International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
Examining Attribute Non-Attendance in Discrete Choice Experiments using a gaze-contingent eye tracking application
Kennet Christian Uggeldahl, Chris N.H. Street, Thomas Lundhede, Søren Bøye Olsen

Last modified: 28 March 2017

Abstract


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 stated non-attendance (SNA) or inferred non-attendance (INA). SNA is based on simply asking respondents a follow-up question regarding which attributes they ignored, if any. Arguably, this method suffers from being a self-reported measure, leaving it susceptible to issues such as inaccuracies in perception and recall, as well as respondents striving to seem consistent. INA is a more sophisticated method of measuring non-attendance, where it is inferred from the econometric models based on the observed choices. Comparisons of the two approaches have found that they might not reach the same conclusions (Scarpa et al. 2013, Mørkbak et al. 2014). Differences in the conclusions based on the two methods have also been explained by the way of asking the non-attendance follow-up question (Alemu et al. 2013), however, it seems like the majority of papers dealing with non-attendance have taken the view that inferring non-attendance is superior to relying on respondents' statements, due in part to some of the issues of SNA discusses above. Nonetheless, the approach of inferring non-attendance is not flawless either, as it rests on the usual modelling assumptions that are subject to uncertainty. Depending on the choice of model, one might find higher or lower rates of non-attendance, depending on the extent to which preference heterogeneity is taken into account (Hess et al. 2013).

A few recent studies have relied on a new approach of measuring attribute non-attendance, one that would seem to overcome some of the problems of the earlier methods by providing an exogenous measure of non-attendance. Eye tracking, i.e. the process of recording what a respondent is looking at, has extensively been used in the psychology and neuroscience literature to understand the decision process underlying choices (see e.g. Glaholt and Reingold (2011) and Orquin and Loose (2013) for recent reviews), and has recently been used to measure visual non-attendance (VNA) in DCEs (Balcombe et al. 2015; Spinks and Mortimer, 2016; Van Loo et al. 2014; Olsen et al. 2015). Eye tracking could provide an exogenous measure of non-attendance, as visual attention is tightly linked with what respondents are actually processing, i.e. what they are attending to. A rubber band analogy has been used to describe the eye-mind relationship, in that attention precedes eye movements to a given target location. However, eye movements and attention can be dissociated, although this generally takes a certain amount of conscious effort (Rayner, 2009).

As exogenous as the eye tracking measure might seem, it might still suffer from a number of issues. Previous literature has shown that viewers can get the gist of a scene from a single brief exposure, during which it would be impossible to move the eyes (Castelhano and Henderson, 2008). Furthermore, the range of the visual field around the fixation from which information can be acquired, has been shown to be wider when viewing a scene compared to when reading. Recent studies suggest that objects in a scene can be identified up to four degrees around the location of the fixation (Henderson et al., 2003), equal to almost 5 cm on a 22 inch screen viewed from 60 cm. In choice sets where color or other illustrations are used to represent the attribute levels, these features could cause the VNA measure to falsely classify some attributes as not attended to, even though the respondent has acquired the information without fixating on it (Olsen et al. 2015).

We designed the choice sets in a way that avoids the possible issues discussed above, with the aim of making the VNA measure as precise as possible. We do this in two ways: Firstly by using words to describe the attribute levels in the choice sets and spacing both the alternatives and the attributes sufficiently far away from each other with respect to the functional field of view as well as measurement error. Secondly, for half of our respondents we utilize a gaze-contingent set-up, where the attribute level information is shown only when the respondent is fixating on it, and masked otherwise. With the gaze-contingent set-up we further tighten the link between VNA and “true” non-attendance; 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.

In one of the largest DCEs incorporating eye tracking so far, a total of 293 respondents completed an incentivized food choice experiment about chocolate. By defining areas of interest based on the on-screen location of each attribute in the choice tasks, we are able to objectively assess whether or not each respondent has visually attended to the presented information in the choice sets. We will present and compare models and WTP estimates where non-attendance is accounted for using SNA, INA and VNA. Different definitions of VNA, as well as the difference of these between the two treatments (gaze-contingent and control) will be tested. In addition to this, the two treatments will be tested for differences in a range of descriptive statistics (focusing particularly on eye-tracking related measures) as well as for differences in error variance. The results could be important in both validating the SNA and INA methods against an exogenous measure of attribute non-attendance, as well as providing insights into the VNA measure itself.  The dataset has been collected very recently, so the final analysis has yet to be done.

 

References:

Alemu, M. H., M. R. Mørkbak, S. B. Olsen and C. L. Jensen (2013). "Attending to the reasons for attribute non-attendance in choice experiments." Environmental and Resource Economics 54(3): 333-359.

Castelhano, M. S., & Henderson, J. M. (2008). “The influence of color on the perception of scene gist.” Journal of Experimental Psychology: Human perception and performance, 34(3), 660.

Glaholt, M. G. and E. M. Reingold (2011). "Eye movement monitoring as a process tracing methodology in decision making research." Journal of Neuroscience, Psychology, and Economics 4(2): 125.

Henderson, J. M., C. C. Williams, M. S. Castelhano and R. J. Falk (2003). "Eye movements and picture processing during recognition." Perception & Psychophysics 65(5): 725-734.

Hess, S., Stathopoulos, A., Campbell, D., Gibson, V. and Caussade, S. (2013). "It's not that I don't care, I just don't care very much: confounding between attribute non-attendance and taste heterogeneity." Transportation, 40(3): 583-607

Mørkbak, M. R., Olsen, S. B., and Campbell, D. (2014). "Behavioral implications of providing real incentives in stated choice experiments." Journal of Economic Psychology, 45: 102–116

Rayner, K. (2009). "Eye movements and attention in reading, scene perception, and visual search." The quarterly journal of experimental psychology 62(8): 1457-1506.

Scarpa, R., R. Zanoli, V. Bruschi, and S. Naspetti. (2013). “Inferred and stated attribute non-attendance in food choice experiments”. American Journal of Agricultural Economics, 95(1): 165-180.

Olsen, S. B., Uggeldahl, K., Jacobsen, C., & Lundhede, T. H. (2015). ”Measuring Attribute Non-Attendance in Stated Choice Experiments using statements, inference and eye-tracking–Does presentation format matter?” In International Choice Modelling Conference 2015.

Orquin, J. L. and S. M. Loose (2013). "Attention and choice: A review on eye movements in decision making." Acta psychologica 144(1): 190-206.

Van Loo, E. J., R. M. Nayga Jr, H.-S. Seo and W. Verbeke (2014). Visual Attribute Non-Attendance in a Food Choice Experiment: Results From an Eye-tracking Study. 2014 Annual Meeting, July 27-29, 2014, Minneapolis, Minnesota, Agricultural and Applied Economics Association.


Conference registration is required in order to view papers.