Learning and fatigue in discrete choice experiments: addressing preference and variance instability

Abstract

In a typical discrete choice experiment respondents are asked to make choices over a series of choice tasks. While the repeated nature of these choices is advantageous from a sampling eciency perspective, discrete choice analysts have, for the most part, neglected the fact that there may be systematic dierences across the tasks as a result of learning and/or fatigue. The few studies that have focused on the topic have assumed that all respondents follow the same pattern of learning and fatigue. Using a latent class modelling approach, we demonstrate that this may not be the case and that patterns of learning and fatigue may actually only be exhibited by a small subset of respondents. Moreover, unlike previous studies that have approached the issue of learning and fatigue from the viewpoint that it influences either the stability in preferences or variances, in this paper we implement a scale-adjusted latent class model to uncover both types of instability simultaneously. Findings from this model highlight the advantages, in terms of model fit and interpretation, that can be achieved when both types of instability are addressed concurrently. We utilise data collected to estimate the existence value of rare and endangered fish species in Ireland as our empirical case-study.