3 Estimating spatial interactions in deforestation decisions

Ongoing decreases in the stock of tropical forest have long been a major concern, due to their implications for biodiversity loss and provision of ecosystem services. Ecological research also provides evidence that even if the stock is held constant, the spatial pattern of forest affects the level of services generated (McCoy and Mushinsky 1994; Twedt and Loesch 1999; Diaz et al. 2000; Parkhurst et al. 2002; Coops et al. 2004; Scull and Harman 2004). A highly fragmented forest made up of small patches may not provide the minimum habitat size that some organisms require. Thus it may offer less protection for species than the same amount of unfragmented forest. It is then important to understand the effects of human activities that fragment standing forest and, as a result, alter the size, the shape, and also the spatial arrangement of habitat. These properties of habitat affect extinction rates of local populations. Standard economic models of rural land use (e.g. agriculture/forest frontiers) will generate predictions of spatial pattern down to the level of detail that their data permit. However, a focus on spatial pattern highlights a question these models do not address: are there spatial dynamics per se? If we look behind observed spatial correlation, do one’s landuse choices actually have any causal impacts upon those made by one’s neighbours? This chapter presents a model of such spatial interactions and then discusses a method to empirically test for their presence using observed deforestation behaviour. Their existence has implications for the stock of forest, its pattern and the effect of policies on forests.