Cascading Behavior of Networks of Smokers/Nonsmokers
Article: http://www.time.com/time/health/article/0,8599,1808446,00.html
The act of smoking cigarettes has been a part of American culture since the early 1900s. In 1914, total tobacco consumption was 7 percent, but by the 1940s, that number drastically increased to 40 percent. Like most trends or fads in society, the network effect of cigarettes made it more attractive to smoke cigarettes as more people decided to start. This type of trend can be classified as a bandwagon effect, a type of positive network effect in which a single person has a higher probability of participating as the proportion of society that participate increases.
The general attitude towards cigarettes changed direction when the Surgeon General issued warnings regarding the negative externalities of cigarettes, such as increased risk of lung cancer, heart disease, and emphysema. Today, more and more people are beginning to quit, and the percentage of American smokers has fallen to 20 percent.
A recent article from Time Magazine suggests that similar to the rise of the popularity of cigarettes, the trend to quit smoking may also be “contagious.” A study conducted by Nicholas Christakis of Harvard Medical School and James Fowler at University of California San Diego was started in 1971 to observe networks containing smokers and nonsmokers. By analyzing these networks, Fowler and Christakis are able to trace patterns among networks containing smokers who decide to quit over the period of their study. They started with networks with equal probability of finding smokers and nonsmokers in the center of the network. By the year 2000, their studies showed that the nonsmokers now outnumbered the smokers, and it was common that smokers in these networks were pushed to the outside of the network.
This change in their sample networks made it apparent that quitting smoking is a new trend in society. The changes in the networks showed that if one person quits smoking, his close friends are 36 percent less likely to be smokers. This influence shows the new network effect opposite to that of the early to mid 1900s. Now, close friends are influencing each other to stop smoking.
Since the network spans beyond this person’s immediate friends, his affect on others outside his group of friends but still in his network are also influenced. The study finds that friends of friends of ex-smokers are 20 percent less likely to smoke. Despite no direct interaction, a smoker quitting can make it more likely for people he is not close with to follow his lead and quit smoking as well. The span of the network affected by one person quitting was found to be much larger than Christakis originally expected. He noted that geographical distance does not affect a person’s influence on people in his network.
These networks show a cascading effect. Like with other probability problems, we can treat a friend quitting smoking almost as a type of signal. This signal would weigh into a friend of an ex-smokers decision to quit smoking. If that friend chooses (or seriously considers) to quit smoking, then his friends are also more likely to quit smoking. This continuous effect represents a cascading trend through the network.
http://www.time.com/time/health/article/0,8599,1808446,00.html
http://www.webmd.com/smoking-cessation/news/20081113/smoking-rate-is-declining-in-us
http://muse.jhu.edu/login?uri=/journals/reviews_in_american_history/v028/28.1laird.html