[EM] SABR voting

[Jameson Quinn]

This is another voting system, very similar to PAR, but hopefully simpler
to describe.

SABR stands for the ratings each voter can give to each candidate:
"Support", "Accept", Blank, or "Reject". The winner is the first candidate,
going in order from most to least "support" ratings, who isn't “strongly
beaten”. To see whether a candidate X is "strongly beaten", give X one
point for each voter who rates them “support” or “accept”, and each other
candidate Y one point for each voter who rates Y above “reject” and at
least as high as X. Any Y who has more points than X "strongly beats" them.
If there’s no candidate who’s not strongly beaten, count all blanks as
"reject" and try again; this guarantees a winner.

This method is monotonic; passes Majority, Mutual Majority, Majority
Condorcet Winner, and Majority Condorcet Loser; does well in center-squeeze
scenarios (by Majority Condorcet Winner); and does well in chicken dilemma
scenarios, without a slippery slope. It obeys a weakened form of
later-no-harm, Later-lower-no-harm (LLNH): adding an “accept” ranking for X
cannot cause a candidate Y you “support” to lose, *unless* X is supported
by more voters than Y.

This method does not meet the favorite betrayal criterion, but I believe
that it has no favorite betrayal incentive in normal scenarios. (I'd like
to find a way to define "normal" so that's rigorous, but I haven't yet got
something I'm ready to share.)

This method does not meet participation and consistency, but it is easy to
define weakened versions of these criteria which it meets.

I'm interested in making a statement of this method that's as clear and
concise as possible, so I'd love to hear any criticism of, suggestions for,
or questions about the definition above.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20161202/3d8579b3/attachment.htm>