A contingent view of narrative content in the age of Content Marketing
Master thesisMSc in Business Administration- Marketing Track
Janneke Flipse-10533117 University of Amsterdam
Master thesis
MSc in Business Administration- Marketing Track Supervisor: Marco Mossinkoff
Statement of Originality
This document is written by Janneke Flipse 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.
Abstract
Content marketing has become a popular trend in the field (Korosec, 2012). Narrative content, an element of content marketing, is seen as the key to success. Academics particularly researched the positive effects of narrative content and the amount of business investments increases considerably (Karr, 2016). In contrast, this thesis tries to undermine the idea of narrative content as the ‘holy grail’. This research project sheds light on moderating roles that can elicit a central way of information processing, compared to a peripheral way, which is often used when narrative content is involved (Petty & Cacioppo, 1983) (Lien & Chen, 2013). As starting point, the effect of narrative content on brand attitude, as an antecedent of purchase intention, is investigated. The moderating roles of the level of product involvement and the level of scepticism are taken into account while investigating this relationship. It is hypothesized that narrative content is more effective if a low involvement product is promoted (Akbari, 2015) and the positive effect of narrative content will disappear if people are sceptical about the content due to an analytical way of information processing (Escalas, 2007). On the basis of an online-experiment, in which 165 people participated, is concluded that narrative content did not significantly affect the attitude towards the brand. The results, concerning the moderating roles, were in line with the proposed hypotheses, however these results were insignificant. As expected, this thesis project revealed a significant correlation between scepticism and type of content. Narrative content is associated with a higher level of scepticism. Future research is required to prove significant results and to reinvestigate to which extent backfire effects occur due to a central way of information processing (Petty & Cacioppo, 1983).
Acknowledgements
Writing a master thesis is an inspiring project. However, it is also an intense project in which you are designated on your own. Successfully completing this research project required perseverance. Without the support of my family, friends and close fellow students it would have been much harder to see the light at the end of the tunnel while conducting this research. They always encouraged me. In particular, I would like to thank Bas Flipse, Marcel Flipse and Milou ter Hoek for the helpful discussions and feedback, which gave me new insights.
A special word of thanks to Marco Mossinkoff, for mentoring this project. You gave me the opportunity to choose a research subject where I feel connected with myself. I also benefited from the interesting discussions that we had and the feedback that you gave me.
Table of Contents
Statement of Originality 2
Abstract 3
Acknowledgements 4
1. Introduction 7
1.1. Practical and theoretical relevance 11
2. Literature Review 13 2.1. Brand attitude 14 2.2. Type of content 15 2.3. Moderating roles 18 3. Methodology 23 3.1. Strategy 23 3.2. Design 24 3.3. Sample 25 3.4. Procedure 26 3.5. Pilot-test 27 3.6. Stimuli 27 3.7. Measurements 29 3.8. Control variables 31 3.9. Pre-test 32 3.10. Data analysis 34 4. Results 38 4.1.Correlation analysis 38 4.2. Factorial ANOVA 39
4.2b. Interaction-effect hypothesis 2 40 4.2c. Interaction-effect hypothesis 3 41
5. Discussion 43
5.1. Conclusion 43
5.2. Practical and theoretical implications 47
5.3. Limitations and future research 48
References 53
Appendices 57
Appendix 1. Online-experiment exported from Qualtrics 57 Appendix 2. Results One Sample T-Tests (pre-test) 62
List of Tables
Table 1. Experimental design 25
Table 2. Sample size 26
Table 3. Facets of involvement 33
Table 4. Descriptive statistics scales Level of scepticism & Brand attitude 36
Table 5. Correlation matrix 39
Table 6. Results Factorial Analysis of Variance (ANOVA) 42
Table 7. Summary hypotheses 45
List of Figures
Figure 1. Conceptual model 11
1. Introduction
Marketing is a business philosophy, which implies ‘a way to approach the market’ (Sheth, 2011). This definition explicitly states that a general way to approach the market does not exist. Marketing thinking is contextual and dynamic. A general rule how to execute marketing overtime or in different contexts cannot be obtained (Sheth, 2011).
Something important has changed in the marketing field. Marketing used to be product-driven and solely focused on transactions. Currently, delivering value has become a key driver in marketing (Achrol & Kotler, 2012). Marketing is customer-driven; it is focused on building a relationship with customers and delivering unique value to them (Achrol & Kotler, 2012). Multiple business scholars and practitioners even argue that adding value to products is not the key to success while competing nowadays. The entire organization needs to view itself and needs to approach the market with a ‘dominant logic’ (Lusch, Vargo & O’Brien, 2007). The service-dominant logic focuses on the co-creation of value with their customers instead of on value distribution, which characterises the ‘goods-dominant logic’. It is about doing ‘things’ together with the customers instead of doings ‘things’ to them (Lusch, Vargo & O’Brien, 2007).
In brief, marketing thinking is dynamic and important transitions take place in the marketing field. Consequently, marketing tactics need to evolve as well (Sheth, 2011). Most marketers are already aware of the fact that consumers prefer a pull strategy rather than a push strategy. This entails that consumers prefer to search for information that is important to them instead of being bombarded with information
(Korosec, 2012). This is where content marketing comes in. Content marketing has become a popular trend in the marketing field (Korosec, 2012). This marketing technique can be defined as ‘a technique of creating and distributing valuable, relevant and consistent content to attract and acquire a clearly defined audience with the objective of driving profitable customer action’ (Contentmarketing, n.d.). Consumers expect more from companies. Their content needs to go beyond simply selling products and needs to add value to consumers’ lives to be relevant (Korosec, 2012).
Multiple illustrations of content marketing already exist. The last Christmas commercial of H&M ‘Come together’ is an example of this. Wes Anderson, a famous filmmaker, created a short movie with an impressive storyline for H&M (Nededog, 2016). The railway guard, Adrien Brody, told his passengers that the train was delayed due to challenging weather and mechanical difficulties. As a consequence, they had to celebrate Christmas in the train. The crew organised a Christmas brunch, which was enjoyed by everyone. In short, it was a great movie to watch, which had nothing to do with selling clothes. Moreover, as a watcher, you do not even recognize that you are watching a commercial until you are exposed to the H&M logo in the final shot (Nededog, 2016). This example illustrates that inspiring stories have replaced selling messages, which solely push and focus on product sales (Contentmarketing, n.d.).
Nowadays, as shown in the above-described H&M commercial, brands need to operate as media companies. Storytelling, in all its forms, is an important aspect of content marketing and is in the centre of ‘new marketing’ today (Pulizzi, 2012).
Stories have power. They are seen as the way to reach for and connect with people on an emotional level. Our preference for stories is embedded in our human basic need to understand patterns of life as a more emotional and personal experience (Smith, 2015). Storytelling is used as a way to create value to consumers with the aim to generate positive customer behaviour (Pulizzi, 2012). 90% of all companies already made use of this new kind of marketing technique in 2012, which gives the impression that this new way of marketing is the key to success. The future would be ‘publishing’ instead of ‘marketing’ (Pulizzi, 2012).
However, is it confirmed that storytelling always leads to profitable customer action (Contentmarketing, n.d.)? Or do organisations imitate each other’s marketing techniques, which causes a hype that lacks any scientific support? This is an important question if you are looking at the amount of investments in content marketing. Marketers are spending more than 25% of their total budget on content marketing and they have the intention to increase the amount of investments considerably (Karr, 2016).
N.B. Storytelling is the popular designation. This concept is described as narrative content in the literature. The same kind of referencing is used in this thesis.
The literature, however, paid less attention to the effect of narrative content on purchase intention. Furthermore, consistency about the strength and direction of this relationship lacks. This is not in line with the increasing focus on marketing accountability (Verhoef & Leeflang, 2011). This implies that it becomes more important for marketers to invest in marketing activities that cause sales. On the
contrary, evidence for the positive causal inference between narrative content and brand attitude can be found (Escalas, 2004b) (Van Laer et al., 2014). Brand attitude and purchase intention are highly correlated constructs, which highlights the importance of the effect of narrative content on brand attitude (Spears & Singh, 2004).
In addition, the literature suggests that narrative content, independently seen from the context, always outperforms. For example in terms of brand attitude, narrative content leads to a more positive brand attitude rather than content that conveys product information in a more straightforward and factual manner (Escalas, 2004b) (Van Laer et al., 2014). Previous research did not focus on the limitations with regard to the positive effects of narrative content. So far, less is known about the different effects for different types of products, such as low involvement products versus high involvement products. This is remarkable, because it has been confirmed that emotional advertising appeals, such as narrative content, have a different effect for low involvement products than for high involvement products (Akbari 2015). Furthermore, less is known about the extent to which backfire effects occur due to the level of scepticism. It already has been supported that narrative content does not always elicit a narrative way of processing. Consumers, who are sceptical about the content, process narrative content in a totally different way (Escalas, 2007).
In sum, research is necessary to obtain an overview of the influence of narrative content on brand attitude, as an antecedent of purchase intention, and the moderating effects of the level of scepticism and the level of product involvement. Based upon
these gaps in the already existing literature, the following research question will be investigated in this thesis:
What is the effect of narrative content in ads on brand attitude, as an antecedent of purchase intention, and what is the role of the level of product involvement and level of scepticism on this relationship?
Figure 1. Conceptual model
H1
H2 H3
1.1. Practical and theoretical relevance
Marketers and practitioners need to get a clear overview of how narrative content operates. As outlined before, it is important to pay more attention to the inference between narrative content and purchase intention. This is due to the increasing pressure on marketers and practitioners; it is a difficult time for marketers to justify their expenditures in terms of a return on investment (Verhoef & Leeflang, 2011). Furthermore, they need to fully understand in which cases backfire effects probably occur. The literature regarding this topic creates the impression that it is always clever to use narrative content. Actually, if this is always the case, the harmfulness of
Content type
Product
involvement scepticism Level of
backfire effects needs to be investigated. This research thesis will fill above-mentioned gaps in the literature by providing an overview of the effect of narrative content on brand attitude, as an antecedent of purchase intention, and by focusing on moderation effects.
In addition, it is also considered to be of high relevance for the literature to fulfil the existing gaps. As discussed earlier, the literature is not complete yet. This is probably due to the newness of content marketing and narrative content (Korosec, 2012). This thesis will contribute to the theory by clarifying several remaining issues. Furthermore, this research project contributes to the existing literature by focusing on the Elaboration Likelihood Model, a dual-process theory of attitudes formation, as an explanation for effects that presumably occur. The Elaboration Likelihood Model outlines two different ways in which information can be processed, the central route and the peripheral route. The type of information processing is determined by several factors, such as someone’s motivation or ability to process the information (Petty & Cacioppo, 1983).
2. Literature Review
This section provides an overview of the existing literature with regard to the fundamental subjects of this thesis. The aim of this section is to elaborate on previous researches to be able to build on it.
Marketing accountability becomes more important in today’s organisations (Verhoef & Leeflang, 2011). This is due to the fact that marketing is in deep trouble. Marketing is often seen as a cost instead of an investment. To regain their positions, marketers need to focus on the accountability of their expenditures (Verhoef & Leeflang, 2011). Because of this reason, sales drive practitioners and marketers nowadays.
In spite of the high relevance of this variable, the literature, with regard to narrative content, paid less attention to the effects on purchase intention. The Elaboration Likelihood Model is a dual process theory that describes the change of attitudes formation. This theory assumes that an attitudinal change, caused by a peripheral cue, cannot predict subsequent behaviour (Petty & Cacioppo, 1983). This is due to the peripheral way of processing, an information processing route, which is used by people who are not able or motivated to process the information. If people use the peripheral route, they form their attitude based upon a cue, such as narrative content. In short, according to the Elaboration Likelihood Model, narrative content cannot lead to behavioural changes due to peripheral processing (Petty & Cacioppo, 1983).
In contrast, evidence for a positive effect of narrative content on behavioural intentions exists as well. Van Laer et al. (2014) confirmed that narrative content increases story-consistent intentions. In addition, Escalas (2004b) found a positive
effect on behavioural intent itself. These articles support the positive effect of narrative content on behavioural intentions. However, ‘behavioural intentions’ is a broad and vague concept.
To conclude, consistency about the influence of narrative content on purchase intention lacks. A deeper understanding of the effects of different types of content is needed to be able to clarify this relationship.
2.1. Brand attitude
Positive effects of narrative content on attitudes, such as the attitude towards the product, advertisement or brand, have been supported in the literature. Escalas (2004a) was able to conclude that mental stimulation, in the form of narrative content, positively affects product and advertisement evaluations. This effect is also due to the peripheral way of processing; people use narrative content as a peripheral cue to form their attitude (Escalas, 2004a) (Petty & Cacioppo, 1983). Van Laer et al. (2014) concluded in their article that narrative content, indeed leads to transportation, and the higher the level of transportation, the more story-consistent affective responses, narrative thoughts, beliefs and attitudes positively increase. This article afresh confirmed that critical thoughts decrease due to narrative transportation.
Moreover, Escalas (2004b) confirmed a positive effect on brand attitude in general. ‘Brand attitude’ can be defined as the extent to which people respond favourably towards the brand. A brand attitude will be considered as more positive if people respond more favourable towards the brand. The opposite holds for a negative brand attitude (Escalas, 2004b). The positive causal inference between narrative content and
brand attitude is in line with the Elaboration Likelihood Model (Petty & Cacioppo, 1983). Information processing, according to the peripheral route, can lead to attitudinal change, because the attitude can be influenced by cues or simple decision rules (Petty & Cacioppo, 1983).
Brand attitude and purchase intentions are two popular dimensions, which are used very often by practitioners and scholars. These constructs are highly correlated which makes it empirically undistinguishable and unnecessary to investigate one more often in this research project (Spears & Singh, 2004). Because of this, this research focuses on brand attitude as an antecedent of purchase intention.
As outlined, a more positive brand attitude will increase the intention to purchase. With this in mind, it is important to look at content types, because of their influence on brand attitude (Escalas, 2004b).
2.2. Type of content
‘Content marketing’ is a broad concept that can be applied to several things (Korosec, 2012). Content marketing is not restricted to one communication channel; it can be used for different channels and the communication message itself is deemed more important. Content marketing is a tactic, which can be used to deliver value to consumers. The aim is to create content that consumers also want to receive. This kind of content engages consumers and sometimes they even want to share it themselves. It can be described as a more soft-sell approach; it promotes a product in an entertaining and indirect way. Content marketing is about ‘you, you’ instead of ‘me, me’ (Korosec, 2012). Empirical evidence for the positive effects of using content
marketing, as marketing tactic, is already available (Ahmad, Musa & Harun, 2016). It seems quite clear that content marketing is an important and effective new marketing tactic that needs to get attention.
This research focuses on narrative content, a common element of content marketing, in advertisements. Narrative content in advertisements conveys product information in a story-like format (Peracchio & Meyers-Levy, 1997). So, the underlying selling message and product information are presented by encumbered episodic prose (Milton, 1974). These advertisements are contextually rich, intriguing and have a conservation-like approach. Most often, narrative content contains details with regard to setting, event actions and characters (Peracchio & Meyers-Levy, 1997). According to Escalas (2004a), narrative content consists of a story element and two different items: chronology and causality. Over time, things occur in a story, which is arranged in episodes. Besides this, narrative thought structures story elements into an organised framework that establishes relationships between the elements and allows for causal inference (Escalas, 2004a). In another article, Escalas (2004b) discussed a third characteristic of narrative content: it shows a personal evaluation of one or more characters. Adaval and Wyer (1998) already defined narrative content in this way; it is seen as a temporally and thematically description of character consumption.
A non-narrative advertisement conveys product information without emphasising a story that can be associated with the product (Polyorat, Alden & Kim, 2007). It communicates fact-laden descriptions of benefits for the consumer and product features (Milton, 1974).
The processing of narrative content works differently compared to non-narrative content. Mental stimulation in the form of narrative content, leads to transportation, a state of immersion. This implies that people become lost in a story. Narrative processing is a totally different way of processing, especially compared to analytical processing (Escalas, 2004a). Analytical processing can also be described as the central route of the dual-process models of persuasion (Petty & Cacioppo, 1983). In this case, people use logical consideration and evaluate arguments critically. It is expected that non-narrative content focuses more on the central/analytical route. If people process the content due to the peripheral route, they use cues or heuristics that enable them to form an attitude quickly. A state of transportation, which is caused by narrative content, ensures that people use the peripheral route while processing the content. Then, people form an attitude based upon affective responses instead of logical consideration (Escalas, 2004a).
The authors Lien and Chen (2013) found empirical evidence for the different way of processing when narrative content is used. These researchers investigated the effect of argument strength for narrative and non-narrative ads. It is concluded that argument strength positively affects attitudes for non-narrative ads. This supports the idea that people use the central/analytical route while processing non-narrative content. This positive effect of argument strength declines for ads that contain narrative content. This result shows evidence for the fact that people process narrative content in a more peripheral route, in which arguments or ad claims are less evaluated by people (Lien & Chen, 2013). In addition, immersion in a narrative will diminish counter-arguing if the content is in contrast with the already existing attitude (Slater & Rouner, 2002).
It is confirmed that narrative content leads to a narrative way of processing due to a state of immersion, also called ‘transportation’ (Escalas, 2004b). A mediator that can explain several positive effects of narrative content is ‘self-brand connections’ (Escalas, 2004b). Narrative processing creates and increases self-brand connections. A link between the brand and the self (the consumer) will be created due to narrative processing. This happens when consumers try to map incoming stories with stories in memory. This will be positive for several outcomes; the more the brand is linked to the person, the more meaningful the brand becomes for that person (Escalas, 2004b).
Based upon previous studies and the Elaboration Likelihood Model, the following hypothesis is drawn up and will be researched:
Hypothesis 1: People who are exposed to narrative content in an advertisement have a more positive brand attitude than people who are exposed to non-narrative content.
2.3. Moderating roles
When diving into the literature, it conveys the impression that narrative content, irrespective of the context, leads to multiple positive outcomes. This is in contrast with, which is discussed before, that marketing strategy as well as marketing tactics are contextual (Sheth, 201). Hence, a more extensive literature analysis is required to conclude if narrative content always outperforms.
Product involvement
Akbari (2015) investigated the effect of different advertising appeals, rational and emotional appeals, on the attitude towards an advertisement and on purchase
intention. Rational advertising appeals are based on logic, facts, information and reasoning. Opposed to this, emotional appeals are intended to generate positive feelings, emotions and to build a brand personality. Even more important, he made a distinction between two different kinds of products: high involvement and low involvement products. High involvement products can be described as utilitarian products, which include high technology products or industrial goods e.g. (Johar & Sirgy, 1991). Low involvement products are seen as hedonic products, these products are value expressive, such as specialty goods or convenience products (Johar & Sirgy, 1991). Akbari (2015) concluded that both types of advertising appeals positively influence purchase intention and advertising attitude. However, emotional appeals have a higher influence for low involvement products and rational appeals work better for high involvement products (Akbari, 2015). These results are of high interest, because narrative content can be described as a more emotional appeal compared to product-relevant information, which is seen as a more rational appeal. Based upon this research, it can be expected that the effectiveness of narrative content depends on the level of product involvement. This has, however, never been tested before.
More evidence for this assumption can be found. Emotional appeals will, independent of the product type, generate a positive outcome. However, emotional appeals do indeed have a more positive influence if it is used to promote a low involvement hedonic product (Geuens, De Pelsmacker & Faseur, 2011). A hedonic product is a value-expressive product; the benefits of consuming the product are especially about the meaning of the product itself (Johar & Sirgy, 1991). This makes an emotional appeal, such as narrative content, more effective in this case.
Reiteratively, the Elaboration Likelihood Model can explain the outcomes (Petty & Cacioppo, 1983). If people lack the motivation or ability, they will process the information by the peripheral route. On the other hand, if they are motivated and able to process the information, they will follow the central route. People will have a higher motivation to process the content if it promotes a high involvement product, due to the extent to which they are involved in the decision-making process and the extent to which they search for information (Laurent & Kapferer, 1985). The opposite happens with regard to a low involvement product (Akbari, 2015). In the case of a low involvement product, people are less involved in the decision-making process and less likely to search for information (Laurent & Kapferer, 1985). As explained earlier, a rational appeal works better if people follow the central route and an emotional appeal is more successful if people use the peripheral route to process content (Petty & Cacioppo, 1983). Whether narrative content actually works this way and how it works for brand attitude remain questionable.
Based upon these findings in the literature, the following hypothesis will be investigated:
Hypothesis 2: The positive effect of narrative content in an advertisement on brand attitude is higher if a low involvement product is promoted than if a high involvement product is promoted.
Level of scepticism
As discussed before, narrative content will cause narrative processing and this leads to several desired outcomes. However, do consumers always elicit a narrative way of thinking?
In 2007, one article highlighted the importance of the moderating role of the level of scepticism towards narrative content. This article states that a narrative or analytical way of processing is not only determined by the content, but also moderated by the extent to which consumers think that the advertiser will manipulate them (Escalas, 2007). In brief, positive effects of narrative content disappear when consumers are sceptical. This is due to a more analytical way of processing (Escalas, 2007). It has to be mentioned that this research solely focuses on a particular element of narrative content, namely ‘narrative self-referencing’. Nevertheless, this assumption has been supported by the article of Wentzel, Tomczak and Herrmann (2010) who concluded that consumers evaluate narrative content more positive due to a narrative way of processing. An analytical processing is used, instead of narrative processing, when scepticism is present. The positive attitude disappears in this case (Wentzel, Tomczak & Herrmann, 2010).
These research projects suggest that narrative content does not always lead to positive evaluations. The positive advantage of narrative content over other forms of content disappears if analytical processing is used. However, the literature paid less attention to the negative effects of narrative content, if it does not elicit narrative processing. It seems reasonable that backfire effects occur if narrative content leads to analytical processing due to critical evaluation (Petty & Cacioppo, 1983). Philips and
McQuarrie (2010) already confirmed that narrative content problematically affects brands if consumers perceive the images as obvious endeavour to sell. However, more research is needed to investigate the extent to which backfire effects, due to scepticism, occur.
Based upon the already existing literature with regard to the moderating role of the level of scepticism, the following hypothesis is formulated and will be investigated in this thesis:
Hypothesis 3: The positive effect of narrative content in an advertisement on brand attitude disappears when people are sceptical about the content.
Taken this together, previous research give support to the causal inference between narrative content and several desired outcomes, such as a more favourable attitude towards the brand (Escalas, 2004b). Initially, the literature gave the impression that narrative content always operates in the same way, irrespective of factors and circumstances. However, Akbari (2015) found evidence for the differential effects of emotional appeals for low involvement products compared to high involvement products. Narrative content can be seen as an emotional appeal that will be processed according to the peripheral route, which makes this content type more effective in case of a low involvement product (Petty & Cacioppo, 1983). Narrative content will not always elicit a peripheral way of processing (Escalas, 2007). An analytical way of processing will be elicited if people are sceptical about the narrative content. The positive attitude will disappear in this situation (Wentzel, Tomczak & Herrmann, 2010). An online-experiment will be executed to investigate the proposed hypotheses.
3. Methodology
This chapter describes how the online-experiment is conducted, how the data is obtained and how the data will be analysed. A pilot-test will be discussed as well. This section also describes the measurements of all variables and discusses a pre-test, which is executed to be able to decide which kinds of products represented a high or low involvement product.
3.1. Strategy
This study is conducted by using an online-experiment, which was hosted by Qualtrics. Qualtrics is a research software company that enables online data collection in a structured and easy manner.
An experiment in general is chosen, because of the opportunity to show stimuli and to investigate causal inferences (‘T Hart, Boeije & Hox, 2009). The experiment, which is executed in this thesis, can be described as a ‘lab-experiment’. A lab-experiment is chosen because of the high level of control over variables and research context, the easiness to investigate causal relationships and to be able to randomly allocate participants into different conditions. This positively contributes to the high internal validity of this study (‘T Hart et al., 2009). On the other hand, lab-experiments score usually low on external validity. This is due to the artificial setting which negatively influences the extent to which the results can be generalized to other situations (‘T Hart et al., 2009).
This problem is partially be solved by using an online-experiment. The participants filled out the survey in a real setting instead of an artificial setting, so at different
places and various time slots. In addition to this, an online-experiment is conducted, because of limitations with regard to money and time. An online-experiment makes it possible to reach a high number of participants in a very limited time period. Now, people were able participate, at any place and at any time, and it took approximately 2-3 minutes of their time. Due to this easy way of participating it is accomplished to reach a high number of participants.
3.2. Design
The experiment had a 2x2 between-subjects design. People were assigned to one of the four conditions on a random basis, so each participant went through only one treatment. Random allocation controls for extraneous variables, which makes this experiment a ‘true experiment’ (‘T Hart et al., 2009). This experiment had a between-subjects design instead of a within-between-subjects design, because of the risk of interferences between conditions. In addition, the chance that participants discover the research goal would be much higher if a within-subjects design was used. The internal validity will be negatively influenced if people recognized the research goal while participating (Schreuder Peters, 2011). Differences in terms of demographic variables between participants are solved by randomly allocating these participants into one of the four conditions. By using this way of allocation, it can already be assumed that all four conditions are quite similar to each other. To confirm this, these conditions were compared on the basis of how they were distributed in terms of gender, educational level and age. As expected, all conditions consisted of a similar set of participants.
Table 1. Experimental design
Narrative content Non-narrative content High involvement Condition 1 (40) Condition 2 (35)
Low involvement Condition 3 (41) Condition 4 (43)
3.3. Sample
At least 160 participants were needed to be able to draw conclusions (de Boer & Van Groningen, 2012). Either a man or woman who is older than 18 years, because of ethical policies, was a potential participant. Social media is used to reach out for these people. This approach is known as the ‘snowball method’ (‘T Hart et al., 2009). Multiple channels, such as Facebook and LinkedIn, are used to persuade people to fill out the questionnaire. Participants of the experiment were also asked to share the link of the experiment with their own social network. On the one hand to be able to reach a high number of people, but on the other hand to reach a more varied group of people instead of by solely focusing on the social network of the author. It took a few weeks before 160 people were willing to participate on a freely basis. In the end it is accomplished by constantly asking people to fill out the questionnaire in an online as well as offline environment.
People who participated in the pre-test were excluded from the main study. If people already participated in the pre-test, they could be aware of the research goal, which can threaten the internal validity of the main study (Schreuder Peters, 2011).
The initial sample consisted of 227 participants. However, 62 people did not completely filled out the questionnaire so these people were removed from the dataset
due to missing data. The data of 165 participants was usable. As elaborated later on, six participants were eliminated while recoding the numerical variable ‘Level of scepticism’ into a categorical variable with two levels. The final sample consisted of 159 people.
Table 2. Sample size
Gender N=
Male 35
Female 124
Total 159
3.4. Procedure
People could participate by using the Qualtrics link of the online-experiment, which was shared online. First of all, they were exposed to an introduction in which the whole procedure was outlined, in which they were asked to read all questions carefully and in which the contact details of the researcher were presented in case of questions, suggestions or complaints. After the introduction, participants were randomly allocated to one of the four conditions. Four different advertisements were created and functioned as stimuli. They were exposed to narrative content of a high or a low involvement product or participants were exposed to non-narrative content of a high or low involvement product.
The participants of all conditions received the same survey after the stimulus. This survey started with one single question, an item of the latent construct ‘Level of scepticism’. After this question, all questions concerning brand attitude were
presented, followed by the other questions, which were inserted to measure the level of scepticism. After these questions, the extent to which the participants were already familiar with the brand of the advertised product was measured. Lastly, the questions with regard to demographic variables were presented to the participants. These questions were asked in the final part of the survey to higher the chance that people were willing to reveal this kind of personal information. On the last page, the participants were asked for the favour to share the link with their own social network and is communicated that it is really appreciated that the person was willing to dedicate time to this research project. The online-experiment, including the introduction page, stimuli, questions and an expression of gratitude, is exported from Qualtrics and included as appendix.
3.5. Pilot-test
Prior to the online-experiment, a pilot-test is executed. The main goal of this pilot-test was to check if all questions and instructions were clear and if not, to get insights in which questions or instructions needed to be clarified. A small group of 10 closely related people, who where not familiar with this field of research neither with conducting research, took a look at the online-experiment and were asked to give as many feedback as possible. Based upon this feedback, several questions and instructions were rewritten to make it more understandable.
3.6. Stimuli
Four different advertisements were created. Two of these show narrative content, the first one in combination with a laptop and the other one promotes a ballpoint pen. Two other advertisements communicate the same information in a non-narrative
format. One of these shows a laptop and the other advertisement promotes a ballpoint pen as well. Based upon a pre-test it is concluded that a laptop represents a high involvement product and a ballpoint pen a low involvement product. To emphasize the difference between the levels of involvement, the consumer prices were presented on the advertisements. To be able to measure the variables ‘Content type’ and ‘Level of product involvement’ in a validated way, all four advertisements are structured in the same way and similar backgrounds and colours are used. To distract participants from the research variables and to try to make the advertisements as believable as possible ‘NO.1’ is added on all ads. Because of the high importance of the brand name due to the dependent variable ‘Brand attitude’, the brand name is described in the content itself and also displayed in the right corner of the advertisement. VAIO is not a very popular and well-known brand, because of this reason VAIO has been used as brand for the advertisements that promote the laptop. The same kind of reasoning is used while choosing a writing instruments brand. It is expected that the majority of the participants are unfamiliar with the brand LAMY. To be able to control for this assumption the level of brand awareness is measured. This is explained in more detail in the upcoming part.
Figure 2. Final design of stimuli
Narrative, high involvement Non-narrative, high involvement
Narrative, low involvement Non-narrative, low involvement
3.7. Measurements
‘Content type’ is operationalized in two different content forms: narrative content and non-narrative content. Narrative content conveys the message by telling a story (Lien & Chen, 2013). The operationalization of narrative content and non-narrative content is based upon the procedures outlined by previous researches (Lien & Chen, 2013) (Peracchio & Meyers-Levy, 1997) (Wentzel, Tomczak & Herrmann, 2010) (Polyorat, Alden & Kim, 2007) (Escalas, 2004b). The narrative advertisements show the personal evaluation of someone who consumes the product and communicate product
features and benefits in a conversation-like story (Escalas, 2004b) (Peracchio & Meyers-Levy, 1997). The non-narrative advertisements convey the same product features and benefits, but now in a factual and more straightforward manner (Peracchio & Meyers-Levy, 1997). This content type cannot be interpreted as a story, because all characteristics of a story lacked (Escalas, 2004a) (Escalas, 2004b). Regarding the internal validity, the benefits/product features are in both kinds of advertisements presented in the same sequential order and have approximately the same length. Content can be communicated verbally or visually (Lien & Chen, 2013). It is chosen to communicate both types of content in a verbal way. In this case it is easier to control for extraneous variables between all conditions, to make sure that the content type alone varied between conditions.
‘Level of product involvement’ is also operationalized in two different types: low product involvement and high product involvement. Both types of products are manipulated in the four different ads. To decide which product represents a high or low involvement product, a pre-test is conducted. The procedure and results of the pre-test will be discussed in an upcoming part. Based upon the results of this test we were able to conclude that a ballpoint pen represents a low involvement product and a laptop represents a high involvement product. These two different products are used in this experiment to be able to measure the moderating role of the level of product involvement on the relationship between content type and brand attitude.
To measure the dependent variable ‘Brand attitude’ a validated measurement scale is used (Campbell, 1995). Four 7-point semantic differential scales (bad/good,
unpleasant/pleasant, low quality/high quality, dislikeable/likable) measured the attitude towards the brand in all four conditions.
The moderator ‘Level of scepticism’ is measured by using a reliable and validated scale (Campbell, 1995). This scale consists of 5 questions, measured on a 7 points scale (1=strongly agree, 7= strongly disagree) and one question, measured on a 7-point semantic differential scale (fair/unfair). The following six statements are presented to the participants in all four conditions: ‘The way this ad tries to persuade people seems acceptable to me.’ ‘The advertiser tries to manipulate the audience in ways that I don’t like.’ ‘I am annoyed by this ad because the advertiser seemed to be trying to inappropriately manage or control the consumer audience.’ ‘I don’t mind this ad: the advertiser tries to be persuasive without being excessively manipulative.’ ‘This ad is fair in what was said and shown.’ ‘I think that this advertisement is unfair/fair.’
3.8. Control variables
Taken the internal validity of this research into account, it is chosen to use already existing brands in the advertisements instead of fictional brands. VAIO and LAMY are used as brands, because these brands are not very popular and well known. It is of high importance that people do not have an already existing attitude towards one of these brands, because this can interfere with the effect of narrative content and level of product involvement on brand attitude. To make sure that the majority of people are not familiar with the brand of the advertised product, a control question is added in the questionnaire. The level of brand awareness is measured by asking the question: ‘Were you already familiar with the brand of the advertised product?’
Three questions were inserted in the survey to measure the demographic variables: gender, age and educational level. This data provides information about the extent to which the results can be generalized to different groups. The aim was to reach a representative sample group that consists of men and women of different age groups and different educational backgrounds. However, it is expected that especially students were willing to participate in this online-experiment. This can negatively affect the external validity of this research.
3.9. Pre-test
To be able to decide which kinds of products represent a high or low involvement product, a pre-test was accomplished. The pre-test consisted of an online survey, which one is also hosted by Qualtrics. The questions, which are presented in the pre-test survey, were based upon the five facets of involvement (Laurent & Kapferer, 1985). Laurent and Kapferer (1985) stated that it would be impossible to measure ‘involvement’ directly. Because of this reason, based upon a review on empirical research and practices, they indicated five antecedents of the ‘involvement’ construct. These facets of involvement are outlined in table 3.
Table 3. Facets of involvement
Five facets of involvement (Laurent & Kapferer, 1985)
1. The perceived importance of the product (its personal meaning).
2. The perceived risk associated with the product purchase, which in turn has two facets:
-the perceived importance of negative consequences in case of poor choice -the perceived probability of making such a mistake
3. The symbolic or sign value attributed by the consumer to the product, its purchase, or its consumption. This differentiates functional risk from psychosocial risk.
4. The hedonic value of the product, its emotional appeal, its ability to provide pleasure and affect.
To decide which product could function as a low involvement product and which product could function as a high involvement product, in the online-experiment, we decided to measure the level of involvement for four different products. Based upon previous research in the literature, a ballpoint pen and shower gel were considered as low involvement products and it was expected that a laptop and car would have been evaluated as high involvement products (Akbari, 2015) (Johar & Sirgy, 1991) (Laurent & Kapferer, 1985).
The following questions were based upon the five facets of involvement (table 3) and were presented to the participant about each product: ‘To which extent are you interested in this product?’ ‘To which extent does this product bring pleasure to you?’ ‘To which extent do this product symbolize your self-concept?’ ‘To which extent will it be a disaster for you if you make the wrong purchase choice?’ ‘To which extent are
you likely to make the wrong purchase choice?’(Laurent & Kapferer, 1985). The last question is reversed, because people are less likely to make the wrong purchase choice if they purchase a high involvement product, because they are highly involved in this purchase process. All questions are measured on a 7 points scale.
The participants could answer the questions by clicking on the link, which was shared with many acquaintances. After one week, 48 people had the intention to participate, but only 39 participants completely filled out the questionnaire.
While comparing the means and standard deviations of all four products, the results showed that a laptop represents a high involvement product
(Question 1: M= 5.18, SD= 1.19) (Question 2: M= 5.05, SD= 1.65) (Question 3: M= 3.92, SD= 1.91) (Question 4: M= 5.23, SD= 1.83) (Question 5: M=1.92, SD= 1.53) and a ballpoint pen represents a low involvement product (Question 1: M= 1.46, SD= 1.59) (Question 2: M= 1.59, SD= 1.89) (Question 3: M= 1.13, SD= 1.81) (Question 4: M=0.92, SD= 1.55) (Question 5: M=3.08, SD= 2.55). One Sample T-Tests were executed to test if the variations in means were significant. All scores, assigned to the laptop or ballpoint pen, differ significantly at the level of 0.01 (2-tailed). The results are included as appendix.
3.10. Data analysis
As an initial step, raw data is exported from survey software Qualtrics and imported into SPSS, a statistical program that is used in this thesis for descriptive statistics and for running analyses. All unusable variables, such as IP-addresses, were immediately eliminated. The data of participants who did not fully participate are eliminated as
well and all variables were checked for missing data or outliers. Missing data or outliers were not present in the dataset. This is probably due to the ‘forced response’ option, which was active for all questions in Qualtrics.
The raw dataset, exported from Qualtrics, did only provide information about the type of stimulus that was presented to the participant. To be able to run statistical analyses for ‘Content type’ and ‘Type of product involvement’, both nominal variables were created in SPSS. For each participant it is coded if they were exposed to non-narrative (value= 0) or narrative content (value= 1) and if they were exposed to a low involvement (value= 0) or high involvement product (value= 1).
The measurement scale for the latent construct ‘Level of scepticism’ consisted of multiple indicators. In this case, people are more likely to ignore different questions and are more likely to give similar responses to each question. To prevent this, three out of the six questions were counter indicative. Due to this, these questions needed to be recoded before a reliability analysis and principal axis factoring analysis could be conducted. This is executed by using the ‘recode into same variables’ option in SPSS. The scores that participants assigned to these questions gave the impression that they were aware of the counter indicative questions. This is probably due to the fact that several words, which indicated that the questions were formulated in an opposite direction, were highlighted in Qualtrics.
The latent constructs ‘Level of scepticism’ and ‘Brand attitude’ are both measured by using multiple items. Before computing these scales, a reliability analysis and principal axis factoring analysis were conducted for both measurement scales. A reliability analysis tests the internal consistency of the measurements. The Cronbach’s Alpha represents the estimator of the internal consistency, Level of scepticism= .86,
Brand attitude= .91. Both values were above .70, which shows that both scales have a high internal consistency. Based upon these Cronbach’s Alpha’s it can be concluded that both scales are highly reliable. The corrected item-total correlations are all above .30, this shows that all items are highly correlated with the total score of the scale. The Cronbach’s Alpha for both scales could not be improved if an item was removed. As a next step, a principal axis factoring analysis was conducted. The Kaiser-Meyer-Olkin measure verified the sampling adequacy, KMO Level of scepticism= .85, KMO Brand attitude= .84. These values meet the requirement of KMO > .60. Bartlett’s test of sphericity was significant for ‘Level of scepticism’ χ2 (15)= 425.40, p< .000.
Bartlett’s test of sphericity was also significant for ‘Brand attitude’ χ2 (6)= 431.38, p<
.000. This means that the correlations between the items were sufficiently large for PAF for both. The exploratory factor analyses with Varimax rotation showed that both scales were unidimensional. This implies that only one component has an Eigenvalue above 1. The Eigenvalue for ‘Level of scepticism’ is 3.57, which explained 59.6% of the total variance. The analysis for the variable ‘Brand attitude’ reveals an Eigenvalue of 3.14, which explained 78.4% of the total variance. All items that measure the level of scepticism of participants were computed into a new variable, by calculating the mean across all items. The same procedure is executed for the items that measure the attitude towards the brand. ‘Level of scepticism’ as well as ‘Brand attitude’ are normally distributed (skewness between -1.0 and +1.0 and Kurtosis between - .5 and + .5).
Table 4. Descriptive statistics scales Level of scepticism & Brand attitude
Scale variable M SD Minimum Maximum
Level of scepticism 3.12 1.03 1.17 6.17
A factorial Analysis of Variance (ANOVA) test will be conducted in this thesis. It suits the measurement levels of the variables and hypotheses best (de Boer & Van Groningen, 2012). However, ‘Level of scepticism’ needed to be recoded into a categorical variable with two levels (non sceptical vs. sceptical) to be able to use this statistical test. Scores between 1.00-3.99 were coded as non-sceptical, scores between 4.01-7.00 were coded as sceptical. Recoding this variable into a categorical variable will reveal information, which provides a more suitable answer to hypothesis 3.
In sum, an online-experiment, with a between-subjects design, is conducted in which 165 people participated. These people were randomly assigned to one of the four conditions in which they were exposed to one of the four different advertisements. The content type and the level of product involvement were manipulated in these advertisements. A pre-test was executed to be able to conclude that a laptop could function as high involvement product and a ballpoint pen could function as low involvement product. After the stimulus, the survey questions were presented to each participant. The latent constructs ‘Brand attitude’ and ‘Level of scepticism’ were both measured by using multiple indicators. For both constructs, a reliability analysis and principal axis factoring analysis were conducted. Both constructs met the requirements, so both scales could be created. A factorial ANOVA will be conducted to test all hypotheses in once. To do so, ‘Level of scepticism’ is recoded into a categorical variable. The results of the factorial ANOVA are discussed in the upcoming section.
4. Results
In this section, a correlation analysis and a factorial Analysis of Variance (ANOVA) will be discussed. Issues that stand out in the correlation matrix will be elaborated on. An answer will be given to each hypothesis by using the results that stem from the factorial ANOVA analysis.
4.1 Correlational analysis
In order to observe the correlation between all variables, a correlation analysis is executed. The correlations are described in the correlation matrix below. A Point Biserial Correlation coefficient is a suitable estimator if a nominal variable and interval variable are involved. However, this function was not available in SPSS. A Point Biserial Correlation is a special case of Pearson Correlation coefficient. Because of this reason, Pearson is used to estimate the correlation between all variables.
Interestingly, the Pearson Correlation coefficient was significant, at the level of 0.01, for the correlation between ‘Brand attitude’ and ‘Level of scepticism’ (r= -.35). The Person Correlation coefficient for ‘Brand attitude’ and ‘Level of product involvement’ revealed also a significant value (r= -.24). ‘Age’ correlated, significant at the 0.01 level, with ‘Level of scepticism’ (r= .26) and ‘Educational level’ (r= -.21). Even more important, ‘Level of scepticism’ and ‘Content type’ did correlate significant at the level of 0.05 (r= .19). This correlation shows that narrative content is associated with a higher level of scepticism, which is in line with previous research (Escalas, 2007). This is of high interest; it is likely that scepticism can lead to undesired and probably even harmful outcomes.
Table 5. Correlation matrix Variables M SD 1 2 3 4 5 6 7 1.Content type .52 .50 2.Level of product involvement .48 .50 .04 3.Level of scepticism .18 .38 .19* .09 (.863) 4.Brand attitude 4.67 1.23 -.05 -.24** -.35** (.91) 5.Age 25.39 8.73 .03 -.08 .26** -.14 6.Gender .78 .41 -.07 -.08 -.07 .10 -.11 7.Educational level 4.67 .67 -.01 -.10 -.01 .03 -.21** -.02
*=Correlation is significant at the 0.05 level (2-tailed). **= Correlation is significant at the 0.01 level (2-tailed).
4.2. Factorial ANOVA
A factorial ANOVA is used as statistical test in this thesis. An ANOVA determines to which extent the population means of two groups differ significantly (de Boer & van Groningen, 2012). This procedure fits the measurement levels of the variables best. All variables are categorical, except for the numerical dependent variable. This test included the nominal variables: independent variable ‘Content type’ (non-narrative vs. narrative), moderator ‘Level of product involvement’ (low product involvement vs. high product involvement) and the moderator ‘Level of scepticism’ (non-sceptical vs. sceptical). The variable ‘Brand attitude’, which was measured on an interval level, functioned as dependent variable in this test. By using a factorial ANOVA, the main-effects as well as interaction-main-effects could be analysed at the same time. To be able to
provide an answer to hypothesis 2 and hypothesis 3, the interaction-effects between the independent variable ‘Content type’ and each moderator, ‘Level of scepticism’ and ‘Level of product involvement’, were taken into account in the analysis.
The Levene test for equal variances did not yield a significant result, so it can be concluded that all population variances are equal, F (7,151)= .77, p= .617. This result meets the requirement of an ANOVA test.
4.2a. Main-effect hypothesis 1
To test hypothesis 1, the main-effect between the independent variable ‘Content type’ (non-narrative vs. narrative content) and the dependent variable ‘Brand attitude’ is analysed. These results did not show a significant effect of the type of content, non-narrative vs. non-narrative, on the attitude towards the brand, F (1,151) = .56, p= .564. This implies that people who are exposed to non-narrative (M= 4.76, SD= 1.21) or to narrative content (M= 4.62, SD= 1.25) do not differ significantly in terms of their attitude towards the brand. This means that hypothesis 1 ‘People who are exposed to narrative content in an advertisement have a more positive brand attitude than people who are exposed to non-narrative content.’ will be rejected.
4.2b. Interaction-effect hypothesis 2
The interaction-effect between ‘Content type’ (non-narrative vs. narrative content) and ‘Level of product involvement’ (low product involvement vs. high product involvement) on ‘Brand attitude’ is analysed in order to test hypothesis 2. The results did not reveal a significant effect, F (1,151) = .45, p= .504. This suggests that people who are exposed to narrative content of a low involvement product (M= 4.99, SD= 1.10) have a more positive brand attitude compared to people who are exposed to
narrative content of a high involvement product (M= 4.24, SD= 1.30), which is in line with the hypothesis. However, this difference is not significant. Due to this, hypothesis 2 ‘The positive effect of narrative content in an advertisement on brand attitude is higher if a low involvement product is promoted than if a high involvement product is promoted.’ will also be rejected.
4.2c. Interaction-effect hypothesis 3
In order to test hypothesis 3, the interaction-effect between ‘Content type’ (non-narrative content vs. (non-narrative content) and ‘Level of scepticism’ (non-sceptical vs. sceptical) on ‘Brand attitude’ is researched. This analysis did not reveal a significant interaction-effect, F (1,151) = .87, p= .354. People who are exposed to narrative content did not score significantly lower on the attitude towards the brand if they are sceptical about the advertisement (M= 3.90, SD= 1.04) compared to people who are not sceptical about the advertisement (M= 4.86, SD= 1.23). Because of this reason, hypothesis 3 ‘The positive effect of narrative content in an advertisement on brand attitude disappears when people are sceptical about the content.’ will be rejected as well.
Table 6. Results Factorial Analysis of Variance (ANOVA) Source SS df MS F p ηp2 Content type .733 1 .733 .564 .454 .004 Content type*Level of product involvement .585 1 .585 .450 .504 .003 Content type*Level of scepticism 1.126 1 1.126 .866 .354 .006 Error 196.378 151 1.301 Total 3734.000 159
SS= Sum of Squares df= degrees of freedom MS=Mean Square
Taken this all together, a correlation analysis and factorial ANOVA were executed. The correlation analysis did reveal some significant correlations. It showed that narrative content is more associated with a higher level of scepticism rather than non-narrative content. This result is of high importance, concerning the backfire effects that probably occur due to a higher level of scepticism. A factorial ANOVA is conducted to test the hypotheses. The results of the main-effect, the influence of narrative content on brand attitude, were not in line with what was proposed. Moreover, the results were insignificant so this hypothesis was rejected. The results, with regard to the interaction-effects, were in line with the hypotheses. A higher value was dedicated to the attitude towards the brand, if a low involvement product was promoted and people score lower on the brand attitude scale if they expressed themselves as sceptical. However, these hypotheses were rejected as well due to insignificant results. The next section will discuss on these results.
5. Discussion
In this chapter will be elaborated on the results. Key points will be highlighted one more time and general conclusions will be discussed. Above this, managerial and theoretical implications will be presented. As a final part, the shortcomings of this research project will be discussed and directions for future research will be provided.
5.1.Conclusion
Storytelling, an aspect of content marketing, is at the centre of ‘new marketing’ (Pulizzi, 2012). It is a popular phenomenon for business practitioners as well as for academics, looking at the increasing amount of business investments and the volume of academic research, which is dedicated to this subject (Karr, 2016). Narrative content, the academic subscription of storytelling, is seen as the key to success in marketing. This thesis stepped into the gap in the literature by investigating if narrative content always outperforms and if this influence is as easy as suggested or if it is limited and more contingent.
The effect of content type, non-narrative versus narrative content, on the attitude towards the brand is researched. ‘Brand attitude’ functioned as antecedent of the willingness to purchase the product in question. This thesis highlighted the effect of content type on purchase intention, because of the increasing importance of accountability in the marketing environment (Verhoef & Leeflang, 2011). Both constructs are highly correlated (Spears & Singh, 2004). Due to this correlation, the effects on brand attitude have been taken into account in this thesis.
This research project tried to investigate in which cases narrative content does not lead to desired outcomes, or even more important, in which cases it can lead to undesired outcomes, due to a high level of scepticism for example. Escalas (2007) sheds light on the moderating role of the level of scepticism. He suggested that the positive effect of narrative content disappears if people are sceptical about the content due to an analytical way of processing. This relationship is investigated in this thesis. The moderating effect of the level of product involvement, on this relationship, is also taken into account. Earlier, Akbari (2015) found evidence for the differential effect of emotional and rational appeals for low involvement versus high involvement products.
The Elaboration Likelihood Model, a dual process theory that outlines two different information-processing routes, the central and peripheral route, is used to understand the differential effects if narrative content is involved (Petty & Cacioppo, 1983). This theory gave insights in how people process information in different circumstances and how this, in turn, influences the effect of narrative content. In addition to this, the Elaboration Likelihood Model was used to come up with moderating factors that could probably elicit a central route of information processing instead of a peripheral route, which is normally used and more effective in case of narrative content (Lien & Chen, 2013).
An online-experiment is conducted, which is hosted by Qualtrics. Each participant was randomly assigned to one of the four conditions, in which an advertisement was shown. SPSS is used to draw conclusion from the raw dataset. On the basis of a factorial ANOVA an answer could be generated to the hypotheses and research
question: ‘What is the effect of narrative content in ads on brand attitude, as an antecedent of purchase intention, and what is the role of the level of product involvement and level of scepticism on this relationship?’
Table 7. Summary hypotheses
Hypothesis Conclusion
H1: People who are exposed to narrative content in an advertisement have a more positive brand attitude than people who are exposed to non-narrative content.
not supported
H2: The positive effect of narrative content in an advertisement on brand attitude is higher if a low involvement product is promoted than if a high involvement product is promoted.
not supported
H3: The positive effect of narrative content in an advertisement on brand attitude disappears when people are sceptical about the content.
not supported
As can be observed in the table above, each hypothesis is rejected. It was hypothesized that narrative content would lead to a more positive brand attitude, compared to people who were exposed to non-narrative content, as suggested by initial research (Escalas, 2004b) (Van Laer et al., 2014). However, the results are not in line with this direction. People who were exposed to narrative content scored even lower on the ‘Brand attitude’ scale. This indicates that their attitude towards the brand is less positive. Despite of the insignificant results, it still suggests that narrative content does not outperform in each situation, which is in contrast with previous research.
