This is a really interesting question. I don't have an answer. Have you looked for any articles on this yet?
If I have time, I'd like to take a big corpus and make a fun data set, run some regressions, and see what factors turn out to be significant. But I doubt I'll have time for that undertaking anytime soon.