Age-based treatment effect heterogeneity and propensity score matching evidence on college interventions

This paper revisits the experimental framework by examining the impact of academic support and financial incentives on college students’ academic performance, with a particular focus on treatment effect heterogeneity by age. Using data from the Student Achievement and Retention Project (STAR), the study applies propensity score matching (PSM) to address residual covariate imbalance and improve causal inference. Average Treatment Effects (ATE) and Intent-to-Treat (ITT) estimates are obtained through OLS and 2SLS models, respectively, while quantile regression is employed to explore heterogeneity across the achievement distribution. The results suggest that none of the interventions, academic support (SSP), financial incentives (SFP), or their combination (SFSP), produce statistically significant improvements in GPA in the full sample. However, the combined intervention (SFSP) exhibits the largest economic magnitude and the most consistent positive effects across both short-term and long-term outcomes. Quantile regression reveals that treatment effects vary across the performance distribution, with mid- to high-achieving students benefiting more than those at the lower tail. Importantly, subsample analysis uncovers substantial heterogeneity by age. Elder (non-traditional) students respond more strongly to all interventions in both the short and long run, although these effects tend to decay over time. In contrast, younger students exhibit weaker immediate responses but more persistent gains. These findings highlight the importance of accounting for demographic heterogeneity in evaluating educational interventions and suggest that policy effectiveness may differ significantly across student populations.