By Dr. Jakina R. Debnam, PhD, Assistant Professor of Economics at Amherst College
If colleges and universities are to be vehicles of socio-economic mobility, diversifying student social networks is key.
From the start of freshman year, students’ social connections have an important effect on their outcomes. Unsurprisingly, the evidence for these peer effects is robust and spans across decades of research. Multiple studies have found that the academic ability of a student’s roommate can impact their own academic performance. A study of students at Berea College finds that a student’s roommate’s high-school GPA predicts one-fifth of the variation in own GPA that the student’s own high school GPA does. Roommate behavior can also help us predict social behaviors such as the likelihood that a student joins a fraternity, takes up smoking, or binge drinks. Stinebecker and Stinebecker find that if a student’s roommate plays video games during college, then that student earns lower grades. Dormmates, classmates, teammates – evidence of peer effects in educational attainment spans throughout a students’ time in college. There is even a documented role of peer effects in the first job that students get after graduation. In a 2002 paper, David Marmaros and Bruce Sacerdote found that among students graduating from Dartmouth College, the post-graduation employment of students is correlated with that of their randomly assigned roommates and hallmates.
Outside of a few rare circumstances (like the assignment of college roommates), students’ peer networks are notoriously difficult to disentangle. Social environments are a function of individual choices and circumstances, and a function of social tendencies, most of which are not random. In fact, if we look at any student network (or at any network!), we see that people who are similar are likely to be friends with each other. This tendency is called homophily.
Homophily holds for commonality across unobservable characteristics (like religion and personality type) and across observable ones (like gender and race). In virtual environments, like the online communities, students’ characteristics are less salient – observable characteristics may play a smaller role in defining student social networks. Low-income students may be more likely to connect with more affluent students. International students may be more likely to connect with domestic ones. This presents an opportunity to help students who might not otherwise interact to influence one another’s outcomes.
Using data from a Salesforce platform used by a college within a large elite university, I compare the way students make friends before and after they arrive on campus. Using this data, I find evidence that students are making more diverse connections within the online community. If I randomly simulate thousands of networks using a process called exponential random graph modeling, I find that when individuals interact exclusively online, they are less likely to sort based on observable characteristics than we might otherwise expect them to be. I find statistically significant evidence that students’ ethnicity does not predict friendship formation during these exclusively online interactions. However, this pattern reverses once students have met.
Further investigations suggest that sorting among students may be driven by the increased propensity to form relationships with others of your own ethnicity once ethnicity is salient (rather than by the decreased likelihood of forming relationships with others from different ethnicities). For example, white students are less likely to form ties with one another in the exclusively-online period, while becoming more likely to connect with one another once on campus. In all demographic categorizations I examine, except students’ citizenship, the importance of categorical similarity for the ways students form relationships increases once they are on campus. Compare this to findings from Mayer and Puller (2008) who find that two college students on Facebook are more likely to form friendships if they are of the same race, major, cohort, and or political orientation.
Further, I find that there is a role for this platform in democratizing access to information – students who may have a lower ability to self-advocate are more likely to use the college’s platform than the average student. Interestingly, I find that first-generation students and students from underrepresented groups tend to be more engaged with the Chatter platform than is suggested by chance. African-American students, who comprise 7.61% of the students in the college, are part of 11% of the connections on the Chatter platform. Students coming from outside of the U.S., who comprise 4% of students in the college, make up 8% of all the connections on the platform. These differences are statistically significant.
My results suggest that online communities for students create an environment where they can make more diverse friendships. The work of others, and common sense, suggests that these friendships can have lasting consequences for students as disadvantaged students gain greater access to the social networks of advantaged ones.
Excited to see what technology can do to help connect your students to resources? Learn more about Education Cloud in this e-book. You’ll see how 8 leading higher education institutions are transforming the student journey.
About the Author
Jakina R. Debnam, PhD is an Assistant Professor in the Economics Department at Amherst College in Amherst, Massachusetts. Jakina works to understand the impact of economic policies and events on human thriving, where human thriving is broadly defined. Her current projects address three themes: 1) consumer responses to anti-soda legislation and campaigns, 2) the primary science of subjective well-being measures, and 3) learning and peer effects in social networks.