Impact of Academic Interaction Networks on Grades & Jobs: Evidence from Four-year University Trajectories

Tuan Q. Phan, Dan Ding, Prasanta Bhattacharya, Edoardo M. AiroldiResearch 2021 · DOI

TL;DR

By analyzing 1.5 million interactions among 6,355 university students over four years, researchers found that larger academic networks significantly boost both grades and post-graduation salaries. Most strikingly, the study identifies a "network premium" where mid-performing students with large networks can actually out-earn high-performing students who have smaller social circles.

What Problem Does This Paper Address?

While the adage "it’s not what you know, but who you know" is common, quantifying how these networks form and their specific impact on academic and professional success is notoriously difficult. Previous research often suffered from three main issues:

  1. Data Scarcity: Relying on small, self-reported surveys rather than objective, long-term interaction data.
  2. The "Reflection Problem": It is hard to tell if students perform well because of their friends (peer influence) or if they simply choose friends who are already like them (homophily).
  3. Course Selection Bias: Students don't just form networks in a vacuum; they choose classes together, which creates a "pre-selection" bias that standard models often ignore.

This paper seeks to disentangle these factors to see if academic interaction networks truly drive performance or if they are merely a byproduct of existing student traits.

Methods

The researchers employed a sophisticated 3-stage Stochastic Actor-Oriented Model (SAOM). Unlike standard regression, this model treats the university experience as a co-evolving process:

  • Stage 1 (Course Selection): Models how students choose classes based on prior performance and existing relationships.
  • Stage 2 (Network Formation): Models how ties are created or dropped within those classes.
  • Stage 3 (Performance Change): Models how a student’s GPA (Cumulative Average Point or CAP) changes over time as a result of their network.

By using this multi-stage approach, the authors could statistically separate the "birds of a feather flock together" effect from actual peer influence.

Datasets

The study utilizes a massive, de-identified longitudinal dataset from the National University of Singapore (NUS):

  • Cohort: 6,355 undergraduate students who entered in 2012 and graduated in 2016.
  • Interactions: 1,574,355 digital interaction traces from the university’s Learning Management System (LMS).
  • Academic Records: 234,265 course enrollment records and term-by-term grades.
  • Career Outcomes: Data on first-job placements, including employment status and starting salaries.

Key Insights and Ideas

  • The Network Premium: The most provocative finding is that social capital can compensate for academic capital. Mid-performing students with high "degree centrality" (lots of connections) achieved higher average salaries than high-performing students with small networks.
  • Homophily vs. Influence: Both are at play. Students are significantly more likely to connect with peers of similar academic ability (homophily), but once connected, their performance tends to converge toward the level of their peers (influence).
  • The "Rich Get Richer": High-performing students benefit from stronger network effects than their lower-performing peers. They are more likely to form dense clusters that reinforce high achievement.
  • Gender and the Labor Market: Interestingly, the data showed that female students were more likely to secure a job quickly but started with lower average salaries than their male counterparts, even when controlling for network size and GPA.

Key Results

  • Model Fit: The 3-stage SAOM outperformed baseline models significantly. The Mahalanobis Distance (MHD)—a measure of model error—was 301.57 for the proposed model compared to a staggering 21,384.20 for the baseline.
  • Network Size Matters: Students in the "High Degree" group (top 25% of network size) consistently outperformed the "Low Degree" group in both GPA and job placement probability.
  • Statistical Significance: The homophily estimate (5.64, p < 0.001) and peer influence estimate (5.79, p < 0.001) were both strongly positive, confirming that networks are both a choice and a driver of change.

Limitations and Future Work

  • Content of Interaction: The dataset tracks that an interaction happened (e.g., a forum post or file share) but not the quality or content of the communication. It’s unclear if students were sharing deep insights or just asking about deadlines.
  • Digital vs. Physical: The study relies on LMS data as a proxy for interaction. While highly granular, it may miss "offline" interactions like study groups in a physical library.
  • The Online Shift: As education moves toward hybrid and remote models, the authors warn that the "incidental" network formation that happens in physical classrooms may be lost. They suggest that future work should focus on how to design digital communities (like forums and virtual tools) that can replicate these powerful network effects.

Summary updated

Abstract

Abstract Student interactions play a critical role in education. Interactions and collaborative partnerships formed within universities can have a profound impact on academic and job-related performance. However, little is known about the mechanisms using which these networks develop inside the classroom, generate network effects, and impact academic and employment outcomes. We analyze 1.5 million de-identified in-class interactions, grades, and job outcomes for 6,355 undergraduates through a four-year education program to show that students with larger networks exhibit better academic and employment outcomes, and that high-performing students benefit from stronger network effects, as compared to their mid- and low-performing students. Our results also suggest the existence of a network premium - mid-performers with larger networks can achieve higher average salaries than high-performers with smaller networks. We adapted a stochastic actor-oriented model (SAOM) to model the underlying mechanisms of course selection, network formation, and peer influence among students. Our research sheds light on the importance of interaction networks as drivers of academic and employment performance, as well as the different mechanisms through which high- and low-performing students benefit from them. As education shifts to online, educators and policymakers should invest in the design of digital interactions and communities (e.g.forums) which can generate and sustain healthy networks among students.
BibTeX
@inproceedings{phan2021impact,
  author = {Tuan Q. Phan and Dan Ding and Prasanta Bhattacharya and Edoardo M. Airoldi},
  title = {{Impact of Academic Interaction Networks on Grades \&amp;amp; Jobs: Evidence from Four-year University Trajectories}},
  booktitle = {Research},
  year = {2021},
  doi = {10.21203/rs.3.rs-646237/v1},
}
Figure