What's New In Science

UW researchers are using a statistical framework for analyzing social networks to explain how sexually transmitted diseases spread and how people find jobs.

In social network analysis, each person in a particular network is represented by a node. Each person, or node, then connects to any number of other nodes through ties, or relationships. If the model focuses on the spread of sexually transmitted diseases, then each tie represents sexual contact between two people. If the model analyzes the flow of job-related information, those connected may only be acquaintances passing on information about potential jobs, said Katherine Stovel, associate professor of sociology.

Stovel is using network analysis to understand the flow of job-related information and assistance between acquaintances.

"We're interested in how these networks operate and how people use them to gain information and labor-market opportunities," Stovel said. "We're really interested in the ways that things above and beyond skill and education might be affecting the labor prospects of different groups of people."

When looking for a job, living in a rural, isolated location or coming from an economically disadvantaged group may make applicants more likely to be in a network that cannot help them find higher paying jobs, Stovel said.

Stovel is also using network analysis to understand adolescent sexual behavior and the spread of STIs. Using data from the National Longitudinal Study of Adolescent Health, a nationally representative study of adolescent health funded by the National Institutes of Health, Stovel and her colleagues modeled a complete sexual network for one high school. They found what many intuitively know: Partners choose each other based on similarities. In addition, the network exhibited a tree-like or chain structure, she said.

"That is a relatively protective structure. It's the kind of thing that would protect a lot of people, even if disease were introduced into this network," Stovel said. "It doesn't mean an individual has no risk of getting disease, but it's unlikely that the whole school would have an epidemic."

Past models have assumed people form relationships randomly, but that doesn't capture the way people form social or romantic ties, according to an article in the July 2004 issue of the American Journal of Sociology.

"That would be like people covering their eyes and randomly choosing the next person who comes by," said Martina Morris, associate professor of sociology and statistics and director of the Center for Studies in Demography and Ecology.

To overcome this difficulty, researchers have spent the past 10 years developing a general statistical framework that enables them to realistically model the way people form romantic or social connections. The newer approach has already helped explain why HIV infection rates can vary widely between different countries, communities or ethnic groups, even when the average number of sexual partners for each group is similar, she said.

"There's not that much difference in the number of partners people have in the U.S. versus Uganda, but the U.S. has had no generalized epidemic of HIV whereas Uganda, at its high point, had a prevalence of nearly 20 percent," Morris said.

The group found that having multiple partners concurrently greatly increased the connectivity of the network.

"So what we're looking at is maybe it's not the number of partners, but the timing and sequence of partnerships, and in particular, deviations from the usual norm of serial monogamy," Morris said. "Imagine if everybody had only two partners, but they had them at the same time, then it's like a big circle with everybody holding hands."

Now the group is translating their results into potential solutions. If partner concurrency greatly increases the spread of HIV, then new HIV prevention strategies may emphasize having one partner at a time.

"Our goal is to begin to identify the networks' structures that are responsible for providing the connectivity in these different settings and then giving policy makers the opportunity to develop context-specific preventions and intervention," Morris said.

Reach columnist Tia Ghose at news@thedaily.washington.edu.