[EM] Addendum to approval voting

[Abd ul-Rahman Lomax]

At 09:58 PM 4/28/2006, Antonio Oneala wrote:
>There could be an option on the ballot, that a voter may or may not 
>mark, to choos one of the candidates a "favorite".  Those with 
>"favorite" vote would then be empowered to unrank any of the 
>candidates that the person had ranked.  The obvious situation in 
>which this would work is when believe that you may be hurting your 
>favorite by ranking one of the other candidates; therefore, if your 
>vote caused them to lose, they can unrank the votes that caused them 
>to lose and win.

I don't think this was stated very clearly. I think what is being 
proposed is what I called "A+PW," which is A+ voting -- Approval Plus 
the indication of a favorite, counted PairWise.

It is not commonly realized that standard plurality with no 
overvoting is a Condorcet-Compliant method. So is standard Approval. 
The only problem is that there are only, in these, two ranks: Marked 
and Not-Marked, or Approved and Not-Approved.

A+ is counted, for the purposes of determining the winner, as 
standard Approval. The extra mark -- or marks, there is not reason to 
prohibit marking more than one Favorite -- are used only for 
informational purposes, and, most notably, for public campaign 
financing where this is an issue. Without this, Approval is problematic.

A+PW is, quite simply, a three-rank method, and, as such, could be 
counted with any of many different methods; but presumably they would 
be either IRV or Condorcet-compliant.

Full Condorcet implies as many ranks as candidates. This is probably 
more ballot complication than necessary. So many practical Condorcet 
proposals do limit the number of ranks (with unmarked candidates 
presumably being considered equally approved for the implicit bottom 
rank). The only question is how much to limit it.

Three ranks, i.e., Favorite, Also Approved, and [Rejected] probably 
will deliver 95% of the Condorcet buck. And be much easier to count, 
even with limited lever machines.

>This addendum will make approval more likely to pick a Condercet winner.

Approval is a Condorcet-compliant method, as is no-overvoting 
plurality. They only seem not to be by assuming voter preferences 
that are not stated on the ballot.... No method can use unstated 
information! Of course, a simple Approval ballot does not allow more 
than two ranks.... the proposal here adds one rank, that's all.

>   It won't simply reduce to plurality because now if your favorite 
> doesn't win, they can simply refuse to withdraw votes from a more 
> moderate candidate on your ballot, thereby denying the other guy a 
> victory that he would've gotten under plurality

I'm not sure what is meant here, except perhaps it is an indication 
that I did not understand the proposal. Is this some kind of Asset 
Voting proposal, where a candidate receiving votes may -- or may not 
-- transfer those votes?

If so, the reader should note FAAV, another method that would 
probably be superior to about anything else on the table. Fractional 
Approval Asset Voting is a variation on Warren Smith's Asset Voting, 
where, instead of voters assigning a fraction to as many candidates 
as they like, the total adding up to one -- only a mathematician! -- 
voters vote Approval style, and the votes are divided among the 
candidates chosen. In standard Approval, this would dilute the impact 
of that voter's vote, but in Asset Voting, candidates receiving votes 
may recast them (and elected candidates in PR elections receiving 
more than the quota may redistribute the excess) to create winners. 
So you can vote for one, effectively either making that candidate win 
or making him or her into a proxy for continued voting to determine 
the winner, or you can vote for more than one, creating an effective 
electoral committee. All very simple to count.

This is an example of what is really a deliberative method, turning 
all candidates receiving votes into a kind of electoral college....

>What do you think?  Is it a good idea?

Sure. Just not, I think, a new one. Doesn't mean that it's not worth 
discussing!

Approval Voting is probably the number one simplest possible reform, 
all it takes is deleting the lines of election code that toss 
overvoted ballots. But it should probably have a Favorite designation 
added to it, if only for purposes of campaign finance. And once that 
is on the ballot, all it takes is a change of counting method to make 
the method more deeply ranked, i.e., to actually rank Favorite over 
Approved. And then, if it turns out to be needed, it would be a small 
step to add another rank or even more.