Análisis estadístico y modelos econométricos de elección discreta con correlación espacial en transporte y economía urbanaaplicación a los modelos de predicción de la localización residencial

Abstract

A new model of discrete choice between alternatives of a geographical nature is proposed in this thesis, the spatially correlated nested logit (SCNL) model. This type of choices are generally characterized by the existence of a high number of alternatives and the presence of correlation, at least spatial, between them. The current literature on discrete choice models considers two approaches for this type of choice, which are analyzed in this thesis. The proposed model combines both approaches without adding unknown parameters, it presents a closed mathematical structure and it is compatible with mixed specifications that allow variations in the tastes of those who decide. The SCNL model is formulated and analyzed in this thesis, where it is also applied to a real case, to model the choice of residential location in an urban context, which is a key problem in transport planning and urban economics. The empirical results with the new model significantly improve those obtained with the current models. In addition, a spatial metric for spatial zoning with different sizes and irregular shapes is proposed, very common in administrative areas of European cities. The models specified with the proposed spatial metric obtained empirical results that were significantly better than those obtained with the rest of the spatial metrics considered.