Analysing Trends and Sources of Variability in Average Travel Distance: New Insights for Trip Distribution Modelling
K Jahanshahi, Y Jin, I Williams, University of Cambridge, UK
An innovative use of SEM to investigate interactions among socio-economic, demographic, accessibility and land use factors in determining patterns and trends in travel distance. This follows our ETC 2009 and 2011 papers.
This presents the final stage of three linked research studies to investigate the factors that influence the characteristics of passenger travel demand:
* The factors influencing trip rates were presented in ETC 2009;
* those influencing total weekly travel time per capita were presented in ETC 2011;
* here we discuss the influences on annual travel distance per capita, in the context of the previous papers, and aim to obtain new insights into the patterns and trends in travel distance per capita that have been overlooked in the existing literature. Also, by controlling for the effect of travel distance, we build on our 2011 study on travel time per capita by investigating the extent to which changes in total travel time per capita are due both to changes in travel distance per capita and to changes in journey speeds, including the impacts of traffic congestion.
Provided we controlled for differences between person types and area types, our earlier studies demonstrated that there was no significant evidence of major changes through the years either in trip rates or of travel time per capita. However, this constancy through the years is not observed for travel distance per capita. Travel distance per capita has increased strongly from the 1970s onwards. This study analyses how much of this increase is due to:
* incidence effects: e.g. an increased proportion of the population falls within the groups that traditionally have travelled furthest (e.g. high income males or rural dwellers)
* behavioural change: those within a homogeneous group now travel further on average than would have been the case in the past.
To carry out these investigations we use an extensive time-series dataset ? the Great Britain National Travel Survey (GB-NTS) and we adopt an innovative approach based on Structural Equation Modelling (SEM).
GB-NTS is an annual series of household surveys across Great Britain designed to provide regular, up-to-date data on personal travel and to monitor changes in travel behaviour over time. GB-NTS covers individuals? travel data for seven days of the survey week and collects a wide range of attributes regarding the individuals and their households, areas of residence and trips. It is one of the most comprehensive surveys of this type in the world and the long term stability in its methodology makes it particularly suitable for studying behavioural trends through the years.
We first analyse the influence of socio-economic, demographic, accessibility and land use factors on the distance travelled by individuals and then investigate the rate of growth through the years in this total travel distance per capita while controlling for the type of person and place. The distance by mode and purpose will also be examined to broaden understanding and to examine the importance of changes in car ownership patterns.
There are many correlations between the characteristics that influence travel distance per capita (e.g. high income households living in picturesque low density villages with high levels of car ownership versus low income households in high density congested inner city areas with low car ownership). SEM modelling has proved to be an effective analysis tool for such situations as it can help to separate out intertwined influences through its use of latent variables, which are described by a set of observed factors and also through exploring the direct effect of factors on the dependant variables as well their indirect effect through their impact on other factors involved in the analysis.
The outcomes from this research include:
* Measurement of the relative independent influence of each of socio-economic, demographic, accessibility and land use factors on average distance travelled by mode per capita; this quantification is helpful to underpin planning policies aiming to reduce carbon emissions;
* Measurement of the further influence of within-group temporal trends towards longer travel distances;
* Analysis of causes of such trends (e.g. changes in travel speed and increasing ability to pay transport costs).
The overall findings from this research provide a solid evidence base from which to identify homogeneous population segments that should be distinguished when constructing trip distribution models. By maintaining distinct population groups that can be demonstrated to have homogeneous within-group travel behaviour, the forecasting reliability of transport models and their realism for policy testing can be improved significantly.
Association for European Transport