The dynamics of land-use in Brazilian Amazon

This paper studies the dynamics of land-use in the Brazilian Amazon using a structural vector autoregressive (SVAR) model. A fixed effect panel data specification is used to control for the heterogeneity in the data. Meanwhile, spatial autocorrelation is also diagnosed by a statistical methodology that allows us to split the model in subsamples (clusters) of more homogenous municipalities. The clustering analysis shows that there are three clusters whose land-use patterns are strongly different in an economical point of view. The first cluster identifies municipalities dedicated to logging, natural resources exploitation and slash-and-burn cultures; the second cluster shows a more diversified agriculture; while the third cluster presents very developed intensive agriculture municipalities.