How Slogans Drive Word-of-Mouth:
The Role of Vagueness and Liking
Author: Daphne Kiani Aisa Fiers Student number: 10853189 Date of submission: 28-01-2016 Version of submission: Final Version
MSc. in Business Administration – Marketing Track Thesis Supervisor: Dr. A.C. Krawczyk
i
Statement of Originality
This document is written by Daphne Kiani Aisa Fiers who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the
supervision of completion of the work, not for the contents.
Acknowledgements
This thesis is written as the concluding step in obtaining my master’s degree in Business
Administration at the University of Amsterdam. Writing this thesis has given me the opportunity of applying some of the knowledge that I have obtained over the course of my Master. Obtaining my master’s degree would not have been possible without the guidance and support of several people. Therefore, I would like to show my gratitude to them.
First and foremost I would like to thank my supervisor Dr. A. C. Krawczyk for providing me with her critical guidance and support, as well as valuable and speedy feedback. A special thanks goes out to Marinda van Eersel, Menno Visser and Nora van Bracht, with whom I have shared the struggles of writing a thesis. For they acted as a soundboard as well as a voice of reason when all the faith in my own ability was lost. And as they were willing to spend several hours of their valuable time to act as the independent assessors of the vagueness in 126 slogans. I also want to express my gratitude to my friends and family for their understanding and infinite support. Specifically to Jacques Fiers for proofreading my thesis multiple times and providing me with his refreshing view on the matter. Last but not least I want to thank all the respondents of my pretest and experiment, for without respondents I could not have written my thesis.
Table of Contents
1 Abstract ... 1 1.1 Theoretical Contribution ... 3 1.2 Managerial Contribution ... 3 2 Introduction ... 4 3 Literature Review ... 7 3.1 Branding ... 7 3.2 Word of Mouth ... 8 3.3 Slogans ... 11 3.4 Conceptual Model ... 16 4 Method ... 17 4.1 Measures... 17 4.2 Slogan Selection ... 19 4.3 Pretest ... 21 4.4 Data Collection ... 23 5 Data Analysis ... 26 5.1 Sample ... 27 5.2 Reliability ... 29 5.3 Manipulation Check ... 315.4 Skewness and Kurtosis ... 32
5.5 Descriptive Analysis and Correlation Matrix... 33
5.6 Hypotheses Testing ... 36 6 Discussion... 44 6.1 Results ... 44 6.2 Theoretical Implications ... 48 6.3 Practical Implications ... 50 7 Conclusion ... 52
8 Limitations and Future Research ... 54
8.1 Limitations ... 54
8.2 Future Research ... 56
9 References ... 59
10 Appendices ... 67
10.1 Adapted Measurement Scales ... 67
10.2 List of Selected Slogans ... 70
10.3 Meta Model ... 74
10.4 Pretest ... 75
10.5 Slogan Selection After Pretest ... 89
10.6 Survey... 90
10.7 Invitation Messages ... 102
10.8 Statistical Tables ... 103
10.9 Reliability of Pretest Scales... 107
List of Tables and Figures
Tables Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7 Table 8 Table 9 Table 10 Table 11 Table 12 Table 13 Table 14 Table 15 Table 16 Table 17 Table 18 Table 19Selected slogans for pretest.
Selected slogans’ mean scores of the measured constructs in the pretest.
Demographical overview of the samples for separate conditions and the entire sample.
Reliability scores for construct scales of the experiment. Kolmogorov-Smirnov and Shapiro-Wilk tests for normality. Skewness and Kurtosis.
Means and Standard Deviations per condition.
Means, Standard Deviations, Correlations and Reliability scores. Hierarchical Regression Model of Vagueness on WOM Intentions. Hierarchical Regression Model of Vagueness on Slogan Liking. Hierarchical Regression Model of Slogan Liking on WOM Intentions. Analysis of the mediating effect of Slogan Liking
Complete list of selected slogans and their respective translations. Mean scores of the Pretest on 4-item Vagueness scale, 2-item Liking scale, and 3-item Brand Attitude scale.
Pearson Chi-Square test for differences in gender distribution between groups.
Pearson Chi-Square test for differences in educational distribution between groups.
Levene’s test for the four conditions regarding age distribution. ANOVA and Welch’s test results for four conditions regarding age distribution.
Levene’s test for four conditions regarding Slogan vagueness.
22 23 28 31 33 33 34 35 37 38 40 41 70 - 73 89 103 103 103 104 104
Table 20 Table 21 Table 22 Table 23 Table 24 Table 25 Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6
ANOVA results for four conditions regarding Slogan Vagueness. Tukey’s test results for four conditions regarding Slogan Vagueness. Levene’s test for four conditions regarding Slogan Liking.
ANOVA for four conditions regarding Slogan Liking.
Tukey’s test results for four conditions regarding Slogan Vagueness. Reliability scores for construct scales of the pretest
Conceptual model
The brands used in the four conditions of the experiment.
The different slogans that respondents of the different groups were exposed to.
Graphical depiction of the mediating effect of Slogan Liking. The direct effect of slogan vagueness on WOM Intentions can be found in
parentheses.
Overview of hypotheses results.
Meta model as presented to raters and as used by Strutton & Roswinanto (2014). 104 105 105 105 106 107 16 24 24 43 43 74
1
1 Abstract
The aim of this thesis is to investigate the effect of strategically using vagueness in slogan on Word-of-Mouth (WOM) intentions and considering the role of slogan liking. Where vagueness is considered the opposite of clarity and preciseness.
The research methodology consisted of several steps. First, the stimuli had to be selected from a long list of 126 slogans that were gathered online. The slogans were assessed on vagueness by independent raters. The 23 remaining slogans were used in the pretest, where they were
measured on vagueness, familiarity, likeability and attitudes toward the underlying brand. Then an experiment was conducted online, which was split into four conditions based on (high/low) vagueness and (high/low) liking. Each of the 297 respondents in the experiment rated the stimuli of their condition on vagueness, likeability and their own intentions to engage in WOM.
The findings show that slogan vagueness has a negative effect on slogan liking and WOM intentions. On the other hand slogan liking has a positive effect on WOM intentions and partially mediates the negative relationship between slogan vagueness and WOM intentions.
The thesis adds to existing theory by combining the fields of branding and WOM. It proves that brand elements, like slogans, can have both positive and negative effects on consumers and their responses to brand related messages. Furthermore, it shows that slogan liking partially mediates the negative relationship between slogan vagueness and WOM intentions. The thesis contributes to managerial considerations when designing slogans, as the strategic use of vagueness requires making tradeoffs in the desired outcomes of brand related messages. Moreover, it demonstrates
2 that managers should strive to create likeable slogans to increase people’s intentions to engage in WOM.
Multiple limitations may have impacted the results of the thesis. Consider time restrictions, the restricted generalizability due to limited representativeness of the sample, the limited usability of the results from the independent raters and the fact that the stimuli did not create maximum variation in slogan vagueness and likeability. Moreover, possible industry and brand effects may have been at play. Furthermore, self-enhancement bias is a risk due to the self-reporting method and the ecological validity was limited due to exclusion of visual cues in the stimuli and
measuring WOM intentions instead of behavior. Lastly, the content validity of the vagueness scale is uncertain.
Several avenues for future research were discovered. Future research should replicate the results controlling for the limitations of this study. The effects of vagueness in slogans on consumer responses should be mapped. A valid and reliable vagueness scale should be developed. Future research could explore what mediates the relationship between slogan likeability and WOM. Moreover, it should examine the effects of other brand elements on WOM and study the role of brand equity and brand-concept consistency when creating vague slogans.
This thesis provides new insights by establishing relationships between slogan vagueness, slogan liking and WOM intentions and improves our understanding of tradeoffs to be made in slogan design.
3
1.1 Theoretical Contribution
The main purpose of the thesis is to combine the fields of branding and WOM and provide relevant insights in particular relationships between them. Little research has been conducted on the relationship between branding and WOM, and even fewer on slogans and WOM intentions. Previous research by Strutton & Roswinanto (2014) is key to this study, as the results show how the strategic use of vagueness in brand slogans can lead to positive consumer responses. Limited research has been done to examine the consequences of the strategic use of vagueness in slogans on consumer responses, and the effect on WOM intentions has yet to be investigated.
Nevertheless, in real-life vague slogans are being launched, without a comprehensive
understanding of the outcomes. This thesis adds to our understanding of the consequences of vagueness in slogans. It examines the effect that vagueness has on people’s intentions to engage in WOM, a different consumer response than those that have been researched before and it will consider the role of slogan liking. It will provide insights into our understanding of the tradeoffs in outcomes that are made in slogan design. The thesis will provide statistical proof for the clarity-liking relationship, as identified by previous research. In addition, it will prove the link between the likeability of a slogan and people’s intentions to engage in WOM through two drivers of WOM. Furthermore, it will provide insights into the mediating effect of the likeability of a slogan in the slogan vagueness and WOM intention relationship.
1.2 Managerial Contribution
This thesis provides relevant managerial insights into the relationship between the strategic use of vagueness and consumer responses. It will show the importance of considering tradeoffs in the outcomes of using vagueness in brand related messages, like slogans. Moreover, providing managers with new insights on factors in slogan design that have the ability to drive WOM.
4
2 Introduction
We live in a world where there is no escaping the constant exposure to brands. Companies strive to build positive customer-based brand equity in the minds of consumers to increase their market value (K. L. Keller, 1993). For example, Apple and Google have successfully created brand equity with a value of over 100 billion (Interbrand, 2014). Brand names, logos and slogans are brand elements that help convey a brand’s identity (Kohli, Leuthesser, & Suri, 2007). The word slogan stems from the Scottish Gaelic word sluagh-ghairm, which refers to a war/battle cry and literally translates to the words army and cry (Miriam-Webster.com, 2015). Slogans are often used to convey the identity of brands in advertising, however their message isn’t always clear and precise. Take the slogan of the Dutch hardware store GAMMA: “Dat zeg ik”, which loosely translates to “That’s what I said”. The slogan is definitely not clear, as it does not specify what it is that was said, and it clearly does not say anything about the brand’s identity whatsoever, with GAMMA being a hardware store.
Vagueness in slogans can have positive effects on consumer responses, like evoked thought, brand attitude and persuasiveness (Strutton & Roswinanto, 2014). However, further research is needed as we lack understanding of the consequences of the strategic use of vagueness in
slogans, specifically the consequences on additional consumer responses. An increase in evoked thought could lead to increased accessibility of the slogan and increased brand attitude could lead to increased emotional or functional motivation to talk about the slogan. Since, accessibility and motivation are drivers of WOM (Berger & Schwartz, 2011; Mazzarol, Sweeney, & Soutar, 2007), this thesis sets out to investigate the relationship between slogan vagueness and WOM intentions. The likeability of a stimuli increases people’s processing of that stimuli (Dahlén &
5 Rosengren, 2005). More processing can lead to greater accessibility of a slogan and in turn increased WOM intentions. Moreover, people talk about things they find interesting to seem more interesting to others (Dye, 2000; Hughes, 2005). If a likeable slogan is seen as interesting, this may increase the social motivation to engage in WOM. However, previous research failed to investigate this relationship, which is why this thesis explores the relationship between slogan liking and WOM. Previous research by Dass et al. (2014) explored the antecedents of slogan liking and showed that the clarity of the message has a positive effect on the likeability of a slogan. Vagueness, in this thesis, is seen as the opposite of clarity and preciseness. Therefore, we explore if a relationship exists between the construct of vagueness and liking. Dass et al. (2014) indicated the theoretical and managerial need for improved understanding of potential tradeoffs in slogan design. By exploring slogan features that have the ability to improve the performance of a slogan on one element, while making concessions on the performance of the slogan on another element. All in all, the aim of this research is to investigate the relationship between slogan vagueness and people’s intentions to engage in WOM. Moreover, it looks at what role slogan liking plays in that relationship. The following research question was proposed:
What is the effect of slogan vagueness on intentions to engage in word-of-mouth and does slogan liking mediate this effect?
An experiment was conducted to answer the proposed research question. The experiment yielded 297 valid responses over four different conditions. Every respondent answered the same
questions. However, per condition the respondent was exposed to a different slogan to create maximum variation in vagueness and slogan liking. The four stimuli were selected from an initial list of 126 slogans, using independent raters to assess vagueness and a pretest to test brand
6 attitude, slogan vagueness, liking and familiarity. The results of the experiment contribute to theory by combining the fields of branding and WOM and by providing new insights in both fields. In addition, the thesis contributes to the considerations that managers have to make when designing slogans for a brand, as creating a vague slogan requires making tradeoffs in the desired outcomes of the slogan.
This thesis consists of several sections. The first section of the thesis consists of the introduction, which introduces the reader to the research topic, theoretical and managerial relevance. The second section provides the critical literature review. It gives an overview of the existing theories and relationships between the fields of WOM and branding. In this section the research gap is identified and the hypotheses and conceptual model are described. The third section consists of the research methodology and describes the measures, process of stimuli selection, pretest and experiment. In the fourth section a detailed overview is given of the statistical analyses that were performed and it includes the evaluation of support for the hypotheses. The discussion of the results is included in the fifth section. Here the findings are discussed and the theoretical and practical implications of the thesis to the fields of WOM and branding are explained. The conclusion to this thesis can be found in the subsequent section. The next section discusses the limitations of this research and reveals important avenues for future research. The final two sections contain the overview of the used literature in this thesis and the appendices, which provide additional (graphical) information to the thesis.
7
3 Literature Review
3.1 Branding
Nowadays, people are surrounded by brands 24/7, which leaves companies striving to build positive brand equity to set themselves apart. Brand equity is “the differential effect of brand knowledge on consumer response to marketing efforts” (K. L. Keller, 1993, p. 8). When
consumers react more favorably to a brand than an unknown/nonexistent brand, a brand is said to have positive brand equity (K. L. Keller, 1993). The increasingly competitive environment poses a threat to brands. With consumers being exposed to an abundancy of brand related
communication, creating lasting impressions with consumers is all the more challenging (Rosengren & Dahlén, 2006). In building brand equity and influencing consumer behavior companies can employ brand elements, use their product and marketing communications, and capitalize on indirect associations (Dahlén & Rosengren, 2005; K. L. Keller, 2005; Laran, Dalton, & Andrade, 2011a). Brand elements are the trademarkable features through which people can identify or differentiate brands (K. L. Keller, 1993; Kohli et al., 2007). Among these brand elements are brand names, logos and slogans. However, each brand element has its own advantages. The core of a brand is its brand name, which is the only essential element for establishing a brand (Kohli, Thomas, & Suri, 2013). Logos have the advantage of being understood worldwide, because of their visual representation. A disadvantage of both brand names and logos is their inability to communicate brand meaning. Slogans, on the other hand, have the ability to explain what the brand is about (Kohli et al., 2013). To create a strong brand it is important to combine multiple brand elements and capitalize on potential synergies (K. L. Keller, 2005). A brand’s identity can be seen as both dynamic and consistent over time, as some
8 elements may need to be changed in order to react to the environment while core elements need to remain the same (Da Silveira, Lages, & Simões, 2013).
A stepwise approach can be taken to build a brand. Firstly, one needs to create a brand identity in order to create awareness. Secondly, it is crucial to build a meaningful brand that has strong, favorable and unique associations in the consumers mind. A meaningful brand can lead to
positive brand responses with consumers, and finally creating intense and active consumer-brand relationships. These steps make up the “customer-based brand equity pyramid” (K. L. Keller, 2001). Brand relationships are created in the last and most valuable brand building block; Resonance. Brand resonance consists of attachment and activity (K. L. Keller, 2001). A consequence of strong attachment and activity in consumer-brand relationships can be repeat purchasing, the existence of a brand community and consumers spreading positive WOM about the brand (K. L. Keller, 2001).
Some research has considered WOM as a part of loyalty (e.g. Zeithaml, Berry, & Parasuraman, 1996), however consistent with more recent studies loyalty will be considered a driver of WOM (Roy, Lassar, & Butaney, 2014). Companies build brand equity to differentiate themselves and highly differentiated brands receive more WOM (Lovett, Peres, & Shachar, 2013).
3.2 Word of Mouth
Word-of-mouth (WOM) can be defined as “communication involving the spoken word, between one friend or relative in a face-to-face situation sharing product information with another” (Bickart & Schindler, 2001, p. 37). In our ever growing online world, a new form of WOM has been developed; online WOM, electronic WOM or eWOM. eWOM can be defined as involving
9 “personal experiences and opinions transmitted through the written word” (Sun, Youn, Wu, & Kuntaraporn, 2006, p. 1106).
There are several differences between traditional WOM and eWOM. On the one hand, traditional WOM is often created in a one-to-one setting that is more personal and interactive and allows the incorporation of non-verbal communication. On the other hand, eWOM is created on the Internet on a one-to-many basis and allows for faster dissemination among a larger group of people (Lovett et al., 2013). eWOM is considered more influential than traditional WOM due to its speed, convenience, reach and the lack of personal interaction pressure (Phelps, Lewis, Mobilio, Perry, & Raman, 2004). The use of online forums, for example, can allow multiple people to view electronic consumer generated content at the same time. Electronic WOM has the ability to influence the judgements of consumers (Bambauer-Sachse & Mangold, 2011). Even though, WOM can be either positive, neutral or negative, this thesis will only focus on positive WOM. Despite the differences between online and offline WOM this thesis will consider them as being part of one construct.
In half of all brand related communications between people, someone refers to marketing actions that they have been exposed to (E. Keller, 2007). One stream of research discusses the
importance of accessibility as a driver of WOM. They argue that for people to talk about brands, these brands need to be top of mind. Furthermore, the accessibility of brands can differ and accessibility can be created by external cues (Berger & Schwartz, 2011). The most talked about brands are those which are most visible (Berger & Schwartz, 2011).
10 Another stream of research discusses the importance of motivation. The desire to help the
receiver, activation of associations and advertising can trigger someone’s motivation to engage in WOM (Mazzarol et al., 2007). They argue that people engage in WOM as a way to express themselves to others (social motive), share their emotions (emotional motive) and to spread information (functional motive) (Lovett et al., 2013). Traditional WOM is deemed more
appropriate for sharing emotions, while eWOM is deemed more appropriate for self-expression and information sharing (Lovett et al., 2013). Moreover, when people talk about brands, they share information that says something about both the brand and themselves (Wojnicki & Godes, 2008). Therefore, partaking in WOM communications about interesting brands makes the person talking about them seem interesting (Dye, 2000; Hughes, 2005).
Considerable research has been done in both the field of branding as well as WOM and attempts have been made to combine the two to examine the relationships between them. Research has shown that WOM can be a driver of sales (Chevalier & Mayzlin, 2006; Reichheld, 2003), product judgements (Herr, Kardes, & Kim, 1991), brand image (Graham & Havlena, 2007; Jalilvand & Samiei, 2012), market share (Graham & Havlena, 2007), and purchasing behavior (Jalilvand & Samiei, 2012; E. Keller, 2007). With the growing ability to be online anywhere at any time, electronic WOM can now influence our judgements right up to the final steps in our decision making process. Previous studies have shown that superior product performance, (dis)satisfaction, loyalty, commitment, trust, and advertising can drive WOM communication (Graham & Havlena, 2007; Mazzarol et al., 2007; Roy et al., 2014; Sundaram, Mitra, & Webster, 1998). Of greatest importance to this research is the proven positive relationship between
11 communication channel” (E. Keller, 2007). Advertising messages often contain brand elements, like slogans. As a result these slogans surface in consumers WOM communications. The
underlying reason for the effectiveness of WOM is that people see it as being reliable information (Bambauer-Sachse & Mangold, 2011).
3.3 Slogans
All brand elements have the same goal: to generate positive consumer reactions to the brand (Laran, Dalton, & Andrade, 2011b). Furthermore, brand elements can influence consumer behavior (Laran et al., 2011a). The focus of this thesis will be on one specific brand element, namely slogans. The term slogan is used, whereas other authors may refer to taglines, mottos, catchphrases or brand signatures.
Research has shown that slogans can have a positive effect on product differentiation, brand awareness, brand evaluations and product beliefs (Boush, 1993; Dahlén & Rosengren, 2005; Katz & Rose, 1969; K. L. Keller, 2003). Slogans can act as a “hook” in the mind of the consumer, tying a meaning to the brand (K. L. Keller, 2005), due to their unique ability to
communicate something about the brand (Kohli et al., 2007). Slogans can successfully contribute to building brand equity, as they can be used to create and uphold a brand’s identity over time (Dahlén & Rosengren, 2005; Kohli et al., 2007). The brand’s identity then improves brand equity through greater brand knowledge in the minds of the consumer through awareness (recognition and recall) and image (brand associations) (K. L. Keller, 1993; Kohli et al., 2007; Mathur & Mathur, 1995). Moreover, communicating a brand’s attributes, it’s positioning or changes to it is frequently done by slogans (Boush, 1993; Dass et al., 2014; Kohli et al., 2007; Laran et al., 2011a).
12 A brand’s identity is oftentimes inseparable from a brand’s name and logo, which makes it difficult to change them without losing the identity. Slogans, on the other hand, are seen as the most dynamic of the brand elements. Slogans can be used for short-term and long-term strategies (Kohli et al., 2007). Nevertheless, companies can change their slogan, when there is a need to communicate a different, new or evolved identity, due to changes in the environment without losing their identity (Kohli et al., 2007).
The use of slogans in advertising as brand building tool has become increasingly popular (Lamons, 1997; Mathur & Mathur, 1995). Slogans can motivate people to engage WOM communication (Strutton & Roswinanto, 2014). Moreover, they can become so popular that people use them in situations that are completely unrelated to the intended purchase or usage situations (Mitchell, Macklin, & Paxman, 2007). In order to be replicated by others and therefore be successful, slogans should be unique, memorable, credible and marketable (Mitchell et al., 2007). The replication of slogans is most often the result of TV advertising (Mitchell et al., 2007). Slogans are the most “talkable” brand element and are frequently used in advertising campaigns. With half of all branded conversations involving brand related communications (E. Keller, 2007) there is a good change that some of these branded conversations include slogans. The ability of slogans to drive WOM is crucial, as the advertising containing the slogans is frequently perceived as being of a persuasive nature by consumers and therefore having decreased effects on influencing consumer responses (Laran et al., 2011a). WOM, however, is often perceived as highly reliable, thus useful information that influences consumer responses
13 (Bambauer-Sachse & Mangold, 2011). Even if the WOM communication is about the same advertisements that were perceived in a negative way at first.
3.3.1 Slogan Vagueness
Slogans consist of words that aim to describe or differentiate a brand. However, not all slogans have a precise and clear message. In this thesis the vagueness of a slogan will be used as the opposite of the preciseness (Cutting, 2007, in Strutton & Roswinanto, 2014), clarity, or vividness of a slogan’s purpose. Vividness has been defined as “referring to the extent [a] mental picture is lifelike or resembles real seeing under actual conditions” (McKelvie, 1995, p. 251) and is seen as similar to clarity and liveliness (Marks, 1995). Various research has been done on the effect of vividness of advertisements (Fennis, Das, & Fransen, 2012; Fortin & Dholakia, 2005; Rossiter & Percy, 1978), however the results are inconsistent. Some research found a positive effect
between vividness and persuasiveness, whereas others found a negative effect or no effect at all. This has led to increasing research on the effect of vagueness. According to Carter and McCarthy (2006, p. 928), vague expressions “deliberately refer to people and things in a non-specific, imprecise way”. Bandler & Grinder (1976) created the Meta model. The Meta model was
initially used to for therapists to find vagueness in the responses of patients. However, the model has been used to identify vagueness in sentences, such as brand slogans (Strutton & Roswinanto, 2014). The primary notion behind the Meta model is that we make models of everything that happens around us, however these models are mere representations of the world not the world itself (Dilts & DeLozier, 2000).
Vagueness is an important construct, because it can be used strategically to make a tradeoff between the effectiveness and relevance of a slogan (Strutton & Roswinanto, 2014). Strutton &
14 Roswinanto (2014) studied the effect of vague slogans on consumer responses and found that the interaction effect of slogan vagueness and length has an effect on evoked thought, brand attitude and persuasiveness. However, the authors didn’t research the effect on WOM as a consumer response, even though an increase in evoked thought could lead to increased accessibility of the slogan and increased brand attitude could lead to increased emotional or functional motivation to engage in WOM. With accessibility and motivation being drivers of WOM, a positive
relationship between slogan vagueness and WOM intentions can be expected, therefore the following hypothesis is proposed:
H1: There is a positive relationship between slogan vagueness and WOM intentions
3.3.2 Slogan Liking
In successfully building a brand, brand managers can put emphasis on the creation of likeable, memorable and meaningful brand elements (Cui, Hu, & Griffith, 2014). Creating a likeable slogan is important, as slogan liking can transfer into brand preference (Vakratsas & Ambler, 1999). Traditionally, brand managers were seen as the sole creators of brand identity. Recently, this view has changed and increasing importance is put on the role of consumers as co-creators of a brand’s identity (Da Silveira et al., 2013). An example of how consumers can influence a brand’s identity is through WOM communications.
Research has shown several elements that influence a slogan’s likeability; clarity of purpose, preciseness, creativity, length, communicating benefits and containing rhymes (Dass et al., 2014; Hugh-Williams, 1996). Others argue that deliberate ambiguity can have positive effects on the liking of slogans (Lagerwerf, 2002) and that intentional use of moderate levels of complexity causes more elaborate processing (Kohli et al., 2007). Still, slogan vagueness can be seen as the
15 opposite of clarity and as clarity positively influences the likeability of a slogan, vagueness is expected to negatively influence the likeability of a slogan. Research has yet to validate the link between slogan vagueness and slogan liking, hence the following hypothesis is proposed: H2: There is a negative relationship between slogan vagueness and slogan liking.
There is a positive relationship between the likeability of a stimuli and the processing of said stimuli (Dahlén & Rosengren, 2005). Increased processing of a slogan, leads to greater accessibility of the slogan. Furthermore, talking about interesting brands makes the person talking about them seem more interesting (Dye, 2000; Hughes, 2005). Something interesting, like a slogan someone likes, may increase someone’s social motivation to engage in WOM. As accessibility can drive WOM and likeability increases the social motivation to engage in WOM, greater slogan likeability is expected to have a positive effect on WOM intentions, therefore the following hypothesis is proposed:
H3: There is a positive relationship between slogan liking and WOM intentions.
Slogan vagueness is expected to have a negative effect on slogan liking. Therefore, the higher the vagueness of the slogan the lower the liking of the slogan will be. In turn, slogan liking is expected to have a positive effect on WOM intentions. So, the higher the slogan liking the higher someone’s intentions will be to engage in WOM. Consequently, a vague slogan is expected to decrease liking, which will result in lower WOM intentions. Therefore the following hypothesis is proposed:
H4: The positive relationship between slogan vagueness and WOM intentions is partially mediated by slogan liking, so that slogan vagueness will decrease WOM intentions through a decrease in slogan liking.
16
3.4 Conceptual Model
Figure 1: Conceptual model
H1: There is a positive relationship between slogan vagueness and WOM intentions. H2: There is a negative relationship between slogan vagueness and slogan liking. H3: There is a positive relationship between slogan liking and WOM intentions.
H4: The positive relationship between slogan vagueness and WOM intentions is partially mediated by slogan liking, so that slogan vagueness will decrease WOM intentions through a decrease in slogan liking.
17
4 Method
The research for this thesis was of explanatory nature. To answer the proposed research question a quantitative study, by means of an experiment, was performed. The data collection consisted of several steps. First, two master students determined the level of vagueness in a list of
pre-selected slogans using the adapted version of Meta model (Bandler & Grinder, 1976) by Strutton & Roswinanto (2014). Second, a pre-test was performed, to ensure maximum variation in slogan vagueness and slogan liking and to control for potential brand effects in the four slogans that were used in the experiment. Third, a survey was used as the data collection instrument to gather cross-sectional data. Four different blocks were prepared in Qualtrics, to create four conditions that provided maximum variation in the independent (slogan vagueness) and mediator variable (slogan liking). Respondents were randomly assigned to one of the four conditions of the experiment, all answering the same questions on slogan vagueness, slogan liking and WOM intentions about differing stimuli. The surveys were administered online, however the
respondents were approached both online and offline. The experiment required a minimum of 232 respondents for the data to be analyzable. Detailed descriptions of all the data collection steps can be found in the following subchapters.
4.1 Measures
All scales used in the pretest and experiment used very similar 7-point Likert scales, for the purpose of creating a consistent survey. Using only Likert scales increases the easiness of use for the respondents, as semantic differential scales are considered “cognitively more complex” (Friborg, Martinussen, & Rosenvinge, 2006, p. 882). Moreover, the respondents were deemed to be more familiar with, and therefore more capable of, answering questions posed using a Likert scale compared to semantic differential scales. Additionally, translations were provided for
18 English words that could have been misconstrued by respondents with limited understanding of the English language.
4.1.1 Slogan Vagueness
Vagueness, as used in this thesis, was defined as the opposite of clarity, preciseness and vividness (Cutting, 2007, in Strutton & Roswinanto, 2014). To measure slogan vagueness as assessed by the participants in the pretest and experiment, the validated four item semantic differential scale by Strutton & Roswinanto (2014) was adapted to a 7-point Likert scale to match the other measures (E.g. The slogan is vague). The original scale has a Cronbach’s alpha of .95 and includes two counter indicative items. The adapted version of the scale can be found in appendix 10.1.
4.1.2 Slogan Familiarity
Familiarity was measured as it is a condition that has to be met in order for something to be liked. Familiarity can be described as “recognizing a person [or construct] as familiar but not being able to recollect who the person [or construct] is or where they were previously
encountered” (Yonelinas, 2002, p. 441). To measure the extent to which the respondents in the pretest were familiar with the slogans, the three item semantic differential scale by Simonin & Ruth (1998) with a Cronbach’s alpha of .80 and .94 measuring brand familiarity, was adapted to a 7-point Likert scale to match the other measures (E.g. I recognize the slogan). The adapted version of the scale can be found in appendix 10.1.
4.1.3 Slogan Liking
Slogan liking can be described as a favorable affective response one has toward a slogan (Peter & Olson, 2010). The validated two item 7-point Likert scale by Dass et al. (2014) was used to
19 measure the slogan liking of the respondents in both the pretest and the experiment. The scale has a Cronbach’s alpha of 0.90 (E.g. Overall, how much do you like this slogan?).
4.1.4 WOM Intentions
WOM intentions represent the likelihood that the respondents will discuss the slogan with a friend or relative. To measure respondents’ WOM intentions in the experiment, an adapted version of the three item 7-point Likert scale by Maxham (2001) will be used in the experiment and has a Cronbach’s alpha of 0.91 (E.g. How likely is it that you would talk to your
friend(s)/family about this slogan?). The adapted version of the scale can be found in appendix 10.1.
4.1.5 Brand Attitude
Brand attitude can be defined as, “the beliefs about attributes of the advertised brand” (Olson & Mitchell, 1981, p. 320). An adapted version of the validated three item 9-point semantic
differential scale by Till & Busler (2000) was used to measure brand attitude in the pretest. The scale has a Cronbach’s alpha of 0.92 and was adapted to a 7-point Likert scale to match the other measures (E.g. I like (Brand X)). The adapted version of the scale can be found in appendix 10.1. 4.1.6 Control Variables
The pretest and experiment both contained three control variables and asked the respondents to answer questions about their gender (nominal variable), age (ratio variable) and educational background (ordinal variable).
4.2 Slogan Selection
A list of slogans was prepared, from which slogans were drawn that provided maximum variation in both the independent and mediator variable. The slogans were gathered from
20 company websites and databases, such as superslogans.nl and Textart.ru. Previous research by Strutton & Roswinanto (2014) has used the same method of slogan preselection. This allowed further selection of existing slogans. Previous research by Strutton & Roswinanto (2014) manipulated real-life slogans to create variation in vagueness among their chosen stimuli. However, for this research real-life slogans needed to provide this variation, as familiarity is a condition that has to be met in order for people to be able to like a slogan (Dass et al., 2014). A total of 126 slogans were gathered for use. A complete list of the selected slogans, including a translation of Dutch slogans, can be found in table 12 in appendix 10.2.
Two master students with a background in Marketing determined the level of vagueness in a list of prepared slogans using the Meta model (Bandler & Grinder, 1976). The Meta model defines three dimensions that help determine the vagueness of a slogan; deletion, generalization and distortion (Dilts & DeLozier, 2000). Deletion is when someone selectively chooses to put greater importance on some parts of an experience and eliminates others (Dilts & DeLozier, 2000). Generalization encompasses the detaching of parts of our model from actual experiences and having them represent the overall category of experiences that the original experience belongs to (Dilts & DeLozier, 2000). Distortion “allows us to make shifts in our experience of sensory data” (Dilts & DeLozier, 2000, p. 734). Previous research on vagueness in slogans has used the Meta model and its ten sub dimensions to assess vagueness in slogans (Strutton & Roswinanto, 2014).
The master students were presented with a useful model for assessing vagueness in slogans as created by Strutton & Roswinanto (2014) based on the Meta model. An overview of the entire model can be found in figure 6 in appendix 10.3. In addition to the model, the master students
21 received five example slogans that were rated on vagueness by the author and were thoroughly discussed before they started the process of assessing the remaining slogans on vagueness. The master students were deemed capable as they have very similar backgrounds and relevant
knowledge on the topic of branding. The slogans were rated on the number of sub dimensions of vagueness present in a slogan. This resulted in a list of slogans ranked on a scale of zero to ten, with zero representing no vagueness and ten representing maximum vagueness. To see if the results of the independent raters showed enough consistency to be used in further steps, the inter-rater reliability was tested applying the most widely used interjudge reliability measure Cohen’s kappa (Cohen, 1960). The results showed that the inter-rater reliability was too low, as Cohen’s kappa was .323 for p =.000. This is smaller than .40, which is interpreted as poor reliability (Fleiss, 1981). Therefore, the same master students in combination with two other master
students, with similar backgrounds and relevant knowledge on branding, were asked to select the slogans that they deemed highly vague and very clear. The slogans that were deemed highly vague and very clear by all of the master students were then used in the pretest.
4.3 Pretest
A pretest was performed, for which the primary objective was to ensure maximum variation in the vagueness and likeability of the selected slogans for the experiment and to control for any potential interfering effect of brand attitudes on the conceptual model. To control for the potential effect of brand attitudes on the conceptual model, the 23 selected slogans that were deemed low/high in vagueness by the master students were measured on brand attitudes. The slogans that were selected to be used in the experiment had to have similar mean scores in brand attitudes, regardless of them being high or low. Additionally the same slogans were measured on slogan familiarity, slogan vagueness and slogan likeability to ensure maximum variation in
22 vagueness and liking of the selected slogans for the experiment. The slogans used in the pretest can be found in table 1 below. The complete pretest, as distributed online, can be found in appendix 10.4.
Table 1
Selected slogans for pretest.
Brand Slogan
Milka Durf teder te zijn
Canon You can
KLM Thuis in de lucht
Gamma Dat zeg ik
EA Sports It’s in the game
Nescafé Open up
Nationale Nederlanden Wat er ook gebeurt Esso Stop een tijger in je tank
UPS Consider it done
Twix Dubbel genieten
Telfort Werkt in je voordeel
Chocomel De enige echte
De Hypotheker Jazeker, (Brand name)
Jumbo Hallo (Brand name)
Beter Bed Retteketet, naar (Brand name) Maybelline Maybe, it’s (Brand name)
Interpolis Glashelder
Volkswagen Das auto
Maggi Een beetje variatie in de Nederlandse keuken Haribo The happy world of (Brand name)
Heineken Biertje?
Coca-Cola Open Happiness
Nike Just do it
The pretest was answered by 53 respondents, of which the majority was female (62%). The mean age of the sample was 25.34 years and the educational background of the majority was a
Master’s degree (62%). The mean scores of the measured constructs for all the slogans used in the pretest can be found in table 14 in appendix 10.5. After careful consideration of the results on all constructs, four slogans were selected that were deemed most suitable. The results for the four selected brands can be found in table 2 presented below.
23
Table 2
Selected slogans’ mean scores of the measured constructs in the pretest.
Heineken Milka De Hypotheker Coca-Cola
Slogan Vagueness 2.25 4.39 3.77 3.57
Slogan Liking 5.39 3.81 4.06 5.28
Brand Attitude 5.66 5.72 5.32 5.61
The slogans selected for use in the experiment created maximum variation in slogan liking. However, the results regarding vagueness were less promising. The slogans by De Hypotheker and Coca-Cola did not significantly differ, as both slogans score closer to average rather than high or low on this construct. Nevertheless, the slogans by these brands were most appropriate for inclusion in the experiment. Furthermore, the brands behind these slogans did not differ significantly on brand attitudes, except for the brand attitude for De Hypotheker. This brand attitude was significantly lower compared to the other three brands, however it was the best slogan for the low vagueness, low liking condition. It is possible that the lower brand attitude for De Hypotheker is a result of differing attitudes towards the industry, as the other brands are all in the fast moving consumer goods industry and De Hypotheker is a mortgage broker.
4.4 Data Collection
This research is of explanatory nature. To answer the proposed research question a quantitative study, by means of an experiment, was performed. A survey was used as the data collection instrument to gather cross-sectional data. Four different surveys were prepared using different slogans in each survey to ensure maximum variation in the independent and mediator variable. The survey was created using Qualtrics and included four blocks for the four conditions and every respondent was randomly exposed to one of the conditions. The survey was administered online. Each variant of the survey contained the same questions, however the stimulus presented in each variant differed. The four brands that were selected after the pretest were Heineken,
24 Coca-Cola, De Hypotheker and Milka. Figure 2 below shows a graphical representation of the brands used in the four conditions.
Low Vagueness High Vagueness High Liki ng Heineken Coca-Cola Low Liki ng De Hypotheker Milka
Figure 2: The brands used in the four conditions of the experiment.
Since the results showed that brand attitude for De Hypotheker was significantly lower than the others, the brand names were eliminated from the experiment to further avoid potential effects that brand attitudes could have on the conceptual model. The stimuli used for the experiment are the slogans belonging to each of the brands. To control for potential influences from the visual presentation of the slogan, the slogans used in the surveys were shown in a similar manner, using the same font, size, color and no additional images or information. The respondents were
randomly assigned to one of the four conditions to prevent for systematic and personal biases created by the researcher (Dean & Voss, 1999). These slogans as well as the specific condition that the brand and slogan belong to can be found in figure 3 below.
Respondent group:
Brand Slogan Condition
Group 1 Heineken Biertje? Low vagueness, High liking
Group 2 Milka Durf teder te zijn. High vagueness, Low liking Group 3 De Hypotheker Jazeker, (Brand name) Low vagueness, Low liking Group 4 Coca-Cola Open Happiness High vagueness, High liking
25 The population of interest for this study consisted of Dutch consumers, because they are the ones being exposed to the slogans in their everyday lives. As the population is large, the sampling frame was unknown and there was a restricted timeframe for data collection, the research was conducted using a non-probability convenience sample (Saunders, Lewis, & Thornhill, 2012). Upon drawing conclusions the limited generalizability of the sample should be considered. The researcher attempted to collect as many respondents as possible during the data collection period; however the minimum number of respondents over all four surveys was 232 respondents in order for the data to be analyzable. The sample size was determined using an existing rule of thumb, as suggested by Green (1991). The rule of thumb suggests that the appropriate sample size for running regression analysis when using one predictor variable is 58 for an expected moderate effect size (.13) using a power of .80 (Alpha = .05). Therefore, every condition required a sample of at least 58 respondents resulting in a minimum total of 232 respondents.
The objective of the experiment was to measure to what extent the respondents deemed the slogan to be vague, liked the slogan and the likelihood that they would talk to their friends or family about the slogan. The surveys contained four parts. Firstly, the respondents were given an introduction. Secondly, the respondents were presented with one slogan and questions measuring slogan vagueness, slogan liking and WOM intentions. The slogan was shown before every question. Thirdly, the respondents were asked to answer demographic questions regarding their gender, age and educational background. Finally, the respondents received a notification that they had completed the survey. A detailed overview of the four surveys can be found in appendix 10.6.
26
5 Data Analysis
Conducting the experiment resulted in a total of 297 unique responses over the four conditions. In order to answer the proposed research question several steps were taken in analyzing the data. Firstly, the sample is described, the representativeness of the sample is assessed and the samples of the four conditions are compared with one another using the demographical information gathered in the experiment. To provide a basic understanding of the sample, see if the results are generalizable to the population and examine whether the randomization of the respondents may have an effect on the results (Field, 2013). Secondly, counter-indicative items are recoded and the reliability of the measurement scales used in the experiment is assessed. The measurement scales need to show high levels of internal consistency, in order to be sure that they provide results upon which conclusions can be drawn (Field, 2013). Thirdly, a manipulation check is performed to confirm the effectiveness of the manipulations. Fourthly, the data is tested for normality to ensure that the conclusions that are drawn from the data are accurate (Field, 2013). Fifthly, a descriptive analysis and correlation analysis are performed to provide a basic
understanding of the results and examine the linear relationships between the variables (Field, 2013). All previous steps are preliminary steps, which are taken in order to assess the suitability of the data for further analysis. The data proved to be suitable for hypothesis testing, the last step of the data analysis. Here, the data are analyzed using three hierarchical regression analyses and the process macro for mediation by Hayes (2013). The hierarchical regression analyses allow us to analyze the predictive ability of the control variables and the independent variable on the outcome variable separately (Field, 2013). The process macro for mediation is used to examine the extent to which slogan vagueness influences WOM intentions through the mediator variable
27 slogan liking (Field, 2013). All steps of the data analysis are described in great detail in the following subchapters.
5.1 Sample
On 13 December 2015 the survey used for data collection was opened. The survey was closed on 21 December 2015 after nine days. The data collection resulted in 322 unique responses, well over the required minimum of 232 respondents. This survey was distributed online through personal e-mail and Facebook and offline by approaching students at the Roeterseilandcampus of the University of Amsterdam (UvA). However, all surveys were filled in digitally. Family
members and close friends were invited through email, some redistributed the email to their family/friends/colleagues. This sample was fairly small, however it most likely resulted in the highest response rate. Other friends and acquaintances were invited through Facebook
messenger, this sample was very large. It resulted in a response rate significantly lower than that of family and friends, yet it was still fairly high due to the personalized messages. All digital invitations to the survey were accompanied by an explanatory message, which can be found in appendix 10.7. Approaching students at the UvA resulted in a high response rate (91%), as most people that were approached agreed to participate in the study (128 people) and only a few refused to participate (12 people). A lower mean age for Facebook and UvA respondents is expected, as well as a higher educational background for UvA respondents. However, respondents are randomly assigned to a condition so the effect will be the same over all four conditions. In total 25 cases had to be excluded as a result of missing data on all variables and a sample of 297 valid responses were gathered.
28 The sample consists of 297 respondents (Mage = 26.53, SDage = 9.9, age range: 16-73), of which 117 (39.4%) are male and 180 (60.6%) are female. The age distribution shows positive
skewness, as a considerable part (77.8%) of the respondents were below the age of 26. The education distribution shows some negative skewness, as the majority (78.8%) of the
respondents have a bachelor’s or master’s degree. Similarity of the sample to that of the Dutch consumer population was assessed using the demographic information (age, gender, educational background) gathered in the surveys. It is important to consider that the sample may not be representative of the Dutch consumer population when drawing conclusions. Table 3 gives more insight into the demographics of the respondents assigned to the different conditions.
Table 3
Demographical overview of the samples for separate conditions and the entire sample.
Condition:
Exposed to Sample
size Gender Age Education
Slogan N Male Female Mean
age < High School High School MBO Bachelor HBO/WO Master Condition 1 Biertje? 73 31 42 24.95 0 12 6 35 20
Condition 2 Durf teder te zijn. 73 23 50 27.56 1 11 7 31 23
Condition 3 Jazeker, (Brand name) 73 30 43 27.04 0 7 8 37 21
Condition 4 Open Happiness 78 33 45 26.56 0 5 6 47 20
Total 297 117 180 26.53 1 35 27 150 84
Since the respondents were randomly assigned to one of four conditions it is important to examine if there are significant differences between samples of the different conditions before further statistical analysis of the data. As significant differences between the samples of the conditions may affect the results (Field, 2013). A Pearson Chi-Square test was run to examine potential differences of the gender and educational background distributions between the samples, as gender and educational background were measured as categorical variables and contained more than two independent groups (Field, 2013). The results show that the gender distributions between the samples did not significantly differ, χ2 (3, N = 297) = 2.556, p = .465
29 and that the educational distributions between the samples did not significantly differ, χ2 (3, N = 297) = 10.802, p = .546.
To test for differences in age distribution between the samples a different test had to be
performed, as age was measured as a numerical variable. Since we are looking at more than two independent groups an ANOVA was performed (Field, 2013). The Levene’s test statistic is significant (F = 3.83, p = .010), which shows that the variances of the groups are statistically different from one another (Field, 2013). Welch’s test is performed to correct for the unequal variances (Field, 2013). The results show that the age distributions between the samples did not significantly differ, F(3,159) = 1.31, p = .272. As the control variables show that there are no significant differences between the samples, it can be concluded that the randomization of
respondents did not result in a response bias and will not affect the results. A full overview of the test results for the differences between the samples can be found in tables 15–18 in appendix 10.8. No missing values in the data were encountered, as all questions in the survey required a response in order to complete the survey. This may have resulted in the 25 cases where no responses was found, however as a result all remaining 297 responses are suitable for analysis.
5.2 Reliability
Before testing the reliability of the measurement scales two items of the vagueness scale (rVague2 and rVague3) needed to be recoded, as they were counter indicative. That entails that agreeing with the statement of these variables represented a low level of vagueness instead of a high level of the construct. Therefore, rVague2 and rVague3 were recoded into different variables and are now coded as Vague2 and Vague3. The same thing was done for the pretest variables and the reliability results of the pretest can be found in table 25 in appendix 10.9.
30 An overview of the Cronbach’s alpha scores for all scales in all conditions of the experiment can be found in table 4 below. The construct scales measuring slogan liking and WOM intentions clearly show sufficient reliability as their Cronbach’s alpha is above the .70 threshold.
Furthermore, for both scales the corrected item-total correlations show that all items have a good correlation with the total score of the scale (greater than .30). Neither slogan liking nor WOM Intentions would benefit from removing an item from the scale as the change in Cronbach’s alpha after deletion never results in a substantial change in reliability (bigger than .10). However, the construct scale for slogan vagueness does not show sufficient reliability as the Cronbach’s alpha repeatedly is below the .70 threshold. Slogan vagueness would benefit from removing an item from the scale, as deleting item four (Vague4) would increase the Cronbach’s alpha to above the .70 threshold in all four conditions. Furthermore, the fourth item of the scale only seems to be a good item in condition two. However, it just barely shows a corrected item-total correlation above the .30 threshold (.315). In the other conditions it does not reach a corrected item-total correlation above .30 and therefore does not have a good correlation with the total score of the scale. Though, guidelines suggest only to delete an item if the Cronbach’s alpha changes with more than .10, which only happens in condition four, it was decided to delete item four (Vague4) from the scale before subsequent analysis was run. After the removal of one item from the vagueness scale, all scales showed a Cronbach’s alpha higher than .70. Therefore, the scales can be considered to have high levels of internal consistency.
31
Table 4
Reliability scores for construct scales of the experiment.
Condition: Exposed to Sample Size Slogan Vagueness (4-item) Slogan Vagueness (3-item) Slogan Liking WOM Intentions Slogan n α α α α Condition 1 Biertje? 73 .685 .732 .871 .857
Condition 2 Durf teder te zijn. 73 .670 .701 .862 .857
Condition 3 Jazeker, (Brand name) 73 .689 .759 .919 .751
Condition 4 Open Happiness 78 .573 .786 .830 .857
5.3 Manipulation Check
Respondents were randomly assigned to one of the four conditions of the experiment, which represent low/high vagueness and low/high liking. A manipulation check, by means of a one-way ANOVA, was performed to confirm the effectiveness of the manipulations. The ANOVA allows us to compare more than two conditions and to compare the differences of their means (Field, 2013). Therefore, it allows us to test if the slogans selected as stimuli for the experiment are good indicators of low/high slogan vagueness and low/high slogan liking. Respondents were asked to rate both their perceived vagueness and liking of the slogans. The results for vagueness show that Levene’s test is not significant (p=.543, so equal variances can be assumed) and that the effect of vagueness is significant, F(3, 293) = 34.47, p = .000. To find out which conditions were significantly different on vagueness Tukey’s post hoc test was run (Field, 2013). The results show that all conditions are statistically different regarding vagueness except for
condition two and three, which is not in line with the intended manipulations. Ideally, condition one and three as well as condition two and four would not have been significantly different from one another, as they were deemed to be both low/high in vagueness. The results for liking show that Levene’s test is not significant (p=.128, so equal variances can be assumed) and that the effect of liking is significant, F(3, 293) = 14.94, p = .000. To find out which conditions were
32 significantly different on vagueness Tukey’s post hoc test was run (Field, 2013). The results show that all conditions are statistically different regarding liking except for condition two and four, which is not in line with the intended manipulations. Ideally, condition two and three as well as condition one and four would not have been significantly different from one another, as they were deemed to be both low/high in likeability. Results of the manipulation check can be found in tables 19-24 in appendix 10.8. Upon drawing conclusions it is important to consider that the manipulations were not completely effective.
5.4 Skewness and Kurtosis
The data was tested for normality using both the Kolmogorov-Smirnov test and the Shapiro-Wilk test and the results can be found in table 5 below. Taking the entire sample in these tests would certainly show non-normality as we expect bimodal distribution (Field, 2013), because these conditions were manipulated to test either low/high degrees of slogan vagueness and slogan liking. Therefore, we expect normal distributions per condition and the data was split per
condition before testing for normality. Nevertheless, most variables showed signs of non-normal distribution, as they reported that the distribution is significantly different from a normal
distribution. However, for all conditions we are dealing with a large sample size and Field (2013) identified that large samples will result in significant results even for minor effects. Which can be seen from the results as almost all effects are significant, but smaller than 0.2 in effect size.
33
Table 5
Kolmogorov-Smirnov and Shapiro-Wilk tests for normality.
Conditions Variables n Kolmogorov-Smirnov Shapiro-Wilk
Condition 1 Slogan Vagueness 73 .135** .939** Slogan Liking 73 .177*** .889*** WOM Intentions 73 .094 .970 Condition 2 Slogan Vagueness 73 .116* .965* Slogan Liking 73 .136** .963* WOM Intentions 73 .108* .962* Condition 3 Slogan Vagueness 73 .108* .972 Slogan Liking 73 .199*** .909*** WOM Intentions 73 .135** .966* Condition 4 Slogan Vagueness 78 .196*** .934** Slogan Liking 78 .155*** .939** WOM Intentions 78 .102* .964* Statistical significance: *p <.05; **p <.01; ***p <.001
According to Field (2013), when dealing with large samples (N>20) one should neither use before mentioned tests nor worry about normality (p. 184). However, to be sure that a normal distribution could be assumed, the skewness and kurtosis levels of the main variables were examined according to a rule of thumb. The skewness and kurtosis results can be found in table 6 below. This rule states that normality can be assumed with skewness and kurtosis levels within the -1 and +1 range (Leech, Barrett, & Morgan, 2005). The main variables show that the
skewness and kurtosis levels are within the aforementioned range. Therefore, no transformation of variables was deemed necessary.
Table 6
Skewness and Kurtosis.
Variables n Skewness Std. Error Kurtosis Std. Error 1. Slogan Vagueness 297 -.197 .141 -.887 .282
2. Slogan Liking 297 -.722 .141 .301 .282
3. WOM Intentions 297 .315 .141 -.691 .282
5.5 Descriptive Analysis and Correlation Matrix
Scale means have been computed for Slogan Vagueness, Slogan Liking and WOM intentions, and they were coded as VagueTOT, LikingTOT and WOMTOT. The scale mean for Slogan Vagueness was computed for a 3-item scale, instead of the original 4-item scale. A complete list
34 of the variables that were used for this thesis can be found in appendix 10.10. A descriptive analysis was performed to provide a basic understanding of the results. The means and standard deviations of the scales per condition can be found in table 7 below. The analysis results show that condition 1 (low vagueness, high liking) reports the lowest vagueness score and the highest liking score as was expected. Moreover, condition 4 (high vagueness, high liking) has the highest mean score for vagueness and the second highest score for slogan liking. However, condition 2 (high vagueness, low liking) and 3 (low vagueness, low liking) seem to deviate from their intended purpose. Condition 2 does show high vagueness, however it also scores relatively high on slogan liking. Furthermore, condition 3 shows a fairly high vagueness score and even though it has the lowest liking score its mean is still well above the 3.5 average of the 7-point Likert scale. As mentioned before in chapter 5.3, it is important to consider that the conditions may not have created maximum variation in the independent and mediator variable when drawing
conclusions.
Table 7
Means and Standard Deviations per condition.
Condition: Exposed to Sample Size Slogan Vagueness Slogan Liking WOM Intentions 3-item scale 2-item scale 3-item scale
Slogan n μ σ μ σ μ σ
Condition 1 Biertje? 73 2.795 1.339 5.336 1.121 4.110 1.498
Condition 2 Durf teder te zijn. 73 4.256 1.282 4.610 1.188 3.388 1.518
Condition 3 Jazeker, (Brand name) 73 4.155 1.332 4.014 1.310 2.954 1.171
Condition 4 Open Happiness 78 4.901 1.268 4.731 1.170 3.077 1.261
The next step in analyzing the data is to review the correlations. This step is done to examine whether relationships exist between variables and what directions (positive/negative) these relationships have (Field, 2013). Therefore, a correlation matrix was created for all combinations of variables. Table 8 gives an overview of the means, standard deviations, correlations and reliability scores for all the constructs and control variables. The correlation matrix shows results
35 for the entire sample (n=297) regardless of the conditions respondents were exposed to. Included in the correlation analysis were the control variables (gender, age and educational background) as well as the main variables Slogan Vagueness, Slogan Liking and WOM intentions.
Table 8
Means, Standard Deviations, Correlations and Reliability scores.
Variables Mean SD 1 2 3 4 5 6 1. Gender 1.61 .489 2. Age 26.53 9.91 -.056 3. Educational background 3.95 .936 -.002 -.177** 4. Slogan Vagueness 4.04 1.51 -.026 .054 .026 (.808) 5. Slogan Liking 4.67 1.28 -.060 -.001 -.012 -.424** (.888) 6. WOM Intentions 3.38 1.43 .073 .031 -.049 -.368** .575** (.840) ** Correlation is significant at the 0.01 level (2-tailed)
The correlation matrix shows that there is one significant correlation between the control variables. A negative correlation is found between age and educational background. Indicating that younger people have higher educational backgrounds compared to older people. However, the strength of the correlation, with a Pearson Correlation Coefficient of r(297)= -.177, p<.01, is too small to consider it a relation at all. The correlation can be explained as the youngest part of the sample was mainly gathered on the UvA campus, resulting in a higher educational
background. Whereas the educational distribution of the older part of the sample contained greater differentiation. No significant correlations were found between the control variables and the main variables.
Several significant correlations were found between the main variables. The results show that Slogan Vagueness is negatively correlated with both Slogan Liking (r(297)= -.424, p<.01) and WOM Intentions (r(297)= -.368, p<.01). The negative correlation between vagueness and liking is expected, as it is in line with the second hypothesis. Furthermore, previous research by Dass et al. (2014) already established the positive relationship between clarity and liking. Since
36 vagueness in this research is considered to be the opposite of clarity a negative relationship could be expected. However, the negative correlation between vagueness and WOM was unexpected. Slogan vagueness was hypothesized to have a positive relationship with WOM Intentions. Moreover, a strong positive correlation was found between slogan liking and WOM Intentions, with a Pearson Correlation Coefficient of r(297) = .575, p<.01. Which was in line with
expectations, as the third hypothesis predicts a positive relationship between slogan liking and WOM intentions. The fourth hypothesis predicts the partial mediating role of Slogan Liking on the relationship between Slogan Vagueness and WOM Intentions, as all variables are correlated amongst each other partial mediation is possible.
5.6 Hypotheses Testing
The hypotheses were tested using two different methods. First, three hierarchical regression analyses are performed, which allow us to analyze the predictive ability of the control variables and the independent variable on the outcome variable separately (Field, 2013). Second, the process macro for mediation by Hayes (2013) is used, as it examines the extent to which slogan vagueness influences WOM intentions through the mediator variable slogan liking (Field, 2013).
First, a hierarchical regression analysis was performed to test the first hypothesis. The hierarchical regression analysis examines the linear relationship between the independent
variable (Slogan Vagueness) and the dependent variable (WOM Intentions) while controlling for the effects of the demographic variables (Field, 2013).
