Can our workouts be shaped by what our friends do?
That question is at the heart of an important new study of exercise behavior, one of the first to use so-called big data culled from a large-scale, global social network of workout routines.
The researchers focused on running, because so many of the network participants were runners. And what they found suggests that whether and how much we exercise can depend to a surprising extent on our responses to other people’s training.
The results also offer some practical advice for the runners among us, suggesting that if you wish to improve your performance, you might want to become virtual friends with people who are just a little bit slower than you are.
There have been intimations for some time that aspects of our lifestyles and health can be contagious. Using data from surveys and postings on social media, scientists have reported that obesity, anxiety, weight loss and other behaviors, including exercise routines, may be shared and intensified among friends.
But those studies had limitations, particularly related to the tendency of people to gravitate toward others who are like them. This phenomenon, which researchers call homophily, makes it difficult to tease out how friends influence each other’s lives. Many of these studies also relied on people’s notoriously unreliable estimations of their behavior, whether it involved eating or exercise.
The new study, published on Monday in Nature Communications, sought to avoid these pitfalls by turning to data from a worldwide social network devoted to sharing objectively measured exercise routines. (The network is not named in the study for contractual reasons, the researchers say.)
People who join this network upload data from an activity monitor, which precisely tracks their daily exercise regimens. They also become virtual friends with others in the network who seem like-minded. Friends then automatically share workout data.
The researchers, from the Massachusetts Institute of Technology’s Sloan School of Management, eventually gathered five years worth of data from about 1.1 million runners from across the globe. Cumulatively, those in the network had run almost 225 million miles during that time.
The identity of the individual runners was masked, but the researchers could tally exactly how often, far and fast each had gone every day for five years. They could similarly map out how often, far and fast their particular friends had run on those same and subsequent days.
Using this data, the researchers noted immediate correlations. Friends tended to display similar training routines day to day and year to year, even if they were separated geographically. But it remained unclear whether the runners were influencing one another’s distance and pace or just hanging out virtually with people who already ran like them.
So the researchers next decided that they should also consider weather. Bad weather can dampen enthusiasm for exercise, the researchers reasoned, so if someone heads out in rotten weather on a day when friends elsewhere have run, the soggy runner presumably has been influenced by what his or her friends had been up to.
Consequently, the researchers also gathered five years’ worth of data from global weather stations and cross-correlated this massive database with information about the 1.1 million runners’ daily workouts.
The results clearly showed that runners do influence one another, the scientists found. Over all, if one person ran for about 10 minutes more than usual on any given day, that runner’s friends would lengthen their workout by approximately three minutes, even if the weather was discouraging. Similarly, if a friend ran faster than usual, his or her friends would tend to pick up their pace in their runs that same day.
The effects were most pronounced, the researchers found, if one runner previously had been just a little slower or less in shape than a friend but now showed signs of overtaking a friend’s performance. The threat of falling behind would prod that friend to run a bit harder.
Gender also mattered. Men generally ran faster or longer if their male virtual friends had done likewise and also if their female friends had, although not to the same extent. But female runners seemed unaffected by male network friends. They altered their training routines almost exclusively in response to changes among their female friends.
In aggregate, these results indicate that, with caveats, “running can be socially contagious,” says Sinan Aral, the David Austin Professor of Management at M.I.T. who led the study.
The impacts “go beyond correlation to causation,” he says. “In general, if you run more, it is likely that you can cause your friends to run more.”
But the findings apply only to people who already are runners, he adds, since the data he and his colleagues used described runners. They cannot tell us whether other types of exercise are equally catching or how to make exercise in general more palatable and contagious among inactive people.
Dr. Aral and his colleagues plan to use other social media data to study those questions soon.