Abstract
Transportation modeling frameworks assume that travelers are economically rational; that is, they choose the lowest-cost alternative to complete a desired trip. The reliability of travel time is of critical importance to travelers. The ability to quantify reliability allows planners to estimate more accurately how system performance influences local travel behavior and to evaluate more appropriately potential investments in the transportation system infrastructure. This paper presents a methodology that makes use of automatic vehicle location data from the regional municipality of Waterloo, Ontario, Canada, to estimate the reliability of transit service. On the basis of these data, the impacts of unreliable service on generalized transit user costs are quantified by use of a simulation model of bus arrivals and passengers' desired arrival times. It is shown that the increasing reliability of arrivals at a station can decrease transit users' generalized costs significantly and by as much as 15% in a reasonably reliable network. It is further posited that the inclusion of uncertainty in the calculation of generalized costs may provide better estimates of mode splits in travel forecasting models. A description of future applications of the model concludes the paper.