AbstractThis paper suggests the Burr distribution as a useful statistical model to represent travel time reliability through studies of the day-to-day variability in travel times on journeys in urban areas. This distribution has a flexible shape and the ability to describe the very long upper tails (and hence significant skewness) seen in observed distributions of travel time variations. This result provides a means to study travel time reliability in more detail, and in line with recent research on indices of travel time reliability. The Burr distribution is algebraically tractable, which means that percentile values can be computed directly. In this way various travel time reliability metrics, such as the FHWA Buffer Index and the Delft skewness parameters can be computed from the fitted Burr parameters. This opens the way to the inclusion of reliability as a consideration in economic analysis, using a metric that can better reflect the nature of travel time variability. A case study example, using real world data, illustrates the proposed computational method for the travel time reliability metrics.