Modeling passenger train delay distributions: evidence and implications

This paper addresses the lack of reliability within the Swedish rail network by identifying passenger train delay distributions. Arrival delays are analyzed in detail using data provided by the Swedish Transport Administration, covering all train departures and arrivals during 2008 and 2009. The paper identifies vulnerabilities by size, space and time in the network. Our results show that the delay distribution seems to be plagued by low probability high impact events. A major share of all delay time is associated with the tail of the delay distribution, indicating that extreme delays cannot be neglected when prioritizing between measures improving rail infrastructure. Delays are not only concentrated in size, but also concentrated in space and time and seem to follow a precise power law with respect to days and an exponential distribution with regard to stations. Moreover, we also examine the link between capacity usage and expected delay over different time scales.