ENHANCING CONVENTIONAL RAIL DEMAND MODELS USING A WIDER RANGE OF EVIDENCE FROM ANALYSIS OF NATIONAL TRAVEL SURVEY DATA



ENHANCING CONVENTIONAL RAIL DEMAND MODELS USING A WIDER RANGE OF EVIDENCE FROM ANALYSIS OF NATIONAL TRAVEL SURVEY DATA

Authors

Mark Wardman, SYSTRA, Peter Connell, Leigh Fisher, Bhanu Patruni, RAND Europe

Description

The aim of this research was to improve the forecasting performance of conventional rail demand models estimated to ticket sales data in Great Britain by enhancing them with behavioural insights using National Travel Survey data.

Abstract

ENHANCING CONVENTIONAL RAIL DEMAND MODELS USING A WIDER RANGE OF EVIDENCE FROM ANALYSIS OF NATIONAL TRAVEL SURVEY DATA

Authors
Mark Wardman SYSTRA
Péter Connell LeighFisher
Alex Coulthard LeighFisher
Bhanu Patruni RAND Europe

The railway industry in Great Britain uniquely has a recommended demand forecasting framework and set of parameters contained in the Passenger Demand Forecasting Handbook (PDFH). This originated in 1986 and has been subject to continual update, improvement and extension, with version 5.1 released in 2012. This paper would report on research that is complete and is being evaluated for inclusion in an updated version of PDFH.

Central to rail demand forecasting is how variables outside of the control of the rail industry, commonly termed external factors, impact upon the demand for rail travel. These variables are key drivers of rail demand, with employment and income recognised as being particularly important but generally with little allowance for other socio-economic factors.

The background to this project, and the reasons why further research on this crucial subject was clearly warranted, was broad acceptance amongst key stakeholders and practitioners in the U.K. that:
• Rail growth forecasts derived from PDFH and official Department for Transport recommendations had not generally performed well in explaining recent growth in rail demand;
• Whilst the current forecasting framework covers the key demand drivers of income and employment, there are other important influential variables which are currently not covered in PDFH;
• Recent econometric studies, which had aimed to provide updated values for existing PDFH parameters and insights into unaccounted influences on rail demand, had not provided entirely convincing findings.
The approach adopted in this study is innovative and enhances significantly the demand models that form the basis not only of PDFH recommendations but which are typical of this type.

We have exploited the behavioural insights that can be provided by the National Travel Survey (NTS). This is, as far as we are aware, a very much under-exploited resource as far as understanding rail demand is concerned, and can explain how rail demand varies with a range of socio-economic variables.

Some previous modelling approaches have aimed to enhance rail demand models by including a wider range of socio-economic variables and more detail to employment and income variables. The problem encountered was that these cannot be freely and accurately estimated primarily due to correlations between variables although data quality is also an issue.

Our approach has been different. We accept that it is not possible to enter a wide range of variables into rail demand models and expect to obtain statistically significant and plausible effects. However, we do recognise that analysis of NTS data can provide such insights.

The approach therefore adopted was to examine how rail trips rates vary with key socio-economic factors and to use the evidence obtained aligned with data on the socio-economic features of different origins and destinations to better explain the propensity to make rail trips in rail demand models.

We have estimated rail demand models to very large data sets over the past 20 years. These models have been enhanced with trip rate evidence relating to age group, employment status, occupation type and household car ownership level.

We also make a number of other improvements to traditional models, such as using NTS evidence on how commuters over time are less likely to use season tickets and allowing employment levels to influence the demand for non-season tickets. We also attempt to address the impact of the digital revolution which some commentators believe is a factor underlying strong demand growth even in recessionary periods.

We will demonstrate the impact on rail forecasting performance using our enhanced models compared to previous models using a range of back-casting methods again based on large amounts of data.

Publisher

Association for European Transport