Are your friends making you fat? Can headaches or smoking be infectious?
Past studies have shown that obesity is “socially contagious” and may spread from person to person in a social network. So if your best friend is overweight, there is a good chance you will gain weight as well. But more recent research by Jason Fletcher, assistant professor at the Yale School of Public Health, and Ethan Cohen-Cole of the Boston Federal Reserve Bank, seems to contradict these conclusions.
Fletcher and Cohen-Cole have produced new data revising “the social network effect” in which they statistically accounted for the effect of identical surroundings. For example, two college students typically have similar lifestyles and may be exposed to the same types of foods.
The researchers took into account the differences in schools and neighborhoods in contributing to the spread of obesity in adolescents. Once they controlled for similar features in a given environment, Fletcher and Cohen-Cole found that the spread of obesity through social networks declined significantly.
“Contextual effects can lead to false inference on social network effects,” the authors wrote in their paper. “The social network effects were significantly reduced.”
But Nicholas Christakis, sociology professor at Harvard, and James Fowler, associate professor at the University of California San Diego, proponents of the social network effect, continue to defend the spread of obesity through social ties.
“Network phenomena appear to be relevant to the biological and behavioral trait of obesity,” Christakis said in an e-mail.
While he concedes that three factors contribute to clustering in networks: (1) homophily (i.e. that birds of a feather flock together), (2) confounding (joint exposure to environmental factors) and (3) induction (peer effects), he argues that his analysis accounted for these features.
In a second follow-up study, Fletcher analyzed traits that are unlikely to be “contagions,” such as height or the tendency to develop headaches.
Fletcher observed significant network effects in the areas of acne, headaches and height. Having a friend with acne problems, for instance, increased an individual’s own odds of having acne.
Since Fletcher assumed that a contagion cannot exist in the areas of height, acne or headaches, he concludes that the statistical analysis did not adequately account for the “large environmental bias.”
But Christakis has another explanation for Fletcher’s finding.
“It is not inconceivable that acne, headaches and even height (in adolescents) might indeed spread from person to person,” he said.
Since data are self-reported, it is possible, he said, that people who have friends with headaches will exaggerate their symptoms more than those who don’t. Similarly, whether a person deems the few pimples on his face to be worthy of report as “acne” may be influenced by his friend’s perceptions. Finally, although height cannot spread in adults, an adolescent’s height might be influenced by his peers to the extent that they exercise the same amount or eat similar foods.
The disagreement between Christakis and Fletcher remains largely technical — both researchers agree that a shared environment and homophily are important factors, but the difficulty lies in finding the social contagion while controlling for these factors.
“Two sets of researchers view this problem in different ways,” Fletcher said. “We wouldn’t have been comfortable with their control for friend selection and shared environment. We tried to show that our concerns are well founded.”
The studies by Fletcher and Cohen-Cole were published in the December issues of the Journal of Health Economics and the British Medical Journal.