Estimating New Destination Choice Models for Commuting and School Travel Using Systematic Data Sources in England

Estimating New Destination Choice Models for Commuting and School Travel Using Systematic Data Sources in England


G Deane, Y Jin, I Williams, WSP Policy & Research, UK


We report new destination choice models for commuting and school travel, which are significantly better in terms of evidence base and calibration methodology


Commuting and school travel have also been a central concern in the development of sustainable transport policies in the UK and many other European countries. Representing travel behaviour of workers, pupils and their accompanying adults realistically is a prerequisite for transport policy analysis, especially in terms of policy interventions such as sustainable travel initiatives, pricing, and infrastructure investment.

However, with few exceptions the emprical evidence base for distribution choice in existing travel demand models have been limited. For example, the commuting models often relate to trips only without any reference to the underlying production-attraction of the employed residents. Also, the existing school destination choice models tend to consider only travel costs and times to and from school, and make little reference to school admissions policy and parental choice which are central to school choice. This tends to produce partial and often unrealistic responses in the travel demand models, particularly under pricing and fares scenarios.

A large number of travel demand models are also defined with large geographic zones, which makes it difficult to represent the shorter distance ranges of school travel.

To date, a major difficulty in improving commuting and school travel modelling has been the lack of appropriate locally observed data for calibration.

We report the design and estimation of new destination choice models for commuting and school travel in the East Midlands region in the UK. This has been made possible through an innovative use of the Census database and the National Pupils Database (which is a systematic administrative dataset held at the UK Central Government Department for Education and Skills).

The model estimation work has been carried out as part of the development of a strategic land use and transport model for the Nottingham-Leicester-Derby area, which is funded by the Highways Agency within the UK Department for Transport and a number of regional partners.

The availability of these data sources has enabled us to estimate new and significantly improved models. These new destination choice models are calibrated, in particular:

1) To provide production-attraction as well as origin-destination travel matrices, in a transparent manner

2) To explore simultaneously those model parameters that tend to be collinear, e.g. travel costs, times, and the impacts of labour market areas and school catchments, taking advantage of the large data sample

3) To define detailed geographic zoning - on average each travel demand model zone contains around 1000 households

The destination choice models have been estimated separately for distinct segments of commuters and pupils, using a logit-based discrete choice model structure, estimated using the statistical estimation package Biogeme.

We will consider the model results (the presentation of results is subject to DfES permission) in the context of model applications, in particular

1) The use of these new and improved models in testing current policy initiatives, such as parental choice of schools, green travel plans, and improvement of public transport services

2) Possible medium to long term improvements in data collection and surveys which may lead to refinement of the destination choice models

3) Possible implications for modelling the distribution of other non-discretionary travel such as employer's business travel.


Association for European Transport