[EM] Completion methods for Smith Sets

[Michael Rouse]

While I was trying to figure out ways to use Approval voting in Smith sets 
as a completion method, I came up with a completion method that has 
probably already been independently discovered half a dozen times or so 
(grin). Perhaps someone will recognize it by the following longwinded 
description:

As a completion method for a single-member election: Choose the Condorcet 
winner, if one exists. Otherwise, find the Smith set (which I usually see 
defined as the smallest set where none of the candidates outside the set 
beat any of the candidates within the set in a pairwise contest). Throw out 
all candidates not in the Smith set.

Once you have the candidates in the Smith set, rank them, starting with the 
candidate that has the highest number of first place votes (the plurality 
winner), then the winner of the combined first and second place votes, then 
the winner of the first, second, and third place votes together, continuing 
until you have one fewer candidate on the list than you started with. Don't 
count last place rankings; if there are remaining places on the "winner" 
list, start with the second candidate on the "plurality" list and continue 
down the line. Ties are ok unless there is one spot remaining; in that 
case, choose the pairwise winner between the two candidates. Throw out the 
loser, adjust the rankings, and begin the process again. Continue until you 
have one candidate.

For a multi-candidate Condorcet completion method, find the largest inner 
unbeaten set and the smallest unbeaten outer set that encloses the number 
of positions available -- in other words, C<=N for the inner set (where C 
is the number of candidates, and N is the number of positions in the winner 
set) and C>=N for the outer set.

If the inner and outer set are the same, those candidates are elected, and 
we are done. If they are different, the ones in the inner set are 
automatically elected, and the ones between the inner and outer set are run 
through the single-member method above, dropping one candidate at a time 
until the number of candidates is reduced to its final number.

For now I won't worry about voting problems with truncated votes or tied 
preferences. I really just wanted to see if someone knew what this method 
is called. If anyone knows its name and what benefits/drawbacks it might 
have, please let me know. I know it's probably not monotonic -- if someone 
has an example, that would be great as well, as would any other info. Thanks!

Mike Rouse
mrouse at cdsnet.net

PS I'll write a bit more in another post where this is described as a 
full-fledged method, and not just as a completion method as above. My 
apologies to IRVing and Borda supporters, but I feel a voting method should 
choose the Condorcet winner if it exists, and should choose from the Smith 
set if it does not. If a method like Nanson's does so, maybe this one will 
as well; if not, then maybe it can serve as a completion method.