Mapping Happiness In American Cities, Using Twitter

Using Amazon’s Mechanical Turk and about a zillion tweets, these maps correlate happiness and location.

Back in 2009, a group of University of Vermont researchers published a surprising paper in the Journal of Happiness Studies. In it, they outlined the process of determining the saddest day in recent memory, the day Michael Jackson died using data culled from blogs. The happiest, they found, was November 4, 2008–the day Obama was elected. The paper’s underlying concept dates back to 1881, when an Irish scientist conceived of an instrument called a “hedonometer” that could measure the happiness of a population. The 2009 study does something similar, using millions of words scraped from the blogs and ranked according to happiness levels. In 2011, they refined the methodology further and applied it to Twitter alone, analyzing 46 million words.

This week, the same researchers published a fantastic set of maps that show how their findings correlate to geographic locations. In other words, it’s a map of how happy Americans are, right down to zip code. There’s plenty of insight to parse in their full paper, but a few highlights include the revelation that Hawaii is the happiest state (Louisiana is saddest). Napa is the happiest city in the country, a place where residents tweet frequently about wine and restaurants. The saddest was the tiny Beaumont, Texas–largely because of the frequent cursing. “In general, cities in the south tended to be less happy than those in the north, with a major contributing factor being the relative abundance of profanity used in those cities,” explains UV postdoc student Lewis Mitchell.

The study uses word clouds to determine happiness. To get an empirical read on what words humans use when they feel good (or bad), the authors devised a test that ranked thousands of words by happiness, based on answers culled from Amazon’s Mechanical Turk. Using those values, they wrote a complex algorithm that analyzed the words in 10 million tweets, geotagging the resulting “happiness values” on a fine-grained map. So, for example, subjects associated negative emotions with curse words, meaning that tweeters who swore more often were ranked as less happy.

Keep in mind that “happy” is a benchmark for happy words, not literal emotional happiness. Like traditional studies about happiness, which ask subjects to self-report how happy they are, you have to assume a certain amount of obfuscation (No, Mom, I’m fine! I’m happy, really!). It’s up to you whether you want to believe that people who use happier words are actually happier. What’s more, a comparison of the data with 2011 census data revealed that people with more education tended to use more complex words, while those without college degrees used words that are emotional, like “you,” “me,” “love,” and “hate.”

“An exciting result of this sort of analysis of social media data is that we’re getting to the stage where we can infer things about groups of people or places just by looking at the words, the atoms of language, that they use in their natural habitats,” Mitchell explained over email. “I think that the potential for this as a real-time survey technique, or as a one number GDP-like metric of well-being is one of the most exciting aspects of the study.” So these maps reveal more than just how “happy” we are. They reveal how different socioeconomic groups talk about themselves–and how, on the Internet, we present ourselves to the rest of the world.

Check out the full post on the Computational Story Lab here. There’s also an awesome appendix map here that updates national happiness every day.

About the author

Kelsey Campbell-Dollaghan is Co.Design's deputy editor.



More Stories