Modelling the effect of transit supply and price structure on mode choice and route choice

Document Type

Conference Paper

Publication Date


Subject Area

place - europe, mode - car, mode - mass transit, ridership - modelling


Sweden, work trips, car, transit, modelling


Regional traffic models used for transportation planning in Sweden, such as the national modelling system Sampers, use different multinomial logit models to forecast travel demand and modal split for different trip purposes. However they do no account for the fact that the value of time varies within a population of travellers making a trip with the same purpose or the fact that the price can differ between different transit lines (bus, regional trains, etc). The cost for transit trips in these models does, as a rule, only depend on the origin and destination of the trip. This makes analyses of the impact of changes in headway, price structure and alignment of route not entirely reliable. This paper develops a new mode choice and transit route choice model for work trips. Two mode choice models, one binomial logit (BNL) and one mixed binomial logit (MXL) model with a lognormally distributed cost parameter, have been estimated using revealed preference data for work trips in Stockholm County during 2005. The mode choice models include the car and transit modes. While the BNL model implies a single value of time, the MXL specification makes it possible to capture some of the variation in the value of time. The transit route choice model rests on the assumption that transit commuters purchase travel passes that are valid for a certain time period, e.g. a month. The travel pass then allows the traveller to use a certain set of transit lines, while others may not be available. The traveller will chose the travel pass that maximizes the traveller's utility given his/hers value of time. For the mode choice, the traveller compares travel cost and time with the chosen pass with the travel cost and time by car. Using the newly developed models, three different methods for calculating the number of car and transit work trips and the distribution of the transit trips on different transit routes have been tested using the EMME software for network assignment. The first method (BASE- BNL) implements the BNL model for mode choice and uses the same algorithm for transit assignment as in the conventional regional traffic models in Sweden where all transit lines are available and have the same cost given origin and destination. The second method (PASS-BNL) includes a deterministic choice of travel pass based on the estimated common value of time in the population and on the cost and travel time with each pass. For the transit assignment only the lines associated with the selected transit pass are available. The third method (PASS-MXL) builds on PASS-BNL, but uses a continuous distribution for the value of time instead of the average value in the population. For the testing of the three methods an area consisting of the central parts of Stockholm County and the neighbouring city of Uppsala was chosen. In the tests, changes were made regarding alignment, headway and price for one of the major transit lines. The results show that all three methods give similar changes of the transit mode share. However, the changes in the number of boardings on the transit line undergoing changes differed substantially between the methods, especially BASE-MNL gave different results that the other two methods. Also calculated changes in ticket income for the public transport operator and the economic benefits of the changes differed substantially between the three methods. Therefore, the results indicate that if the interest is in overall mode share and overall travel flows the conventional method in Swedish transport modelling will suffice but if the interest is more detailed, for example concerning boardings and ticket income from a certain transit line, or the total benefit of price change, the PASS-MXL (and sometimes the PASS-BNL) method will give more reliable results.


Permission to publish the abstract has been given by AET, copyright remains with them.