Several commenters have noted below that all the cool moral theologians say you can vote for a bad candidate against a worse candidate. That's significant for a probabiliorist like me, so let me poke at this idea a bit.
What pops out first is the diciness of the colloquial way of expressing the idea -- viz., "to vote for the lesser of two evils." Now, you can never licitly choose to do evil yourself, so clearly moral theologians discern some moral distance between the act of voting for a "bad" candidate and the bad acts the candidate would commit if and only if he were elected. Moral theologian Kevin Miller puts it this way:
...the evil act is voting for the candidate qua evil. But no one is absolutely evil. If you're voting for the candidate qua better than the alternative, then you're not voting for him qua evil.
Here I'll grant that voting for a candidate qua better than the alternative is possible. Once that's granted -- or more precisely, once it's granted that voting for a candidate is not necessarily voting for a candidate's bad positions – we can dust off the Principle of Double Effect apparatus.
By the Principle of Double Effect, an act is permissible, despite having a morally certain bad effect, if a) the act itself is not immoral per se; b) the bad effect is not intended; c) the bad effect is not the means to the intended good effect; and d) the good effect outweighs the bad effect.
On the question of voting for a bad candidate to prevent the election of a worse candidate, I've just granted that a) the act itself is not immoral per se. Is the bad effect intended? It is if the bad effect is the election of the bad candidate, so we should cast the bad effect as the implementation of the bad candidate's bad policies. (This will be important for understanding what the moral act of voting really is.) This bad effect is not intended, which we can see because we would be perfectly happy if it never occurred, so (b) is satisfied. The implementation of the bad candidate's bad policies is not the means of preventing the worse candidate's worse policies, so we're fine with (c).
Now, does the good effect outweigh the bad effect? Here, I think, we need to be careful.
The bad effect, as I've said, is the implementation of the bad candidate's bad policies; the good effect is the prevention of the implementation of the worse candidate's worse policies. But we can't simply subtract the one from the other and say, "The net result is less badness with the bad candidate than with the worse candidate." The Principle of Double Effect doesn't speak about net results, it speaks about effects.
If we return to my example of the Dystopian Party candidate vs. the Atheist Dystopian Party candidate: the bad effect of voting for the former is the execution and consumption of everyone over 30; the good effect is those accused of possessing religious faith are not executed and consumed until they reach 30. While it's clear the Dystopian Party candidate is better, it's not so clear that the good effect of his election outweighs the bad. (To remove all lack of clarity, we could make the second choice an Anti-Parrot Dystopian Party, identical to the Dystopian Party except it also calls for the extermination of parrots in the wild.)
So while it may be possible to vote for a bad candidate over a worse candidate, a straightforward application of the Principle of Double Effect doesn't suggest it's always permissible.
Corners cut in this post include a failure to address what it means for a candidate to be "bad" in a morally significant sense and a failure to observe that the election of a candidate is not a sufficient cause of the implementation of his policies.