Modeling deforestation and development
in the Brazilian Amazon
[I]t is now time for those interested in deforestation to shift the direction of research away from descriptive accounts and a priori reasoning and toward the careful empirical analysis needed to document the relationships involved and to measure their magnitudes.
(Robert T. Deacon 1995)
This chapter will describe our econometric model of land clearing and economic development in the Brazilian Amazon. First, however, we briefly review some of the methodological approaches that have been used in previous empirical studies of the topic and discuss the strengths and weaknesses of each. These studies can be broadly categorized as crosscountry studies, regional-level analyses (such as our own), Geographical Information System (GIS) studies, and micro-level studies. For a more in-depth discussion of alternative models of tropical deforestation, we refer readers to Barbier and Burgess (2001), who also provide an excellent bibliography of recent studies.
In principle, cross-country analyses permit investigations into the relationships between the rate of deforestation and macroeconomic and institutional factors such as economic growth, population growth, openness, trade policies, political regime, indebtedness, devaluation rates, inequality, education, inflation, etc. Many of these factors vary only at the national level, and thus can be analyzed only in a cross-country context. However, serious problems beset these models, including the poor quality of data and the heterogeneity between countries. Most of these analyses are based on FAO estimates of forest cover or forest loss. These data, however, are considered highly unreliable (e.g. Rudel and Roper 1997; Kaimowitz and Angelsen 1998) and alternative estimates vary greatly. For example, the FAO estimates and the World Bank estimates of the rate of deforestation in Indonesia during the 1980s differ by a factor of 3. Furthermore, many countries reported no loss of closed-canopy forest at