Folks may not agree with my particular conclusions about voting in the upcoming Presidential election, but maybe we can make some progress in mutual understanding of what constitutes a proportionate reason for engaging in remote material cooperation with evil. Inspired by an inquiring commenter, I give you the following:
Suppose we are contemplating doing act X in order to block a big evil E, where X is not intrinsically evil but doing it involves remote material cooperation with evil.
A proportionate reason to do X obtains when (1) X is reasonably effective in stopping E without being excessive, and (2) stopping E does not produce evils and disorders graver than E.
Folks tend to make a reasonable case for (2): that is, they make a reasonable case (lets stipulate, in case you disagree) that McCain winning does not produce evils and disorders graver than those which would follow from Obama winning.
But there is a very strong tendency to ignore (1) completely, treating an act of voting as if it were precisely the same thing as making McCain win by fiat. That isn't the kind of thing that voting is though: it has very little actual efficacy in making one's favored candidate win, and yet it has quite a bit of efficacy in exercising influence over the person who votes himself and those within his immediate sphere of influence.
So whether or not there is a proportionate reason to vote for a candidate depends on understanding not only the outcome dependent results of the act (that is, results which follow from McCain winning over Obama), but also the act's outcome independent results, as well as their relative importance.
Their relative importance will in turn depend on how much influence the act actually has over the outcome: if the act has very little influence over the outcome, then its outcome-independent effects will dominate moral evaluation of the act. This is the part that people find very counterintuitive, because we think of voting as just being about the outcome. Nevertheless, because the efficacy of a vote in determining the outcome of a national election is so infinitesimal, a proper moral evaluation of it is going to be completely dominated by its outcome independent effects.