Our Tweets Suggest Massachusetts Isn’t the Happiest of States
Red is happier. Blue is less happy. Image via the Vermont Complex Systems Center study.
Just how happy are we compared to the rest of the country? A new study from the Vermont Complex Systems Center turned to Twitter to answer that question, looking at 10 million geotagged tweets from 2011 to find out which states are the happiest. Massachusetts falls right in the middle—both nationally where we come in 25th—and regionally, where we fall behind Maine (No. 2), Vermont (No. 5), and New Hampshire (No. 8) but ahead of Rhode Island (No. 32) and Connecticut (No. 38).
Researchers measured the happiness of each tweet using a word list created by Amazon’s crowd-sourced laborers at Mechanical Turk’s Language Assessment. The use and frequency of words that Mechanical Turk workers determined to be positive—like rainbow, nice, hope, and beautiful—caused the tweet’s happiness score to go up. The use of words determined to be negative—like bored, stupid, hate, and ugly—caused its score to go down. Researchers admitted that this method failed to take the context of the words into account, and that it struggled to determine the emotional context of smaller samples. But with sufficiently large samples, they said it gave reliable—if improvable— results.
The researchers also note that the findings were, on the whole, consistent with other existing measures of happiness. Hawaii, which came in at No. 1, has been ranked No. 1 in the Gallup-Healthways wellbeing survey for three consecutive years. Utah, Colorado, New Hampshire, and Vermont were also at the top of both lists. It’s not a perfect correlation, though. Louisiana, which came in around the middle at No. 36 in the Gallup-Healthways survey, was ranked dead last at No. 51 in the Twitter study. (It included the District of Columbia.) Massachusetts also ranked lower in the Twitter study—it dropped down 10 places from its No. 14 spot in the Gallup-Healthways ranking. And Maine, which came in at No. 23 in the Gallup-Healthways survey, sky rocketed to second place in the Twitter study.
What gives? As the study authors admit, there’s legitimate concern over how well its Twitter data set can be said to represent the happiness of the greater population. Only 15 percent of adults that use the Internet regularly use Twitter, and 18- to 29-year-olds are more highly represented on Twitter than in the general population. Also, just because a tweet is geotagged to a particular state doesn’t mean that it’s from someone who lives there. Tourists traveling to Hawaii and tweeting about its “beautiful beaches” (both positive words) or to Napa Valley and tweeting about its “awesome wine” (also positive words) are going to drive a state’s ranking up. And those traveling to Foxborough and tweeting about how much they hate losing to the Patriots (with hate and losing counting as negative words) are going to drive our state’s ranking down.
The study also doesn’t account for the idea that some people might express emotion in some states or within some demographic groups differently than in others. The word “boo,” for instance, was scored as a negative word in the study, even though it can also be a slang word used as a term of endearment. The word “nasty” also received a negative score, even though in some corners, its used to describe something as cool. Profane words were also scored negatively, skewing the happiness readings of tweets like “that s— is f—ing awesome.”
You’ve got to wonder then whether the sample size means that researchers were right to determine meaning while discarding context. Before we’re able to draw any substantial conclusions from it (don’t go moving to Maine yet) or grow too concerned about our middling ranking, they’re going to have to work out some of the kinks in their methods. Even so, it’s another cool representation of how researchers ought to be using the interesting and vast new data set that is now widely available to anyone with an internet connection thanks to Twitter.
