messages towards meat consumption on the intention to reduce meat consumption: the mediating role of personal

messages towards meat consumption on the intention to reduce meat consumption: the mediating role of personal

norms and the moderating role of biospheric values

By

Albert Vlietstra

University of Groningen Faculty of Economics and Business

Master’s thesis

MSc Marketing Management

January 10, 2020

Supervisor: dr. J.I.M. de Groot Second supervisor: dr. J.W. Bolderdijk

Reitemakersrijge 6-34 9711 HT Groningen

+31625207999 a.vlietstra.1@student.rug.nl

S3222357

Abstract

This paper examined the effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption. Specifically, the mediating role of personal norms towards meat consumption and the moderating role of biospheric values on that effect was studied. The results of the one-way between-subjects design (N = 332), in which participants received either a static or dynamic descriptive normative message towards the reduction of meat consumption in the Netherlands, showed no significant differences between the effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption. Furthermore, personal norms towards meat

consumption mediated the relationship between descriptive normative messages towards meat consumption and the intention to reduce meat consumption. Finally, biospheric values were positively and significantly related to the intention to reduce meat consumption. However, our findings showed that the main effects of biospheric values and descriptive normative message framing (static versus dynamic) on the intention to reduce meat consumption were qualified by a moderator effect. That is, results showed that while dynamic descriptive normative messages were equally effective to encourage the intention to reduce meat consumption regardless of the extent to which a consumer valued the environment or biosphere, the static descriptive normative message was increasingly effective to encourage the intention to reduce meat consumption the stronger one’s biospheric values. The findings and implications of the present results are discussed in relation to popular normative theories in the field.

Keywords: static versus dynamic descriptive normative message towards meat consumption,

personal norms, biospheric values, pro-environmental behavior.

Table of contents

Abstract ... 1

Table of contents ... 2

1. Introduction ... 4

2. Literature review ... 8

2.1. Intention to reduce meat consumption ... 8

2.2. Static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption………...8

2.3. Personal norms as a mediator between descriptive normative messages towards meat consumption and the intention to reduce meat consumption ... 11

2.4. Biospheric values as a moderator on the effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption ... 13

2.5. Conceptual model ... 17

3. Methodology ... 18

3.1. Population and sampling method ... 18

3.2. Research design………...………..…..19 3.3. Materials……,………..….20

3.4. Measures ... 22

3.5. Procedure……….….23

3.6. Factor analysis and Reliability analysis………24

3.7. Assumptions parametric tests ... 26

4. Results ... 30

4.1. Descriptive statistics………...…30

4.2. Pearson correlation analysis ... 31

4.3. The effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption ... 32

4.4. Personal norms as a mediator between descriptive normative messages towards meat consumption and the intention to reduce meat consumption……….33

4.5. Static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption: the moderating role of biospheric values ... 36

4.6. Overview accepted / rejected hypotheses ... 40

5. Conclusion and discussion ... 41

5.1. Conclusion ... 41

5.2. Discussion ... 41

5.3. Practical implications ... 45

5.4. Limitations and future research ... 45

References ... 47

Appendix ... 61

Appendix A: Survey……….61

Appendix B: Descriptive statistics one-way ANCOVA ... 70

Appendix C: Descriptive statistics descriptive normative messages towards meat consumption ... 71

Appendix D: Descriptive statistics personal norms towards meat consumption ... 72

Appendix E: Descriptive statistics biospheric values ... 73

Appendix F: Descriptive statistics socio-demographics ... 74

Appendix G: Results Pearson correlation analysis ... 80

Appendix H: Descriptive statistics two-way ANCOVA ... 81

Appendix I: Slides defense ... 82

1. Introduction

With the expectation that the earth will count 9.6 billion persons in 2050 (United Nations, 2017) and since stabilization of the population growth is highly improbable within this century

(Gerland, Raftery, Ševčíková, Li, Gu, Spoorenberg, and Bay, 2014), feeding the globe while maintaining the integrity of the ecosystem will be challenging. Food consumption has an

essential impact on the environment. It is a significant source of the emissions of greenhouse gas, even more than all of the transportation systems around the world combined (Godfray, Aveyard, Garnett, Hall, Key, Lorimer, and Jebb, 2018). There is a difference regarding the impact on the environment between different types of nutrition. In particular food consumption related to the livestock sector is a significant contributor to climate change because it impacts the environment in several ways, such as greenhouse gas emissions, water footprints, land footprints, and several forms of pollution (Lentz, Connelly, Mirosa, and Jowett, 2018). With a predicted increase of 72% in global meat consumption in 2030 comparing to 2000 (Fiala, 2008), the related effects are expected to increase significantly. This is alarming since climate change is detrimental to all living organisms on earth and economies, too (United Nations, 2019).

Individuals can reduce these environmental problems by adopting more pro-environmental behavior, which is conduct that is minimally damaging for the environment and can be even beneficial (Steg and Vlek, 2009). Acting in a pro-environmental way includes setting aside immediate short-term self-interests for the sake of society’s or long-term environmental interests (De Groot and Steg, 2009). Hence, this behavior seems at least partially founded in moral

reasoning and norms; that is, behaving pro-environmentally usually involves social and moral considerations (Thøgersen, 1996). Eating less meat can be seen as an act of pro-environmental behavior because it minimizes climate change and loss of biodiversity (Stoll-Kleemann and Schmidt, 2017). Therefore, this study tries to find how people can be influenced by eating less meat as an act of pro-environmental behavior.

One way to influence the eating behavior of people is to present social norms that guide eating

behavior (Cialdini, Kallgren, and Reno, 1991). Social norms are the usual behavioral rules that

direct human interactions with other human species, and show us what behavior is prevalent or

common in particular circumstances (De Groot, Abrahamse, and Jones 2013). Aronson, Wilson, and Akert (2005) defined social norms as implicit used rules within social groups that are deemed to be acceptable values, beliefs, and behaviors of group members within that social group. To induce conformity, social norms must be salient (Cialdini et al., 1991). Saliency stands for how visible or noticeable something is (Gallagher and Updegraff, 2011).

Making social norms salient can be done by presenting normative messages and has proven to be a successful tool in influencing attitudes, intentions, and subsequent behavior to reducing meat consumption (Peattie, 2010). Normative messages can steer people into acting in line with social norms because they make a norm salient in a specific situation (Peattie, 2010). This study makes use of normative messages since it seems to be an effective way to make a norm salient.

Normative messages often discriminate between descriptive and injunctive normative messages.

While descriptive normative messages make salient one’s perceptions of usual and familiar behavior, injunctive normative messages make salient one’s perception of which behavior is disapproved or approved by other individuals (Cialdini et al., 1991). It is important to distinguish between both types in normative message framing since both of them represent different sources of interest (Stok, 2014). Descriptive social normative messages, such as ‘90% of the people in the Netherlands eat meat daily’ activate individuals by providing them social information.

Injunctive normative messages such as ‘90% of the people in the Netherlands consider eating meat as normal’ steer people into particular conduct via evaluation (Stok, 2014).

Although both descriptive and injunctive normative messages can be successful in promoting pro-environmental behavior such as reducing meat consumption (Sparkman and Walton, 2017), this research will focus on descriptive social normative messages rather than injunctive

normative messages. Descriptive normative messages refer to situations where behavior from other people can happen and where that behavior is visible, whereas injunctive normative messages motivate people’s behavior more in a general context in various situations (Cialdini et al., 1991). Therefore, focusing on descriptive normative messages is more suitable since meat reducing meat consumption is typically behavior that can happen and what can be visible.

Moreover, people find it less difficult to comply with descriptive normative messages. By merely

copying other’s behavior, less cognitive exertion is asked than when complying with injunctive normative messages (Morris, Hong, Chiu, and Liu, 2015).

More recently, another distinction has been made of descriptive normative messages, namely static and dynamic descriptive normative messages. A static descriptive normative message refers to the current state of the norm, whereas a dynamic norm explains how certain behaviors of groups have been changed throughout the years (Sparkman and Walton, 2017). Behavioral changes (dynamic norms in groups) can inspire people by adapting their behavior to the new practice (Mortensen, Neel, Cialdini, Jaeger, Jacobson, and Ringel, 2019; Sparkman and Walton, 2017). Despite promising research regarding the effectiveness of dynamic normative messages towards reducing meat consumption, empirical research remains scarce. Hence, this study will replicate and extend the work of Sparkman and Walton (2017) to gain more insight into the robustness and potential of using static versus dynamic normative messages towards reducing meat consumption.

First, even though researchers have looked at how static versus dynamic descriptive normative messages towards meat consumption could effectively influence reducing meat consumption, they have not yet examined the underlying motivational process yet. Earlier research suggested that an essential underlying mechanism regarding pro-environmental behavior such as reducing meat consumption is the role of personal norms, caused by guilty feelings and pride (Bamberg, Hunecke, and Blobaum, 2007; Onwezen, Antonides, and Bartels, 2013). Personal norms are different from social norms since they explain inner standards relating to a specific response, rather than to rules enforced externally (Kallgren, Reno, and Cialdini, 2000). This study will add to the existing literature by examining whether the effectiveness of descriptive normative

messages towards meat consumption is successful because of the role of one’s personal norms towards meat consumption.

Second, researchers have not yet looked at under what conditions emphasizing static versus

dynamic descriptive normative messages towards meat consumption on the intention to reduce

meat consumption might be even more effective. Various researches suggest that values are

relevant in environmental contexts because they adequate prognosticate pro-environmental

behavior (Nigbur, Lyons, and Uzzell, 2010; Stern and Dietz, 1994). Values are desirable trans- situational objectives and serve as navigated principles in a person’s or other societal entities

‘lives (Schwartz, 1992). Within values, biospheric values are considered the best forecast of pro- environmental behavior (De Groot and Thøgersen, 2019; Schultz, Gouveia, Cameron, Tankha, Schmuck, and Franěk, 2005) because they refer to concerns towards the quality of the

environment and nature for its interest (De Groot and Thøgersen, 2019). This paper will add to the existing literature by examining the impact of biospheric values on the effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption. It will also be studied whether the effect of the static descriptive normative message on the intention to reduce meat consumption will be low regardless of one’s biospheric values. Finally, we study whether the dynamic messages towards meat consumption will be more impactful the weaker one's biospheric values. Altogether, the following research questions have been composed:

(1) What is the effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption, (2) What is the mediating role of one’s personal norms towards meat consumption in that relationship, and (3) What moderating impact have biospheric values on that effect?

In sum, this research contributes to the field of static versus dynamic descriptive normative messages towards meat consumption by focusing on the underlying motivational process (mediator) and boundary conditions (moderator) of why/when they are most effective in

encouraging people to reduce their meat consumption. Knowing this will help marketers or other

influencers in making their messages aiming for reducing meat consumption more effectively.

2. Literature review

2.1. Intention to reduce meat consumption

A higher number of people indicate that they intend to reduce their meat consumption (Schenk, Rössel, and Scholz, 2018). For example, 46% of the people in the Netherlands intend to eat less meat (AD, 2018; Voedingscentrum, 2018). One of the most common reasons for reducing meat consumption lies in animal welfare concerns (Sanchez and Sabaté, 2019). Less common reasons are environmental, ethical, and personal benefits related to health (Schenk et al., 2018).

Although people intend to reduce meat consumption, their actual behavior shows the opposite because global meat production and consumption is still rising. Since 1961, the manufacturing of meat has increased 4 - 5 fold (Ritchie and Roser, 2017). The prediction for the upcoming years shows similar statistics. Where in 2010, 277 million tons of meat has been consumed, the forecast for 2021 is 332 million (Shahbandeh, 2018).

This paper will focus on the intention to reduce meat consumption since the intention to perform a certain behavior is an indication of the willingness of a person to perform a particular act (Ajzen, 2002). Moreover, the intention is regarded as the best predictor of that behavior (Ajzen, 1991). Indeed, research in meat consumption has shown that behavioral intention also represents the most accurate predictor of meat consumption (Zur and Klöckner, 2014).

2.2. Static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption

Recent research suggests that static and dynamic descriptive normative messages towards meat

consumption can be efficient in influencing the intention to reduce meat consumption and

subsequent behavior (Sparkman and Walton, 2017). A static descriptive normative message

refers to the current state of the norm, such as ‘’six percent of the Dutch population does not eat

meat’’(Sparkman and Walton, 2017). Although they can instigate various processes, for instance,

perceptions about what is accurate, efficient, and appropriate (Cialdini and Goldstein, 2004), the problem with these types of messages is that the desired behavior (reducing meat consumption) mostly represents a minority of the population. As a result, they can be even detrimental

sometimes in cases of high levels of unwanted behavior (Cialdini et al., 1991; De Groot et al., 2013). When people notice that the majority of the people engage in the undesired conduct (meat consumption), it encourages them to change their behavior, but only into the undesired direction because ‘that is normal’ (Sparkman and Walton, 2017). Hence, emphasizing that just six percent of the Dutch population does not eat meat (Voedingscentrum, 2018) can work contra-productive.

To conquer the shortage of static descriptive normative messages, using a dynamic descriptive normative message rather than the static counterpart, can be beneficial (Sparkman and Walton, 2017; Loschelder, Siepelmeyer, Fischer, and Rubel, 2019). A dynamic descriptive normative message explains how specific behaviors of groups have been changed throughout the years (Sparkman and Walton, 2017). For conduct not yet being the entrenched norm, such as reducing meat consumption, dynamic normative messages informing that an increased number of

individuals are adopting the desired behavior can induce people to select the desired behavior (Loschelder et al., 2019). Moreover, behavioral changes (dynamic norms in groups) can inspire people by adapting their conduct to the new behavior (Mortensen et al., 2019).

Dynamic descriptive normative messages towards meat consumption encourage reducing meat

consumption. Sparkman and Walton (2017) concluded that a dynamic descriptive normative

message towards meat consumption was more effective in reducing meat consumption than a

static descriptive normative message. Despite the prevalence of a static descriptive normative

message (‘’recent research has shown that 30% of Americans make an effort to limit their meat

consumption. That means that three in ten people eat less meat than they otherwise would’’-

Sparkman and Walton, 2017, p. 1665) in favor of the undesired behavior (consuming meat),

people were more willing to reduce meat consumption after reading a dynamic descriptive

normative message. That message was: ‘’Recent research has shown that, in the last five years,

30% of Americans have now started to make an effort to limit their meat consumption. That

means that, in recent years, three in ten people have changed their behavior and begun to eat

less meat than they otherwise would’’ - Sparkman and Walton, 2017, p. 1665). Results showed a

reduction of 9.73% in participants‘ intention to reduce meat consumption after reading the static descriptive normative message in favor of lowering meat consumption. However, after reading the dynamic descriptive normative message, this reduction was increased to 28,5% (Sparkman and Walton, 2017), indicating that a dynamic descriptive normative message is more effective in intentions to reduce meat consumption than using a static descriptive normative message.

There are two arguments why dynamic descriptive normative messages might be more effective than static descriptive normative messages. Firstly, pre-conformity plays a significant role in predicting behavior (Loschelder et al., 2019; Sparkman and Walton, 2017). Pre-conformity refers to the future descriptive norm. When a particular action happens more, individuals will adhere to the evolving norms as if it happens today since they anticipate to continual change (Sparkman and Walton, 2017). That is, individuals’ representations of norms are responsive to information beyond the present (Shrum, 2009). Human species are inclined to conform towards social norms because adjusting towards a social group is intrinsically rewarding (Gallagher, 2019). Hence, individuals will react more positively as they see that the norm is changing (Higgs, 2015).

Secondly, people’s interpretations of norms are susceptible to imaginary and fictional worlds (Shrum, 2009). Fictional work can have such an effect on perceptions of reality that individuals can utilize them to inspire positive change in behavior (Paluck, 2009). Considering the potential positive effects of fictitious worlds, it can be suggested that dynamic descriptive normative messages can lead people to believe in such ‘continuously change’. This can inspire them to adopt the changed norm themselves.

Concluding, dynamic descriptive normative messages indicating a trend towards reducing meat consumption seem to be more effective in encouraging people to reduce their intentions to eat meat than a static descriptive normative message indicating that most people are eating meat.

Hence, the following hypothesis is derived:

H1: A dynamic descriptive normative message towards meat consumption has a stronger impact

on the intention to reduce meat consumption than a static descriptive normative message towards

meat consumption.

2.3. Personal norms as a mediator between descriptive normative messages towards meat consumption and the intention to reduce meat consumption

Literature suggests that normative social messages are effective in changing pro-environmental intentions, such as the intention to reduce meat consumption, because the activation of a social norm directly impacts a personal norm (Kim and Seock, 2019). According to the Norm

Activation Theory (Schwartz, 1977), one is expected to behave pro-socially, including pro- environmentally (e.g., Thøgersen, 1996), after reflecting on one’s personal norms. So, the conduct of individuals is primarily driven by triggered personal norms (De Groot et al., 2013).

Personal norms are connected to the self-concept (how we see ourselves) and interpreted as a moral duty to perform a specific behavior (Schwartz, 1973, 1977; Steg and De Groot 2019).

Personal norms differ from social norms since they explain inner standards relating to a particular response, rather than to rules enforced externally. Behavioral control is governed by internal rather than external mechanisms (Kallgren et al., 2000). A personal norm is, leastwise somewhat, originated from individuals’ conscious reasoning and contemplation, regardless of social expectations (Thøgersen, 2009).

Research showed that feelings of pride arise when individuals comply with their personal norms, whereas perceptions of guilt grow when individuals do not comply with their personal norms (Bamberg et al. 2007; Onwezen et al., 2013). Because of this essential underlying motivational process, personal norms are a significant predictor of behavioral intentions (Bamberg et al.

2007). Earlier studies have shown that their intention towards various pro-environmental behaviors, such as reducing car usage (Nordlund and Garvill, 2003) or higher public transportation usage (Bamberg et al., 2007), are stronger as well.

Personal norms are internalized social norms (Thøgersen, 2006). The more intensely an individual internalizes a social norm, the stronger impact that norm will have on pro-

environmental behavior. It is suggested that personal norms mediate the effect of social norms on

pro-environmental intentions (Thøgersen, 2006). For example, a study by Nayum and Klöckner

(2014) about people’s considerations about buying fuel-friendly cars found that personal norms

mediated the influence of social norms on purchase behavior. Some other studies in various behavioral contexts, such as compensating for environmental conservation (López-Mosquera, García, and Barrena, 2014), and buying organic foods (Thøgersen and Ölander, 2006), proved that personal norms attenuated the impact of social norms on pro-environmental behavior. The outcome of these researches suggests that social norms affect personal norms, which in turn affect pro-environmental behavior.

Personal norms are relevant in a pro-environmental context (De Groot and Steg, 2009). Also, previous research outlines that personal norms mediate the effect of social norms on pro- environmental behavior (Thøgersen, 2014). This paper assumes that personal norms towards reducing meat consumption will also be relevant in mediating the effect of descriptive normative messages on the intention to reduce meat consumption. Cialdini et al. (1991) claimed that one’s personal norm solely will have a significant impact on the intention to act pro-environmentally when it is made salient before conducting an experiment. However, personal norms can be triggered after reading a normative message (Bruynzeel, 2019). Moreover, recent research suggested that normative messages towards reducing meat consumption directly impacts the development of one’s personal norms towards meat consumption, and indirectly the intention to reduce meat consumption (Amiot, Boutros, Sukhanova, and Karelis, 2018).

Concluding, descriptive normative messages towards meat consumption seem effective in changing the intention to reduce meat consumption because they help to develop one’s personal norms towards meat consumption, which in succession affects the intention to reduce meat consumption. Therefore, the following hypothesis is composed:

H2: Personal norms towards meat consumption mediate the relationship between descriptive

normative messages towards meat consumption and the intention to reduce meat consumption.

2.4. Biospheric values as a moderator on the effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption

This section introduces biospheric values as a potential moderator on the effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption since they are also relevant in environmental contexts (Nigbur et al., 1994).

Reducing meat consumption can be characterized as a social dilemma because self-serving interests and the interests of society are often at odds (Dawes and Messick, 2000). When facing a moral dilemma, such as eating meat, people need to decide whether to go for the direct and self- interest consequences (often more detrimental for the environment) or to go for the indirect and uncertain advantages for the society when acting pro-environmental (De Groot and Steg, 2009).

For example, when an individual eats less meat, which is environmentally beneficial (for example, reducing pollution), it can also imply some individual disadvantages. A downside is, for instance, decreased taste. Thus, the societal and environmental benefits of reducing meat consumption are often long-term and uncertain, but in addition seem to be, to some extent, at odds with self-interests (De Groot and Steg, 2009).

Research on social dilemmas has found that behaviors incorporating social dilemmas depend on the person’s values (Messick and McClintock, 1968; Van Lange, Joireman, Parks, and Van Dijk, 2013; Van Lange and Visser, 1999). Values are desirable trans-situational objectives and serve as navigated principles in a person’s or other societal entities‘ lives (Schwartz, 1992). Feather (1995) defines values as abstract systems involving people’s beliefs regarding end states what they desire. Values are adequate prognosticators regarding pro-environmental behavior (Stern and Dietz, 1994; Nigbur et al., 2010) and explain several attitudes and behaviors of individuals in an environmental context (Dietz, Kalof, and Stern, 2002; Schultz and Zelezny, 1999).

Within value orientations, self‐transcendence and self-enhancement values are most relevant to

pro-environmental behavior because both values forecast concern for the environment (Schultz

et al., 2005; De Groot and Thøgersen, 2019). Values of universalism and benevolence describe

self-transcendence, values of performance and power characterize self-enhancement (Schultz et al., 2005). Additional research confirmed that self-transcendence and self-enhancement values are also connected to meat consumption (Abrahamse, 2019).

Research analyzing the association between values and environmental worries has shown that self-transcendence values positively and self-enhancement values negatively predict general worries regarding environmental issues (Schultz et al., 2005). Also, self-transcendence values are positively correlated with self-indicated pro-environmental behavior (Karp, 1996), whereas values of self-enhancement are negatively correlated with self-indicated pro-environmental behavior (Stern and Dietz, 1994). Other researchers found that people endorsing strong self- transcendent values are generally more inclined to behave pro-environmentally than individuals upholding weak self-enhancement values, while the opposite is true for self-enhancement values (De Groot et al. 2016; Engqvist Jonsson and Nilsson, 2014; Steg and De Groot, 2012). For instance, individuals who are willing to reduce their meat consumption are more likely to endorse self-transcendence rather than self-enhancement values (Allen and Hung Ng, 2003).

Also, most studies have shown that people who strongly endorse self-transcendent are more likely to “cooperate” in behaviors involving a social dilemma, while the opposite is true for people who more strongly endorse self-enhancement values (Kalof, 1999; Schultz et al., 2005;

Thøgersen and Ölander, 2002). Therefore, self-transcendent values are more likely to be positively and self-enhancement values negatively related to the intention to reduce meat consumption (De Groot and Thøgersen, 2019). Hence, this research focused on self- transcendence values specifically.

The self-transcendence values distinguish between two specific values: altruistic and biospheric values (De Groot and Thøgersen, 2019). Altruistic values refer to a concern with the welfare of other human beings, which means that someone who strongly endorses altruistic values will more likely intend to reduce their meat consumption depending on whether the societal benefits of obeying to this behavior outweigh the societal costs (De Groot and Thøgersen, 2019).

Biospheric values refer to concern for the quality of nature and the environment for its own sake

(De Groot and Thøgersen, 2019). When someone strongly endorses biospheric values this person

will decide to reduce their meat consumption when the benefits outweigh the costs for the environment and nature (De Groot and Steg, 2009). Biospheric values are shown to be even more relevant to pro-environmental behavior than altruistic values because most pro-

environmental behaviors are strongly motivated by environmental and biospheric concerns (De Groot and Thøgersen, 2019; Schultz et al., 2005). Furthermore, as reducing meat consumption is strongly associated with substantial environmental benefits, and consequently regarded as a typical pro-environmental behavior (Stoll-Kleemann and Schmidt, 2017), this paper focused on biospheric values specifically. The stronger people value aspects of the biosphere and

environment for its own sake, the more likely they will weigh the perceived environmental costs and benefits of reducing their meat consumption over other (dis)advantages of reducing meat consumption (Austgulen, et al., 2018). Concluding, the following hypothesis is derived:

H3: Biospheric values positively impact the intention to reduce meat consumption.

The effectiveness of normative messages towards meat consumption on the intention to reduce meat consumption depends on one’s biospheric values. Various studies have shown that individuals having higher levels of biospheric values have a higher awareness of the possible adverse consequences of not acting pro-environmentally than people subscribing lower levels of biospheric values (De Groot et al., 2008; Schultz et al., 2005).

Also, individuals with higher levels of biospheric values are driven to comply with their

environmental self-identity and consider themselves as a person who acts pro-environmentally

(Ruepert, Keizer, Steg, Maricchiolo, Carrus, Dumitru, and Moza, 2016). The environmental self-

identity is the degree someone sees themselves as a person who performs pro-environmentally

(Van der Werff, Steg, and Keizer, 2014). Reducing meat consumption is connected to acting in

line with one’s environmental self-identity (Stoll-Kleemann and Schmidt, 2017). Hence, people

with stronger levels of biospheric values will not be that responsive towards normative messages

towards meat consumption since they probably already lower their meat consumption. Therefore,

their intention to reduce meat consumption will not show significant differences after reading

either a static or a dynamic descriptive normative message towards meat consumption.

In contrast, individuals with lower levels of biospheric values have less vigorous environmental self-identities (Ruepert et al., 2016). For instance, they experience less intense feelings of moral duties to perform pro-environmentally, such as reducing car usage, than people endorsing higher levels of biospheric values (De Groot et al., 2012). Therefore, they are more prone to ‘listen’ and adapt to normative messages than people endorsing higher levels of biospheric values because they will use their ‘external compass’ (Steg, Perlaviciute, Van der Werff, and Lurvink, 2014).

However, since the static descriptive normative message towards meat consumption is clearly against the desired behavior (reducing meat consumption), they will notice that the majority eat meat, and thus consider eating meat as normal (Mortensen et al., 2019; Sparkman and Walton, 2017). Therefore, it will not encourage them to change their undesired behavior (consuming meat). Hence, reading the static message will not significantly increase their intention to reduce meat consumption. On the other hand, they will show significant higher intentions to reduce their meat consumption after reading the dynamic descriptive normative message towards meat

consumption. Paragraph 2.2 has discussed two arguments about why these messages might be more effective than their static counterparts.

In sum, while the effect of the static descriptive normative message towards meat consumption on the intention to reduce meat consumption will be low regardless of one’s biospheric values, the dynamic message will be more impactful the weaker one's biospheric values. Concluding, the following hypotheses are derived:

H4: Biospheric values positively moderate the effect of descriptive normative messages towards meat consumption on the intention to reduce meat consumption.

H4a: The effect of the static descriptive normative messages towards meat consumption on the intention to reduce meat consumption will be low, regardless of one’s biospheric values.

H4b: The effect of the dynamic descriptive normative messages towards meat consumption on

the intention to reduce meat consumption will be more impactful the weaker one’s biospheric

values.

2.5. Conceptual model

Summarizing, this study examines the effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption. Also, this research investigates whether the effect of descriptive normative messages towards meat consumption on the intention to reduce meat consumption is mediated by one’s personal norms towards meat consumption. Furthermore, the main effect of biospheric values on the intention to reduce meat consumption will be studied. Finally, the moderating role of biospheric values on the effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption will also be studied. More specifically, we will test whether the effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption is qualified by biospheric values (moderator). The following hypotheses (see Figure 1) are derived:

Figure 1: Conceptual model.

.

H3 H4

H2 H2

H2

Descriptive normative messages H1

towards meat consumption:

Static versus dynamic

Biospheric values

Intention to reduce meat consumption

Personal norms

3. Methodology

3.1. Population and sampling method

The population of this research comprised any adult Dutch flexitarian or meat-eater. A flexitarian is a person who consumes at least one day per week a meal without meat (Sanchez and Sabaté, 2019). Raphaely and Marinova (2014) described a flexitarian as a part-time vegetarian who tries to reduce his or her meat consumption. Since vegans or vegetarians comprise individuals who do not eat meat at all, it is decided to leave those people out of the study. People who do not eat meat cannot reduce their meat consumption. Therefore, participants were excluded from the results when they regarded themselves as a vegan (n = 3), vegetarian (n = 7), or when they indicated the frequency of their weekly meat consumption with ‘never’ (n = 1).

Furthermore, to ensure participants have had a reasonable opinion about their meat consumption and ownership about what they buy and eat, only adults were allowed to participate. People under the age of eighteen are still not wholly competent to make thoughtful choices (Arshagouni, 2006). Consequently, two participants below the age of eighteen were excluded from the results.

Also, respondents who did not respond to all of the required questions (n = 15) were also excluded from the data set. Furthermore, no outliers were found. Thus, no data points differed significantly from the overall pattern of the data set. Therefore, there was no risk of skewed results and or incorrect inferences (Moore, McCabe, and Craig, 2012). Finally, since this study studies Dutch citizens, people from outside the Netherlands (n = 4) were also excluded from the results when they indicated their nationality with ‘other’.

After removing respondents who did not meet the requirements of the study (n = 30), the final sample existed by 332 people. With having 332 participants, enough respondents were collected.

First, based on the Dutch population of 17,397,620 citizens (CBS, 2019), using a confidence interval of 90%, and margin of error of 5%, 273 or more completed surveys were needed to draw persuasive conclusions (Calculator.net, 2019). Second, a sample size of more than 300

participants ensured a ‘good’ sample size (Comrey and Lee, 2013). Furthermore, the large

number of respondents contributed to enlarged statistical power, lowered the estimated error of the research, and contributed to a more precise representation of the population (VanVoorhis and Morgan, 2007). Finally, enough respondents were collected to satisfy the general criterion of 300 cases for performing factor analysis (Tabachnick and Fidell, 1996).

Convenience sampling has its limitations, such as not being entirely representative of the entire population (Martínez-Mesa, González-Chica, Duquia, Bonamigo, and Bastos, 2016). However, there are two significant arguments on why/when they can be appropriate. First, it is almost impossible for studies like Master’s theses to equally randomize in huge populations like the Netherlands because students have limited resources, workforce, and time (Etikan, Musa, and Alkassim, 2016). Second, convenience sampling is useful in detecting trends, for instance, identifying trends in reducing meat consumption (Etikan et al., 2016). Therefore, participants were indirectly approached via social media (Facebook, WhatsApp, and LinkedIn) and directly contacted at the University of Groningen and University Library with tablets and smartphones.

3.2. Research design

This study comprised a one-way between-subjects design with one independent variable (descriptive normative messages towards meat consumption) with two levels (a static versus a dynamic message). The between-subjects design assured that comparisons between different groups of subjects could be made (Blumberg, Cooper, and Schindler, 2014).

Furthermore, to ensure that the variations in the experimental conditions were caused by

manipulation and not due to external factors, participants were randomly allocated to either a

static or a dynamic descriptive normative message towards meat consumption. Therefore, this

paper was able to draw reliable cause-and-effect conclusions from the data set, indicating good

internal validity (Somekh and Lewin, 2005). After that, having more females (n = 184) than

males (n = 148) in the sample was not an issue since the critical factor of this inquiry was not

gender. Then, a large effect size was assured since more than thirty respondents per condition

were collected, representing the required minimum number of respondents for an AN(C)OVA

for a power of 80%, (Cohen, 1988). Also, a control group was not included since this study

focused on the (dis)similarities between the different conditions instead of comparing the outcomes towards a point of reference (Somekh and Lewin, 2005). Therefore, only the two experimental conditions were used for this research. Finally, the survey was translated into Dutch to prevent loss in translation (Van Nes, Abma, Jonsson, and Deeg, 2010). The method used was back- and forward translation. First, two independent translators translated the items from English into Dutch. For the forward reading, two people having Dutch as their mother language were asked to bring the survey ‘back’ from Dutch into English. Consequently, the quality of the questionnaire remained secure (Tsang, Royse, and Terkawi, 2017). Data were collected between November 6

and November 15

, 2019.

3.3. Materials

The independent variable was manipulated by presenting two descriptive normative messages towards meat consumption: a static and a dynamic message. Some considerations impacted the design of the messages. Firstly, normative messages should relate to a reference group

(Neighbors, O'Connor, Lewis, Chawla, Lee, and Fossos, 2008). Individuals use reference groups for comparing and guiding their behavior and are an essential determinant of the behavior of an individual (Eisenstadt, 1954). For example, perceptions of individuals regarding reference groups impact decisions relating to products and brands (Bearden and Etzel, 1982). Moreover, it influences the effect of social norms on individuals because it teaches people how they are expected to act in a certain situation. The more people identify themselves with a reference group, the higher the likelihood that they conform themselves to the standards of that group (Neighbors et al., 2008). Hence, the reference group in this research was the Dutch population.

To ensure that the impact of the reference group on the intention to reduce meat consumption was controlled, both conditions contained a message referring to the Dutch population.

Secondly, the messages were manipulated by presenting a percentage of the Dutch people who are trying to reduce its meat consumption, representing the reference group of the participants.

Within the normative message framing, the displayed information should be believable

(Berkowitz, 2004). When the information provided is questionable or improbable, people are not

likely to adopt the desired behavior (Berkowitz, 2004). Therefore, the Dutch newspaper

Algemeen Dagblad was used to find a credible percentage since 59% of the Dutch citizens trust the Dutch newspapers (Van Dongen, 2019). Furthermore, to make the messages even more credible, the source Voedingscentrum (2018) was included in the messages because that is also a trustworthy source. Their advice is based on an independent scientific advisory body that

advises, for instance, the Dutch government (Voedingscentrum, 2018). Since 46% of the people in the Netherlands intend to eat less meat (AD, 2018, Voedingscentrum, 2018), this research considered 46% a believable percentage. Therefore, the message of Sparkman and Walton (2017), referring to 30% of Americans, was modified into 46% regarding Dutch people.

Additionally, percentages and numbers and terms such as ‘recent research has shown’ were used to make the messages more convincing (Tormala, Brinol, and Petty, 2006). Finally, due to the subtleness of the manipulation, some parts of the messages were portrayed in bold letters to create attention towards the most relevant information (Kühl and Eitel, 2016). Altogether, the two different normative messages towards meat consumption are shown below:

Condition 1: Static descriptive normative message towards meat consumption.

Recent research by the Voedingscentrum (2018) has shown that 46% of the people in the Netherlands make an effort to limit their meat consumption. That means that in the past year, and also this year, four in ten people in the Netherlands have started eating less meat than they otherwise would.

Condition 2: Dynamic descriptive normative message towards meat consumption.

Recent research by the Voedingcentrum (2018) has shown that, in the last five years, 46% of

people in the Netherlands have started to make an effort to limit their meat consumption. That

means that in recent years, four in ten people in the Netherlands have adjusted their behavior

and started eating less meat they otherwise would.

3.4. Measures

Every participant of this study answered questions regarding the dependent variable (intention to reduce meat consumption), the independent variable (descriptive normative messages towards meat consumption), mediating variable (personal norms), the moderating variable (biospheric values). Also, questions regarding the confounding variables and socio-demographics were asked. The survey is presented in Appendix A.

First, the construct intention to reduce meat consumption (dependent variable) was measured by adopting the scale of Ajzen (1991), see Appendix B. After reading one of the two descriptive normative messages towards meat consumption (manipulated construct), participants indicated their intention to reduce meat consumption by answering three statements. Respondents used a seven-point Likert scale from 1=’Strongly disagree’ to 7=’Strongly agree’.

Second, the variable descriptive normative messages towards meat consumption (continuous construct) was measured by adopting a similar scale of Pedersen, Grønhøj, and Thøgersen (2015), see Appendix C. There is not a widely accepted scale yet for measuring descriptive normative messages towards meat consumption. Therefore, similar statements were created.

Participants indicated via a seven-point Likert scale from 1=’Strongly disagree’ to 7=’Strongly agree’ how they agreed with the presented static and dynamic descriptive normative messages.

The first three statements represented static descriptive normative messages, statement four to six dynamic descriptive normative messages.

Third, the construct personal norms towards meat consumption was measured by using the statements of Steg et al. (2005), see Appendix D. Participants indicated via a seven-point Likert scale from 1=’Strongly disagree’ to 7=’Strongly agree’ how they agreed with the presented seven statements regarding their personal norms.

Fourth, the variable biospheric values was measured based on the work of De Groot and Steg

(2008), see Appendix E. Participants indicated via a seven-point Likert scale from 1=’Extremely

unimportant’ to 7=’Extremely important’ four values on how vital that value was for them.

Furthermore, socio-demographics were measured to acquire some more insight into the sample of this research, see Appendix F. More specifically, participants indicated their current food diet (regarding themselves as vegan, vegetarian, flexitarian, or a meat-eater) and the frequency of their weekly meat consumption (never, seldom, sometimes, often, or very often). The last two variables were moreover used as covariates because both variables could affect the dependent variable. It seems plausible that flexitarians or people who seldom or sometimes eat meat have lower intentions to reduce meat consumption than meat-eaters or people who (very) often eat meat since they already lower their meat consumption (Raphaely and Marinova, 2014).

Consequently, flexitarians or people who seldom or sometimes eat meat are a significant predictor of how much less meat an individual intends to eat. Both covariates were not manipulated but just measured.

Furthermore, participants indicated their nationality (Dutch, other). After that, gender (male, female, other, or I would rather not answer this question) and age (open question) were asked.

Thereafter, participants’ highest educational level (less than high school degree, high school degree, Intermediate Vocational Education / MBO, Higher Vocational Education / HBO,

University-level Bachelor’s degree, University-level Master’s degree, doctorate or I would rather not answer this question ) was asked. Then, respondents indicated their current work situation (full time working, part-time working, jobless but looking for a job, jobless and not looking for a job, student, homemaker, self-employed, retired, unable to work or I would rather not answer this question). Finally, their yearly income (less than €20,000, €20,000 to €34,999, €35,000 to

€49,999, €50,000 to €74,999, €75,000 to €99,999, over €100,000, or I would rather not answer this question) was asked.

3.5. Procedure

When starting the questionnaire, respondents were asked to agree with the terms of the research.

They were informed that this research aimed to investigate motives in meat consumption under

Dutch adult flexitarians and meat-eaters and to gain some insight into their social norms,

personal norms, and biospheric values. Their answer would be processed anonymously, and

there were no right or wrong answers.

After accepting the conditions, participants were first asked to answer the sociodemographic questions about their nationality, gender, age, educational level, current occupation, current food diet, and frequency of meat consumption per week. Then, when they met the requirements of the study, people were randomly allocated to one of the two experimental normative messages immediately followed by statements measuring their intention to reduce meat consumption.

Then, respondents answered the six statements regarding static and descriptive norms, personal norms, and biospheric values. After completing the questionnaire, participants were thanked for their cooperation.

3.6. Factor analysis and Reliability analysis

A factor analysis investigated first whether the items of the variables measured the underlying dimensions of the constructs adequately (Malhotra, Hall, Shaw, and Oppenheim, 2006). To proceed as one construct, the KMO-statistic must be ≥ .50, because lower values indicate that further factor analysis is not recommended (Malhotra et al., 2006). Also, Bartlett's test of

sphericity should show a significant p-value (p < .05) because insignificant p-values suggest that further analysis with the same data is not that useful (Malhotra et al., 2006). After that, the

communalities were analyzed, showing the variance of each parameter of the construct compared to the number of items of the construct (Malhotra et al., 2006). They should be all > .40 because low values suggest that certain items do not properly fit with the construct. Therefore, they should be excluded from the scale and further analysis (Malhotra et al., 2006). Finally, a

Reliability analysis consulting Cronbach’s alpha (α) showed whether the scales proved internally consistent results. A factor is strong and reliable enough when α ≥ .60 (Malhotra et al., 2006).

First, the KMO-statistic (.76) and the statistically significant result of Bartlett's test of sphericity

(p < .01) allowed factor analysis for the construct intention to reduce meat consumption. After

that, the communalities showed that the value of the items was all above .40, whereas the α -

value (.92) indicated a reliable scale of Ajzen (1991). Hence, the construct intention to reduce

meat consumption continued as one construct having three items.

Second, the KMO- statistic (.69) and the statistically significant result of Bartlett's test of sphericity (p < .01) allowed factor analysis for the construct descriptive normative messages towards meat consumption. After that, the communalities showed that the value of the items was all above .40. Then, the Reliability analysis indicated a slightly unreliable scale (α = .59) of Pedersen et al. (2015). So, one or more items were likely to be removed from the scale to increase internal consistency. To decide which items could best be removed, a ‘Cronbach’s Alpha if idem deleted’ was performed. This analysis showed that removing one statement (‘’If other Dutch people show that they eat less meat, then I would also like to eat less meat’’) would improve to a borderline alpha of .62. Therefore, this statement was removed from the scale.

Concluding, the construct descriptive normative messages proceeded as one construct having five items.

Third, the KMO - statistic (.90) and the statistically significant result of Bartlett's test of sphericity (p < .01) allowed factor analysis for the construct personal norms towards meat consumption. After that, the communalities showed that the value of the items was all above .40, whereas the α - value (.90) indicated a reliable scale of Steg et al. (2005). Hence, ‘personal norms’ was kept as one construct containing seven statements.

Finally, the KMO- statistic (.77) and the statistically significant result of Bartlett's test of

sphericity (p < .01) allowed factor analysis for the construct biospheric values. After that, the

communalities showed that the value of the items was all above .40, whereas the α - value (.80)

indicated a reliable scale of De Groot and Steg (2008). Hence, the construct biospheric values

proceeded as one construct having four values.

3.7. Assumptions parametric tests

First, a Pearson correlation analysis was conducted because correlations provide insight into the question if and how strong the relationship between the main variables of the conceptual model were correlated (Malhotra et al., 2006). Some assumptions were checked. First, the data proved homoscedasticity after applying Levene’s test and inspection of the P-P plot (Statistics Solutions, 2019). Second, linearity was concluded after a visual review of the scatterplot. Third, no outliers were found, so there was no risk of skewed results and or incorrect inferences (Lumley, Diehr, Emerson, and Chen, 2002). Thereafter, the assumption of normality has been violated after applying the Kolmogorov-Smirnov (p < .01) and Shapiro-Wilk test (p < .01). Having a non- normal sample can be problematic because it might lead to misleading results. This research nevertheless continued with the analyses because non-normal data only slightly impact type I error rates. Moreover, normality can be assumed in samples higher than thirty respondents (Khan and Rayner, 2003) and tests will still show good results when the distribution is just slightly non- normal (Statistics Solutions, 2019). Then, every result in the dependent variable was paired with an observation of the independent variables, assuring paired observations, and adequate overall coefficients (Pal, 2017). Finally, every construct contained more (N = 332) than the

recommended number of fifty cases. Therefore, it was allowed to make inferences regarding the relationships between them (VanVoorhis and Morgan, 2007).

Second, a one-way ANCOVA detected whether there were statistically significant differences between the effects of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption while controlling for current food diet and frequency of weekly meat consumption (= H1). Dummy variables tested the two different conditions, representing the static descriptive normative message (n = 165) and the dynamic message (n = 167). Some assumptions were checked. First, the assumptions of having a

continuous-scaled dependent variable and having a categorical independent variable with two or

more conditions have been met. Second, covariates can be categorical (Statistics Solutions,

2019). Third, cases were independent due to the randomization of the participants to either one

of the two levels of the independent variable (Statistics Solutions, 2019). Fourth, the dependent

variable should be normally distributed across the population (Malhotra et al., 2006). Although

violating the assumption normality of residuals, this research continued interpreting.

Furthermore, the data proved homoscedasticity after applying Levene’s test. The statistically insignificant results confirmed that the population variances were equal: F(1,330) = 0.32, p = .57. After that, there was a statistically insignificant interaction effect between the

independent variable and covariates current food diet (p = .81) and frequency of weekly meat consumption (p = .87). So, the effect of both covariates did not depend on receiving a static or a dynamic descriptive normative message towards meat consumption (Statistics Solutions, 2019).

Then, covariate current food diet and frequency of weekly meat consumption were both included in the model because they did not highly correlate with each other: r = .52, p < .001. Lastly, no significant outliers were found in both conditions of the manipulated independent variable in terms of the intention to reduce meat consumption (Statistics Solutions, 2019).

Third, a Hayes’ Process Macro Mediation Test model 4, using the 95% CI from 5,000

bootstrapped samples (2009) investigated the mediating role of personal norms towards meat consumption (continuous construct) on the effect of descriptive normative messages towards meat consumption on the intention to reduce meat consumption (= H2). The mediation analysis controlled for current food diet and frequency of weekly meat consumption. It was allowed to use this method because bootstrapping methods do not require normally distributed data (Haukoos and Lewis, 2005).

Fourth, a Hayes’ Process Moderation Test model 1 using the 95% CI from 5,000 bootstrapped samples (2009) examined the main effect of biospheric values on the intention to reduce meat consumption was also studied (= H3). The test also examined whether biospheric values

positively moderated the effect of static versus dynamic descriptive normative messages towards meat consumption (manipulated construct) on the intention to reduce meat consumption (= H4).

The moderation analysis controlled for gender, current food diet, frequency of weekly meat consumption, and highest educational level. Some assumptions were checked. First, the

dependent variable and both independent variables were continuous. Second, the Durbin-Watson statistic (DW = 2.00) proved the independence of observations (Statistics Solutions, 2019).

Third, there was a statistically significant linear association between both predicting variables

and the output variable: R² = .19, F(3,328) = 25.64, p < .001. More specifically, the results

showed b = 8.26, p < .05 for static versus dynamic descriptive normative messages and b = 0.90, p < .001 for biospheric values. Moreover, the dependent variable and the interaction term

showed also a statistically significant linear relationship: b = - 0.38, p < .05. Then, the

assumption of homoscedasticity has been met after conducting the Levene’s test and inspection of the P-P plot. Thus, the error variances were all equal regarding the combination of the

independent variable and the moderator (Statistics Solutions, 2019). Furthermore, the VIF-score (VIF = 1.00) between static versus dynamic descriptive normative messages and biospheric values indicated that multicollinearity was ruled out (Aiken, West, and Reno, 1991). So, the moderating variable and the independent variable were not very closely related to each other, allowing distinguishing their effects in the regressions. After that, no outliers were found in the dataset. Lastly, this research continued performing the analysis despite the violation of the normality of residuals.

Furthermore, a two-way ANCOVA examined whether there was a statistically significant two-

way interaction effect between static versus dynamic descriptive normative messages towards

meat consumption and lower versus higher levels of biospheric values in terms of the intention to

reduce meat consumption. The two-way ANCOVA controlled for gender, current food diet,

frequency of weekly meat consumption, and highest educational level. More specifically, the

analysis examined whether the effect of the static descriptive normative message was low,

regardless of one’s biospheric values (=H4a). Also, it was examined whether the dynamic

descriptive normative message towards meat consumption on the intention to reduce meat

consumption was more impactful the weaker one’s biospheric values (=H4b). Some assumptions

were checked. First, the assumptions of having a continuous-scaled dependent variable and

having two categorical independent variables with two or more conditions were met. Second,

covariates can be categorical (Statistics Solutions, 2019). Then, cases were independent due to

the randomization of the participants to either the static or the dynamic normative message

(Statistics Solutions, 2019). After that, this research continued interpreting despite violating the

assumption normality of residuals. Furthermore, the data proved homoscedasticity after applying

Levene’s test: F(3,33) = 1.351, p = .26. The statistically insignificant results confirmed that the

population variances were equal (Statistics Solutions, 2019). Then, there was a statistically

insignificant interaction effect between static versus dynamic descriptive normative messages

and covariate gender (p = .94), current food diet (p = .60), frequency of weekly meat

consumption (p = .87), and highest educational level (p = .71). So, the effect of the covariates did not depend on receiving a static or a dynamic descriptive normative message towards meat consumption (Statistics Solutions, 2019). Also, there was a statistically insignificant interaction effect between lower versus higher levels of biospheric values and covariate gender (p = .91), current food diet (p = .50), frequency of weekly meat consumption (p = .06), and highest educational level (p = .34). So, the effect the covariates did not depend on having lower versus higher levels of biospheric values. Therefore, all the covariates mentioned above were included in the model. Lastly, no significant outliers were found in all of the possible four combinations between the levels of the two independent constructs.

Finally, before conducting the two-way ANCOVA, the median split method transformed the continuous construct biospheric values into a dichotomous construct consisting of two groups (De Coster, Gallucci, and Iselin, 2011). Dummy variables were created existing of people endorsing lower and higher levels of biospheric values. Scores below the median (23) were labeled as ‘lower levels of biospheric values’ (n = 142), values above 23 as ‘higher levels of biospheric values’ (n = 190). Various researchers have criticized the use of median splits, having concerns such as loss in information and decreased power (Cohen, 1983; De Coster et al., 2011).

However, there are three significant arguments about why they can be useful. Firstly, they are advantageous regarding the simplification of the results of the construct, analyses, and the final presentation of the results. Dividing a group into two categories makes it easier to draw

conclusions and detect differences (De Coster et al., 2011). Secondly, there was no risk of misleading results because multicollinearity was ruled out (Iacobucci, Posavac, Kardes,

Schneider, and Popovich, 2015). Thirdly, statistical results will remain valid when the research

occurs in an experimental field or a natural situation (Iacobucci et al., 2015). Since (reducing)

meat consumption is typically behavior that can happen and what can be visible, this paper

deemed this research setting as a natural occurrence.

4. Results

4.1. Descriptive statistics

Descriptive statistics regarding socio-demographics are briefly discussed below and presented in Appendix F. The final sample existed by 332 people. Equal to having slightly more females (8.7 million) than males (8.6 million) in the Netherlands (Statista, 2019), slightly more females (55.4%) than males (44.6%) participated in this study. Participants were between 18 and 73 years old (M = 26.55, SD = 0.00), denoting a reasonably young sample compared to the average age of the Netherlands of 42 years in 2015 (Plecher, 2019).

Furthermore, the majority of the sample were highly educated, as almost 43% had a Bachelor's or Master's degree on University level or had joined Higher Vocational Education / HBO (37.3%). Since only fifteen percent of the Dutch citizens completed a Bachelor's or Master's degree on University level, the highest educational level of the sample deviated from the average of the population (Maslowski, 2018).

Then, the majority of the sample was a student (59%), worked full time (22.6%), or worked part- time (13%). Also, most participants (65.4%) earned less than €20,000, whereas 12.3% earned

€20,000 to €34,999. Finally, 11.2% earned €35,000 to €49,999, and approximately six percent earned more than €50,000. The average yearly income of the sample was considerably lower than the average income of the Dutch population of €36,000 (Van Elk, Jongen, and Kloot, 2019).

Overall, the present sample seemed to be not fully representative of the Dutch population

(reference group).

4.2. Pearson correlation analysis

The correlations (N = 332) are briefly discussed below and presented in Appendix G. The correlations implied some suggestions concerning the developed hypotheses. There was a significant, moderate positive correlation between descriptive normative messages towards meat consumption (continuous construct) and the intention to reduce meat consumption: r = .29, p <

.001. There was also a significant, moderate positive correlation between descriptive normative messages towards meat consumption and personal norms, r = .22, p < .001. In addition, a

significant, strong positive correlation was detected between personal norms and the intention to reduce meat consumption: r = .69, p < .001. Thus, the stronger one’s personal norms, the higher the intention to reduce meat consumption. The main constructs show a significant correlation with each other, implying that the conditions for the expected main effects of descriptive normative messages towards meat consumption on the intention to reduce meat consumption as well as personal norms as a potential mediator in relation to this relationship are plausible.

Furthermore, biospheric values and the intention to reduce meat consumption did also

statistically significantly and positively correlate with each other, showing a relatively strong correlation: r = .42, p < .001. Thus, the higher one’s biospheric values, the higher one’s intention to reduce their meat consumption. So, the third Hypothesis was likely to be accepted.

Then, confounders current food diet (r = - .58, p < .001) and frequency of weekly meat

consumption (r = - .48, p < .001) did also significant, strong negative correlate with the intention to reduce meat consumption. Finally, there was also a significant, strong negative correlation between confounders current food diet (r = - .44, p < .001) and frequency of weekly meat

consumption (r = - .55, p < .001) and personal norms. Negative correlations refer to relationships

between two variables in which one variable decreases as the other increases, and contrariwise

(Cohen, 1988). Concluding, both confounders did have predictive power on the intention to

reduce meat consumption and personal norms. Therefore, they were included as covariates in the

further analyses.

4.3. The effect of static versus dynamic descriptive normative messages towards meat consumption on the intention to reduce meat consumption

A one-way ANCOVA was performed including static versus dynamic descriptive normative messages towards meat consumption as the manipulation variable, the intention to reduce meat consumption as the dependent variable, and current food diet and frequency of weekly meat consumption as covariates. The output of the analysis is presented in Appendix B.

In contrast to Hypothesis 1, participants reading the static descriptive normative message (M = 4.55, SD = 5.25) showed slightly higher intentions to reduce their meat consumption than participants reading the dynamic descriptive normative message (M = 4.47, SD = 5.23), see Figure 2. However, these differences were not statistically significant while adjusted for the two covariates: F(1,523) = 0.09, p = .77. Thus, the dynamic descriptive normative message towards meat consumption did not have a stronger impact on the intention to reduce meat consumption than the static message, hereby rejecting Hypothesis 1.

Figure 2: Intention to reduce meat consumption after reading the static versus the dynamic

descriptive normative message towards meat consumption.