How Should the Passenger Rail Market Be Segmented for Demand Forecasting Purposes?
W Wingate, Arup, UK; A Meaney, M Shepherd, Oxera, UK; J Cartmell, A Spencer, Department for Transport, UK
This paper develops a new approach to segmenting the passenger rail market for use in demand forecasting. By applying a series of in-depth techniques, a robust segmentation is proposed for developing new forecasting guidance.
Any assessment of a market?be it for the purposes of understanding how demand is likely to develop over time, for understanding the likely impact of a change in policy, or developing evidence on whether a merger should be allowed to proceed?requires evidence on demand responses. In many applications, these demand responses are represented using the concept of elasticity?which reflects how demand will change for a given change in a demand driver (fares, journey times, incomes, etc).
However, elasticities will be different for different products, and in different geographies. Currently, in the Passenger Demand Forecasting Handbook (PDFH), which presents demand forecasting guidance for the British passenger rail market, the market is segmented according to ticket types and a number of geographic markets. It is not necessarily the case, though, that guidance in the Handbook is provided consistently, or, more importantly, that the market segmentation has been consistently estimated.
In terms of products, the segmentation used is broadly consistent through the Handbook, with elasticities provided for first and standard class travel, for tickets booked in advance, for off-peak tickets, and for season tickets. However, the definitions of products in these categories has changed through time, particularly with the increasing use of yield management by train operators. In addition, a product-based segmentation is not necessarily consistent with the needs of some users of the Handbook, especially policymakers.
Geographical segmentation in the Handbook has, in many cases, not changed in the two decades or more of its development. As a result, geographical markets are often inconsistently defined, and do not necessarily reflect today?s market, which has changed considerably in the intervening period.
In light of these issues, one of the key aims of a study currently being undertaken by Arup and Oxera for the UK Department for Transport, the Passenger Demand Forecasting Council, and Transport Scotland is to provide empirical evidence on market segmentation in the passenger rail market. This paper presents the approaches used and initial recommendations on a revised market segmentation.
The revised geographic market segmentation is based on a four-pronged approach:
? An initial market assessment, reviewing trends and correlations in the rail market to provide hypotheses for an econometric investigation;
? Refining the hypotheses using cluster analysis?where rail flows are clustered together using variables that, ex ante, we consider to impact on elasticities?and then estimating elasticities for each cluster using standard econometric techniques;
? Using 'within model' econometric techniques as an alternative approach to generating clusters, and which also estimate elasticities for each segment;
? Testing the clusters identified using within-sample forecasting, and against expert judgements, previous evidence, and prior beliefs.
Within each geographic market segment identified, the study also provides evidence on elasticities by ticket type and?using a mapping from ticket types to journey purposes?by journey purpose.
In order to develop fit-for-purpose demand forecasting guidance the team will need to trade off model fit against usability, across both geographic and product market segments. There is little point in developing a complex approach to segmentation if it does not then meet the requirements of demand forecasting practitioners. Therefore, the authors will provide a clear description of how final recommendations will be reached based on the models estimated and tested using the above process.
Association for European Transport