Next Article in Journal
The Influence of Structural Conditions and Cultural Inertia on Water Usage and Landscape Decision-Making in a California Metropolitan Area
Previous Article in Journal
Evaluation of Integrated Air Pollution and Climate Change Policies: Case Study in the Thermal Power Sector in Chongqing City, China
Article Menu

Export Article

Sustainability 2017, 9(10), 1743; doi:10.3390/su9101743

Visual Attention to Eco-Labels Predicts Consumer Preferences for Pollinator Friendly Plants
Hayk Khachatryan 1,*Orcid, Alicia L. Rihn 2Orcid, Benjamin Campbell 3, Chengyan Yue 4, Charles Hall 5 and Bridget Behe 6
Food and Resource Economics Department, Mid-Florida Research and Education Center, University of Florida, Apopka, FL 32703, USA
Mid-Florida Research and Education Center, University of Florida, Apopka, FL 32703, USA
Agricultural and Applied Economics, University of Georgia, Athens, GA 30602, USA
Departments of Applied Economics and Horticultural Science, University of Minnesota, St. Paul, MN 55108, USA
Department of Horticultural Sciences, Texas A&M University, College Station, TX 77843, USA
Department of Horticulture, Michigan State University, East Lansing, MI 48824, USA
Correspondence: Tel.: +1-407-410-6951
Received: 4 August 2017 / Accepted: 16 September 2017 / Published: 27 September 2017


Recent declines in pollinator populations have led to widespread concern due to their impact on food/crop production and the environment. Contrary to growing interest in the use of insecticides in urban landscapes, the relationship between pollinator-related eco-labels, visual attention, and preferences for plants is less understood. The present study combines eye tracking and stated preference experiments to examine the effects of pollinator-related labels on consumers’ preferences and willingness to pay. Results show that the pollinator-friendly attribute positively correlates with consumers’ purchasing decisions and visual attention supports that relationship. Implications of mandatory labeling of pesticide content for the horticultural industry are discussed.
conjoint analysis; eye tracking; ordered logit model; ornamental plants; willingness to pay

1. Introduction

Eco-labels are voluntary credence attributes defined by a third party that differentiate products based on their environmental impact [1]. Credence attributes are not directly observable from viewing (i.e., search attributes) or interacting with a product (i.e., experience attributes), and therefore must be communicated at the point of sale to customers. Labeling credence attributes (i.e., eco-labels) allows consumers to carefully weigh all of the products’ traits and select the product that best meets their needs [2]. Currently, 465 eco-labels exist worldwide with 203 in the U.S. alone [3]. Eco-labels increase consumers’ trust, preferences and willingness to pay (WTP) [4,5,6]. However, the extent to which eco-labels impact purchase intent is less understood [4]. Previous literature shows that excessive information in retail settings reduces the effectiveness of point-of-sale signage because customers selectively attend to visual information [7,8].
To account for consumers’ visual behavior, previous studies have used eye tracking technology to address in-store signage [9] and sustainability labels [8]. Visual attention is a key component in consumers’ decision-making processes since information must be visually noticed to influence choice [7,10]. For instance, Reutskaja et al. [7] note consumers often choose the best-seen alternative. Factors that influence consumers’ visual attention during the decision making process can be framed as top-down (i.e., “goal-driven”) and bottom-up (i.e., “stimulus-driven”) [10]. Top-down stimuli are the focus of this study because they dominate when consumers are making choices [11] or are presented with unfamiliar labels (such as pollinator-related eco-labels [12]).
Eye tracking metrics are becoming more prevalent in consumer behavior research addressing eco-labels and sustainability [5,8,13]. For example, Van Loo et al. [8] reported that consumers who value sustainably produced coffee fixate more on sustainability labels (i.e., United States Department of Agriculture (USDA) organic, Rainforest Alliance, carbon footprint, Free Trade) and pay premiums for coffee with those labels. In the ornamental plant industry, Behe et al. [9] found that consumers interested in production methods (sustainable, energy-saving, water-saving) had increased visual attention to the production methods. Rihn et al. [5] determined consumers’ visual attention to sustainable production methods on plants varied by product end use. To date, eye tracking metrics have not been utilized to address consumer behavior toward pollinator-related eco-labels.

2. Background

Recently, pollinator insects have received considerable attention due to declining populations. This is concerning because of their role in crop/food production, the environment, and the economy [13,14]. Pollinator insects pollinate 70% of the world’s food crops [14], resulting in increased food quantity and quality [15]. Gallai et al. [13] determined that, in 2005, pollinator insects contribute 153 billion euros per year (about 195 billion USD) to global food crop production with insect-pollinated crops being valued at 761 euros per ton (about 970 USD per ton). Global food crop supply would not meet current or projected world consumption levels without insect pollination [13]; an issue that will become more critical over time, given expected world population growth. Pollinators also benefit the environment through increased biodiversity, wildlife food availability, and landscape aesthetics and contributed to the prevention of soil erosion and water runoff [15,16].
Despite the importance of pollinator insects, very few studies investigate consumer preferences for pollinator-friendly products but instead focus on the value of pollinator services and/or conservation efforts [17,18,19,20,21]. In 2008, UK households were willing to pay 1.37 pounds sterling per week (cumulatively 1.77 billion pounds sterling per year or roughly 3.5 billion USD) to protect bees [19]. Another study demonstrated that UK consumers would pay 13.4 pounds sterling/year per taxpayer (roughly 21.6 USD/year) to protect pollinator insects and their habitat [17]. U.S. consumers were willing to pay 4.78–6.64 billion USD to conserve monarch butterflies and their habitat [18]. Consumers value conservation measures to aid pollinator insects; however, research at the retail level assessing the impact of pollinator-related eco-labels on consumer behavior are scarce [20,21].
Currently, the U.S. does not have a pollinator-related eco-label. Instead, the promotion of pollinator-friendly products is the responsibility of individual green industry stakeholders (e.g., retailers), resulting in a plethora of pollinator-related labels (e.g., pollinator friendly, neonics-free, bee friendly, and so forth [20]), which may reduce the effectiveness of pollinator eco-labels. Despite this issue, Wollaeger, Getter, and Behe [21] determined pollinator-related labels positively impact consumers’ WTP for ornamental plants. Similarly, Rihn and Khachatryan [20] found that the presence of a pollinator-related label had a positive impact on U.S. consumers’ purchase likelihood for plants regardless of wording. However, the impact of actually viewing the pollinator-related label cannot be determined from these studies. Nor was the presence of additional eco-labels (i.e., production method, origin) factored into the experiments. More information may improve product value by increasing consumer understanding [22], or it may reduce label effectiveness by increasing cognitive load [23]. The current study builds on the assumption that consumers want to aid pollinators [17,18,19,20,21] by incorporating visual attention measures and alternative eco-labels.
To examine consumers’ purchase likelihood and WTP for pollinator-friendly plants, we used a rating-based conjoint analysis (CA) in combination with eye tracking technology. Ordered logit models estimated purchase likelihood while controlling for visual attention and socio-demographic variables. Recently, visual attention metrics have been incorporated into CA. A review article of CA literature by Agarwal et al. [24] recommends using eye tracking technology in CA to explore decision-making strategies. A key assumption of CA is that participants evaluate all available attributes and choose the product that provides them the most utility. Yet several studies have shown consumers selectively attend to relevant attributes to reduce their cognitive load, which influences WTP and utility [12,17,25,26]. For instance, Meißner, Musalem and Huber [11] found consumers focus on positive attributes of the chosen product and negative attributes of rejected products. Previous studies have addressed this selectivity using self-reported attendance measures [27] or latent class models [28]. However, participants often overstate their attendance [27] while latent class models can be problematic in that as the number of attributes increase, the number of latent classes increase exponentially [28]. Eye tracking analysis overcomes the difficulties by accurately recording what attributes participants view, which improves model accuracy, reduces bias, and accounts for attribute non-attendance [8,23,26,29]. Bundesen, Habekost, and Kyllingsæk’s [30] neural theory of visual attention states that consumers are very selective about what information is viewed/used in decision making due to limited cognitive capacity. As a result, the stimulus that are visually attended to are subject to filtering and sorting prior to processing so that only a fraction (approximately 2%) of the visual field is used when making a decision [29,30]. Neural theory of visual attention indicates that more processing (i.e., visual attention) is devoted to important stimuli while reduced processing occurs for less important stimuli [8,12,30]. Here, eye tracking technology is utilized to assist in capturing what attributes consumers visually attend to while making purchasing decisions.

3. Objectives and Hypotheses

Following the conceptual framework that links eco-labels with consumer preferences and WTP price premiums, the overall objective of this study was to assess consumers’ preferences, specifically for the pollinator-friendly attribute on ornamental plants. First, since consumers are willing to pay taxes to fund pollinator insect conservation programs [17,18,19] and are willing to pay premiums for plants with bee-friendly production labels [21], we hypothesize that a pollinator-friendly eco-label will be positively correlated with participants’ purchase likelihood (H1).
Building off hypothesis 1, we hypothesize that if consumers are interested in aiding pollinators through purchasing ornamental plants that are pollinator friendly, one would expect that their visual attention to that attribute would increase. For instance, Van Loo et al. [8] found sustainability-minded individuals fixated on sustainable labels more than those who are not interested in sustainability. Other studies found similar results with consumers selectively viewing important, relevant attributes [11,12,17,25,26]. Thus, hypothesis 2 is that consumers’ visual attendance to the pollinator-friendly eco-label will positively impact consumers’ purchase likelihood (H2).
The rest of the manuscript is organized as follows. The following section provides an overview of the research methodology, followed by the empirical results, and a discussion of implications and limitations of the study.

4. Methods

4.1. Product Selection

Ornamental plants were chosen as the experimental product for several reasons. First, environmental groups have recently targeted the horticulture industry for their use of controversial insect control measures (i.e., neonicotinoid pesticides) that may negatively impact pollinator insects [1,31]. Due to inconclusive research about the risks of neonicotinoid pesticides on pollinator insects [32], the United States Environmental Protection Agency [31] currently does not require labeling of neonicotinoid-based pesticides on plants sold in the U.S. With this in mind, our results may have implications for the commercial horticulture industry in terms of the impact of mandatory labeling of neonicotinoid pesticides on consumer preferences. Secondly, ornamental plants directly impact pollinator insects’ health due to providing nutrients and habitat [17,32]. Therefore, encouraging the sales of plants that aid pollinators could improve pollinator insect health [33]. Lastly, ornamental plants are sold with minimal packaging and branding, which reduces preconceived preferences at the onset of the experiment [34].

4.2. Visual Attendance Metrics

Previously, stated attribute non-attendance has been used to account for consumers’ attribute selectivity; however, discrepancies exist between stated and actual attribute attendance [27]. Eye tracking metrics can be used to remove this discrepancy by accurately recording what consumers view [23,24]. Fixation counts (when the eye is still/focused for 200–500 ms) are one means of measuring visual attendance since fixations are positively correlated with information acquisition and processing [34], attribute importance/relevance to task [9,12], and decision-making [10]. According to Balcombe et al. [29], more than one fixation is needed to constitute visual attendance since the first fixation is random. In this study, two fixations were deemed an acceptable point for the visual attendance metrics. Similar to Balcombe et al. [29], one visual attendance variable was generated per attribute, where 1 equals greater than two fixations and 0 otherwise.

4.3. Experimental Design

A rating-based CA experiment and eye tracking technology were used to assess the impact of the pollinator-friendly attribute on consumers’ purchase likelihood. A rating-based CA was chosen for several reasons. First, a rating-based CA experiment minimized additional visual clutter and isolated the effects of visual attendance to attributes on participants’ purchasing behavior. Secondly, rating-based CA experiments are comparable to the choice-based approach in terms of reliability, validity, and predictive power [35,36,37]. Lastly, rating-based CA and gaze data have been successfully used to investigate consumer purchasing behavior toward ornamental plants [9]. Thus, the rating-based CA approach aligned well with the research goals. The University of Florida’s Institutional Review Board approved all experimental procedures (2014-U-0539).
The CA experiment scenarios were comprised of five product attributes: plant type; price; pollinator friendly; production method, and; origin (Table 1). Plant types (petunia—Petunia × hybrid Juss., pentas—Pentas lanceolata Forssk., and hibiscus—Hibiscus rosa-sinensis L.) were selected after consulting with retail experts to identify the most common landscape plants sold in the research area (Florida). The three price points reflect the low-end prices present in big box stores and high-end prices found in niche/specialty stores (i.e., independent garden centers). Regarding the pollinator-related attribute, in-store labels, and signage varies greatly. To encompass all of the options (i.e., attracts bees, butterflies, etc.), pollinator friendly was selected to represent this attribute. Plants were labeled as pollinator friendly or not rated. Production method options included certified organic, organic production (i.e., produced using organic methods but not certified), and conventional. Following previous studies that incorporate organic and conventional production methods, we added organic production to assess production methods that are between the two extremes (i.e., certified organic and conventional). Regarding origin of production, in-state represented the closest area, followed by domestic, and then imported.
A full factorial design would result in 162 possible product options (Table 1). To reduce participant fatigue, Statistical Package for the Social Sciences (SPSS) software was used to generate a fractional factorial design with 16 product scenarios. Each scenario was displayed on a 24-inch computer screen with 1920 × 1080 pixel resolution. Each scenario consisted of five identical plants with the attributes presented on above-plant signs with consistent font style, size, and color (Figure 1A). Attribute sign location was randomized between the scenarios to eliminate order effect. For each image, fixation counts (FC) were collected for the attributes. In order to collect FC data, areas of interest (AOI) were generated. An AOI is a geometrical outline used to define a specific region within the image (i.e., each attribute) where researchers want to collect the visual metric recordings (Figure 1B).
As participants viewed the scenario images, a Tobii X1 Light Eye Tracker (Tobii Technology, Stockholm, Sweden) recorded their eye movements. The eye tracking camera was stationary and located at the bottom of the monitor. After each scenario, participants indicated their purchase likelihood on a seven-point Likert scale (1 = very unlikely; 7 = very likely). A fixation cross was shown for 3 s between scenarios so all participants fixated on the same location prior to each image. Lastly, the questionnaire questions included sociodemographic questions.

4.4. Sample Demographics

Local newspaper advertisements, Craigslist posts, and printed fliers were used to recruit participants. Fliers were distributed through independent garden centers, public gardens, and community boards. Participants were screened to insure they had purchased plants within the past year. A total of 108 people participated in the experiment, of which, 104 (96% of the sample) successfully had their eye movements recorded. A sample size of 104 was deemed acceptable since comparable studies have used substantially fewer participants [7,12,26]. Participants averaged 53 years of age and had a household income between $51,000 and $60,000 (Table 2). The average household size was two people and 38.5% of participants were male. Most participants were in a relationship/married and indicated that pollinator-friendly promotions would improve their plant purchasing preferences. Florida Census statistics are also provided for comparison purposes [38]; however, statistical inferences cannot be determined due to lack of standard deviation data. But, compared to Florida as a state, the study appeared to overrepresent females, smaller households, and higher incomes, which is consistent with core plant consumers [39]. Specifically, gardeners tend to be female, over 45 years old, college educated, married, in a two-person household, and have a household income of over $50,000 [39].

4.5. Experimental Procedure

At the beginning of the experiment, the Tobii X1 Light Eye Tracker was calibrated to each participant using the Tobii Studio five-point calibration method. Next, instruction slides explained the experimental procedure and defined the attribute levels for participants. Then, to control attributes not included in the study, participants were told all plants were the same size and had the same care requirements. Fixing the unobserved attributes improves confidence that the observed responses are due to differences in the provided attributes [40]. Participants then evaluated an example product to become familiar with the experimental process. The experiment was not timed, and the participants proceeded to the subsequent slides at their own pace in order to eliminate the effects of time pressure on choice and eye movements [7,10]. The experiment lasted approximately 30 min (with 10–12 min spent on the CA experiment) and participants were compensated $30.

4.6. Ordered Logit Model

Purchase likelihood scale was used as the dependent variable. Since the dependent variable was ordinal, an ordered logit model was estimated. Following Long and Freese [41], the model was derived from a measurement model by mapping a latent variable y ranging from − to to an observed variable y . Considering the J number of categories in the ordinal measure, the relationship between observed and latent variable can be shown as:
y i = m   if   κ m 1 y i < κ m   for   m = 1   to   J
where κ s are thresholds (or cutpoint boundaries for each m category) in the distribution of y* that once crossed result in a category change. The extreme categories 1 and J can be represented by the following open-ended intervals κ 0 = and κ J = , which translated into our purchase likelihood rating with seven categories can be shown as:
y i = { 1   2   7     i f i f i f     κ 0 = y i <   κ 1     κ 1 y i <   κ 2     κ 6 y i <   κ 7 =  
Based on the measurement model above, the structural model then can be defined as [41]:
y i = x i β + ε i
where x i is a row vector of values for the i th observation, β is a column vector of structural parameters, and ε is the random error term. To estimate the model using maximum likelihood (ML) method, a specific form of the error distribution must be assumed. Although other distributions were considered in previous research (for example, McCullagh [42]) for the ordered logit model, the ε is conventionally assumed to have a logistic distribution with a mean of 0 and variance π 2 / 3 , with the following probability distribution λ ( ε ) = exp ( ε ) / [ 1 + exp ( ε ) ] 2 , and cumulative distribution Λ ( ε ) = exp ( ε ) / 1 + exp ( ε ) functions.
The assumption of the distribution of the error term allows relating probabilities of outcomes ( y ) given values of x , as shown in the following equation [40]:
Prob ( y i = m | x i ) = Prob ( κ m 1 y i > κ m | x i )
Substituting x i β + ε i for y i in Equation (4) leads to the probability of any observed outcome y i = m given x i to be generalized as the difference between cumulative distribution functions evaluated at any given m values:
Prob ( y i = m | x i ) = F ( κ m x i β ) F ( κ m 1 x i β )
where F indicates the cumulative distribution function. Given the formulation in Equation (5), the probability of observed value of y i (i.e., purchase likelihood rating) for the i th observation can be represented as:
p i = { Prob ( y i = 1 | x i , β ,   κ   )   if   y = 1   Prob ( y i = m | x i , β ,   κ   )   if   y =   m Prob ( y i = 7 | x i , β ,   κ   )   if   y =   7
The likelihood equation can be represented as:
L ( β ,   κ   | y ,   X ) = i = 1 N p i
After multiplying over cases where y is observed to equal j , and taking logs, the log likelihood function becomes [41]:
l n   L ( β ,   κ   | y ,   X ) = j = 1 J y i = J l n [ F ( κ m x i β ) F ( κ m 1 x i β ) ]
Upon estimation of the ordered logit model, the coefficient estimates were used to calculate WTP for attributes using the following equation [42,43]:
WTP =   ( β a t t r i b u t e β p r i c e )

5. Empirical Results

5.1. Visual Attendance and Ordered Logit Model Coefficient Estimates

Table 3 summarizes the total fixation counts and visual attendance metrics. The highest mean fixation and percent attendance was for plant (7.892; 76.3%) and the lowest was for the plant sign (1.752; 28.0%). Other attributes that received more visual attention included import, organic, conventional, and domestic. The medium and low prices and in-state production received less visual attention. The pollinator-friendly and high-price attributes received an intermediate amount of visual attention at 2.832 (47.7% attendance) and 2.911 (attendance 49.2%), respectively.
Model 1 included the bivariate product attribute and sociodemographic variables while Model 2 built on Model 1 by adding the visual attendance variables (Table 4). The lower Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) values and higher Adjusted Pseudo/McFadden R2 value for Model 2 indicates improved fit to Model 1. Across Models 1 and 2 the estimated attribute coefficients are similar in terms of sign and significance. As expected, price was negatively correlated with participants’ purchase likelihood ratings. The positive coefficients for hibiscus and pentas imply that the participants preferred these plants over the base alternative (petunia). The pollinator-friendly attribute increased participants’ purchase likelihood when compared to plants without the pollinator-friendly attribute, supporting the first hypothesis that the pollinator-friendly eco-label will be positively correlated with participants’ purchase likelihood. This finding is consistent with the literature that reported public support for protecting pollinator insects [17,21]. Compared to conventional production methods, participants were more likely to purchase certified organic or organically produced plants, similar to the findings in previous studies [21,44,45]. Regarding origin, in-state and domestic origins improved participants’ purchase likelihood when compared to imported plants. The origin results are substantiated by literature showing consumers value products from closer origins [44], even if they are uncertain about economic or environmental ramifications. Overall, the estimated attribute coefficients suggest the data is robust since the results align with the previous literature.
Several individual sociodemographic variables also impacted participants’ purchase likelihoods (Table 4). In both models, older participants were more likely to purchase the plants than younger individuals. Additionally, participants who had completed a higher level of education were less likely to purchase the plants. The significance of gender, household size and income varied between the models. In Model 1, participants with higher incomes were more likely to purchase the plants. Additionally, as participants’ household size increased, they were less likely to purchase the plants. The coefficient for gender was not significant. In Model 2, men were more likely to purchase plants, while household size and income were not significant.
Unlike Model 1, Model 2 also included visual attendance metrics. All visual attendance results are in comparison to participants who did not visually attend to the attribute. Participants who visually attended to the plant images (Plant attendance) were more likely to purchase the plants than those who did not view the plant images; however, visual attendance to the plant identification sign (Plant sign attendance) was not significant (Table 4). Visual attendance to the low price point (Low price attendance) improved participants’ purchase likelihood, while visual attendance to the medium and high price points decreased purchase likelihood. This result aligns with the attribute coefficient results, which indicate that price and purchase likelihood are inversely correlated. Participants who visually attended to the pollinator-friendly attribute (Pollinator attendance) were more likely to purchase the plant than those who did not, supporting the second hypothesis that consumers’ visual attendance to the pollinator-friendly eco-label positively impacts purchase likelihood. Visual attendance to the certified organic production (Certified organic attendance) attribute was also positively correlated with increased purchase likelihood, while visual attendance to organic production (Organic attendance) decreased purchase likelihood. Visual attendance to conventional production (Conventional attendance) was insignificant. Regarding origin, only the domestic origin (Domestic attendance) was significant and positive.

5.2. Willingness-to-Pay Estimates

Model 1 and 2 estimates were used to generate premiums participants were willing to pay for the different attributes (Table 5). Specifically, participants were willing to pay the highest premium for plants with in-state ($5.96–$6.01) or domestic origins ($4.63–$4.67) when compared to imported plants. Participants were also willing to pay premiums for plants produced using organic production ($4.02–$4.12) or certified organic ($2.98–$3.06) when compared to conventional production methods. The origin and production method WTP results are consistent with previous WTP estimates for origin [44] and organically grown ornamental plants [46], suggesting the reliability of the findings. Desirable plant types also generated premiums at $3.93–$4.05 and $2.39–$2.45 for hibiscus and pentas, respectively. Lastly, participants were willing to pay $1.81 to $1.84 more for pollinator-friendly plants.

6. Discussion, Implications and Limitations

Previous research efforts showed that consumers value pollinator-related conservation measures [17,18,19], but the extent to which pollinator-related promotions attract consumers’ visual attention and influence their behavior remains unknown. The present study contributes to the literature by combining a CA and eye tracking technology to investigate the effect of pollinator-friendly attributes on consumers’ purchase likelihood for ornamental plants. Eye tracking technology was used as an explicit data generation mechanism to analyze participants’ visual attendance to attributes. Previous studies have demonstrated that incorporating visual attendance measures reduces bias results [27,28]. Here, we found statistically significant relationships between attendance and purchase likelihood. Overall, our results indicate pollinator-related promotions improve consumers’ purchase likelihood and generate $1.81–$1.84 price premiums. This suggests there is demand for pollinator-friendly products and that in-store pollinator-related promotions could benefit the green industry supply chain members (growers, intermediaries, retailers). However, it should be noted that other eco-labels also generated positive part-worth utilities (i.e., production method, origin) and further research is needed to better understand the relationships between different types of eco-labels.
An additional contribution of our research is the combination of eye tracking analysis with CA as a means to explicitly measure attribute attendance. Previously, Agarwal et al. [24] suggested pairing the two methods. Similar to Balcombe et al. [29] and Van Loo et al. [8], who used eye tracking in choice experiments, we outlined an approach of combining the two methods and using the data in regression analysis. Results indicate visual attendance positively influences consumers’ purchase likelihood for pollinator-friendly plants, which supports Bundesen, Habekost, and Kyllingsæk’s [30] neural theory of visual attention. Specifically, that more visual processing is devoted to important stimuli (i.e., pollinator friendly), which in turn impacts consumers’ purchasing decisions. This implies visual attention to in-store signage affects consumers’ purchasing behavior and demand [2], and pollinator-related promotions can be used to attract more consumers.
Although the results provide interesting implications, there are several limitations that need to be acknowledged. First, the data was collected using stated preference measures; consequently, it was subject to hypothetical commitment bias. However, the positive and statistically significant estimates for the pollinator-friendly attribute appear realistic given the increased attention to pollinator health [17,18,19,21]. Additionally, while participants control their choices, eye movements are much more difficult to regulate and accurately reflect what participants view [10]. The alignment of the eye tracking measures and the CA estimates suggest reliability. A second limitation was that to facilitate using eye tracking technology, on-site individual participation was required. Therefore, the sample was localized and data was collected in the lab setting using one plant type per image. In the retail setting, multiple products would be in the same visual field at one time as well as additional visual stimuli. The adding of additional visual clutter would likely influence visual attendance measures. In-store trials and retail observations could be used to overcome this limitation.
Overall, our results could assist policymakers, horticultural firms, retailers, and researchers. In the horticulture industry, stakeholders (e.g., growers, retailers, wholesalers, etc.) could utilize the findings as feedback to align their production methods and product offerings with consumer preferences, provided that pollinator-friendly product options are economically feasible. Additionally, since public opinion influences demand and subsequently affects producer welfare, results could benefit policymakers as they determine legislation pertaining to pollinator-related labels (i.e., neonicotinoid-free labeling). Results could also aid future research by providing an additional means (i.e., eye tracking recordings) of handling attribute attendance in CA and choice experiments with more explicit data.


The authors gratefully thank the Florida Department of Agriculture and Consumer Services (FDACS) for funding this project.

Author Contributions

Project design/research question development: Hayk Khachatryan, Alicia L. Rihn, Benjamin Campbell, Charles Hall, Bridget Behe—Data collection: Hayk Khachatryan, Alicia L. Rihn, Benjamin Campbell—Data analysis: Hayk Khachatryan, Alicia L. Rihn—Manuscript preparation: Hayk Khachatryan, Alicia L. Rihn—Editing manuscript: Hayk Khachatryan, Alicia L. Rihn, Benjamin Campbell, Chengyan Yue, Charles Hall, Bridget Behe.

Conflicts of Interest

The authors declare no conflict of interest.


  1. U.S. Department of Agriculture—National Agricultural Library. Sustainable Agriculture: Definitions and Terms. Related Terms. Available online: (accessed on 2 February 2016).
  2. Lusk, J.L.; Fox, J.A. Consumer demand for mandatory labeling of beef from cattle administered growth hormones or fed genetically modified corn. J. Agric. Appl. Econ. 2002, 34, 27–38. [Google Scholar] [CrossRef]
  3. Ecolabel Index. Available online: (accessed on 2 February 2017).
  4. Atkinson, L.; Rosenthal, S. Signaling the green sell: The influence of eco-label source, argument specificity, and product involvement on consumer trust. J. Advert. 2014, 43, 33–45. [Google Scholar] [CrossRef]
  5. Rihn, A.; Khachatryan, H.; Campbell, B.; Hall, C.; Behe, B. Consumer preferences for organic production methods and origin promotions on ornamental plants: Evidence from eye-tracking experiments. Agric. Econ. 2016, 47, 599–608. [Google Scholar] [CrossRef]
  6. Taufique, K.M.R.; Siwar, C.; Talib, B.; Sarah, F.H.; Chamhuri, N. Synthesis of constructs for modeling consumers’ understanding and perception of eco-labels. Sustainability 2014, 6, 2176–2200. [Google Scholar] [CrossRef]
  7. Reutskaja, E.; Nagel, R.; Camerer, C.F.; Range, A. Search dynamics in consumer choice under time pressure: An eye tracking study. Am. Econ. Rev. 2011, 101, 900–926. [Google Scholar] [CrossRef]
  8. Van Loo, E.J.; Caputo, V.; Nayga, R.M., Jr.; Seo, H.; Zhang, B.; Verbeke, W. Sustainability labels on coffee: Consumer preferences, willingness-to-pay and visual attention to attributes. Ecol. Econ. 2015, 118, 215–225. [Google Scholar] [CrossRef]
  9. Behe, B.K.; Campbell, B.L.; Khachatryan, H.; Hall, C.R.; Dennis, J.H.; Huddleston, P.T.; Fernandez, R.T. Incorporating eye tracking technology and conjoint analysis to better understand the green industry consumer. HortScience 2014, 49, 1550–1557. [Google Scholar]
  10. Orquin, J.L.; Mueller-Loose, S.M. Attention and choice: A review on eye movements in decision making. Acta Psychol. 2013, 144, 190–206. [Google Scholar] [CrossRef] [PubMed]
  11. Meißner, M.; Musalem, A.; Huber, J. Eye tracking reveals processes that enable conjoint choices to become increasingly efficient with practice. J. Mark. Res. 2016, 53, 1–17. [Google Scholar] [CrossRef]
  12. Ares, G.; Gimenez, A.; Bruzzone, F.; Vidal, L.; Antunez, L.; Maiche, A. Consumer visual processing of food labels: Results from an eye tracking study. J. Sens. Stud. 2013, 28, 138–153. [Google Scholar] [CrossRef]
  13. Gallai, N.; Salles, J.M.; Settele, J.; Vaissiere, B.E. Economic valuation of the vulnerability of world agriculture confronted with pollinator decline. Ecol. Econ. 2009, 68, 810–821. [Google Scholar] [CrossRef]
  14. Klein, A.-M.; Vaissiere, B.E.; Cane, J.H.; Steffan-Dewenter, I.; Cunningham, S.A.; Kremen, C.; Tscharntke, T. Importance of pollinators in changing landscapes for world crops. Proc. R. Soc. B Biol. Sci. 2007, 274, 303–313. [Google Scholar] [CrossRef] [PubMed]
  15. Hanley, N.; Breeze, T.D.; Ellis, C.; Goulson, D. Measuring the economic value of pollination services: Principles, evidence and knowledge gaps. Ecosyst. Serv. 2015, 142, 137–143. [Google Scholar] [CrossRef]
  16. Wratten, S.D.; Gillespie, M.; Decortye, A.; Mader, E.; Desneux, N. Pollinator habitat enhancement: Benefits to other ecosystem services. Agric. Ecosyst. Environ. 2012, 159, 112–122. [Google Scholar] [CrossRef]
  17. Breeze, T.D.; Bailey, A.P.; Potts, S.G.; Balcombe, K.G. A stated preference valuation of the non-market benefits of pollination services in the UK. Ecol. Econ. 2015, 111, 76–85. [Google Scholar] [CrossRef]
  18. Diffendorfer, J.E.; Loomis, J.B.; Ries, L.; Oberhauser, K.; Lopez-Hoffman, L.; Semmens, D.; Semmens, B.; Butterfield, B.; Bagstad, K.; Goldstein, J.; et al. National valuation of monarch butterflies indicates an untapped potential for incentive-based conservation. Conserv. Lett. 2014, 7, 253–262. [Google Scholar] [CrossRef]
  19. Mwebaze, P.; Marris, G.C.; Budge, G.E.; Brown, M.; Potts, S.G.; Breeze, T.D.; Macleod, A. Quantifying the value of ecosystem services: A case study of honeybee population in the UK. In Proceedings of the 12th Annual BIOECON Conference ‘From the Wealth of Nations to the Wealth of Nature: Rethinking Economic Growth’, Venice, Italy, 27–28 September 2010. [Google Scholar]
  20. Rihn, A.; Khachatryan, H. Does consumer awareness of neonicotinoid insecticides influence their preferences for plants? HortScience 2016, 51, 388–393. [Google Scholar]
  21. Wollaeger, H.M.; Getter, K.L.; Behe, B.K. Consumer preferences for traditional, neonicotinoid-free, bee-friendly, or biological control pest management practices on floriculture crops. HortScience 2015, 50, 721–732. [Google Scholar]
  22. Teisl, M.F. What we may have is a failure to communicate: Labeling environmentally certified forest products. For. Sci. 2003, 49, 668–680. [Google Scholar]
  23. Spinks, J.; Mortimer, D. Lost in the crowd? Using eye-tracking to investigate the effect of complexity on attribute non-attendance in discrete choice experiments. BMC Med. Inf. Dec. Mak. 2016, 16. [Google Scholar] [CrossRef] [PubMed]
  24. Agarwal, J.; DeSarbo, W.S.; Malhotra, N.K.; Rao, V.R. An interdisciplinary review of research in conjoint analysis: Recent developments and directions for future research. Cust. Needs Solut. 2015, 2, 1–22. [Google Scholar] [CrossRef]
  25. Hensher, D.A.; Rose, J.M. Simplifying choice through attribute preservation or non-attendance: Implications for willingness to pay. Transp. Res. Part E Logist. Transp. Rev. 2009, 45, 583–590. [Google Scholar] [CrossRef]
  26. Vidal, L.; Antunez, L.; Sapolinski, A.; Gimenez, A.; Maiche, A.; Gaston, A. Can eye-tracking techniques overcome a limitation of conjoint analysis? Case study on healthfulness perception of yogurt labels. J. Sens. Stud. 2013, 28, 370–380. [Google Scholar] [CrossRef]
  27. Kragt, M.E. Stated and inferred attribute attendance models: A comparison with environmental choice experiments. J. Agric. Econ. 2013, 64, 719–736. [Google Scholar] [CrossRef]
  28. Scarpa, R.; Gilbride, T.J.; Campbell, D.; Hensher, D.A. Modelling attribute non-attendance in choice experiments for rural landscape valuation. Eur. Rev. Agric. Econ. 2009, 36. [Google Scholar] [CrossRef]
  29. Balcombe, K.; Bitzios, M.; Fraser, I.; Haddock-Fraser, J. Using attribute importance rankings within discrete choice experiments: An application to valuing bread attributes. J. Agric. Econ. 2015, 65, 446–462. [Google Scholar] [CrossRef]
  30. Bundesen, C.; Habekost, T.; Kyllingsbæk, S. A neural theory of visual attention: Bridging cognitiona neurophysiology. Psychol. Rev. 2005, 112, 291–328. [Google Scholar] [CrossRef] [PubMed]
  31. US-EPA. Colony Collapse Disorder: European Bans on Neonicotinoid Pesticides. United States Environmental Protection Agency. Available online: (accessed on 9 January 2015).
  32. Fairbrother, A.; Purdy, J.; Anderson, T.; Fell, R. Risks of neonicotinoid insecticides to honeybees. Environ. Toxicol. Chem. 2014, 33, 719–731. [Google Scholar] [CrossRef] [PubMed]
  33. McIntyer, N.E.; McIntyer, M.E. Effects of urban land use on pollinator (Hymenoptera: Apoidea) communities in a desert metropolis. J. Appl. Theor. Biol. 2001, 2, 209–218. [Google Scholar]
  34. Collart, A.J.; Palma, M.A.; Carpio, C.E. Consumer response to point of purchase advertising for local brands. J. Agric. Appl. Econ. 2013, 45, 229–242. [Google Scholar] [CrossRef]
  35. Pieters, R.; Warlop, L.; Wedel, M. Breaking through the clutter: Benefits of advertisment originality and familiarity for brand attention and memory. Manag. Sci. 2002, 48, 765–781. [Google Scholar] [CrossRef]
  36. Elrod, T.; Louviere, J.J.; Davey, K.S. An empirical comparison of rating-based and choice-based conjoint models. J. Mark. Res. 1992, XXIX, 368–377. [Google Scholar] [CrossRef]
  37. Green, P.E.; Srinivasan, V. Conjoint analysis in marketing research: New developments and directions. J. Mark. 1990, 54, 3–19. [Google Scholar] [CrossRef]
  38. U.S. Census Bureau. Quick Facts. Available online: (accessed on 10 February 2015).
  39. National Gardening Association. The Impact of Home and Community Gardening in America; Butterfield, B., Ed.; National Gardening Association, Inc.: South Burlington, VT, USA, 2009; pp. 1–17. [Google Scholar]
  40. Johnson, R.D. Making judgements when information is missing: Inferences, biases, and framing effects. Acta Psychol. 1987, 66, 69–82. [Google Scholar] [CrossRef]
  41. Long, J.S.; Freese, J. Regression Models for Categorical Dependent Variables Using STATA; A Stata Press Publication, StataCorp, LP: College Station, TX, USA, 2006. [Google Scholar]
  42. McCullagh, P. Regression models for ordinal data (with discussion). J. R. Stat. Soc. 1980, 42, 109–142. [Google Scholar]
  43. Hole, A.R. A comparison of approaches to estimating confidence intervals for willingness to pay measures. Health Econ. 2007, 16, 827–840. [Google Scholar] [CrossRef] [PubMed]
  44. Yue, C.; Hugie, K.; Watkins, E. Are consumers willing to pay more for low-input turfgrasses on residential lawns? Evidences from choice experiments. J. Agric. Appl. Econ. 2012, 44, 549–560. [Google Scholar] [CrossRef]
  45. Curtis, K.R.; Cowee, M.W. Are homeowners willing to pay for “origin-certified” plants in water-conserving residential landscaping? J. Agric. Res. Econ. 2010, 35, 118–132. [Google Scholar]
  46. Schimmenti, E.; Galati, A.; Borsellino, V.; Ievoli, C.; Lupi, C.; Tinervia, S. Behaviour of consumers of conventional and organic flowers and ornamental plants in Italy. HortScience 2013, 40, 162–171. [Google Scholar]
Figure 1. Example scenario image and areas of interest (AOI). Note: Dashed lines designate areas of interest (AOI), which correspond to the pollinator-friendly, price, production method, origin, plant sign, and plant image attributes. (A) Original Scenario Image; (B) Areas of Interest.
Figure 1. Example scenario image and areas of interest (AOI). Note: Dashed lines designate areas of interest (AOI), which correspond to the pollinator-friendly, price, production method, origin, plant sign, and plant image attributes. (A) Original Scenario Image; (B) Areas of Interest.
Sustainability 09 01743 g001
Table 1. Conjoint analysis experimental attributes and attribute levels to measure consumer preferences for eco-labels on plants.
Table 1. Conjoint analysis experimental attributes and attribute levels to measure consumer preferences for eco-labels on plants.
AttributeDefinitionAttribute levels
Plant typeType of plant shown in the scenario imagePetunia (Petunia x hybrid Juss.)
Pentas (Pentas lanceolata Forssk.)
Hibiscus (Hibiscus rosa-sinensis L.)
Price aPrice per plant$10.98
Pollinator Describes if the plant benefits pollinatorsPollinator friendly
No label—not rated
Production methodHow the plants were producedCertified organic b
Organic production c
OriginWhere the plants were producedIn-state (“Fresh from Florida”)
Domestic (“Grown in U.S”) d
Import (“Grown outside U.S.”)
a Price points were determined based on retail observations in central Florida (i.e., big box stores and independent garden centers). b Certified organic was described as “the plants are certified as organically produced”. c Organic production was described as “the plants are produced in an organic manner but are not certified organic”. d Grown in U.S. was described as “plants are produced in another U.S. state other than Florida” to differentiate the in-state and domestic attributes.
Table 2. Sociodemographic summary statistics of Floridian participants from a 2014 study.
Table 2. Sociodemographic summary statistics of Floridian participants from a 2014 study.
VariableDescription of VariablesMean (Std. Err.)Florida a Mean
AgeAverage age (in years) of participant52.782 (1.633)40.7
GenderGender of participant 0.385 (0.048)0.499
1 = male
0 = female
Income2013 gross household income of participants5.010 (0.298)$47,309
1 ≤ $20 K
2 = $21–$30 K
3 = $31–$40 K
4 = $41–$50 K
5 = $51–$60 K
6 = $61–$70 K
7 = $71–$80 K
8 = $81–$90 K
9 = $91–$100 K
10 ≥ $100 K
EducationHighest level of education completed3.880 (1.649) 26.2% Bachelor’s degree or higher
32.2% Bachelor’s degree or higher
HouseholdNumber of people in household1.870 (0.135)2.58
Pollinator purchaseLikelihood that a pollinator-friendly plant label would change the consumers’ purchasing decision.3.913 (0.893)Not available
1 = very unlikely6.6%—very unlikely or unlikely
2 = unlikely
3 = undecided 18.5%—undecided
4 = likely 75.0%—very likely or likely
5 = very likely
a Source: U.S. Census Bureau [23].
Table 3. Participants’ fixation count means and visual attendance to specific plant attributes, by attribute.
Table 3. Participants’ fixation count means and visual attendance to specific plant attributes, by attribute.
Fixation CountAttendance
AttributeMean (Std. Dev.)Mean (Std. Dev.)
Pollinator2.832 (1.691)0.467 (0.288)
Low price2.572 (1.808)0.417 (0.301)
Medium price2.517 (1.851)0.410 (0.332)
High price2.911 (1.649)0.492 (0.277)
Certified organic2.968 (1.831)0.390 (0.262)
Organic3.445 (2.245)0.530 (0.356)
Conventional3.252 (2.176)0.544 (0.330)
In-state2.688 (1.592)0.454 (0.296)
Domestic3.006 (2.016)0.483 (0.314)
Import4.490 (2.741)0.627 (0.341)
Plant7.892 (5.294)0.763 (0.274)
Plant sign1.752 (1.656)0.280 (0.290)
Table 4. Ordered logit regression coefficient estimates of Florida consumers’ purchase likelihood for ornamental landscape plants with different eco-labels.
Table 4. Ordered logit regression coefficient estimates of Florida consumers’ purchase likelihood for ornamental landscape plants with different eco-labels.
Model 1 (n = 108)Model 2 (n = 104) a
VariableCoef. (Std. Err.)Coef. (Std. Err.)
Price−0.176 (0.028) ***−0.186 (0.029) ***
Hibiscus0.690 (0.114) ***0.759 (0.116) ***
Pentas0.420 (0.113) ***0.463 (0.114) ***
Pollinator friendly0.318 (0.094) ***0.343 (0.095) ***
Certified organic0.537 (0.113) ***0.555 (0.114) ***
Organic production0.723 (0.128) ***0.751 (0.129) ***
In-state1.056 (0.118) ***1.104 (0.120) ***
Domestic0.813 (0.123) ***0.863 (0.124) ***
Sociodemographic variables
Age0.011 (0.003) ***0.009 (0.003) **
Gender0.181 (0.095)0.298 (0.106) **
Household−0.122 (0.038) ***−0.055 (0.043)
Income0.049 (0.017) **0.003 (0.018)
Education−0.175 (0.030) ***−0.158 (0.033) ***
Eye tracking variables
Pollinator attendance---0.923 (0.329) **
Low price attendance---0.487 (0.215) *
Medium price attendance---−1.418 (0.218) ***
High price attendance---−0.993 (0.300) ***
Certified organic attendance---1.054 (0.339) **
Organic attendance---−1.015 (0.203) ***
Conventional attendance---0.416 (0.236)
In-state attendance---−0.225 (0.261)
Domestic attendance---1.476 (0.312) ***
Import attendance---0.211 (0.243)
Plant attendance---0.780 (0.212) ***
Plant sign attendance---−0.023 (0.200)
Threshold parameters
1−3.757 (0.476)−3.072 (0.509)
2−2.559 (0.468)−1.826 (0.501)
3−1.808 (0.466)−1.034 (0.499)
4−1.434 (0.466)−0.638 (0.499)
5−0.412 (0.464)0.456 (0.499)
60.723 (0.464)1.666 (0.501)
Number of obs.17281664
Log likelihood−2691.334−2612.1372
Prob > χ 2 <0.001<0.001
McFadden/Pseudo R20.04760.0756
Akaike Information Criterion (AIC)5420.6675286.274
Bayesian Information Criterion (BIC)5522.0195451.639
***, **, * Indicate p-values of ≤0.001, ≤0.010, and ≤0.050 when compared to the base variables. a Four participants were unable to have their eye movements recorded reducing the sample size to 104 for Model 2. Note: Marginal effect estimates were calculated based on the ordered logit model results and are available upon request.
Table 5. Floridian consumers’ willingness-to-pay premiums (in U.S. dollars) for plants with different eco-labels.
Table 5. Floridian consumers’ willingness-to-pay premiums (in U.S. dollars) for plants with different eco-labels.
Model 1Model 2
AttributePremium (Std. Err.)Premium (Std. Err.)
Hibiscus$3.926 (0.882)$4.072 (0.855)
Pentas$2.389 (0.731)$2.486 (0.704)
Pollinator friendly$1.808 (0.632)$1.838 (0.602)
Certified organic$3.056 (0.819)$2.976 (0.773)
Organic production$4.117 (0.981)$4.028 (0.925)
In-state$6.011 (1.197)$5.922 (1.129)
Domestic$4.629 (0.979)$4.629 (0.932)
Sustainability EISSN 2071-1050 Published by MDPI AG, Basel, Switzerland RSS E-Mail Table of Contents Alert
Back to Top