Choice set pruning in residential location choice modelling: a comparison of sampling and choice set generation approaches in greater London

The discrete choice analysis of residential location choice forms an important part of land use-transport modelling systems but gives rise to a number of significant modelling challenges, one of which is the choice set formation problem. A number of alternative approaches exist to addressing the choice set formation problem, but to the best of our knowledge, there have been no efforts to empirically compare their performance using real choice data. In this paper, we propose to address this gap by examining the performance of several choice set formation methods within the context of residential location choice in London. The paper presents a brief review of the relevant theoretical and empirical literature on the treatment of the choice set formation problem in the context of spatial choice and sets out the characteristics of the specific modelling approaches implemented in the empirical research. In particular, we compare a recently proposed hazard-based screening model with more conventional pure statistical treatments. The empirical research uses a specially constructed data-set on residential location choice, developed by the authors for the Greater London area, which draws on information from a number of separate data sources (including revealed preference data from the London Area Travel Survey). The alternative choice set formation approaches are assessed in terms of their prediction performance on a hold-out validation sub-sample and also in terms of estimated model parameters.