Thresholds and indifference in stated choice surveys

Abstract

One typical aim of choice experiment designs is utility balance, that is, the alternatives defined within each choice set should have similar choice probabilities; otherwise, choice is too easy and little information about preferences may be obtained. Therefore, in a good design respondents may often find themselves close to indifference and thus perception thresholds may be an issue. We propose a discrete choice model to examine the behaviour of individuals with indifference thresholds, i.e. that would make them perceive two or more alternatives as almost identical in stated choice (SC) experiments. Such thresholds may be stochastic, differ among the population and even be a function of socio-economic characteristics and choice conditions. Two estimate this model we need SC data including an “indifference option”, so that respondents are not forced to choose when finding that the two alternatives are equally attractive. Our formulation allows estimating the parameters of a threshold probability distribution using information about choices. As an illustration, the model is applied both to synthetic and real data; results clearly show that when indifference thresholds exist, using models without them can lead to errors in estimation and prediction.