1. INTRODUCTION\ud Since the late 1960s, transport demand analysis has been the context for significant developments in model forms for the representation of discrete choice behaviour. Such developments have adhered almost exclusively to\ud the behavioural paradigm of Random Utility Maximisation (RUM), first proposed by Marschak (1960) and Block and Marschak (1960). A common argument for the allegiance to RUM is that it ensures consistency with the fundamental axioms of microeconomic consumer theory and, it follows,\ud permits interface between the demand model and the concepts of welfare economics (e.g. Koppelman and Wen, 2001). The desire to better represent observed choice, which has driven developments in RUM models, has been somewhat at odds, however, with the frequent assault on the utility maximisation paradigm, and by implication\ud RUM, from a range of literatures. This critique has challenged the empirical validity of the fundamental axioms (e.g. Kahneman and Tversky, 2000; Mclntosh and Ryan, 2002; Saelensmide, 1999) and, more generally, the\ud realism of the notion of instrumental rationality inherent in utility maximisation (e.g. Hargreaves-Heap, 1992; McFadden, 1999; Camerer, 1998). Emanating from these literatures has been an alternative family of so-called\ud non-RUM models, which seek to offer greater realism in the representation of how individuals actually process choice tasks. The workshop on Methodological Developments at the 2000 Conference of the International Association for Travel Behaviour Research concluded: 'Non-RUM models\ud deserve to be evaluated side-by-side with RUM models to determine their practicality, ability to describe behaviour, and usefulness for transportation policy. The research agenda should include tests of these models' (Bolduc and McFadden, 2001 p326). The present paper, together with a companion paper, Batley and Daly (2003), offer a timely contribution to this research\ud priority. Batley and Daly (2003) present a detailed account of the theoretical derivation of RUM, and consider the relationships of two specific RUM forms;\ud nested logit [NL] (Ben-Akiva, 1974; Williams, 1977; Daly and Zachary, 1976; McFadden, 1978) and recursive nested extreme value [RNEV] (Daly, 2001 ; Bierlaire, 2002; Daly and Bierlaire, 2003); to two specific non-RUM forms;\ud elimination-by-aspects [EBA] (Tversky, 1972a, 1972b) and hierarchical EBA [HEBA] (Tversky and Sattath, 1979). In particular, Batley and Daly (2003) establish conditions under which NL and RNEV derive equivalent choice\ud probabilities to HEBA and EBA, respectively. These findings would seem to ameliorate the concern that the application of RUM models to data generated by non-RUM choice processes could introduce significant biases. That\ud aside, substantive issues remain as to how non-RUM models can best be specified so as to yield useful and robust information in both estimation and forecasting contexts, and how their empirical performance compares with\ud RUM models. Such issues are the focus of the present paper, which applies non-RUM models to a real empirical context.
The results below are discovered through our pilot algorithms. Let us know how we are doing!
- Tversky, A. (1972a) Choice by elimination. Journal of Mathematical Psychology, 9, pp341-367.
- Tversky, A. (197213) Elimination by aspects: a theory of choice. PsychologicalReview, 79 (4), pp281-299.
- Tversky, A. and Sattath, S. (1979) Preference trees. Psychological Review, 86 (6) pp542-573.
No related research data.
No similar publications.