Poster #54 - Heterogeneous Treatment Effects in Cross-Cultural College Housing for Economically Disadvantaged Students: A Machine Learning Approach

• In Event: Friday Poster Reception

Abstract

This study employs machine learning techniques to examine whether the positive effects of international roommate pairings vary across different student subgroups at an institution in the Appalachian region of Eastern Kentucky serving a student population of predominantly economically disadvantaged students. Using generalized random forests and randomization inference as a robustness check, I analyze institutional data spanning 2000-2015 (n=6,665) to identify patterns of effect heterogeneity. Despite substantial variation in variable importance scores, the results reveal no statistically significant heterogeneous treatment effects across student subgroups. This pattern of equitable benefits, coupled with significant overall positive effects on early academic performance that diminish over time, suggests that international roommate pairings leverage existing campus infrastructure at minimal additional cost that benefits disadvantaged US domestic students broadly. The findings have important implications for institutions seeking to leverage diversity initiatives to enhance educational outcomes while maintaining commitments to equity.

Author

• Hsin-Ta Tsai, Georgia Institute of Technology & Georgia State University
  Presenting Author