Effects of Fuel Price Fluctuation on Activity-travel Behaviour by Transit and Slow Modes: Evidence from a Pseudo Panel Data



Effects of Fuel Price Fluctuation on Activity-travel Behaviour by Transit and Slow Modes: Evidence from a Pseudo Panel Data

Authors

D Yang, H Timmermans, Eindhoven University of Technology, NL

Description

This study aims analyzing the effects of fuel price fluctuations on activity-travel patterns by transit and slow mode.

Abstract

The impact of fluctuation in fuel prices on people’s activity-travel behavior has been a topic of interest in travel behavior research over the past forty years. Several literature reviews have provided overviews of fuel price elasticity on travel demand (Oum et al., 1992; Sterner and Dahl, 1992; Goodwin, 1992; Lee, 1998; Graham and Glaister, 2002, 2004; Holmgren, 2007; Hensher, 2008; Wardman and Grant-Muller, 2011). The main focus of these previous studies has been on the effects of changes in fuel price and income on either the amount of fuel bought or the amount of traffic, measured in terms of vehicle-miles traveled (VMT). Except for personal vehicle miles traveled, elasticities of public transit ridership such as rail systems, bus, heavy rail and light rail have also been explored (Currie and Phung, 2007; Haire and Machemehl, 2007; Kennedy et. al., 2007; Stover, and Bae, 2011; Wadud et.al., 2007; Yanmaz-Tuzel and Ozbay, 2010). These studies have typically estimated aggregate cross-elasticities of public transportation demand with respect to fuel price and compared these elasticities both at the national level and international level. However, little is known about the effects of fuel prices’ fluctuation on transit and slow mode at the individual or household level.

Further, prior elasticity studies involved either time series data, cross sectional data only or pooled time series/cross-section data (usually comparisons of countries). These data have different potential shortcomings. For studies at the country or sub-nation level, time series data are easy to use. However, such data do not closely reflect individual or household adaptation behavior. On the other hand, time-series data are quite difficult to obtain for disaggregate analysis of individual or household travel behavior. As discussed by many researchers (e.g., Hanly et al., 2002), cross-sectional data inherently limit the study of dynamic effects. Moreover, annual data used in prior research limit interpretation as fuel prices change much more frequently.

To contribute to this line of work, this paper utilizes repeated cross-sectional monthly data from a travel survey in the Netherlands to estimate a transit ridership model. This study aims analyzing the effects of fuel price fluctuations on activity-travel patterns by transit and slow mode. Given the importance of dynamics for the problem at hand and the non-existence of panel data in the Netherlands, a pseudo-panel approach is used to estimate a dynamic model of individual travel distance by transit and slow modes.

Starting from early 2002, strong fluctuations in retail fuel prices can be observed in the Netherlands, providing a good opportunity to examine consumer response to changes in fuel prices. The pseudo panel data for this study was created using the 2004-2009 Dutch MON data (travel survey) on individual activity-travel behavior. Cohorts were constructed by grouping individuals from cross-sectional observations according to cohort subdivisions: year of birth, gender and two types of fuel, (diesel and petrol). Then, the averages of each cohort were treated as individual observations in the pseudo panel. Using an activity-based framework, a distinction is made between three categories of activities: compulsory (school and work-related trips), maintenance (shopping, delivery of goods), and leisure (social or recreational trip, tours or hiking). For each category, elasticities of travel distance by transit and slow mode for fluctuations in fuel prices are analyzed.

The model relates travel distance by transit and slow mode to a cohort-specific generation effect, gender, personal net-income per year, number of trips per person, child-ratio, monthly fuel price and living environment (urbanization level of municipality). It is dynamically specified so that the effects on transit and slow mode of changes in these factors can be analyzed over time, providing estimates of both short-run and long-run elasticities. Lags in adjustment of car driving distance to changes in the explanatory variables are specified by one monthly lag effects. It reflects individual adjustment costs and habit persistence. The paper extents previous work on elasticity of travel distance by transit and slow mode by estimating elasticities on monthly data and by considering differences in travel purpose: compulsory, maintenance and leisure. The paper presents the results of an estimation of dynamic transit ridership models for the resulting 28 cohorts.

Publisher

Association for European Transport