Examining Attribute Non-Attendance in Discrete Choice Experiments using a gaze-contingent eye tracking application

The Behavioral Economics group at Department of Food and Resource Economics (IFRO) invite to open seminars with a range of subjects within behavioural economics.

Kennet Christian Uggeldahl, PhD student at Department of Food and Resource Economics, University of Copenhagen, will give a presentation on Examining Attribute Non-Attendance in Discrete Choice Experiments using a gaze-contingent eye tracking application, joint work with Chris Street, Thomas Lundhede, and Søren Bøye Olsen. The seminar might tip a bit towards choice modelling.


When respondents answer discrete choice experiments (DCE), they might choose to ignore part of the information provided to them in the choice sets. This has been labelled attribute non-attendance and is a decision processing strategy which has received considerable attention in the DCE literature. In the majority of the literature dealing with attribute non-attendance, this is measured using either statements made by the respondents, or it inferred from the econometric model. In this study, we use eye-tracking to measure of non-attendance. As exogenous as a non-attendance measure based on eye-tracking might seem, it might still be possible for respondents to acquire information from the visual field around the point of fixation.

In this study, we design the experiment with an aim of making the visual non-attendance measure absolute. For half of our respondents, we use a gaze contingent design, where the point of fixation determines what information the respondents see on screen. This means that if the respondent has not fixated on the attribute, the information has not been visible at all, and thus it cannot have been attended to. The data consists of 293 respondents answering an incentivized food choice experiment about chocolate.

Our results show two main conclusions. Firstly, even though respondents have not looked at an attribute (and it thus has not been shown to the respondent), this still seems to impact their choices. This result is rather troubling, and we have yet to figure out what might cause this. Secondly, even though respondents look at an attribute, this does not necessarily mean that the attribute has any importance for them when making their choice. This is still work in progress, and the final analysis is yet to be done.

Bring your own lunch, and we provide knowledge and network.

For further information, contact: Catrine Jacobsen, cj@ifro.ku.dk