Image courtesy Mitchell et al, PLoS ONE
Published May 29, 2013
The town of Beaumont is known as "Texas … with a little something extra."But the industrial town along the Gulf Coast now has a more dubious distinction: It's been named the saddest city in America—at least, if you're measuring sadness on Twitter.
That's according to a group of researchers at the Vermont Complex Systems Center, who analyzed over 80 million words from more than ten million geotagged tweets written throughout 2011. The results of their study, published Wednesday in the journal PLoS ONE, showed that the happiest tweeters in the U.S. live in Napa, California, and their sad counterparts live mostly in the Rust Belt and along the Gulf Coast border.
"You can infer a lot of information about an area based on what people are writing on Twitter," says Christopher Danforth, a mathematician and a co-author of the study.
Danforth explains how his team measured the emotional state of a tweet: They created a simple computer algorithm to analyze the words within the tweets themselves. Each word was measured on a happiness scale, which his team had previously created using paid workers from Amazon's Mechanical Turk service. The workers were asked to score more than 10,000 common English words on a happiness scale from 1 to 9. Words like "laughter," "love," "rainbow," and "smile" made the top of the list; at the very bottom—unsurprisingly—were words like "terrorist," "ugly," "cancer," "die," and "fatal."
A Geography of Happiness
Using that list, researchers then collected tweets from more than 300 separate cities and towns across the United States and created an algorithm to assess how frequently "happy" words occurred vs. how frequently "sad" words occurred in different places. For example, people in Napa were much more likely to tweet the word "hope" than were their counterparts living along the Gulf Coast.
"The differences in the words people used told us a lot about the cities themselves," says Lewis Mitchell, a mathematician and the study's lead author. "Essentially we were able to create a geography of happiness."
Many of the places at the very top of the list—Hawaii, Maine, and Napa—are also top vacation spots. A previous study by the same researchers indicated that people tend to use less-negative words when they're far away from home. But other places near the top of the list—like Green Bay, Wisconsin, and Spokane, Washington—aren't really tourist destinations.
The researchers say they plan to look at tourism's role in a future study. They also plan to analyze tweets in other languages. The current study looks only at tweets written in English, which could skew data in parts of the United States where many people tweet in Spanish.
In addition, the researchers plan to look at profanity more closely. Their current findings suggest that one of the major driving forces in a city's happiness—or lack thereof—is how frequently people use curse words in their tweets.
"People curse more and more as the day goes on," says Danforth, "but there are definitely places where profanity is more common. In the South, more people are cursing on Twitter. It's a tapestry of negative words."
He notes that many of the cities close to the bottom of their happiness list also rank low on other lists that measure factors like health outcomes and quality of life.
"The people at the bottom of our list live in states that are more socioeconomically depressed and where more natural disasters occur," he says. "There are higher rates of poverty, and the median incomes are lower."
This might explain why places like Beaumont and Shreveport, Louisiana, have sadder tweets. But it doesn't explain one surprising finding: Tweets across the country are getting sadder, in general.
"If you go through all of the demographics since 2008, it's getting sadder everywhere," says Mitchell. "There's a strong downward trend. We don't know why this is."
He recently made a Twitter account—@geographyofhapp—that tracks the happiest and saddest cities on Twitter on a daily basis. But his own personal Twitter account—@dr_pyser—remains cheerfully optimistic.
"I try to be more conscious of what I'm talking about online and the way I talk about it," says Mitchell. "I try to put my best self out there."
Rick Perry for Gov is enough reason to want to jump out of a 10 storied building. I feel sorry for anyone living in texas
I think that emotions are very hard to quantify in any serious way as NLP algorithms are very far from perfect. Additionally, besides any errors in methodology, I don't think there's any realistic way to take accurate stats on Twitter. If you look at http://www.buyfacebookfansreviews.com there are hundreds of sites out there where companies can buy Twitter followers. Many of them use questionable means to promote social media pages and I think this completely distorts any potential stats that could be found from this without normalizing for this effect. It's still interesting to think about though, but I don't think I'd have anywhere close to a supreme measure of confidence in this type of study as presented.
Josh: The questionable methodology of this study aside, the results are consistent with the Gallup/Healthways Well-Being Index.
Both studies point to a lack of economic opportunity and increasing economic inequality as major factors in unhappiness. The recession certainly hit the southern and midwestern states harder than others, so it is not at all surprising these states have a greater number of unhappy people. You can read the results of the Gallup/Healthways study and its the methodology here: http://www.well-beingindex.com
Interesting but trite. Seems it would be far too easy to skew these numbers. One demographic could use specific words repetitively. And seems to not take into account sacasm or context.
ted cruz should make you feel better...if you have obamacare!
Special Ad Section
Our new Phenomena space blogger, Nadia Drake, will be bringing you tales from beyond our home planet.
Video of the Day
During one of the worst droughts in centuries, California's top snow surveyor is sounding the alarm.