Addressing Contextual and Location Biases in the Assessment of Protected Areas Effectiveness on Deforestation in the Brazilian Amazônia
Abstract
Using a remotely sensed pixel data set, we develop a multilevel model and propensity scoreweighting with multilevel
data to assess the impact of protected areas on deforestation in the Brazilian Amazon. These techniques
allow taking into account location bias, contextual bias and the dependence of spatial units. Our results show
that the hierarchical structure of the database matters and should be considered in the assessment of protected
areas effectiveness. Our results also suggest that protected areas have slowed down deforestation between 2005
and 2009, whatever the type of governance. The effectiveness of protected areas differs according to socioeconomic
and environmental variables measured at municipal level. For instance, indigenous protected areas are
found to be marginally more efficient than sustainable use areas and integral use areas. Protected Areas that
were more recently implemented are also found to avoid more deforestation than older ones. This corroborates
the idea that recently created protected areas in the Brazilian Amazon have a greater agricultural potential.