Does change in respondents’ attention affect willingness to accept estimates from choice experiments?

This study conducts a discrete choice experiment to estimate turfgrass producers' willingness to accept (WTA) values using different logit models and specifications to capture respondents' attention. We first estimate the mixed logit model and a generalized multinomial logit model with and without eye-tracking variables to demonstrate the importance of accounting for individuals' differing levels of attention during an experiment. Our study finds that marginal WTA values are biased when individuals' attention changes are not properly accounted for in the model specification. This finding leads to our second objective, to determine whether attention changes can be fully captured in the absence of eye tracking data by testing six alternative model specifications. All six models are able to detect learning and fatigue effects but are unable to fully capture changes in attention. Of the six alternative models tested, the two models that implement panel data offer more reliable and significant results, suggesting the type of data and model specification used may play an important role in diagnosing attention changes when compared to various heterogeneity models.