The drivers of deforestation are the subject of many spatially explicit studies with considerable policy impact, yet few studies account for spatial dependence, thus neglecting spillover effects. In this work, we use high-resolution remotely sensed land cover change maps, extended with socioeconomic panel data for 141 municipalities in the state of Mato Grosso, Brazil, to investigate the role of agriculture in deforestation from 2006 until 2016. Our econometric model specifically accounts for spatial indirect effects from the dependent and explanatory variables, thus avoiding biased and inconsistent estimates. We identify indirect spillover effects from croplands and direct effects from cattle as significant deforestation drivers. Neglecting to explicitly account for spatial dependence considerably underestimates deforestation pressure of soy production. We conclude that spatial dynamics play a crucial role in deforestation and need to be considered in econometric studies, in order to facilitate informed policy decisions.

ACCESS PAPER