The effect of positive versus negative descriptive norms and
personal norms on the intention to reduce meat consumption
explained by cognitive dissonance
by
Rob van der Wal
University of Groningen
Faculty of Economics and Business
MSc Marketing Management
09-06-2020
Supervisor: dr. J.J.M. de Groot Second supervisor: dr. J.W. Bolderdijk
Hora Siccemasingel 348 9721HZ Groningen
+31 6 11436274
2
Abstract
3
Table of content
Abstract ... 2
1. Introduction ... 5
2. Literature review... 8
2.1 Reducing meat consumption: Intention to eat less meat ... 8
2.2 Social norms: Influence of descriptive norms on the intention to eat less meat ... 8
2.3 The impact of personal norms on the intention to eat less meat ... 10
2.4 The moderating role of personal norms on negative/positive descriptive norms and cognitive dissonance ... 12
2.5 The present study ... 14
3. Methodology ... 15
3.1 Population and sampling method ... 15
3.2 Research design and procedure ... 16
3.3 Materials ... 17
3.4 Measures ... 18
3.5 Plan of analysis ... 20
3.6 Factor analysis and reliability analysis ... 22
4. Results ... 24
4.1 Descriptive statistics ... 24
4.2 The effect of positive versus negative descriptive norms on the intention to reduce meat consumption ... 24
4.3 Personal norms as a moderator on the relationship between positive/negative descriptive norms and intention to reduce meat consumption ... 26
4.4 Personal norms as a moderator on the relationship between positive and negative descriptive norm and cognitive dissonance ... 28
5. Conclusion and discussion ... 29
4
5.2 Discussion and practical implications ... 30
5.3 Limitations and future research ... 33
References ... 36
Appendix ... 45
Appendix A: Questionnaire ... 45
Appendix B: Communalities cognitive dissonance construct ... 51
5
1. Introduction
Already in the 1970s, the public was warned about climate change (De Wereld Morgen, 2015). Since then some regions of the world have already faced the consequences of climate change (Edenhofer, et al., 2014). It should not come as a surprise that climate change will have an impact on all life on earth, especially on vulnerable and poor people (Mayor, 2009). An important driver for this problem is the urge from people to keep consuming until the point that it is becoming unsustainable (Graca, Godinho, & Truninger, 2019).
Evidence that human consumption patterns are moving beyond the carrying capacity of the earth and is becoming unsustainable is growing (Hansen, 2005; Brulle & Young, 2007; Thøgersen, 2014). There are almost no consumption behaviours that have zero impact on the natural environment but the food sector, amongst others, is one of the most problematic sectors concerning the pressure caused on the environment (European Environment Agency, 2010). One specific area that can be described as highly taxing is the agricultural sector (Carlsson-Kanyama & Gonzalez, 2009).
The agricultural sector is currently responsible for almost 25% of all greenhouse gas emissions worldwide (Smith, et al., 2014). The amount of emissions coming from the agricultural sector is higher than all the transport over land, sea and air combined (Godfray, et al., 2018). Moreover, nearly 80% of all land on earth is currently used to grow livestock (United Nations Food and Agriculture Organization, 2010). However, all this land only produces about 20% of the total supply in calories worldwide (Ritchie & Roser, 2020). Next to these unsustainable production processes, there is the expectancy that the world’s population will grow to around 10 billion in 2050 (United Nations, 2019), what will result in even more problematic pressure on the environment. Taken together, reducing the growth of this sector by changing people’s diet can help alter the sustainability battle the earth is faced with.
6 will focus on the specific underlying processes of intentions to reduce meat consumption rather than actual behaviour.
Reducing meat consumption can be regarded as a typical social behaviour because it is influenced by standards of society (Thøgersen, 1996). The perceptions of these standards of society are referred to as ‘social norms’ (Cialdini, Kallgren, & Reno, 1991). Indeed, social norms have been acknowledged as an important determinant to understand and change social behaviour(al intentions) (Sharps & Robinson, 2016; Collins, Thomas & Robinson, 2019), especially when they are made salient (De Backer & Hudders, 2015). However, the evidence of when and how social norms are effective at changing social intentions and behaviours, such as meat consumption, is less clear (Higgs, Liu, Collins, & Thomas, 2019). Therefore, this present study will focus on when social norms saliency is most effective at increasing a consumer’s intention to reduce meat consumption. More specifically, social norms have been argued to especially be effective because they can activate one’s internal set of moral rules (Thøgersen, 2006), as translated in personal norms.
7 as well because a misalignment between social norms and personal norms will activate a process of cognitive dissonance reduction.
Cognitive dissonance is a state of mind that people experience when their cognition is opposite of the behaviour that they carry out (Festinger, 1957). When people perceive dissonance, they are driven to resolve any inconsistencies that they experience (Festinger, 1957). People can do so by changing their current cognition or by generating a new cognition (e.g. changing social norms towards reducing meat consumption). In the present paper, we argue that personal norms are not only relevant because they directly predict subsequent intentions and behaviours, but they also influence behaviour(al intentions) because they influence how they perceive/interpret external cues, such as salient descriptive norms via the process of cognitive dissonance reduction. More specifically, when an individual has a strong moral belief about a certain topic (e.g. meat reduction), the level of support they get from their surroundings should not impact them severely (Schultz, Messina, Tronu, Limas, Gupta & Estrada, 2016). This suggests that the effect of making salient social norms, that explains what others do, will probably less strongly influence the behaviour(al intentions). Indeed, research confirms that personal norms significantly impact the effect of social normative messages: if an individual has a stronger moral belief on how to behave, this person will be less affected by social norms (Schultz et al., 2016). We believe that, based on cognitive dissonance, the misalignment between social norms and personal norms will result in a lower intention to reduce meat consumption than when social norms and personal norms are in line with one another. However, the few studies that have shown that personal norms moderate the effect of social norms (Schultz et al., 2016; Stok, De Ridder & De Wit, 2012; Van Herpen, Van Trijp & Van Amstel, 2012), do not explain why this happens. We will examine whether this effect occurs because of a misalignment between social norms and personal norms, which ultimately, results in feelings of cognitive dissonance. If there is any merit in these assumptions, it can provide guidelines to practitioners to design effective behaviour change campaigns as to how to reduce the misalignment between personal norms and social norms. Furthermore, it can benefit the research on the theory of cognitive dissonance by finding empirical evidence for the influence of cognitive dissonance on the effectiveness of social norms on intention. Therefore, the present study will investigate the following research question:
8
2. Literature review
2.1 Reducing meat consumption: Intention to eat less meat
There is evidence that industrialized countries are eating less meat than they have done before (Fresco, 2009). For example, 52% of a Finnish sample and around the same number of American consumers intend to decrease their meat consumption in the near future (Latvala, et al., 2012; Johns Hopkins Center for a Liveable Future, 2019). However, what people intend to do and what people actually do is quite often the opposite of each other. A lot of people want to change their diet, but fail to do so (Latvala, et al., 2012). While their intention is good, their ultimate course of action is the exact opposite of what they intend to do. Even though intentions will not always translate in actual behaviour, intentions are deemed to be a good predictor of future behaviour in general (Ajzen, 1991; Sheeran & Norman, 1999).
Research shows that intention is a strong predictor of dietary behaviours, such as the intention to reduce meat consumption. For example, McEachan and colleagues, (2011), show that future dietary behaviour can be better predicted by measuring people’s intention. Also, Coker and Van der Linden (2020), find intention to be an effective predictor of dietary behaviour. Therefore, it comes as no surprise that intention to reduce meat consumption is regarded as a strong predictor for reducing meat consumption (Zur & Klöckner, 2014)
To examine how intentions to reduce meat consumption are formed and through what they are influenced, this paper will examine the underlying processes of intentions to reduce meat consumption rather than actual behaviour. This study will examine the process of why personal norms moderate the social norms- meat reduction behaviour. As these theoretical relationships are expected not to be different for intentions than for actual behaviour, the present study, therefore, focuses on behavioural intention as a proxy for behaviour.
2.2 Social norms: Influence of descriptive norms on the intention to eat less meat
As a society, we have established certain beliefs that we think of as correct or incorrect and we all should act upon those beliefs. These beliefs can be described as a number of unwritten rules of behaviour, which are commonly accepted within a group or society at large (Cialdini et al., 1991). Taken together, these unwritten rules of behaviour are called social norms (Cialdini & Goldstein, 2004; Berkowitz, 2004).
9 with the social norm (e.g., Goldstein et al., 2008; Perkins, 2002; Sharps & Robinson, 2016; Collins et al., 2019). It is undisputed that social norms, when made salient, are an important predictor of dietary behaviour. However, there are also times that social norms do not effectively predict behaviour (Rimal, 2008).
Unlike harming the environment, eating meat is in many cases a well-accepted part of people’s lives and therefore not many people will feel that they act outside the acceptance of the current social norms (De Backer & Hudders, 2015). Although there is evidence that social norms can impact behaviour, there is also evidence that shows that social norms do not always influence behaviour. A meta-study on social influence approaches shows that the effect sizes of social norm related approaches (i.e. making salient social norms) are often fairly small or even absent (Abrahamse & Steg, 2013). For example, Schultz and colleagues (2007) showed that when people already adhere to the social norm, a social norm directed at energy consumption in their study, people increased instead of decreased their energy consumption due to the normative messages. Since social norms do not always influence behaviour or influence the behaviour negatively, it is important to know which social norms are most effective in what circumstances. Within the field of social norms, two types of norms are particularly interesting, the injunctive norm and the descriptive norm (Cialdini & Reno, 1990). The injunctive norm refers to certain beliefs and or rules that society has come up with and tell us what is morally approved or disapproved. The descriptive norm tells us what is commonly done by the majority of a group or society. It describes what is typical or normal. It gives people reason to believe that adapting to this behaviour is the most sensible thing to do (Cialdini & Reno, 1990).
10 rather than injunctive norms towards reducing meat consumption as a relevant type of social norms.
The present research also makes a distinction between two types of descriptive norms, being a positive (e.g. the majority of people approves reducing meat consumption) and a negative descriptive norm (e.g. the minority of people does not approve reducing meat consumption), because if the norms operate independently of each other, making salient a negative descriptive norm towards reducing meat consumption should undermine the intention to reduce meat consumption, while on the other hand making salient the positive descriptive norm towards reducing meat consumption should promote the behavioural intention to reduce meat consumption (Staunton, Louis, Smith, Terry & McDonald, 2014). To test these assumptions, the following hypothesis is derived:
H1: Positive descriptive social norms will result in stronger intentions to reduce meat consumption than negative descriptive social norms.
2.3 The impact of personal norms on the intention to eat less meat
Most people want to act in line with what they think of themselves. This relates to the self-concept and is attached to personal norms (Schwartz, 1973). Personal norms help people to be motivated to act in a way that is compatible with their self-image (Schwartz, 1973). For example, if somebody is restricting him/herself from eating junk food out of dietary reasons and later on finds him/herself eating at a hamburger restaurant this can induce feelings of guilt because he/she has violated their personal norm (Onwenzen, Antonides & Bartels, 2013). Personal norms are referred to as a mental construct that refers to feelings of moral obligations to act ‘correct’ (e.g. decrease meat consumption to preserve the earth) and help to motivate people to act in a way that is in line with their moral obligations (Schwartz, 1973). The main difference between personal norms and social norms is that usually, people adhere to social norms because of social pressure and to personal norms because people feel that it is the morally right thing to do (Thøgersen, 2019).
11 the personal norm determines whether, and to what extent, people are willing to give attention to a new environmental problem like the overconsumption of meat. This directly influences the probability that people will change behaviour and to what extent they are able to handle the problem (Stern, Dietz, Kalof, & Guagnano, 1995). Previous work has shown that when people have a stronger personal norm, they are more likely to intend to purchase organic food (Thøgersen & Ölander, 2006), or engage in other types pro-environmental behaviour (Brown, Ham, & Hughes, 2010).
Normative social influence has a clear impact on meat and non-meat related contexts. It is, however, less clear which factors possibly moderate this process (Rimal & Real, 2005). Cialdini et al., (1991) found that normative messages are most effective when the descriptive and injunctive norm are aligned which resulted in the strongest motivation to engage in a behaviour. This means that the effect of normative messages is moderated by the injunctive and descriptive norm. We assume, based on research, that a similar mechanism could exist for a moderating role of personal norms and descriptive norms. For example, Göckeritz, Schultz, Rendón, Cialdini, Goldstein & Griskevicius, (2010), showed that the stronger one’s personal norm towards energy conservation, the lower the correlation between the descriptive norm and conservation behaviour. Furthermore, Schultz et al., (2016), showed that people who possess a strong personal norm are less influenced through normative messages compared to people who do not have a strong personal norm. They were able to find a moderator effect of personal norms on a aligned message (descriptive and injunctive norm both present in the message), but were not able to find a moderating effect of personal norms on a descriptive normative message, nor did they made a distinction between a positive/negative and their message was framed only as a static normative message.
12 misalignment between social norms and personal norms, which ultimately, results in feelings of cognitive dissonance. Therefore, the following hypothesis are derived:
H2: Personal norms have a positive effect on the intention to reduce meat consumption
H3: Personal norms negatively moderate the effectiveness of positive and negative descriptive norms on the intention to reduce meat consumption (i.e. the main effect of a negative versus a positive descriptive social norm will be more likely to be present the less strong one’s personal norm towards meat consumption is)
2.4 The moderating role of personal norms on positive/negative descriptive norms and cognitive dissonance
According to Festinger (1957), people are driven to resolve any inconsistencies that they experience. This phenomenon is a state of mind called cognitive dissonance. People experience cognitive dissonance whenever they believe that their cognition is opposite of the behaviour that they actually carry out (Festinger, 1957). It is an aversive state of mind which people want to reduce. In that sense, it can be related to by feelings of hunger or the urge to reduce thirst. There are two ways to reduce the aversive feelings of cognitive dissonance according to Festinger (1957). This can be done by lowering the importance of the dissonance creating cognition and by changing the current cognition or by generating a new cognition. By doing this, the discrepancy between beliefs and behaviour diminishes or entirely disappear, creating a less aversive state of mind (e.g. changing attitudes towards animal welfare).
The consumption of meat can be seen as morally troublesome behaviour since it violates the concerns for the welfare of animals which is regarded as a specific case of cognitive dissonance in which a belief and a practice are in conflict (Loughnan, Haslam & Bastian, 2010). According to the theory of cognitive dissonance (Festinger, 1957), this perceived inconsistency between a cognition (i.e. animals suffer) and an action (i.e. eating meat) results in an aversive state of mind (Buttlar & Walther, 2019). The cognitive dissonance of omnivores needs to be reduced to avoid the aversive state of mind cognitive dissonance brings to people (Waytz, Gray, Epley & Wegner, 2010).
13 change. This explanation does not take into account the influence of social norms on one’s personal norms. It may be that people ignore a salient social norm when that social norm is not in line with one’s moral compass. If a social norm conflicts with a personal norm it can be easier ignored and therefore will be less influential to change behaviour.
Research found that cognitive dissonance indeed influences behaviour (Bastian, Loughnan, Haslam & Radke, 2011; Elkin & Leippe, 1986). The present study argues that personal norms can moderate the relationship between social norms and behaviour because a misalignment between social norms and personal norms will activate a process of cognitive dissonance reduction (i.e. changing the importance of the perceived social norm). Research showed that when the minority shows the wanted behaviour (e.g. you are compliant when you eat meat) this often leads to less wanted behaviour instead of an improvement of the majorities behaviour (Stok et al., 2012; Van Herpen et al., 2012) and when descriptive social norms are made salient in a specific context, people feel the urge to comply with such norms (Mollen et al., 2013). However, the present study argues that this does not have to be the case if one’s personal norm is strong in relation to the wanted behaviour (e.g. reduce meat consumption as people also have, at least to some extent, their internal moral compass to rely on when deciding to reduce their meat consumption (Thøgersen & Ölander, 2006). If one’s personal norm is strong, making the descriptive norm salient will have less influence on the intention/behaviour of the individual due to perceived cognitive dissonance.
14 their feelings of responsibility by maintaining a personal distance and by claiming that it is the norm to eat meat.
The present study assumes that, based on cognitive dissonance, the misalignment between social norms and personal norms will result in a lower intention to reduce meat consumption than when social norms and personal norms are in line with one another. Schultz and colleagues (2016), showed that people with a strong personal norm towards the desired behaviour (e.g. reduce meat consumption) ‘set aside’ any social norm because when the norm is positive it is already aligned with their personal norm and, therefore, no dissonance is experienced and the social norm will activate the personal norm. When the social norm is negative, the present study assumes that dissonance will be experienced because then the social norm is not in line with one’s personal norm. To decrease these feelings of dissonance, people will set aside the salient negative descriptive norm. The lower one’s personal norm towards the desired behaviour ( e.g. reduce meat consumption), the more likely that these people will not experience any dissonance with social norms regardless of whether the social norm is positive or negative. The present study assumes that this is the case because those people don’t really know whether they believe the desired behaviour is something that they feel any moral obligations to act in a certain way at all. In that case, they do not have an internal moral compass which means that they do not experience any dissonance at all. To test this assumption, the following hypothesis is derived: H4: A stronger misalignment between a positive/negative descriptive norms and personal norms will result in feelings of cognitive dissonance (i.e. personal norm negatively moderate the effectiveness of positive and descriptive norms resulting in more perceived feelings of dissonance)
2.5 The present study
15 Figure 1: Conceptual model
3. Methodology
3.1 Population and sampling method
Data was collected through the use of an electronic questionnaire. A questionnaire gives the possibility to draw large amounts of data from a large number of respondents. It makes quantification of results easier. The questionnaire was spread out via social media and email. All participants were invited to voluntarily answer the questionnaire anonymously and confidentially. All answers were used for research purposes only.
Convenience sampling was the method of data collection because it gives the opportunity to obtain basic data and trends without the difficulties of having a fully representative sample (Blumberg, Cooper, & Schindler, 2014). With this method, there is a risk that the sample is less representative but it has the upside of drawing inferences out of an easy to reach sample within strict time limits for this study.
16 people do not consume meat or animal-based products at all, they cannot intend to decrease meat consumption and are therefore excluded from participating. Furthermore, this study chose to only allow for adult participants as people under the age of eighteen were considered to not be fully able to make independent judgements about meat consumption and what they eat as most of them still fall under the jurisdiction of their parents (Casey, Getz & Galvan, 2008) (n= 10 removed). Next, all participants that did not respond to all the required items were removed (n= 79), as well as participants in the pilot study to check whether the flow of the questionnaire was correct (n= 5). Finally, this study only comprised Dutch citizens which resulted in the exclusion of another 17 respondents. The target population was anyone within reach of the researcher’s personal, social media and email network.
Next, we included two confounding variables, that are, frequency of meat-eating and diet choice. It is harder for someone to reduce meat consumption when that person only eats meat two times a week compared to someone who eats meat every day (i.e. frequency of meat consumption). Similarly, diet choice was included as a confounding variable because flexitarians already reduced their meat consumption to some extent (compared to meat eaters). Both confounding variables were controlled for during the study which enabled us to examine the true effects of descriptive norms, cognitive dissonance and personal norms on intentions only.
To have enough statistical power, a general rule of thumb says that around 30 participants are needed per cell size in an analysis that measures group differences (e.g. ANOVA,) and at least 50 participants for regression analyses (e.g. moderation analysis)(Carmen, Van Voorhis & Morgan, 2007). If these conditions are not met, the study runs the risk of not being able to distinguish an effect size from luck. With at least 30 participants per cell and at least 50 participants for the regression analysis, the statistical power of 0.8 or higher, corresponding to an 80% or more chance that there is indeed a real effect instead of luck is generally satisfied (Blumberg et al., 2014). In other words, the power of the test is the probability of being able to reject the null hypothesis. After applying all the exclusion criteria a total of 279 participants were taken into account for the analysis. This amount of participants satisfies the condition to have enough statistical power to make inferences about the results.
3.2 Research design and procedure
17 comparisons between two groups or subjects can be made (Blumberg et al., 2014). To make sure that any variation in the experimental conditions between groups is created due to the experiment and not due internal factors, all participants were randomly allocated to either the positive or the negative descriptive social norm condition. The dependent variable was intention to reduce meat consumption (Hypothesis 1, 2 and 3) and cognitive dissonance (Hypothesis 4), and, the moderator variable was personal norms.
All participants were asked to agree with the terms of this study when they started the questionnaire. Next, they were informed about the purpose of this study. Participants were told that the aim was to investigate what people’s opinion is on a news article from the Voedingscentrum and what their opinion is to eating meat because we wanted to compare the opinions about the Dutch Voedingscentrum to the Voedingscentrum of other countries. Furthermore, it was explicitly stated that there were no right or wrong answers to prevent socially desirable answers.
All participants that proceeded (i.e. that were not excluded based on the exclusion criteria) were asked to fill in some sociodemographic questions regarding their nationality, age, gender, education level, annual income, diet choice and the frequency of meat-eating per week. Then, participants were randomly allocated to either the positive or the negative normative message condition. Next, they received questions related to their intention to reduce meat consumption, the manipulation check (descriptive social norm), their personal norm towards meat consumption and their perceived cognitive dissonance. Finally, they were thanked and debriefed.
3.3 Materials
The independent variable, descriptive norms, was manipulated by presenting two normative messages towards meat consumption: a negative and positive message. The positive message was formulated as a gain, whereas the negative message was formulated as a loss. The messages were based on the research of Staunton et al., (2014). Participants in the salient positive descriptive norm condition were presented with:
18 Participants in the salient negative descriptive norm condition were presented with:
‘’Recent research by the Voedingscentrum (2019) has shown that around 20% of the Dutch population is either trying, or considering to make an effort to limit the amount of meat they consume. This means that less than a quarter of people like you have started eating less meat than they otherwise would.’’
The percentages in the different messages were not based on actual research. This is because research showed that around half of the population of different studies is interested in reducing meat consumption. For example, Latvala et al. (2012) showed that around 52% of a Finish sample already decreased their meat consumption or was considering to do so within the near future and a survey among American consumers showed that around 50% of American consumers are interested in reducing their meat intake (Johns Hopkins Center for a Liveable Future, 2019). This is comparable to 46% of Dutch people claiming to be interested in reducing meat consumption (Voedingscentrum, 2018). The differences between the groups in terms of percentages in these studies are so small that it is unlikely that there is a noticeable effect of the group percentages. Therefore, it was chosen to create a condition where the effect of a positive versus a negative condition is more likely to be detected. We chose to create a condition using percentages of 80% (positive condition) versus 20% (negative condition). Since these percentages are not based on actual evidence, it was a potential confounding variable in this study. To correct for this, we included a question concerning how reliable the participants believed the message was because some of the participants may already know the true figures around intended meat reduction in the Netherlands. It was investigated in later analysis whether this item was a confounding variable in the sample. Furthermore, to increase the credibility of the message, the source of the research for this study was changed into the Voedingscentrum (2019). The Voedingscentrum is a credible and well-known source of information on dietary behaviour in the Netherlands and therefore likely to be perceived as trustworthy.
3.4 Measures
19
decrease my meat consumption in the near future.‘’ For all items a seven-point Likert scale was used, ranging from 1=’strongly disagree’ to 7=’strongly agree’. The whole questionnaire can be found in Appendix A
The manipulation check, descriptive social norm towards meat consumption, was measured by adopting scales from several scholars. Since there is no standardized scale to measure descriptive normative messages, this study created statements that are similar to the research it was adopted from. More specifically, the difference between positive and negative descriptive norm was based on research from Staunton et al., (2014), combined with research from Mollen et al., (2013), Burger et al., (2010), and Pedersen et al., (2015). The manipulation check was measured by using six items, under which: ‘’Reducing meat consumption is a trend in the Netherlands.’’ For all six items a seven-point Likert scale was used, ranging from 1=’strongly disagree’ to 7=’strongly agree’.
The moderator personal norms towards meat consumption was measured based on research from Göckeritz et al. (2010), Schultz et al. (2016), and Nolan et al. (2008). For example, Göckeritz et al. (2010), used a four-item scale to measure personal norm towards energy consumption. The items of all three pre-mentioned scholars were transformed for this research purpose into items regarding personal norm towards meat consumption. The seven personal norm items were mixed with four general items regarding meat consumption, such as ‘’I eat meat because I like the taste of it,’’ to disguise the aim of the study. The seven items regarding personal norm towards meat consumption and these general questions were all measured using a seven-point Likert scale, ranging from 1=’strongly disagree’ to 7=’strongly agree’.
20
3.5 Plan of analysis
Prior to the main analysis, we prepared the data, including checking descriptive statistics and dummy coding the experimental variable. Descriptive statistics of all constructs were shown because presenting the raw data makes it hard to visualize what the data was showing. It enables the presentation in a meaningful way and makes the interpretation of data easier (Keller, 2011). A dummy code was created to assign participants to either the negative (dummy= 0) or the positive (dummy= 1) descriptive norm condition. This is important because it allows to make inferences about the differences between the two experimental conditions in later analysis. A factor analysis along with a reliability check was conducted to investigate whether the items of all constructs measure the same underlying dimension in the set of items Malhorta, Hall, Shaw & Oppenheim, 2006). For example, all items regarding the descriptive norm towards meat consumption have to measure that underlying dimension to make inferences in the later analysis. It is important that the KMO statistic, which is produced by the factor analysis, is ≥.50. Only then the items from a construct can proceed as one construct. Any number below .50 is regarded not to be reliable enough to make inferences about in later analysis. Furthermore, the Bartlett’s test of sphericity has to show a significant p-value (p < .05) to be considered useful for later analysis (Malhorta et al., 2006). Next to that, the communalities were produced, which shows the variance of each parameter of an item compared to the total number of items that belong to the specific construct. Communalities have to be > .40 to be considered reliable because lower values indicate that some items do not belong in the same construct (Malhorta et al., 2006). Finally, to measure the reliability of all Likert-type scales, a Cronbach α test was used. It measures the internal consistency of all items within a specific construct. It measures whether all items of the theoretical construct (e.g. the six personal norm items) measure the same underlying construct as a group (e.g. personal norm construct). To ensure internal consistency, the Cronbach α has to be at least 0.6 to be considered reliable (Malhorta et al., 2006).
21 analyses with and without controlling for diet choice and frequency of meat consumption, but we will only report both analyses in case the conclusions in relation to rejecting or accepting the hypotheses differ. The assumptions for a one-way ANOVA and ANCOVA are the same in general, except for the fact that in a one-way ANCOVA for each independent variable, the relationship between a dependent variable and covariate is linear. All other assumptions are the same for both types of analysis. First, the lines that express a linear relationship have to be parallel which suggests homogeneity of regression slopes (Malhorta et al., 2006). Second, the Levene’s test for equality of variance which measures whether the population of a sample is equal or not and shows whether the data is homoscedastic or not (Keller, 2011). If the Levene’s test is positive (P < 0.05) this means that variances in the groups are not homogeneous and the assumptions for an ANCOVA are not met. Finally, a Shapiro Wilk test of normality indicates whether or not the data was normally distributed for the dependent variable.
After the one-way ANOVA and ANCOVA, Hayes’ process moderation test (Hayes, 2017) was used to examine the main effect of personal norms towards meat consumption on the intention to reduce meat consumption (Hypothesis 2) and the interaction effect of positive/negative descriptive norms and personal norms (Hypothesis 3) on the intention to reduce meat consumption. First, the basic assumptions of linear regression were checked. In a linear regression analysis, there must be a linear relationship, multivariate normality, no or little multicollinearity, no autocorrelation and homoscedasticity. A Durbin-Watson statistic showed independence of observations, whether there is a linear association between the dependent and independent variable and whether there is no autocorrelation of variables (Malhorta et al., 2006). Furthermore, multicollinearity assumptions were checked using the Variance Inflation Factor (VIF). If multicollinearity is present, the independent variable that causes multicollinearity should be removed and the VIF statistic has to checked again to see whether the problem with regards to multicollinearity is solved (Malhorta et al., 2006). Finally, the Levene’s test for equality of variances tested the assumption of whether or not the data is homoscedastic.
22 categorical variables with two or more conditions. Next to that, there should be independence of observations, meaning that there is no relationship between the observation in each group or between the groups themselves, and multicollinearity and homoscedasticity were checked.
3.6 Factor analysis and reliability analysis
A factor analysis was conducted to investigate whether the items of the different variables measured the same underlying dimensions of all constructs adequately (Malhorta et al., 2006). First, The KMO- statistic (.76) and the significant Bartlett’s test of sphericity (p < .001) allowed the construct intention to reduce meat consumption to be factored. The communalities of all items in the construct were above .40. ‘’I intent to decrease my meat consumption in the near future’’ (.87), ‘’I want to reduce my meat consumption in the following weak’’ (.89) and ‘’I intent to eat less meat in the same meal in the following month’’ (.91). Next to the factor analysis, the Cronbach α was computed and the combination of the three items to measure intention to reduce meat consumption had an α-value of .94, meaning that the three items together make a strong factor which can be used for later analysis. As such the construct intention to reduce meat consumption continued as one construct with three items within the construct.
Second, the KMO-statistic (.88) and the significant Bartlett’s test of sphericity (p < .001) allowed the construct personal norm towards meat consumption to be factored. Important to note is that next to the items regarding personal norms towards meat consumption, there were also general questions regarding meat consumption in the same construct. All general questions were removed, leaving seven items that were used to measure personal norm towards meat consumption. Next, the communalities were checked and showed that the values of all items were above .40. ‘’I feel morally obliged to eat plant-based food instead of meat’’ (.62), ‘’I feel it is morally right to reduce my meat consumption as much as possible’’ (.61),’’ I feel guilty when I eat meat’’ (.55), ‘’There are no moral obligations that I consider when eating/consuming meat’’ (.51), ‘’People like me should persist in reducing meat consumption’’ (.66), and ‘’I get a bad conscience if I choose meat instead of a plant-based alternative’’ (.54). The following reliability analysis showed that all items are internally consistent, α = .68. Thus, the construct personal norm towards meat consumption continued being one construct with having seven items.
23
will be easier for me to do the same’’ (.74), ‘’I believe that my friends and family members are trying to reduce their meat consumption’’ (.48), ‘’the majority of Dutch people are trying to reduce their meat consumption’’ (.74), ‘’Reducing meat consumption is a trend in the Netherlands’’ (.72), ‘’the majority of the Dutch population eats meat’’ (.76), ‘’the minority of the Dutch population is trying to reduce their meat consumption’’ (.68). Next to the factor analysis, the produced Cronbach α showed a value of .34 meaning that the underlying dimensions are not internally consistent. Therefore, the items ‘’the minority of the Dutch population is trying to reduce their meat consumption, the majority of the Dutch population eats meat, if other people I know are eating less meat it will be easier for me to do the same’’ and ‘’I believe that my friends and family members are trying to reduce their meat consumption’’ were removed from the reliability analysis. The two remaining items ‘’the majority of Dutch people are trying to reduce their meat consumption’’ and ‘’reducing meat consumption is a trend in the Netherlands’’ showed a Cronbach α value of .60, which is enough to meet the threshold value of .60 to continue as a construct in later analysis.
24
4. Results
4.1 Descriptive statistics
Descriptive statistics with regards to socio-demographics can be found in Appendix C. Some relevant descriptive statistics are discussed below and presented briefly. The total sample contained 279 participants of whom 137 (49.5%) indicated to be male and 140 (50.5%) of whom indicated to be female. Two participants were not willing to reveal their gender in this study. Furthermore, the average age of the participants was 39, which is around the average of the Dutch population of 42 years (CBS, 2019). The sample was relatively highly educated as 71.7% of the participants owned a diploma in Higher Vocational Education (HBO) or higher. Compared to the average education level in the Netherlands, this sample contained about two times the percentage of people that have graduated Higher Vocational Education or higher (Onderwijsincijfers, 2019). Most participants earned between €20.000 and €50.000 (€54.5%), which is comparable with the Dutch population of an average annual income of €36.500 (CPB, 2019). Overall, based on the socio-demographics, except for the educational level, this sample seemed to be quite representative of the Dutch population. Furthermore, 94 participants indicated to be a flexitarian (33.7%) and 185 participants indicated themselves as meat-eater (66.3%). Out of the total sample, around 75% indicated that they eat meat every weak on a regular basis (often or more). These percentages matter as they show the differences in food diet and the frequency of meat consumption and that can have implications for further analysis.
4.2 The effect of positive versus negative descriptive norms on the intention to reduce meat consumption
25 of normality indicated that the data is not normally distributed within the sample for descriptive norms towards meat consumption (p< .001) and intention to reduce meat consumption (p< .001). Although the data was not normally distributed within the sample, the analysis did continue as the sample size is sufficiently large (n = 277), because with such a large sample, the sample means can be assumed to be approximately normally distributed (Malhorta et al., 2006).
Next to checking the assumptions, both diet choice and frequency of meat consumption were tested to see whether there is an interaction effect between the independent variable and the covariates. We checked the interaction effect because there must be no interaction between the independent variable and the covariate. First, the covariate diet choice showed an insignificant interaction effect (p= .70) and there was also an insignificant interaction effect for the frequency of meat consumption (p= .12), meaning that there is no dependable effect of both covariates on receiving a negative versus a positive descriptive normative message towards meat consumption (Malhorta et al., 2006). A correlation check was done to make sure that both covariates were not similar and showed that both variables do not correlate too much (r= .62, p< .001). In other words, both covariates are largely independent of each other and do not measure the same underlying construct. Therefore, both variables can be included in the model. Next to the covariates, the item that measured to what extent participants deemed the manipulation text credible was investigated to check whether it was a confounding variable. A linear regression model was computed in which each predictor was added separately to investigate the change in the effect of each predictor. The credibility check proved not to be a confounding variable for both the negative and positive descriptive norm condition. The effect of each predictor became stronger after adding the credibility check or remained virtually unchanged. Only when adding a variable that causes a decrease in the effect of the predictor variable, the added variable is considered to be a confounding variable.
26 that these differences were significant at the 10% confidence level (F(1,277)= 3.26, p= .07) (Daihiru, 2008).
To find out whether this borderline significant statistical result was not caused by the addition of covariates, we also ran a one-way ANCOVA to check whether there is a difference in the outcome statistic. While controlling for both food diet and frequency of meat consumption the outcome statistic became insignificant (F(1,277)= 1.92, p= .17). The fact that the outcome statistic of the one-way ANCOVA became insignificant means that the covariates have an impact on the intention to reduce meat consumption. In conclusion, the results of both analyses are mixed: positive descriptive social norms do result in stronger intentions to reduce meat consumption than negative ones, but only when not controlling for covariates, meaning that Hypothesis 1 is partially supported. A complete overview of the conceptual model including the results of all analyses can be found in Figure 2.
4.3 Personal norms as a moderator on the relationship between positive/negative descriptive norms and intention to reduce meat consumption
Hayes’ moderating procedure (2017) was performed to test the main effect of personal norms towards meat consumption on the intention to reduce meat consumption (H2) and the interaction effect of personal norms and positive/negative descriptive norms on the intention to reduce meat consumption (H3). The results of this procedure are reported in Table 1. The analysis was performed first without controlling for food diet and frequency of meat consumption, and, to further validate these results, with these potential confounding variables. Several assumptions were checked prior to the main analysis. The bell-shaped histogram indicated that the assumption of normality was not violated. The VIF-score between negative/positive descriptive norm and personal norm towards meat consumption was 1.00, meaning that there was no concern for multicollinearity. The Durbin-Watson statistic (DW= 1.95) showed independence of observations. The Levene’s test of homogeneity of variance showed that the variances can be assumed to be equal (F(1,275)= 2.20, p= .14). Finally, the scatterplot showed linearity. As no assumptions were violated, we continued with the main analysis.
27 explanatory power of the model as well. A positive effect of personal norms on the intention to reduce meat consumption was found: the stronger one’s personal norm towards meat consumption, the stronger the intention to reduce meat consumption (B= 0.34, p< .001). Hypothesis 2 is therefore accepted. The moderating effect of personal norms and negative/positive descriptive norms towards meat consumption contributed significantly to the model as well (B= 0.16, p< .05) (See Table 2).
To further test Hypothesis 3, we analyzed the direction of the moderator by looking at the conditional effects of the focal predictor at low and high values of the moderator (See Table 2). For those with low personal norms towards meat consumption (1 SD below the mean), positive/negative descriptive norm had a strong influence on intention to reduce meat consumption (B= -1.91, p< .01). For those with high personal norms towards meat consumption (1 SD above the mean), positive versus negative descriptive norms did not significantly influence the intention to reduce meat consumption. (B= 0.39, p = .54). Thus, when one’s personal norm decreased, the effect of a positive/negative descriptive norm on the intention to reduce meat consumption becomes stronger. This finding was confirmed by analyzing the Johnson-Neyman significance region to determine where the moderator effect stops to be significant. The lower the personal norm, the higher the impact of a negative/positive descriptive norm on intention. At around 22 personal norm units, negative/positive descriptive norm were significantly related (B= -.89, p= .05). However, this effect became stronger with the lowest personal norm (7 units), B= -3.33, p< .01. Thus, personal norms negatively moderate the effect of positive and descriptive norms on the intention to reduce meat consumption, hereby supporting Hypothesis 3.
Table 2: Moderation analysis of personal norms (moderator) on descriptive norm-intention to reduce meat consumption relationship (n = 277)
Variable relationships B SE P 95% CI R² F
(4,272) Outcome variable: intention model
summary
< .001*** .45 44.88
Negative/positive descriptive norm 0.80 0.31 < .05* [0.19, 1.42]
Personal norms 0.34 0.05 < .001*** [0.24, 0.43]
Interaction effect of PN*DN 0.16 0.07 < .05* [0.03, 0.29]
28 Frequency of meat consumption -0.67 0.28 < .05* [-1.23, -0.12]
1 SD < mean -1.90 0.67 < .01**
1 SD > mean 0.39 0.64 ns
Note: *** p<.001, **p<.01, *p<.05, ns = not significant
4.4 Personal norms as a moderator on the relationship between positive and negative descriptive norm and cognitive dissonance
To test Hypothesis 4, Hayes’ moderating procedure (2017) was performed (see Table3). The histogram showed a bell-shaped form and therefore showed a normal probability curve and thus resembles a normal distribution of the dependent variable. Also, the Shapiro-Wilk test showed normality of distributions (p= .21) Third, the Durbin-Watson statistic (DW= 1.01) showed independence of observations. Fourth, the Levene’s test of homogeneity of variance showed that the variances can be assumed to be equal (F(1,277)= 2.01, p= .16). Finally, linearity was proven by looking at the scatterplot.
29 Table 3: Moderation analysis of personal norms (moderator) on descriptive norm-cognitive dissonance relationship (n = 277)
Variable relationships B SE P 95% CI R² F
(4,272) Outcome variable: Cognitive
dissonance model summary
< .001*** .10 6.15
Negative/positive descriptive norm 0.87 0.92 ns [-0.95, 2.68]
Personal norms 0.77 0.38 < .05* [0.01, 1.52]
Interaction effect of PN*DN -0.03 0.04 ns [-0.11, 0.05]
Diet choice 0.73 1.77 ns [-2.75, 4.22]
Frequency of meat consumption -0.26 0.83 ns [-1.89, 1.37] Note: *** p<.001, **p<.01, *p<.05, ns = not significant
Figure 2: Overview conceptual model with results of the analysis
30
5. Conclusion and discussion
5.1 Conclusion
The goal of the study was to investigate whether and to what extent positive/negative descriptive norms influence intention to reduce meat consumption, with personal norms as a moderator. Next to that, this study aimed at investigating whether cognitive dissonance arises when a social norm and a personal norm are misaligned. The present paper found that positive versus negative descriptive norms indeed impact behaviour. Both the ANOVA and moderation analysis showed that positive descriptive norms do result in stronger intentions to reduce meat consumption. However, when we added two covariates (diet choice and frequency of meat consumption) to the analysis, the outcome statistic of the ANCOVA became insignificant. Therefore, Hypothesis 1 was only partially accepted. The main effect of personal norms on the intention to reduce meat consumption showed to be present, meaning that a stronger personal norm towards meat consumption results in a stronger intention to reduce meat consumption. More importantly, this paper found that personal norms significantly moderate the effect of positive/negative descriptive norms on the intention to reduce meat consumption, adding to the understanding of the process of normative influence and thus adding evidence to existing literature that personal norms can moderate behaviour(al intentions). This paper showed that when one’s personal norm towards meat consumption is less strong, the effect of a positive/negative descriptive norm towards meat consumption on the intention to reduce meat consumption becomes stronger. Finally, no moderating effect was found in the relationship between positive/negative descriptive norms towards meat consumption and cognitive dissonance. Thus, the underlying process of the moderating role of personal norms on social norms and intentions is likely not attributed to cognitive dissonance, meaning that there probably is a different explanation for the moderator effect.
5.2 Discussion and practical implications
31 consumption). Since the moderator variable influences the strength of the relationship between positive/negative descriptive norm towards meat consumption on the intention to reduce meat consumption, it is possible that the moderator ‘overrides’ the main effect which potentially caused a less strong main effect. Nonetheless, our findings can be a starting point for interventions aimed at helping people to change their eating habits. Practitioners could not only inform people on what they should eat, but they should also create a message that highlights a positive descriptive norm as that is what most influences intention, rather than a negative descriptive norm.
32 Contrary to expectations, misalignment of social norms and personal norms did not result in cognitive dissonance. We expected that the descriptive norm would be ‘set aside’ by respondents when the descriptive norm became in conflict with one’s personal norm creating a process of cognitive dissonance reduction (as it is easier to change or set aside your beliefs in relation to the social norm than it is to change your beliefs in relation to personal norms). Aligning social norms is generally found to have a larger impact on behaviour than when norms are not aligned (Cialdini et al., 2006). We expected a similar mechanism for alignment of descriptive social norms and personal norms but with the results of the analysis, we were not able to confirm this assumption. We did find that the stronger one’s personal norm towards meat consumption, the more dissonance is experienced, but a misalignment between social norms and personal norms towards meat consumption did not provoke experienced dissonance and, therefore, participants did not ‘set aside’ or changed their beliefs in relation to the social norm. This is also confirmed by the absence of a main effect of positive/negative descriptive norm towards meat consumption in relation to cognitive dissonance. It is, therefore, likely that there is an alternative explanation as to how and when cognitive dissonance arises. For example, Göckeritz and colleagues (2010) claimed that inconsistencies between normative beliefs (e.g. misalignment between social norms and personal norms) diminishes the pressure for individuals to conform to the desired behaviour which could be an explanation as to why no cognitive dissonance was observed in this study.
33 hand, it could be that the individuals in this sample are that much concerned about the issue that they refuse to show cognitive dissonance and want to maintain their current beliefs.
Finally, a last explanation for the absence of perceived cognitive dissonance may be that it is hard to measure cognitive dissonance (Chen, 2008). According to Harmon-Jones (2000), people can experience difficulties in reporting their experienced dissonance. Although we used Elliot and Devine’s, (1994) scale to measure cognitive dissonance, it could be that because we asked people to think about their emotions and by asking them to write their emotions paper, we possibly put them in to a rational rather than an emotional state of mind.
Concluding, the present paper added to the general understanding of the effect of positive versus negative descriptive norms towards meat consumption and the relationship with personal norms towards meat consumption on the intention to reduce meat consumption. Our findings suggest that a positive descriptive norm results in stronger intentions to reduce meat consumption compared to a negative descriptive norm. However, the present paper was not able to fully distinguish the factors that ultimately influence intentions to reduce meat consumption when diet choice and frequency of meat consumption were controlled for.
Also, the expected perceived cognitive dissonance was not present in the present paper when the descriptive social norm and personal norm were not in line with each other. We were, however, able to find a clear boundary condition for when a descriptive norm is less effective compared to a personal norm. That is, the stronger one’s personal norm, the less effective a descriptive norm, regardless of whether that norm is positive or negative. These findings can be useful for marketers and policymakers that want to target a population with either low or high personal involvement as that determines whether they should use a descriptive norm or that they should focus on people’s personal involvement. Also, making use of a positive descriptive norm seems to be more effective compared to the utilization of a negative descriptive norm in future campaigns where the target groups personal norm is low.
5.3 Limitations and future research
34 settings with participants are hard to generalize as the ‘real’ world varies in background settings like time, context and characteristics (Lynch, 1999). In other words, results from a ‘lab’ setting experiment such as this study are often regarded to be ‘unrealistic’ that does not have enough relevance for a ‘real’ world setting (Falk & Heckman, 2009). However, this does not mean that research that does not come from the field has no relevance whatsoever. Experiments that are complemented with information that is derived from other research methods has more explanatory power than research that is based on only one method (Falk & Heckman, 2009). Therefore, future studies should examine whether these relationships produce similar outcomes when they are focussed on actual behaviour instead of behavioural intentions.
Secondly, another issue with self-reported behaviour in social studies is that individuals tend to over-report behavioural intentions (Brener, Billy & Grady, 2003). The fact that individuals over-report self-reported behaviours stems from a social desirability component in people’s nature (Edwards, 1957). Overestimating results from self-reported behaviour studies is a threat, especially in health related studies such as this one (Brener et al., 2003), however, Milfont (2009), showed that socially desirable responding does not have a strong impact on the way that people answer a questionnaire. Furthermore, many previous studies have used a similar mechanism to measure intentions because intentions are believed to be closely linked to actual behaviour (McEachan, et al., 2011; Cocker & Van der Linden, 2020; Zur & Klöckner, 2014). Since intentions and actual behaviour are believed to be closely linked, making inferences about intentions rather than behaviour will likely result in comparable insights. Even more so, most studies described in the present paper that investigated the role of social and personal norms based themselves on behavioural intention (Abrahamse & Steg, 2013; Schultz et al., 2007; Schultz et al., 2016; Thøgersen & Ölander, 2006; Brown et al., 2010). As the present study used validated measures from these studies it is easier to compare these results with other similar research. It is, however, important to assess whether future studies produce similar results when examining the relationship between social and personal norms on intention.
36
References
Abrahamse, W., & Steg, L. (2013). Social Influence Approaches to Encourage Resource Conservation: A meta-analysis. Global Environmental Change, 1773-1785.
Aiking, H., & De Boer, J. (2018). The Next Protein Transition. . Trends in Food Science & Technology, 1-8.
Ajzen, I. (1991). The theory of Planned Behaviour. Organizational Behaviour and Human Decision Processes, 179-211.
Ajzen, I. (2006). Constructing a TPB Questionnaire. Retrieved from http://www.people.umass.edu/aizen/pdf/tpb.measurement.pfd.
Bamberg, S., & Schmidt, P. (2003). Incentives, Morality, Or Habit? Predicting Students’ Car Use for University Routes With the Models of Ajzen, Schwartz and Triandis. Environment and Behavior, 264-285.
Bastian, B., Loughnan, S., Haslam, N., & Rakde, H. R. (2011). Don't Mind Meat? The Denial of Mind to Animal Used for Human Consumption. Personality and Social Psychology Bulletin, 247-256.
Berkowitz, A. (2004). The Social Norms Approach: Theory, Research and Annotated bibliography. Social Norms Theory and Reserach, 1-47.
Blumberg, B., Cooper, d., & Schindler, P. (2014). Business Research Methods. 4th edition: McGraw-Hill.
Brener, N. D., Billy, J. O., & Grady, W. R. (2003). Assessment of Factors Affecting the Validity of SelfReported Health-Risk Behavior Among Adolescents: Evidence From the Scientific Literature. Journal of Adolescent Health, 436-457.
Brock, B. (2019). Changing Ethically Troublesome Behaviour: The Causes, Consequences, and Solutions to Motivated Resistance. Social Issues and Policy Review, 63-92.
37 Brulle, R. J., & Young, L. E. (2007). Advertising, Individual Consumption Levels, and the
Natural Environment. Sociological inquiry, 522-542.
Burger, J., Bell, H., Harvey, K., johnson, J., Stewart, C., Dorian, K., & Swedroe, M. (2010). Nutritious or Delicious? The Effect of Descriptive Norm Information on Food Choice. Journal of Social and Clinical Psychology, 228-242.
Buttlar, B., & Walther, E. (2019). Dealing With the Meat Paradox: Threat Leads to Moral Disengagement from Meat Consumption. Appetite, 73-80.
Carfora, V., & Caso, D. (2017). Correlational Study and Randomised Controlled Trial for Understanding and Changing Red Meat Consumption: The Role of Eating Identities. Soc Sci Med, 244-252.
Carfora, V., Caso, D., Sparks, P., & Conner, M. (2017). Moderating Effects of Pro-environmental Self-identity on Pro-Pro-environmental Intentions and Behaviour: A Multi-behaviour Study. Journal of Environmental Psychology, 92-99.
Carlsson-Kanyama, A., & Gonzalez, A. (2009). Potential Contributions of Food Consumption Patterns to Climate Change. American Journal of Clinical Nutrition, 1704-1709. Carmen, R., Van Voorhis, W., & Morgan, B. L. (2007). Understanding Power and Rules of
Thumb . Tutorials in Quantitative Methods for Psychology, 43-50.
Casey, B., Getz, S., & Galvan, A. (2008). The Adolescent Brain. Developmental Review, 62-77.
CBS. (2019, December). Bevolking; kerncijfers. Retrieved from
https://opendata.cbs.nl/statline/#/CBS/nl/dataset/37296ned/table?fromstatweb
Chen, K. M. (2008). Rationalization and Cognitive Dissonance: Do Choices Affect or Reflect Preferences. Cowles Foundation Discussion Paper, 1-19.
Cialdini, R. B., & Reno, R. R. (1990). A Focus Theory of Normative Conduct: Recycling the Concept of Norms to Reduce Littering in Public Places. Journal of Personality and Social Psychology, 1015-1026.
38 Cialdini, R., Demaine, L., Sagarin, B., Barret, D., Rhoads, K., & Winter , P. (2006). Managing
Social Norms for Persuavive Impact. Social Influence, 3-15.
Cialdini, R., Kallgren, C., & Reno, R. (1991). A Focus Theory of Normative Conduct: A theoretical refinement and reevaluation of the role of norms in human behavior. Advances in experimental social psychology, 201-234.
Coker, E., & Van der Linden, S. (2020). Fleshing Out the Theory of Planned Behavior: Meat Consumption as an Environmentally Significant Behaviour. Current Psychology, 1-10. Collins, E., Thomas , J., & Robinson, E. (2019). Two Observational Studies Examining the Effect of a Social Norm and a Health Message on the Purchase of Vegetables in Student Canteen Settings. Appetite, 122-130.
CPB. (2019). Kerngegevenstabel. Retrieved from
https://www.cpb.nl/sites/default/files/omnidownload/Kerngegevenstabel%202017-2020%2C%20plus%20koopkracht_0.pdf
Dahiru, T. (2008). P-value, A True Test of Significance? A Cautionary Note. Annals of Ibadan postgraduate medicine, 21-26.
De Backer, C., & Hudders, L. (2015). Meat Morals: Relationship Between Meat Consumption, Consumer Attitudes Towards Human and Animal Welfare and Moral Behavior. Meat Science, 68-74.
De Correspondent. (2017, February 28). Reconstructie: Zo Kwam Shell Erachter dat Klimaatverandering Levensgevaarlijk is (en ondermijnde het alle serieuze oplossingen). Retrieved from https://decorrespondent.nl/6262/reconstructie-zo-kwam-shell-erachter-
dat-klimaatverandering-levensgevaarlijk-is-en-ondermijnde-het-alle-serieuze-oplossingen/690128758-e657cfa2
De Wereld Morgen. (2015, December 28). Oliebedrijven Wisten in Jaren 1980 al van
Klimaatprobleem Door Aardolie. Retrieved from
https://www.dewereldmorgen.be/artikel/2015/12/28/oliebedrijven-wisten-in-jaren-1980-al-van-klimaatprobleem-door-aardolie/
39 Edwards, A. (1957). The Social Desirability Variable in Personality Assessment and Research.
New York: Holt, Rinehart and Winston.
Elkin, R., & Leippe, M. (1986). Physiological Arousal, Dissonance, and Attitude Change: Evidence for a Dissonance-arousal Link and a ''don't remind me'' Effect. Journal of Personality and Social Psychology, 55-65.
Elliot, A. J., & Devine, P. G. (1994). On the Motivational Nature of Cognitive Dissonance: Dissonance as Psychological Discomfort. Journal of Personality and Social Psychology, 382-394.
European Environment Agency. (2010). The European Environment: State and Outlook 2010. Copenhagen, Denmark: European Environment Agency.
Falk, A., & Heckman, J. (2009). Lab Experiments are a Major Source of Knowledge in the Social Sciences. IZA Discussion Paper, 1-14.
Farrokhi, F., & Mahmoudi-Hamidabad, A. (2012). Rethinking Convenience sampling: Defining Quality Criteria. Theory and Practice in Language Studies, 784-792.
Festinger, L. (1957). A Theory of Cognitive Dissonance. Evanston: IL: Row, Peterson.
Francis, J. E., & al., e. (2004). Constructing Questionnaires Based on the Theory of Planned Behaviour: A Manual for Health Service Researchers. Retrieved from http://openaccess.city.ac.uk/1735/1/TPB%20Manual%20FINAL%20May2004.pdf Fresco, O. (2009). Challenges for Food System Adaption Today and Tomorrow. Environmental
Science & Policy, 378-385.
Göckeritz, S., Schultz, W., Rendón, T., Cialdini, R., Goldstein, N., & Griskevicius, V. (2010). Descriptive Normative Beliefs and Conservation Behavior: The Moderating Roles of Personal Involvement and Injunctive Normative Beliefs. European Journal of Social Psychology, 514-523.
Godfray, H., Aveyard, P., Garnett, T., Hall, J., Key, T., Lorimer, J., . . . Jebb, S. (2018). Meat Consumption, Health, and the Environment. Science, 243-253.
40 Graca, J., Godinho, C. A., & Truninger, M. (2019). Reducing Meat Consumption and Following Plant-based Diets: Current Evidence and Future Directions to Inform Integrated Transitions. Trends in Food Science & Technology, 380-390.
Hansen, J. (2005). “Is There Still Time to Avoid ‘Dangerous Anthropogenic Interference’ With Global Climate?". American Geophysical Union, Fall meeting 2005.
Harland, P. (2007). Situational and Personality Factors as Direct or Personal Norm Mediated Predictors of Pro-Environmental Behavior: Questions Derived from Norm-Activation Theory. Basic and Applied Social Psychology, 323-334.
Harmon-Jones, E. (2000). Cognitive Dissonance and Experienced Negative Affect: Evidence that Dissonance Increases Experienced Negative Affect Even in the Absence of Aversive Consequences. Personality and Social Psychology, 1490-1502.
Higgs, S., Liu, J., Collins, E., & Thomas, J. (2019). Using Social Norms to Encourage Healthier Eating. Nutrition Bulletin, 43-52.
Johns Hopkins Center for a Liveable Future. (2019). Is The Movement To Eat Less Meat Actually Making A Difference? Retrieved from https://www.huffpost.com/entry/eat-
less-meat-environmental-effect_l_5d39d84fe4b020cd99501f2d?guccounter=1&guce_referrer=aHR0cHM6Ly93
d3cuZ29vZ2xlLmNvbS8&guce_referrer_sig=AQAAAJmY--
YClj7_VxWBDWIzBD7nazy9tZca0S-E95PP4C1q6WdXc9oyTjtRRZ2rfmfvpIOhpoTUm2yfCeWxq
Keller, G. (2011). Managerial Statistics, international edition. Cengage Learning.
Larsen, S. (2016). The Relative Importance of Social and Personal Norms in Explaining Intentions to Choose Eco-Friendly Travel options. International Journal of Tourism Research, 159-166.
Latvala, T., Niva, M., Mäkelä, J., Pouta, E., Heikkilä, J., Kotro, J., & Forsman-Hugg, S. (2012). Diversifying Meat Consumption Patterns: Consumers' Self-reported Past Behaviour and Intentions for Change. Meat science, 71-77.
