[EM] Combining election methods

[Michael Rouse]

I've been looking at many of the election methods described on the web 
(thanks, Google! :)  and I haven't seen any versions that calculate the 
winner two different ways and compare the result, using a third method as a 
tie-breaker if necessary.

For example, we could use method A to figure out an election winner A1, and 
then method B to choose B1. If the results match, then that person is the 
winner. If each method chooses someone different, then use a third method C 
to choose between A1 and B1.

If we were to use a ranked ballot and were to add a cutoff line voters 
could choose between approved and unapproved candidates, we could use both 
Approval and Borda to look at the voting data. If they select the same 
winner, great; if not, the top choice  from each method gets compared 
directly just like it was a two-person race. Another possibility would be 
to take the Smith set of the data first. If there were a circular tie, run 
the Smith set through Approval and Borda, and if there was still a tie do a 
plurality test on the two remaining candidates (the process would go from X 
number of candidates to 3 to 2 to 1 in that case).

Of course, you could choose your own voting methods to use in comparison -- 
it might be interesting to match strengths and weaknesses between methods. 
Admittedly, calculating the winner would be much more difficult than simple 
plurality, but that's what computers are for.  Since this method sorts 
candidates (allowing for ties in ordering, if desired) and divides the 
group between acceptable and unacceptable candidates, it is only a tiny bit 
more complicated than straight Borda.

Mike Rouse
mrouse at cdsnet.net