Simultaneously Accounting for Inter-alternative Correlation and Taste Heterogeneity Among Long Distance Travellers Using Mixed Nested Logit (MXNL) Model So As to Improve Toll Traffic and Revenue Forecasts.
Collins Teye-Ali, Peter Davidson Consultancy, Peter Davidson, Peter Davidson Consultancy, Rob Culley, Peter Davidson Consultancy
In this paper we investigate the potential of using Mixed Nested Logit (MXNL) models to simultaneously account for the above two phenomena and the distribution of willingness to pay for toll roads, using a Stated Preference (SP) dataset from toll route choice experiments conducted during a recent toll route study in Nigeria.
Properly accounting for inter-alternative correlation and taste heterogeneity in a choice process is crucial in capturing the underlying choice behaviour. A model capable of capturing these two phenomena is expected to have high predictive power and more accurate measure of key policy variables such as value of time.
The existence of unobserved inter-alternative correlation is usually accounted for by the use of GEV models with the nested logit model being the modeller's 'workhorse'. However the assumptions underlying these models imply that all sampled individuals have identical choice process (ie are homogeneous). Although each individual makes different choices because of either differences in personal characteristics (e.g., income, age, or gender) or unobserved, but systematic, behaviour (maybe rules) that he/she uses for making judgments about choice alternatives. While it seems desirable to allow different individuals to have different taste parameters, we rarely have sufficient data to analyse behaviour in this level of detail. Even in situations where sufficient data exist; it may not be computationally efficient to estimate different set of parameters for each individual. A practical approach for accommodating this unobserved heterogeneity is to use mixed logit models, in which the taste parameters are assumed to follow some joint probability distribution function with a set of parameters characterising the distribution. This approach effectively allows each individual to have different taste parameters using the distribution. Additionally, the mixed logit model provides the means of estimating the distribution of key policy variables such as willingness-to-pay, rather than simple averages as averages do not tell the full story. This distribution provides maximum information to access drivers’ behaviour under various levels of value of time.
In this paper we investigate the potential of using Mixed Nested Logit (MXNL) models to simultaneously account for the above two phenomena and the distribution of willingness to pay for toll roads, using a Stated Preference (SP) dataset from toll route choice experiments conducted during a recent toll route study in Nigeria. By Mixed Nested Logit model we mean the model which combines the mixed logit with the nested logit estimated simultaneously. Results reveal the presence of both correlation and different taste heterogeneity. The estimation results for the combined mixed nested logit model is presented compared with individual estimation results for nested logit on its own, mixed logit on its own and multinomial logit. We also presented formula and values for elasticities.
This paper is unique in that there does not seem to be much work in using this combined mixed nested logit model approach to understanding long distance travellers’ behaviour in the context of road pricing. This paper opens up this area for investigation and shows the additional explanation we can potentially achieve to improve our models forecasting ability.
Association for European Transport