News Release

Smoking is addictive, but quitting is contagious

Peer-Reviewed Publication

Harvard Medical School

BOSTON, Mass. (May 21, 2008)—Over the last 30 years, the number of smokers in the U.S. has steadily decreased—a tribute to the efforts of public-health workers everywhere. And while this fact is unarguable, less obvious are the social and cultural forces that lead an individual to kick the habit. In fact, when someone crumbles that last empty pack of their favorite unfiltered brand and vows to never buy another, he might not realize that he is less like the heroic individual grasping his own boot straps and more like a single bird whose sudden left turn is just one speck in the larger flock.

These are the findings of a massive longitudinal study spanning 32 years: people quit smoking in droves. Through reconstructing the social network of 12,067 individuals, researchers from Harvard Medical School and the University of California, San Diego have discovered that smoking cessation occurs in network clusters and is hardly the isolated decision it might feel like to the individual quitter.

“We’ve found that when you analyze large social networks, entire pockets of people who might not know each other all quit smoking at once,” says Nicholas Christakis, a professor in Harvard Medical School’s Department of Health Care Policy, who, along with U.C. San Diego researcher James Fowler, authored the study. “So if there’s a change in the zeitgeist of this social network, like a cultural shift, a whole group of people who are connected but who might not know each other all quit together.”

The study, which was funded primarily by the National Institute on Aging, appears in the May 22 issue of the New England Journal of Medicine.

Over the last few years, Christakis, who is also a professor of sociology in the Faculty of Arts and Sciences at Harvard University, and Fowler have been analyzing data from the Framingham Heart Study (an ongoing cardiovascular study begun in 1948), recreating the social patterns contained within the study data to see how health correlates with an individual’s social network.

The researchers derived information from archived, handwritten administrative tracking sheets dating back to 1971. All family changes for each study participant, such as birth, marriage, death, and divorce, were recorded. In addition, participants had also listed contact information for their closest friends, as well as coworkers and neighbors. Coincidentally, many of these friends and coworkers were also study participants. Focusing on 5,124 individuals, Christakis and Fowler observed a total of 53,228 social, familial, and professional ties.

Last year, they reported on how obesity spreads through social networks. Using the same data, they decided to analyze smoking cessation trends within that same population.

The first and most striking finding was the discovery that, from the larger network perspective, people quit smoking as groups and not as individuals.

“When you look at the entire network over this 30-year period, you see that the average size of each particular cluster of smokers remains roughly the same,” says Fowler. “It’s just that there are fewer and fewer of these clusters as time goes on.”

They were able to quantify the person-to-person effects of smoking cessation among married couples, siblings, friends, and coworkers. In addition, they also discovered “quitting cascades” that advanced from person-to-person-to-person. (See end of release for statistical chart.)

Christakis illustrates this point by describing a small network containing three individual smokers, persons A, B, and C. The first person, A, is friends with B, and B is friends with C, but A and C do not know each other. If C quits smoking, A’s chances of not smoking spike 30 percent, regardless of whether or not B smokes. The middle individual, it would appear, might act as a kind of “carrier” for a social norm.

Education also seems to matter. We are more influenced by the quitting behavior of others if those people are highly educated. To add a further twist, we are also more influenced by others if we ourselves are more educated.

Says Christakis, “We see by this that the educated are not only more influential, but they are also more easily influenced.”

And finally, Christakis and Fowler discovered that smokers are increasingly marginalized throughout social networks.

“If you look back at 1971, smokers and non-smokers alike were at the centers of social networks,” says Fowler. “For people running companies and having parties, smoking was irrelevant. But during the ‘80s and ‘90s we saw a dramatic shift of smokers to the periphery of the social network. Contrary to what we might have thought in high school, smoking has become a supremely bad strategy for getting popular.”

This marginalization of smokers appears to occur across all educational and economic demographics.

According to the researchers, this is an additional concern. Social marginalization leads to poor health. Smoking then is not only bad for your physical health but for your social health as well.

“What people need to understand is that because our lives are connected, our health is connected,” says Christakis.

“Policy makers have an understandable tendency to treat people as atomized individuals, and to anticipate the impact of their policies accordingly,” says Duncan Watts, professor of sociology at Columbia University, who was not involved with this study. “What this study—like the authors’ previous study on obesity—points out clearly, however, is that individuals do not behave as atoms, but as part of a network. Although simple to state, the consequences of that observation are profound, and will require us to fundamentally rethink our intuition about the world.”

“The culture of individualism is so strong that we sometimes forget how powerfully and silently social networks and those around us influence our health,” said Richard Suzman, director of National Institute on Aging’s Division of Behavioral and Social Research. “If decisions to quit cascade through social networks, then this study has provided public health campaigns a powerful new methodology with which to influence behaviors.”

###

Written by David Cameron

SMOKING CESSATION, RANKED BY RELATIONSHIP:

  • When one spouse quits smoking, the other spouse’s chances of continuing smoking decrease 67 percent.

  • When a friend quits, a person’s chances of continuing decrease by 36 percent.

  • Among co-workers in small firms, 34 percent. (Data for larger firms was insignificant.)

  • Among siblings, 25 percent.

Additional Contact Information:

Harvard University Faculty of Arts and Sciences
Amy Lavoie
amy_lavoie@harvard.edu
617.496.9982

National Institute on Aging
Margaret Vaughn
vaughnms@mail.nih.gov
301.451.8408

Full citation:

New England Journal of Medicine, May 22, Vol. 358, No. 21, 2249-2258
“Dynamics of Smoking Behavior in a Large Social Network”
Nicholas Christakis(1,2) and James Fowler(3)

1-Harvard Medical School, Department of Healthcare Policy, Boston, MA
2-Harvard University, Department of Sociology, Cambridge, MA
3-University of California, San Diego, Department of Political Science, San Diego, CA

Harvard Medical School (http://hms.harvard.edu/hms/home.asp) has more than 7,500 full-time faculty working in 11 academic departments located at the School's Boston campus or in one of 47 hospital-based clinical departments at 18 Harvard-affiliated teaching hospitals and research institutes. Those affiliates include Beth Israel Deaconess Medical Center, Brigham and Women's Hospital, Cambridge Health Alliance, Children's Hospital Boston, Dana-Farber Cancer Institute, Forsyth Institute, Harvard Pilgrim Health Care, Hebrew SeniorLife, Joslin Diabetes Center, Judge Baker Children's Center, Immune Disease Institute, Massachusetts Eye and Ear Infirmary, Massachusetts General Hospital, McLean Hospital, Mount Auburn Hospital, Schepens Eye Research Institute, Spaulding Rehabilitation Hospital, and VA Boston Healthcare System.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.