Random utility (RU) theory is the most popular framework for the modeling of transport demand. RU models simulate the choice of a decision maker among a set of feasible alternatives (choice set); their operational use requires that the analyst is able to specify correctly the choice set for each individual. Early applications of RU models essentially ignored this problem, assuming that all decision-makers chose from a pre-specified choice set. This assumption may be unrealistic in many practical cases and cause significant misspecification problems (Stopher, 1980; Williams and Ortuzar, 1982). Therefore, in many practical cases, the choice set formation should be simulated as well as the choice within the choice set. Different approaches have been proposed for the choice set simulation (Twersky 1972; Mansky, 1977; Swait and Ben Akiva 1987; Ben Akiva and Boccara 1995; Cascetta and Papola 2001). Most of these models are based on prespecified constraints or specific attributes concerning the individual or the alternative and follow either a deterministic or a probabilistic approach. The purpose of this paper is twofold. On the one hand, it proposes an operational model to introduce the concept of dominance among alternatives in the framework of choice set and random utility modeling. This concept which has long been recognized in collective choice models has been overlooked in individual choice modeling. The paper also proposes operational destination choice models including dominance variables, taking into account also spatial effects such as mental maps and intervening opportunities, never included in random utility theory models. In more detail, the general framework consists in defining dominance variables to be used in probabilistic choice set simulation models; these variables are constructed by specifying a rule to define comparable alternatives, a rule to define a dominated (or dominant) alternative between a pair of comparable alternatives and a method of using such information. Concerning the destination choice context, different possible specifications of dominance variables were proposed and different choice set simulation models estimated on available data. Estimation results show a generally high significance of dominance variables and a considerable improvement in the goodness of fit statistics of the models. For the covering abstract please see ITRD E135207.
Samenvatting