The relationship between travel time variability and road congestion

Abstract

During the last few years, increasing attention has been paid to the problem of travel time variability caused by congestion on urban roads. Several countries have decided or are considering to include the cost of travel time variability in their cost-benefit analyses, including the United Kingdom, Holland and Sweden. Travelers’ valuations of travel time variability has been investigated in several studies. However, studies that try to provide quantitative methods to forecast travel time variability (or rather, effects on variability of various investments or policy measures) are still rather scarce. The present study is an attempt to develop such a method. Using data from Stockholm’s automatic camera system for travel time measurements, the relationship between congestion levels and travel time variability is investigated. We show that there is a stable and typical relationship between the relative standard deviation of travel time (standard deviation divided by travel time) and the relative increase in travel time (travel time divided by free-flow travel time). Using this, we estimate a function that can be used to predict how changes in congestion affect the standard deviation of travel time. This function has been used together with standard traffic assignment software (emme/2) to forecast the socioeconomic benefit of reduced travel time variation caused by a proposed road investment (the western bypass in Stockholm). This paper also investigates the distribution of travel times across days, given a certain road and a certain time of day. It turns out that when congestion gets severe, travel times are approximately normally distributed. For low levels of congestion, the travel time distribution is skewed: most days, the travel time is roughly equal to the free-flow travel time, while it for a few days is longer – perhaps due to incidents or random disturbances.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 21 citations exist upstream.