To the heterogeneity of which model parameters is the traditional CBA methodology the most sensitive? Can we hypothesize that the VoTTS is the most influential one when certain user preference parameters are correlated with each other?
Travelers may differ in the way they evaluate a wide range of trip attributes such as time loss, travel time variability, crowding inconvenience, other comfort factors, and externalities imposed on fellow residents. Our working hypothesis at the current phase of this project is that variation in the VoTTS parameter is the most influential aspect of heterogeneity, so that allowing for a probability density of VoTTS may have the most significant impact on policy assessment. A shortcoming of his approach is that the input parameters are drawn from a single case study. In the first stage of the present research project our plan is to improve the methodology of Monte Carlo based sensitivity analyzes by introducing correlation between randomly selected input parameters,
Which parametrical distribution describes heterogeneity in VoTTS the best? Can we replace point estimates first with parametrical distributions, and then potentially with non-parametric ones?
This problem statement refers to the mixed logit approach in which the travel time parameter of the discrete choice model’s underlying utility function is replaced with a random variable representing heterogeneity among respondents. The random variable can be specified either with a standard parametric distribution, or as a non-parametric kernel in certain cases. Our goal is to validate this hint based on Hungarian data, and test other specifications, including non-parametric ones that allow less regular shapes for this distribution.
In a public transport context, can we disentangle heterogeneity in VoTTS from the impact of varying degrees of crowding during data collection?
Assume that consumers value travel time homogeneously, but this valuation is affected by the degree of crowding associated with the service that the discrete choice data collection is based on. Under such conditions, we do observe heterogeneity in the value of travel time unless we control for the density of crowding during the experiment. That is, crowding and travel-related time loss may be confounding determinants of travel disutility in the discrete choice framework. However, this threat is normally neglected in the literature of travel time and crowding cost estimation, even if heterogeneity is allowed for in one or the other attribute. This research project could contribute to the literature with the first experiment in which heterogeneity in travel time and crowding valuation are controlled for simultaneously.