Endogenous land use and the Ricardian valuation of climate change

The Ricardian technique uses cross-sectional variation in the capitalized value of climate in land to infer the agricultural costs or benefits of dynamic climate change. While a practical approach for predicting the consequences of global warming with readily available data, it may yield biased results when land-use decisions depend on the climate attributes being valued and when land has unobserved attributes that differ with the use to which it is put. This paper illustrates the conditions under which such a bias will occur, describes an empirical model that corrects for it, and estimates that model with agricultural census data from Brazil. The approach, moreover, allows constraints on adjustment to be explicitly incorporated into the Ricardian framework, relaxing one of that technique's most conspicuous assumptions.