Endogenous Targeting and the Additionality of Conservation

The world has lost one-third of its forests, with those in the tropics facing the most rapid decline despite their substantial ecological and climate benefits. Protected areas (PAs) have become the primary policy instrument to curb deforestation, yet they are often established where deforestation pressure is relatively low and potential conservation gains are limited. We evaluate how endogenous policy targeting shapes the additionality of PAs established in Bolivia between 1991 and 2023. We employ a staggered difference-in-differences design, matching treated and control units on a novel measure of predicted deforestation risk in the absence of protection, generated using a Random Survival Forest model. This framework allows us to evaluate treatment effects across the distribution of baseline deforestation risk. Our estimates indicate that, on average, PAs reduce deforestation by approximately 0.19 percentage points (pp), corresponding to a 68% reduction relative to the national deforestation rate over the study period. However, average treatment effects mask substantial heterogeneity across the deforestation risk distribution, with no evidence of additionality in low-risk areas and the largest effects emerging under high deforestation pressure, where PAs reduce deforestation by up to 0.50 pp. Furthermore, while PAs are disproportionately established in low-risk areas, we find limited evidence that this reflects systematically greater biodiversity or carbon gains, or that protection in high-risk areas substantially hinders economic development. Overall, our findings suggest that endogenous targeting is a key determinant of conservation additionality, highlighting the importance of prioritising conservation in high-deforestation-risk regions.