Investigating travel behaviour dynamics and their incorporation into transport models
Project details
Full project title: Investigating travel behaviour dynamics and their incorporation into transport models
Sponsor: Engineering and Physical Sciences Research Council
Principal investigator: Dr Kiron Chatterjee
Principal researcher: Dr Kang-Rae Ma
Start date: October 2004
Finish date: September 2006
Project briefing sheet: Download the briefing sheet document (PDF)
Project report: Download the final project report document (PDF)
Project summary
Longitudinal methods can be used to study how people's behaviour changes over time and to develop models that can forecast this. They are commonly used in other fields (e.g. health, education) but have been used to a limited extent in transport where it is usual practice to compare different people's behaviour at a single point in time and infer how behaviour may change in the future from this.
This research examined the use of longitudinal travel data sets to obtain a better understanding of behavioural dynamics and to develop dynamic relationships of travel behaviour. New data collection involved a four wave panel survey of the travel behaviour of residents of Crawley, West Sussex, before and after the introduction of a new guided bus service. There have been few transport-specific multi-period panel data sets collected and the collection of this one represents a unique addition in that it coincided with a major transport intervention, thus incorporating dynamic response phenomena of key interest.
Two main forms of dynamic analysis were conducted to investigate
- the timing of behaviour change
- the transition in bus usage over course of survey period.
Duration modelling was used to analyse the factors that influenced the time taken to first use or 'adopt' the new bus service. The specific methodological contributions of the project were to apply duration analysis to 'adoption' of a new transport mode, which is not known to have been conducted before, and to apply a new version of the duration model, the split population duration (SPD) model, that relaxes the assumption that all people will ever use the new travel mode option. Discrete choice modelling was applied to model the transition in bus usage using various model specifications to capture the dynamics involved. Findings were obtained on the importance of socio-economics, life events and level of service changes that would not have been possible using cross-sectional data.
The latest methodological approaches to handling the challenging issues of distinguishing unobserved heterogeneity and state dependence and of specifying initial conditions were used. It is asserted that the analysis provides a more sophisticated effort at panel data modelling for travel demand analysis than has otherwise reported in the transport literature.
The implications of behavioural dynamics for forecasting were explored through sensitivity testing with different lag response values in the mode choice sub-component of the Dynamic Urban Model developed by Steer Davies Gleave. This showed how the evolution of travel demand is affected over a longer period than the specific value of the lag period assumed.
Drawing on the findings of the research project and synthesising these across the different elements of the project, a set of guidelines has been developed which addresses motivations for dynamic modelling, panel survey methodology, model estimation methods and application of dynamic relationships into modelling systems. A greater realisation is emerging that longitudinal studies can improve our understanding of behavioural change and ability to predict future travel demand but it will take a concerted effort to introduce more longitudinal data collection studies and to develop the expertise needed to analyse the data.
During the course of the research, a seminar was organised (hosted by the Department for Transport) to examine longitudinal methods. Presentations and workshop notes from the seminar are available for download.