[Election-Methods] Improved Approval Runoff

[Abd ul-Rahman Lomax]

At 10:48 PM 8/20/2007, Kevin Venzke wrote:
>That's still pretty strange... What about IRV with equal rankings allowed?

Well, I suggested it long ago as a simple improvement. Voters can 
essentially vote it as Approval if they want.

In an Approval election, if all the candidates you approve are not 
going to win, your vote has been wasted. (Unless, of course, it is 
Asset Voting or some PR scheme). Having a fallback vote makes sense.

However, I wouldn't personally choose IRV as the ranked method to 
use. Why in the world not use a Condorcet method, if you want ranked?

The trick that I've proposed to make Range methods MC compliant could 
also be used with IRV. Let the IRV election play out, then reanalyze 
the ballots fully and see if anyone beats the IRV winner pairwise. 
Since you need only compare the pairs including the IRV winner, the 
counting is simplified. This would detect the Condorcet winner 
(though it might detect more than one candidate beating the IRV 
winner -- but that should be rare, since IRV does usually pick the 
Condorcet winner unless there are a lot of candidates.)

You could either award the victory to the one who beats the pairwise 
winner -- in some fashion -- or hold a runoff. A real runoff is the 
ultimate challenge, the proof that the winner is acceptable to a 
majority, at least comparatively. (A true test would be a pure Yes/No 
ratification.)