Estimating the Potential Modal Split of Any Future Mode Using Revealed Preference Data

Mode choice behaviour is often modelled by discrete models, in which the utility of each mode characterized mode-specific parameters reflecting how strongly that depends on attributes such as travel speed and cost, a constant value. For new modes, function models are not known difficult to estimate basis stated preferences data/choice experiments cannot be estimated revealed preference data. This paper demonstrates data can used model without using constants parameters. establishes method analyze any demonstrated OViN 2017 dataset with trips throughout Netherlands multinomial nested logit model. results alternative specific or parameters, rho-squared 0.828 an accuracy 0.758. The from this calculate future modal split shared autonomous vehicles electric steps, leading potential range 24–30% 37–44% when model, 15–20% 33–40% An overestimation occurs due partial similarities between different transport modes It therefore concluded better suited for estimating than To authors’ knowledge, first time steps has been calculated existing unlabelled modelling approach.