Someone made a quantitative study about the /r/fatpeoplehate ban recently and their conclusion was that the honeypot theory isn't valid and that banning hateful subs causes a reduction in overall hateful behavior instead of just shuffling it around.