Blaming the brand : the impact of brand failures on consumers’ trust in the brand

Blaming the Brand:

The impact of brand failures on consumers’ trust in the brand.

Name: Lisa Veen

Student number: 10655476

Date: 27-01-2017

MSc. in Business Administration: Marketing Track

Statement of Originality

This document is written by Lisa Veen 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 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.

Preface

Dear reader,

Before you lies my master thesis ‘Blaming the Brand: the impact of brand failures on consumers’ trust in the brand’. This master thesis has been written to fulfil the graduation requirements of the Business Administration Marketing Program at the University of Amsterdam.

While writing this foreword, I realize that not only will I be completing my Master’s degree programme, but also my days as a student have come to an end. I look back on those years with great happiness. For the past six months I was engaged in researching and writing this thesis, and now I have the honour to present you this paper. This period has not always been easy, though thanks to some great people I was able to stay motivated and to complete this master thesis.

Firstly, I would like to express my sincere gratitude to my supervisor Dr. M. Mossinkoff for his time, and insightful comments. Secondly, I would like to thank all the respondents for their participation. Without their cooperation I would not have been able to conduct this research. Furthermore, I would like to thank my father and friends for their encouragement and kind words during these months. In particular I would like to thank Suzanne Felix. It was always very helpful to debate issues regarding rigorously conducting an experiment with you.

I hope you enjoy your reading!

Yours Sincerely,

Lisa Veen

Table of Contents

ABSTRACT ... 4 LIST OF TABLES AND FIGURES ... 5 1. INTRODUCTION ... 6 1.1 BACKGROUND ... 6 1.2 RESEARCH QUESTION ... 9 1.3 RELEVANCE ... 9 1.3.1. Managerial relevance ... 9 1.3.2. Academic relevance ... 10 1.4. RESEARCH METHOD ... 10 1.5 READING GUIDE ... 11 2. THEORETICAL FRAMEWORK ... 12 2.1 BRAND FAILURES ... 12 2.2 BRAND TRUST ... 14 2.3 BRAND BLAME ... 15 2.4 RELATIONSHIP BETWEEN BRAND FAILURES AND BRAND TRUST ... 17 2.5 CONCEPTUAL MODEL ... 22 3. RESEARCH METHOD ... 23 3.1 PARTICIPANTS ... 23 3.2 DESIGN AND PROCEDURE ... 24 3.2.1 Pre-Test Stimuli ... 24 3.2.2 Post-Test only Experiment ... 25 3.3 MEASUREMENTS ... 27 3.3.1 Back-translation procedure ... 28 3.3.2 Brand Trust ... 28 3.3.3 Brand Blame ... 28 3.3.4 Purchase Intention ... 29 3.4.5 Control Variables ... 29 3.4 STATISTICAL PROCEDURE ... 29 3.5 VALIDITY AND RELIABILITY ... 30 3.5.1 Internal Validity ... 30 3.5.2 External Validity ... 32 3.5.3 Reliability ... 33 4. RESULTS ... 34 4.1 MANIPULATION CHECKS ... 34 4.2 DESCRIPTIVE STATISTICS AND CORRELATION ANALYSIS ... 34 4.3 HYPOTHESES TESTING ... 38 5. DISCUSSION ... 42 5.1 INTERPRETATION RESULTS ... 42 5.2 THEORETICAL AND MANAGERIAL IMPLICATIONS ... 44 6. CONCLUSIONS ... 46 6.1 CONCLUSION ... 46 6.2 LIMITATIONS AND FUTURE RESEARCH ... 47 REFERENCES ... 50 APPENDIX ... 57 APPENDIX A: PRE-TEST ... 57 APPENDIX B: QUESTIONNAIRE ... 60

Abstract

Brand failures can have serious consequences for consumers’ trust in a brand. Yet, focus has been on brand-related product failures, although there might be more brand-entities that cause a failure and negatively impact consumers brand trust. I examine and compare the possible influence of two brand-related failures, that is CEO misconduct and product failure, on brand trust and test whether this relationship is mediated by brand blame. A three between-subjects experimental design was set up to analyse consumers level of trust in a brand after a brand-related failure. A total of N=156 respondents participated in this study. The results show that consumers trust in the brand is only affected when they attributed a certain amount of blame to the brand. This discovered effect is stronger for consumers who were exposed to a product failure. In the product failure condition consumers attribute more blame to the brand and in turn their trust in the brand decreased significantly more than in the CEO condition.

Keywords: Brand Failures; Brand Trust; Brand Blame; CEO misconduct; product failure; Purchase Intention

List of Tables and Figures

Figures

Figure 1: Conceptual Model

Figure 2: Conclusion of tested Conceptual Model

Tables

Table 1: Means, Standard Deviations, Correlations and Reliabilities Table 2: Descriptive Statistics Group 0: Control Condition

Table 3: Descriptive Statistics Group 1: CEO misconduct Table 4: Descriptive Statistics Group 2. Product failure Table 5: Results of Brand Failure as predictor for Brand Trust Table 6: Results of Brand Failure predicting Brand Blame

Table 7: Regression results for Brand Blame as a mediator of the relationship between Brand Failure and Brand Trust.

1. Introduction

1.1 Background

Brand-related failures, such as product failures caused by the brand manufacturer or company misconduct, induce a challenge for marketers, since failures are often hard to control. Also, the nowadays more pressing and knowledgeable customers and the attentive media increase the visibility of the occurrence (Klein & Dawar, 2004). Not only do failures result in financial cost for the company, especially noteworthy, failures can deteriorate consumers’ trust, brand equity and consumers’ intention to buy the brand in the future (Laufer Gillespie, McBride, and Gonzalez, 2005) as such jeopardizing the entire business (Reilly, 1993; Siomkos, 1999).

In September 2015, the American Environmental Protection Agency (EPA) and the California Air Resources Board (CARB) discovered that Volkswagen, one of the world’s largest car manufacturer, is not meeting the federal diesel-emission requirements. The company installed an emission software on over 11 million diesel cars designed to cheat on the emission test (O,Boyle & Adkins, 2015). As a consequence, a recall of Volkswagen’s vehicles took place. The car company suffered a market share drop, a change in consumers’ brand evaluation, and a decline in the willingness to purchase a Volkswagen vehicle (O,Boyle & Adkins, 2015). In August 2016, Samsung had to deal with the consequences of a product failure. The batteries of the Samsung Galaxy Note 7 exploded or spontaneously inflamed (FD, 2016). Samsung replaced 2,5 million damaged models. Unfortunately for the brand manufacture, the same failure occurred with the replaced apparatus. Subsequently Samsung abolished the Galaxy Note 7. As a result of the latter events, Samsung faced a decline in stake- and shareholders level of trust in the brand, a damaged reputation, and a severe market share drop (FD, 2016). These latter examples illustrate the costs for the brand due to those deficiencies, and the need to comprehend consumer behavior subsequent to failure.

Dawar and Pillutla, (2000) dispute that mistakes might start a brand crisis, that is the development and spreading of undesirable brand images to the mass in turn harming trust in a brand. Not by any means can a company or brand avoid crises for long, since crises in companies are considered to be adversities waiting to happen (Fink, 1986). Additionally, crises are considered to be a fundamental part of a brand’s life, and should therefore be studied (Yannopoulou, Koronis, and Elliot, 2011). As several researchers have focussed on a brand-caused product-harm crisis as an antecedent setting to investigate post effects on consumers’ judgements, other settings with brand-relatedness have been a challenge for researchers (Ping, Ishaq, and Li, 2015). Additionally, Song, Sheinin, and Yoon (2016) stress the need to further investigate failures with various types of brand-relatedness initiating a brand crisis, such as misconduct by the Chief Executive officer (CEO), and identify consumers’ reactions to the brand. In particular, to my knowledge no studies have manipulated brand crisis in a way that the post effect of the failures caused by different brand-entities are compared. Though, understanding the possible ambiguous effect is important as choosing an inappropriate response strategy might severely damage the brand (Laufer & Coombs, 2006).

Regardless of the recurrence and severe impact of a brand in crisis, Dawar & Lei (2008), amongst others, argue little scholarly research exists to shed light on the impact of brand crisis on brand trust. Brand trust is vulnerable since it is based on consumers’ judgements of the brand. For that reason, information about events outside the marketing managers control, can cause major unexpected shifts in the level of trust (Yannopoulou et al., 2011). Furthermore, Dawar and Pillutla (2000) argue that the construct brand trust during a brand crisis has not been investigated adequately, since previous researchers have mainly focused on corporate reputation in combination with trust instead of threatening trust as an individual brand effect. Additionally, scrutiny of the construct trust in the consumer-brand

area is lacking (Delgado-Ballester, 2004). Nevertheless, Yannopoulou et al. (2011) investigated brand trust in a product-harm crisis setting, though the authors did not compare consumers’ reactions to various types of brand-related failures, that is to say how different brand failures causing a brand crisis will likely lead to variation in the level of brand trust scores.

As brands are of tremendous economic importance to companies I recognize the need for further investigation of the concept of brand failures on consumer behavior, in which I am specifically interested in the variations in consumers’ brand trust scores, when investigated through different brand failure settings. Also of great value is to understand the extent, to which the variation in the level of trust is explained by brand blame, as consumers’ attitude diminishes when object blame increases, (Weiner, 2000).

Moreover, in this research I explore how consumers’ trust in a brand is affected and how much variation is explained by the blame attributed to the brand during a brand in crisis. In other words, for the purpose of this paper I investigate two brand-related failures initiating a brand crisis. The brand-related failures that will be analysed are CEO misconduct and product failure. Additionally, the context in which the brand crises will be tested is the insurance industry. The insurance industry is considered to be an opportune one for the study of brand trust as their offerings – insurances – are intangible. Consumers have to believe that the company will deliver what they promise. In other words, consumers trust is a requirement for insurance companies and exploring brand failures within this industry will likely lead to apparent changes in the level of brand trust, which in turn increases the opportunity to better comprehend influences on brand trust. To conclude, in this research I attempt to uncover the possible broader scope of brand crisis and its effects on consumer behavior.

1.2 Research Question

The purpose of this research is to explore and identify consumers’ reactions to brand failures. As such, the following research question is formulated:

How will brand failures affect trust in a brand?

1.3 Relevance

1.3.1. Managerial relevance

Failures generate a big challenge for marketing managers. Moreover, when a brand is surrounded by bad news, due to product recall, bad behavior from a brand endorser, not following the rules, and other similar events, this can do severe damage to a brand. Especially since nowadays consumers live in a digital age in which acquiring information from, among other things, social media, online news forums and smartphones, has never been so easy. These latter channels are an opportunity for companies, but also a threat, since information about failures is harder to control, in turn altering consumers’ perception of a brand (Monga & John, 2008). As such, brand failures can have severe consequences for brand trust in turn unfavourably impacting consumers’ willingness to (re)purchase the brand (Dawar & Pillutla, 2000). Therefore, acquiring in-depth knowledge about the impact of various types of brand-related failures on the level of brand trust is of the essence. Also, brand- and marketing managers should understand how much blame the consumer is likely to attribute to the brand, as in cases of high brand blame an effort strategy should be adopted, though in cases of less blame a voluntary recall should be put in motion (Laufer & Coombs, 2006). Therefore, the results of this research may provide a foundation for recovery strategies.

1.3.2. Academic relevance

Previous researchers have focused on identifying the effects of brand-caused product failures initiating a brand crisis on consumer reactions (Ping, Ishaq and Li, 2015; Soon et al., 2016; Yannopoulou et al., 2011). Though the need for further research regarding the impact of various brand-related failures causing a brand crisis on the possible changes in brand trust scores is stressed (Soon et al., 2016). Yet, despite several scholars acknowledge the importance of studying the role of brand trust, there is a lack of empirical studies that explicitly investigated the trust concept in the ‘consumer-brand’ domain (Delgado-Ballester, 2004). Therefore, acquiring in-depth knowledge about the impact of various types of brand-related failures on the level of brand trust is of the essence. In particular, brand failure has not been manipulated to the extent that the effects of various brand-related failures on consumers’ brand trust are compared. Even though, comprehension of the possible variation in effect is important as an inadequate response strategy can seriously harm a brand (Laufer & Coombs, 2006). By means of this research I try to extend and add knowledge in the field of marketing and consumer behavior.

1.4. Research Method

By means of an experiment I researched consumers’ perceived level of brand trust post brand failures, in order to formulate a rigorous answer to the research question. An essential advantage of conducting an experiment is the ability to draw causal inferences and to examine underlying mechanisms between brand failures and consumer reactions. The design of the experiment is a three between-subjects design, that is, brand failure: CEO misconduct and product failure, and a control condition. Respondents were randomly assigned to one of the three groups. Data was collected and analysed via multiple regressions. Altogether 156 respondents participated in this experiment.

1.5 Reading Guide

This thesis encompasses six chapters, whereof the first covered the introduction of the research topic. In the second chapter I discuss relevant scholarly literature in order to develop a theoretical framework. Chapter three covers the methodology of this research and in chapter four I report the results of the analysed data. In the last two chapters I draw conclusions based on the results, discuss the limitations and give directions for future research.

2. Theoretical Framework

This chapter is concerned with reviewing relevant literature about brand failures and consumer behavior, and addresses the expectations of this research. In the beginning the key concepts of brand failures is outlined. Subsequently the theoretical foundations of brand trust in relation to brand-related failures are discussed, followed by analysing relevant literature about the role of brand blame. Based on this literature review a conceptual model is proposed and hypotheses are formulated.

2.1 Brand Failures

“The good news about brands is that people know who you are. The bad news is that if something goes wrong, everyone knows” (Knowledge@Wharton, 2005)

The concept ‘branding’ was, amongst others, designed to protect products from failure. In the 19th century, companies like Heinz, and Quaker Oats were even more concerned about consumer’s judgments towards mass-produced products. These companies started to develop brand identities not solely to differentiate their products, but to encourage consumers’ confidence in the idea of factory-produced goods (Haig, 2005). Due to branding, consumers could now place trust in the brand itself instead of placing trust in, for example, the merchant.

Nowadays, in the 21st century, it is the ‘brand’ that faces difficulties. If a product fails, the brand is likely to receive the blame. In other words, if mistakes are made the brand is at fault (Haig, 2005). Additionally, when failures involving brands occur it creates media and consumer awareness and sensitivity to such failures. As such, a brand failure cannot be perceived as an isolated incident, as it is rather an opportunity for the media to spread the failure, in turn initiating a brand crisis (Yannopoulou et al., 2001).

Brand crises arise out of scandals or factual negative information about a brand. As a consequence, consumers’ trust and confidence in a brand can be damaged (Dawar & Lei, 2009). The damaging effects for the brand are nowadays possibly reinforced due to the digital era in which failures are more visible. As such, new media, like social network sites, has caused the consumer to be more influential and opinionated, as new media increases the ability to share information. In conclusion, due to the Internet and other technologies consumers cannot be led into an erroneous conclusion and perceptions are altered in turn damaging the brand (Yannopoulou et al., 2011).

Previous researchers have focused on analysing brand-caused product failures causing a brand crisis, that is a product failure caused by the brand manufacturer, and scrutinized for consumer’s response to companies’ recovery attempts and market position after a failure was well publicized (Hes, 2008). Though, rarely studied are brand-related failures that are less likely to be controlled by the brand, like negative news about the wrongdoing of a CEO and the effect on consumer behavior (Song et al., 2016). However, past research did investigate and compare brand-caused product failures with consumer-caused failures (Folkes, 1984; Hess, 2008; Yoon, 2013; Song et al., 2016). Nevertheless, those studies did not look at brand-failures in the broadest sense of the word, which is, investigating and comparing brand-failures caused by various types of brand-related entities, who can severely damage consumers’ trust in a brand. Though gaining more in-depth knowledge about brand-related failures is of the essence, as consumers demonstrate to have strong equity-related reactions to brand failures like, complaints, anger, longing for revenge, appeal for an apology and demand for compensations (Folkes, 1984).

To conclude, it is hard for brands to avoid any form of failure as failures are considered to be an indispensable part of a brand’s life (Yannopoulou et al., 2011). Furthermore, these failures are nowadays probably reinforced due to the Internet and other

technology. Therefore, understanding consumer behavior and in turn adequately knowing how to handle such a failure is of the essence. Especially, since failures can cause sudden and long-lasting change in consumers’ perceptions about a brand.

2.2 Brand Trust

The concept of trust is at an early stage of understanding within marketing and consumer research compared to other fields of study that is, sociology, psychology and philosophy (Yannopoulou et al., 2011). One of the reasons for a moderate understanding might derive from the difficulty in conceptualizing trust as it is built on both cognitive and affective foundations (Gurviez & Korchia, 2003), consequently resulting in an ambiguous theoretical stance on trust. Alternatively, trust is not sufficiently comprehended because the brand is intangible (Delgado-Ballester, 2004).

As researchers have not reached a consensus on the concept of trust, several ideas are established within the marketing research. For Luhmann (1979) trust is a function of experience and high-perceived risk. In other words, in order for trust to evolve, history as a reliable framework is needed (McAllister, 1995). In this manner, the development of brand trust is a result of consumers increased brand knowledge due to repeated exposure and interaction with a brand. On the other hand, to become apparent trust needs conditions of high-perceived risk. More specifically, the trusting parties have to be vulnerable in order for trust in brands to be developed (Elliot & Yannopoulou, 2007). However, others define brand trust as customers believe in the brand’s promise and the ability of the brand to live up to that promise by performing its function as expected by the customer (Chaudhuri & Holbrook, 2001; Ha & Perks, 2005, Jin, Line and Merkebu, 2015). In contrast, Delgado-Balleseter (2004) imply brand trust is concerned with two dimensions: brand reliability, which is concerned with consumers’ belief that the brand realizes its value promise, and brand

intentions, which is based on consumers’ belief that the brand is not going to take advantage of the consumer, hence, the brand puts the consumers interest first instead of its own interest when a failure occurs. This latter dimension comprises benevolence and honesty (i.e. cognitive and emotional) (Larzelere & Huston, 1980). Accordingly, Gurviez and Korchia (2003) underscore that the breakdown of the concept trust remains unsettled, and conclude that within the marketing research three acceptations of trust can be found: one-dimensional, two-dimensional, and three-dimensional.

Yannopoulou et al, (2011) emphasize the delicate nature of brand trust as it is based on consumers’ judgements. As a consequence, brand trust is vulnerable since factors, outside of the marketing management’s control, such as consumers being exposed to negative publicity can cause extensive and unforeseen variations in the level of brand trust. Accordingly, brand brand failures can have considerable consequences for brand trust (Dawar and Pillutla, 2000).

2.3 Brand Blame

Since an inadequate response to a brand failure strategy can seriously harm a brand, understanding the impact of different brand-related failures on marketing constructs like ‘brand blame’ is of the essence. More specifically, how do people explain causes for occurrences by for instance attributing blame to an actor after an incident (Weiner, 2000). So, what happens when the CEO is acting in a manner that undermines the consumers’ perception of brand? How much blame are consumers likely to attribute to the brand after a brand failure, and what will be the impact on consumer’s trust in the brand? Also importantly, to what extent does the impact of the latter event differ from the impact of a product failure? To adequately understand perceptual implications of failure, it is important to comprehend consumers’ attribution of blame when a failure occurs (Song et al., 2016). In this particular

case, understanding possible mechanisms that cause consumers to blame the brand for a failure.

In ambiguous settings, such as brand failures, the most common cognitive process of consumers is ‘attribution’ (Folkes, 1984). Consumers tend to unconsciously construct blame attributions when encountered with harmful or defective products (Folkes, 1984; Folkes & Kotsos, 1986). According’s Weiner (2000) attribution model there are three determinants of attributions that lead to an overall assessment of blame: (1) Locus of causality, that is the event that triggered the failure; (2) Stability, which can be enduring or temporary; and (3) Controllability of the failure, that is control within the actor or outside the actor. As such, if the failure is perceived as being internal, stable and controllable, consumers tend to blame the actual actor, though if the failure is perceived as being external, temporary and uncontrollable consumers are likely to blame external factors such as the brand (Folkes, 1984). Subsequently, the attribution theory deals with the concept of explanation, namely why a particular event, state or outcome happens, and the consequences of the causality (Weiner, 2000). Several researchers applied Weiner’s model in failure contexts, and found that consumers’ attribution of blame to the perceived cause altered their affect and attitude towards the company (Folkes & Kostos, 1986; Jorgensen, 1994). Though important to note, the objective of this study is to investigate the possible mediating effect of brand blame on the relationship of brand failure and brand trust. Therefore, the focus is not on ‘how’ consumers arrive at blame attributions, hence assessing the three causal dimensions of attribution (Folkes, 1988; Weiner, 2000), but on ‘the role’ of consumer’s brand blame attribution on consumer’s reactions to brand failures.

Another cognitive process that shape consumers’ perception about a brand is the ‘halo effect’. A halo effect is “the bias due to a measure that spills over to another measure (Klein & Dawar, 2004, p. 204). In other words, consumers use global evaluations, like overall

attitude of a brand, to make judgements about specific attributes. For example, Volvo’s performance might spill over onto consumers’ judgment about its safety. However, the halo effect might also backfire as consumers (negative) overall attitude of a brand might spill over to their judgments about other brand attributes (i.e. characteristics that define the brand) (Beckwith & Lehmann, 1975).

2.4 Relationship between Brand Failures and Brand Trust

Based on the reviewed literature, assumptions about a possible relationship between brand failures and brand trust arise.

Failures are considered to be a fundamental part of a brand’s life and are therefore hard to avoid. When a Chief Executive Officer of a brand is associated with negative publicity he or she will most likely hurt the brand and cause a brand crisis. Larcker and Tayan (2016) discovered a significant and long-lasting negative impact of CEO’s misconduct on company reputation. Even though the company might claim to be genuine post-crisis, imprudent decisions of a CEO severely damage the brand (Palmer, 2008). Furthermore, Forest (2011) argues that some CEO identities are inseparably intertwined with the reputation of the company (e.g. Steve Jobs with Apple). Since a CEO is often perceived as the face of an enterprise (Larcker & Tayan, 2016) and acknowledged to be an important asset to convey a brand-consistent message (Rehmeta & Dinnie, 2013), I argue that most of the CEOs can be considered to be a brand spokespersons, if not ‘the brand’, though in both cases strongly associated with their companies’ brands, and therefore, brand related entities. Due to this interconnectedness, I expect consumers will attribute a certain amount of blame to the brand instead of solely blaming the CEO for the failure (Song et al., 2016). Though the latter presumption about CEO’s brand-relatedness will be pre-tested before conducting the experiment.

The product-harm crisis setting deemed to be an opportune setting to study consumer perceptions about the brand, because a product-harm crisis causes often large and sudden change in consumers’ beliefs about the brand (Klein & Dawar, 2004). Consequently, researchers can study and identify influences on consumers’ brand impressions, which is important, as these altered perceptions can be long-lasting. Additionally, the possible severe consequences of product failures strongly contrast with the relative lack of research in this particular area (Klein & Dawar, 2004; Song et al., 2016). Based on the aforementioned, I choose product failure as the second brand-related failure to evaluate its influence on brand constructs like brand blame and brand trust. Some researchers explicitly draw a clear distinction between product failures and product-harm crisis (Dawar & Lei, 2009; Laufer Gillespie, McBride, and Gonzalez, 2005) though I do not try to draw a dichotomy between these two concepts. Rather, I consider both as a continuum on which product failures almost always automatically result is a product-harm crisis. Therefore, I will use both terms simultaneously.

For the objective of my research I utilize a three-dimensional perspective of brand trust to adequately comprehend the concept of trust and its fragility during a brand in crisis. Accordingly I adopt Gurviez and Korchia’s (2003) definition of brand trust: “ a psychological variable mirroring set of accumulated presumptions involving credibility, integrity and benevolence that a consumer attributes to the brand” (p.3). Credibility attributed to the brand is based on the ability of the brand to meet consumers’ functional expectations, hence, meet the terms of the expected performance. Integrity is concerned with the honesty of the brand regarding its promises. Lastly, benevolence is about placing consumers interest ahead of its self-interest. A benevolent brand decreases the perceived risk for consumers as the brands offering is based on a fair exchange. To conclude, my overall approach to brand trust follows

Gurviez and Korchia (2003) definition in order to adequately scrutinize the impact of brand failures on brand trust. Especially since trust is considered to be delicate and have

Luhmann (1979) claims trust is dogmatically unstable, as the first disappointment often results in the downfall of trust and consequently the consumer-brand relationship changes as trust diminishes or is eliminated. However, when a consumer buys a defective new product, it most likely reduces the brands credibility, though it does not harm the brands integrity. In other words, trust in a brands technical performance may differ despite of trust in the honesty and goodwill of the brand. Nevertheless, product failure results in a declined overall level of consumer’s brand trust (Gurviez and Korchia, 2003).

The Internet and other technology has possibly lead to the empowerment of the consumer, since they can communicate and share their opinions easily, and increase the visibility of brand failures. Therefore, nowadays the impact of brand failure might be even more harmful resulting in higher levels of variation in brand trust. Gurviez and Korchia, (2003) postulate that when consumers become aware of the failure via media, consumers reshape their beliefs regarding the failure and consequently their brand perception and brand trust are altered. In addition, occurrences of lower than average levels of trust turned out to be a result of extensive negative publicity about failures (Romaniuk & Bogomolova, 2005). Also, Yannopoulou et al. (2011) discovered a significant relationship between the broadcast of brand failures and its direct negative affect on brand trust. Accordingly, I hypothesize that:

H1: There is a negative relationship between Brand Failure and Brand Trust

H1a: Consumers’ level of trust will be lower in the product-failure setting than in the CEO

As mentioned before, the halo effect is a cognitive process that shapes consumers’ perception about a brand due to spill-over of a biased measure to another specific measure (Klein & Dawar, 2004). For the purpose of this paper I used a fictitious brand and a non-existing CEO to measure consumers’ reactions. Therefore, the halo effect might not be apparent enough in order to assess the impact of the effect after a brand failure. Weiners (2000) theory of attribution might be the best theory to comprehend why consumers are likely to blame the brand for a failure instead of for instance the CEO. Though, as mentioned before, I am interested in understanding the explanatory nature of consumers’ brand blame attribution, not the possible influence of the antecedents of attribution, as this is out of scope. Moreover, my particular interest lies in understanding how much variation of brand trust following brand failure, is explained by brand blame. Song et al., (2016) validated that brand blame mediated the relationship between product failure and overall brand evaluations. Therefore I assume that:

H2: The negative relationship between Brand Failure and Brand Trust is mediated by Brand

Blame

H2a: There is a negative relationship between Brand Failure and Brand Blame H2b: There is a positive relationship between Brand Blame and Brand Trust

When consumers perceived that the brand manufacturer could possibly control the failure, there reactions about a brand were most negative (Choi & Mattila, 2008). For that reason I expect that unethical behavior by a CEO is less harmful for the brand, hence consumer attribute less blame, compared to a product failure as I expect that consumers perceive a brand-caused product failure as being more controlled by the brand manufacturer. Based on the above I hypothesize:

H2c: This mediating effect is higher for product failure than CEO misconduct

Some researchers discovered that the level of blame attributed to the brand differ depending on the perceived level of severity of the outcome, since the difference in the level of severity incites divergent reactions (Chang, Tsai, Wong, Wang, and Cho, 2015; Song et al., 2016). However, other researchers validated that consumers also showed strong brand-blame attribution and devalued the brand in cases of less severe failures (Bitner, 1990; Folkes, 1984). An increase in the degree of outcome severity of the brand failure should not per se increase the level of impact on consumers’ perception about a brand. Therefore, marketing practitioners should develop strategies that address all brand-related failures irrespectively of the level of severity. To stay within the scope of my research severity will function as a control variable, hence, is held constant.

Judgements, like blaming the brand, have a domino effect on brand evaluations and consumers buying intentions. Therefore, attributing blame to the brand can have long-lasting effects on consumer behavior (Klein & Dawar, 2004). Consumers purchase intention serves as a validation for the ‘domino effect’ of brand blame on brand trust. In other words, purchase intention is not the focal dependent variable, but is intended to validate the knock-on effects of brand blame. Additionally, as brand failures have long-lasting consequences for consumer’s brand evaluations, and it is known that brand evaluations impact consumer’s buying intention, a positive relationship between brand trust and purchase intention is expected.

H3a: Purchase Intention will be lower in the product failure setting than in the CEO

misconduct setting

2.5 Conceptual Model

The following model illustrates the assumptions of this research:

Figure 1. Conceptual model

Brand Failure: -  CEO misconduct -  Product Failure Brand Trust Brand Blame Purchase Intention H1 - H2a + H3 + H2 H2b-

3. Research Method

In this section I discuss the empirical set up for the research and how data was collected. The first part covers the characteristics of the assembled sample. Subsequently, the pre-test of the used stimuli, the experimental design and the procedures are discussed. In the third paragraph I reviewed the measures of the variables relevant for this research. Next, I outlined the statistical procedure for testing the hypothesized relationships. Lastly, the validity and the reliability of this research are discussed. A complete overview of the pre-test and the questionnaires can be found in the appendix.

3.1 Participants

The target population was the Dutch consumer. The rationale for selecting solely the Dutch consumer, instead of every consumer worldwide, is to control for – in advance – other factors that might influence the generalizability of the results such as cultural differences and the possible differences in code of conduct, per country, of the insurance businesses. A probability sample was used in order to give every consumer a known change of being selected. The respondents were approached via e-mails and by posting the link of the study on several social media platforms.

A total of 156 respondents completed the questionnaire of which 55% were female and 45% were male. Furthermore, the age of the participants ranged from 16 to 61. Taking all the respondents together the average age was M = 29.35, SD = 19.88. A majority of the sample (72,4%) had an academic degree (WO), 14,1% completed University of Professional Education (HBO), 0,6% had an Intermediate Vocational Education (MBO), 9% pre-university (vwo), and 3,2% completed a Senior High programme (havo).

3.2 Design and Procedure

The objective of this research design is to characterize the essential type of evidence needed to answer the research question as unambiguously as possible (De Vaus, 2001). For the purpose of this research, an experiment was set up to gain relevant evidence for answering the research question. The main advantage of conducting research by means of an experiment is the ability to draw causal inferences and to research, in this particular case, the underlying mechanisms between brand failures and consumer behavior. The stimuli material used for this research was pre tested.

3.2.1 Pre-Test Stimuli

The purpose of this pre-test was to test the fictitious newspaper articles for readability and credibility and to assure the CEO is perceived as brand related, as this a premise for this research. Newspaper articles were used to present the information about the context, as people regard newspaper articles – in the broad sense – to be credible as a source of information (Newhagen & Nass, 1989). The pre-test was send to participants in the research population (N= 20). A five-point Likert scale (1= Completely Disagree to 5= Completely Agree) was used to check CEO’s brand-relatedness and the fictitious messages for readability and credibility. As expected the CEO (M = 4,72, SD = 0,59) is considered to be brand related (p<. 00). Furthermore, both newspaper articles; CEO misconduct (M = 4,60, SD = 0,59) and product failure (M = 4,55, SD = 0,76) were perceived as readable (p< .01). And both newspaper articles about CEO misconduct (M = 4.05, SD = 0.69) and product failure (M = 4,15, SD = 0.81) were perceived as credible (p<. 01). At the end of the survey was an opportunity for respondents to leave additional comments. Based on those comments I made some linguistic corrections in the stimuli material.

3.2.2 Post-Test only Experiment

The classic experimental design is the most known experimental design, in which a pre- and a post-test are conducted with a group with intervention (i.e. treatment group) and a group without intervention (i.e. control group). For this research a post-test only experiment was constructed in order to discover the effects of brand failures on brand trust. By using a post-test only experimental design, changes between groups cannot be analysed. Instead, the focus of the analysis is on the post-test differences between groups. Nevertheless, when respondents are randomly allocated to their groups and these groups are large enough, the differences in outcome of a post-test should be the same as the differences in change scores of a pre- and post-test of a classical experiment (De Vaus, 2001).

Furthermore, the experimental design is implemented in a laboratory context. The main advantage of a laboratory context is the insurance that the treatment and control groups are exposed to the same environment, except for the stimuli. Moreover, the laboratory context facilitated the ability to control the setting. The only difference between the groups was the intervention or lack of intervention (De Vaus, 2001).

As mentioned before, this research is designed to test the effects of the independent variable brand failure on the dependent variables brand trust and as second dependent variable purchase intention with brand blame as mediating variable. Moreover, this experiment compares the effect of one brand-related failure, CEO failure and that of another type of brand-related failure, a product failure. Hence, the variable brand failure is manipulated. Participants were exposed to one of the three between-subjects conditions, that is to a newspaper article about CEO failure, a newspaper article about product failure or no news paper article, all with corresponding questionnaires. For the purpose of this study a fictitious company’s brand name was used, namely FDHA. The reason for choosing a fictitious company name was to rule out the possibility of contamination of the manipulation by

respondents holding pre-existing associations about the brand. Furthermore, previous research demonstrated that the effects of the manipulation were dampened because of the use of realistic stimuli (Folkes, 1988; Sparkman & Locander, 1980; Yalch & Yoshida, 1983).

At the beginning, all respondents (i.e. treatment and control groups) read the following statement:

The following information concerns a real and well-known Dutch insurance company. An organization that has more than 16 years of experience in health-, life- and damage insurances. For the purpose of this research we call the company FDHA.

Since I used a fictitious brand, I provided the respondents – after the statement – with information about the company’s brand values and promises, and consumer reviews. By exposing the subjects to this information I created an opportunity for respondents to develop any perception about the brand, hence, obtain brand knowledge. Moreover, respondents needed to form prior beliefs about the brand in order to be able to measure the effect of the failure on the level of brand trust. Accordingly, Kang, Manthiou, Sumarjan, and Tang (2016) argue consumers can form prior brand knowledge via diverse sources such as the Internet, advertising, word-of-mouth, and reviews. After reading the company information, respondents were asked some questions regarding the text to ensure that the participants read the brand values and promises, in turn increasing internal validity.

Subsequent to the introduction, respondents in the CEO failure condition read a newspaper article in which the CEO of insurance company FDHA engaged in bad behavior, that is committing fraud. The CEO had stolen company money, namely 15 million euros and spent that money on a luxurious villa and boat. In order to obtain all that money, the CEO unnecessarily fired employees and customers received less or no compensation when they

filed claims. In the product failure condition, respondents read a newspaper article in which customers of insurance company FDHA feel betrayed as they claim that the car insurance is failing. More specifically, most insured did not receive any financial compensation when they had filed a claim about their damaged car. Furthermore, respondents read that in the past FDHA had dealt with a related scandal and the company had to rewrite their policy conditions. Additionally, the respondents, assigned to one of the two treatment groups were asked to finish a choice task, after reading the newspaper article. The multiple-choice task was set up to check if the manipulation worked as intended, that is to check whether the subjects read the newspaper articles.

In the control condition respondents were not exposed to a newspaper article. Subsequently, subjects were asked to fill in a questionnaire that included questions regarding brand blame, brand trust, purchase intentions, and demographics. All questions were asked in the aforementioned order. The items used for the questionnaire were coherent with items of previous research on blame attributions and brands (Gurviez & Korchia, 2003; Klein & Dawar, 2004; Moon, Chadee, and Tikoo, 2004; Song, Sheinin, and Yoon, 2016). After finalizing the questionnaires, respondents could read additional information regarding this research. Furthermore, subjects were informed that they participated in an experiment and that the company and failure scenarios were fictitious to take ethics into account.

3.3 Measurements

Items from existing measurement scales were selected for the questionnaire. In order to measure the variables rigorously scales with a high (α>.8) Cronbach’s Alpha were selected, as the alpha indicated the reliability of the scale. All items were retrieved from English studies and needed to be translated to Dutch as the research sample is the Dutch consumer. Consequently, a back-translations procedure is of the essence.

3.3.1 Back-translation procedure

The items used in the pre-test and the actual questionnaire contained items extracted from English measurement scales. Since the target population is the Dutch consumer, I translated the English items to Dutch. To ascertain that the content remains unaffected, two persons back-translated the items. Discrepancies were corrected.

3.3.2 Brand Trust

To measure the level of perceived trust in the brand, a scale of Gurviez and Korchia (2003) was adopted (Cronbach’s α=0.89). According to the authors the concept trust consists of three dimensions, that is, credibility, integrity and benevolence. The eight items used to measure trust were: Credibility (1) This brand’s products make me feel safe; (2) I trust the quality of this brand’s products; (3) Buying this brand’s products is a guarantee; Integrity (4) This brand is sincere with consumers; (5) This brand is honest with its customers; (6) This brands expresses an interest in its customers; Benevolence (7) I think this brand renews its products to take into account advances in research; (8) I think that this brand is always looking to improve its response to consumer needs. A seven-point Likert scale was used to measure the items (1=Completely Disagree to 7=Completely Agree)

3.3.3 Brand Blame

In order to measure the level of blame attributed to the brand, the scale of Klein and Dawar (2004) was used (Cronbach’s α=0.86). The scale existed of three items: ‘What is FDHA level of responsibility for the CEO/Product failure?’, ‘Should FDHA be held accountable for the CEO/Product failure?’, and ‘This incident is the fault of FDHA’. All items were answered by means of a seven-point Likert scale (1=Completely Disagree to 7=Completely Agree).

3.3.4 Purchase Intention

Purchase intention was measured by means of three items of the scale from Moon, Chadee and Tikoo (2004). Those three items were: ‘I will purchase the FDHA car insurance’, ‘Given a choice, my friends will choose the FDHA car insurance’, and ‘I would like to recommend a FDHA car insurance’ (Cronbach’s α= 0.86). All items were measured via a seven-point Likert scale (1= Completely Disagree to 7= Completely Agree).

3.4.5 Control Variables

In this research the results were controlled for three control variables, that is, Gender, Age, Education and Severity. Questions about these variables were included at the end of the survey.

3.4 Statistical Procedure

Data was assembled by means of an online experiment with survey. The questionnaire was available at the 19th of December 2016, and was closed at January 3th 2017. Respondents were randomly assigned to one of the three between-subjects conditions. To analyse the collected data I used the program ‘Statistical Package for the Social Science (SPSS). The first step was to clean the dataset and check for missing values to prepare the data for the actual analysis. Frequency tables were established to screen for any errors in data entry. The discovered errors were corrected. Missing values were deleted listwise and as a result only the conditions without any missing data in any variable were analysed (N=156). The variables gender (0=female, 1=male) and brand failure (1=CEO, 2= product) were recoded into dummy variables. After cleaning the data set, scale reliabilities, descriptive statistics and normality checks were performed. All Cronbachs Alphas of the scales brand blame, brand trust and

purchase intention were relatively high (see Table 1). Also, normality checks (i.e. Skweness and Kurtosis) were conducted. All variables were sufficiently normally distributed therefore parametric tests could be performed.

Manipulation checks were computed to test whether the manipulation was effective. The manipulation was assessed by means of multiple one-way ANOVA’s and a multiple-choice task. After the checks, a correlation matrix was constructed to assess the linear associations between the variables. Descriptive statics of the three conditions (i.e. CEO misconduct, product failure and control) was presented and two planned comparison analyses, to compare the means between the groups, were undertaken. Hypotheses were tested by means of hierarchical regressions.

3.5 Validity and Reliability

3.5.1 Internal Validity

The occurrence of internal validity is when solely the independent variable influences the dependent variable. Any alternative variable, known as confounding variables, influencing the dependent variable is a threat for internal validity. In other words, by means of an experimental design, researchers should strengthen the logical rigor of a causal explanation, by getting rid of alternative explanations for an association between the treatment and dependent variable (De Bunt & Nencel, 2012).

There are several factors that could have been a threat to internal validity. One of them is the occurrence of a maturation effect. This effect is a result of natural processes like boredom, irritation and sleepiness, which arises during the experiment and influences the dependent variable (De Vaus, 2001). The latter problem however most often occurs in experiments that take hours to complete, which is not the case for this research. Based on the

findings of pilot tests, this experiment on brand failures took about 5-10 minutes to be completed (i.e. depends on the group the respondents were assigned to).

Secondly experimental mortality might have endangered internal validity as this mortality arises when some respondents fail to complete the entire experiment (De Bunt & Nencel, 2012). A total of 203 respondents started the questionnaire however 156 completed the surveys. A total of 70 respondents were randomly assigned to the CEO failure condition. Though after the newspaper article 56 subjects proceeded and completed the questionnaire, hence 14 respondents dropped out. The same event occurred in the product failure condition. In this condition a total of 65 respondents started the experiment though nine respondents dropped out after the newspaper article. The rest, hence 56, of the respondents finished the survey. In the control condition 60 respondents started the questionnaire however after the introductory text 15 respondents dropped out and the remaining 44 completed the task.

A final issue concerning internal validity is selection bias. The latter can occur when an experiment has more than one group and the groups are significantly different at the beginning of the experiment. For the purpose of this research a three-group experiment was designed and respondents were by means of randomization assigned to one of the three groups. Multiple Pearson chi-square tests demonstrated there were no significance differences between the distribution of each group based on gender χ2 (2) = .59, p>.05, age χ2 (68) = 70,29 p>.05, and education χ2 _{(10) = 9,75, p>.05. Hence, all groups were equivalent in the}

beginning.

To increase internal validity, a pre-test of the stimuli was conducted. By means of the pre-test I tested whether participants found the fictitious messages credible and readable. Additionally, respondents had the opportunity to leave additional comments regarding the experiment and based on those comments I made some linguistic corrections. Also, after

finishing the experimental set-up I ran a pilot test in order to understand the time necessary to complete the task and to fine-tune the experiment.

Lastly, after being exposed to the information about the company’s brand values and promises, the respondents were asked some questions to assure the respondents read the information, in turn increasing internal validity.

3.5.2 External Validity

The external validity is the ability of generalizing experimental findings beyond a specific study. A main issue of external validity is the concerns about the possibility to generalize from participants in the experiment to an entire population.

First off, the response rate of this study was 76.8 %. This number is considered to be high based on how the survey was administered (i.e. Email and social media platforms) and on the purpose of this research (i.e. to measure effects), in turn increasing external validity (De Bunt & Nencel, 2012). Though, to be able to check whether the sample is representative for the target population, the Dutch consumer, demographics were asked at the end of the questionnaire. As initially everybody in Holland is considered a consumer, the researched population is extensive. According to the Central Bureau for Statistics (CBS), 49% of the Dutch population is men and 50% is women. In this research 55% of the participants were women and 45% were men. The age of the researched sample varies from 16 to 61 with an average age of 27. In total 82% of the Dutch population lies within this age category. Important to note is that most of the respondents had an academic degree 72,4% (WO), though only 11,6% of the Dutch population achieved an academic degree. Accordingly, the sample is not entirely representative for researched population due to the skewness of education, that is 72,4% of the participants is highly educated. As a consequence, the results are not generalizable for the entire population.

3.5.3 Reliability

Reliability is about the degree to which the same results and the consistency of the measurement tools are yielded when the research is repeated. The measurement scales Brand Trust, Brand Blame and Purchase intention all had high Cronbach’s Alphas, hence, high reliabilities (see Table 1). To further take reliability into account, respondents were told that: participating in the experiment was voluntarily; the purpose of the study; by whom the study is conducted; their answers would be treated confidentially; respondents were free to stop participating at any desirable moment; and there was an opportunity to mail additional questions regarding the research.

4. Results

This chapter starts with a discussion about the manipulated stimuli, followed by an analysis of the descriptive statistics and the correlation matrix. Lastly the results of the tested hypotheses are reviewed.

4.1 Manipulation Checks

Manipulation checks were utilized in order to assess the success of the manipulation. The check demonstrated that respondents in both failure conditions read the newspaper articles (100 % N=112). These results are based on a multiple-choice task in which the respondents were asked to identify the main subject of the newspaper article. Furthermore, the control condition was also intended to verify that the manipulated factor caused variation in the dependent variables. As such, multiple one-way ANOVA’s are computed to assess whether the manipulated variable brand failure influenced brand trust and purchase intention. Based on the results, the control group significantly differed in the level of brand trust compared to the brand failure groups; F(2,153) = 62.01, p <.001, and significantly differed in level of purchase intention compared to the brand failure groups; F(2,153) = 88.500, p <.001. The aforementioned indicates that the manipulation worked as intended because the intervention caused variation in the dependent variables. Hence, participants’ reactions are altered due to the intervention.

4.2 Descriptive Statistics and Correlation Analysis

Table 1 demonstrates an overview of the descriptive statistics, correlations and scale reliabilities of the two treatment groups taken together (N=112). The independent variable was coded to 1=CEO and 2=Product.

Table 1

Means, Standard Deviations, Correlations and Reliabilities

Note: N=112. Reliabilities are reported along the diagonal * Correlation is significant at the 0.05 level (two-tailed) ** Correlation is significant at the 0.01 level (two-tailed)

The correlation matrix displays the degree of linear association between variables. A first observation, is that the control variable severity positively correlates with brand blame (r =.24, p<0.01), and negatively correlates with brand trust (r = -.31, p <.01) and purchase intention (r = -.30, p <.01), which indicates that severity might be a confounding variable that could have affected the results. Consequently, controlling for this variable is of the essence as it could have compromised internal validity. Severity also correlates with brand failure (rpb =

-.32, p <.01), however, this is a point-biserial correlation. The indication of the coefficient depends on which category is assigned to which code (Field, 2009). Therefore, no judgements can be made about the direction of the relationship. The strongest positive correlation in the matrix is between purchase intention and brand trust (r =.57, p <0.01), which indicate that if brand trust decreases, purchase intention decreases correspondingly. The relationship between brand blame and brand trust is negatively correlated signifying that if brand blame increases,

Variable M SD 1 2 3 4 5 6 7 8 1. Gender (0=female, 1=male) .55 .49 – 2. Age 27.34 8.33 -.22* – 3. Education 5.45 1.14 .04 .21* – 4. Severity 6.99 2.24 -.12 -.00 .05 – 5. Brand Failure (1=CEO,

2=Product)

1.50 .50 .07 -.05 -.05 .32** –

6. Brand Blame 4.68 1.18 -.19* -.01 -.04 .25** .24** (.79)

7. Brand Trust 2.66 .90 -.05 .07 .01 -.31** -.22* .00 (.89)

trust in the brand decreases. Brand failure correlates with brand trust (rpb =-.22, p <.05) but as

mentioned before nothing can be said about the direction of the relationship as the variable is a discrete dichotomy, hence, exists of two categories. Noteworthy, almost all correlations are not very strong since the values are not close to -1 or +1.

The descriptive statistics of the three groups are also presented separately in order to enhance the understanding of possible differences between the two groups, and to assess to what extent the level of brand trust changed after the intervention. Table 2 shows the descriptive statistics of the control group, Table 3 illustrates the descriptive statistics of the CEO misconduct group, and Table 4 demonstrates the descriptive statistics of the product failure group.

Table 2

Descriptive Statistics Group 0: Control Condition

Table 3

Descriptive Statistics Group 1: CEO misconduct

Variable N Min. Max. M SD

Gender (0=female, 1=men) 44 1 2 1.53 .51

Age 44 18 59 29.30 9.69

Education 44 3 7 5.5 1.00

Brand Trust 44 1 6.5 5.04 1.28

Purchase Intention 44 1 3.67 4.66 1.34

Variable N Min. Max. M SD

Gender (0=female, 1=men) 56 1 2 1.52 .50

Age 56 16 56 27.77 8.04 Education 56 2 6 5.5 1.06 Severity 56 2 10 6.29 2.02 Brand Blame 56 1 7 4.33 1.23 Brand Trust 56 1 5 2.88 .82 Purchase Intention 56 1 3.67 2.11 .74

Table 4

Descriptive Statistics Group 2. Product failure

Several observations should be addressed. First of all, the level of brand trust is higher in the control condition (M= 5.04) compared to CEO misconduct condition (M= 2.88) and product failure condition (M= 2.47). To profoundly compare the different means, two planned comparison were conducted. Planned contrasts revealed that brand failures significantly decreased brand trust compared to no control condition, t(74) = 28,40, p <.001. The same mean comparison was done for purchase intention. In the control condition respondents were more willing to purchase a product of the insurance company (M = 4.66) than in the CEO misconduct condition (M= 2.11) and product failure condition (M= 2.23). Planned contrasts revealed that brand failures significantly decreased purchase intention compared to control condition, t(67) = 21.69, p < .001. Lastly, the results from Table 1 showed that respondents in the product failure treatment perceived the failure to be more severe (M= 7.71) than the CEO misconduct group (M= 6.29).

Noteworthy, the control condition served as a means to assess to what extent consumers’ brand trust and purchase intention were altered compared to no failure condition. After the latter analyses, the control condition was filtered and not addressed in any further examinations. Thus, only the particpants in the manipulated conditions (N=112) were included in the analyses, as I am interested in effect differences between those two failures.

Variable N Min. Max. M SD

Gender (0=female, 1=men) 56 1 2 1.59 .50

Age 56 16 61 26.91 8.58 Education 56 2 6 5.39 1.21 Severity 56 1 10 7.71 2.24 Brand Blame 56 1.67 6.67 4.91 1.16 Brand Trust 56 1 5.88 2.47 .94 Purchase Intention 56 1 5.67 2.23 1.07

4.3 Hypotheses Testing

The first hypotheses H1 and H1a, forecasted a negative relationship between brand failure and brand trust, and that the level of trust will be lower after a product failure than CEO misconduct. A hierarchical multiple regression was calculated and the results are showed in Table 5.

Table 5

Results of Brand Failure as predictor for Brand Trust

Brand Trust Variable B SE β t R R2 DR Step 1 .31 .10 .10 Age .01 .01 .05 .50 Gender -.12 .18 -.07 -.70 Education .00 .08 .00 .05 Severity -.13 .04 -.31 -3.29** Step 2 .33 .11 .01 Age .01 .01 0.5 .47 Gender -.10 .18 -.06 -.57 Education -.00 .08 -.00 -.01 Severity -.11 .04 -.27 -2.72** Brand Failure -.22 .18 -.12 -1.24 Note: N = 112. *p <.05, **p <.01

The variables age, gender, education and severity were entered at step one of the regression to control for the effects of demographics and perceived severity of the failure. The multiple regression disclosed that at step one, control variable severity significantly contributed to the model, F(4,105) = 2.85, p<.01. The first model explained 10% of the variation in brand trust. Step two revealed that the added variable brand failure was not a significant predictor of brand trust, F(1,104) = 1.54, p =.22. H1 and H1a are not accepted. In other words, there is no negative relationship between brand failure and brand trust.

The set of hypotheses H2, H2a, H2b and H2c, assumed that the relationship between brand failure and brand trust would be mediated by brand blame, and that this mediation effect would be larger in the product failure condition than the CEO misconduct condition.

Four steps discussed by Baron and Kenny (1986), Judd and Kenny (1981), and James and Brett (1984) in establishing mediation are undertaken. All four steps have to be met to be able to conclude that brand blame completely mediates the relationship between brand failure and brand trust. The first step is concerned with the relationship between the independent variable brand failure and the dependent variable brand trust. As tested in the first hierarchical regression (see Table 5), no significant effect was found, F(1,104) = 1.54, p =.22. Table 6 displays step 2, that is the hierarchical regression analysis for brand failure predicting brand blame.

Table 6

Results of Brand Failure predicting Brand Blame

Note: N = 112. *p <.05, **p <.01

The multiple regression disclosed that brand failure contributed significantly to the model, F(5,104) = 3.13, p<.05 and accounted for an additional 3% of the variation in brand blame. The third step is concerned with the relationship between brand blame and brand trust and the fourth step controls if the effect of brand failure on brand trust is zero. Both steps are demonstrated in Table 7 as the effects are estimated in the same equation.

Brand Blame Variable B SE β t R R2 DR Step 1 .31 .10 .10 Age .01 .01 -.05 .47 Gender -.13 .18 -.07 -.72 Education .04 .20 .02 .20 Severity -.13 .04 -.31 -3.29** Step 2 .36 .13 .03 Age .01 .01 .04 .44 Gender -.11 .18 -.06 -.60 Education .04 .20 .02 .18 Severity -.11 .04 -.27 -2.74** Brand Failure -.22 .18 -.12 -1.24*

Table 7

Regression results for Brand Blame as a mediator of the relationship between Brand Failure and Brand Trust.

Note: N = 112 *p <.05, **p <.01

The multiple regression disclosed that brand blame contributed significantly to the model, F(1,103) = 6.42, p<.05 and accounted for an additional 7% of the variation in brand trust. A total of 40% variance in brand trust was explained by the whole model. In other words, participants brand trust decreased B = -.19 for each increase in the amount of blame attributed to the brand after a brand failure, and the level of brand trust was B =-.19 lower after brand blame for a product failure compared to CEO misconduct. Even though the condition of step one was not met, mediation has occurred as some researchers argue that step one is not required to determine a mediation (Kenny, Kashy, and Bolger, 1998; Shrout & Bolger, 2002). Noteworthy, it is a complete mediation as the effect of brand failure on brand trust only exist

Brand Trust Variable B SE β t R R2 DR Step 1 .31 .10 .10 Age .01 .01 -.05 .47 Gender -.13 .18 -.07 -.72 Education .04 .20 .02 .20 Severity -.13 .04 -.31 -3.29** Step 2 .33 .11 .01 Age .01 .01 .04 .44 Gender -.11 .18 -.06 -.60 Education .04 .20 .02 .18 Severity -.11 .04 -.27 -2.74** Brand Failure -.22 .18 -.12 -1.24 Step 3 .40 .16 .12 Age .00 .01 .03 .36 Gender -.19 .18 -.11 -1.1 Education .03 .19 .01 .15 Severity -.09 .04 -.23 -2.34** Brand Failure -.13 .18 -.07 -.76 Brand Blame -.19 .07 -.25 -2.53**

when brand blame is added to the equation, so the indirect effect is the total effect (Kenny, 2016). Based on the analyses H2, H2a, H2b and H2c are accepted.

The last hypotheses, H3 and H3a predict that brand trust has a positive relationship with purchase intention, and that the level of purchase intention will be lower after a product failure than CEO misconduct. Table 8 depicts the results of the hierarchical regression of the relationship between brand trust and purchase intention.

Table 8

Results of Brand Trust as predictor of Purchase Intention

Note: N = 112 *p <.05, **p <.01

p <.01

A significant regression equation was found F(1,104) = 38.73, p <.01), and after controlling for age, gender, education and severity, brand trust explained an additional of 26% of variance in purchase intention. Hypothesis 3 can be accepted. Respondents purchase intention decreased B =.53 for each decrease in the level of brand trust. The level of purchase intention was lower after CEO misconduct B =.53 than after a product failure. H3a is rejected.

Purchase Intention Variable B SE β t R R2 DR Step 1 .34 .12 .12 Age 0.1 .01 .09 .99 Gender -.11 .18 -.06 -.60 Education .21 .20 .10 1.10 Severity -.13 .04 -.32 -3.44 Step 2 .60 .36 .24 Age .01 .48 .07 .87 Gender -.04 .01 -.02 -.27 Education .19 .15 .09 1.11 Severity -.07 .17 -.16 -1.92 Brand Trust .53 .04 .52 6.22