The appropriate inclusion of attitudes in discrete choice models
This thesis takes a cross-disciplinary approach to investigating preference formation. Using preference theory from economics and psychology, a combined statistical methodology is developed to enable integration and comparison of two techniques for eliciting values and preference for water resource options for Perth, Western Australia. Previous work by Ajzen, Brown and Carvajal (2004) and Spash et al (2006) use a similar approach, however that work is limited in the comparisons which can be made between differing preference elicitation formats, because of the data set used as well as their choice of analytical technique. I propose to use a framework developed in transport economics by Ben-Akiva et al (1999), which incorporates structural equation modelling in a discrete choice model.
In stated preference techniques, which are used in environment, health, marketing and transport economics, the use of latent constructs to explain an economic value estimate (willingness to pay) has been increasing in recent times. Analysis of these constructs has been limited to factor analysis even though they are superior structural equation modelling techniques being used in fields such as psychology.
The idea to use structural equation modelling in stated preference studies was first presented by Ben-Akiva et al (1999) in transport economics for travel demand models. The technique overcomes an issue identified by Train, McFadden and Goett (1987), that using factor analysis to generate latent variables, for inclusion in discrete choice models, produces biased parameter estimates. However it has been overlooked in previous stated preference studies in environmental, as well as other, valuation literature (Burton, 2008) and is yet to be adopted in environmental stated preference studies.