R-Tresis: developing a transport model system for regional New South Wales

This paper sets out a demand modelling framework for the development of a regional transport and land use model system (R-Tresis), to be implemented for New South Wales (Australia). Traditionally, the focus of such a model system has been major metropolitan areas such as Sydney, where we have developed Tresis (Hensher, 2002). Given the growing concern about regional accessibility to many service classes, there is a need for a modelling capability that can be used to prioritize and guide policy decisions in regions that are often described as remote, rural, low density and small town. In developing a framework that is capable of integrating both demand and supply elements of transportation and land use activity, we recognized the challenges in developing primary data sources, and the high likelihood of a reliance on secondary data sources. This suggested an alternative approach to demand modelling that was not dependent on choice models; namely a suite of continuous choice models in which we capture the actual activities undertaken by each mode on both the demand and supply side at high spatial resolution.     

Evaluating the Behavioural Performance of Alternative Logit Models: An Application to Corporate Takeovers Research

Abstract:  Econometric models involving a discrete outcome dependent variable abound in the finance and accounting literatures. However, much of the literature to date utilises a basic or standard logit model. Capitalising on recent developments in the discrete choice literature, we examine three advanced (or non-IID) logit models, namely: nested logit, mixed logit and latent class MNL. Using an illustration from corporate takeovers research, we compare the explanatory and predictive performance of each class of advanced model relative to the standard model. We find that in all cases the more advanced logit model structures, which correct for the highly restrictive IID and IIA conditions, provide significantly greater explanatory power than standard logit. Mixed logit and latent class MNL models exhibited the highest overall predictive accuracy on a holdout sample, while the standard logit model performed the worst. Moreover, the analysis of marginal effects of all models indicates that use of advanced models can lead to more insightful and behaviourally meaningful interpretations of the role and influence of explanatory variables and parameter estimates in model estimation. The results of this paper have implications for the use of more optimal logit structures in future research and practice.     

A comparison of responses to single and repeated discrete choice questions

According to neoclassical economic theory, a stated preference elicitation format comprising a single binary choice between the status quo and one alternative is incentive compatible under certain conditions. Formats typically used in choice experiments comprising a sequence of discrete choice questions do not hold this property. In this paper, the effect on stated preferences of expanding the number of binary choice tasks per respondent from one to four is tested using a split sample treatment in an attribute-based survey relating to the undergrounding of overhead low-voltage electricity and telecommunications wires. We find evidence to suggest that presenting multiple choice tasks per respondent decreases estimates of expected willingness to pay. Preferences stated in the first of a sequence of choice tasks are not significantly different from those stated in the incentive compatible single binary choice task, but, in subsequent choice tasks, responses are influenced by cost levels observed in past questions. Three behavioural explanations can be advanced – weak strategic misrepresentation, reference point revision, and asymmetric value learning. The evidence is contrary to the standard assumption of truthful response with stable preferences.     

Cost thresholds,cut-offs and sensitivities in stated choice analysis: Identification and implications

Within the discrete choice literature, there is growing recognition that some respondents do not process all attributes when evaluating their choice outcomes. Worryingly, the cost attribute is often among those attributes that are likely to be ignored by respondents. We use probabilistic decision process models (similar in form to latent class models, but where we define the classes to describe specific heuristics) to facilitate situations where respondents adopt cost thresholds and cut-offs. We further develop this model to address the potential confounding between preference heterogeneity and processing heterogeneity by simultaneously allowing for a segmentation of respondents based on their sensitivities to cost. Results, based on an empirical dataset on the existence value of rare fish species in Ireland, provide further confirmation that a share of respondents did not attended to cost. Importantly, however, when heterogeneity to cost levels is accounted for the inferred incidence of complete non-attendance is markedly lower, to the extent that when cost thresholds and cut-offs are also accommodated it almost disappears. This modelling approach leads to significant gains in model fit and has important implications for welfare analysis.     

Recovering costs through price and service differentiation: Accounting for exogenous information on attribute processing strategies in airline choice

The entry of low cost airlines has thrown out a challenge to all airlines to find ways of attracting passengers, through a mix of fare discounting, greater frequency, improved flight times and no-frill's levels of on-board service. These competitive strategies have an impact on cost recovery. As airlines seek business in an increasingly heterogeneous passenger market, a greater understanding of what matters to potential passengers in choosing an airline grows in importance. Traditional studies of passenger airline choice assume that all attributes matter, but some to a lesser extent. What happens to the empirical evidence on willingness to pay when specific attributes are totally ignored by particular passengers? In this paper, we examine the impact of individual-specific attribute processing strategies (APS) on the inclusion/exclusion of attributes on the parameter estimates and behavioural outputs of models of airline service and fare level choice. Modelling practice assumes that whilst respondents may exhibit preference heterogeneity, they employ a homogenous APS with regards to how they process the presence/absence of attributes of stated choice (SC) experiments. We demonstrate how information collected exogenous of the SC experiment on whether respondents either ignored or considered each attribute of the SC task may be used in the estimation process, and how such information may be used to provide outputs that are attribute processing strategies segment specific.     

HEV Choice Models as a Search Engine for the Specification of Nested Logit Tree Structures

This paper illustrates how the heteroskedastic extreme value (HEV) model can serve as an effective search engine for identifying appropriate tree structures in hierarchical choice models, particularly the nested logit. This use of the HEV model exploits its ability to estimate unique variances, and hence unique scale parameters, for each alternative in a choice set. The analysis of variance can reveal tree structures that may not be obvious to analysts who tend to base their search strategy on intuitive tree structures. The reliance on behavioural intuition may miss out on the identification of the 'best' tree in an econometric sense. This note illustrates how the HEV model is used to search for the hierarchical domain in which a statistically preferred nested logit model is positioned.     

An exploratory analysis of the effect of numbers of choice sets in designed choice experiments: an airline choice application

Stated choice (SC) experiments are increasingly adopted as the empirical source of information on how individuals respond to current and potential travel contexts. The accumulated experience with SC data has been heavily conditioned on analyst prejudices about the acceptable complexity of the data collection instrument, especially the number of profiles (or treatments) given to each sampled individual (and the number of attributes and alternatives to be processed). It is not uncommon for analysts to impose very stringent limitations on the complexity of an SC experiment. A review of the literature suggests that very little is known about the basis for rejecting complex designs or accepting simple designs. In this paper, we develop a complex design as the basis for an SC study, producing a fractional factorial of 32 rows. However, we then truncate the fraction by administering 4, 8, 16, 24 and 32 profiles to a sample of individuals in Australia and New Zealand faced with the decision to fly (or not to fly) between Australia and New Zealand by alternative airlines and fare regimes. Statistical comparisons of elasticities (an appropriate behavioural basis for comparisons) suggest that the empirical gains within the context of a linear specification of the utility expression associated with each alternative in a discrete choice model may be quite marginal.     

Restating modal investment priority with an improved model for public transport analysis

We compare analytically and numerically the optimised performance of different urban public transport modes for three objectives: total cost minimisation, profit maximisation, and welfare maximisation. We find that under optimal operation, buses provide lower waiting time and operator costs, therefore the only possible advantage for rail is providing a lower in-vehicle time cost if trains are faster. Using Australian data, we found that bus rapid transit provides a lower total cost, higher profit and welfare, up to their capacity. Introducing passengers dislike of crowding and train-specific attraction elements might make Heavy Rail the most cost effective mode for high levels of demand.     

Searching for policy priorities in the formulation of a freight transport strategy: a canonical correlation analysis of freight industry attitudes

An efficient and effective freight transport strategy can be aided by early professional contributions from key stakeholders. One broad group who have historically been given limited opportunity to influence the drafting of a freight strategy, are commercial road users and shippers who manufacture and distribute goods. Utilising a data set collected in Australia in 1996 from a sample of organisations involved directly and indirectly in road freight transportation, views were sought on road infrastructure changes, new road infrastructure, non-road infrastructure needs and transport policies. An optimal scaling approach using non-linear canonical correlation is implemented to search for structural relationships between the underlying policy and infrastructure dimensions and the various industry categories. This framework provides a powerful mechanism for identifying differences among stakeholders in terms of their support for or opposition to specific policies. Results reveal the considerable differences in attitudes associated with the component parts of the freight industry.