[EM] 2-step CR: A proposal

[Tarr, Adam]

I sent this out yesterday, but I forgot that my send address had changed.
Joe Weinstein made a lot of the same points I make below.

Alex wrote:

> This seems to retain the advantages of 3-level approval, but 
> gives voters more flexibility.  There is a slightly stronger 
> incentive to insincerely rate somebody equal to favorite, but 
> using a wide scale should substantially mitigate that.

It should be better, but in a close election the distortion caused by
additional candidates still matters.  What if your second-favorite candidate
loses to your least favorite because you (and many others like you) rate the
compromise candidate a 9 or 8?  The best strategy would generally still be
to give every candidate a 10 or a 0.

If we want to make a ranked election that plays like approval, then we
should use the "majority choice approval" or "Bucklin done right" or
whatever it's being called.  It's a really good idea, and moreover there's
no reason we can't expand it out from 3 slots to 10 or however many we like.
So the system would be:

- Voters can rank candidates in first place, second place, and on down, as
far as they like.  Tied rankings are permitted.

Counting procedure:

1) Count the first place votes.  If any candidate has a majority, the
candidate with the most votes is the winner.

2) If no candidate has a majority, then count the first and second place
votes for each candidate.  If any candidate has a majority, the candidate
with the most votes is the winner.

3)  Repeat 2) until a winner is declared or all votes have been counted.  If
all votes are counted, the candidate with the most votes wins (or
alternatively, have some other election a-la Demorep).

In the end, this system plays out a lot like Approval voting, only the
multiple levels give the voters a chance to hedge their bets a bit in a
close multi-candidate race.

-Adam

----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em