International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
Choice models with psychosocial attributes. Revealing some facts.
Alejandro M. Tudela, Arnoldo Tapia, Juan Antonio Carrasco

Last modified: 28 March 2017

Abstract


It is getting more common to find discreet choice models that incorporate attributes oriented to describe the person’s behaviour, in terms of personality, attitudes, affection and other individual’s dimensions. There is general agreement that the incorporation of these attributes help to understand and explain behaviour, apart from improving the general goodness of fit. Nevertheless, there is not a systematic approach to collect these data, Researchers tend to create ad hoc instruments to measure these attributes for later inclusion in their models, forgetting or not knowing that there is a plethora of tools and theories from psychology which might help them to collect the appropriate data for the desirable analysis, avoiding some misleading interpretations and conclusions.

 

Interpersonal Behaviour Theory by Triandis is one of the theoretical frameworks which can be used to explain observed conduct. According to this theory, comportment is the interaction result of three aspects:

- Intention to develop a conduct, which rests on attitudes, affections and social factors,

- Habit related to the development and repetition of that conduct in different scenarios, and

- Situational factors, associated with the socio-demography of the individual, the characteristics of the available alternatives and the restrictions the individual might face to develop a conduct, which might influence someone’s final decision, being potentially quite different to the primary intention.

 

Missing some of these individual’s internal and external dimensions might lead to the wrong interpretation, conclusions and forecasting when dealing with choice models. The goal of this paper is to show how the incorporation of these aspects might help us to reveal the effective role of some attributes and factors on transport mode choice. A latent variables approach is used for their incorporation in the discreet choice models, given the nature of the psychosocial explanatory variables,

 

Data collected the years 2011 and 2012 are initially used to carry on this research. This data was gathered in the Collao area, Concepcion district, Chile, corresponding to a mode choice revealed preference experiment. Initial estimations using the modes attributes and some respondent socio-demographic characteristics showed that the two-year responses were subject to different scale factors. Nevertheless, when psychosocial attributes were incorporated, then the scaling effect disappears. Besides, the incorporation of these attributes into the modelling allows obtaining more reasonable coefficients, according to the economics theory and the trips context. The simultaneous incorporation of these psychosocial aspects and some socio-demographic factors, through a MIMIC approach, permitted getting even more plausible models. These different models were used to forecast demand, given certain development scenarios, showing that results are quite sensible to the attributes being considered, affecting the decision making process resting on those results.

 

Ongoing research is oriented to combine the aforementioned data with a second set of information collected the years 2007 and 2008. This data was about revealed mode choice information and was gathered among university workers, utilizing a rather similar questionnaire to the one used years 2011 y 2012. Given the time span between those two data sets, a scaling effect on responses is expected. A step by step modelling approach will be used, looking for the detection of differences between the data sets, and revealing the actual role of the psychosocial and socio-demographic data on mode choice.

 

Hybrid discreet choice models will be estimated. In all cases, a simultaneous method for the determination of coefficients has been and will be used, avoiding the introduction of biases in the parameters due to model estimation. Besides, an endogeneity analysis will be carried out during the modelling such that its presence can be detected and corrected if necessary.

 

Expected results should provide some insights with respect to the role on mode choice of the individual intrinsic characteristics as well as his/her socio-demography, determining their role on the results of the forecasting process.

Conference registration is required in order to view papers.