Social Networks' Sway May Be Underestimated
Monday, May 26, 2008; Page A06
Facebook, MySpace and other Web sites have unleashed a potent new phenomenon of social networking in cyberspace. But at the same time, a growing body of evidence is suggesting that traditional social networks play a surprisingly powerful and underrecognized role in influencing how people behave.
The latest research comes from Nicholas A. Christakis, a medical sociologist at the Harvard Medical School, and James H. Fowler, a political scientist at the University of California at San Diego. The pair reported last summer that obesity appeared to spread from one person to another through social networks, almost like a virus or a fad.
In a follow-up to that provocative research, the team has produced similar findings about another major health issue: smoking. In a study published last week in the New England Journal of Medicine, the team found that a person's decision to kick the habit is strongly affected by whether other people in their social network quit -- even people they do not know. And, surprisingly, entire networks of smokers appear to quit virtually simultaneously.
Taken together, these studies and others are fueling a growing recognition that many behaviors are swayed by social networks in ways that have not been fully understood. And it may be possible, the researchers say, to harness the power of these networks for many purposes, such as encouraging safe sex, getting more people to exercise or even fighting crime.
"What all these studies do is force us to start to kind of rethink our mental model of how we behave," said Duncan Watts, a Columbia University sociologist. "Public policy in general treats people as if they are sort of atomized individuals and puts policies in place to try to get them to stop smoking, eat right, start exercising or make better decisions about retirement, et cetera. What we see in this research is that we are missing a lot of what is happening if we think only that way."
For both of their studies, Christakis and Fowler took advantage of detailed records kept between 1971 and 2003 about 5,124 people who participated in the landmark Framingham Heart Study. Because many of the subjects had ties to the Boston suburb of Framingham, Mass., many of the participants were connected somehow -- through spouses, neighbors, friends, co-workers -- enabling the researchers to study a network that totaled 12,067 people.
When researchers analyzed the patterns of those who managed to quit smoking over the 32-year period, they found that the decision appeared to be highly influenced by whether someone close to them stopped. A person whose spouse quit was 67 percent more likely to kick the habit. If a friend gave it up, a person was 36 percent more likely to do so. If a sibling quit, the chances increased by 25 percent.
A co-worker had an influence -- 34 percent -- only if the smoker worked at a small firm. The effects were stronger among the more educated and among those who were casual or moderate smokers. Neighbors did not appear to influence each other, but friends did even if they lived far away.
"You appear to have to have a close relationship with the person for it to be influential," Fowler said.
But the influence of a single person quitting nevertheless appeared to cascade through three degrees of separation, boosting the chance of quitting by nearly a third for people two degrees removed from one another.
"It could be your co-worker's spouse's friend or your brother's spouse's co-worker or a friend of a friend of a friend. The point is, your behavior depends on people you don't even know," Christakis said. "Your actions are partially affected by the actions of people who are beyond your social horizon" -- but in the broader network.
In addition, the researchers found that the size of smokers' own networks did not change over time, even though the overall number of smokers plummeted, from 45 percent to 21 percent of the population during that time. The researchers realized that what happened was that entire networks of smokers would quit almost simultaneously.