Why choice set size matters

Research in discrete choice modelling techniques has taken for granted the effects of choice set size in discrete choice experiments. Methodologically, the current state of practice is for researchers to strive to minimise the level of cognitive burden in a choice experiment at all costs, primarily with the goal of improving the statistical efficiency of modelling procedures and other measures of model fit. Furthermore, many of these methodological studies seek to make inferences about individual consumer behaviour using between-subjects designs. Concurrently, market researchers and transport planners are interested in how to model consumer behaviour in real world contexts which are characterised by varying levels of choice complexity. Prior research has not answered the question of how choice set size affects consumer decision making processes. Cognitive psychologists and behavioural economists have long argued that decision making processes are driven by two cognitive systems; System 1 and System 2. These two systems describe how decision makers may fail to optimally maximise their utility from seemingly intuitive choices. The research presented in this thesis explores these possibilities using a within-subjects design discrete choice experiment which systematically varies the number of alternatives presented to consumers in a public transport context. Conditional multinomial, mixed, and factor analytic specifications of logistic regression are estimated which show how attribute importance, preference heterogeneity and attribute trade-off strategies change as dependent on choice set size. The factor analytic model presented here is a brand new way to model latent sources of variation in consumers’ choice behaviour that has not been before seen in the literature. The models specified here show that as choice set size changes, the patterns of correlation with a latent factor can change significantly. Implications for theory and practice are discussed