Measuring travel time reliability and risk: A nonparametric approach

Abstract

Abstract The reliability of travel time significantly affects individual travelers’ decision-making behaviour and hence in turn influences the overall travel demand at the macroscopic level. The travel time reliability ratio (TTRR), defined to be the ratio of the value of travel time variability to the value of travel time, is an important metric for measuring such reliability. In this paper, we first show that the TTRR is closely related to a widely used risk measure in financial economics, i.e. conditional value at risk (CVaR). Then based on the connection between the TTRR and CVaR, we develop a nonparametric approach to estimate the TTRR. In the literature, to compute the TTRR, it usually needs to assume a specific statistical distribution for the travel time. This can produce a misleading result when this assumption goes awry due to the potential complexity of travel time distributions. Based on the relationship between the TTRR and CVaR, this paper proposes a new nonparametric method, i.e. the kernel density estimation method, to overcome this problem. We show that this new nonparametric method is robust in terms that it does not depend on any assumptions about the shape of the travel time distribution. The simulation studies demonstrate that the proposed method outperforms the existing methods and substantially improves the numerical accuracy. Finally, a practical example is used to illustrate the proposed method.