Setting up a continuous panel for collecting travelling information: discussion on methodological issues

Abstract

Modelling travel behaviour has always been a major area of concern in transportation research. Since 1950, due to the rapid increase in car ownership and car use in Western Europe and in the US; several models of transport mode, route choice and destination have been used by transportation planners. Some drawbacks of these first models are clearly the focus on individual trips, where the interrelationships (spatial, temporal, intra-household) between trips and their characteristics are ignored. This clearly shows why household travel data is a critical component of the travel-demand forecasting process. The data are typically generated through a household-based survey in which a sample of the population records their travel patterns over a given time period. This information is combined with socio-demographic information about the sample to develop relationships between individual/household characteristics and their observed travel patterns. It seems only logical that the choice of a transport mode can rely heavily on the weather or other seasonal components, which can perfectly be encompassed by a longitudinal survey. However, nearly all household travel surveys that are conducted in the past were “snapshots” of travel behaviour in a region. Therefore, the models that are developed up to now can capture cross-sectional variation, i.e. variation among individual respondents, but no changes to individual or household behaviour over time. In order to anticipate future travel demand, there is need for dynamic models based on longitudinal data. Traditionally, travel data on households and individuals is collected every five years in Flanders on two thousand five hundred households that have to report their travel behaviour for two consecutive days. This means that if we want to investigate the changes from one wave to another, we can only speak in terms of moving averages over five years. Therefore, the idea was suggested of setting up a continuous panel on a smaller amount of people, that are asked to report their travel behaviour (in relation to the other household members) for two randomly chosen days each week by means of an internet-based survey. In this way, we can have an initial idea about the changes in travel behaviour much faster. However, this smaller sample might not be a good representation for the people of Flanders, since more than 50% of the people over fifty do not have a computer at home. Although everybody is granted access to the internet (at the library), we may assume that this will cause bias. We suggest to use administrative data that are collected by the government, such that the results of the sample can be weighted in order to be more representative. The aim of this paper is to discuss some of these methodological issues that will arise if one transfers from a multiple cross-sectional method to a more continuous approach of collecting data. Some of the issues that will be addressed are the survey method, the sample rotation and refreshment sampling and the handling of non-response. This will prove to be a very important issue in this type of research, since filling in the questionnaire has proved to be quite burdensome in previous cross-sectional studies. So drop-outs and non-response will need to be dealt with in an appropriate way. Future research will mainly focus on the analysis of the data, since we will have to deal with data that are clustered in households as well as correlated over time.