Eye Tracking to Model Attribute Attendance

The literature on choice experiments has been dealing with ways to refine preference elicitation from subjects and predictive power of models. Technological advances such as eye tracking has improved our understanding on how much of the attributes and attribute levels presented to participants is being considered in the decision making process in these kind of experiments. This study investigates subjects’ degree of attendance to attributes and how it influences their choices. The amount of time the subjects spent observing each attribute, relative to all available information on each choice set is used to estimate the attribute attendance. This indicates the revealed attendance to the attributes in the experiment. A simple econometric approach compares the parameter estimates from revealed attribute attendance adjusted models using data from an eye tracking device and a model endogenously inferring the probabilities of using information from each attribute in the choice. The results show that the assumption that participants use all the available information to make their decisions produces significant differences in the parameter estimates, leading to potential bias. The results also illustrate that model fit and predictive power is greatly increased by using revealed attendance levels using eye tracking measures. The most significant improvement however, is to endogenously infer attribute attendance; even more so with revealed attendance indicators.