Comparing Community Structure to Characteristics in Online Collegiate Social Networks

We study the structure of social networks of students by examining the graphs of Facebook “friendships” at five American universities at a single point in time. We investigate each single-institution network’s community structure and employ graphical and quantitative tools, including standardized pair-counting methods, to measure the correlations between the network communities and a set of self-identified user characteristics (residence, class year, major, and high school). We review the basic properties and statistics of the pair-counting indices employed and recall, in simplified notation, a useful analytical formula for the z-score of the Rand coefficient. Our study illustrates how to examine different instances of social networks constructed in similar environments, emphasizes the array of social forces that combine to form “communities,” and leads to comparative observations about online social lives that can be used to infer comparisons about offline social structures. In our illustration of this methodology, we calculate the relative contributions of different characteristics to the community structure of individual universities and subsequently compare these relative contributions at different universities, measuring for example the importance of common high school affiliation to large state universities and the varying degrees of influence common major can have on the social structure at different universities. The heterogeneity of communities that we observe indicates that these networks typically have multiple organizing factors rather than a single dominant one.

💡 Research Summary

The paper presents a systematic investigation of Facebook friendship graphs from five American universities, captured at a single point in time, to understand how online community structures relate to self‑reported student attributes. After cleaning the raw data—removing self‑loops, duplicate edges, and isolating each institution’s network—the authors apply the Louvain modularity‑maximization algorithm to detect communities within each campus network. The resulting partitions vary in number and size, reflecting differences in campus size, residential patterns, and social density.

Four categorical attributes are examined: residence (on‑campus vs. off‑campus housing), class year (freshman through senior), major, and high‑school origin. For each attribute, the authors treat the attribute labels as a “ground truth” partition and compare it to the algorithmic community partition using a suite of pair‑counting similarity measures: Rand index, Adjusted Rand index, Jaccard coefficient, and Fowlkes‑Mallows index. They derive the expected value and variance of the Rand index under a random labeling model and present a compact analytical formula for the corresponding z‑score, enabling rapid statistical significance testing even for large graphs.

To quantify the relative importance of each attribute, the study converts the z‑scores into a normalized “relative contribution” metric (each attribute’s z‑score divided by the sum of all four z‑scores for that university). This allows direct comparison across institutions. The findings reveal a heterogeneous landscape: at large state universities, shared high‑school affiliation contributes markedly (often >20 % of the total explanatory power), suggesting that pre‑college social ties persist in the online environment. In contrast, at selective private colleges, major affiliation dominates, accounting for up to one‑third of the community structure, indicating that academic interests drive online clustering more strongly there. Class year consistently shows a strong signal across most campuses, but its dominance is attenuated at very small, research‑intensive schools such as Caltech, where residence exerts a larger influence.

The authors argue that no single attribute universally explains community formation; instead, multiple forces interact, producing overlapping and nested community layers. They discuss how these patterns likely mirror offline social organization while also highlighting the unique affordances of online platforms—such as reduced geographic constraints—that can amplify certain ties (e.g., high‑school alumni networks). The paper concludes with suggestions for future work, including longitudinal analyses to capture network evolution, cross‑platform comparisons (e.g., Twitter, Instagram), and qualitative studies to interpret the meaning behind detected communities. Overall, the study offers a robust methodological framework for linking structural network analysis with demographic metadata, providing insights valuable to sociologists, data scientists, and university administrators interested in the digital social fabric of student life.