Here is my paraphrase of an argument that seems to be popular in some circles on days like this:
In an election involving two unsatisfactory candidates one of whom is morally certain to win (for the purposes of this post, call them the two major party candidates), not to vote for one of the two candidates is equivalent to voting for the less satisfactory one.
But voting for the less satisfactory one is wrong.
Therefore, one ought to vote for one of the two candidates morally certain to win.
And voting for the less satisfactory one is wrong.
So one ought to vote for the less unsatisfactory of the two candidates morally certain to win.
The problem with this perfectly valid argument is that its first premise is false. Not to vote for one of the two candidates is not equivalent to voting for the less satisfactory one.
You can see where the premise comes from, of course. If one were to vote for one of the two major party candidates, one would vote for the less unsatisfactory one. Therefore, the reasoning goes, not voting for one of the two major party candidates "costs" the less unsatisfactory one a vote, which is numerically equivalent to giving the less satisfactory one a vote.
There are three problems with this. First, the voter is probably well aware of the numerical implications of his vote. If his decision to not vote for one of the two major party candidates already accounts for the purely conceptual (because it is relative to no existing reality) cost to the less unsatisfactory candidate, then pointing this out to him won't affect his decision.
Second, given that one is not going to vote for one of the two major party candidates, it doesn't matter what follows from the supposition that one is. The negation of a true statement implies anything you like. P->((~P)->Q) for any P and Q.
Third, a certain numerical equivalence (in this case, the difference in vote count between the two major party candidates) by no means implies an overall prudential, much less moral, equivalence. Of all the effects, public and private, of a single vote, perhaps the least significant is that on the margin of victory. To treat the margin of victory as the most (some would seem to have it the only) significant factor in determining how to vote is to misunderstand the act of voting.