Measuring soil quality is extremely difficult, yet it has clear economic importance. In particular, there is a great deal of empirical interest in the dynamics of soil quality evolution when land managers respond to policies and other incentives. Yet current methodologies for measuring changes in agricultural land quality are largely static and rely heavily either on incomplete measures such as proxy variables, or ad hoc indexes of selected soil characteristics. Moreover, much empirical work relies on static econometric techniques or simulation models. In this paper, we develop a means to infer soil quality changes from input and output data using a dynamic production function model. Using data from field experiments, we estimate the model in a way that allows the recovery of a dynamic measure of soil quality whose evolution depends on variations in management practices. Our methodology and findings will help provide firmer empirical foundations for analyses of the economic implications of land degradation and the soil quality implications of agricultural policies.