I don't think that this is a good idea. It's very similar to the Accept Rate, which got shot down because people were voting on a post based on previous behavior of the OP. If someone has posts with an average score of -2, they have a much smaller chance of asking a good question(they'll probably be banned by this point, at any rate), but that does not mean that they are incapable of doing it.
I feel that this will encourage voting on the person for their average scores, rather than voting on the individual posts. Also, quantity is an important factor when determining these things (see Blankman). He has a ton of posts; a lot of them are highly upvoted, but the truth is, the vast majority of them are simply mediocre or bad.
Also, when you say that it will encourage people to improve or delete low-scoring answers, a lot of the time, when people delete low scoring answers, it triggers the post ban. It also may be helpful to keep those answers around, so people know what not to do.
If this is going to be implemented at all, I suggest we ignore the mean and go with the median; it's a much better representation of a person's asking and answering ability, since a few highly upvoted or downvoted posts will throw this completely out of whack. I'm not really familiar with the Data Explorer, but I'm trying to learn it to compose a query for this, but I'm sure someone else could do it a lot more quickly than I could has done it here (Change postTypeId
to 1
for question)
It's not a bad idea, I just fear that it could cause a lot of serial voting sprees.