This has been making the rounds. Based on 150,000 geocoded tweets from June 2012 to April 2013, filtered 1st by use of word, and then manually whether it was used in a negative or derogatory fashion.
Obviously this suffers from selection bias as it only includes people who bother to tweet, and those who aren’t ashamed to do it publicly. There’s also the usual population density distortion (last map below), which would be compounded by cell phone coverage out west. So, basically, this is another pretty visualization of social media meta data that doesn’t really mean much of anything. To be honest I’m surprised they only found 150,000 hateful tweets in 11 months. (The author’s FAQ is an interest read)
An addictive collection of beautiful charts, graphs, maps, and interactive data visualization toys -- on topics from around the world.