Unveiling the Dynamic Impact of Protected Areas: An Event Study Analysis to Assess Conservation Effectiveness

Previous studies estimating the effect of the creation of protected areas (PAs) on forest conservation suffer from biases due to staggered protection and to unobservable drivers of protection’s effectiveness. We address these biases by using a cohort-time refined effect estimator in an event study with Amazon Basin data from 2003 to 2020. Which also unveils meaningful dynamic patterns that remained so far hidden in previous papers’ aggregate effects. Our findings show that PAs’ effects on deforestation and fires were at least halved by the aforementioned biases, being also deflated in 13% and inflated in 16%, respectively, by the failure to control for concomitant and synergistic anti-deforestation policies. We also found strong evidence of forward-looking behaviour by deforesters, with deforestation becoming larger inside protected land two years before protection. This suggests that local agents rush to deforest after learning that the likelihood of being sanctioned will rise with protection. A gradual increase of the effect with the ageing of PAs confirmed that enforcing protection is subject to learning. Also notably, effects were heterogeneous. Whereas both moderately and severely restricted PAs avoided fires, only severely restricted avoided deforestation. In addition, whereas neither national nor subnational conservation unit PAs have reduced deforestation, national units reduced fires but subnational increased them. Indigenous lands reduced deforestation and fires. Results urge policymakers to plan the creation of PAs not merely seeking to change the tenure of land but mainly to align expectations of deforesters to national conservation goals.