Estimating Spatial Interaction Models using Panel Data: a Generalized Maximum Entropy Formulation

Flow data are viewed as cross-classified data, and spatial interaction models are reformulated as log-linear models. According to this view, we introduce a spatial panel data model and we derive a Generalized Maximum Entropy – based estimation formulation. The estimator we propose has the advantage of being consistent with the underlying data generation process and eventually with the restrictions implied by some non sample information or by past empirical evidence by also controlling for collinearity and endogeneity problems.