Do emotional threat messages improve our health? : the effect of different emotional threat messages on the binge drinking intentions of people from 18 until 28 years old, moderated by self-efficacy increasing health me

(1)

Do emotional threat messages

improve our health?

The effect of different emotional threat messages on the binge drinking

intentions of people from 18 until 28 years old, moderated by self-efficacy

increasing health message elements

.

Timo Schmal 10244034 Master Thesis

Graduate School of Communication

Master programme Communication Science Thesis supervisor S.C.M. Welten

(2)

2

Abstract

Binge drinking is a serious problem amongst young adults that must be dealt with effectively and requires proper communication. Fear ads are often used in alcohol campaigns to influence the binge drinking behaviour. It is however questionable if these fear appeals result in the intended behaviour. Overall, the empirical findings about the effectiveness of fear appeals are inconsistent and contradictory. This study hypothesized that shame and guilt ads work more effectively in influencing the attitude towards binge drinking and binge drinking intensions than fear ads, moderated by self-efficacy. In addition, it was expected that guilt ads would work more effectively than shame and fear ads. 270 respondents (Mage= 22) participated in a

study with a 2 (self-efficacy element: present vs. absent) x 3 (message: fear vs. shame vs. guilt) design. Results suggest that it does not matter which emotion (shame, fear and guilt) is

induced. Consequently, all emotions do not differ in the effect on binge drinking or the attitude towards binge drinking. In addition, the emotional threat appeals do not impact the intention to binge drink when self-efficacy is added to a threat message that consists of emotional content. The additional analysis shows that shame and guilt ads induce more susceptibility than the fear ad.

(3)

3

Preface

I would not have been able to do this thesis without the support, inspiration and kindness of many individuals. A special thanks to my thesis supervisor. Stephanie, your enthusiasm, guidance and support meant a lot to me. I admire your passion in this field! A special thanks to my three close friends: Jeroen, Nynke and Nina who supported me during this challenging time. My family for always believing in me and my boyfriend, who helped me collecting my data.

(4)

4

Introduction

“My name is Kathleen Quartaro. My daughter, Ali, died at the age of 18 after a night of binge drinking. Ali was a freshman at the University of Wisconsin-Milwaukee, in the Honors Program at the School of Business on a scholarship. On the morning of February 21, she was found dead on the upper floor of an off campus apartment-a known party house. She had been drinking shots of hard liquor, passed out and lay unconscious for hours while others partied around her, afraid to get her help. She was a beautiful, physically fit girl. But her body could not handle what the students were doing that night.” (Choose Responsibility, 2014).

The number of young people ending up in hospitals because of alcohol abuse is increasing disturbingly. With 2300 cases in 2010, the number of hospitalised youngsters increased to 3100 cases in 2011, according to hospital records. Furthermore, results from "Statistics Netherlands" in 2009 showed that heavy drinking substantially occurs (17.6%) in the age group of 15 to 28 years (ref). Nine out of ten young adults between 18 and 30 years old say that they drink on a regular basis with an average of 7.7 glasses per occasion (CBS, 2004), even though health organisations recommend a maximum of five glasses. These health organisations consider drinking more than five glasses of alcohol in a short amount of time as "binge drinking" (Medisch Centrum Haaglanden, 2010). Binge drinking mostly occurs at a relatively young age, where such behaviour could seriously affect the future lives of the young adults involved. Risks include, amongst others: alcohol intoxication, risky sexual behaviour, driving under influence of alcohol, and increased levels of violence and victomization. Furthermore, long term effects may include a variety of illnesses, such as increased liver problems and different forms of cancer, such as oesophageal cancer (Wereld Kanker Onderzoek Fonds, 2014). Binge drinking is a serious problem amongst young adults that must be dealt with effectively and requires proper communication.

(6)

6 A common technique in health communication is the use of fear appeals in the message in order to bring attention to social problems (Peters, Ruiters, & Kok, 2012; Witte, 1992; Witte, 1994). "Fear appeals are persuasive messages designed to scare people by describing

terrible things that will happen to them if they do not do what the message recommends.”

(Witte, 1992, p 329). It is however questionable if these fear appeals result in the intended behaviour. Overall, the empirical findings about the effectiveness of fear appeals are inconsistent and contradictory (Witte, 1992; Peters, Ruiters, & Kok, 2012). Messages that contain fear appeals could also trigger too much anxiety or fear, and may therefore not lead to desired responses (Peters, Ruiters & Kok, 2012). Therefore, using fear appeals as a basic emotion may seem peculiar, as fear usually triggers a survival instinct (e.g. running away from a threat) (Tracy & Robins, 2004). In the case of health communication, applying fear appeals in messages would cause people to rather ignore a message than to keep ones attention.

Therefore, this study will focus on emotions that draw someone's attention. Two of these so-called self-conscious emotions are shame and guilt. It is to be expected that these emotions are more effective in changing the behaviour than fear emotions, due to the fact that self-conscious messages triggers a person's self-reflection and trigger the individual’s involvement with the message (Tangney, Steuwig, & Mashek, 2006; Agrawal & Duhachek, 2010).

Consequently, when viewing a self-conscious message, it is expected that the viewer will think more about the message, which will increase the chance of a change in the person's behaviour.

In the scope of binge drinking, gaining more insight into the effects of these different

emotions in health messages could be important in order to reduce this drinking behaviour in the age group of 15 to 28 years old. To find out which emotions work more effectively in

(7)

7 alcohol campaigns, a person's intension to participate in binge drinking will be assessed by comparing the effects of self-conscious emotions with fear appeals.

Another important factor to take also into consideration is the influence of adding a self-efficacy element in the message. Self-self-efficacy refers to an “individual’s belief in his or her

ability to perform the recommended response” (Witte, 2000). Self-efficacy strongly

influences whether the behaviour will be performed when using a fear appeal (Witte, 1992; Witte, 1994; Peters, Ruiters, & Kok, 2012). As a result, it is expected that the self-efficacy element is also more effective for other emotional appeals in changing the behavioural intention. Therefore, this study will investigate if the use of self-efficacy in combination with different emotional messages has different effects on the intentions to binge drink. This leads to the following research question:

To what extent do different types of threat emotional appeals (fear, shame and guilt) affect the binge drinking intentions of people from 18 to 28 years old? And how is this effect moderated by self-efficacy elements in the health message?

Figure 1: Conceptual model

Threat emotional appeals: - Fear appeals

- Shame appeals - Guilt appeals

Intentions to binge drink less

Self-efficacy: yes or no

(8)

8

Theoretical framework

Nowadays it is hard to keep the attention of the audience due to the clutter of advertisements. One of the most used persuasion theories in health messages is the fear appeal (Maddux, Rogers, 1983; De Hoog et al., 2007) Yet, fear studies suggest that the arousal of the fear emotion is needed to trigger attention and to motivate young adults in this case to undertake action regarding binge drinking. However, a lot of conflicting viewpoints are given (Witte, 2000).

Witte and Allen (2000) and Peters, Ruiter and Kok (2012) discussed in their reviews the effectiveness of fear appeals in different studies. Studies in the past were especially focused on the severity of the threat in health messages and how that effected the attitude towards the behaviour. Later in the seventies this way of thinking changed and a distinction was made about the way an individual responded to the threat: danger control (how to avoid the risk?) and fear control (how to reduce the feelings of fear?) (Levental, 1970). In case of danger control the individual will perform the health behaviour to decrease the risks and control the danger. Reducing fear is possible by ignoring the health message or by saying that these risks will only happen to someone else. Leventhal (1970) failed to find empirical evidence when danger and fear control processes would be initiated, but the model changed the current thinking about fear appeals. The Protection Motivation Theory (Rogers, 1983) was able to explain when and why fear appeals work, but fail to explain when and how fear appeals fail.

Subsequently, the use of fear appeals in health campaigns are not always effective in changing the attitude and the behavioural intentions. Each of the prior models only explains a part of how fear-appeal works.

(9)

9 However, it remains unclear when and how fear appeals fail (Witte, 2000). Witte (1992) integrated older models into the Extended Parallel Process Model (EPPM).

The Extended Parallel Process Model

The EPPM is a ‘fear appeal’ theory that shows how individuals respond to health risk messages (see figure 2).Witte (2000) distinguishes two evaluation phases after being

confronted with the fear appeal. In the first evaluation phase, or first appraisal, the perceived fear level will be based on the individuals who perceived severity and susceptibility.

Perceived severity is an individual’s belief about the seriousness of the threat (e.g., imagine your binge drinking behaviour would make you choke in your own vomit), while

susceptibility is an individual’s belief about his or her chances to experience the threat (e.g., how likely is it that I will choke in my own vomit due to binge drinking?). When one of these two components is rated as low, the individual will not be motivated to process the health risk message any further and there will be no response to the health message. When the perceived threat is moderate or high, the threat is taken seriously. Accordingly, the individual takes the risks seriously and based on the example believes that he or she could choke in his or her own vomit due to binge drinking and is therefore highly susceptible to the threat. The developed fear will motivate the individual to do something about the threat and reduces the fear. If the threat (severity and susceptibility) is high, a person is motivated to go to the second appraisal (efficacy appraisal), which consists of the response efficacy and the self-efficacy appraisal (Witte, 1994).

How individuals deal with the threat depends on the efficacy appraisal, which is an evaluation of the efficacy of the recommended response and consists of two elements. The element response efficacy is the individual's belief concerning the effectiveness of a

(10)

10 recommended response (to what extent is a certain response effective in reducing the fear?). The other element, self-efficacy, is the individual's belief about their ability to perform the behaviour as recommended by the health message (to what extent am I capable to do

something about this threat?). In case of high self-efficacy or response efficacy, an individual has the feeling he or she has enough efficacy to avoid the risks and this leads to danger

control. In this case the individual will perform the recommended behaviour. Applied to binge drinking thoughts, for self-efficacy this could imply: "When I make appointments on

forehand with friends to drink less than five glasses of alcohol, I will be able to avoid the risk of choking in my own vomit." A response-efficacy could be: "I believe that choking in my own vomit could be avoided by making appointments on forehand with friends to drink less than five glasses of alcohol. When the self-efficacy and response efficacy are high rated, the individual will be motivated to decrease the danger by performing the recommended

behaviour (danger control).

In case of low self-efficacy or response efficacy the risks will be avoided by fear control (how to reduce the feelings of fear?). In this case the individual believes he or she is incapable of performing the recommended response. Fear control also occurs when an individual realizes that he or she cannot prevent a serious threat from occurring (threat is too high), because he or she believes the recommended behaviour is ineffective. In case of fear control an individual decreases fear by maladaptive responses, such as denial and avoidance (defensive motivation). Figure 2 shows how a fear appeal can lead to one of the three responses (no response, danger control and fear control) (Peters, Ruiter and Kok, 2012).

(11)

11 First evaluation phase second evaluation phase Process

Figure 2: The Extended Parallel Process Model (Witte, 1992)

Review of fear studies

Peters, Ruiter and Kok (2012) reviewed previously performed fear studies and found that only six studies showed a significant interaction between threat and efficacy on the behavioural change. In these studies self-efficacy and the perceived threat were manipulated. All the other reviewed studies turned out to have inconsistent and contradicting results. This could be explained by the fact that studies did not manipulate the self-efficacy, flawed methodology, the studies had different outcome measurements than behaviour, and the studies did not take into account the differences in efficacy between different populations.

Threat appraisal - susceptibility - severity High: - susceptibility - severity (= fear) Low: - susceptibility - severity (= disregard) message Efficacy appraisal - self-efficacy - response-efficacy high - self-efficacy - response-efficacy (= danger control) Low: - self-efficacy - response-efficacy (= fear control) Message acceptance Message rejection No reaction

(12)

12 For example, self-efficacy for exercising four times in a week is in general higher in

populations consisting of people who are 20 to 40 years old compared to populations with people in the ages of 60 to 80. Consequently, the studies containing inconsistency or

contradictories were excluded from the review. Thus, very few studies supported the evidence for the effectiveness of threat appeals in health messages and most of these studies state that using fear appeals in health messages is at best ineffective and leads at worst to health

defaulting behaviour, unless the message contains an element that increases the self-efficacy or response efficacy (Peters, Ruiter and Kok, 2012). As a result, fear appeals have a relatively weak effect on attitude, intensions and behaviour (Witte, 2000).

Fear as emotion

To assure that a health message is effective, using fear as an emotional tool to trigger interest, seems an odd choice. Odd, because fear is a basic emotion, which often triggers an immediate reaction, caused by the biological based nature of these emotions. Applying fear as a tool in health campaigns would mean that, in a situation where an individual is exposed to a message that contains a threat, a first response will be to turn away from that message (Tracy & Robins, 2004).

Therefore the focus within this study will not be on fear appeals, but on self-conscious emotions such as shame and guilt. These self-conscious emotions increase a person's involvement with a message, triggers self-evaluation and might therefore be more effective than fear appeals (Tangney, Steuwig, & Mashek, 2006; Agrawal & Duhachek, 2010). As a result, when being exposed to a self-conscious message it is expected that the viewer will think more about the message and will be more motivated to change his or her behaviour. The self-conscious emotions fit the Elaboration Likelihood Theory (ELM), when self-conscious

(13)

13 emotions evoke more elaboration (Cacioppo & Petty, 1985). According to the ELM

(Cacioppo & Petty, 1985) more elaboration leads to more persuasion and in this case to behavioural change. The effectiveness of conscious emotions is confirmed by Agrawal & Duhacheck (2010). Agrawal & Duhacheck (2010) mentioned that self-conscious emotions may be particularly persuasive tools for reducing risky behaviour, such as binge and underage drinking, due to the strong personal implications of these emotions. However, little research has been done in this field (Agrawal & Duhacheck, 2010).

Self-conscious emotions

What makes emotions conscious is that they require awareness and involve self-evaluation. Self-conscious emotions are seen as more complex than basic emotions when looking at the required cognitive ability needed to experience them (Tangney, 1999). Self-evaluation can only be experienced when the self-awareness is more developed. For example, infants do not have the capacity to experience shame or guilt, because they are not aware of themselves or able to reflect on themselves (Tangney, Steuwig and Mashek, 2006; Agrawal, Duhachek, 2010). Since fear is a basic emotion, very few cognitive capacities are needed to experience the emotion (Tracy & Robins, 2004). Cognitive processing is needed for a person to enable the evaluation of one’s self in a negative way (Tangney, Steuwig and Mashek, 2006; Agrawal, Duhachek, 2010). Subsequently, it is expected that self-conscious emotions lead to more elaboration of the message. In sum, it can be expected that using self-conscious

emotions as tools in health messages yield better results than when using basic emotions. However, there are also differences between these self-conscious emotions, which could have different effects on the attitude towards binge drinking. These differences should be taken into consideration.

(14)

14

Shame and guilt

In order to specify the different effects of self-conscious emotional appeals on binge drinking behaviour, shame and guilt will be analysed separately. Both emotions share several

characteristics (Tangney, 1999): (1) shame and guilt are both emotions that give people the feeling they can control the socially undesirable outcomes (guilt much more than shame). (2) people are motivated to eliminate the negative state the emotions triggers and (3) the negative feelings are often linked with aspects of one’s self (Tangney, 1995).

Shame and guilt are difficult to distinguish from each other. However, studies have found some distinctions between these two emotions. Shame often involves negative feelings about the stable, global self, while guilt involves feelings about a specific behaviour or action taken by someone (Tangney & Dearing, 2002; Lewis, 1971). In case of guilt a person might feel bad about a thing he or she has done and in case of shame a person evaluates one’s self negatively (Tangney, 1999). For instance, in case of shame, when someone made a fool of himself while being drunk, the self-evaluation might change negatively the next day (e.g, “I feel so stupid”). However, when experiencing guilt an individual might have said inappropriate things when being drunk and might feel bad about the things he or she said. Additionally, when

experiencing shame someone feels humiliated, because they view themselves through the eyes of another and the individual realises that the behaviour causing the emotion does not fit with his or her ideal self. In contrast, guilt makes someone take into account what his or her action means to others (Baumeister, Stilwell, and Heatherton, 1994).

Duhachek, Agrawal and Han (2012) studied the effects of shame and guilt on binge drinking. They found that guilt is associated with a higher level of self-efficacy, because guilt increases the desire to compensate the bad or harmful behaviour. However, they found that shame is

(15)

15 associated with a low level of self-efficacy, because a person cannot control the situation in most cases. Consequently, a defensive reaction helps in maintaining a positive self-image. This is supported by Tracy and Robins (2004), who state that shame often promotes avoidance behaviour, while guilt often promotes apology behaviour. This also contradicts with fear which lacks self-efficacy to change the attitude or the behavioural intensions. Therefore, it is expected that guilt (high self-efficacy) works more effectively to decrease the binge drinking intensions than shame (low self-efficacy) and fear (no self-efficacy) do. This assumption leads to the following hypotheses:

H1a: Guilt works more effectively in changing the attitude towards binge drinking than

shame and fear.

H1b:Guilt works more effectively in decreasing the binge drinking intensions than shame and

fear.

Self-efficacy

Mentioned before is that self-efficacy is extremely important for the effectiveness of threat appeals (fear, shame and guilt) on the binge drinking intensions. Therefore, the self-efficacy will be manipulated within this study, because fear needs self-efficacy to be effective. In case of shame the evaluation of one’s self is negative. Normally, shame leads to defensive

responses because someone feels they do not control the situation. Thus, by increasing the self-efficacy an individual might get the feeling that he or she is able to control the situation with a result that the person will be motivated to repair his or her self-image. Guilt is already associated with efficacy, thus the assumption is made that guilt does not need the self-efficacy to be effective. These assumptions lead to the following hypotheses:

(16)

16 H2a: The effect of emotional threat appeals on the attitude to binge drink less is moderated by

self-efficacy such that adding self-efficacy leads to a positive attitude to binge drink less for shame and fear, but not for guilt.

H2b: The effect of emotional threat appeals on the intensions to binge drink less is moderated

by self-efficacy such that adding self-efficacy leads to stronger intensions to binge drink less for shame and fear, but not for guilt.

(17)

17

Method

Design

An experimental survey is conducted to investigate the influence of shame, guilt and fear appeals on the intention to binge drink. The hypotheses were tested using a 2 (adding self-efficacy element vs. not adding a self-self-efficacy element) x 3 (message: fear vs. shame vs. guilt) design.

Pretest

Before performing the main questionnaire, a pre-test was developed in order to check whether participants experience the targeted emotions (fear, shame and guilt). Ninety participants from the University of Amsterdam and the Hogeschool van Amsterdam were randomly assigned to one of the three conditions (fear, shame or guilt) and filled in the questionnaire(Mage= 21.XX,

SD = 2.10; 61.10% females, 38.90 % males. To avoid biased results this sample was not used

in the actual study. Before participating with this study participants were being told that the questionnaire was about advertisements.

Three sets of advertisements were designed with different themes to find out which set is most effective in inducing the three emotions (fear, shame and guilt) as intended. For each of the three emotions different ads were designed within one theme to ensure that the themes were kept constant and that the emotions were a response to the manipulations (Appendix 1).The chosen themes within the sets were: Vomiting, Drunk Driving and Blurred lines.

Consequently, each respondent was exposed to three fear, shame and guilt ads of each theme to measure which theme evoked all three emotions the best. There was a particular interest to choose the themes that could be applied for shame, guilt and fear.

(18)

18 For example, an ad about drunk driving where a person could have killed a family due to his or her drunk driving behaviour (guilt), could have killed him or herself (fear) or everyone knows that he or she has been driving drunk (shame).

In the theme Vomiting the ads were designed as follows: for the guilt condition, the first ad consisted of a girl who was vomiting in the toilet and got help from a friend. The ad text was: “Your stupid drinking behaviour ruined the night of your friends as well. In the fear ad, a boy was lying on street in a strange position in his own puke.” The ad text was: “Imagine your drinking would made you choke in your own vomit”. In the shame condition the picture of the ad showed a girl who puked in the toilet and friends were pointing at her, with the text:

“Imagine that someone would see you like that!”

For the Blurred lines theme, an ad was designed for the fear condition with the text: “You do not remember what happened last night”. The picture showed a girl who was sleeping on a bed with money in her pants. For the shame condition the following text was used: “Everyone knows that you have slept with that idiot!” The picture consisted of a girl who sits ashamed on the ground. For the guilt condition, the picture consisted of a situation where you see a boy regretting his cheating on a girl with the following text: “You have hurt your partner by sleeping with someone else”.

The formats of all these ads were the same, but the texts and pictures were different from each other, except for the slogan: “Alcohol! Know your limits!”. Taking into consideration when designing the ads was that the picture and the text in the ad needed to connect with the emotions.

The participants were asked to what extent they felt several emotions (shame, guilt, fear, disgust, anger and sadness) on a seven-point Likert scale ranging from 1 (not at all) to 7 (very much) after being exposed to an ad (Rottenberg, Ray, & Gross, 2007). As control variable

(19)

19 disgust, anger, and sadness were also measured after the exposure to one of the three sets to make sure that the control emotion is constant in all three advertisements. The participant also needed to select the emotion he or she felt the most when viewing the ad. The severity and susceptibility towards binge drinking were also measured and how credibility and

professional the ads were on a seven-point Likert scale ranging from 1 (not at all) to 7 (very much).

The experienced severity towards binge drinking is measured with one question, such as: “The ad makes me realise how serious the effects of alcohol might be” (Witte, 1992). The experienced susceptibility to the threat is measured with one question, such as: “This could

happen to me as well”(Witte, 1992). Furthermore, the credibility was measured with the

question: “How credible would you rate this ad?” and finally, the professionalism of the ad was measured with the question: “How professional would you rate this ad?”

Finally the participant needed to list any factors or circumstances that would make it easy or enable to drink less than five standard glasses of alcohol and any factors or circumstances that would make it difficult or prevent the participant from drinking less than five standard glasses of alcohol (Ajzen, 2014). This to measure which factors might be most effective to use in the self-efficacy message.

An ANOVA was conducted with the emotion manipulation as independent variable and the evoked emotion as dependent variable. By means of a Tukey post-hoc test, the mean scores were compared. Table 1 shows the main results for set 1: vomiting. The shame ad induced significantly more shame than the fear and the guilt ad. The guilt and fear ad did not differ significantly from shame ratings. The guilt ad induced significantly more guilt than the fear and shame ad.

(20)

20 The fear and shame ad did not differ significantly on guilt ratings. Moreover the fear ad induced significantly more fear than the guilt and shame ad. The guilt and shame ad did not differ significantly on fear ratings. Concerning the control emotions (disgust, anger and sadness), only the guilt ad induced less disgust than the fear and shame ads.

Table 1. Set 1“vomiting ads”: Mean scores for the different evoked emotions per set and condition.

Condition Evoked

emotion Fear Guilt Shame F (2, 87) p M (SD) M (SD) M (SD) Shame 3.00(2.05)a 3.30(1.89)a 4.67(1.65)b 6.799 .002 Guilt 2.63(1.97)a 3.83(2.02)b 2.57(1.72)a 4.190 .018 Fear 3.30(1.56) a 1.57(0.68)b 2.23(1.46) b 13.757 <.001 Disgust 4.37(1.87) a 2.77 (1.68)b 4.57(1.54) a 10.093 <.001 Anger 2.53(1.50) a 1.93(1.23) a 2.37 (1.77) a 1.250 .292 Sadness 2.90(1.78) a 1.93(1.08)b 2.10(1.49)ab 3.644 .030 _______________________________________________________________

Note: Different subscripts letters mean that conditions differ significantly on the depended variable with p < .05.

The difference of the induced severity and susceptibility per emotional threat appeal ad are also measured (see table 2). The fear ad in set 1 induces significantly more severity than the guilt and shame ad. The shame and guilt ad did not differ significantly on the severity ratings. The guilt and shame ad induced both significantly more susceptibility than the fear ad. The shame and guilt ad did not differ significantly on the susceptibility ratings.

(21)

21 Table 2 shows also the credibility and the professionalism of the ads. The shame ad

significantly rated more credible than the fear and guilt ad. The fear and guilt did not differ significantly on the credibility ratings. The fear ad and the shame ad significantly rated more professional than the guilt ad. The fear and shame did not differ significantly on the

professional ratings.

Table 2. Set 1“vomiting ads”: Mean scores for evoked severity, susceptibility to the threat and the credibility and professionalism of the ad per set and condition.

Evoked emotion Fear Guilt Shame F (2, 87) p M (SD) M (SD) M (SD) Severity 4.27(1.51)a 2.70(1.62)b 2.97(1.52)b 8.768 <.001 Susceptibility 2.30(1.37) a 3.99(1.95)b 3.63(1.86) b 7.614 <.001 Credibility 4.00(1.26)a 3.53(1.50)a 5.10(1.39)b 9.636 <.001 Professionalism 3.53(1.33)a 2.60(1.10)b 3.93(1.17)a 9.656 <.001 _____________________________________________________________________

Table 3 shows the result for the ads of set 2 Drunk Driving There are no differences between the shame, fear and guilt ad in evoking shame. Subsequently, this means that the shame ad does not induce more shame than the fear and the guilt ad. There were no significant

differences between the shame and guilt ad in evoking guilt. The guilt and shame ads induced both more guilt than the fear ad. There were no significant differences between the shame, fear and guilt ad in evoking fear. This means that the fear ad does not induce more fear than the shame and guilt ad. Concerning the control emotions (disgust, anger and sadness), only the guilt ad induced more sadness than the shame ad.

(22)

22

Table 3. Set 2“drunk driving” ads: Mean scores for evoked emotions per set and condition.

Evoked

emotion Fear Guilt Shame F (2, 87) p M (SD) M (SD) M (SD) Shame 2.77(1.55)a 4.63(2.51)a 3.93(2.21)a 2.701 .730 Guilt 2.63(1.52)a 4.723(2.12)b 4.57(2.06)b 15.164 <.001 Fear 4.53(1.66) a 4.27(2.43) a 3.77(2.39) a 0.949 .391 Disgust 4.13(2.06) a 4.33(2.52) a 3.90(1.99) a 0.297 .744 Anger 3.50(2.10) a 4.83(2.39) a 4.50(2.11) a 2.970 .057 Sadness 4.07(1.91)ab 4.70(2.44)a 3.37(1.97)b 2.923 .059

The difference of the induced severity and susceptibility are also measured (see table 4). The fear, guilt and shame ads in set 2 did not differ significantly from the severity ratings. The shame ad induced significantly more susceptibility than the guilt ad. The guilt and fear as the fear and shame did not differ significantly from susceptibility ratings. Moreover the fear, guilt and shame did not differ significantly on the credibility ratings and on professional ratings.

(23)

23

Table 4. Set 2“drunk driving ads”: Mean scores for evoked severity, susceptibility to the threat and the credibility and professionalism of the ad per set and condition.

Evoked

emotion Fear Guilt Shame F (2, 87) p M (SD) M (SD) M (SD) Severity 4.17(1.99)a 5.10(1.87)a 4.83(1.60)a 1.962 .147 Susceptibility 1.93(1.311)ab 1.33(.96)a 2.27(1.34) b 4.547 <.001 Credibility 4.20(1.63) a 4.80(1.81)a 5.00(1.29)a 2.066 .134 Professionalism 4.23(1.46)a 3.73(1.68) a 4.43(1.33)a 1.744 .181 _________________________________________________________________

Table 5 shows the result for the ads of set 3 blurred lines. There were no significant

differences between the shame, fear and guilt ad in evoking shame. This means that the shame ad does not induce more shame than the fear and the guilt ad. The guilt ad induced

significantly more guilt than the fear and shame ad.The fear and shame ad did not differ significantly from guilt ratings. There were no significant differences between the fear, shame and guilt ad in evoking fear. Consequently, the fear ad does not induce more fear than the shame and guilt ad.

Concerning the control emotions (disgust, anger and sadness), only the guilt ad induced more angry feelings than the shame ad. As a result, the shame ad needs to evoke less angry feelings to make sure the control emotion is equally induced within this set. The fear and shame ads did not differ significantly on anger ratings. Also the fear and guilt ad did not differ

significantly from anger ratings. The guilt ad induced more sadness than the shame ad. The fear and shame ad as the fear and guilt ad did not differ significantly on sadness ratings.

(24)

24

Table 5. Set 3“blurred lines ads”: Mean scores for evoked emotions per set and condition.

Evoked

emotion Fear Guilt Shame F (2, 87) p M (SD) M (SD) M (SD) Shame 3.83(2.20) a 4.37(2.33) a 4.63(1.65) a 1.151 .321 Guilt 2.73(1.72) a 5.27(1.82)b 2.87(1.70) a 20.004 <.001 Fear 2.73(2.00) a 3.27(1.93) a 2.40(1.52) a 1.714 .186 Disgust 3.10(2.28) a 4.00(2.35) a 3.33(1.95) a 1.351 .264 Anger 3.00(2.21)ab 3.63(2.27)b 2.30(1.69) a 3.110 .050 Sadness 2.63(2.03)ab 3.87(2.49)a 2.33(1.63)b 4.595 .013

Note: Different subscripts letters mean that conditions differ significantly on the depended variable with p < .05.

The fear, guilt and shame ads in set 3 did not differ significantly from the severity, susceptibility, credibility and professional ratings (see table 6).

(25)

25

Table 6. Set 3“blurred Lines”: Mean scores for evoked severity, susceptibility to the threat and the credibility and professionalism of the ad per set and condition.

Evoked

emotion Fear Guilt Shame F (2,87) p M (SD) M (SD) M (SD) Severity 3.47(1.57)a 3.53(1.10)a 2.67(1.47)a 2.432 .094 Susceptibility 2.80(1.71)a 2.30(1.80) a 3.00(1.80) a 1.242 .294 Credibility 4.57(1.57)a 4.067(1.46)a 3.70(1.62)a 2.357 .101 Professionalism 3.53(1.66)a 3.13(1.41)a 2.97(1.27) a 1.204 .305 _________________________________________________________

In the second question participants needed to point out which emotional appeal, they were experiencing when seeing the ad. For set 1 the fear and shame ad were inducing the intended emotion (see table 4). The fear ad in set 1 was evoking the highest fear emotion (66.76%) and the shame ad the highest shame emotion (93.33%). Unlike the guilt ad, which induced the shame emotion more than the guilt emotion (53.33%)(indicated in red) and guilt was induced with 46.47%. Therefore, it can be concluded that the guilt ad should be replaced in an

advertisement, which evokes less shame and more guilt.

Table 7. Set 1“vomiting”: Induced emotion in % per condition.

Induced emotion in % per condition

Fear Guilt Shame

Fear 66.67% 3.33 26.66%

Guilt 0.00% 46.47% 53.33%

(26)

26 Table 8 shows the results for the ads in set 2 Drunk Driving. For set 2 fear and guilt emotions were inducing the intended emotion. The fear ad induced the highest fear (83.33%) and the guilt ad induced the highest guilt (70.00%). Unlike the shame ad, which was inducing the guilt emotion the most (53.33%), fear (26.67%) and guilt the least (20.00%). As a result, the guilt ad induced the guilt emotion the least of the three emotions. The whole advertisement should be replaced by guilt, because it the two other emotions were more induced by this ad.

Table 8. Set 2 “drunk driving ads”: Induced emotion in % per condition.

Fear 83.33% 26.67% 10.00%

Guilt 20.00% 70.00% 10.00%

Shame 26.67% 53.33% 20.00%

Table 9 shows the results for the ads in set 3 Blurred Lines. For set 3 the shame and guilt ads were inducing the intended emotion. The guilt ad induced the highest guilt (76.67%) and the shame ad induced the highest shame (90.00%). Unlike the fear ad, which was inducing the shame emotion the most (76.67%), guilt (13.33%) and fear the least (10.00%). Consequently, the fear ad induced the fear emotion the least of the three emotions. The whole advertisement should be replaced by fear, because it the two other emotions were more induced by this ad.

(27)

27

Table 9. Set 3 “blurred lines ads”: Induced emotion in % per condition.

Fear 10.00% 13.33% 76.67%

Guilt 3.33% 76.67% 20.00%

Shame 6.67% 3.33% 90.00%

In conclusion, the first set should be used for the actual questionnaire. In contrast to set 2 and 3 the guilt emotion did evoke a bit more shame, but did not evoke fear at all. Therefore, the advertisement should be changed partly, while the ads in set 2 and 3 should be replaced completely, because they were both evoking the intended emotion the least of the three emotions. This is in accordance with the results from the first question. Concerning the control emotions (disgust, anger and sadness), only the guilt ad induces less disgust than the fear and shame ads. Consequently, a more disgusting picture is needed to make sure the control emotion is constant in all three advertisements (shame, guilt and fear) within this set. For the main investigation the same picture will be taken with the manipulation of vomit to make disgust constant in all three conditions.

Participants also needed to write down any factors or circumstances that would make it easy or enable them to drink less than 5 glasses of alcohol on an event. These answers will be used for the actual questionnaire to include an effective self-efficacy element in the ad.

Table 10 shows that the most frequent answers are: Friends (make arrangements in advance with friends or others not to drink too much) (55 times), little money or expensive drinks (34

(28)

28 times) and (important) obligations the next day (33 times). For example, when a participant wrote down as answer for the question what would make it easier or what would enable them to binge drink less, “I would drink less if I have and exam the next day or when I have to work next day”, consequently it was coded as obligations the next day. When the participant answered the high prices of the alcoholic drinks or they give as reason they did not have enough money, consequently it was coded as little money or expensive drinks.

Table 10.“all ads”: Rated self-efficacy elements for all conditions.

Self-efficacy elements Percentage (%)

Friends 61.00%

Little money or expensive drinks 37.78% (Important) obligations the next day 36.67%

The element “friends” was rated the highest for factors which makes it easier or difficult to binge drink less. Thus most effective in the self-efficacy messages would be to mention friends in the

recommendation. Besides the fact that the majority of respondents gave “friends” as an answer, it is also the only factor of the three (little money/expensive drinks and obligations the next day), which the participant can manipulate himself. In the beginning of the evening, for example, a person could make an agreement with their friends to not drink more than 5 glasses of alcohol on that occasion.

Obligations the next day, little money or expensive drinks cannot be manipulated as easily by the respondent himself.

In summary, the theme of the ads was chosen for the main study based on results that showed in which theme the intended emotions were triggered the most (shame, guilt and fear). Therefore set two was chosen.

(29)

29 The format and the layout of the ads for the shame and fear emotion are the same as the pre-test to minimize the effect of confounding results, except for the fact that the self-efficacy message was added for three conditions, such as: “Make an agreement with friends to drink no more than 5

(30)

30

Main study

A 2 (self-efficacy element: present vs. absent) x 3 (message: fear vs. shame vs. guilt) design was used to examine the effects of shame, guilt and fear messages on the binge drink intention. The ads in different conditions were manipulated in order to trigger the intended emotions (see appendix 3). 158 participants from the University of Amsterdam and the Hogeschool van Amsterdam filled in the questionnaire and 150 travellers in the train from Amsterdam to Utrecht and back. The respondents were randomly assigned to one of the six conditions to avoid biased results. Thirty respondents filled out the questionnaire inadequately, by forgetting one or two questions and were therefore left out from further analyses. Eight of these thirty respondents were not used for further analyses, because they were too old to participate (above 28 years). In each condition 50 participants were collected. (Mage= 22.XX,

SD = 2.01, 61.10% females and 38.90% males). Most respondents were studying at an

University (65.60%), followed by HBO (29.60%), HAVO/VWO (11.1) and lastly MBO (3.70%). This sample was chosen, because previous studies indicated that this group is highly relevant to test the effect on binge drinking behaviour (Duhachek, Agrawal, & Han, 2012). Prior to participation all participants signed an informed consent and read the factsheet in which they were warned for possible shocking content. Randomly they were assigned to one of the six conditions and they were first exposed to the ad before answering the questions.

(31)

31

The intention to stop binge drinking

The intention to stop binge drinking was measured with four questions, like “I do not intent to drink more than five glasses of alcohol on one occasion in the future.” (Witte, 1994) The items were measured on a seven-point Likert scale from 1 (not at all) to 7 (very much) (Rottenberg, Ray, & Gross, 2007). A principal components factor analysis with Varimax rotation indicated that the four items form one scale. Only one component has an eigen value above 1 (EV=3.53, R2= .84). A reliability analysis showed that the items indeed formed a reliable scale (Cronbach’s alpha = .93, M = 3.35, SD = 1.57).

Attitude towards binge drinking

The attitude towards drinking less than five glasses of alcohol on an occasion is measured with a nine item semantic scale, like “unfavorable/favorable”, “a bad idea/a good idea”, “not fun/fun”, “boring/not boring” (Fishbein & Ajzen, 1974). The items were measured on a seven-point Likert scale from 1 (not at all) to 7 (very much). A principal components factor analysis with Varimax rotation indicated that the nine items form one scale. Only one

component has an eigen value above 1 (EV=3.93, R2= .44). A reliability analysis showed that the items indeed formed a reliable scale (Cronbach’s alpha = .73, M = 4.70, SD = 1.00).

Self-efficacy

The experienced self-efficacy to drink less than five glasses of alcohol on an occasion is measured with 4 questions, such as: “I am capable of drinking less than 5 standard glasses of alcohol an occasion, I am convinced I would be able to drink less than 5 standard glasses of

(32)

32 alcohol an occasion”, “I think I will succeed in drinking less than 5 standard glasses of

alcohol an occasion” and “It is in my own hands if I would be able to drink less than 5 standard glasses of alcohol an occasion” (Witte, 1994). The items were measured on a seven-point Likert scale from 1 (not at all) to 7 (very much). A principal components factor analysis with Varimax rotation indicated that the four items form one scale. Only one component has an eigen value above 1 (EV = 2.72, R2= .68). A reliability analysis showed that the items indeed formed a reliable scale (Cronbach’s alpha = .84, M = 5.83, SD = 1.39).

Response efficacy

The response efficacy to drink less than five glasses of alcohol on an occasion is measured with 2 questions, such as: “Negative situations could be prevented by drinking less than 5 standard glasses of alcohol” and “I am capable in drinking less than 5 standard glasses of alcohol an occasion and prevent negative situations” (Witte, 1994). The items were measured on a seven-point Likert scale from 1 (not at all) to 7 (very much). A principal components factor analysis with Varimax rotation indicated that the two items form one scale. Only one component has an eigen value above 1 (EV = 1.51, R2= .75). A reliability analysis showed that the items indeed formed a reliable scale (Cronbach’s alpha = .68, M = 5.48, SD = 1.45).

The severity, susceptibility, reliability and professionalism of the ad were measured in the same way as in the pre-test, on base of a seven point Likert scale from 1 (not at all) to 7 (very much).

(33)

33

Manipulation checks

In order to test the success of the manipulation, a manipulation check was conducted to find out whether the advertisements indeed triggered the intended emotions of fear, anger, shame and guilt and the control emotions (disgust, anger and sadness).The items were measured on a seven-point Likert scale from 1 (not at all) to 7 (very much).

Table 11: Mean scores for evoked emotions per condition

Evoked

emotion Fear Guilt Shame F (2,267) p M (SD) M (SD) M (SD) Shame 2.21 (1.43)a 2.41(1.65)ab 2.87(1.92)b 3.714 .026 Guilt 2.00(1.25)a 2.27(1.48)a 1.98(1.30)a 1.366 .257 Fear 2.46(1.55) a 1.79(1.10)b 1.97(1.38) b 6.120 .003 Disgust 3.85(1.90)a 3.40 (1.87)a 3.64(1.80)a 1.339 .264 Anger 2.25(1.48)a 1.91(1.40)a 2.07 (1.44)a 1.265 .284 Sadness 2.34(1.51) a 1.59(1.11)b 2.08(1.58)ab 6.550 .002

An ANOVA was performed with the emotion manipulation as independent variable and the evoked emotion as dependent variable. By means of a Tukey post-hoc test, the mean scores are compared. The shame ad induces significantly more shame than the fear. The guilt and fear ad do not differ significantly on shame ratings. The shame and guilt ad do also not differ significantly on shame ratings. Additionally the guilt and shame ad do not differ significantly on guilt ratings.

(34)

34 There are significantly no differences between the fear, shame and guilt ad in evoking guilt. Consequently, this means that the guilt ad does not induce more guilt than the shame and fear ad. Concerning the control emotions (disgust, anger and sadness), only the fear ad induces more sad feelings than the guilt and shame ad.

Furthermore, the additional manipulation check question is performed, in which the

respondents need to point out which emotional appeal they were experiencing when viewing the ad.

Table 12: Induced emotion in % per condition

Induced emotion in % per

condition

Fear 52.17% 16.48% 13.80%

Guilt 10.87% 32.97% 9.20%

Shame 35.87% 50.55% 77.01%

The participants needed to point out which emotional appeal, they were experiencing when while being exposed the ad. In contradiction to the previous emotion check all emotional appeals (fear, guilt and shame) were inducing the intended emotion (see table 12). The fear ad was evoking the highest fear emotion (52.17%), the guilt ad the was evoking the highest guilt emotion (32.97%) and the shame ad the highest shame emotion (77.01%).

(35)

35

Results

Attitude towards binge drinking

The results of the first hypothesis are not in line with the expectation that guilt works more effectively in changing the attitude towards binge drinking than shame and fear. A two-way ANOVA with the condition (shame, fear and guilt) as independent variable, self-efficacy as moderator and attitude towards binge drinking as dependent variable was conducted. No significant results are found. The effect of the emotional conditions (shame fear and guilt) on binge drink intensions (main effect) is not significant F(2,264) = 1.40, p = .249. Therefore hypothesis 1a is rejected.

Furthermore, the results of the second hypothesis are not in line with the expectation that the effect of emotional threat appeals on the attitude to binge drink less is moderated by self-efficacy. Adding self-efficacy leads to a positive attitude to binge drink less for shame and fear, but not for guilt. A two-way ANOVA with the condition (shame, fear and guilt) as independent variable, self-efficacy as moderator and attitude towards binge drinking as dependent variable was conducted. The effect of the emotional conditions (shame fear and guilt) on binge drink intensions (main effect) is not significant F (2,264) = 1.40, p = .249. Moreover the main effect for self-efficacy on binge drink intensions is not significant: F (1, 264) = .06, p = .801. Furthermore the interaction effect between adding self-efficacy to the message and emotional appeal is also not significant: F (2,264) = .052, p = .950. Thus hypothesis 2a is rejected.

(36)

36

Table 13: The means and standard deviations of different emotions with and without a self-efficacy element in the message.

Self-efficacy not present in message Self-efficacy present in message Emotion M SD M SD Fear 4.64 .77 4.55 .93 Shame 4.84 1.18 4.84 1.09 Guilt 4.68 1.06 4.69 .91

Binge drinking intensions

The results of the first hypothesis are not in line with the expectation that guilt works more effectively in changing the binge drink intentions than shame and fear. A two-way ANOVA with the condition as independent variable, self-efficacy as moderator and binge drink intensions as dependent variable is conducted. No significant results are found. The main effect of emotional conditions on binge drink intensions is not significant F(2,264) = 0.76, p = .468. Therefore hypothesis 1b is rejected.

The results of the second hypothesis are not in line with the expectation that the effect of emotional threat appeals on the intensions to binge drink less is moderated by self-efficacy such that adding self-efficacy leads to stronger intensions to binge drink less for shame and fear, but not for guilt. A two-way ANOVA with the condition as independent variable, self-efficacy as moderator and binge drink intensions as dependent variable was conducted. No significant result are found. The main effect of the emotional conditions on the binge drink intentions is not significant F (2,264) = 0.762, p = .468. The main effect for self-efficacy on the binge drink intensions is also not significant: F(1, 264) = .007, p = .932. Furthermore the

(37)

37 interaction effect between adding self-efficacy to the message and emotional appeal is also not significant: F (2,264) = .659, p = .518. Therefore hypothesis 2b is rejected.

Table 14:The means and standard deviations of different emotions with and without a self-efficacy element in the message.

Binge drink intensions Self-efficacy not present in message Self-efficacy present in message Emotion M SD M SD Fear 2.18 1.21 2.26 1.52 Shame 2.65 1.68 2.37 1.09 Guilt 2.22 1.54 2.47 1.74

(38)

38

Additional analyses

Since the EPPM (Witte, 1992) is the foundation of the present study, additional analyses were conducted to find out how the different emotions (shame, guilt and fear) influences the

severity, susceptibility, self-efficacy and response-efficacy.

Susceptibility

A two-way ANOVA is conducted with emotions as independent variable, self-efficacy as moderator and the susceptibility as dependent variable. By means of a Tukey post-hoc test, the mean scores of the different emotions are compared. The main effect of the emotional conditions on susceptibility is significant F (2,264) = 31.38, p < .001. The main effect of self-efficacy on susceptibility is not significant: F(1, 264) = .14 , p = .712. Furthermore the

interaction effect between adding self-efficacy to the message and emotional appeal is also not significant: F (2,264) = .863, p = .423.

Table 15: Experienced susceptibility for shame, guilt and fear, moderated by self-efficacy.

Susceptibility Self-efficacy not present in message Self-efficacy present in message Emotion M SD M SD Fear 1.95 1.06 2.06 1.26 Shame 3.53 2.06 3.67 2.11 Guilt 4.23 1.96 3.75 2.01

Severity

A two-way ANOVA is conducted with emotions as independent variable, self-efficacy as moderator and the severity as dependent variable. By means of a Tukey post-hoc test, the mean scores of the different emotions are compared. The main effect emotional conditions on the severity is not significant F (2,264) = 1.96, p = .143. The main effect for self-efficacy on severity is not significant: F(1, 264) = 2.85 , p = .092. The interaction effect between adding

(39)

39 self-efficacy to the message and emotional appeal is also not significant: F (2,264) = .463, p = .463.

Table 16: experienced susceptibility for shame, guilt and fear, with as moderator self-efficacy.

Severity Self-efficacy not present in message Self-efficacy present in message Emotion M SD M SD Fear 3.10 1.66 3.52 1.68 Shame 3.38 1.63 3.49 2.12 Guilt 2.62 1.78 2.25 1.85

Severity and susceptibility

The shame ad induces significantly more susceptibility than the fear ad and the guilt ad induces also more susceptibility than fear. The guilt and shame ad do not differ significantly on experienced susceptibility ratings. Severity has no significant relationship with fear, shame or guilt. That means that no emotion induces more severity.

Table 17:Mean scores for evoked emotions per condition.

Evoked susceptibility

and severity Fear Guilt Shame F (2,267) p M (SD) M (SD) M (SD) Susceptibility 2.01(.186)a 4.00(.186)b 3.60(1.91)b 31.714 .026 Severity 3.35 (.187) 2.92(.19) 3.44(.193) 2.099 .125

Note: Different subscripts letters mean that conditions differ significantly on the dependent variable with p < .05.

Experienced self-efficacy

A two-way ANOVA is conducted with emotions as independent variable, self-efficacy as moderator and the experienced self-efficacy as dependent variable. By means of a Tukey

(40)

40 post-hoc test, the mean scores of the emotions are compared. The main effect of the emotional conditions on the experienced self-efficacy is not significant F (2,264) = 1.25, p = .288. Moreover the main effect of adding self-efficacy in the ad on the experienced self-efficacy is not significant: F(1, 264) = 2.47 , p = .619. Furthermore the interaction effect between adding self-efficacy to the message and emotional appeal is also not significant: F (2,264) = 1.72, p = .182.

Table18: Experienced self-efficacy for shame, guilt and fear, with as moderator self-efficacy.

Self-efficacy Self-efficacy not present in message Self-efficacy present in message Emotion M SD M SD Fear 5.70 1.24 5.81 1.33 Shame 6.18 1.14 5.87 1.59 Guilt 5.49 1.43 5.95 1.54

Experienced response-efficacy

A two-way ANOVA is conducted with emotions as independent variable, self-efficacy as moderator and the experienced response-efficacy as dependent variable. By means of a Tukey post-hoc test, the mean scores of the emotions are compared. The main effect of emotional conditions on the experienced response-efficacy is not significant F (2,264) = 1.25, p = .288. The main effect of self-efficacy on response-efficacy the is not significant: F (1, 264) = 2.47 ,

p = .619. The interaction effect between adding self-efficacy to the message and emotional

(41)

41

Table 19: Experienced response-efficacy for shame, guilt and fear, with as moderator self-efficacy

Response-efficacy Self-efficacy not present in message Self-efficacy present in message Emotion M SD M SD Fear 5.32 1.33 5.73 1.44 Shame 5.58 1.35 5.27 1.52 Guilt 5.26 1.43 5.73 1.57

Credibility of the ad

A two-way ANOVA is conducted with emotions as independent variable, self-efficacy as moderator and the credibility of the ad as dependent variable. By means of a Tukey post-hoc test, the mean scores of the emotions are compared. The main effect of the emotional

conditions on the credibility of the ad is significant F (2,264) = 9.83, p < .001. The main effect of the self-efficacy on the credibility of the ad is not significant: F (1, 264) = .006 , p = .939. The interaction effect between adding self-efficacy to the message and emotional appeal is also not significant: F (2,264) = 1.37, p = .256.

Table 20: Experienced credibility for shame, guilt and fear, with as moderator self-efficacy.

Credibility Self-efficacy not present in message Self-efficacy present in message Emotion M SD M SD Fear 5.32 1.53 3.96 1.86 Shame 5.00 1.60 4.60 1.73 Guilt 4.57 1.68 4.59 1.72

Professionalism of the ad

A two-way ANOVA is conducted with emotions as independent variable, self-efficacy as moderator and the professionalism of the ad as dependent variable. By means of a Tukey post-hoc test, the mean scores of the emotions are compared. The main effect of the emotional

(42)

42 conditions on the professionalism of the ad is significant F (2,264) = 3.21, p < .042. The main effect of self-efficacy on the professionalism of the ad is not significant: F(1, 264) = .2.96 , p = .087. In addition the interaction effect between adding self-efficacy to the message and emotional appeal is also not significant: F (2,264) = 1.09, p = .339.

Table 21: Experienced professionalism of the ad for shame, guilt and fear, with as moderator self-efficacy.

Professionalism Self-efficacy not present in message Self-efficacy present in message Emotion M SD M SD Fear 3.07 1.50 3.23 1.56 Shame 3.70 1.86 3.81 1.60 Guilt 2.98 1.44 3.73 1.76

(43)

43

Professionalism and credibility of the ad

An ANOVA with the emotion manipulation as independent variable and credibility as

dependent variable shows the mean emotion scores for the different conditions. By means of a Tukey post-hoc test, the mean scores are compared. The shame ad is significantly more credibility than the fear ad. Moreover there are no significant differences between the shame and guilt ad in credibility. Additionaly, the guilt ad is significantly more credible than the fear ad.

An ANOVA with the emotion manipulation as independent variable and professionalism as dependent variable shows the mean emotion scores for each condition. By means of a Tukey post-hoc test, the mean scores are compared. The shame ad is significantly more professional rated than the fear ad. There are no significant differences between the shame and guilt ad in rating the professionalism of the ad.

Table 22: Mean scores for evoked emotions per condition

Evoked

credibility and professionalism

F (2,267) p M (SD) M (SD) M (SD) Credibility 3.75(1.71)a 4.78(1.67)b 4.58(1.69)b 9.83 <.001 Professionalism 3.15 (1.53) a 3.76(1.72)a 3.34(1.63) a 3.21 .040

(44)

44

Binge drink intensions

In addition a regression analysis is conducted to test the different effects of the experienced fear, shame and guilt emotions on binge drink intensions, while being exposed to the ad. The ANOVA shows that the regression model is not significant: F(3,266) = 1.63, p =.182.

Subsequently, there are no significant relationships between the binge drink intensions and the experienced emotions shame, fear and guilt when being exposed to the ad.

(45)

45

Conclusion and Discussion

This experimental survey investigated whether the different types of emotional threat messages 'shame', 'fear' and 'guilt' have different effects on the binge drinking intensions of people from 18 till 28 years old, moderated by self-efficacy. The results of this experimental survey are in contradiction with the hypotheses and the findings of previous literature on this subject. In contrast to the expectation that the use of the guilt emotional appeals in a health message would lead to a greater decrease in binge drinking intensions than fear and shame (Tangney, Steuwig and Mashek, 2006; Agrawal, Duhachek, 2010), the current study found that it does not matter which emotion (shame, fear and guilt) is induced. Consequently, all emotions did not differ in the effect on binge drinking. This also applies to the attitude that individuals have towards binge drinking.

An important factor to take into consideration is the influence of adding a self-efficacy element in the message. According to the Parallel Process Model (Witte, 2000), the expectation is that adding a self-efficacy element in a threatening message will lead to a decrease in binge drinking intensions. As a result, it is expected that the self-efficacy element is also more effective for self-conscious emotional appeals in changing the behavioural intention. However, the present research did not find support for this model. The emotional threat appeals did not impact the intention to binge drink when the self-efficacy is added to a threat emotional message. Moreover, the use of different emotions in a threat emotional message with added self-efficacy showed no significant difference in the attitude towards binge drinking.

In line with previous studies, no change of behavioural intentions or attitude towards binge drinking was found for the emotion ‘shame’ when a self-efficacy element was missing in the message. It should be noted that the self-efficacy element was not effective and no effects

(46)

46 could be found. Moreover, the additional analyses showed that the experienced self-efficacy is not effective enough to find any effects for emotional appeals on binge drink intensions. Nevertheless, this contrasts the expectations since this might be expected for a shame appeal in a health message, because shame triggers a defensive reaction to maintain a positive self-image (Duhachek, Agrawal and Han, 2012) or avoidance behaviour (Tracy & Robins,2004).

Peters, Ruiter and Kok (2012) mentioned in their review that most studies using fear appeals in health messages are at best ineffective and lead at worst to health defaulting behaviour, unless the message contains an element that increases the self-efficacy or response

efficacy. The latter is in line with the results of this study, because the manipulation of self-efficacy did unfortunately not show to be of significant influence in this study.

Furthermore, it is difficult to distinguish between the emotions ‘shame’ and ‘guilt’, as has already been shown in previous studies (Tangney & Dearing, 2002). This might explain why no differences were found between the emotions, since shame and guilt did not differ

significantly on shame or guilt ratings.

A possible explanation for the lack of effects could be the result of a self-efficacy manipulation which did not work out. The additional analysis showed that with the fear emotion the susceptibility rating for the ad was relatively low (M=1.95) in comparison to shame (M=3.53) and guilt (M=4.23). A low level of susceptibility might explain why the binge drink intensions did not change for the fear condition. According to the EPPM (Witte, 1994), the susceptibility to the threat should be high to be as effective as the self-efficacy. Both elements were low in this case, thus in line whit the results of previous studies. This resulted in no effects on binge drink intensions.

(47)

47 There are several limitations to this study, which are of interest for further research. First of all, it should be taken into account that the experimental survey only meets the requirements regarding a good internal validity to a limited extend. Since all respondents completed the questionnaire at university, other school or on public transport, it was not possible to let all respondents fill out the questionnaire under the same circumstances. This means that it is possible that some unknown factors have influenced the final results. Thus, in short, the environment was not controlled in this study.

Another limitation of this study is the manipulation of self-efficacy. Bartholomew (2011) noted that to enhance self-efficacy, a simple recommendation (e.g., the recommendation to call a phone number) is not ineffective. The efficacy manipulations should include at least a paragraph of text (Peters, Ruiter and Kok, 2012). Thus the manipulation in this study has been done incorrectly based on the findings of these authors and this might explain the lack of effects, which is a possible explanation for the lack of measured effects.

Additional research is needed focussing on the effects of shame and guilt ads on binge drinking in combination with self-efficacy. This because the additional analysis shows that both shame and guilt ads induced more susceptibility than the fear ad. More susceptibility could lead to more effects on binge drink intensions according to the EPPM (Witte, 1994). Therefore, it is expected that shame and guilt might be indeed more effective than fear ads.

A possible explanation for the lack of effect from the different emotions could be caused by the fact that not all the emotions of the ads induced the intended feeling. Consequently, the emotional appeals in the ad might have not been explicit enough. There were no significant differences between the fear, shame and guilt ad in evoking guilt. Additionally the guilt and shame ad do not differ significantly on guilt ratings. Consequently, this means that the guilt ad does not induce more guilt than the shame and fear ad.

Do emotional threat messages improve our health? : the effect of different emotional threat messages on the binge drinking intentions of people from 18 until 28 years old, moderated by self-efficacy increasing health me