Boarnet paper presented at IATBR
Organized by the Choice Modelling Centre at the University of Leeds, the 14th International Conference on Travel Behaviour Research was held July 19-23, 2015 at the Beaumont Estate in Windsor, UK. On Sunday July 19, 2015 a paper presentation on “Day-to-Day Travel Variability and ‘Optimal’ Duration of Travel Survey: Moving Beyond the Single-Day Convention” by Wei Li, Douglas Houston, Marlon Boarnet, Han Park, Gavin Ferguson was presented.
The abstract is available below, the paper is now available to conference attendees:
Day-to-Day Travel Variability and ‘Optimal’ Duration of Travel Survey: Moving Beyond the Single-Day Convention
Wei Li, Douglas Houston, Marlon Boarnet, Han Park, Gavin Ferguson
As the emphasis of transportation planning was shifting from capacity expansion to management of travel demand, improved knowledge of travel patterns and variability during multiple days became necessary. The single-day travel survey (SDTS), in which each individual or household only reports one day’s travel activity, represents the archetypal duration of data collection among travel surveys conducted at the national, state or regional levels. The basic assumption in the conventional SDTS approach is that travel activities are repetitive from day to day, and if travel is reported for a randomly chosen day out of some longer time period, then an unbiased sample of behavior for that period can be obtained. While previous studies have challenged the SDTS approach, the implication of survey duration on travel parameter estimates is still largely unknown; accordingly, there is little agreement on the “optimal” travel survey duration for accuracy of travel demand modeling. By analyzing a sample of seven-day travel logs from the city of Los Angeles, California (USA) during 2011-2012, we aim to contribute the understanding of intrapersonal day-to-day travel variability (IDTV) and the implication of travel survey duration on parameter estimates.
Our final sample included 2,395 person-days from 352 individual respondents. Our analytical methods included linear regressions and Monte Carlo simulation experiments. Our linear models revealed that a number of factors significantly influenced IDTV, such as gender, age, income, and automobile ownership. However, the observed characteristics could only explain a very small portion of IDTV: 2% for the total trip count, 5% for the private vehicle trip count, 12% for the bus/train trip count, 4% for the walking/biking trip count, and 4% for the duration of walk/biking. There was a tremendous amount of residual IDTV that was hard to explain by observable characteristics; this may pose a risk to jeopardize the validity of results from single-day surveys.
In order to better understand the risk and elucidate the influence of survey duration on travel parameter estimates, we also conducted Monte Carlo simulation experiments, which enabled us to contrast travel variables measured from the seven-day master sample with those from subsamples of a shorter period (one to six days). We found that it is possible for a less-than-seven-day period to generate parameter estimates close enough to the seven-day period; the “optimal” duration for a travel survey may be dependent on specific travel variables to measure and desired accuracy levels. Regarding the total trip count and the private vehicle trip count, three days of survey duration may produce estimates that are very close to the estimates from the seven-day survey. Walking/biking trips may need at least five days of survey duration to produce estimates that are likely (about 90% chance) to approach those from the seven-day survey. Bus/train trips may be subject to the largest degree of uncertainty among all modes in terms of the repetitiveness of travel; therefore, we recommend the seven-day duration for bus/train trips; estimates about bus/train trip frequencies from a less-than-seven-day survey duration may be associated with high risks of bias.
In conclusion, the conventional one-day approach is highly likely to produce biased parameter estimates due to the intrapersonal day-to-day travel variability. We recommend that transportation professionals and policymakers consider shifting from the conventional one-day approach towards the multi-day approach; surveys which aim to reveal trip frequency information comprehensively from a wide range of modes should consider data collection from the seven-day duration.
Conference registration is required in order to view papers.