A multitrait-multimethod model linking consumers’ priorities and preferences in public transport usage

• Presents a multitrait-multimethod specification of mixed logit. • Sources of preference heterogeneity common and unique to two separate preference elicitation tasks (best worst object case and stated choice) are investigated. • Results show heterogeneity for alike attributes between the two tasks tends to converge on the same source(s) of heterogeneity. • The extent to which best worst and stated choice can be considered interchangeable is discussed. The interrelatedness of stated choice and best worst experiments motivates the core question of the current research; that is, does a shared behavioural process given rise to responses to both preference elicitation methods. The basic premise of the paper is that a common, underlying behavioural process gives rise to the responses of decision makers to both preference elicitation methods; and therefore, the patterns of preference heterogeneity between both methods can be mapped. We investigate this using a multitrait-multimethod specification of mixed logit which has the following two defining characteristics: firstly, the partitioning of the unobserved sources of preference heterogeneity into components that are general to the methods and second, components that are unique to each method. The core findings show that common underlying sources of preference heterogeneity do give rise to decision makers’ responses in both stated choice and best worst experiments (following a specifically theorised stucture). Our primary empirical contribution is – for within subjects deigns where data is generated from the same respondents across two or more preference elicitation methods – that the preference heterogeneity for alike attributes in each of the elicitation methods can be empirically shown to converge on the same source. This remains an assumption in much prior research of this type, but until now, method effects related to the use of different preference elicitation tasks have confounded the interpretation of jointly estimated models that combined attribute priorities and preferences. The implications of our work for transport researchers is that best worst and stated choice experiments can be used in complementary ways – given the empirical support for this core assumption.