Can Payments for Ecosystem Services schemes reduce deforestation? A robust evaluation example from the Bolivian Andes.

There is growing interest in Payments for Ecosystem Services (PES) as a habitat conservation approach. Key questions remaining are 1) the extent to which conservation funded is additional (would not have occurred in absence of the scheme), and 2) the extent to which habitat degradation leaks away from enrolled land, and 3) whether improvements in outcomes of interest can be robustly attributed to the intervention. We explore these questions for Watershared, a widely implemented payment for watershed services scheme in Bolivia which incentivises upstream farmers to cease deforestation and exclude cattle from riparian forest to protect habitat and water quality. To robustly evaluate the effectiveness of the scheme at reducing deforestation, we made use of a Randomised Control Trial set-up, where 128 communities were divided randomly into treatment and control categories, and treatment community households were offered contracts on a voluntary basis. We evaluated additionality of enrolled land using 509 participating households’ responses to a survey on individual land parcels. Leakage was examined by evaluating deforestation at the community scale, thereby accounting for deforestation next to enrolled land. Intervention effect on deforestation was assessed using openly available Global Forest Change data to model the relationship between the PES scheme and control + treatment deforestation after implementation (2011-2016), accounting for competing predictors such as past deforestation (2000-2010) and geographical factors (e.g. slope, distance to road and river). We did not detect an effect of the scheme in an intention-to-treat analysis where control and treatment communities were included as a binary variable, which could plausibly be influenced by two factors. Firstly, only 13% of the area or 22% of contracts included in the survey were additional, showing a predominance of business-as-usual, and secondly, land enrolment across treatment communities was highly variable (0-80% of community area), limiting the use of binary analysis. Indeed, significant negative relationships between deforestation and % enrolled land were apparent both at the community (% deforestation) and the pixel (deforestation probability) scale of analysis (p=0.02, p<0.001), suggesting that the scheme reduces deforestation given adequate uptake. Importantly, this negative relationship also shows that leakage is not sufficient to offset conservation gains of the PES scheme. In a hypothetical scenario where the study communities all have 60% land enrolled, the gains in forest area would be ca 1000 ha compared with a 0% scenario, equating to >30% reduction in deforestation. We conclude that despite the small proportion of land enrolled being additional, measurable reductions in deforestation have occurred due to the PES scheme. Given the rapid spread in PES schemes but limited robust studies exploring their effectiveness, this research lends support to their wider use.peerReviewe