Empirical versus exogenous spatial weighting matrices: an entropy-based intermediate solution

The classical approach to estimate spatial models lays on the choice of a spatial weights matrix that reflects the interactions among locations. The two main alternative proposals to specify the structure of these weights may be called exogenous and endogenous procedures. In this paper, we propose an intermediate solution by using entropy econometrics. The basic idea is that, once an exogenous weighting matrix is specified, we could modify our initial specification using information contained in the sample. We evaluate the comparative performance of this entropy-based estimation by means of Monte Carlo simulations and also show an illustration with an empirical example.