• No results found

The influence of affect induction on the processing

N/A
N/A
Protected

Academic year: 2023

Share "The influence of affect induction on the processing "

Copied!
65
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

MSc Business Economics

Neuroeconomics

Master Thesis

The influence of affect induction on the processing

of sustainable information and subsequent sustainability

choice

by

Maximilian Zimmermann

Student 13307312

August 2021 15 ETCS March 2021 to

August 2021

Supervisor/Examiner: Examiner:

Jan Engelmann Jantsje Mol

(2)

Statement of Originality

This document is written by Student Maximilian Zimmermann who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

UvA Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Table of Contents

Introduction... 5

Previous Literature... 7

Sustainability in Consumer Decision Making ... 7

Emotions and Ethical Decision Making ... 10

Taking Attention Into Account ... 13

Materials and Methods ... 16

Subjects ... 16

Mouselab Web Application ... 16

Experimental Design ... 17

Emotion Induction ... 17

Decision Making Task ... 18

Procedure of Decision Making Task ... 19

Sustainability Consciousness Scale (SCS) ... 20

Payment Structure ... 20

Data Analysis ... 21

Results ... 26

Descriptive Statistics ... 26

Descriptive Attentional Analysis ... 27

Main Analysis ... 28

Effects of Mood Induction on Attention ... 28

Effects of Mood Induction on Decision Making ... 29

Effects of Attention on Decision Making ... 30

Joint Effects of Mood Induction and Attention on Decision Making ... 31

Specific Emotions and Sustainable Information Processing and Decision Making ... 36

Discussion ... 37

Limitations ... 41

Conclusion ... 43

Bibliography ... 45

Appendix ... 54

(4)

Abstract

Climate change and its implications for human kind have been publicly debated topics in the last decades. Recently, this development has intensified with increasing number of public protests demanding a concerted, coordinated effort against climate change. One potential strategy to combat climate change is a more sustainable lifestyle of individuals. Hence, it is important to investigate which factors influence our sustainable decision making. However, research concerning sustainable decision making and the factors that influence it in consumer settings is lacking, even though it has increased in recent years. Therefore, this experimental study examines how incidental affective states influence sustainable information processing and subsequent product choice. Participants were asked to perform a decision making task after they were induced with affective states, in which they had to inspect different attributes of a sustainable and a non-sustainable product. Following this they completed a trial by choosing a product. Negative incidental affective states, in comparison to positive incidental affective states, lead both to an intensified attentional focus of participants to sustainability attributes of sustainable food products and an increase of sustainable product choice.

Additionally, it was shown that attentional focus on either the conventional or sustainable product predicts product choice. Lastly, incidental affective states were only evidenced to partially moderate the relationship between attentional focus and product choice. Still, there is a need for further research and further evaluation of the topic as limitation apply to this research.

(5)

Introduction

Climate change is on the rise and a climate crisis looms. According to the United Nations, global emissions of carbon dioxide (CO2) have increased by almost 50% since 1990.

Additionally, 20 of the last 22 years were declared as the warmest years since the beginning of temperature recordings, of which 2019 recorded an all-time high temperature (World Meteorological Organization, 2021). There are several causes for this development. One of them is an increased use of fossil fuels, such as coal, which is needed, amongst others, in order to produce electricity. Another reason is deforestation. A specifically prominent case is the Amazonas rainforest in South America, where deforestation has hit an all-time high in 2020, during which 11.088 square kilometers of rainforest were destroyed (BBC, 2020; WWF, 2021).

As rainforests store carbon dioxide, decreasing deforestation aids in slowing down climate change. However, in countries such as Brazil deforestation often occurs in order to create living space for livestock. This connects to the third main reason for climate change: an increasingly intensive agriculture (WWF, 2021). The agriculture sector was responsible for 10% of the total greenhouse gas emissions in the European Union (EU). While the EU has managed to decrease emissions from this sector by nearly 25% in the last two decades, this trend is going in an opposite direction when inspected from a global perspective: crop and livestock production have risen by 10% in the last ten years (European Environment Agency, 2021).

Climate change has a severe impact on human and animal life on the planet. Direct effects of climate change include not only the immediate contribution to rising sea levels, which endangers human cities and communities in coastal areas, but also higher maximum and minimum temperatures on land as well as in the water, with this endangering thousands of species (MyClimate, 2021; WWF, 2021). Indirect effects of climate change include an increase in hunger and water crises, health risks for humans, economic damage through dealing with climate change, increasing spreads of pests and pandemics, a loss of biodiversity, ocean acidification and the need for adaptation in all kinds of areas, such as agriculture, tourism and infrastructure (MyClimate, 2021).

Urgent action is needed to combat this crisis. Recently, protesters around the world have demanded an intensified fight from governments against climate change. The so-called

“Fridays for Future” protesters demand strong approaches from governmental institutions in order to fight the current climate crisis. Some governments have already implemented measures: For example, the Dutch Government plans to reduce the Greenhouse gas emissions - largely responsible for a shift in climates temperature - by 95% up until 2050 (Government of

(6)

the Netherlands, 2021). Furthermore, several countries are introducing (or are debating introducing) climate friendly policies, such as making flights less attractive to customers through increasing prices (Schulte, 2021). However, protesters claim that this action is not sufficient in order to stop climate change to a degree which is necessary for a livable future (Fridays for Future, 2021).

One approach to decrease greenhouse gas emissions, as it is demanded by the protesters, is that governments implement strict climate policies to regulate individual and corporate behaviors. While an effective strategy, it should not be neglected that individuals can contribute to decreasing carbon dioxides through a more sustainable lifestyle. In the 2015 Paris Agreement, the worldwide community has reached an agreement to implement measures so that temperatures do not increase more than 1.5 degrees Celsius to pre-industrial ages (United Nations, 2021). In order to reach this goal, a shift in sustainable awareness and subsequently sustainable behavior is needed, at individual as well as societal level.

So far, there has been scarce research concerning sustainable decision making.

However, as evidenced above, it is relevant to understand which factors influence sustainable decision making in order to gain a larger subject specific understanding and to eventually implement measures which might shift behaviors and decision making towards sustainability.

Therefore, this thesis investigates how positive and negative affective states, as well as specific emotions, influence the attention to and the processing of sustainable information on food products and how it is influencing subsequent choice behavior. In doing so, this research contributes insights into sustainable decision making and investigates if affect is, as in other areas of decision making (e.g. Engelmann & Hare, 2018), an important factor influencing the judgement of participants.

Results of the conducted experiment investigating this topic include that participants who were induced with negative affect made substantially more sustainable decisions in comparison to the group which was induced with positive affect. Additionally, negatively induced participants also attended more to the sustainability information of the sustainable product in comparison to the positively induced group. A third finding indicates, that participants who pay more attention to sustainability attributes of sustainable food options have a higher likelihood of choosing the sustainable food. Lastly, it was investigated if positive and negative affective states moderate the relationship between attention times to sustainable information and choice behavior. However, the present data did not only show partial evidence of this impact of affect.

(7)

This paper will continue by introducing previous literature concerning sustainability issues in consumer decision making, as well as literature concerning emotions in ethical and moral decision making in the ‘Previous Literature’ section. Additionally, the hypotheses forming the base of the experiment are presented in this section. Afterwards, information concerning participants, payment scheme, experimental design as well as the structure of the data analysis is going to be presented in the ‘Methods and Materials’ section. After explaining the experimental results in the ‘Results’ section, a discussion of these results follows in the

‘Discussion’ section, including a presentation of limiting factors.

Previous Literature Sustainability in Consumer Decision Making

Consumer decision making has been of interest in several disciplines for decades. Researchers from the fields of psychology, marketing and behavioral economics have investigated the process of how individuals form their decisions in a consumer setting. The theoretical basis for these investigations was set by research on individual decision making (e.g. Amadae, 2007;

Bernoulli, 1738; Tversky & Kahnemann, 1992).

One of the latest models - stemming from the discipline of neuroeconomics - has further contributed to clarify consumer decision making theory. Fehr & Rangel (2011) present a computational model which describes five decision making steps for simple choices from a neuroeconomic perspective. When individuals enter situations in which they have to decide between two items, e.g. during grocery shopping, the model claims that the brain computes a decision value for each of the options at the time of choice as a first step. A decision value is a signal that is computed at the time of choice, predicting the hedonic value of each option.

Importantly, these decision values are computed by integrating information about attributes of the products. For example, price is an attribute in a shopping situation. The second step of the decision making process is the brain computing an experienced utility signal at the time of consumption, which is distinct from the decision value as it is an outcome-related signal and is used by the brain to update future decision values. Next, the brain evaluates the distinct decision values with so-called drift diffusion models (DDM), by comparing the decision values and coming to a decision. Lastly, the model denotes that the computation and comparison of the decision values are attention-dependent. The brain can vary the deployment of computational resources, which is important due to their scarcity. Attention can impact the decision values in

(8)

two ways. On the one hand, it can influence how attributes are weighted and computed, while on the other hand it can play a role in how decision values are compared at the time of choice.

In recent years, sustainability information attributes have gained importance within research on the consumer decision making process. Products are defined as sustainable when they have a “positive environmental and/or social impact because they are produced with concern for human and natural resources” (Brach, Walsh & Shaw, 2018). In light of the aforementioned climate crisis, it is important to maintain conditions in which humans and nature co-exist (Bangsa & Schlegelmilch, 2020). This realization is spreading in broader society as well, as customers increasingly buy items which are labelled sustainable. Between 2018 and 2019, the sale of sustainable food products increased in nearly each food category and even doubled for dairy products (Van Gelder, 2020). However, the market share of sustainable products in general remains, even though it is increasing, low. For example, sustainable certified coffee holds only 12% of the global general market share in 2012. Still, in Western countries the certified coffee share was with up to 40% higher. Producers have realized the demand for sustainable products and increased food item certifications as well. The European Commission reported 129 food information schemes related to sustainability in the European Union (EU) (Van Loo et al., 2015). Sustainable information are often communicated to consumers via these sustainability labels, which often include different dimensions of sustainability, ranging from animal welfare to carbon footprint. The goal of these labels is to enable customers to make sustainable food choices. Interestingly, the number of ecolabels has increased substantially during the last years (EU Ecolabel, 2021).

As interest has risen from the consumer as well as on the producer perspective, it is not surprising that the role of sustainable product characteristics in the consumer decision making processes has been investigated in various studies. Researchers find that ethical product attributes are especially important in the choice stage of a decision making process.

Additionally, they can also influence initial information search behavior, as ethical attributes involve customers more in the search for information (Zander & Hamm, 2012). Lastly, the formation of consideration sets is influenced, as consumers are forced to create trade-offs between different product types. Taking all of these together, Aprile, Caputo & Nayga (2012) claim that labels boost sustainable and prevent conventional product purchases. Nonetheless, it has also been found that the effect depends on the displayed logo type as well, showing that fair trade as well as carbon footprint labeling increase purchase intent (Bangsa & Schlegelmilch, 2020). However, there exists also evidence which claims that sustainability labels might not be as influential as suggested. Tebbe & von Blanckenburg (2018) found that consumers are willing

(9)

to pay for sustainable labeled plant based products, but not for corresponding meat products.

Furthermore, Ghvanzide, Velikova, Dodd & Oldewage-Theron (2017) claim that price and nutritional information on food products are more influential attributes as sustainability information. Silva, Bioto, Efraim & Queiroz (2017) even claim that most consumers would ignore sustainability labeling on food products and only consider price, brand and taste.

Therefore, O’Rourke & Ringer (2016) tried to establish an explanation for the previous mixed results, finding that individuals who have shown previous commitment to sustainability related questions show a significant relationship between sustainability information and their purchase intent. In summation, sustainability information can both enhance and weaken consumer preferences depending on the attributes that customers consider important. As customers make trade-offs between different attributes, they do not always live up to their moral entitlement (Bangsa & Schlegelmilch, 2020).

As mentioned in the computational model of Fehr & Rangel (2011), attention plays a key role in the consumer decision making process. Attention is defined as the “degree to which consumers focus on specific stimuli within their range of exposure” (Bialkova et al., 2014, p.

67). Van Loo et al. (2015) were the firsts to conduct an experiment in order to investigate the role of visual attention for consumer behavior and choice when being exposed to sustainable certified coffee. The authors argue that focusing on attention is important as it is a prerequisite for information processing, which takes place when the label is being noticed. Within the experiment, 6500 participants were presented with coffee products which were described as a combination of different sustainability labels and price. Amongst these labels were the Fair Trade, the Rainforest Alliance, USDA Organic and Carbon Footprint label. During the experimental task, participants were asked to make eight choices in total. Additionally, they had to rate coffee attributes, such as taste, as well as filling in a questionnaire about their concern for sustainability. The results demonstrate that the deployed visual attention relates to the subjective product importance indicated within the surveys. Furthermore, fixating on sustainability attributes shows that the participant has a higher preference for these attributes, resulting in a higher willingness to pay for the item. The authors also show that price has the highest fixation rate, while both the carbon footprint label as well as the USDA Organic label were fixated on less.

In another experimental study by Van Loo, Grebitus & Verbeke (2021), the authors examined the relationship between nutritional and sustainable information and attention and choice. In their experimental design participants were asked to choose a granola bar concept, which varied in five different aspects, including sustainability, price and nutrition claims. Here,

(10)

the authors hypothesized that higher visual attention for sustainability claims will lead to a higher likelihood of choosing the sustainable product option. Additionally, they claim that sustained attention to attributes predicts that customers value these attributes more. Indeed, the authors were able to confirm these hypotheses. Their results showed that visual attention led to a higher likelihood of choosing the respective option and that a higher food product price relates to a lower choice probability for that product. Hereby, the authors establish a relation between attention, choices and reaction times. This is in line with previous research by Bialkova et al.

(2014), which had already predicted that continued visual attention relates to higher probability of choosing related products. Here, the authors tested if attention mediates the effect of nutrition label information on consumer choice and reported that labels – which that attract more attention - had an mediating effect on consumer choice.

Emotions and Ethical Decision Making

Decision making involving sustainable attributes is a sub-form of moral and ethical decision making (EDM), whereas ethics is defined as the degree to which society defines a matter as right or wrong, thereby being a key element influencing consumer behavior (Escadas, Jalali &

Farhangmehr, 2019; Öberseder & Schlegelmilch, 2010). Chan & Harris (2011) title sustainability “one of the greatest moral challenges facing us in this century”. Therefore, it is relevant to take previous research of ethical and moral decision making into account in order to derive implications for sustainability related issues.

One of the first to provide a framework for moral decisions, Kohlberg (1976), describes moral judgements as a merely rational process, arguing that they arise as a product of conscious deliberation. Another early model of moral cognition and behavior was provided by Rest &

Narváez (1994). The authors describe a four-step process, starting by the rise to awareness of a moral issue, continuing with judging the moral issue, subsequently forming a decision intention and finally making the moral issue decision.

However, it is nowadays widely accepted that affective influences play a profound role in moral cognition as well (Haidt, 2001; Pinker, 1997). Affect is the quality associated with a feeling state. This quality is described along a goodness-badness dimension and often labelled as valence, while arousal is the second component describing affect. In general, affect provides us with information about the environment, where positive information signals indicate that surroundings are safe and that no specific action is needed, while negative affect signals the opposite (Västfjäll et al., 2016). Several forms of affect exist, one of them being mood. Mood is a stable but mild affective state and doesn’t focus at any specific item. It thereby differs from

(11)

emotions, which are another form of affect but which are rather intense and do focus on a specific matter (Gaudine & Thorne, 2001). Emotions are “multidimensional feelings” and thereby reflect reactions of humans to their surroundings, for example in shopping situations (Achar et al., 2016; Krishnakumar & Rymph, 2012). Some theorists even go so far as to prescribe emotions a dominating role in moral decision making. Previous research has shown that emotions can impact rationality in several stages of Rests (1984) model. For example, people can infer information from their emotions, a possible influence within the judgement stage (Prinz, 2016).

However, even though there is overarching agreement that emotions do play a role, there exists an ongoing controversy amongst researchers around how emotions interact with cognition during moral decision making (George & Dane, 2016; Pascual, Rodrigues &

Gallardo-Pujol, 2013). Greene et al. (2009) proposes the dual process model of moral decision making which describes both an automatic, emotional, as well as a conscious, controlled reasoning system, stating that both play essential and occasionally competitive roles. Thereby, the theory connects conscious reasoning processes to utilitarian moral judgement and emotional processing to deontological judgement, which is referred to as the “Central Tension Principle”

of the theory (Greene, 2014). Other authors argue that one system - which integrates conscious reasoning as well as emotional inputs - is the basis of moral decision making, thereby postulating a “cognitive and emotional integration theory”. (Bluhm, 2014; Moll, De Oliveira- Souza & Zahn, 2008; Valdesolo & Desteno, 2006).

Even though the question of how cognition and affect interact during EDM is not yet resolved, there has been plenty of research investigating how affect influences ethical decision making. The last decade has shown how this relationship has been a topic of steadily growing interest: publications within this area of research have doubled from 2004 to 2007 as well as from 2007 to 2011 (Lerner, Li, Valdesolo & Kassam, 2015). Especially the relationship between emotions and consumer decision making involving ethics has been investigated, showing that ethical attributes are essential to consumer behavior (Gaudine & Thorne, 2001).

In general, there exist two affect types which are known to influence decision making:

The first, integral affect (also known as anticipatory affect) arises from making decisions or judgements and is therefore, as a direct part of the choice process, decision-dependent. For example, a customer could choose to buy chocolate and feel happiness anticipating its taste (Achar, So, Agrawal & Duhachek, 2016; Engelmann & Hare, 2018; Lerner et al., 2015).

Integral affect has a two-fold potential to influence decision making: first, it can guide decision making by providing reactions to stimuli, thereby indicating if stimuli are rather beneficial or

(12)

not. Damasio (1994) hypothesized that integral responses are formed in response to so-called

“somatic markers”. With this, Damasio (1994) described that either negative or positive markers could be linked to an image of a future outcome, thereby respectively either describing an alarming or an incentivizing situation. Secondly, integral responses can function as biases, thereby disrupting decision making. For example, people might feel afraid to fly and therefore travel by car, even though the base rate for deathly accidents is higher by this form of transportation (Lerner et al., 2015).

The second form of affect, incidental affect, is decision-independent, thereby arising from factors which are external to the judgement. Incidental affect is relevant as it can carry over to scenarios in which it influences decision making. For example, a customer could feel angry during a shopping scenario due to an unrelated, previous fight with their partner (Achar et al., 2016; Garg, Inman & Mittal, 2005). Therefore, incidental affect can function as a bias as well and consequently influence consumer responses (Achar et al., 2016).

Previous research in EDM has investigated the influence of these types of affect and has shown mixed results. Specifically early EDM studies have focused on valence-based affective influences on decision making (Singh, Garg, Govind & Vitell, 2018). Gaudine and Thorne (2001) hypothesize that the induction of positive affect will make individuals select more ethical choices. Moreover, Zolotoy, O’Sullivan, & Seo (2021) suggest in their article that positive affect amongst employees at firms headquarters lead to an increased corporate giving and influences corporate philanthropy. Lastly, research conducted by Escadas et al. (2019) hypothesized that positively (derived from ethical behaviour) as well as negatively anticipated emotions (derived from unethical behaviour) will positively influence ethical consumer awareness and ethical behaviour. This is due to the fact that humans try to seek out positive emotions while simultaneously avoiding negative emotions. The results indicate that anticipated negative emotions reflect the largest response by inducing higher levels of ethical awareness and behaviour. Guzak (2015) conducted an experiment in which he investigated incidental affective states influence on EDM as well. He exposed university students to movie clips, thereby inducing positive, neutral and negative mood. Afterwards, participants were exposed to a hypothetical ethical dilemma, to which they could respond to in a utilitarian and consequentialist or in a rights and non-consequentialist manner. Here, the author hypothesizes that decision makers in a negative emotional state differ in comparison to decision makers in a positive affective state. Specifically, a positive or neutral affective state leads decision makers to act along a consequentialist approach to an ethical problem, whereby the consequentialist theory of value “judges the rightness or wrongness of an action based on the consequences that

(13)

the action has” (Guzak, 2015; Stanford Encyclopaedia of Philosophy, 2003). Opposing this, a negative affective state leads to more non-consequentialist decision making approach to ethical problems, rather judging “the rightness or wrongness of an action based on properties intrinsic to the action, not on its consequence” (Guzak, 2015). Typically, a utilitarian response to an ethical problem is considered a less ethical choice, as it has been shown that it correlates with egocentric attitudes, not showing concern for the greater good and being less lenient for moral transgressions (Kahane et al., 2015). Guzak (2015) finds a significant effect between the decision means of negative and positive affective state. Additionally, the author finds that participants in a negative state, compared to the positive group, use a non-consequentialist approach in order to answer ethical problems, indicating that different decision making processes are employed depending on the subjects mood. As Guzak (2015) induced affect incidentally in his study, the following hypothesis is derived from the experiment:

H1 : Participants in an incidental positive affective state will make more sustainable product decisions in comparison to participants in a negative affective state.

Taking Attention Into Account

A study investigating the impact of emotion on attention to foods was conducted by Motoki, Saito, Nouchi, Kawashima & Suguira (2019). The researchers claim that visual attention is influenced by the interplay of integral as well as incidental emotional states and were the first to investigate the interaction between both affective states and attention times simultaneously. As part of their experimental design, participants had to attend to food comparisons which differed in food category (healthy or hedonic food) or in packaging attributes, while their reaction to these attributes was measured with an eye-tracking device.

Here, the researchers argue that integral affect arises from hedonic food. Before being exposed to the food comparisons, the participants are initially induced with an incidental affective state.

The researchers hypothesize that incidental anxiety (in comparison to anger and neutrality) will increase visual attention to hedonic foods and that anxiety-induced participants attend more to the pictures than to the nutritional information. The first, but not the second hypothesis could be confirmed.

However, next to this study only a limited number of studies have investigated how visual attention is influenced by incidental affective states, specifically for attention duration and acquisition frequency. Most of the literature has focused on the role of emotion on attentional scope. Specifically for the domain of sustainability, barely any research on the

(14)

relationship between emotion and visual attention has been conducted. Also in the field of EDM, which was previously used to argue for predictions concerning sustainability, this research scope is rare. However, it seems intuitive that a negative incidental state would rather disengage subjects from caring about sustainable issues, not only in experimental tasks but also in real life, thereby decreasing attention to sustainable information. Yet, it might also be the case that subjects who already feel bad, do not want to further increase this feeling this by choosing a non-sustainable product. Even though the direction of the effect is therefore unsure, it seems reasonable that a difference in attention span to sustainability related information, depending on if participants are in a positive or negative incidental affective state, exists.

H2: Participants in a negative incidental affective state will differ in deployed attention time to sustainability information of the products to participants in a positive incidental affective

state.

As mentioned earlier, previous literature has shown that sustained attention to sustainable information can predict the choice for sustainable products (Krajbich, Armel &

Rangel, 2010; Van Loo et al., 2015, 2021). However, other research claims that sustainability information may not have the same importance that other research grants it. Therefore, it is hypothesized that attention to sustainable attributes of each product will predict which choices will be made.

H3: The likelihood to choose the sustainable product increases with sustained attention to sustainability attributes of the sustainable food product in comparison to the non-sustainable

food product.

Lastly, it might also be the case that affect induction does not directly influence either attention or choice, but indirectly influences the relationship between the both, thereby taking on a moderating role.

H4: Incidental affective states moderate the effect of attention towards sustainable attributes on consumers choice

While the valence-based approach has dominated research of emotional influences and decision making for decades, in recent years the focus has shifted on investigating the influence

(15)

of specific discrete emotional states (Singh et al., 2018). The reason behind this development is that valence based models cannot explain all affective influences on judgement. Distinct emotions can differ crucially in different ways such as appraisals, processing depth, different kind of brain activation and central nervous system activity (Lerner et al., 2015). Therefore, it can be the case that different emotions with the same valence have an opposing influence on decision making (Singh et al., 2018). In order to classify the possible influence of each specific emotion, Lerner et al. (2015) proposed the appraisal tendency framework. It suggests that each emotion is defined by two to three key appraisal dimensions, which make it possible to predict the influence of an emotion on judgement: the main appraisals include categories such as attentional activity, certainty, control, pleasantness, anticipated effort and responsibility. Here, an assumption of the framework is that emotions of the same valence can have distinct influences on judgements, while emotions of opposite valences can have the same.

Additionally, Engelmann & Hare (2018) suggest that specific emotions influence (risky) decisions through their motivational-specific nature, regardless of their valence. The authors base their claim on neuroscientific evidence, which indicates that approach-oriented emotions show an increasing activation within the brains’ valuation system, while avoidance and withdrawal-related emotions support activation in the insula and at the same time decrease activation in the valuation system.

The research shift from valence to specific emotions can also be observed in research concerning ethical decisions as more and more researchers have investigated the influence of specific emotions on ethical decisions. For example, a study by Singh et al. (2018) investigated the effects of incidental anger and fear on ethical decisions. The authors hypothesized that fear would lead to higher levels of ethical judgements, while anger will lead to lower levels of ethical judgement. The results of the study confirmed the predictions, showing that distinct emotions with the same valence can have different impacts on ethical decision making. Another study by Yip & Schweitzer (2016) investigated the effect of incidental anger and sadness on ethical behavior. The authors came to similar conclusions and postulated that an incidental state of anger leads to less ethical behavior by increasing deception. These results stand in comparison to sadness, which did not show these effect. Furthermore, a study performed by Zhang et al.

(2020) examined the effect of anxiety on ethical decision making. Here, the results showed that the experimental group induced with an anxious incidental affective state acted more unethically than a neutrally induced contrast group. These results are supported by a study of Kouchaki & Desai (2015) which similarly found that incidental anxiety leads to more unethical acts. Lastly, previous literature has produced mixed results for the specific emotion of disgust.

(16)

While Winterich, Mittal & Morales (2014) states that disgust does increase unethical behavior, a replication study by Kugler, Noussair & Hatch (2020) cannot find a relationship between disgust and unethical behavior, even though statistical power was higher due to a large sample size.

In order to incorporate the effects of specific emotions in this study, the previously stated hypothesis will be tested by replacing the affect condition with the specific emotions that seem to have an influence on EDM in previous literature.

Therefore, the following hypotheses emerge when taking this previous research into account:

H1.1 : Participants in angry or anxious affective states will make less sustainable product decisions in comparison to participants in a negative affective state, while sadness will lead

to more sustainable product decisions.

H4.1: Specific emotional states moderate the effect of attention towards sustainable attributes on consumers choice.

Materials and Methods Subjects

In total, 107 participants took part in the experiment. The recruitment process included requesting fellow students, friends and family to participate as well as addressing contacts via the social media platform LinkedIn. After inspecting the data, four participants were excluded from the analysis as they didn’t finish the experiment. Therefore, 103 responses were recorded.

Mouselab Web Application

The MouselabWeb application was used during this experiment as a process tracing tool in order to acquire information about attention times to certain attributes as well as choices of participants between two options. The tool has been developed in order to give insights into information acquisition processes and to understand decision making of participants. Therefore, MouselabWEB is a proxy for eye-tracking measurements. This relation is given, as participants acquire information with their mouse similarly as with their eyes, namely by moving to a location, pausing and processing the obtained information. Advantages of this tool are its’ open source technology, meaning that it can be run on existing webpages without experimenter

(17)

interventions. Additionally, the recordings of attention times and decisions are automatic and the possibility of online research allows for large samples of heterogenous participants (Willemsen & Johnson, 2009, 2010).

In general, MouselabWEB displays boxes under which information are hidden. In order to be able to gain access to the information hidden under the boxes, participants have to hover a computer mouse over the box. As long as the mouse remains on the box, the information are visible, but when the mouse moves away from the box, the information will be hidden or blurred again. The boxes typically represent distinct attributes, of which the participant is supposed to gain knowledge. By observing how participants obtain the required information, it is possible to deduct inferences about cognitive processes. Additionally, it might be possible to predict choices from the processed attentional data. Choices are indicated after participants have inspected all the boxes. The time participants spent on the boxes, as well as the specific decision, is automatically recorded by the application (Willemsen & Johnson, 2010).

Experimental Design

Participants were first given time to read the experimental instructions as well as to perform an instruction trial with the MouselabWEB application, in order to assure task familiarization.

Additionally, participants were asked to indicate demographic variables such as age, occupation and socio-economic factors such as education and wage, diet preferences, gender and nationality. These information can be found in Appendix B and C.

Emotion Induction

Displaying affective movie clips is an effective emotion induction method. Other methods to induce emotions include affective memory recall or reading brief, affective texts (Engelmann

& Hare, 2018; Mills & D'Mello, 2014). Affect induction through video presentation was selected in this experiment as previous research has shown it to be the most successful approach when being compared to other methods, such as sounds, pictures, social interactions or facial expressions (Westermann, Spies, Stahl & Hesse, 1996). Subjects were randomly divided into two groups, while each group was exposed to a different movie clip. The first groups’ movie clip is displaying distinct and beautiful landscapes of the earth. The goal of showing this video was to induce neutral to slightly positive, incidental affective states. Contrarily, the second group was exposed to a video which displayed possible consequences of climate change in case humanity does not manage to counteract it. This video was rather dramatic and had the goal to induce negative, incidental affective states. The participants were not able to skip this essential

(18)

experimental step, as the button to continue with the rest of the experiment only displayed at the respective end of each video. Screenshots of the movie clips can be found in Appendix D.

After watching the movie clips, participants continued the experiment by filling in the Positive and Negative Affective Schedule (PANAS) in order to control the emotion manipulation (Appendix E). The PANAS contains twenty emotional states of which the perceived extent has to be rated from 1 (very slightly) to 5 (extremely). Ten of these emotional states are used to indicate positive affect (e.g “enthusiastic”, “proud”), while the other ten emotional states indicate negative affect (e.g. “irritated”, “nervous”). These states are summed and therefore form the individual positive and negative affect score.

Additionally, four specific emotional states were included within the questionnaire to control for effects which can stem from these specific emotions, thereby accounting for more recent scientific developments and findings established in previous literature (e.g. Engelmann

& Hare, 2018; Singh et al., 2018). These additional specific emotions included anger, sadness, disgust and anxiety.

Decision Making Task

Before the affect induction took place, participants already had the possibility to become accustomed to the setup of the experimental task. Therefore, subjects continued straightaway with the decision making task after being exposed to the affect induction. The task was administered via the application of MouselabWeb on a website which was programmed for the experiment. In total, each participant had to perform 52 trials, while each trial was a food comparison of a more sustainable and a less sustainable food option. The 52 trials were displayed randomly to each participant, in order to avoid order effects. 13 food comparisons were shown four times each, during which the prices varied. Specifically, one of these four food comparison displayed real-life prices which were retrieved from a known Dutch supermarket.

However, in the other three comparisons the more sustainable option was respectively 5%, 15%

and 25% more expensive than in the original food comparison. Additionally to the price attribute, sustainability attributes of each food item were shown, including the CO2 emissions, the water scarcity footprint as well as the level of animal welfare for each respective product.

The displayed information were provided by the database of the Eaternity application, which takes into account how foods are produced, transported and packaged, as well as from a Dutch governmental report concerning the environmental impact of foods. (Eaternity, 2021; RIVM, 2021). The sustainability information were displayed in the form of a sustainability logo, indicating the written information of CO2 emission values, the water scarcity footprint and

(19)

animal welfare indications. Additionally, a star system, ranging from one to five stars, indicated the quality of the food product in each of the sustainability categories. Here, five stars indicated the best value while one star showed the worst rating. The design of the sustainability logo was inspired by the logos created by Eaternity. The hidden price information were presented in written form, while the images of the food products were pictures taken from the online webpage of a known Dutch supermarket. The set-up of the experimental task is illustrated in Figure 1.

Procedure of Decision Making Task

During each trial two food products were shown to participants. Figure 1 shows that the MouselabWeb application made it possible to hide the corresponding information under boxes over which the participant had to hover with their mouse in order to extract the information. On the left side of the screen, it was indicated which information were hidden behind which boxes.

Specifically, five boxes were shown which showed the initially hidden information of the three sustainability items, the product price and a picture displaying the food item, respectively for each of the two food products (Fig. 1). This amounted to ten boxes which were shown in total.

The goal of this experimental set-up was for the participant to inspect the boxes in order to retrieve information while the attention is being tracked through the MouselabWeb application at the same time. It was only possible to inspect one of the hidden boxes at a time. After participants had studied the attributes of each food product in the comparison, they had to decide for either the more sustainable or the less sustainable option. After choosing an option, the page would refresh with another product. After 52 trials, the participants were re-directed to the Sustainability Consciousness Scale (SCS). An overview of the values of all trials can be found in Appendix A, while the SCS can be found in Appendix F.

(20)

Fig 1. Trial of the Experimental Task

Sustainability Consciousness Scale (SCS)

After participants finished the decision making task, the last part of the experiment included a survey assessing their sustainability consciousness. However, due to time constraints the short form of the SCS was administered. This included 20 questions assessing different dimensions of sustainability consciousness, which had to be rated on a five Point-Likert scale. By assessing the subjects’ sustainability consciousness, it is possible to investigate if emotion induction might only have effects in subgroups, e.g. for participants with a low sustainability awareness.

Payment Structure

In the explanatory instruction sheet in the beginning of the experiment, each participant was informed about the payment structure. In total, three participants were compensated for their efforts with 10 EUR each. These winners were randomly raffled after the data collection was finished. As each of these randomly drawn participants had made 52 choices, participants were told in the beginning of the experiment, that one of their choices was going to be drawn as well.

This choice was made between two products, therefore the winners received the product they chose, yet had to pay for it. Therefore, the final payment was the product item and the 10 EUR, from which the product price was subtracted. With this, an payment scheme was established

(21)

with which participants had the impression that every single choice could be a hypothetical purchase as every participant and every trial had the potential to be drafted as the winner.

Data Analysis

The data analysis of this experiment involved several steps. Initially, a data table for the experimental data and for the remaining questionnaires was created. Two approaches to achieve the appropriate experimental data structuring were used: First, a script in Python was written, using the data analysis tool Pandas, which structured the experimental data for further statistical processing. Second, the MouselabWEB application offers a datalyser, which is a tool that automatizes the data structuring. As the end products of both approaches were similar, they were combined for the data analysis. Thereby, the advantage of the datalyser was that it not only summarizes the total time participants have looked at a product attribute, similarly to the self-written program, but also recorded how often each product information was acquired. This is an important factor as attentional data can be described in terms of fixation duration as well as fixation frequency (Rayner, 2009). Furthermore, previous research has shown that both of these factors are highly correlated and at the same time accurate measurements and predictors of attention (Schulte-Mecklenbeck et al., 2013). Additionally, the tool immediately excludes participants who do not finish the experiment and does not take fixation times at a threshold of 200ms into account, as it is argued that participants might not consciously perceive information in such short time frames (Willemsen & Johnson, 2010). However, the datalyser wrongly excluded one participant. By adopting data from the python script, the data was saved. After finishing the pre-processing of the data and finalizing the experimental data table, statistical testing followed. All of these statistical tests were performed in R Studio.

In order to test the first hypothesis, the direct impact of emotion induction on choice was examined. Here, it was investigated if the emotion induction within the two experimental groups had a different effect on choice behavior, in the sense that a group chose the sustainable option more frequently. First, a Shapiro-Wilk test was performed in order to test the normality assumption of independent t-test. The results was significant (p < 0.01). Therefore, first a Mann Whitney U test was performed in order to test if the data stem from two distinct populations.

Following this, the analysis continued with examining the direct impact of affect induction on fixation times to sustainable food products. Here, the second hypothesis states that the attention times will differ to sustainable attributes between affect conditions. Therefore, several Mann-Whitney U tests - after having checked for normality – were performed, which

(22)

investigated if relative attention time to different attributes stems from another population in one group in comparison to the other.

Here, the relative attributes constitute the subtraction of attended time to the information of the non-sustainable product from the attended time to information of the sustainable product, divided by 1000 in order to rescale the unit to seconds. For example, the relative CO2 variable is calculated as follows:

𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 =(𝐴𝑡𝑡𝑒𝑛𝑡𝑖𝑜𝑛 𝑡𝑜 𝐶𝑂2 𝑆𝑢𝑠𝑃𝑟𝑜𝑑𝑢𝑐𝑡 − 𝐴𝑡𝑡𝑒𝑛𝑡𝑖𝑜𝑛 𝑡𝑜 𝐶𝑂2 𝑁𝑜𝑛𝑆𝑢𝑠𝑃𝑟𝑜𝑑𝑢𝑐𝑡) 1000

This calculation form is equal for all relative attributes within this and following equations, if attention times are replaced for the respective attribute. The value of this equation describes participants attention to the attributes of either the sustainable or non-sustainable food product.

Specifically, it describes the participants attention to sustainable product attributes, in this case CO2 emissions, subtracted from participants attention to non-sustainable product attributes.

This shows how much more participants attend to the sustainable product attribute.

Additionally, the result is divided by 1000 to transform the unit of reaction times from milliseconds to seconds. Next to the relative CO2 value, relative animal welfare, relative water scarcity footprint as well as relative food image are included as products attributes and attention measures. Additionally, the difference in price, the gender and age of the participants as well as the trial are included as control variables.

The next statistical steps involved the introduction of regression analyses. Here, two possibilities to analyze the collected experimental data are appropriate. First, it is possible to analyze the data by applying a logistic regression, in which the standard errors are clustered in order to account for within-cluster correlation or heteroscedasticity. This was accomplished with the “rms” R package. A second possibility to analyze the experimental data is to introduce a Generalized Linear Mixed-Effects Model (GLMM), in which random effects are integrated via maximum likelihood estimation. This, on the other hand, was accomplished using the R package “lme4”. Therefore, first the data was analyzed via logistic regression with clustered standard errors (Model 1-7). Afterwards, a GLMM analyses was conducted in order to investigate if the results differ with this distinct statistical approach (Model 1.1 – 4.1).

In order to test the third hypothesis, stating that the likelihood to choose sustainable products increases with sustained attention to sustainability attributes, a logistic regression analysis, for which the standard errors were clustered per subject, was performed (Model 1):

(23)

log(𝑐ℎ𝑜𝑖𝑐𝑒) = 𝛼 + 𝛽1⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶02 + 𝛽2⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽3⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽4⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒 + 𝛽5⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐹𝑜𝑜𝑑 + 𝑃𝑟𝑖𝑐𝑒𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑇𝑟𝑖𝑎𝑙

(1)

Following the logistic regression, also a GLMM was introduced in order to account for random effects of subjects:

log (𝐸[𝑐ℎ𝑜𝑖𝑐𝑒| ]) = 𝛼 + 𝛽1⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽2⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽3

⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 + 𝛽4⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑜𝑜𝑑 + 𝛽5 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑠𝑢𝑏𝑗𝑒𝑐𝑡

(1.1)

A mixed model is defined in the equation above. The dependent variable (choice) is distributed according to the exponential family where its expectation E is related to the linear predictor through the link function, conditioned on the random effect . The attentional measures, such as relative attention to water scarcity, animal welfare, CO2 emissions and food images as well as the control variables age and gender are the predictors and the respective beta represent the distinct fixed effects. Additionally, the variables subject symbolizes the random effects design matrix, which describes a random intercept for each participant while  represents the respective random effects. The results of the this and the following GLMM analyses can be found in the Figure 3.

As both groups have been irreversibly induced with emotional states before they performed the decision making task, it is important to take into account that not including the induced mood as an experimental factor within the analysis might lead to an omitted variable bias. However, this is still done in an attempt to reproduce findings from previous research (Krajbich et al., 2010; Van Loo et al., 2015, 2021). Therefore, a second regression analysis was performed in which the condition of the different experimental groups was added into the model (Model 2):

log(𝑐ℎ𝑜𝑖𝑐𝑒) = 𝛼 + 𝛽1⋅ 𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 + 𝛽2⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽3⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽4⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 + 𝛽5⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑜𝑜𝑑 + 𝛽6 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒

+ 𝑃𝑟𝑖𝑐𝑒𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑇𝑟𝑖𝑎𝑙

(2)

With this, it is possible to investigate the main effect of affect condition on choice. Adding the affect condition to the model, reveals further information about the role of the affect induction in forming a sustainable or non-sustainable choice, thereby contributing evidence in order to

(24)

answer the first hypothesis. Additionally, Model 2 doublechecks the main effects which were reported in Model 1, and demonstrates if the affect condition influences the previously stated attention coefficients. Again, a GLMM was introduced, in which the affect condition was added as well in order to account for random effects, thereby comparing two distinct statistical models.

log (𝐸[𝑐ℎ𝑜𝑖𝑐𝑒| ]) = 𝛼 + 𝛽1⋅ 𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 + 𝛽2⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽3

⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽4⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 + 𝛽5⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑜𝑜𝑑 + 𝛽6 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑠𝑢𝑏𝑗𝑒𝑐𝑡

(2.1)

As a next step, the sustainability consciousness score (SCS) of participants was integrated in an extended model (Model 3):

log(𝑐ℎ𝑜𝑖𝑐𝑒) = 𝛼 + 𝛽1⋅ 𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 + 𝛽2⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽3⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽4⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 + 𝛽5⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑜𝑜𝑑 + 𝛽6 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒 + 𝑆𝐶𝑆 + 𝑃𝑟𝑖𝑐𝑒𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑇𝑟𝑖𝑎𝑙

(3)

Sustainability consciousness is an important control variable to add as it might be that choice behavior mainly depends on the awareness of participants about sustainability issues. The sustainability consciousness score is derived from the short version of the Sustainability Consciousness Scale (SCS – S). This regression was additionally analyzed as a GLMM.

log (𝐸[𝑐ℎ𝑜𝑖𝑐𝑒| ]) = 𝛼 + 𝛽1⋅ 𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 + 𝛽2⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽3

⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽4⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 + 𝛽5⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑜𝑜𝑑 + 𝛽6 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒 + 𝑆𝐶𝑆 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑠𝑢𝑏𝑗𝑒𝑐𝑡

(3.1)

Lastly, the fourth hypothesis is tested which states that affect induction moderates the relationship between product choice and attention times to product attributes. The foregoing investigations looked at direct effects of emotion induction, while this analysis focuses on an indirect effect. Logistic moderation regression analysis is applied in order to examine the possible indirect effect of emotion induction (Model 4):

log(𝑐ℎ𝑜𝑖𝑐𝑒) = 𝛼 + 𝛽1⋅ 𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 + 𝛽2⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽3⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽4⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 + 𝛽5⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑜𝑜𝑑 + 𝛽6 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒 + 𝛽7

⋅ (𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟) + 𝛽8⋅ (𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑊) + 𝛽9

⋅ (𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐶𝑂2) + 𝑆𝐶𝑆 + 𝑃𝑟𝑖𝑐𝑒𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑇𝑟𝑖𝑎𝑙

(4)

Here, interactions terms between the affect induction and the respective attention capturing variables were created in order to investigate if attention to sustainability attributes differs

(25)

within the affect conditions and if that difference has an effect on the final product choice. As done before, the interactions were also analyzed as a GLMM in order to account for random effects between participants, not only for non-independence as in the logistic regression (Model 4.1)

log (𝐸[𝑐ℎ𝑜𝑖𝑐𝑒| ]) = 𝛼 + 𝛽1⋅ 𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 + 𝛽2⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽3

⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽4⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 + 𝛽5⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑜𝑜𝑑 + 𝛽6 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒 + 𝛽7⋅ (𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟) + 𝛽8

⋅ (𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑊) + 𝛽9⋅ (𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐶𝑂2) + 𝑆𝐶𝑆 + 𝑃𝑟𝑖𝑐𝑒𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑇𝑟𝑖𝑎𝑙 + 𝑆𝐶𝑆 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑠𝑢𝑏𝑗𝑒𝑐𝑡

(4.1)

As it was mentioned in previous literature, specific emotions are used more often in recent research (Singh et al., 2018). Therefore, specific emotions were measured during the experiment as they might have a more distinct influence on choice behavior. In the analysis, all regression analyses in which the affect condition was featured were replaced with discrete, specific emotions, therefore testing their specific effect (Model 5). As analyzing specific emotions is an additional analysis, it will only be performed by logistic regressions with clustered standard errors.

log(𝑐ℎ𝑜𝑖𝑐𝑒) = 𝛼 + 𝛽1⋅ 𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝐸𝑚𝑜𝑡𝑖𝑜𝑛 + 𝛽2⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽3

⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽4⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 + 𝛽5⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑜𝑜𝑑 + 𝛽6

⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒 + 𝑆𝐶𝑆 + 𝑃𝑟𝑖𝑐𝑒𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑇𝑟𝑖𝑎𝑙

(5)

Additionally, specific emotions were interacted with the attentional measures in order to observe if they moderated the relationship between attention and choice. Before conducting regression with interaction terms serving as moderator terms, all continuous variables were first mean-centered in order to decrease multicollinearity (Model 6):

log(𝑐ℎ𝑜𝑖𝑐𝑒) = 𝛼 + 𝛽1⋅ 𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝐸𝑚𝑜𝑡𝑖𝑜𝑛 + 𝛽2⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽3

⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽4⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 + 𝛽5⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑜𝑜𝑑 + 𝛽6

⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒 + 𝛽7⋅ (𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝐸𝑚𝑜𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟) + 𝛽7

⋅ (𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝐸𝑚𝑜𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑊) + 𝛽8

⋅ (𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝐸𝑚𝑜𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐶𝑂2) + 𝑆𝐶𝑆 + 𝑃𝑟𝑖𝑐𝑒𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑇𝑟𝑖𝑎𝑙

(6)

Lastly, the interactions of specific and attentional measures were interacted again with the condition, in order to observe if an moderation effects might occur condition specific only.

(Model 7)

(26)

log(𝑐ℎ𝑜𝑖𝑐𝑒) = 𝛼 + 𝛽1⋅ 𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 + 𝛽2⋅ 𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝐸𝑚𝑜𝑡𝑖𝑜𝑛 + 𝛽3⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 + 𝛽4

⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑛𝑖𝑚𝑎𝑙𝑊𝑒𝑙𝑓𝑎𝑟𝑒 + 𝛽5⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐶𝑂2 + 𝛽6⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑜𝑜𝑑 + 𝛽7

⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑃𝑟𝑖𝑐𝑒 + 𝛽8⋅ (𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑊𝑎𝑡𝑒𝑟 ⋅ 𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝐸𝑚𝑜𝑡𝑖𝑜𝑛) + 𝛽9⋅ (𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐴𝑊 ⋅ 𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝐸𝑚𝑜𝑡𝑖𝑜𝑛) + 𝛽10

⋅ (𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ⋅ 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐶𝑂2 ⋅ 𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝐸𝑚𝑜𝑡𝑖𝑜𝑛 ) + 𝑆𝐶𝑆 + 𝑃𝑟𝑖𝑐𝑒𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 + 𝐴𝑔𝑒 + 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝑇𝑟𝑖𝑎𝑙

(7)

Only results which differ from the valence-based analysis were reported in the results section.

As previously mentioned, fixation frequency and fixation duration are both predicting participants attention. Therefore, the aforementioned models were all tested with the fixation frequency instead of fixation duration of the distinct attention measures, thereby replacing the attentional predictors of choice. Results that varied from the ones of fixation duration were reported and are otherwise to be found in Appendix H. The models representing fixation frequency are called Model 8-12.

Results Descriptive Statistics

The 103 participating subjects were randomly assigned to two experimental groups, thereby forming the positive-induced affect group (Group 1, N = 52) and the negative-induced affect group (Group 2, N = 51). In both groups, slightly more females than males participated, forming 54% of the participants in group 1 and 59% in group 2. Additionally, the 27,8 years age average is higher in Group 2 than in Group 1, where the average is 23,8 years. The sample is biased towards higher education, as 45% of the participants had obtained a Master Degree (or equivalent) and 34.9% a Bachelors’ degree (or equivalent) at a university in the past as their highest level of education. The rest of the sample either obtained a high school degree (11.6%), finished an apprenticeship (4.8%) or received a PhD (0.9%). Mostly, participants were still students and just a minority worked in a job. Furthermore, the origin of most participants was primarily from Western Europe, mostly stemming from Germany (71,8%) or the Netherlands (14.6%). Other nationalities that took part in the survey were Austria, Azerbaijan, Belgium, Brazil, Canada, Luxemburg, Portugal, Romania, South Africa, Spain, Ireland, Sweden and Turkey.

As mentioned before, manipulation checks were performed by administering the PANAS in order to observe if the emotion induction was successful (Table 1). Therefore, we can observe the differences in positive and negative affect, as well as the difference in other

Referenties

GERELATEERDE DOCUMENTEN

According to these Recommendations member states have to identify risks, and develop policies and domestic coordination to address them; detect and pursue

Drie ouders van Spirit en vijf ouders van Altra hebben verteld over de draaglast die zij zien bij hun kind minimaal 1 maand na de afsluiting van het SoS traject.

De natuurlijke vijanden uit de oorsprongsgebieden van japanse vliegen, Japan en Noord-Amerika, kunnen niet zonder meer worden ingevoerd als klassieke biologische bestrijding.. Wel

Results showed incidental exposure to aversive cues biased selection of behaviours with learned aversive outcomes over behaviours with positive outcomes, despite a prefer- ence

The proposed nondimensional repre- sentation is a new simplicity of the nominal model that does not exist in more accurate or complex existing models (e.g., [10], [11] in the

He noted that while section 3(3)(a) of PAJA empowers the administrator to exercise discretion to give a person whose rights are materially and adversely affected by

Omdat de lijnen evenwijdig lopen zijn de richtingscoëfficiënten