Service Failure Recovery on Social Media:

Master Thesis

Dual Masters Award in Advanced International Business

Management and Marketing

Service Failure Recovery on Social Media:

How Observers Perceive Organizational

Complaint Handling

Ralph Gumann

Submission Date: December 5th_{, 2016} Word Count: 14,662

Newcastle University University of Groningen Supervisor: Dr. Ana Javornik Drs. Ad Visscher

A

Unsatisfied customers increasingly use social media networks such as Facebook to voice their complaints directly to companies. Often, these complaints are posted publicly, as complainers feel to put more pressure on companies when others can observe the complaint handling. With the help of an experiment, this study looked into different response styles of companies and how they result in cognitive and behavioural outcomes of observers. Results suggest that deontic justice theory is applicable and show that different to directly affected customers, interactional justice has the largest effect on satisfaction with complaint handling of observers. The experiment revealed that organizational complaint responses conveyed in high levels of conversational human voice lead to significantly higher outcomes of observers in terms of corporate image, word-of-mouth intention and purchase intention compared to responses conveyed in low levels of conversational human voice or no responses at all. Furthermore, only high conversational human voice responses prevented negative outcomes for observers after the exposure to the complaint and the organizational complaint handling. In contrast, organizational responses to social media complaints conveyed in low levels of conversational human voice as well as no responses at all, are likely to negatively influence observers.

L

C

2.1 Service Recovery on Social Media ... 14

2.1.1 Background of Service Recovery Research ... 14

2.1.2 Service Failure and Recovery ... 15

2.1.3 Shift to Complaints on Social Media ... 16

2.2 Observing Audiences on Social Media ... 17

2.2.1 Shift in Communication and involved Actors ... 17

2.2.2 Terminology ... 18

2.2.3 Theoretical Background and previous Research on Observers ... 20

2.3 Organizational Complaint Handling on Social Media ... 22

2.3.1 Response Strategies and Dimensions ... 22

2.3.2 Satisfaction with Complaint Handling ... 24

2.3.3 Organizational Webcare on Social Media ... 26

2.3.4 Conversational Human Voice ... 27

2.4 Conceptual Model ... 31

3 Methodology ... 32

3.1 Research Paradigm ... 32

3.2 Research Design ... 33

3.2.1 Research Approach ... 33

3.2.3 Manipulations ... 34

3.2.4 Stimulus Material Development ... 34

3.2.5 Measurement Constructs ... 37

3.3 Data Collection and Framework for Analysis ... 38

3.3.1 Questionnaire Design and Structure ... 38

3.3.2 Pre-test ... 39

3.3.3 Sampling ... 40

3.4 Ethical Considerations ... 41

4 Findings ... 42

4.1 Preparatory Analysis ... 42

4.1.1 Data Screening and Cleaning ... 42

4.1.2 Sample Characteristics ... 42

4.1.3 Variable Composition and Testing ... 46

4.1.4 Descriptive Analysis ... 48

4.1.5 Test for Non-Response Bias ... 48

4.2 Main Analysis ... 48

4.2.1 Manipulation Check ... 48

4.2.2 Perceived Justice on Satisfaction with Complaint Handling ... 50

4.2.3 Satisfaction with Complaint Handling on Outcomes ... 51

4.2.4 CHV on Perceived Justice and Satisfaction with complaint Handling ... 52

4.2.5 Overall Analysis of Experiment ... 52

4.2.6 Overview of Hypotheses Testing Results ... 57

5 Discussion ... 58

5.1 Recapitulation of Research Goals ... 58

5.2 Service Recovery Evaluation of Observers ... 58

5.3 The effect of Conversational Human Voice on Observers ... 60

5.4 Limitations ... 64

6 Conclusion ... 65

6.1 Concluding remarks ... 65

6.2 Contribution and Implications ... 65

Appendix ... 88

Appendix A – Stimulus Material ... 88

Appendix B - Preparatory Analysis ... 92

Appendix C - Manipulation Check ... 98

Appendix D - Hypotheses 1: Multiple Regression ... 101

Appendix E - Hypothesis 2: Linear Regressions ... 103

Appendix F - Hypothesis 3: Linear Regressions ... 106

Appendix G - Hypothesis 4 and 5: MANOVA ... 108

L

T

Table 2.1: Overview of Terminologies ... 19

Table 2.2: Complaint Handling Dimensions ... 23

Table 3.1: Comparison of Research Philosophies in Business Research ... 32

Table 3.2: Overview of Conditions ... 34

Table 3.3: Measurement Scales ... 38

Table 3.4: Conversational Human Voice Construct for Conditions... 40

Table 4.1: Socio-Demographic Sample Characteristics ... 43

Table 4.2: Study specific Sample Characteristics ... 44

Table 4.3: Respondents by Home Country ... 45

Table 4.4: Overview of Cronbach’s Alpha ... 47

Table 4.5: Descriptive Statistics and Pearson’s Correlations between Variables ... 48

Table 4.6: Tukey-Kramer Test Comparisons for Manipulation Check ... 49

Table 4.7: Regression coefficients and standard errors ... 51

Table 4.8: Mean and Std. Deviation Values for MANOVA ... 54

Table 4.9: Univariate ANOVAs (Tests of Between-Subjects Effects) ... 54

Table 4.10: Tukey Post-Hoc Analysis of MANOVA ... 55

Table 4.11: Paired Sample Tests for each Condition ... 56

L

F

Figure 2.1: Actors in Social Media Complaint Settings ... 18

Figure 2.2: Conceptual Model ... 31

Figure 4.1: Social Media Usage Frequency of Respondents ... 44

Figure 5.1: Mean Differences Cognitive Outcomes ... 60

Figure 5.2: Mean Differences Behavioural Outcomes ... 61

L

A

Abbreviation Meaning

CHV Conversational Human Voice

CI Corporate Image

PCA Principle Component Analysis

PI Purchase Intention

SATCOM Satisfaction with Complaint Handling

A

The present master thesis marks the end of my studies and the completion of my master’s degree at Newcastle University and the University of Groningen. I would like to thank my supervisor in Newcastle, Dr. Ana Javornik for her valuable feedback and support and for making a seamless transition from my previous supervisor. I also would like to thank my supervisor in Groningen, Drs. Ad Visscher for his feedback and support.

1. Introduction

1 I

1.1 Background

The advent of web 2.0 technologies such as social media has given consumers new opportunities to get in contact with companies (Hennig-Thurau et al., 2010). In 2015, 74% of the Fortune 500 companies had a Facebook presence (Ganim Barnes, Lescault and Holmes, 2015). Not seldom, customers use Facebook to complain or spread negative word-of-mouth directly on companies’ social media sites (Babić Rosario et al., 2016). One reason for this increase in social media complaints is the perceived cost of complaining. When consumers complain online the perceived costs in terms of effort and time are not only lower compared to traditional communication channels (Sharma

et al., 2010), but also consumers feel to be in a more powerful position vis-à-vis the

company. This is because complaints on social media are visible for other consumers as well and therefore complaining customers think they can put companies on the spot (Grégoire, Salle and Tripp, 2015).

For companies this poses new challenges as consumer-generated content is perceived to be more credible by other consumers (Bickart and Schindler, 2001; van Noort and Willemsen, 2012) and negative word-of-mouth has been found to affect consumers’ brand evaluation (Laczniak, DeCarlo and Ramaswami, 2001), purchase intention (Bansal and Voyer, 2000) and sales (Chevalier and Mayzlin, 2006). There is also a bias insofar that negative word-of-mouth has stronger effects compared to positive word-of-mouth (Chen and Lurie, 2013). In addition, the online setting of these service recovery scenarios presents further challenges to companies as the lack of human elements in computer-mediated communication hinders the development of trust (van Noort and Willemsen, 2012). This challenge is even amplified, as switching costs for consumers in online shopping are low (Chebat, Davidow and Borges, 2011; Matos, Henrique and Alberto Vargas Rossi, 2007).

1. Introduction

brand awareness, brand loyalty and stimulate purchase intention (Lipsman et al., 2012). And indeed, several studies have shown a positive relation between social media activities of companies and these previously mentioned customer outcomes (Benthaus, Risius and Beck, 2016; Kim and Ko, 2012; Kumar et al., 2016). A “post” in social media contexts refers to an electronic utterance aimed at other users to create engagement or potentially starting a discussion while “comments” are responses to posts (Lillqvist, Louhiala-Salminen and Kankaanranta, 2016).

Frequently, these company posts intended to create consumer engagement, are being captured by unsatisfied customers who post a complaint directly as a response to a company’s post. Often, these complaints then receive “likes” or comments from other users which leads the Facebook algorithms to list these customer complaints as “top comments”. This has the consequence that these complaints are directly listed beneath the company post, visible for every other user. In the month of July in 2016 for example, the Dutch airline KLM published 27 posts on their Facebook page in the form of photos, videos and links which received an average of 12,000 likes and 255 comments. Out of those 27 published posts, 13 or 48% displayed a complaint or any form of negative word-of-mouth as “top comment” (Sociograph, 2016). These kind of complaints pose a multiplex challenge to companies in their service recovery efforts, as companies not only have to consider the complaint itself but also understand the implications for other consumers who observe the complaint and the subsequent complaint handling (Schaefers and Schamari, 2016). Van Vaerenbergh, Vermeir and Larivière (2013) for instance, show that observing an unsuccessful service recovery, leads consumers to form negative perceptions of the company’s service quality.

1.2 Research Questions

1. Introduction

experience is abundant (Brocato, Voorhees and Baker, 2012; Grove and Fisk, 1997; Miao, 2014; Rosenbaum and Massiah, 2007). However, little research has been conducted on how observers perceive and react to service failures and recovery attempts that happened to others (Mattila, Hanks and Wang, 2014). This is especially true for online and social media settings.

This lack of attention is surprising because while only one focal customer is directly affected by the service recovery, on social media considerably more observers will read, evaluate and draw conclusions from the organizational response and complaint handling (Lee and Song, 2010). Additionally, social media is increasingly becoming important for consumers as a tool to make purchase decisions (Hudson et al., 2016) and for companies to reach and interact with customers (Lipsman et al., 2012; Schamari and Schaefers, 2015). In the past, this neglect could perhaps be explained by stakeholder theory in so far that observers represent secondary stakeholders, meaning that there is no relationship yet, and therefore are of lesser importance to companies (Frooman, 1999; Mitchell, Agle and Wood, 1997). However, in today’s transformed marketing environment with empowered consumers, secondary stakeholders can no longer be neglected (Deighton and Kornfeld, 2009; Karaosmanoğlu, Banu Elmadağ Baş and Zhang, 2011). Schaefers and Schamari, p. 204 (2016) state that due to the nature of social media, other consumers are likely to be influenced by the interactions of complainants and companies; and that such “spill-over effects” on observing audiences should be further researched.

1. Introduction

customers’ social media postings/conversations and generate appropriate responses when warranted?”.

The present thesis aims to contribute to fill this gap in the literature by specifically investigating how consumers perceive observed service recovery scenarios on social media and how this may affect them in their cognitive evaluations and future behaviours. In regards to the first mentioned question by Ostrom et al. (2015), the issue of potential spill-over effects on observers is of particular interest to determine the effectiveness of social media-based recovery efforts. Therefore, the following research questions is posed:

RQ 1: Do observers perceive and evaluate service recovery encounters on social media in similar ways as directly affected customers?

Addressing the second question by Ostrom et al. (2015), the issue to generate appropriate organizational responses to social media complaints, considering both the complaining customer but also the implications for observers is of utmost importance. Recently, the concept of conversational human voice has received considerable attention in the research of organizational online communication and complaint handling (Schamari and Schaefers, 2015; van Noort and Willemsen, 2012). Conversational human voice can be defined as “an engaging and natural style of organizational communication as perceived by an organization’s publics based on interactions between individuals in the organization and individuals in publics” (Kelleher, 2009, p. 177).

The idea behind this concept is that companies appear to speak less in a ‘corporate voice’ which is associated with profit-driven intentions, but more ‘human-like’ and therefore consumers can more easily form relationships with brands and perceive their communication as friendly (Dijkmans et al., 2015). Consequently, the second research question will investigate if organizational complaint responses conveyed with a conversational human voice can be an effective strategy to counter potential negative effects for observers on social media:

2. Literature Review

2 L

R

2.1 Service Recovery on Social Media

2.1.1 Background of Service Recovery Research

Service recovery has been studied for more than four decades, but still companies struggle with various aspects of it (van Vaerenbergh and Orsingher, 2016). For example, the 2015 National Customer Rage Study shows that customer satisfaction with complaint handling is no higher than it was in 1976 (CCMC, 2015). Even though the reasons for this might be multiplex, it shows that the topic remains relevant and worth studying. According to Kunz and Hogreve (2011, p. 244), there are two reasons why complaint handling and service recovery will “continue to be of major importance for research and management”: (1) effective service recovery management will remain important as a competitive advantage for companies and (2) new forms of service delivery and technologies demand different service recovery strategies, for example in online settings.

Research in the service failure and recovery domain is manifold. Operation management researchers investigate how companies can develop service recovery systems to improve recovery processes (Johnston and Michel, 2008; Smith, Fox and Ramirez, 2010; Smith, Karwan and Markland, 2009), human resource management researchers are concerned with the question how employees can be best trained for service recovery encounters with customers (Koc, 2013; Liao, 2007; van der Heijden et

al., 2013), and marketing researchers aim to understand how customers perceive and

2. Literature Review

2.1.2 Service Failure and Recovery

Service failures are an inevitable part of every service provided. This is due to unique characteristics of services and the complexities of service delivery (Balaji, Jha and Royne, 2015; Hart, Heskett and Sasser, 1990). Services are intangible, (i.e. they can only be evaluated by customers during consumption), services are heterogeneous (i.e. every service encounter is unique) and services are produced and consumed at the same time (Parasuraman, Zeithaml and Berry, 1985; Subramony and Pugh, 2014).

A service failure (real or perceived) occurs when the service received does not meet the expectations of a customer (Andreassen, 2001; Maxham, 2001). The actions service providers undertake in response to such failures are referred to as service recovery (Kelley, Hoffman and Davis, 1993). A service recovery encounter can consequently be defined as an exchange between a customer who experienced a service failure and a company which attempts to provide the customer with a gain in order to restore the customer’s loss (Smith, Bolton and Wagner, 1999). The notion of exchange is taken from social exchange and equity theory (Hatfield, Walster and Berscheid, 1978; Homans, 1961). It can be argued that service recovery encounters entail utilitarian exchange aspects such as money, goods or time as well as social and psychological symbolic aspects such as esteem and status (Smith, Bolton and Wagner, 1999).

2. Literature Review

2.1.3 Shift to Complaints on Social Media

The internet has drastically changed how consumers interact among each other and with companies (Hennig-Thurau et al., 2010). Deighton and Kornfeld (2009, p. 4) state that the “digital innovations of the last decade made it effortless, indeed second nature, for audiences to talk back and talk to each other’’. Especially through the advent of Web 2.0 technologies and its new media channels such as Facebook, Twitter, YouTube and blogs, have given consumers extensive options to create, share and comment on brand related content. This has become known as electronic word-of-mouth (Gruen, Osmonbekov and Czaplewski, 2006) and can come in the forms of Facebook posts, tweets, reviews, blog posts, “likes”, “pins,” images and video testimonials (Babić Rosario et al., 2016). According to Babić Rosario et al. (2016, p. 297), electronic word-of-mouth represents “one of the most significant developments in contemporary consumer behaviour” and has turned consumers into “web-fortified” decision makers.

Web 2.0 and especially social media has also changed how customers complain to companies (Naylor, Lamberton and West, 2012; van Noort and Willemsen, 2012). Complaints are a special form of (negative) mouth. While negative word-of-mouth represents any negative statement about a company, its service or products, complaints are directed directly at the company with a certain goal (Kowalski, 1996). Einwiller and Steilen (2015, p. 196) define complaints as “an expression of dissatisfaction for the purpose of drawing attention to a perceived misconduct by an organization and for achieving personal or collective goals”.

2. Literature Review

consumers”. Consumers are aware of these large audiences which potentially receive their negative word-of-mouth messages when they post complaints publicly to companies (van Noort and Willemsen, 2012).

However, it is important to note that not all cases of service failures lead to public online complaints. Typically, customers turn to public complaint channels, when they see no other help or due to a series of failures, for example when the service failure was followed by a failed service recovery (Grégoire, Tripp and Legoux, 2009). This is called double deviation and can substantially hurt the customer relationship (Basso and Pizzutti, 2016; Grégoire and Fisher, 2008; Joireman et al., 2013).

2.2 Observing Audiences on Social Media

2.2.1 Shift in Communication and involved Actors

The previously discussed changes brought by digital innovations have also affected the way of complaint communication. In the past, complaining was a dyadic communication between a company and the complaining customer held via non-public channels such as telephone, letters, emails or in person. Nowadays, when consumers use company pages on social media to voice their complaints, complaining has become a triadic communication as it also involves other stakeholders who observe and potentially even engage in the complaint handling process (van Noort et

al., 2015).

2. Literature Review

Figure 2.1: Actors in Social Media Complaint Settings

The dark grey boxes represent actors in the complaint setting, as previously described. The light grey boxes represent posts or comments made by these actors. The solid lines indicate the target of the post or comment and the number on each arrow shows its typical order. Lastly, the dotted lines indicate that observers can see the original complaint post from the complainant, comments from repliers as well as the reaction to the complaint from the company.

2.2.2 Terminology

Previous research has made use of various terminologies to refer to other people involved in service encounters next to the focal customer. Usually, ‘focal customer’ refers to the customer who is in focus of the respective research, which in most cases is the customer who engages in the service encounter with a company.

Depending on research focus and context, different terms are used to describe other customers who are indirectly part of the service encounter through their presence. These may be customers waiting next in line, or simply other people observing the service encounter for example in a restaurant (Grove and Fisk, 1997). Table 2.1 shows examples of different terminologies.

Complainers Complaint _{Post Facebook page}Company

2. Literature Review

Table 2.1: Overview of Terminologies

These other customers can then be further distinguished into customers who are only present at the service encounter but do not interact with the focal customer and customers who interact with the focal customer. The first case is often referred to as ‘mere social presence’ whereas the latter case is referred to as ‘interactive social presence’ (Argo, Dahl and Manchanda, 2005).

For the context of online social media settings, Naylor, Lamberton and West (2012) have introduced the term ‘virtual presence’. Again, a distinction can be made between solely observing users and interacting users. Naylor, Lamberton and West (2012, p. 105) define ‘mere virtual presence’ as the “passive exposure to a brand’s supporters experienced in […] social media contexts”. Accordingly, ‘interactive virtual presence’ refers to users who actively engage in the exchange through comments about their own opinion, or support for the complainant or the company (Lee and Song, 2010; Schaefers and Schamari, 2016).

The present thesis solely focuses on online social media settings and within that on observers who do not actively interact with the complaint situation. Therefore, the example of Lee and Song (2010) will be adopted by using the simple terms: complainant to refer to the Facebook user who posted the complaint and observers to refer to the audience which observes the complaint handling.

Terminology Source

Bystanders Steinhoff and Palmatier, 2016; Yi He, Qimei Chen and

Alden, 2008

Customer next in line Cowley, 2005; van Vaerenbergh, Vermeir and Larivière, 2013

Observing consumer Cowley, 2005; Mattila, Hanks and Wang, 2014; Schamari

and Schaefers, 2015

Observing customer Mattila, Hanks and Wang, 2014; Wan, Chan and Su, 2011;

Yi, Gong and Lee, 2013

Other consumers Miao, 2014; Thakor, Suri and Saleh, 2008

Other customers Brocato, Voorhees and Baker, 2012; Grove and Fisk, 1997;

2. Literature Review

2.2.3 Theoretical Background and previous Research on Observers

Social Comparison Theory

Steinhoff and Palmatier (2016) researched loyalty programs as a marketing tool. They apply social comparison theory (Festinger, 1954) to see how loyalty programs affect both target consumers (who receive loyalty bonuses) and observers (who witness others receiving a bonus but don’t receive one themselves). Their results show that observers compare their input-output ratio in the relationship with the brand to the perceived input-output ratio of the focal customer. This often leads to feelings of unfairness and unjust treatment, which can cause strong negative reactions for the observer (Samaha, Palmatier and Dant, 2011).

Observational Learning Theory

Van Vaerenbergh, Vermeir and Larivière (2013) in their research on service recovery’s impact on customers ‘next-in-line’ applied observational learning theory (Bandura, 1977; Bandura, 1986). According to the theory, two types of observational learning can be distinguished. The first is imitation and of lesser interest for the present study. The second however, is called vicarious learning and describes how an observer’s attitude and behaviour changes similar to the focal customer’s attitude and behaviour (van Vaerenbergh, Vermeir and Larivière, 2013). For example, if an observer sees the negative reactions of a focal customer as a result of an unsatisfactory service recovery, the observer draws similar negative conclusions, such as inferences about minor service quality. It might appear obvious that the observation of a negative service recovery leads to negative quality inferences, however the authors find evidence that this relationship is moderated by locus of control attributions which means who was to blame for the service failure. Interestingly also, the results show no positive effects of observing a successful service recovery (compared to a failure free service delivery), only negative effects in cases of an unsuccessful service recovery. Attribution Theory

2. Literature Review

word-of-mouth communications (Laczniak, DeCarlo and Ramaswami, 2001; Lee and Song, 2010). Cowley (2005) investigated the fundamental attribution error from the view of consumers ‘next in line’. She found that the observer’s allocentric-idiocentric orientation, which is the “psychologically equivalent to collectivist and individualist cultural orientations” (Cowley, 2005, p. 140), plays an important role in observers’ situational evaluations. While allocentric observers consider the service providers’ situational constraints when attributing blame, idiocentric observers blame the service provider regardless of the situation. Generally, idiocentrics try to avoid disappointment and therefore lower their expectations of their own service encounter (Cowley, 2005).

Deontic Theory of Justice

Mattila, Hanks and Wang (2014) researched how observers react to service failures and subsequent recoveries happened to others in terms of emotions, justice perceptions and behavioural intentions. They applied the deontic theory of justice, which is based on the assumption that everyone deserves equal treatment in similar situations and that individuals have an “inherited predisposition to be sensitized to unfair treatments” (Mattila, Hanks and Wang, 2014, p. 554) and therefore care about the wrongdoings happened to others, not because of self-interest but because it is ‘the right thing’ (Cropanzano, Goldman and Folger, 2003). The results of Mattila, Hanks and Wang (2014) show that observers who witness unfair treatment of others showed negative emotions, had lower fairness ratings and reported lower levels of intention to return. Interestingly, these outcomes hold true when the observer’s own service encounter was not flawed and even are amplified when observer’s own service was flawed. Consequently, the results support the claims of deontic justice theory.

Implications for the Present Research

2. Literature Review

Vermeir and Larivière, 2013). Libai et al., p. 269 (2010) emphasise that especially online environments enable consumers to observe and learn about the behaviour of others, which will make observational learning a “fundamental form of C2C [customer to customer] interaction”. Zhang (2010) shows that mere observations can also lead to inferences about product quality. For the purpose of this study, two theories are of particular interest. Firstly, justice theory, and within it deontic justice theory which helps to establish whether observers perceive organizational complaint handlings similar to directly affected focal customers. Secondly, observational learning theory is specifically suited to explain cognitive and behavioural outcomes after observers witnessed a service recovery on social media.

2.3 Organizational Complaint Handling on Social Media

2.3.1 Response Strategies and Dimensions

The first decision companies have to make when dealing with complaints is whether or not to reply. “No response” strategies mean that the company does not react or take any action in regards to the complaint (Lee and Song, 2010). Following this strategy means that companies try to separate themselves from the complaint which can be useful if it is unclear who is to blame for the complaint or if potential responses may cause even more severe reactions by the complainant or the public (Lee and Song, 2010; McLaughun, Cody and O’Hair, 1983). However, no responses are most likely only accepted by strong supporters of the brand; while other observers may draw negative conclusions about the brand (Lee and Song, 2010).

2. Literature Review

complaint handling on social media, Einwiller and Steilen (2015) found that accommodative strategies are positively related to satisfaction with complaint handling but defensive response strategies are not. Moreover, Lee and Song (2010) found evidence that compared to no answers, defensive answers led observers to think that the company was at fault. Further, accommodative responses were associated with positive company evaluations.

Davidow (2003) provides often quoted dimensions of organizational complaint handling. He identified six dimensions that are important for organizational complaint handling as listed in Table 2.2. These give a broad overview of organizational response options and dimensions which are relevant to understand how consumers evaluate

Table 2.2: Complaint Handling Dimensions

Dimension Definition

Timeliness The perceived speed with which an organization responds

to or handles a complaint

Facilitation The policies, procedures, and structure that a company has in place to support customers engaging in complaints and communications

Redress The benefits or response outcome that a customer receives from the organization in response to the complaint

Apology An acknowledgement by the organization of the complainant’s distress

Credibility The organization’s willingness to present an explanation or account for the problem

Attentiveness The interpersonal communication and interaction between the organizational representative and the customer

2. Literature Review

2.3.2 Satisfaction with Complaint Handling

Orsingher, Valentini and Angelis, p. 170 (2010) define satisfaction with complaint handling as “the customer’s evaluation of how well a service company has handled a problem”. Furthermore, the general framework behind satisfaction with complaint handling, is the confirmation / disconfirmation paradigm of the satisfaction literature (Oliver, 1980), meaning that customers compare their perception of the quality of the complaint handling with their expectations they had before.

Antecedents of Satisfaction with Complaint Handling

Justice theory is the prevailing applied framework in the literature to assess the drivers of satisfaction with complaint handling as confirmed by various meta-analyses (Gelbrich and Roschk, 2011; Orsingher, Valentini and Angelis, 2010). Justice perception is the subjective evaluation of a company’s complaint handling by a customer (Smith, Bolton and Wagner, 1999). In regards to complaint handling, there are three relevant dimensions that can be distinguished. These relate back to the organizational complaint handling strategies as discuss in chapter 2.3.1.

Distributive Justice refers to the customer’s fairness perception of the offered redress

to offset the loss experienced through the service failure (van Vaerenbergh and Orsingher, 2016). This redress can be a monetary compensation or a psychological compensation, for example by providing an apology. Procedural Justice refers to the customer’s perception of the means (procedures, policies, etc.) which were applied in the complaint handling. This also includes the flexibility of the procedures, time to handle the complaint, and the customer orientation of the procedures (Orsingher, Valentini and Angelis, 2010). Interactional justice refers to the customer’s perception of the way he or she was treated during the service recovery. This can be realized by favorable employee behavior such as providing explanations or being friendly (Gelbrich and Roschk, 2011). This dimension also explains why customers can still be dissatisfied with the handling of a complaint even though the service failure was resolved.

2. Literature Review

Orsingher, Valentini and Angelis (2010) examined 50 studies and Gelbrich and Roschk (2011) reviewed 87 studies. Distributive justice was found to have the biggest influence. Nevertheless, all dimensions proved to be relevant in predicting customers’ satisfaction with complaint handling. However, the present study is interested in observers and their evaluations. Therefore, it is drawn on what Spencer and Rupp (2009) call ‘third-party justice’. In that line, O’Reilly, Aquino and Skarlicki (2016) argue based on the deontic model of justice that people not only seek justice for their own interest but also for others because it is the right thing to do and that standards of fairness should be upheld. Moreover, people sometimes even go further in punishing ‘justice violators’ even though they are not direct victims of the injustice. In this vein, the present study argues similar to Mattila, Hanks and Wang (2014), that observers perceive organizational complaint handling similar to directly affected customers in the way that distributive, procedural, and interactional justice plays a major role in determining satisfaction with complaint handling. Therefore it is hypothesized that:

H.1a Distributive justice in service recoveries on social media has a positive influence on satisfaction with complaint handling of observers.

H.1b Procedural justice in service recoveries on social media has a positive influence on satisfaction with complaint handling of observers.

H.1c Interactional justice in service recoveries on social media has a positive influence on satisfaction with complaint handling of observers.

Consequences of Satisfaction with Complaint Handling

Among others, research on service recovery or the outcome of satisfaction with complaint handling, often focuses on (re)purchase intention (Holloway and Beatty, 2003; Maxham, 2001), word-of-mouth intention (van Vaerenbergh, Lariviere and Vermeir, 2012) and corporate image (Andreassen, 2001; Mostafa et al., 2015)Therefore it is hypnotized that:

H.2 Higher levels of satisfaction with complaint handling will positively

influence observers’ (a) purchase intention, (b) word-of-mouth intention, and (c) corporate image.1

2. Literature Review

2.3.3 Organizational Webcare on Social Media

The previous subchapters have discussed several changes brought by information and Web 2.0 technologies that impact company – customer relationships and affect observers on social media. These included shifts in perceived costs of complaining as well as shifts of power due to public complaining. In addition, the shift from dyadic to triadic communication allows for observing audiences to get involved.

Companies have reacted to these shifts by implementing ‘webcare’ teams which monitor (negative) brand related content on social media and potentially react to it. The goal is not only to provide complaining customers with a response, but also to mitigate negative effects on larger observing audiences by restoring, or at least improving brand evaluations of both the complainant and the observers after a complaint (van Noort et al., 2015; van Noort and Willemsen, 2012). Webcare can be defined as the “act of engaging in online interactions with (complaining) consumers, by actively searching the web to address consumer feedback (e.g., questions, concerns and complaints)” (van Noort and Willemsen, 2012, p. 133). It has become a “pivotal part of brand management” (Schamari and Schaefers, 2015, p. 20) and can be regarded as a tool to support customer relationships, manage brand image and reputation, as well as to create consumer engagement. When brands show attentiveness and responsiveness in their webcare, they have the chance to restore satisfaction, corporate image, brand evaluations and limit negative consequences such as negative word-of-mouth (Chevalier and Mayzlin, 2006; Lee and Song, 2010; van Laer and Ruyter, 2010).

2. Literature Review

platforms, they are also active on other platforms such as Facebook, as the example of KLM shows. In their study, van Noort and Willemsen (2012) compared reactive response strategies (responses to a specific consumer request) and proactive response strategies (responses to complaints without a request from the consumer) across different social media platforms. Their results show that proactive webcare is more effective on brand-generated platforms, whereas reactive webcare is effective on brand-generated platforms but also on non-brand-generated platforms such as Facebook. Furthermore, their hypothesis that “conversational human voice serves as an underlying mechanism for the differential effects of webcare strategy on brand evaluations across brand-generated and consumer-generated platforms” (van Noort and Willemsen, 2012, p. 138) was confirmed.

2.3.4 Conversational Human Voice

The concept of conversational human voice originally goes back to Levine et al. (2000) in which Searls and Weinberger (2000, p. 110) raise the question: “How can a large company have conversations with hundreds of millions of real people?”. Contemporary research on conversational human voice predominantly refers to Kelleher and Miller (2006) and Kelleher (2009) and makes use of their developed measurement scale. According to Kelleher (2009, p. 176) the above posed question, “echoes the challenge” for companies to communicate online with large audiences. In the past, companies often adopted a ‘corporate voice’, meaning that they would speak in ‘one voice’ and with ‘one identity’ which let their communications appear formal, persuasive and profit-driven (Dijkmans et al., 2015; Levine et al., 2000).

2. Literature Review

on interactions between individuals in the organization and individuals in publics” (Kelleher, 2009, p. 177). It is characterised by “being open to dialog, welcoming conversational communication, and providing prompt feedback” but also “communicating with a sense of humour, admitting mistakes, treating others as human, and providing links to competitors” (Kelleher and Miller, 2006, p. 399). Particularly the last mentioned characteristics are traditionally not associated with company communications. This leads to the hypothesis that:

H.3 Higher levels of conversational human voice in organizational responses to social media complaints, will positively influence (a) justice

2. Literature Review

Previous studies have found strong evidence for the effect of conversational human voice in organizational communication. The perception of conversational human voice is significantly related to relational outcomes such as trust, satisfaction, control mutuality and commitment (Kelleher, 2009; Kelleher and Miller, 2006). In their study on crisis communication through organizational blogs, Yang, Kang and Johnson (2010) showed that openness to dialog is essential in crisis communication and leads to positive post-crisis outcomes such as positive attitudes towards an organization and word-of-mouth intentions. Park and Cameron (2014) also show that conversational human voice leads to the perception of social presence and interactivity in online communication. Sung and Kim (2014) report that creating a company social media site is not sufficient to generate positive public relations, but that companies have to put effort in interacting with customers. Their results show that consumers evaluate brands more positively when they are highly interactive on social media and employ less promotional messages.

Dijkmans et al. (2015) investigated in a longitudinal study whether and to what extent exposure to a company’s social media activities over time is beneficial for corporate reputation, and whether conversational human voice mediates this relation. The results show that conversational human voice indeed mediates the relation. The authors conclude that conversational human voice fosters corporate image in online communications and will continue to grow in importance for a number of reasons: (1) conversational human voice in social media nurtures dialogic relationships, which are effective in building dynamic and enduring relations with consumers. (2) It creates interactivity in communications, which positively influence attitudes toward the company, perceived company credibility, and organization reputation. (3) Lastly, through conversational human voice, companies are perceived as more human by consumers and therefore it becomes easier to build relationships. Particularly interesting for this thesis is that Dijkmans et al. (2015) note that even observers who read messages conveyed through high levels of conversational human voice are likely to perceive the company as more human-like.

2. Literature Review

strategies and platform type and that this effect is moderated by conversational human voice. They also found that brands are evaluated better when they respond to complaints rather than adopting a ‘no response’ strategy. Van Noort and Willemsen, p. 138 (2012) conclude that the implication of their study is that conversational human voice should “receive focal attention” in the development of effective webcare strategies. Combining these findings with the theoretical backgrounds and findings on observers discussed in chapter 2.2, it is thus hypothesized that:

H.4 Organizational responses to social media complaints conveyed through higher levels of conversational human voice (compared to low levels and no response) will positively influence observers’ (a) satisfaction with complaint handling (b) word-of-mouth intention, (c) purchase intention, and (d) corporate image.

H.5 Organizational responses to social media complaints conveyed through lower levels of conversational human voice will be better perceived by observers then no responses at all in terms of (a) satisfaction with

2. Literature Review

2.4 Conceptual Model

The previously hypothesised relationships are depicted in the conceptual model for this study, as shown in Figure 2.2. As stated in hypothesis 1, it is assumed that similar

Figure 2.2: Conceptual Model

to the complaint evaluation of directly affected customers, the perception of justice is an antecedent to satisfaction with complaint handling for observers. Accordingly, satisfaction with complaint handling will positively affect behavioural and cognitive outcomes such as word-of-mouth intention, purchase intention and corporate image. As hypothesis 3 states, it is further assumed that the level of conversational human voice conveyed in the organizational complaint response relates to the justice perception and satisfaction with complaint handling of observers.

Lastly, hypothesis 4 and 5 are not directly displayed in the model as they represent the overall experiment, combining the response strategies in terms of conversational human voice, or not to respond on the left side with the behavioural and cognitive outcomes on the right side.

Behavioural and Cognitive Outcome Tone of Organizational Complaint Response Satisfaction with Complaint Handling No Response Perceived Justice Procedural Distributive Interactional Corporate Image Purchase Intention Word-of-Mouth Intention Low Conversational Human Voice High Conversational Human Voice

Assumed positive relationship Assumed negative relationship

H.1 H.2

3. Methodology

3 M

3.1 Research Paradigm

Research paradigms help to describe the way social phenomena are understood and explanations from them can be gained (Saunders, Lewis and Thornhill, 2012). As shown in table Table 3.1 there are various different approaches to business research available (Saunders, Lewis and Thornhill, 2012, p. 140). The present study follows the research paradigm of positivism, as it aims to conduct a deductive explanatory research.

Table 3.1: Comparison of Research Philosophies in Business Research

Ontology Epistemology Axiology

Pragmatism External, multiple, chosen to best answering of research question

Either or both observable phenomena and subjective meanings can provide accept-able knowledge dependent upon the research question.

Values play a large role in interpreting results, the researcher adopting both objective and subjective points of view

Positivism External, objective and independent of social actors

Only observable phenomena can provide credible data, facts. Focus on causality and law-like generalisations, reducing phenomena to simplest elements

Research is under- taken in a value-free way, the researcher is independent of the data and maintains an objective stance Realism Is objective. Exists

independently of human thoughts and beliefs or knowledge of their existence (realist), but is interpreted through social conditioning (critical realist) Observable phenomena provide credible data, facts. Insufficient data means inaccuracies in sensations (direct realism). Alterna-tively, phenomena create sensations which are open to misinterpretation (critical realism). Focus on explaining within a context or contexts

Research is value laden; the researcher is biased by world views, cultural experiences and upbringing. These will impact on the research

Interpretivism Socially constructed, subjective, may change, multiple

Subjective meanings and social phenomena. Focus upon the details of situation, a reality behind these details, subjective meanings

motivating actions

3. Methodology

3.2 Research Design

3.2.1 Research Approach

With regards to the previously described research paradigm, a quantitative research approach appears to be best suited to test the hypothesis developed in the second chapter. The literature review also described that there are several organizational strategies available for companies to respond to social media complaints. In order to adequately reflect these different response strategies, an experimental research design is chosen. Experiments are a very common method to conduct research on consumer behaviour (Peighambari et al., 2016). Moreover, experiments allow to explore causal relationships by manipulating independent variables, and holding dependent variables constant (Zikmund et al., 2010). That is the idea behind this study by investigating the effects of conversational human voice on observers. Further, a between-subjects experimental design is chosen, because compared to a within-subject design, the validity of the results is usually higher and the possibility of demand characteristics is substantially reduced (Zikmund et al., 2010).

3.2.2 Industry and Social Media Platform Selection

3. Methodology

Facebook presence (Ganim Barnes, Lescault and Holmes, 2015). Also, compared to other social media networks such as Twitter, Facebook allows for longer messages, which enables brands to better engage in communications and therefore show more conversational human voice in their responses.

3.2.3 Manipulations

In order to perform the experiment, three separate conditions were deemed appropriate. One condition for a response conveyed in a high conversational human voice, one condition with a low conversational human voice and lastly, one condition with no response. The no response condition has a somewhat hybrid function in the experiment. On the one hand it functions as a control group, on the other hand it also functions as a response strategy, namely not to respond at all to the complaint. The next subchapter will further describe how the three conditions, displayed in Table 3.2 were developed and how they differ.

Table 3.2: Overview of Conditions

3.2.4 Stimulus Material Development

In order to perform the experiment in a realistic manner, an adequate complaint to an airline company’s Facebook post had to be found. Therefore, the website Sociograph (2016) was used to data mine the Facebook postings of various international airlines (Emirates, Qatar Airways, Singapore Airlines, KLM, Ryan Air, Eurowings). The first three were selected because they represent the world’s three best rated airlines (Skytrax, 2016), KLM was selected because of their claim to be the industry’s leader on social media (KLM, 2015) and Ryan Air and Eurowings were selected, because they represent, in contrast to the other airlines, so called ‘low cost carriers’ which means that their focus is less service oriented and more low cost driven.

3. Methodology

This resulted in a list with company posts ranging from a few dozens to several hundred cases depending on the social media activity of the respective brand in the time period from January until August 2016. Only posts initiated by the respective airline were considered. Each list was then sorted in descending order by “Likes” as the goal was to find complaints which reached a large number of observers (Brettel et

al., 2015).3 _{One particular Facebook post by the airline ‘Emirates’ was identified which} received around 58 thousand likes and more than 23 thousand shares.4 _{In said post,} Emirates promoted its Airbus A380 and attempted to create user engagement among its followers by asking to share the post if quote: “you have flown on the Emirates A380”.

This post was then commented by a user complaining about her recent trip. The comment in turn was ‘liked’ by 18 other Facebook users which resulted in the comment being displayed as the ‘top comment’ for the post. A test in the form of brief semi-structured interviews among six students (50% female, Mage=23, SD=1.11) was conducted in order to assess if the complaint would be suitable for the experiment. The test revealed that the complaint was not suitable for this study, as participants perceived it as incoherently formulated and confusing. As a result it was decided to create a fictional complaint.

In order to gain an understanding of common formulations and writing styles of social media complaints, the previously obtained lists of airline Facebook posts was used again. According to the U.S. Department of Transportation, in 2015 nearly half of complaints in the airline industry were related to flight problems (36.1%) and baggage problems (13.4%) (Bowen and Headley, 2016). Therefore, with the goal to represent those two complaint categories, the top-listed cases were coded to identify style and writing patterns to use them for the creation of the fictional complaint. The original complaint as shown in Appendix A, was then taken and adapted with the findings from the coding procedure. This resulted in a complaint that reads as follows:

#emirates has the most disappointing customer service ever. I’ll never trust them with my holidays again. They delayed our flight for one day and then gave us a

3 _{‘Likes’ only counts user that react to a post, but not users who merely see the post. For a review on}

different Facebook measures see Brettel et al. (2015)

4 _{As of October 10}th _{2016. The original post can be found under:}

3. Methodology bad hotel.

And when we finally arrived back home, of course none of our bags arrived. No one at the airport or on the phone could tell me where my bags are. NEVER AGAIN. I advise everybody don’t choose Emirates.

The original complainant used the hashtag ‘#emirates’ supposedly to draw more attention to her issue, which was taken over. The rest of the fictional complaint is largely inspired by other similar complaints shown in Appendix A.2. The original name of the complainant was replaced with a generic name for anonymity reasons but also because previous research shows that perceived similarity to focal customers can have an influence on observers (Wan, Chan and Su, 2011). The same procedure was then repeated to identify company responses with high levels of conversational human voice, reflecting communication characteristics described by Kelleher (2009).5 _{In that} procedure, KLM in particular appeared to communicate in higher levels of conversational human voice, which indicates that they are indeed leading the industry on social media which is in line with findings of Dijkmans et al. (2015). The resulting fictional response intended for the ‘high conversational human voice condition’ then read as follows:

Dear Brenda, we understand that this has been an inconvenient situation for you, and we can imagine that this is not how you envisioned your experience.

Naturally, we’re doing our utmost to ensure that flights are on time.

Unfortunately, due to the high safety requirements and complex influencing factors in air traffic, it is never entirely possible to rule out delays.

We would like to check the status of your delayed luggage, therefore we kindly ask you to send us a private message via this link http://bit.ly/QR-tell-us. Please include your booking code and full name as stated on your ticket/passport. We’ll figure this out, Brenda.

For the ‘low conversational human voice condition’ no fictional response was created but an original response from Qatar Airways was used, which can be seen in Appendix A.2. The response reads as follows:

Hello Brenda Sanders,

3. Methodology

It is important to note that in both conditions the underlying company strategy is the same, namely to redirect the complaint handling to non-public domains. This is a very common strategy because there is a “reluctance of publicly handling complaints” on social media as Einwiller and Steilen (2015, p. 201) note. While this reluctance is of lesser interest for this study, it is central for a successful manipulation that there is no bias stemming from different organizational strategies. Both underlying strategies try to redirect the conversation, however in the high level condition the company shows empathy, offers an explanation, shows more attentiveness and addresses the complainant with her first name.

With the help of Adobe Photoshop CC 2015, the complaint as well as both responses were then placed in a screenshot of the actual Facebook post to let it appear real. The third condition ‘no response’ looked exactly the same, except that there was no response from Emirates. All stimuli can be viewed in Appendix A.1. In a second test among seven different students (43% female, Mage=25, SD=1.76) participants were then shown the fictional complaint and response together with three real ones taken from the Emirates Facebook page and asked if they perceive any material as unrealistic. This was not the case.

3.2.5 Measurement Constructs

3. Methodology

Table 3.3: Measurement Scales

Construct Adapted from

Conversational Human Voice Kelleher (2009), Kelleher and Miller (2006) Distributive Justice Mostafa et al. (2015), Tax, Brown and

Chandrashekaran (1998)

Procedural Justice Mostafa et al. (2015), Tax, Brown and Chandrashekaran (1998)

Interactional Justice Mostafa et al. (2015), Tax, Brown and Chandrashekaran (1998)

Satisfaction with Complaint

Handling Tax, Brown and Chandrashekaran (1998)

Corporate Image Andreassen (2001)

Word-of-Mouth Intention Brown et al. (2005)

Purchase Intention Coyle and Thorson (2001)

3.3 Data Collection and Framework for Analysis

3.3.1 Questionnaire Design and Structure

The questionnaire was created with the help of the online tool ‘Qualtrics’. The design was mainly based on the recommendations of Brace (2008) and Bradburn, Sudman and Wansink (2004). All instructions and questions were carefully formulated to avoid ambiguity, double-barrelled questions, and misunderstandings among respondents. Following Podsakoff et al. (2003) to reduce method biases, respondents were assured that there were no right or wrong answers, and that they should answer to the best of their knowledge. To reduce order biases, question and item orders were randomized whenever possible (Podsakoff et al., 2003).

3. Methodology

their perception of conversational human voice, perceived justice (distributive, procedural, interactional), satisfaction with complaint handling, word-of-mouth intention, purchase intention and corporate image.

Before proceeding to the last block an instructional manipulation check recommended by Oppenheimer, Meyvis and Davidenko (2009) was included to screen out random clicking. Their proposed ‘blue dot test’, which is a “validated and well-established technique” (Stroebe, Postmes and Spears, 2012, p. 683), was slightly adapted as shown in Appendix H.4. Finally, the last block asked about respondents’ socio-demographic characteristics such as gender, age, home country, education and occupation. The full questionnaire can be found in Appendix H.

3.3.2 Pre-test

A pre-test among 18 people (44% female, 16% English native speakers, Mage=26, SD=7.14) confirmed that the manipulations in the levels of conversational human voice among the conditions had the intended effects. All three conditions significantly differed in their perception of conversational human voice. Further to that, follow-up interviews with selected participants resulted in some refinement of various wordings and formulations.

The most noteworthy result from the pre-test however is that the vast majority of respondents that was assigned to the ‘no response’ condition reported that they were highly confused about items intended to measure procedural justice and interactional justice such as “The employee who answered the post understood exactly the problem”, when there was no response from Emirates. The same partly applied to items from the conversational human voice scale, such as “Emirates uses a communication style that resembles actual conversations”.

3. Methodology

Consequently, two separate constructs were developed as shown in Table 3.4. The analysis section will indicate which construct was used.

Table 3.4: Conversational Human Voice Construct for Conditions

CHV (complete) CHV (reduced)

CHV_1 Emirates is open to dialogue Emirates is open to dialogue CHV_2 Emirates would admit a mistake Emirates would admit a mistake CHV_3 Emirates addresses criticism in an objective

manner Emirates addresses criticism in an objective manner

CHV_4 Emirates treats me and others as human Emirates treats me and others as human CHV_5 Emirates invites people to conversation Emirates invites people to conversation CHV_6 Emirates uses a communication style that

resembles actual conversations

CHV_7 Emirates tries to communicate in a human voice

CHV_8 Emirates tries to be interesting in communication

CHV_9* Emirates uses a sense of humour in communication

CHV_10* Emirates provides links to competitors CHV_11 Emirates attempts to make communication

enjoyable

* Item was later removed as a result of principal component analysis, see chapter 4.1.3

3.3.3 Sampling

3. Methodology

1961). Considering the time and money constraints of this project however, these drawbacks will be attempted to be mitigated by a relatively large sample size.

As the target population was social media users and in particular Facebook users, respondents were approached via a personal private message asking them to participate in the study. The message did not reveal what the study was about in order to limit biases stemming from demand characteristics. Receivers of the message were also asked to forward the questionnaire to family and friends, which is referred to as snowball sampling (Biernacki and Waldorf, 1981).

3.4 Ethical Considerations

4. Findings

4 F

4.1 Preparatory Analysis

4.1.1 Data Screening and Cleaning

The questionnaire was open for responses from October 28th _{until November 7}th 2016 and obtained 501 responses. In a first step, as recommended by Tabachnick and Fidell (2014) the accuracy of the data file was checked by proofreading all cases for potential mistakes. 96 respondents did not complete the questionnaire and were therefore removed listwise from the data set, as there was no pattern recognizable behind the missing values. 7 respondents indicated in the screening questions that they are not familiar with Facebook and were therefore removed. Lastly, 32 out of the 405 respondents who completed the questionnaire failed the instructional manipulation check described in chapter 3.3.1. This is a failing rate of 8%, which is in upper end of typical rates of highly inattentive responding which varies from 3-9% (Maniaci and Rogge, 2014). In result, a data set containing N=366 cases was obtained with cell sizes of 126 for condition A, 124 for condition B, and 116 for condition C. Even though conditions were presented equally, these deviations in total numbers resulted from the exclusion of cases as described beforehand.

4.1.2 Sample Characteristics

4. Findings

Table 4.1: Socio-Demographic Sample Characteristics

Condition A Condition B Condition C Total

N % N % N % N % Total Sample 126 34.4% 124 33.9% 116 31.7% 366 100% Gender Female 60 47.6% 62 50.0% 54 46.6% 176 48.1% Male 66 52.4% 62 50.0% 62 53.4% 190 51.9% Age Ranges 0-24 Years 42 33.3% 35 28.2% 27 23.3% 104 28.4% 25-34 Years 71 56.3% 82 66.1% 76 65.5% 229 62.6% 35-44 Years 5 4.0% 5 4.0% 5 4.3% 15 4.1% 45-54 Years 3 2.4% 0.0% 3 2.6% 6 1.6% 55-64 Years 5 4.0% 1 0.8% 4 3.4% 10 2.7% 65+ Years 0.0% 1 0.8% 1 0.9% 2 0.5% Education Bachelor’s Degree 59 46.8% 63 50.8% 48 41.4% 170 46.4% Master’s Degree 44 34.9% 37 29.8% 44 37.9% 125 34.2% Work-related Education 9 7.1% 13 10.5% 11 9.5% 33 9.0% Primary or High School 11 8.7% 6 4.8% 7 6.0% 24 6.6%

Other 2 1.6% 4 3.2% 4 3.4% 10 2.7%

Doctoral Degree 1 0.8% 1 0.8% 2 1.7% 4 1.1%

Occupation

Employed for wages 72 57.1% 68 54.8% 71 61.2% 211 57.7%

Student 40 31.7% 45 36.3% 32 27.6% 117 32.0%

Out of work (looking for work) 7 5.6% 5 4.0% 5 4.3% 17 4.6%

Self-employed 5 4.0% 3 2.4% 7 6.0% 15 4.1%

Housewife / househusband 2 1.6% 2 1.6% 0.0% 4 1.1%

Retired 0.0% 1 0.8% 1 0.9% 2 0.5%

4. Findings

Figure 4.1: Social Media Usage Frequency of Respondents

Table 4.2: Study specific Sample Characteristics

Condition A Condition B Condition C Total

N % N % N % N % Total Sample 126 34.4% 124 33.9% 116 31.7% 366 100% Facebook Account Yes 126 100% 121 97.6% 112 96.6% 359 98.1% No 0.0% 3 2.4% 4 3.4% 7 1.9% Complaint Experience Yes 94 74.6% 89 71.8% 88 75.9% 271 74.0% No 32 25.4% 35 28.2% 28 24.1% 95 26.0%

Familiarity with Emirates

Yes 113 89.7% 114 91.9% 104 89.7% 331 90.4%

No 13 10.3% 10 8.1% 12 10.3% 35 9.6%

Lastly, Table 4.3 shows the home country distribution of respondents. The three most represented countries were Germany (54.6%), United States of America (17.5%), and the Netherlands (7.9%). In total, Western Europe made up for 74.0% of the sample and North America 18.3%. According to Calder, Phillips and Tybout (1981) and Reynolds, Simintiras and Diamantopoulos (2003), homogeneous samples are desirable in consumer and marketing research when the goal is to test theory.

1,11% 0,84% 1,11% 10,58% 12,26% 59,61% 3,34% 7,52% 3,62% 0% 10% 20% 30% 40% 50% 60% 70% Once a month

4. Findings

Table 4.3: Respondents by Home Country

Condition A Condition B Condition C Total

N % N % N % N % Total Sample 126 34.4% 124 33.9% 116 31.7% 366 100% Western Europe 94 25.7% 90 24.6% 87 23.8% 271 74.0% Germany 68 18.6% 68 18.6% 64 17.5% 200 54.6% Netherlands 13 3.6% 5 1.4% 11 3.0% 29 7.9% United Kingdom 4 1.1% 8 2.2% 7 1.9% 19 5.2% Italy 3 0.8% 3 0.8% 0.0% 6 1.6% Spain 1 0.3% 1 0.3% 2 0.5% 4 1.1% Austria 2 0.5% 2 0.5% 0.0% 4 1.1% Switzerland 1 0.3% 1 0.3% 0.0% 2 0.5% France 2 0.5% 0.0% 0.0% 2 0.5% Belgium 0.0% 1 0.3% 1 0.3% 2 0.5% Sweden 0.0% 0.0% 1 0.3% 1 0.3% Finland 0.0% 0.0% 1 0.3% 1 0.3% Ireland 0.0% 1 0.3% 0.0% 1 0.3% Northern America 23 6.3% 23 6.3% 21 5.7% 67 18.3%

United States of America 23 6.3% 23 6.3% 18 4.9% 64 17.5%

Canada 0.0% 0.0% 3 0.8% 3 0.8%

Eastern Europe 4 1.1% 5 1.4% 2 0.5% 11 3.0%

Bosnia and Herzegovina 1 0.3% 2 0.5% 0.0% 3 0.8%

Czech Republic 0.0% 1 0.3% 1 0.3% 2 0.5% Hungary 0.0% 1 0.3% 1 0.3% 2 0.5% Slovakia 1 0.3% 0.0% 0.0% 1 0.3% Russian Federation 1 0.3% 0.0% 0.0% 1 0.3% Poland 0.0% 1 0.3% 0.0% 1 0.3% Montenegro 1 0.3% 0.0% 0.0% 1 0.3%

Asia and Near East 3 0.8% 3 0.8% 1 0.3% 7 1.9%

4. Findings

4.1.3 Variable Composition and Testing

Principle Component Analysis

In order to test the validity of the applied scales, a principal component analysis was conducted. An analysis with forced factor extraction showed that five components accounted for 79% of the cumulative variance, after two items of the conversational human voice construct were removed (see Appendix B.1). All three outcome measures, namely, word-of-mouth intention, purchase intention, and corporate image loaded on one component. Therefore, those were separately analysed in a subsequent analysis, which confirmed that all items then loaded on different components. A subsequent analysis of all three justice measures resulted in the removal of one item in the procedural justice scale in regards to the response time which is likely due to respondents not paying attention to when the complaint was posted and when it was answered.6

4. Findings

Cronbach’s Alpha

The constructs of the questionnaire’s items were tested for their reliability. All scales had a high level of internal consistency, as determined by Cronbach’s alphas between .829 and .969 shown in Table 4.4. Values above .7 are desirable (DeVellis, 2009). The full results can be found in Appendix B.3.

Table 4.4: Overview of Cronbach’s Alpha

Construct Cronbach’s Alpha

Conversational Human Voice (complete) .928

Conversational Human Voice (reduced) .917

Distributive Justice .897

Procedural Justice .829

Interactional Justice .900

Satisfaction with Complaint Handling .963

Corporate Image .917

Word-of-Mouth Intention .969

Purchase Intention .883

Variable Composition