THE EFFECT OF THINKING TIME ON ATTRIBUTE NON-ATTENDANCE IN DISCRETE CHOICE EXPERIMENTS: AN EYE-TRACKING STUDY
To compare visual attention to attributes in a discrete choice experiment between participants who could progress through choice tasks at their own pace and those who were subject to a minimum “thinking time”.
An efficient design was used to construct 25 generic binary choice tasks for different types of primary care appointments. Attributes included time before appointment, type of health professional seen, length of appointment and convenience. Participants were randomly assigned to either a self-paced group or a group subject to a computer-enforced 25 second delay on each choice task. The eye movements of all 43 participants were tracked during completion of choice tasks using a specialised video camera. Eye-tracking data were analysed for visual attribute non-attendance and then incorporated into mixed logit models of choice data.
Self-paced participants showed significantly higher visual attribute non-attendance compared to computer-paced participants across all attributes (p ≤0.01). Most visual attribute non-attendance was of one level only, and no participant consistently ignored the same attributes over all choice tasks. Differences in visual attribute non-attendance corresponded with differences in preferences between self-paced and computer-paced groups. The length of primary care appointment was a significant influence on choice for computer-paced participants, but not in self-paced participants. When marginal rates of substitution were examined between groups, the importance of seeing a doctor of choice was nearly three times greater for computer-paced participants than for self-paced participants. Models that used eye-tracking data to define participants as visual non-attenders showed improved fit compared to a standard model assuming full attendance.
These results indicate that the most common format for administering discrete choice experiments leads to substantive visual and cognitive non-attendance. Future studies should consider imposing a minimum thinking time to ensure attendance to all attributes. Eye-tracking offers potential to identify and account for attribute non-attendance.