YOU MAD? Regional variations in anger expressed in tweets

“You mad?” – The internet, it seems, represents a perfect place to vent anger, start a flamewar, and engage in uncivil behavior. Especially in unmoderated comment forums, the combination of distance and anonymity can be a driving force for vitriolic behavior (see Why on is Everyone on the Internet So Angry?)

In this exploration into my big data set of geo-tagged tweets, I examine anger and hatred. Unlike internet comment threads, twitter users are not anonymous. Yet anger on twitter remains a concern for many. Researchers have compared the anger response in twitter as psychologically similar to road rage – it acts at more of a distance, it happens more quickly, and it is more explosive. Furthermore, of all the emotions expressed on social media, it seems that anger is the quickest emotion to spread. In a short time, anger and rage can spread within several degrees of separation.

In this analysis, I have more modest aims:

  • Track anger expressed by usage of the words “anger” and “hate” and their multiple forms (e.g., anger, angry, angriest, etc.)
  • Track anger expressed by emotions ( ? ? ? ? ? ? )
  • Examine how word and emoticon expressions of anger corroborate each other
  • Examine regional variations expressions of anger


Analysis of 24 million geo-tagged tweets show that on average, 0.78% tweets are using forms of the word “hate” or “anger.” That’s not a lot. Of course, there are other ways to express anger in a tweet, and this is just a rough tool. On average, 0.30% of tweets use the anger emoticons. The image shows the hot-spot analysis of regional differences in the distribution of tweets (words and emoticons). Fully red sections use anger statistically significantly more in tweets, fully blue sections use anger less in tweets. The map shows a few things:

  • There is significant regional variation in the use of anger signifiers, and these regional difference represent stable and relatively large regions (not county by county)
  • There is significant correspondence between word and emoticon signifiers. That’s good news, both seem to indicate the same thing

How big are those regional differences? In some cases, the differences are small. Some of the orange and red areas are only slightly above the nationwide average (0.90% vs. 0.78%). Sometimes the difference is stark. For example, portions of Michigan and Texas (at 1.5%) are six times more likely to express anger than portions of Hawaii (at 0.25%), and four times more likely as a state like New Jersey (at 0.5%).

Why do these difference exist? I do not know. Perhaps there are some regional differences in expectations for civility, or different norms for if (or how) the same emotion is conveyed online. Perhaps people in different parts of the country are simply less angry than other parts. I did try lining up this map with some socio-economic indicators like crime-rate, unemployment rate, income distribution, and others. I didn’t get very far.

If you have any ideas, let me know in the comments!

Leave a Reply

Your email address will not be published. Required fields are marked *