This article originally appeared on Tonic
Twitter has 328 million monthly active users. But between the news updates, debates, unsolicited opinions, and meme wars, all those people are revealing enough about their personal lives to give a clue about how long they'll live, according to a new study in the American Journal of Public Health .
Beyond showing correlations between what people say they do on social media and how that's mirrored by real life data, the goal of the project was to demonstrate the potential of this method to spot other health trends in something close to real time.
"I think Twitter can be utilized to spot changes quickly, especially because Twitter is one of those platforms people turn to when major events happen to find out updates, sometimes even before they are broadcasted on the news," says study author Quynh Nguyen, a professor of epidemiology and biostatistics at the University of Maryland. "Other research teams not constructing small area summaries may use a week's, month's, or even day's worth of data."
For this study, Nguyen and her team analyzed a year's worth of geotagged tweets from April 2015 to March 2016—nearly 80 million posts from more than 600,000 accounts. The tweets were combed for references to 1,430 popular foods and 376 types of exercise and physical activity (while ruling out mentions where people seemed to be watching or talking about sports instead of actually participating in them).
To gauge mood, the researchers ran the tweets through machine learning software, which could determine whether the millions and millions of tweets were either happy or not-happy with 78 percent accuracy. "We chose to examine the prevalence of happy tweets because we hypothesized that socially expressed positivity would be related to health outcomes," Nguyen says. "However, that meant that the 'not happy' category would include both neutral and sad tweets."
As it turned out, only 19 percent of all the tweets over the year expressed happiness. Despite social media platform's early reputation as people recounting their lunch, only four percent of tweets mentioned food, and less than one percent of tweets talked about alcohol. People tweeted about exercise 2 percent of the time.
Thanks to the large sample size, however, even those small percentages resulted in millions of tweets to analyze. Using Twitter's publicly available geolocation data, the researchers were able to compare users' happiness, eating habits, and workout frequency to county-by-county data on health outcomes from Centers of Disease Control and Prevention and the Department of Transportation.
"Counties with more happy tweets had fewer premature deaths," Nguyen says. Premature deaths simply mean someone died before the age of 75, and the study didn't examine the cause of those deaths. But researchers also noted a relationship between more happy tweets and healthy lifestyles. "Areas with more happy tweets also had a bit more food tweets and physical activity tweets," she says.
What's more, in areas where people tweet most often about workouts or going for a run, those counties saw 714 fewer premature deaths per 100,000 than places where people tweet less about exercise. Obesity rates were also as much as 2.5 percent lower in the counties where social media users blabbed frequently about their meals and workouts. Tweeting about drinking booze also correlated with more alcohol-related deaths. Counties where they tweet about alcohol the least had nearly 4 percent fewer fatal drunk driving incidents.
Some previous studies have shown a connection between happiness and a longer life. One 2010 study in the Netherlands found that happy people between the ages of 65 and 85 were 22 percent less likely to die over the following 15 years than unhappy people. Other studies have found that a better mood is linked to higher levels of physical activity. "To increase happy tweets in a community, we could invest in food and recreational activity resources," Nguyen says.
According to the study, the happiest US state (at least on Twitter) was Montana, followed closely by Tennessee, Utah, New Hampshire, Arkansas, Maine, Colorado, and New York, in that order. Louisiana had the lowest proportion of happy tweets, followed by North Dakota, Oregon, Maryland, Texas, Delaware, West Virginia, and then Ohio.
The least happy month of the year on Twitter? That would be April, with only 15 percent of tweets sounding cheery.