[EM] Modeling turkey raising in strategic nomination simulations

[Kristofer Munsterhjelm]

On the subject of strategic nomination, I think that adding turkey 
raising as a category to JGA's two self-interested categories could be 
useful. We could also make the tests more general by considering sets of 
candidates instead of just individual candidates. So the categories 
would be:

- If one or more candidates of the initial set drop out, and the winner 
changes in a way that the candidates who dropped out all prefer to the 
original winner, then the method has a strategic exit incentive in that 
election.

- If one or more candidates of the reserve set enter the election, and 
the winner changes to someone they all prefer, then the method has a 
strategic entry incentive in that election.

- Consider a candidate A in the initial candidate set, as well as a 
subset T of the set of reserve candidates (who can join but haven't 
yet). If, when these candidates T are introduced, the winner changes in 
a way that A prefers, then the method has a turkey raising incentive by 
A in that election.

- The method has a turkey raising incentive in an election if there 
exists such an initial candidate A and turkey set T.

This would be a kind of candidate analog of nonmonotonicity, because the 
presence of A in the initial candidate set should (if the method is 
"monotone") push the outcome towards A precisely to the extent that 
voters support A. Thus the introduction of someone who A doesn't like 
shouldn't draw the outcome *closer* to A.

-km