Online vs. Face-to-Face Learning in Higher Education: A Causal Meta-Analysis of Propensity Score Matching Evidence

The rapid expansion of online learning in higher education, particularly following the COVID-19 pandemic, has intensified debate about its effectiveness relative to traditional face-to-face instruction. However, much of the existing literature relies on non-randomized designs vulnerable to self-selection bias, and prior meta-analyses have not systematically examined how Propensity Score Matching (PSM) implementation quality influences synthesized findings. This study presents a meta-analysis of 12 PSM-based studies comparing online and in-person learning outcomes, following PRISMA 2020 guidelines.Using a random-effects model, results indicate a small, non-significant overall effect (Hedges’ g = –0.05, 95% CI [–0.26, 0.17], p = .68), suggesting broad equivalence between instructional formats. Subgroup analyses show that online learning is associated with higher academic engagement (g = 0.13, 95% CI [0.06, 0.21], p < .001), while achievement and persistence outcomes do not differ meaningfully. Heterogeneity is substantial (I² = 99.66%) and is not explained by outcome type or PSM design quality, indicating that instructional design and learner characteristics outweigh delivery modality. Guided by Self-Determination, Self-Regulated Learning, and Social Presence theories, we propose the Design–Engagement Nexus to explain these patterns. Sensitivity analyses confirm robustness to unmeasured confounding and publication bias.