This research increases the overall understanding of a subset of interactions that influence patterns of behaviour:​ intra-household interactions and spatial interactions. Understanding these interactions and their influence on travel behaviour is essential to developing policy sensitive models. These models provide much of the prerequisite information for evidence based planning. This understanding is framed through a review of existing best practices in travel demand modelling and a discussion of how independent models can be integrated within a larger framework.

The thesis presents four empirical models that incorporate either one or both interactions. A model of joint household mode choice which incorporates decisions regarding joint travel and vehicle allocation is estimated. To accomplish this, the model uses a novel choice set formation technique. This is the first modelling to accomplish these tasks simultaneously. Models of transit station location choice are estimated using a newly proposed model structure, the spatially weighted error correlation model. These spatial models are used to compare the difference between independent (transit accessed by driving alone) and joint (transit accessed by being dropped off) spatial choices. The presented model outperforms existing models used to capture the same trend and provides new behavioural insights. A model of chauffeur allocation and location choice for daycare trips is estimated. This model incorporates spatial constraints in this decision through the application of a stochastic frontier model to generate task allocation specific spatial choice sets. This allows for a reduction in the sampling frame for spatial choices. Finally, a model of high school escort and household mode choice is estimated using a parallel constrained choice logit formulation. This empirical investigation generalizes the parallel constrained choice logit to allow for any number of household members to be considered. Concurrently, this empirical analysis highlights a flaw within an existing “best practice” modelling framework.