Abstract/Summary

For many years, economic models have assumed that people make choices by paying full attention to all the information describing their choices. These economic models have offered a good approximation of individual decision-making and were shown to be successful at predicting people’s choices. More recently, economists have come to agree that limitations in the domain of human attention are not only economically important but relevant in examining how people make choices. Recent technological advances that gather data about the information that people acquire during their decisionmaking have facilitated the incorporation of human cognitive limitations into economic models. A tool which has become available for economists is mouse-tracking technology. While arguably providing an important source of economic data, mouse-tracking has had a low uptake in the economic literature more generally and in the choice modelling literature more specifically. Consequently, little is known about its potential to enrich data coming from Discrete Choice Experiments. Moreover, mouse-tracking tools such as Mouselab have been criticized for potentially interfering with participants’ behaviour due to their occluded design, but there is no direct empirical evidence to prove this. To address these gaps, this thesis investigates the potential of using mouse-tracking in economic research as a tool to gather additional insights into human behaviour. The specificities of Mouselab, which involve imposing a cognitive cost to participants, also provide the context to empirically examine the relevance of the Rational Inattention theory in the context of a hypothetical survey applied to nutritional labelling. This thesis models the choices that people make under additional cognitive costs imposed by mouse-tracking. The data were collected using an online Discrete Choice Experiment (DCE) with embedded mouse-tracking technology. The DCE asked respondents to make choices between different food baskets as described by the UK’s Traffic Light System for food labelling. Participants made their choices in two different and subsequent treatments: a classical DCE where all attributes were visible, and a mouse-tracked DCE where most attributes were hidden. Inference about the preference parameters of respondents was conducted using a Bayesian approach. Key findings are that mouse-tracking does not appear to interfere in a significant way with choices made as part of a DCE. Willingness-to-Pay (WTP) estimates from the mouse-tracked DCE were correlated with WTP estimates in the classical DCE and a model merging data from the classical and hidden experiments had a higher predictive validity than the models that treated each experiment separately. Mouse-tracking appears to provide additional and useful insights into human behaviour in a similar way to eyetracking. However, limitations in relation to the size of mouse-tracking data and the complexity of a mouse-tracked experiment need to be recognised. Mouse-tracking data also appears to confirm previous research in relation to how consumers value and use nutritional information: the amount of attention spent on a nutrient is weakly related to how that nutrient is valued. Therefore, while tools that register and quantify attention can contain useful information about people’s preferences, such methods should be used cautiously when attempting to make a connection between attention and value.