Using Empirical Data to Find the Best Measure of Travel Time Reliability

The value of travel time savings is often the largest benefit from transportation projects and has been studied extensively. Recently, additional attention has been paid to the fact that travelers also benefit from reliable travel times. The value of reliability (VOR) has usually been estimated through stated-preference data or survey-based revealed-preference data. In this research, empirical data were used to estimate VOR. One concern about estimating VOR from empirical data was the lack of a definitive measurement for reliability. Should it be the standard deviation (SD) of travel time, the 95th percentile, or another measure? Data from Katy Freeway in Texas, where travelers chose between tolled but generally more-reliable lanes and free but generally less-reliable lanes, were used in an attempt to find the best measurement of reliability that could lead to the best explanation of travelers’ lane choice. Multinomial logit models were used to estimate travelers’ lane choice on the basis of trip attributes, including travel time, many measures of travel time reliability, and tolls. Models including only travel time and tolls yielded reasonable results and values of time ($2.78/h, $9.09/h, and $10.52/h for off-peak, shoulder, and peak periods, respectively). However, adding reliability to the models caused many of them to have counterintuitive results, and concluding which measure was the best was impossible. In addition, the results of this research suggested that reliability might not be an influential factor in the lane choice decision on managed lanes, at least when travelers had reasonable knowledge of their potential travel time.