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

This is the author accepted manuscript. The final version is available from the American Economic AssociationPoster presented at the American Economic Association and Allied Social Science Associations (ASSA) Annual Meeting 2024Previous studies assessing the effectiveness of protected areas (PA) in conserving natural landscapes have seldom explored within variation as means for attenuating the influence of non-observables. That would avoid a bias, as revealed by the post-matching differences-in-differences strategy employed in this paper, of at least 1.8 fold the bias avoided by matching alone, the usual approach in literature. Another source of bias commonly ignored is the staggered implementation of PAs. This was addressed by matching protected land pixels at group-time level with pixels that were never protected, with group defined by the year in which protection began. Which also pointed to a considerable bias. The pixel-level dataset analyzed covered the entire 6 million km² of the Amazon Basin over 18 years. It included two metrics of environmental performance, deforestation and fires, the latter a source of forest degradation. PAs’ effectiveness was attested after mitigating the aforementioned biases. Nevertheless, the impact was small for fires, and both metrics exhibited significant heterogeneity across levels of government and protection stringency. Subnational PAs had a stronger impact on containing fires, but no impact on deforestation. Conversely, national PAs had a larger impact on deforestation and no clear effect on fires. Whereas severely restrictive PAs avoided deforestation but not fires, the moderately restrictive did not avoid any of them. The results were robust to Rosenbaum's hidden bias test. Important policy recommendations follow, whose implementation would avoid large-scale emissions of greenhouse gases. First, enforcement should be strengthened against non-subsistence fires in national PAs and against deforestation in subnational PAs. Second, inside PAs, the replacement of agricultural burnings could be subsidized and fire control requirements be better enforced. Third, policymakers should collect and publish georeferenced information on policies implemented alongside PAs to provide a more accurate evaluation of PAs