6. Conclusions and discussion
A major task involved in setting up a strategic travel demand model such as MetroScan relates to data collection and imputation. This is particularly true when it comes to linking discrete choice models that are estimated on separate datasets and linked together using the concept of maximum expected utility measure. This paper has presented a disaggregated application of discrete choice models to simulate workplace location choices conditioned on residential location choices while informing (or conditioning) the choices of mode and time of day for commuting. Modelling of workplace choices in this way overcomes the challenge of choice set explosion in size due to multidimensional choices (i.e., residence, workplace and commuting mode) being considered but it requires these models to be linked. The paper has shown how the collection of common variables and external data has been used in linking these models and obtaining model parameters that are behaviourally meaningful and intuitively appealing. With respect to the drivers of work location choice, the model detects the presence of both agglomeration and spatial completion forces, with the latter effect being stronger than the former effect. This highlights the importance of using separate accessibility measures if both agglomeration and spatial competition effects are to be captured for a particular planning scenario. Competing destinations models, however, may still be satisfactory if the net effect is of interest.
