Modelling residential location and travel decisions using detailed revealed preference data

Choices regarding residential location are closely linked with travel behaviour. Mathematical models of residential location choice and travel decisions can be used to quantify how these interdependent decisions are influenced by the location and transport attributes and the socio-demographic characteristics of the decision making household or individual. While, revealed preference (RP) data is the most dependable and unbiased source of data to capture the interdependencies among the residential and travel decisions – missing information and coarse spatial and temporal resolution of such data makes it very challenging to use it for developing detailed residential choice and travel behaviour models. This study aims to model household residential and travel decisions and their interdependencies capturing some of the crucial behavioural modelling issues. The residential decision of a household is typically a two-step process: residential mobility decision and residential location choice. Existing models have weaknesses in terms of capturing the geographical scale of the residential mobility decision (i.e. whether to move local, regional or national level) and its impact on household travel decisions. Models to predict the geographic scale of the residential mobility have been developed in this research using the British Household Panel Survey (BHPS) dataset. Further, while capturing the role of residential mobility on car ownership and mode choice decisions, existing studies have considered each direction of shift in car ownership change (e.g. gaining first car, gaining additional car, etc.) and mode choice (e.g. switching from car to public transport, car to active travel, etc.) in separate models. To fill in this research gap, this study attempts to jointly explore the multiple dimensions of changes in a single econometric model. On the location choice aspect, this work also provides important behavioural insight into how the residential location preferences of two major housing markets (ownership and renting) are different from each other. The London Household Survey Data (LHSD) is combined with the Ward Atlas Data (WAD) of Greater London area and travel distance data from the London Transport Studies Model (LTSM) to get a comprehensive set of factors influencing the zonal level choice of residential locations. The residential location preferences modelled in this work are complex due to unobserved choice set for individuals and the large size of the universal choice set. The probabilistic approach and heuristic based methods available in the literature are likely to have weaknesses in terms of capturing behaviourally realistic choice sets in the context of residential location choice. This research makes advancement in the context of choice set generation by proposing an improvement of the state-of-the-art= semi-compensatory choice set construction technique. The proposed technique has better performance over other available semi-compensatory techniques. The empirical results using the RP data provide insights for urban and transport planners by enabling them to better predict the residential and travel decisions in alternative policy scenarios.