Visual attention, measured by eye tracking, increasingly attracts behavioral and experimental economists who investigate consumer preferences and willingness to pay (WTP) for products with different attributes. When modeling consumer preferences for product attributes in choice experiments and experimental auctions, the conventional economic approach assumes that consumers process all of the presented information and tradeoff among different product alternatives [1
]. However, recent evidence from marketing literature and consumer neuroscience literature (a growing research area) suggests that consumers apply heuristic decision rules in processing product information. For instance, consumers might not attend to all product attributes, i.e., attribute non-attendance [2
]. Some studies show consumers are more likely to select the choice that attracted their attention/gaze first or last, i.e., gaze cascade effect [20
]. Other studies demonstrate consumers primarily evaluate products with specific attributes that are important to them (i.e., attribute focus or lexicographic choice). These findings implicitly acknowledge that visual attention is a crucial measure that should be taken into account when analyzing consumer preferences [24
There is a new trend of using eye-tracking technology to obtain information on attribute non-attendance in choice experiments. Given that the primary approach when exploring consumer choice and preference heterogeneity is through modeling product attributes, incorporating visual attention into attribute non-attendance (called visual attribute non-attendance by Van Loo et al. (2018) [18
]) has attracted considerable research interest among experimental economists. Meanwhile, other important visual attention patterns that are well documented in marketing literature (such as the gaze cascade effect, centrality bias, and attribute focus) have received relatively little attention in economic fields.
To address this research gap, in this manuscript eye-tracking was combined with non-hypothetical second-price auctions to elicit participants’ preferences and WTP for gardening products in central Florida. In this study, the attribute focus (Hess et al. (2010) [25
] termed it lexicographic choice) and first and last gaze cascade effects were empirically tested while consumer preferences for labels disclosing the presence and absence of specific types of insecticides (neonicotinoids) were examined. The endogeneity issue of visual attention due to unobserved individual characteristics (such as an existing interest in pollinator conservation or concerns about potential environmental impacts of neonicotinoid use) was considered. Lastly, how visual attention patterns may help explain consumer WTP (in the form of bid values) for different neonicotinoid labeling formats and information framing were addressed.
The rest of the paper is organized as follows. In Section 2
, a brief overview of relevant eye tracking studies is provided followed by the hypotheses and study contributions. Section 3
introduces research methodology. It consists of a detailed description of the experimental design including an areas of interest (AOI) explanation, visual attention measures, and data description as well as the theoretical and empirical framework. Results are discussed in Section 4
. The last section concludes by discussing the implications and limitations of the study.
2. Review of Related Literature and Hypothesis
The literature discussed in this section begins with an overview of eye tracking studies, followed by endogeneity issues, knowledge, and participant involvement. Lastly, visual attribute attendance and the gaze cascade effect will be discussed. Throughout the literature review, the relevant hypotheses will be presented.
Eye tracking as an experimental tool has recently gained traction in several disciplines. Eye tracking has been extensively used in marketing literature to address visual marketing strategies related to brands and product attributes [26
]. Numerous studies have shown that visual attention patterns (e.g., fixations and duration) can be predictive of choice since a longer duration or more fixations can lead to higher choice likelihood [20
]. Meanwhile, visual attention patterns can be affected by individual values, which reflect personal preferences. Hence, the degree to which visual attention plays a role in determining how individuals make choice decisions and reveal preferences is of great interest to economists.
Visual attention measurements (e.g., fixation counts, fixation durations) have recently been utilized as explanatory variables in econometric models to explore the correlations between visual attention and individual choice behavior [19
]. However, the visual attention estimate effects on individual choice are likely biased due to potential endogeneity issues related to the visual attention variable(s) [35
]. For instance, if an unobserved personal characteristic affects both the individual’s visual attention and plant purchasing decisions (the dependent variable in this research), the estimated effect of the visual attention variable could be contaminated by the effect of the unobserved characteristic(s) resulting in bias.
Currently, very few studies utilize eye tracking and account for endogeneity. Takahashi et al. (2018) [35
] was an exception who combined eye tracking with a randomized controlled trial (RCT) to explicitly address this issue. Respondents’ (unobserved) environmental concerns and exposure to certification information influenced both their visual attention and purchasing behaviors. Furthermore, a sustainable graphic representation attracted more visual attention and increased purchase likelihood by 22%. This suggests that endogeneity should be acknowledged and addressed (when possible) in studies utilizing eye tracking.
Participants’ existing knowledge has been shown to influence behavior. Frequently, knowledge is measured either as subjective or objective knowledge [36
]. Subjective knowledge reflects what the consumer thinks s/he knows while objective knowledge is their actual knowledge [37
]. Subjective knowledge is quantified using a scale where the respondent indicates his/her self-perceived knowledge while objective knowledge is frequently revealed using quiz questions [37
]. In general, consumers who report high subjective knowledge frequently exhibit low objective knowledge but are quick to project their self-perceived knowledge on the topic [39
]. Previous research suggests that both types of knowledge potentially aid in lessening the endogeneity issue and should be considered in studies [37
Beyond knowledge, participant involvement with topics related to the study’s focus likely also impact their behavior. Involvement with products has been shown to impact consumer perceptions, product purchasing behavior, motivations, comprehension, and visual attention to relevant information [41
]. Specifically related to eye tracking, several studies have found that consumers’ visual attention increases for stimuli that is directly related to their interests [43
]. One measure of involvement in pollinator health is conservation activities. Previous studies have addressed WTP for pollinator conservation at a national level rather than viewing current household conservation-related activities. For instance, studies estimate that UK and US consumers are willing to donate money or pay taxes to support pollinator conservation [46
]. Yet, studies highlight the importance of household pollinator conservation activities to increase habitat availability and foraging opportunities [49
]. The participation in conservation activities at the household level may indicate increased interest and concern with overall pollinator health. This implies that including household pollinator conservation activities could impact behavior but has yet to be included in empirical analysis.
Given that endogeneity can bias results, the inclusion of additional explanatory variables is one means of addressing this issue (i.e., endogeneity caused by missing variables). Knowledge and involvement (measured by participation in pollinator conservation activities) largely reflect individuals’ increased interest in pollinator health and concerns about neonicotinoids. To date, the two factors have not been combined in empirical analysis but may provide valuable insights into consumer behavior. One could argue that the two factors are related in that someone who is more knowledgeable is likely to be more involved than someone who is less knowledgeable. Hence, the following hypotheses were tested:
Visual attention is endogenously determined by individual knowledge and activities related to neonicotinoid insecticides and pollinator conservation that reflect individual interests or concerns about pollinator conservation.
A higher level of knowledge about the potential impact of neonicotinoids will lead to greater visual attention to the associated attribute (e.g., labels disclosing the absence or presence of neonicotinoids).
A higher level of involvement with pollinator conservation activities will lead to greater visual attention to the associated attribute (e.g., labels disclosing the absence or presence of neonicotinoids).
More recently, eye tracking has been used to obtain information on attribute attendance/non-attendance in choice experiments [17
]. These studies attempt to establish a direct link between visual attention and attribute attendance/non-attendance and further infer the relationship between visual attention and individual preference. For instance, Krucien et al. (2017) [53
] showed that the relationship between visual attention and individuals’ preferences depends on the product attribute type. Preferences for “harder to process” attributes were more influenced by changes in visual attention relative to “easier to process” attributes. Behe et al. (2014) [45
] determined that visually attention increases on information that is important to consumers. On the other hand, Balcombe et al. (2017) [52
] and Van Loo et al. (2018) [18
] reported weaker relationships between visual attention and individual preferences since accounting for visual attribute non-attendance did not significantly influence WTP. Balcombe et al. (2017) [52
] revealed that visual attention data was not a good indicator of individual stated attendance. Eye tracking data was useful in understanding attribute non-attendance but less informative with regard to WTP. Similarly, Van Loo et al. (2018) [18
] concluded that not all visually ignored attributes were truly ignored based on the observed respondents’ choice behaviors and using visual attention to identify attribute attendance was not trivial.
While it is relatively new in the field of economics, the use of eye tracking has been widely applied in the fields of marketing and psychology. Studies have consistently shown subjects pay more attention to chosen alternatives than non-chosen alternatives [10
]. Hess and Hensher (2010) [10
] found that consumers attached less importance to certain attributes and persistently selected the alternatives that had better specific attributes. Meiϐner et al. (2016) [57
] showed that high-valued alternatives or important attributes received increased attention. Hess et al. (2010) [10
] termed this pattern as lexicographic choice, and Meiϐner et al. (2016) [57
] framed it as an alternative focus/attribute focus and investigated how attribute and alternative focus potentially affected subjects’ final choice.
In parallel, information literature has recognized the positive-negative asymmetry in human behaviors when evaluating the impact of information on consumer valuations of a good [58
]. Numerous studies confirmed that negative information has a stronger impact on consumer behavior than positive information [60
]. Building upon the asymmetry of the information effect (i.e., negativity effect), the present study investigates the link between attribute focus and individual bidding behavior by testing the symmetry of attribute focus. Specifically,
Greater attention on attributes communicating positive environmental information (e.g., label disclosing the absence of neonicotinoids) will increase consumer’s WTP.
Greater attention on attributes communicating negative environmental information (e.g., label disclosing the absence of neonicotinoids) will decrease consumer’s WTP.
In addition, Shimojo et al. (2003) [22
] proposed the concept of gaze cascade effect described as an increased likelihood that observers’ gazes were directed toward the chosen object. In subsequent studies, Changizi and Shimojo (2008) [21
], Simion and Shimojo (2006) [23
], and Atalay et al. (2012) [20
] demonstrated a link between observers’ final gaze and the chosen object. Meanwhile, Reutskaja et al. (2011) [64
], Fisher and Rangel (2014) [65
], and Meiϐner et al. (2016) [57
], showed that the feature on which participants first fixated was more likely to be chosen. These results suggest that the first and last gaze might influence choice. In this study, gaze cascade effects were tested by investigating whether the first and last gaze fixation on a label increased or decreased participants’ WTP for that attribute.
Consumer’s first gaze fixation on a particular attribute (i.e., label) will increase (decrease) the consumer’s bid value and WTP if the attribute communicates positive (negative) environmental information.
Consumer’s last gaze fixation on a particular attribute (i.e., label) will increase (decrease) the consumer’s bid value and WTP if the attribute communicates positive (negative) environmental information.
By testing the above-mentioned hypotheses, this manuscript contributes to the literature in several ways. First, hypotheses 1 and 2 verify if consumers’ visual attention patterns are endogenously determined by testing the influence of visual attention, knowledge, and pollinator conservation activities on bid values. Secondly, the manuscript addresses if increased visual attention to important attributes (such as neonicotinoid labels) carries through to the final decision (hypotheses 3a and 3b). According to the noncompensatory processing model in consumer product choice literature, consumers either screen out alternatives that have an undesired feature/attribute or consider only those that have a desired feature/attribute [57
]. By evaluating visual attention to more/less desirable features, one can assess how this information influences bidding. Visual attention provides useful information regarding participants’ decision-making behavior; however, it is not sufficient to identify whether the information was ignored or entered into the decision-making process [18
]. In other words, the relationship between visual attention and consumer preferences may be weak [52
]. Here, the link is investigated. Lastly, using a left-right paradigm (in contrast to horizontal paradigm), evidence addressing the gaze cascade effect on participants’ valuation is explored. Currently, Atalay et al. (2012) [20
], Gilbride and Allenby (2004) [66
], and Meiϐner et al. (2016) [57
] all found initial gaze had minimal effects on product choice when items were placed in a horizontal paradigm.
Incorporating visual attention measures in consumer preference studies could provide useful information regarding participants’ decision making and bidding behavior. In general, participants valued labels disclosing the absence of neonicotinoids and were willing to pay a higher price premium for those products. Participants did not differentiate information formats when a plant was grown with neonicotinoids. This finding offers some interesting implications for the ongoing debate about mandatory disclosure of neonicotinoids on labels in the U.S. Given consumers’ positive response to products free of neonicotinoids, one anticipates a movement away from neonicotinoid use along the supply chain in the ornamental horticulture industry. In addition, policy makers may also consider a prudent policy approach by encouraging voluntary disclosure of the absence of neonicotinoids on labels since consumers prefer neonicotinoid free products.
However, the link between visual attention and consumers’ attribute valuation and ultimate decisions is not necessarily strong. This study demonstrated that visual attention patterns were endogenous to individual knowledge about neonicotinoids and their engagement in pollinator conservation activities. The dichotomous groups exhibited distinct visual attention patterns in the process of determining their bids. While participants who self-reported as knowledgeable paid more attention to neonicotinoid labels, not knowledgeable participants paid more visual attention to the image of the (entire) product. Even though individual knowledge and activities about neonicotinoid were found to have no significant impact on participants’ WTP, leaving them in the error term would likely bias the estimated effect of visual attention on individual choice. This occurred because visual attention patterns were clearly correlated with individual knowledge and pollinator conservation activities, which reflect individual interests or concerns about pollinators. Further, policy makers may consider increasing public awareness of neonicotinoids and engaging consumers as part of a large pollinator conservation movement, which will likely change consumers’ attention on information, thus their shopping behaviors.
There was some evidence of an asymmetric impact of attribute focus on consumers’ WTP. The negative impact of greater visual attention to negative features (i.e., treated with neonicotinoids) was evident in the present study. Specifically, even though consumers may have visually attended to both positive and negative information, it was the negative information that carried through to the final decision, leading to a stronger impact of negative information on consumer behavior identified by in information literature [60
]. This finding provided supporting evidence for negativity effects of information.
This finding may provide some insights on noncompensatory processing [66
]. The results point to the possibility that the participants may have screened out product alternatives with undesired features. This process may play a larger role than strategies that consider alternatives with desired features in noncompensatory processing. First and last gaze cascade effect was confirmed by a significant negative impact of participants’ first fixation to labels disclosing the presence of neonicotinoids on their bid values. A first fixation on neonicotinoid labels decreased participants’ WTP for labels disclosing the presence of neonicotinoids (i.e., treated with neonicotinoid), but had no impact on their WTP for labels disclosing the absence of neonicotinoids (i.e., neonicotinoid-free text and logo).
Further, the results provide additional support to Balcombe et al. (2017) [52
] that the relationship between visual attention and preference could be weak. Participants may have visually attended to some (important) attributes; however, this information may or may not necessarily enter into their decision making process. On the other hand, it is possible that participants may have visually ignored certain (important) attributes which were actually used in their decision process. Van Loo et al. (2018) [18
] recently acknowledged that it was more challenging to quantify attribute non-attendance using eye-tracking and to further infer the impact of visual attribute non-attendance on preference.
There are several limitations to the present study. The endogeneity of visual attention data was approached using proxy variables to factor out the unobserved individual environmental interests from the error term. Further research could find valid instrument variables (IVs) for visual attention measures and use a two-stage estimation method such as 2SLS. Secondly, in contrast to previous findings (e.g., [35
]) indicating that consumers place less value on the appearance of a food label logo, we found participants were willing to pay a higher price for a logo relative to the neonicotinoid-free text. Caution is needed when interpreting the logo price premiums. The “Better Bee Certified” logo does not explicitly display the information that is presented by the neonicotinoid-free attribute. Higher premiums for this logo could result from participants’ broader interpretation of this logo. Lastly, to test the first and last gaze cascade effect and circumvented the central fixation bias, the position of information (i.e., labels) was pre-set in the left-right paradigm. It is likely that the position and the content of the information jointly affect visual attention patterns, but we were not able to separate them in this study. Future studies may consider having control and treatment groups by switching the positions of labels to investigate which plays a larger role. Beyond these limitations, the results call for more attention to experimental design when using eye tracking and have implications for future studies to determine how to adequately use visual attention data in understanding its relationship with individual decision making.