Dave Ketchum wrote: > I find random ballots acceptable for resolving true ties, > assuming the authority conducting the election agrees. I do not find > them acceptable as an excuse for not doing what is possible with > Condorcet vote counts.
Even in the case of a true tie, or a race so incredibly close that the authorities cannot declare a winner with any reasonable certainty (e.g. a margin of less than 100 votes out of a few million, or something like that), I see no need for random selection. One possible method is to work from pre-existing districts of equal population (e.g. state legislative districts in a gubernatorial race) and see which of the tied candidates wins in each district (use a pairwise comparison in each district for 2-way ties, for ranked methods, and straight vote counts for methods like Approval). The method is deterministic, and based only on the votes cast on election day, not the way a coin falls or whatever. The rationale for using the districts is spotty, but no spottier than the rationale for flipping the coin. Alex ---- For more information about this list (subscribe, unsubscribe, FAQ, etc), please see http://www.eskimo.com/~robla/em
