No abstract

Traditional household travel surveys ask respondents to report their travel behaviour for a 24-hour period, although it is well known that travel patterns vary from day to day. While this provides an indication of average household behaviour, or travel on an average weekday, evidence suggests this may not be the most cost-effective way to collect the data because day-today travel variability is substantial, requiring larger sample sizes. In addition, collecting multiday data provides a richness of information that simply cannot be captured with a 1-day survey -offering insights into, for example, differences in weekday versus weekend travel, the longerterm travel impacts of flexible work hours, and trip substitution and cycling patterns that emerge over the course of a week or more. Despite the intuitive appeal of multiday surveys, there are few examples and little information on sampling issues and sample size requirements. With this in mind, this paper explores the reasons given for not doing multiday surveys (which centre around respondent fatigue), why these issues are fast becoming irrelevant (through the use of new passive data recording technologies), the sample size implications of extending a survey, and the potential for estimator efficiency and cost savings when conducting multiday surveys (even accounting for the cost of new technologies). Using GPS-based travel distance data from Adelaide, Australia, we find that the reductions in sample size are large, and that collecting multi-day data is feasible -offering a richness not available in one-day data, along with cost-effective gains in estimator efficiency. (248 words)

This paper analyzes the variability of travel behavior by day of the week and among individual persons, using data from three waves of a panel in South Australia. A previous analysis had been based on the first wave of this panel. The analysis herein provides further insights into the potential benefits of using multiday data for modeling travel behavior and opens up the potential of undertaking a variety of analyses that are not currently possible with 1-day data. As in the first analysis, this paper examines the effect of additional days of measurement on such measures as the number of daily trips, the travel time per trip and per day, and the travel distance per day and per trip. The study analyzes the change in the means of each of these values as additional days of data are considered and also the changes in the estimates of the variances of each statistic. This study also separates out the effects of weekdays and weekend days and considers the implications of these data on sample sizes for modeling purposes and the likely effects on the goodness of fit of models built on multiday data versus the more traditional 1 day of data.