From Online to IRL: Zicklin Prof Shows How Facebook Connections Influence Elections
January 30, 2025A few years ago, Ecem Basak, PhD, was poring over Facebook’s Social Connectedness Index (SCI), which measures the strength of social ties between different geographic areas as represented by Facebook friendship connections. As her cursor hovered over an interactive map showing links between distant U.S. counties, a lightbulb went off in her head.
“I noticed the map looked a lot like similar ones I’d seen displaying the 2016 U.S. election results,” says Dr. Basak, who is an assistant professor in the Zicklin School’s Paul H. Chook Department of Information Systems and Statistics.
For example, because of Black migration to Chicago after World War II, Illinois’ Cook County has historical ties to two counties along the southern Mississippi River, reflected in strong Facebook connections. Basak observed that the two areas exhibited similar voting trends as well: “Counties that were more socially connected yet were geographically distant appeared to have similar election results. I wondered if Facebook friendships had influenced how people voted.”
Basak decided to test her hypothesis. The results were presented recently in “The Role of Online and Geographically Distant Social Networks in Political Decision-Making,” a paper she coauthored that was published in the Journal of the Association for Information Systems. Examining the impact of Facebook ties between two counties on voting outcomes, the study found that the more social connections two counties shared, the more likely they were to vote similarly for major party candidates in the 2016 U.S. presidential election.
But could it be instead that people’s political similarities determined the intensity of their Facebook friendships? In other words, is it possible that Basak and her coauthors had it backwards? To rule out reverse causality, they employed a common statistical method known as an instrumental variable estimation. They used the U.S. highway network as an instrument, arguing that highways fostered lasting social connections that influence voting patterns, rather than directly affecting them. Their diagnostic analysis yielded robust results.
“By using this variable, we determined that the causality goes only one way,” Basak explains.
Basak hypothesizes that granular data on social connections at the county level—perhaps incorporating information from more recent social platforms such as Instagram or Snapchat—could be used by political parties to sway voters.
“Maybe if all those volunteers knocking on doors in Bucks County, Pennsylvania, had also knocked on doors in another county with social connections to Bucks County, the election results would have been different,” she muses.