Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/89247 
Year of Publication: 
2013
Series/Report no.: 
Tinbergen Institute Discussion Paper No. 13-209/VIII
Publisher: 
Tinbergen Institute, Amsterdam and Rotterdam
Abstract: 
We compare three stochastic user equilibrium traffic assignment models multinomial probit, nested logit, and generalized nested logit), using a congestible transport network. We test the models in two situations: one in which they have theoretically equivalent coefficients, and one in which they are calibrated to have similar traffic flows. In each case, we examine the differences in traffic flows between the SUE models, and use them to evaluate policy decisions, such as profit-maximizing tolling or second-best socially optimal tolling. We then investigate how the optimal tolls, and their performance, depend on the model choice, and hence, how important the differences between models are. We show that the differences between models are small, as a result of the congestibility of the network, and that a better calibration does not always lead to better traffic flow predictions. As the outcomes are so similar, it may be better to use computationally more efficient logit models instead of probit models, in at least some applications, even if the latter is preferable from a conceptual viewpoint.
Subjects: 
stochastic user equilibrium
traffic assignment
probit
generalized nested logit
tolling
JEL: 
C63
R41
R48
Document Type: 
Working Paper

Files in This Item:
File
Size
794.55 kB





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.