Sequential Models for the Mobility Decisions: Experimentation for Vehicle Holding Choices
F Russo, G Chilà, DIMET, Mediterranea University of Reggio Calabria, IT
This note presents a vehicle transition choice model, which simulates, according to the sequential approach, the variation of user decisions over time in terms of permanence or transition of the actual system state, considering earlier decisions.
The demand models used in literature, in the field of a behavioural approach, generally simulate the user choices through a discrete choices models.
The consolidated approach isn?t explicitly able to simulate the variation of choice probability, in consequence of the different events, which characterize the evolution of transportation system. So, in this note, we define static demand models the models that give the choice probability of the single alternative, independently of the actual choice of the decision-maker, relative to the actual system condition, and dynamic demand models the models that give the choice probability according to the actual and the previous system condition, then considering evolution system and earlier decisions.
The need to introduce dynamic models regards different main decisions, among which it?s possible to recall: the path choice, for a private transportation system user, and the run choice, for a transit system user, considering the travel choices; the ownership choice and particularly the ownership decision of vehicles, considering the mobility choices.
Predominantly, the models used in this area of research are members of the family of discrete choice models, derived from the random utility theory. This theory is based on the hypothesis that every individual is a rational decision maker, maximizing utility relative to his own choices.
Different discrete choice models are obtained from different specifications of the joint density of the random vector. The models used in literature are: the Generalized Extreme Value (GEV) model, proposed by McFadden (1978); the Probit model (Daganzo, 1979); the ordered GEV (Bhat, 1998) model and the Mixed Multinomial Logit model (McFadden and Train, 1996). The most used from researchers and practitioners is the Multinomial Logit model, which has been introduced by Domencich and McFadden (1975) and formalized by Ben Akiva and Lerman (1985).
These mathematical models have been applied in a way that doesn?t give the possibility to represent adequately the dynamic of the choice process, emphasizing the systematic and invariant aspect of choice behaviour, which are assumed to dominate the choice process (Ben-Akiva et al., 2002). In particular, they don?t represent explicitly the influences exercised by the previous decisions on the actual choice.
In literature, the known models that consider the effects of the previous choices are the car ownership models. Many of these studies (Manski and Sherman, 1980; Train, 1986) simulate the car ownership choices using the Multinomial Logit model and introduce the parameter transaction search cost, in order to consider the influence exercised by the previous choices. Hensher and Le Plastrier (1985) develop a series of linked choice models to explain household vehicle ownership and adjustments in ownership over time. The effects of the experience are considered introducing the brand loyalty and the parameter experience index, which is a summary of retrospective utility. Train and Winston (2004) simulate the vehicle type choice using a Mixed Multinomial Logit model. They represent the effects of the experience through the brand loyalty and the random components, able to consider their endogeneity.
However, these models are based on static structures, that can be defined pseudo-dynamic.
The transition matrix (Gottman and Roy, 1990), which represent the variation of user decisions over time, and the sequential model, which represent the time ? dependencies of path choice (Russo, 1999), has suggested the vehicle transition choice model. According to the sequential approach, the model simulates the permanence or the transition of the actual system state. It is different from the pseudo-dynamic models, because it simulates explicitly the choice of an alternative, in a given period of time, conserving or modifying the choice set relative to the previous period. The time-dependencies are considered introducing some attributes that are function of the past state.
At first, the proposed model has been applied to the holding vehicle, in order to compare the results obtained with the experimentation of the pseudo-dynamic models, used in literature.
The model has been specified, calibrated and validated using:
a database relative to the socio-economic evolution of a sample family, which captures dynamic longitudinal effects;
a database relative to the technical classification of vehicles, defined by a company of car rent;
a database relative to the technical-performances characteristics of vehicles, obtained by a specialized car review published in Italy.
The results obtained from the experimentation of the transition model are presented in the paper with the results obtained through the application of the models used in literature.
The comparison gives favourable index for the transition model in relation to the other experimented models.
Association for European Transport