The dupe author should however be required to confirm it is really a dupe, since mistakes have been made before...
Everyone has different criteria for how they apply their vote. Just because a question is closed as a duplicate, that doesn't mean that a user's vote necessarily translates to the other question.
What happens if the closed post got hammered with down-votes for some crazy reason (or because it was a dupe)? Does the new question get saddled with those, too?
And what happens if the closed post is re-opened? Do you take the transferred votes away? Then you have to track original-versus-transferred votes (possibly from multiple locations).
Much craziness that far outweighs any potential benefit.
Closing is more transient than votes are. As you mentioned, mistakes do happen. And on the flip side we can't ever guarantee that the dupe author will return or will confirm it even if it really is a duplicate (be it stubbornness, spite, or ignorance). What about if the author does confirm, but people vote and reopen it anyway? And how would it handle when a question is closed as a duplicate of multiple questions? Or when people who voted on the duplicate already voted on the target(s)?
It's going to be a lot of effort and calculation for a system that I don't see having a large valuable usage.