Menu
Home
Log in / Register

 
Home arrow Economics arrow Bounded rational choice behaviour
< Prev   CONTENTS   Next >

1.6. Applications

Compared to disciplines such as marketing, the development and application of semi-compensatory and non-compensatory models of choice behaviour has received much less attention in transportation research and similar disciplines. Foerster (1979) was one of the first studies, comparing the performance of additive, lexicographic, maximum, conjunctive-additive and conjunctive-lexicographic choice processes in the context of transportation mode choice decisions in Chapel Hill, North Carolina. The collected data contained information about mode choice, perceptions of speed and costs of car and transit for different trips and importance ratings of time and cost in choosing a transportation mode. He applied the models both at the level of attributes and at the level of subjective ratings. The results of the comparisons, based on Cochran's generalization of McNemar two sample correlated proportions test, indicated that transportation mode choice decisions are not made according to compensatory decision rules. Conjunctive and lexicographic models outperformed additive models. The subjective variants tended to be better than the objective variants.

In contrast, Timmermans (1983) found evidence of underperformance of noncompensatory decision rules compared to additive and multiplicative models in the context of choice of shopping centre in the Netherlands. Choice sets consisted of four to five shopping centres, while 11 attributes were selected.

Several studies have examined the application of hybrid compensatory, noncompensatory models in the context of housing choice. Borgers et al. (1986) modelled the screening phase using conjunctive decision rules, while a multinomial logit model was used to predict housing choice within each choice set. In a similar vein, Kaplan et al. (2012) represented the conjunctive heuristic with a multidimensional mixed ordered-response probit model and the utility-based choice with an error components logit model. In both these studies, the hybrid model performed well. Rashidi, Auld, and Mohammadian (2012) used a very similar two-step approach in which housing alternatives are evaluated and screened based on household priorities, lifestyle and housing preferences. The consideration set is modelled based on household average work distance using a hazard-model (Weibull distribution). A multinomial logit model was used to predict the probability of finding a new residential location, considering the simulated consideration set of the decision-maker. The authors concluded that their modelling approach is capable of generating highly accurate consideration sets. Young and Richardson (1983) addressed the same problem, but applied the Elimination-by-Aspects model. Earlier, they applied this model for modelling freight modal choice (Young, Richardson, Ogden, & Rattray, 1982).

Morikawa (1995) in a study of vacation choice found that a model with probabilistic choice sets, in which the screening stage was based on a set of logistically distributed thresholds outperformed a classic choice model.

Basar and Bhat (2004) modelled airport choice and excluded an airport from the choice set if its utility was below some threshold. They allowed this threshold to vary across individuals, assuming the random threshold to be standard logistically distributed. They found that their model outperformed a classic multinomial logit model for both the estimation and a validation sample.

Cantillo and de Dios Ortuzar (2005) compared their semi-compensatory model with correlated attribute thresholds against the multinomial logit model using two data sets: a simulated data set, and an SP survey on route choice for car trips between Santiago and Valparaiso, Chili. They concluded that fully compensatory models can lead to serious errors in prediction and estimation, and therefore marginal rates of substitution, if individuals do use non-compensatory decision-making processes. Their study indicates that the effect of allowing for correlation between thresholds is marginal. Allowing thresholds to be a function of socio-demographic characteristics and choice conditions improved model performance.

Zheng and Guo (2008) also applied a two-stage model structure, but their problem concerned destination choice behaviour. They assumed that distance constraints define the feasible area for destination choice and that individuals perceive a spatial choice set as a contiguous areas around their home (or trip origin). They used an ordered probit model for the choice set generation stage, with all zones within an individual's distance threshold being included in the consideration set. A multinomial logit model is used to predict the probability that a destination within an individual's consideration set is chosen. The model was estimated to predict the destination choice of shopping, restaurant, and recreational trips in Napa County, California. The authors argue that their model appears to capture behavioural principles that are not accounted for by the MNL model. However, they did not find a significant difference between the two models in terms of goodness-of-fit.

Zhu and Timmermans (2008, 2010, 2011) suggested a model, which they called the heterogeneous heuristic model, which includes several of the mechanisms discussed in this chapter. First, a set of accumulative thresholds is defined for each attribute to vary degrees of filtering and attribute attendance. Next, based on the attributes that are considered, a conventional valuation equation is applied. Third, a further simplifying mechanism is that the outcome utilities are classified, and using an overall threshold, choice alternatives are distinguished into acceptable and non-acceptable alternatives. The process automatically generates different heuristics. The final step is then concerned with the choice of heuristic, which is modelled as a function of the amount of mental effort, risk perception and expected outcome. Variations of this model have been applied to different aspects of pedestrian movement.

 
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
 
Subjects
Accounting
Business & Finance
Communication
Computer Science
Economics
Education
Engineering
Environment
Geography
Health
History
Language & Literature
Law
Management
Marketing
Philosophy
Political science
Psychology
Religion
Sociology
Travel