to tread more lightly than one elected with a true 
mandate. Approval, even if it does not determine the winner except in 
the presence of a Condorcet cycle, probably a rare occurrence.


>Each voter ranks candidates, from best to worst.  Ranking two candidates
>as  equally liked is permitted.  Truncation is permitted - acceptable to
>omit the least liked candidates as equally disliked.

Note that if equal ranking is permitted, the method is an Approval 
method. Approval is a ranked method where only two ranks are allowed: 
Yes or the implied No. (Some Approval forms explicitly have Yes or No 
votes for each candidate, but these lose the utter simplicity of 
basic Approval, and then raise the spectre of what one does with the 
blanks, a matter which seems to be of considerable controversy.)

>      I insist on permitting truncation because forcing voters to go 
> beyond their desires gets noise rather than information - when some 
> theorist demands that voters study rejects in more detail, I 
> recommend more effort in sorting out which possible winner is more attractive.

Not permitting truncation would involve considering ballots as 
spoiled which are not complete, I think I remember reading that this 
is actually done in some countries. Personally, I find it just as 
offensive as spoiling ballots because the voter marked too many 
candidates.... Definitely, truncation should be allowed, and should 
have a simple and rational meaning.

There are two possible meanings: truncation on a ranked ballot means 
that the voter ranks the candidate below all ranked candidates, and 
equally with all other unranked candidates. If it is an Approval 
method, an unmarked candidate would similarly be considered not approved.

The other meaning possible would be that truncation is an abstention 
in every pairwise consideration of the unranked candidate. The 
consequences and implications of this are, however, problematic, and 
I think voters would not expect this. Presently, not marking a 
candidate is effectively a vote against that candidate (as long as 
the voter votes for at least one). Turning that into an abstention 
would be confusing.