International Choice Modelling Conference, International Choice Modelling Conference 2015

Font Size: 
Impulsivity and task complexity in a latent variable choice model applied to environmental valuation
Petr Mariel, Jürgen Meyerhoff

Last modified: 11 May 2015

Abstract


This paper analyzes the issue of choice task complexity represented by five design dimensions (the number of choice sets, alternatives, attributes, attribute levels and their range) on discrete choice model outcomes in the context of environmental valuation. The case study is based on a nation-wide online survey carried out in Germany about land use changes. The survey incorporates aspects of land use changes, such as the share of forests or land conversion as well as different biodiversity attributes. Respondents were requested to value the land use changes in a 15km surrounding of their place of residence. Similar to Hensher (2004), for example, a design of designs approach was used resulting in sixteen different split samples. The effect of task complexity on model outcomes is analyzed in a plain Random Parameter Logit (RPL) and in a Hybrid Choice Model (HCM) which incorporates respondents’ attitudes (Ben-Akiva et al., 2002). HCM is expected to provide better insights into individuals’ choice behavior recognizing their latent nature while at the same time avoiding problems such as endogeneity and measurement error when incorporating these attitudes directly as explanatory variables in choice models.

Apart from that virtue, HCM allows for decomposition of the preference heterogeneity into a purely random part and a part related to attitudes (Vij and Walker, 2012). It allows a deeper understanding of the role of socio-demographics, and, subsequently, better policy recommendations. In our case, the additional information incorporated to the choice model is represented by responses to attitudinal questions related to impulsivity. It is a multifactorial construct that comprises, among other things, a tendency to act on a whim, displaying behavior characterized by little or no forethought, reflection, or consideration of the consequences.

We measured impulsiveness using an item battery  developed by Kovaleva et al (2012). We expect that respondents who reveal a higher degree of impulsivity use different informaion processing strategies and thus, there are more likely to choose the status quo alternative, especially when they face a choice set with a higher dimensionality. Therefore, HCM allows analysis of the interaction between task complexity and impulsivity in a comprehensive hybrid choice model framework.

The HCM applied includes structural equations based on the random utility theory, structural equation for the latent variable and measurement equations which use the values of the attitudinal indicators as dependent variables. The model takes into account the repeated choice nature of the data as well as the ordinal nature of the indicators. 

Our results confirm findings from other studies indicating that respondents’ choices are, apart from the attributes of the alternatives, related to individual characteristics such as attitudes or part of their personality trait. These multifactorial constructs allow for an analysis of the existing preference heterogeneity through the linking of attribute coefficients to socio-demographic variables by the use of these underlying constructs. These results are compared to a plain RPL model in which the same socio-demographic variables affect the means of the random parameters. RPL and HCM outcomes are very similar, nevertheless, as expected, the HCM leads to more precise estimations as additional information on attitudinal indicators is used by the estimation procedure.

The paper raises also some methodological issues related to HCMs. The first issue is related to the definition of the latent constructs used in hybrid models. The definition of these construct should be based on well proved set of questions and, if possible, it should be based on a theoretical background similar to “locus of control” or “environmental beliefs”. The applications should also include exploratory analysis of the indicators which confirm the dimensions of the underlying constructs. The paper decomposes the overall heterogeneity to show what part of it is linked to the latent attitude construct.

The second issue is related to validation of the HCMs. An additional validation apart from simple coefficients testing and LogL comparison is usually not included in DCM applications. The paper highlights the importance of validation and shows a possible way to validate HCMs based on out-of-sample forecasting. 

References:

Ben-Akiva, M., McFadden, D., Train, K., Walker, J., Bhat, C., Bierlaire, M., Bolduc, D., Boersch-Supan, A., Brownstone, D., Bunch, D.S., Daly, A., De Palma, A., Gopinath, D., Karlstrom, A. Munizaga, M. (2002). Hybrid choice models: progress and challenges. Marketing Letters  13(3), 163–175. DOI 10.1023/A:1020254301302

Hensher DA (2004) Identifying the influence of stated choice design dimensionality on willingness to pay for travel time savings. J Transp Econ Policy 38:425–446

Kovaleva, Beierlein, kemper, Rammstedt (2012) Eine Kurzskala zur Messung von Impulsivität nach dem UPPS-Ansatz: Die Skala Impulsives-Verhalten-8 (I-8). GESIS-Working Paper, Mannheim

Vij, A., Walker, J., 2012. Hybrid choice models: holy grail… or not?. In: Paper Presented at the  3th International Conference on Travel Behaviour Research, Toronto.



Conference registration is required in order to view papers.