Switching to open access is a long bet: longitudinal causal effects in a single-institution natural experiment

This study estimates the causal effects of switching to open access (OA) within a single university, treating an institutional excellence initiative as a natural experiment that sharply altered local costs, support, and contractual constraints for OA publishing. To reduce self-selection bias, a strict “switcher” identification strategy was applied: switchers are researchers with low baseline OA adoption before the policy shift who subsequently exhibit sustained high OA publishing behavior, while controls remain consistently low throughout. Using annual panel data for 2016–2025, dynamic effects were estimated with Longitudinal Bayesian Expansion Trees (LongBet) that separates baseline citation trajectories from heterogeneous, time-varying treatment responses in a panel setting. Results indicate delayed and increasing effects consistent with citation accumulation: early post-policy estimates are small and uncertain, but the estimated ATT becomes clearly positive over time, reaching a large and statistically credible gain by 2025 (approximately + 52.3% annual citation flow for switchers relative to their counterfactual trajectory). Effects are heterogeneous by career stage and field; the largest average gains are observed among earlier-career researchers, with field-specific patterns.