By Oliver Gladfelter
Just before this school year began, President Weinberg penned a letter to the class of 2021 with ten pieces of advice. Coming in at #6, he encouraged incoming students to “engage a wide range of people” and to “seek out people who are different from you.” While I fully agree with Weinberg on this, I also understand the human tendency to default to interacting with those who think, act, and even look similar to themselves. So to what extent are Denison students following Weinberg’s suggestion and diversifying their networks?
To investigate this, we’ve opted to explore the composition of social networks on campus. Each student has a network of individuals whom they interact with more often than not – how similar or different these individuals are to the student provides a measure for “social segregation.” For example, if students only hung out with those of the same race or the same political ideology, we’d argue Weinberg’s advice isn’t being followed too well. While the possibilities were endless, we decided to look into three factors which may influence students’ social networks: fraternity/sorority involvement, race, and political partisanship.
In a survey administered to the student body in February of 2018, we asked students to name some of their closest friends at Denison. Students could pick up to five names from a list of all current students. We also asked respondents about their race, political party, and if they’re involved in with a fraternity or sorority. Because respondents’ named friends may have also taken the survey, we were able to collect demographic information about students’ social networks.
34.4% of students in the sample are involved in a fraternity or sorority, so we’d expect the average social network to contain about the same proportion of Greek members. In reality, networks included an average of 37.7% Greek members. This means that if a random Denison student were to list their five best friends, we’d expect at least two of those friends would be in a fraternity or sorority.
Of course, the number of Greek members named depends on whether the respondent was involved in a fraternity or sorority. The networks of Greek students were 73.1% Greek, on average. For the networks of non-Greek students, this average drops to 19.5%.
Not surprisingly, the racial makeup of social networks shows similar patterns. Networks included an average of 75.6% white students. So if a random Denison student listed four of their closest friends, it’s a fair bet that three of them would be white. This network average is nearly the same as the survey sample – 75.4% of students involved in this study identified as white.
The race of the respondent strongly influences what their network looks like: among white students, 86.3% of their social network is also white. For students of other races, 45.9% of their network is white.
Political Party Identification
In a previous post, I encouraged students to reach out to those who may hold different political viewpoints. With that in mind, exploring if partisans are grouping together in the same social networks is particularly important. 73.2% of the sample were Democrats and 17.9% were Republicans. Meanwhile, the average network contains 76.1% Democrats and 14.5% Republicans. This low number for Republicans doesn’t tell the most uplifting story, however it may just be a supply problem, considering most of the campus leans left. We can tease this theory out by comparing partisans’ network averages.
On average, the social networks of Democratic students typically consisted of 80% other Democrats and just 12.2% Republicans. So would we expect to see the opposite trend on the other side of the aisle? Not quite – the social networks of Republican students consist of nearly 64.6% Democrats and 19.8% Republicans.
Clearly, Denison partisans prefer to include their own kind in their network. Of course, this seems to be partially a matter of supply; even for Republicans, it’s far more likely to find a Democrat in their social network than another Republican because there are not enough Republicans to go around. Another explanation is that Republicans just really do prefer Democrats over their own kind, but I’m not betting any money on that one. More than likely, it’s the result of Republicans having far fewer options – a Democrat at Denison could easily construct a network of their own kind. A Republican would have to be a lot more selective.
In each demographic examined, students show a strong preference to their own when it comes to developing friendships (called “homophily”). This may suggest Weinberg’s advice isn’t being faithfully followed, however there’s a huge caveat here I’d be remiss not to mention – even if two individuals aren’t best friends, they could still bump into each other for political discussion. Even if a student’s social network is fairly homogeneous, there are still plenty of areas on campus to engage across differences, such as classes, student organizations, and residence halls. Yet in the social context, Denison students appear to gravitate towards their own.
Following graduation, Oliver Gladfelter will be heading off to New York City, where he will continue procrastinating and writing about data (in that order). Oh, and he’ll also have a job.
1. Average demographic values were calculated by aggregating the answers of each friend then dividing by the number of friends who answered the demographic questions. The percentage of white individuals making up a network, for example, was defined as the number of named friends who classified themselves as white divided by the number of named friends who answered the race question. When calculating averages for network demographics, we subsetted the data set to include only respondents who had at least two of their closest friends also take the survey. This cut down the sample size from 540 to 220.
2. The trend line for percentage of Democrats included in a network had a statistically significant relationship (p < .05) with partisan identification. The percentage of Republicans included in a network had an insignificant relationship with partisan identification, although the mean averages still indicate a pattern.