BAYESIAN ESTIMATION OF A LAND USE MODEL

Abstract

Typically, the parameters of land use models are stepwisely estimated and calibrated because of some practical difficulties, e.g., satisfying sign conditions. However, this estimation method lacks the reliability from the theoretical viewpoint as the parameters that should be estimated simultaneously are separately estimated one by one. This study is the first attempt to mitigate this problem. We aim to introduce a Bayesian approach into the estimation of land use models. Specifically, we conduct a whole procedure of an econometric analysis using land use models; for a given land use model, we simultaneously estimate their parameters including the variable selection, confirm the reproducibility of the data, and conduct the counterfactual experiment. Both the estimation result and its comparison with the result by the conventional method suggested the superiority of Bayesian approach as well as its potential for expansion.