Applications of Spatial Multinomial Logit Model to Transportation Planning

This paper presents the process of derivation and development of a spatial multinomial logit model and its application to a housing type choice problem. Over the past few years, a relatively small body of research was developed that tries to capture the spatial and temporal dependencies across decision-makers and alternatives. While temporal dependencies are often considered especially in dynamic models, there has been relatively little work in the literature on incorporating spatial dependencies into qualitative dependent variables and discrete choice models, leading to inconsistent estimates. Part of the reason is that space is in general more complex to deal with than time. The basic idea presented in this paper is that decision-makers may influence each other, resulting in correlated choice behaviour over space. In this paper, spatial dependency terms are implemented in a standard multinomial logit framework. The results show that the spatial terms are statistically significant in the model and improve model fit.