Six reasons why supply‐oriented indicators systematically overestimate service quality in public transport

Abstract

Supply‐oriented measures of quality lead to a systematic overestimate of quality as experienced by travellers in public transport. An example is a train with an average occupation rate for seats being 50%, where, nevertheless, the occupation rate observed by travellers is much higher when some parts of the trajectory are busy. Similar examples are discussed for waiting times at stops, probabilities of arriving in time, probabilities of getting a connection and walking distances to bus stops. A plea is then made for putting more effort in measuring demand‐oriented quality measures.

Acknowledgements

The author thanks Johan van Dalen, Mark van Hagen, Niek van Trigt and Bert van Wee for constructive comments. Also gratefully acknowledged are two anonymous referees for helpful comments.

Notes

A typical example of the ways in which the excess waiting time is incorporated in quality reports is provided by Transport for London. For each route, the scheduled waiting time is given, based on timetables, complemented by excess waiting due to departures from the timetables. For buses, one gets values for scheduled waiting times ranging from 4.0 to 6.0 minutes, with corresponding excess waiting time values from 0.6 to 1.5 minutes. Note that there is a difference in the way bus and train are treated. For buses, excess waiting time incorporates both variations in scheduled headway (e.g. when several routes converge, even if each runs its own regular headway pattern) and day‐to‐day variations in running. For rail, the variations in service headways, as they are used in modelling work, tend to be based on scheduled service patterns. For the benchmarking of railway companies, the variations, however, are based on actual realisations of the timetable (see also the section ‘Arriving in Time’).

The public performance measure is used among other measures for benchmarking purposes by comparing the various railway operators in the UK.

There are some limitations: below certain minimum fare levels (related to reduced fares and/or short distances), travellers will not get compensation.

A related reason might be that direct measurement of passenger flows is expensive, so that in many cases, model‐based estimations have to be made about how many travellers would be in a specific train.

It might sound implausible that timetable coordination has an adverse effect on expected delay in this multimodal trip. The positive effect of the coordination is, of course, that the waiting time at the station in the case of certain arrivals and departures becomes lower. In the present example, it decreases from an expected 15 minutes to 5 minutes. Of course, when not only the timetables are coordinated, but also the actual operations, the adverse effect on delays may be mitigated: when the bus waits for the delayed train, the adverse effect on reliability will disappear. However, slack in timetables, in general, will not allow much flexibility in this respect.

A necessary nuance is that in space, travellers may not be uniformly distributed, so that in certain cases, irregular stop distances might have a favourable effect on walking distances to the stops.

Or put differently, the average distance of points in the circle with radius r to the centre is the weighted mean of the distances between 0 and r, where the weight is proportional to r since the length of a circle is proportional to r. Note that uniform density is assumed.

In some countries, such quality measurements by passengers are part of the contract between government and the transport company that obtained the license.