Estimating the loss of agricultural productivity in the Amazon

We propose a procedure to obtain a general estimate of the rate at which agricultural productivity declines on newly cleared land in the Brazilian Amazon. This estimated parameter has two advantages over conventional estimates. First, it is a general, average estimate that can be used in macro-scale economic analysis. Second, the estimate is derived from regional data accessible to economists rather than from remote scientific stations. In the first stage a model is estimated that tracks the transition of land use from period to period for each municipality, allowing the process to vary according to different characteristics of each municipality and time period. From this land use transition model the percentage of crop land in each municipality that is recently cleared, 5-years old, 10 years old or previously used for other purposes is calculated. These land vintage estimates are used with labor as inputs in a Cobb–Douglas production function which is estimated using GLS. The estimated elasticities are allowed to vary by relevant municipality characteristics and are then converted into a measure of productivity for each land vintage. It is shown that the productivity of land drops in the first 5-years after clearing land and stabilizes thereafter. Several economic arguments are given to support the empirical results. © 1999 Elsevier Science B.V. All rights reserved.