Stated preference analysis of travel choices: the state of practice
read more
Citations
Stated Preference Approaches for Measuring Passive Use Values: Choice Experiments and Contingent Valuation
Incentive and informational properties of preference questions
The economics of urban transportation
A comparison of stated preference methods for environmental valuation
Forecasting new product penetration with flexible substitution patterns
References
Discrete Choice Analysis: Theory and Application to Travel Demand
Conjoint Analysis in Consumer Research: Issues and Outlook
Conjoint Analysis in Marketing: New Developments with Implications for Research and Practice:
Related Papers (5)
Frequently Asked Questions (17)
Q2. What are the future works in "Stated preference analysis of travel choices: the state of practice" ?
There are many challenges still to be faced in making the existing set of tools both more user-friendly and capable of assisting in the resolution of further issues emanating from state of the art research.
Q3. What methods can be used to estimate the response variable?
Approaches B, D and F can use regression based estimation methods such as generalised least squares because the response variable is continuous.
Q4. What is the main consideration when evaluating new alternatives?
When new alternatives are being evaluated, making the attribute levels believable (and deliverable) becomes a primary consideration.
Q5. How did the model be used to predict market share?
The modelling of this data using standard regression-based estimation procedures required simulation of choiceenvironments in order to predict market share.
Q6. What is the way to scale the variance of the unobserved effects in the SP?
To scale the variance of the unobserved effects in the SP component relative to the RP component, a sequential or a simultaneous scaling approach can be used.
Q7. What scale is used to represent an underlying continuous distribution of interval scaled rates?
Analysts typically select a 5 or 10 point scale (and occasionally 100 points), to represent an underlying (i.e. latent) continuous distribution of interval scaled rates.
Q8. What is the way to preserve a large number of design attributes?
One way of preserving a large number of design attributes is to partition the attributes into generic groups, with each group defined by elemental attributes, and to design a number of linked hierarchical experiments (e.g. Hensher 1991, Louviere and Gaeth 1987, Hague Consulting Group 1988, Kroes and Sheldon 1988, Timmermans 1988).
Q9. What software is readily available for logit modelling?
Specialised software is readily available for logit modelling such as ALOGIT (Hague Consulting Group, the Netherlands), PCLOGIT which superseded BLOGIT (Institute of Transport Studies, University of Sydney), NTELOGIT (Intelligent Marketing Systems - Canada and Econometric Software - Australia) and HLOGIT (ITS, Sydney).
Q10. What is the popular view of the view that individuals are more capable of ordering alternatives than reporting?
Rank order (non-metric) data is popular with analysts who subscribe to the view that individuals are more capable of ordering alternatives than reporting, by a rating task, their degrees of preferences.
Q11. What is the advantage of the direct translation of the choice responses into predictions?
Choice responses are directly translated into predictions, through the application of discrete-choice models such as multinomial logit (MNL), and are also relatively easier for the respondent.
Q12. What is the way to reduce the information loss in a choice metric?
In recognition of this information loss, a number of studies have investigated ways of maximising the information content of a response metric while both maintaining the ability of the respondent to handle a more difficult task and have the capability of estimating a model which can provide useful predictive outputs in the form of market shares (and attribute elasticities) (e.g. Elrod et al 1992, Ben-Akiva et al 1992).
Q13. Why was the interest in the "New" approach to travel behaviour modelling slow?
Widespread interest in this "new" approach to travel behaviour modelling, however, was slow in developing, in part due to the high agenda interest in the development of discrete-choice models and activity approaches to the study of the continuous sequences of human actions over a period of time (see Hensher and Stopher 1979)1.
Q14. What is the way to use ratings data in the derivation of choice probabilities?
A preferable approach to utilising ratings data in the derivation of choice probabilities is to treat the observed ratings as a non-linear rating scale in an ordered response model which defines points on the observed rating scale as thresholds (Henry 1982, Winship and Mare 1984, Crask and Fox 1987).
Q15. What is the way to limit the task to the first preference?
If the request for ranking or rating responses may jeopardise the cooperation acrosss the replications of the experiment, it is more important to limit the task to the first preference choice.
Q16. What is the procedure used to translate rank order data into choice observations?
A procedure proposed by Chapman and Staelin (1982) for translating rank order data into choice responses, referred to as 'rank explosion', enables one to translate the full depth of R ranks into R-1 choice observations.
Q17. What is the probability of a decision maker selecting an alternative out of the available set of alternatives?
The probability of a decision maker selecting an alternative out of the available set of alternatives is defined as the probability that the observed and unobserved indirect utility of an alternative is greater than or equal to the observed and unobserved indirect utility of each and every other alternative in the choice set:Probj = Prob{(Vj+εj) ≥ (Vj ′+εj ′); j∈J; j≠j ′}Particular assumptions on the distribution of the unobserved effects within the sampled population lead to a particular functional form of the discrete choice model (see below).