The influence of customer loyalty on word of mouth : an agent and an observer perspective.

(1)

Master thesis

The Influence of Customer Loyalty on Word of Mouth:

an Agent and an Observer Perspective

Author: Francis Voet - 10458018

University of Amsterdam, Faculty of Economics and Business MSc Business Studies – Marketing track

Final version

Under supervision of: J. Demmers MSc Second assessor: Prof. Dr. W.M. van Dolen

(2)

The main topic of this study is the relationship between customer loyalty (CL), forms of word of mouth (WOM), brand attitude (BA) and the moderating role of CL on the latter two concepts. The two forms of WOM in this study are endogenous and exogenous WOM. The former is WOM generated by customers about their experiences with the core product and/or company; the latter is WOM generated by customers in response to tweets regarding specific online actions of companies. The e-banking business is the main context of this study. In an online survey, with a 3x3x3 mixed design experiment incorporated, respondents

participated in three separate experiments where each experiment related to one particular condition which contained positive, negative or neutral. The objectives of this research are to discuss how CL affects 1) the propensity to spread different forms of WOM, and 2) the impact of different forms of WOM on BA. The first part tests the ‘agent’ relationship, where the customer is regarded as the agent that engages in WOM. Results indicate that CL is only positively linked to endogenous neutral WOM intention. The second part tests the ‘observer’ relationship, and regards the customer as an observer of WOM. Results show that loyal customers have more favorable BA. There is also evidence that exogenous positive WOM (PWOM) is positively linked to BA. Endogenous PWOM scored the highest on BA and leads to significant interaction. CL has a moderating effect on the relation between exogenous PWOM, endogenous PWOM and endogenous neutral WOM and BA.

(4)

Chapter 1: Introduction

1.1 Change in communication

There are different platforms where people can complain, share their thoughts or talk about products and services. In the last ten years, there have been a lot of changes in regard to interpersonal communication (Sonnier, McAlister & Rutz, 2011). Instead of in person or over the telephone, conversations nowadays take place via new technologies, of which “the

popularity […] increases with consumers” (Sonnier et al., 2011, p. 702). Today, increasing numbers of people share online content with others via Internet-based social media that form an integral part of modern society.

The Internet was originally created as “a platform to facilitate information exchange between users” (Kaplan & Haenlein, 2010, p. 60). Once the Internet was established, various social networking applications were developed, including social networking sites like

Wikipedia (2001), Facebook (2004) and Twitter (2006), where people can exchange

information (Kaplan et al., 2010). These sites have been growing rapidly since their inception, and are now widely in use (Kwak, Lee, Park & Moon, 2010). These social online platforms allow people to have their own personal profiles where they can follow other users and are in turn followed by others (Kaplan et al., 2010). In the past few years, the concept of ‘social media’ has become a familiar phenomenon, where “participation, sharing, and collaboration” (Kaplan et al., 2010, p. 65) are key concepts of interpersonal communication. Forrester Research (Kaplan et al., 2010) has investigated ordinary users who make use of social media. They found that 75% of Internet surfers in the second quarter of 2008 used social media by joining social networks, contributing or reading reviews. Six types of social media can be distinguished, namely: blogs, content communications, collaborative projects, social networking sites, virtual game worlds and virtual social worlds (Kaplan et al., 2010). The

(5)

focus of this research is on the social networking site Twitter. The impact of social media use for customers and companies is discussed in the following sections.

1.2 Consequences for customers

Customers become more connected to one another through high-tech devices, such as smart phones and social media (PQ Media, 2009). Despite developments in technology, customers can be uncertain about the quality of products or services of a company before or after buying and using. Online recommendations, where customers document their

experiences and share these with others, are an important source for customers to gather information about a product or service and determine what a customer thinks. Liu defines word of mouth (hereafter WOM) clearly and simply as “informal communication among consumers about products and services” (2006, p. 74). Various empirical studies depict interpersonal communication, or WOM, as a potentially impactful communication channel and an effective source of information for consumers (e.g. Sonnier et al., 2011). East,

Hammond and Lomax state that WOM “is a powerful influence on consumer behavior” (2008, p. 215), e.g. brand choice. Casaló, Flavián and Guinalíu (2008a) have also shown WOM to have a dominant influence on consumer behavior. According to their research, customers spread positive WOM (hereafter PWOM) to fellow customers to encourage brand choice, in contrast to customers who spread negative WOM (hereafter NWOM) to discourage brand choice. Consequently, WOM can help customers in their decision making process when acquiring a product or service and in choosing a particular brand (e.g. Keller & Lehmann, 2006). Behaviors and attitudes of customers can be shaped by WOM communication and is “one of the most widely accepted notions in consumer behavior” (e.g. Brown & Reingen, 1987, p. 350).

East et al. state that “WOM is often the major reason for brand choice” (2008, p. 215), but it is unclear how PWOM and NWOM impact this affect. WOM may include how

(6)

(dis)satisfied customers are with their purchasing experiences (Brown et al., 1987), which may consequently influence their future purchases and willingness to tell other people about their experiences, either in a positive or negative way.

1.3 Consequences for companies

Companies can respond to technological developments by way of various means and actions. According to Kaplan et al., companies should interact with their customers and adopt an active attitude towards them (2010). Because of the increasing use of social media,

decision makers in companies are forced to think about how the company can make profitable use of it (Kaplan et al., 2010). Today, most companies are active in offline, online and social media that offer various ways to communicate with customers. Communication takes place via a two-way communication model, a process called “transactional communication”, which focuses on conversation and dialogue (Duncan & Moriarty, 1998, p. 4). The focus of this research is a close, interactive and long-term relationship (defined by a series of transactions) between a company and its customers (Duncan et al., 1998). For example, both parties post messages on the Internet and respond to each other’s posts.

In the past, companies could oversee all shared online content that regarded them. The companies had the control, knowledge and opportunity “through strategically placed press announcements, good public relations managers” (Kaplan et al., 2010, p. 60) and “strategic firm behavior” (Mayzlin, 2006, p. 156) to alter publicly posted information. However, companies nowadays have increasingly less or even no control over customers’ thoughts, feelings and shared online content that pertain to their businesses (Keller et al., 2006). Instead, companies are observers, “having neither the knowledge nor the chance – or, sometimes, even the right – to alter publicly posted the comments provided by their customers” (Kaplan et al., 2010, p. 60).

Both traditional (mostly offline) and new methods of communication (mostly online)

(7)

may generate WOM. It seems that communicating via social networking sites is less expensive and more efficient than traditional communication tools, such as the telephone, radio or print advertisements (Kaplan et al., 2010). Villanueva, Yoo and Hanssens (2008) show that WOM is a low cost channel for gathering and keeping customers. Customers spontaneously give feedback about their user experiences and companies’ actions. Another advantage of new technologies is that WOM can be spread faster and reach more people than traditional communication tools.

As several studies have shown (e.g. Chevalier & Mayzlin, 2006), WOM can affect consumers’ behavior. Therefore, it is important for companies to be aware of the

consequences of spreading WOM. Presently, there is little evidence concerning the effect of WOM in growing company performance over time (Trusov, Bucklin & Pauwels, 2009). Nevertheless, various influences of WOM have been found in the literature. For example, PWOM is likely to reduce promotional costs of the company (Sundaram, Mitra & Webster, 1998; Holmes & Lett, 1977). Also, it expedites the acceptance of the brand (Holmes et al., 1977). WOM, by influencing the decision making and buying process of customers, impacts future product-market results (Keller et al., 2006). Sonnier et al. (2011) found evidence that positive, negative and neutral WOM on the Internet have an impact on sales performance. Godes and Mayzlin (2004) and Keller and Lehmann (2006) confirm these findings and suggest that conversations about products and services encourage more sales and influence companies’ profits. According to research by Godes and Mayzlin (2009), companies attempt to stimulate WOM among existing and potential customers to create awareness and spreading PWOM. As Luo states: “WOM is one of the most effective tools for generating sales and future cash flows” (2009, p. 148).

In contrast, companies also face some negative consequences due to WOM, e.g. when customers are dissatisfied about their experiences of using a product or service. Luo (2009)

(8)

argues that the frequency of NWOM about a company directly influences the long-term financial prospects of the company, such as stock market activity (stock prices) and reduction of future cash flows. NWOM may be disadvantageous for the company because it discourages potential consumers to make use of a particular product or brand in a way that damages the financial position by reduced firm sales and may damage the reputation or image of the company (Holmes et al., 1977; Richins, 1983).

According to Rist, WOM is the most influential information source when making a purchase for customers of any age group (2005). By 2013, companies are expected to spend over 3 billion dollars on WOM marketing, which encourages customers to exchange

information about services and products via online and offline tactics (PQ Media, 2009 in Sonnier et al., 2011). Companies appreciate WOM and an increasing number of companies encourage customers to share their experiences and opinions with or without monetary incentives because they hope people will share their online feedback with others (e.g. Villanueva et al., 2008; Berger & Milkman, 2012). The customer is then considered as the agent that engages in WOM. In the efforts of companies to encourage WOM, there is an ongoing debate about whether companies should focus on loyal (Casaló et al., 2008a) versus less loyal (Godes et al., 2009) customers because little is known about how customer loyalty (hereafter CL) affects the propensity of customers to spread WOM.

1.4 Problem definition

This research will be performed in the context of the e-banking business and is about the way banks should handle online customer feedback on their Twitter accounts. The topic of this study is the relationship between CL, forms of WOM, brand attitude (hereafter BA) and the moderating role of CL on the latter two concepts.

The primary objective of this study is to investigate the relationship between CL and WOM. As described, the literature disagrees on whether a company should focus on more or

(9)

less loyal customers. Godes et al. (2009) state that it is more effective for companies to focus on less loyal customers to spread WOM, while Casaló et al. (2008a) suggest that companies should focus on high loyal customers because this is positively and directly related to higher levels of PWOM. According to Dick and Basu (1994), the link between loyalty and WOM has not been studied empirically. This study will research this important gap in the literature. The second objective is to examine how CL affects the impact of different forms of WOM on BA. Little research has been done to determine the way customers perceive the WOM posted by other customers or companies, the customer is then considered as an observer of WOM.

The research question in this study consists of two parts. The first part considers the ‘agent’ relationship, regarding the customer as the agent that engages in WOM and focuses on how CL affects the tendency of customers to spread different forms of WOM. The second part considers the ‘observer’ relationship, regarding the customer as an observer of WOM who is exposed to WOM of others (either the company or another customer) and attempts to explore how CL affects the impact of different forms of WOM on BA. This leads to the following research question:

“How does customer loyalty affect 1) the propensity to spread different forms of WOM, and 2) the impact of different forms of WOM on brand attitude?”

The research question can be divided into the following sub questions:

Agent relationship

1. What is the effect of customer loyalty on different forms of WOM?

Observer relationship

2. What effect does customer loyalty have on brand attitude? 3. What is the effect of different forms of WOM on brand attitude?

(10)

4. How does customer loyalty moderate the relationship between different forms of WOM and brand attitude?

These questions will ultimately estimate to what extent CL leads to the improvement of BA. Additionally, the first part will demonstrate the tendency to spread WOM relative to CL. The second part will measure the respective influence of CL on the impact of forms of WOM on BA. Understanding which form of WOM is sent by which customers can lead to a specific kind of BA, which can help companies avoid negative results. The outlined research- and sub questions are summarized in the two-step conceptual model below.

1. Agent relationship

Figure 1. Agent relationship

2. Observer relationship

Figure 2. Observer relationship

As described, the first model regards the customer as the agent that engages in WOM. The relationship between CL and forms of WOM is considered the ‘agent’ relationship. The

(11)

second model regards the customer as an observer of WOM. The influence of CL on the impact of different forms of WOM on BA is considered the ‘observer’ relationship.

1.5 Contributions

Growing interest in online business has raised critical questions among companies about what they can do to influence WOM and how employees should respond in order to improve or retain their business results and image. Empirical studies have given considerable attention in the marketing literature to CL, WOM and BA separately, but little is known about the relationship between these concepts. According to research by Sundaram et al. (1998), little attention is paid to the consequences of WOM. WOM has a bigger impact on reinforcing or forming attitudes when succeeding in reaching an individual (Engel, Kollat & Blackwell, 1968). The e-banking business is the main context of study because while there have been studies about CL and PWOM development, few studies have been conducted into the formation of these variables in the online business (Casaló et al., 2008a). Evidence found by Herr, Kardes and Kim (1991) suggests that WOM may affect pre-usage attitudes about a product; this study will investigate BA as consequence of WOM in the observer relationship. The two forms of WOM that can be distinguished in this study are endogenous and exogenous WOM. The former is WOM generated by customers about their experiences with the core product and/or company, and the latter is related to WOM generated by customers in response to specific online actions of companies (Godes et al., 2004, 2009). Little research has been done on the two different forms of WOM, but both forms will be explained in more detail in section 2.1.

This study will make several important contributions. There is a rising competition between companies who expand via the Internet and use social media sources to communicate with their customers. The primary objective of this research will be to contribute to scientific knowledge about the ‘agent’ relationship between CL and forms of WOM by investigating

(12)

whether (less) loyal customers will post positive, negative or neutral tweets regarding

customer experiences or in response to posted actions by their primary bank. The effect of the degree of CL on BA will be examined. Another contribution to scientific knowledge concerns the ‘observer’ relationship between forms of WOM and BA by analyzing the moderating role of CL. The two forms of WOM will be categorized to see whether and which conditions have an impact on BA and it will be examined whether CL affects the strength of the relationship between conditions of WOM and BA.

The source of WOM information discussed in this research is Twitter, a micro-blogging application that allows the user to broadcast text-based posts of 140 or less

characters (Kaplan et al., 2010). An interesting question for companies is whether they can get customers to talk about their products and services in such a way that it impacts company profits positively. From a managerial point of view it is important that a company receives feedback from participants who deliver a high benefit to the company (Godes et al., 2009). The results of this research have important implications for the strategy of companies to create CL and to know whether and on which form of WOM the company should focus to increase BA.

The rest of this research will proceed as follows. Chapter 2 will present the theoretical background to the hypotheses. Chapter 3 will describe the setup of the research design. In chapter 4 the results will be discussed. The research concludes in chapter 5 with a discussion of the results, limitations, possibilities for future research and a conclusion.

(13)

Chapter 2: Theoretical framework

This chapter includes an in-depth literature review concerning the variables relevant to this study. After the literature has been discussed, this chapter will list the hypotheses of the research and outline the conceptual models with the expected relationships.

2.1 Definition of variables Word of mouth

Chapter 1 includes a brief description of what is meant by word of mouth – WOM in short. According to East et al., WOM is a quick and interactive way of communication that consumers use to spread informal advice (2008). Oliver (1980) describes WOM as post-purchase evaluations by consumers to describe whether expectations have been met and whether they are satisfied after consumption. Arndt (1967) states that the WOM

communication process is valuable for customers for seeking social support in (not)

purchasing a product or service, as well as providing a risk decrement by group action. Bone (1995) and Richins (1984) describe WOM as a form of interpersonal communication where participants are no marketing sources (Bone, 1995), but customers who share their personal experiences about a product or a firm (Richins, 1984).

Marketing practitioners recognize the importance of the WOM communication process. Several authors confirm this and argue that this process is a very powerful tool and have a marked influence on people’s behavior (e.g. Bansal & Voyer, 2000; Buttle, 1998; East et al., 2008). This observation is strengthened by research by Taylor (2003, p.1), who states that “according to a McKinsey&Co. study, 67% of U.S. consumer-goods sales are based on word-of-mouth”. Consumers appreciate WOM because fellow consumers are seen as more honest than other information sources (Casaló et al., 2008a). Customers prefer to rely on

(14)

WOM because personal sources are considered as a more trustworthy and solid informal communication tool in making purchase decisions than more formal, organized information communication sources, such as advertising campaigns (e.g. Day, 1971; Bansal et al., 2000). Furthermore, WOM is considered a persuasive communication source where customers can share their comments anonymously (Mayzlin, 2006). Despite many consumers basing their choice on WOM, not every user spreads WOM after consumption. The next subsection will explain the concept of online WOM.

Electronic WOM

When consumers share WOM online it is called “electronic word of mouth

communication”, or eWOM for short (Gruen, Osmonbekov & Czaplewski, 2006). The biggest difference between eWOM and WOM is that eWOM is not “face to face, direct, oral and ephemeral” (Buttle, 1998, p. 243). Hennig-Thurau, Gwinner, Walsh and Gremler define eWOM communication as “any positive or negative statement made by potential, actual, or former customers about a product or company, which is made available to a multitude of people and institutions via the Internet” (2004, p. 39). Bickart and Schindler (2001) offer a clear distinction between WOM and eWOM: the former is about physically sharing product information with another through conversation; the latter is about posting personal

experiences and opinions about products online. Consumers share online WOM by giving their thoughts, opinion or review on products or services in different categories (e.g. music, films, television, books etc.). Online WOM can be useful to identify product defects,

consumers’ preferences or correcting mistakes (Jansen, Zhang, Sobel & Chowdury, 2009). In this research, eWOM is the main data collection source; this research will focus primarily on eWOM concerning bank services. Bank users are involved in online banking and can give their opinion after using a product or service of a bank.

(15)

Endogenous vs. exogenous WOM

WOM in general refers to the sharing of experiences after consumption of products or services of a company. In the literature however, WOM is categorized into two different forms: endogenous and exogenous. There are few studies that investigate the two different forms of WOM concurrently, but various studies have primarily focused on endogenous WOM (e.g. Godes et al., 2004, 2009). Godes et al. (2009) clearly define endogenous WOM as “characterized by conversations that occur naturally among consumers as a function of their experiences with the product” (p. 723) “and/or firm” (Godes et al., 2004, p. 5). Keller et al. (2006) define this individual experience as the feedback from product or service use and product satisfaction; the feedback covers what a consumer thinks, based on experiences with the core services and products of a company, and develops naturally. Research by Sundaram et al. (1998) shows that reasons of individuals to engage in WOM are considerably related to their consumption experiences. An example of endogenous WOM may be a tweet regarding a personal experience of a consumer with a company, e.g. feedback from product use and satisfaction (Keller et al., 2006). The above example is related to endogenous WOM because the customer spreads WOM about personal experiences with a core product or service of the company.

Exogenous WOM can be considered the opposite of endogenous WOM and refers to “WOM created as the result of [a] firm’s actions” (Godes et al. 2009, p. 723). Keller et al. (2006) suggest that exogenous WOM is related to the experiences of others (through WOM) and judgments of professionals. An example of exogenous WOM may be a tweet from a customer in response to e.g. a tweet regarding a promotional or innovative action from a company. The action of the bank can contain positive, negative or neutral sentiment. An example of a positive action is that the bank gives away a free product if the consumer saved a certain amount. An example of a negative action is that the bank still has not been able to

(16)

solve interferences with regard to the payment system. An example of a neutral action is that the bank explains the launch of a new payment function. Exogenous WOM includes positive, negative and neutral tweets from customers in response to an action from the bank.

This research will distinguish between two forms of WOM, based on the duality explained above: “endogenous WOM” and “exogenous WOM”. Various tweets from

customers will be discussed to clarify the difference where the former focuses on tweets about customer experiences while the latter focuses on tweets in response to companies’ actions. The next subsection will explain the concept of online sentiment.

Online sentiment

According to research by Anderson, every message contains a certain value of

sentiment and “the sentiment of word of mouth [may] be positive, neutral or negative” (1998, p. 6). Thus, WOM, online reviews or tweets can be evaluated and classified in positive (PWOM), negative (NWOM) and neutral subjective expressions (Cui, Mittal & Datar, 2006). According to research by Buttle (1998), PWOM is the result of a satisfactory balance between expectations and perceptions, where NWOM is an unsatisfactory balance between those feelings. In other words: PWOM arises when performance exceeds expectations and NWOM arises when performance disappoints users. Examples of PWOM, broadly defined, are “pleasant, vivid, or novel experiences: recommendations to others; and even conspicuous display” where examples of NWOM includes “behaviors such as product denigration, relating unpleasant experiences, rumor, and private complaining” (Anderson, 1998, p. 6). According to research by Balahur and Steinberger (2009), it is not necessary to find examples of neutral WOM because it functions as the default. Sundaram et al. (1998) show the motives why customers spread WOM with a certain sentiment. According to them customers engage in PWOM for “altruistic, product involvement, and self-enhancement reasons” and in NWOM

(17)

for “altruistic, anxiety reduction, vengeance, and advice seeking reasons” (1998, p. 527). This research will use three variables of online sentiment: besides positive and negative tweets, neutral examples will also be investigated. The first two cases of sentiment are subjective, whereas the third case of sentiment is non-subjective (Balahur et al., 2009). In this research, the goal of online sentiment analysis is to discover the opinion of an individual about a certain product, service or action of a bank company. These opinions are used as examples for the three experiments in the online survey. Sentiment analysis of online content is called ‘sentiment’, ‘affect’ or ‘polarity’ classification (Cui et al., 2006). The impact of the sentiment of WOM will be discussed in the next subsection.

Impact PWOM, NWOM and neutral WOM

As described, WOM can have a strong impact on customers’ behavior and on

companies’ performances. It seems obvious that people who make use of a product or service exercise the most influence on non-users and have a great impact on whether a product or service will be a success or failure. Besides, it seems logical that a customer gains a positive attitude towards a product, service or company after receiving PWOM, and a negative attitude after receiving NWOM. According to O’Malley (1998), this presumption is superficial, since it is not possible to assume that satisfied customers will stay loyal and dissatisfied customers will not. As described, WOM is often a primary motivation for brand choice where PWOM encourages brand choice and NWOM discourages brand choice (East et al., 2008). Arndt (1967) found evidence for the findings that both PWOM and NWOM have a powerful impact on the behavior of consumers; PWOM increased the probability of purchase, where NWOM decreased the probability of purchase of products or services. East et al. (2008) found

evidence in their research regarding familiar brands that participants resist NWOM on brands they are very likely to adopt, and resist PWOM on brands they are unlikely to adopt.

(18)

Conversations about a product or service have an impact on evaluations of the usage and affect choice behavior by influencing the attitude of consumers and their purchase decisions (Bone, 1995).

It remains unclear whether either PWOM or NWOM is more influential. There is little evidence put towards the impact of PWOM compared to NWOM, possibly because it is difficult to make precise measurements in this field (East et al., 2008). However, different studies found contrary results regarding this supposition. According to East et al. (2008), the impact of PWOM is greater than NWOM in the context of familiar brands. A recent study by Nguyen and Romaniuk (2012) builds on the research of East et al. (2008) and concludes that PWOM is more influential than NWOM for TV programs. Conversely, several authors found opposite results, stating that the impact of negative information is greater than positive information and that NWOM is therefore more influential than PWOM (e.g. Arndt, 1976; Ahluwalia, Burnkrant & Unnava, 2000; Chevalier et al., 2006; Assael, 2004). For example, Arndt (1976) and Ahluwalia et al. (2000) found that NWOM is more effective and impactful on the decision making process than PWOM and can be considered more useful by consumers. Due to the influence of WOM on customers and companies, the latter should stimulate

PWOM about their products (Chung & Darke, 2006).

Customer loyalty

Annually, companies lose 15 to 20 percent of their customers; when a company retains 5 or more percent of its customers, profits can increase by almost 100 percent (Reichheld & Sasser, 1990). Companies should do their best to create and satisfy loyal customers by “[tracking] the ones [they] lose, to learn how to keep customers” (Reichheld et al., 1990, p. 105). O’Malley (1998) described the focus on customer retention as loyalty marketing. Edvardsson, Johnson, Gustafsson and Strandvik (2000) investigated the relationship between

(19)

forms of firms’ loyalty and their effect on company performance. They found evidence that the relative effect of loyalty is positive for service companies, where the presumption is that companies must earn their loyalty. Reichheld et al. (1990) state that the longer the relationship between a customer and the company lasts, the more profits increase. For product companies however, loyalty can have a negative effect on company performance because these

companies maintain their customers and achieve CL by lowering their prices (Edvardsson et al., 2000). Dick et al. (1994) state that sustainable competitive advantage could be created by CL. Loyalty may be established in different ways; for example, by considering a company as the first choice over others, by buying products and services from this company repeatedly, or by doing more business with the company in the future (Zeithaml, Berry & Parasuraman, 1996).

Loyal customers are valuable to companies because they repeatedly buy a number of products or services of the same brand. Loyalty may be defined as “a consumer’s

predisposition to repurchase from the same organization again” (Edvardsson et al., 2000, p. 918) because a consumer has a favorable attitude towards the brand or organization (Dick et al., 1994). Casaló, Flavián and Guinalíu (2008b, p. 328) define loyalty as “a non-random behavior, expressed over time, which depends on psychological processes and closeness to brand commitment”. CL is defined by Dick et al. as: “the strength of the relationship between an individual’s favorable attitude that is high compared to potential alternatives and repeated patronage are required” (1994, p. 99-100). Behavior is called loyal when WOM develops from customers’ belief that “the quantity of value received from one supplier is greater than that available from other suppliers” (Hallowell, 1996, p.28). Thus, CL can be considered as the relationship between the favorable attitude toward an entity (core product or service of a particular bank in this research) and repeat purchase (Dick et al., 1994). Loyalty is called high

(20)

when the customer has a high repeat purchase rate and a favorable attitude or strong preference for a brand (e.g. O’Malley, 1998; Dick et al., 1994).

The figure below shows how attitude and repeat patronage translate to loyalty.

Figure 3. Four specific conditions related to loyalty (Dick et al., 1994).

The relationship between favorable attitude and repeat patronage is mediated by situational factors and social norms. For example, situational factors such as promotion actions of competitive brands or lack of variety may have an impact on decreasing loyalty. In 1980, Ajzen and Fishbein formulated the theory of reasoned action. This theory is based on the assumption that “human beings are usually quite rational and make systematic use of the information available to them” where “people consider the implications of their actions before they decide to engage or not engage in a given behavior” (1980, p. 5). A person’s behavior is determined by the intention to perform the behavior, which is in turn determined by two factors. The first one, attitude toward the behavior, relates to the “individual’s positive or negative evaluation of performing the behavior” (1980, p. 6). The second factor is social influence, which is referred to the ‘subjective norm’ in the literature, and posits the notion that “individuals will intend to perform a behavior when they evaluate it positively and when they believe that important others think they should perform it” (1980, p.6).

There are several forms of loyalty distinguishable in the literature (e.g. Dick et al., 1994), such as service loyalty (services), vendor loyalty (industrial goods), store loyalty (retail

(21)

establishments) and brand loyalty. Most marketing studies on loyalty have focused on brand loyalty, where customers repeatedly purchase packaged goods (Dick et al., 1994). In general, loyal customers may be seen as those who repeatedly purchase a product or service because they are committed to the brand and are willing to pay more for a known and trusted service or product (Reichheld et al., 1990). Assael (1992) states that loyalty is not just about repeat purchase; commitment is equally important. According to Godes et al. (2009), highly loyal customers like to live in social networks where “others are also loyal to the firm” or “others are aware of but not interested in the firm’s products” (2009, p. 722). According to various authors, CL can be divided into two dimensions: recommendation and patronage, i.e. repurchase intention (e.g. Lam, Shankar, Erramilli & Murthy, 2004). According to research by Gronholdt, Martensen and Kristensen, recommendation is the “intention to recommend the brand/company to other consumers” where repurchase intention is “the customer’s intention to repurchase” (2000, p. 511). Individuals can be considered as more or less loyal customers, depending on their willingness to recommend and purchase intentions (Zeithaml et al., 1996). It is remarkable to note that recommendation plays a central role in both CL and WOM. A practical example in regard to CL is described by Casaló et al. (2008a), which shows that consumer satisfaction with previous communications of the e-banking service website is linked to CL to this website.

Only a fraction of customers are 100% loyal to a single brand and full loyalty seems hard to achieve (O’Malley, 1998). The majority of consumers have a few (two or three) brands within any product category that they are loyal to and which they regularly buy, which is called polygamous loyalty (Dowling & Uncles, 1997; O’Malley, 1998). Polygamous loyalty is reflected in different product- and service markets such as breakfast cereals and soft drinks, but also in business airline travel and fast-food outlets (Dowling et al., 1997).

Prior research (Chaudhuri & Holbrook, 2001; Jacoby & Kyner, 1973; O’Malley, 1998)

(22)

has shown that a distinction can be made between loyalty as a behavioral or attitudinal measure. Behavioral – or purchase – brand loyalty is characterized by repeated purchases of the brand (Chaudhuri et al., 2001; Godes et al., 2009) or intentions to repurchase (Hallowell, 1996); frequency and recentness of purchase are important indicators of behavioral brand loyalty (O’Malley, 1998). Casaló et al. (2008b) describe attitudinal measure as affective loyalty that includes a psychological link. According to Hallowell (1996), this measure is based on consumer feelings that create an attachment to products or services; the degree of CL is based on these raised feelings. Attitudinal brand loyalty contains “a degree of

dispositional commitment to the brand in terms of some unique values associated with the brand” (Chaudhuri et al., 2001, p. 82). Opperman (2000) states that it is hard to measure the psychological aspects of loyalty, which is why this study will examine behavioral measure of loyalty, rather than attitudinal loyalty.

Brand attitude

Before explaining the concept of BA, attitude in general will be discussed first. Mitchell and Olson define attitude as “an individual’s internal evaluation of an object such as a branded product” and is considered as a “useful predictor of consumers’ behavior toward a product or service” (1981, p. 318). Berger and Mitchell define attitudes as “evaluations of a product or brand” (1989, p. 269). However, consumers can frequently purchase a product or service while not having a favorable attitude toward a brand. Or vice versa: consumers can have a favorable attitude toward a brand but not purchase their products or services repeatedly because they prefer an equivalent brand or have a higher favorable attitude toward other brands (Dick et al., 1994). Oliver (1981) states that attitudes of consumers are relatively enduring and are less situational focused.

This research focuses on evaluations of customers about their primary bank company.

(23)

The concept BA is defined by Mitchell et al. as the “beliefs about attributes of the advertised brand” (1981, p. 320). Based on above definition, BA can be defined as “a consumer’s overall evaluation of the brand” (Faircloth, Capella & Alford, 2001, p. 62). According to research by Berger et al. (1989) BA is a very well-known and important research topic in consumer behavior because it often forms the basis for consumer behavior (Keller, 1993). However, many studies have failed to find evidence that “brand attitude consistently predicts purchase intentions or behavior” (e.g. Faircloth et al., 2001, p. 64).

According to research by Olson and Mitchell (2000), BA may have an impact on ratings of belief strength. Park, Maclnnis, Priester, Eisingerich and Iacobucci define BA strength as “the positivity or negativity of an attitude weighted by the confidence or certainty with which it is held” (2010, p. 1), which suggest that BA can be positive or negative.

2.2 Hypotheses

This section will define all hypotheses of this study. A remark that must be made is that since little research has been done on the two different forms of WOM, it is difficult to describe expectations concerning these variables. In this study, it is assumed that the tweet about a positive action of the company leads to positive sentiment, the tweet that contain a negative action of the company leads to negative sentiment and the tweet that contain a neutral action of the company leads to neutral sentiment. Therefore, H2 and H5 focus only on the results regarding exogenous PWOM (PWOM after a tweet about a positive action of the company), exogenous NWOM (NWOM after a tweet about a negative action of the company) and exogenous neutral WOM (neutral WOM after a tweet about a neutral action of the

company). Because of the sequence of the manipulations in the survey, the agent relationship focuses first on endogenous WOM and then exogenous WOM; in the observer relationship exogenous WOM is discussed first, endogenous WOM second.

(24)

Agent relationship

The hypotheses toward this relationship test the effect of CL on the tendency of (less) loyal customers to spread different forms of WOM. As described above, it is difficult to describe expectations toward this relationship because the two forms of WOM have hardly been investigated. A direct relation between CL and the two forms of WOM manifest itself either positively or negatively. The assumption is that both forms of WOM will contain either positive, negative or neutral sentiment.

As described, highly loyal customers are likely to live in an environment where others are also loyal to the company, or where others are aware of the products of the company but not interested themselves (Godes et al., 2009). However, it is not obvious that highly loyal customers spread important WOM. According to research by Godes et al. (2009) the most loyal customers do not always deliver the most impactful WOM. Their research confirmed that it may be more meaningful to target less loyal customers “in terms of their ability to create meaningful, incremental WOM” because this is more impactful for the company than when highly loyal customers do it (2009, p.728-729). They found that for a product with a low initial awareness level, WOM from a less loyal customer that occurs between strangers has a bigger influence at driving sales. There are two strategies distinguishable about

spreading WOM through customers which may affect company sales: a company could adopt NBC’s strategy for spreading WOM through its loyal customers, or Lee Dungarees’ strategy for spreading WOM through less loyal customers. The former is often preferred in the business press (Godes et al., 2009). However, according to Godes et al. (2009), targeting less loyal customers who spread WOM is most effective for a company because “their

recommendations are more likely to be received by people who are currently less experienced with or less informed about the firm’s products” (p. 737). This idea is based on the

supposition that “less loyal customers are effective because their network is more likely to

(25)

consider information about the product to be new” (Godes et al., 2009, p. 734). However, the expectation is that these results will not persist when the customer becomes increasingly loyal the product (Godes et al., 2009).

It is clear that CL measures the likelihood of a customer repurchasing a service or product. Several authors have a different opinion regarding the reasons why customers spread WOM. According to research by Shoemaker and Lewis (1999), customers spread WOM to bring others to a product or service, where loyal customers are more willing to spread WOM. Sundaram et al. show that customers spread WOM for personal reasons and are “related to consumption experiences” (1998, p. 1). For example, one of the reasons to spread PWOM is involvement with the product, whereas NWOM is spread in the spirit of revenge. According to research by Mangold, Miller and Brockway (1999), there are two reasons for people to spread WOM. The dominant reason for WOM communication is that individuals need advice about products and services. The second reason is that information about products and

services arise during a dialogue. In this study ‘input WOM’ refers to WOM which can be accessed as an information source for purchase, where ‘output WOM’ refers to WOM after buying or consumption experience (Buttle, 1998). The agent relationship focused on ‘output’ WOM and the observer relationship on ‘input’ WOM. Hennig-Thurau et al. state that

customers spread eWOM due to their “desire for social interaction, desire for economic incentives, their concern for other consumers, and the potential to enhance their own self-worth” (2004, p. 39).

The importance of CL is not conclusively defined in any of the literature and there is no clear answer on whether companies should focus less or more on loyal customers for creating endogenous or exogenous WOM with a particular sentiment. In this research, customers may spread endogenous WOM about their personal experiences with a core product or service of their bank company and occurs naturally. Customers may spread

(26)

exogenous WOM about tweets regarding actions of their primary bank company (Godes et al., 2009).

The next hypotheses examine which form of WOM is created by (less) loyal

customers. This will help companies determine whether they should focus on two of the main objectives of online managers, namely “the development of customer loyalty and positive word-of-mouth” (Casaló et al., 2008a, p. 399). Generally, loyal customers spread PWOM and share their experiences with other online users (Bowen & Chen, 2001). Chi and Qu (2008) found that customers, who are likely to purchase a product or service because they have satisfied feelings about it, are more eager to spread PWOM. Casaló et al. (2008a) found a positive relationship between CL and PWOM. They additionally found support for their hypothesis that “greater customer loyalty is directly and positively related to greater levels of positive WOM with respect of a website in the e-banking business” (p. 404-408). Furthermore, they found that perceived CL is a key determinant of PWOM development in the e-banking context.

Based on what is described above, the following expectations are made. It seems that less loyal customers have more intention to spread endogenous WOM since they buy and use products or services of the company less often and will be more surprised by their experiences. It is expected that loyal customers are less likely to post something about a positive or

negative experience with the bank and more likely to post something about a neutral experience with the bank than less loyal customers.

Since loyal customers are more likely to repeatedly buy the same products and services in the future it is expected that they will have more intentions to spread exogenous WOM than less loyal customers because they want to help the company. Thus the expectation is that loyal customers are more likely to respond to positive and neutral tweets of the bank and are less likely to respond to negative tweets of the bank than less loyal customers. This

(27)

results in the following hypotheses:

H1a. Negative relation between CL and endogenous PWOM intention H1b. Negative relation between CL and endogenous NWOM intention H1c. Positive relation between CL and endogenous neutral WOM intention H2a. Positive relation between CL and exogenous PWOM intention

H2b. Negative relation between CL and exogenous NWOM intention H2c. Positive relation between CL and exogenous neutral WOM intention Observer relationship

The hypotheses toward this relationship test how CL affects the impact of different forms of WOM on BA. First, the direct relationship between CL and BA is examined. Highly loyal customers repeatedly buy a number of products or services of the same brand because they have a positive attitude towards the organization or brand (Dick et al., 1994). According to research by Zeithaml et al. (1996), customers are considered highly loyal when they have the willingness to recommend and have high purchase intentions because they have a positive attitude toward a brand (O’Malley, 1998; Dick et al., 1994). Therefore, this study proposes a direct relationship between CL and BA. Based on what is described above, it is expected that loyal customers have more favorable BA. This result in the following hypothesis:

H3. Positive relation between CL and BA

The next hypotheses test the influence of the two forms of WOM on BA. As described above PWOM encourages brand choice; it increases the probability of customers purchasing

(28)

products or services, where the reverse is true for NWOM (East et al., 2008; Arndt, 1967). Besides, Keller et al. (2006, p.754) state: “both personal experience (feedback from use and product satisfaction) and the experience of others (through word of mouth and ‘expert’ ratings) also determine what a customer thinks of a brand”. According to research by Oliver (1980), WOM has an impact on the attitude of the customer after purchase. It seems logical that customers gain a favorable BA after receiving PWOM, where the reverse seems true for NWOM. Mitchell et al. (1981) associated advertisements of brands with visual stimuli. The two brands associated with an image of a kitten and a sunset over the ocean were more

positively evaluated and generated more positive BA than the brand association with a neutral abstract painting. Arndt (1967) found evidence that PWOM creates a positive attitude towards the firm and the brand. Mitchell found evidence for these assumptions and states that

sentiment affects BA, where a more positive BA is formed when the advertisement includes positively or neutrally evaluated photographs. The reverse was true for photographs with negative sentiment (Mitchell, 1986). Based on above, it is expected that people who see exogenous and endogenous positive or neutral WOM will have more favorable BA than people who see exogenous or endogenous NWOM. This lead to the following hypotheses:

H4a. Positive relation between exogenous PWOM and BA H4b. Negative relation between exogenous NWOM and BA H4c. Positive relation between exogenous neutral WOM and BA H5a. Positive relation between endogenous PWOM and BA H5b. Negative relation between endogenous NWOM and BA H5c. Positive relation between endogenous neutral WOM and BA

(29)

In the previous two hypotheses (H4 + H5) the effects of different forms of WOM

on BA are tested. The question that is raised is whether these results would differ for less or more loyal customers. Hypothesis 6 tests whether CL is a moderator for the effects of different conditions of the forms of WOM on BA, i.e. whether different conditions of the

forms of WOM have a stronger or weaker effect on BA for loyal customers. Based on previous hypotheses it is expected that both positive and neutral conditions of either form of

WOM have a stronger effect on BA for loyal customers. A weaker effect on BA for loyal customers is expected for the negative conditions in both forms of WOM. This lead to the following hypothesis:

H6. The effects of different forms of WOM on BA are moderated by CL

2.3 Conceptual models

The outlined hypotheses are summarized in the conceptual model below. This study presents a conceptual model from the perspective of the individual customer. This study tries to explain the relationship between several variables. In the first part of the research question, where the agent relationship is measured, the independent variable is CL and the dependent variable is the two distinguished forms of WOM.

(30)

1.AAgent relationship

Figure 4. Expected relative agent relationships

In the second part of the research question, where the observer relationship is measured, the independent variable are the two distinguished forms of WOM and the

dependent variable is BA. The moderating variable is CL that may moderate the effect of the two distinguished forms of WOM on BA. This moderator is chosen to test which conditions of the forms of WOM have a stronger or weaker effect on BA for loyal customers. Aside from the research questions, the second relationship will also examine the direct effect of CL on the dependent variable BA.

(31)

2.OObserver relationship

Figure 5. Expected relative observer relationships

Chapter 3: Data and method

This explanatory research will analyze quantitative data gathered through an online survey. This method of study was chosen because of the main advantages of an Internet questionnaire: low costs and easily accessible information. All variables in this research are operationalized by several scales and investigated via survey questions in order to clarify relationships among variables. The questionnaire was made in Qualtrics, an online survey platform that directly exports the results to SPSS (Qualtrics, 2013). The questions are based on the primary bank of the respondent. Some research was conducted into the social media

(32)

platform Twitter to find relevant examples of tweets that distinguish between the forms of WOM. These examples are included in the survey as experimental stimuli.

3.1 Method

The online survey focused first on the ‘agent’ relationship where the effect of CL on endogenous- and exogenous WOM intentions is examined. The second part of the research focused on the ‘observer’ relationship where the influence of CL on the effect of both forms of WOM on BA is studied.

Stimuli

Because eWOM plays a central role in this study, research into the social media network Twitter was conducted in order to find examples of endogenous and exogenous WOM. The examples were gathered via Twitter Analytics, a program that analyzes tweets from the social networking site Twitter and moves them directly to Excel. The subsequent influx of data is then evaluated by the Excel Add-in PowerPivot; this program gathers 1500 results per query per day and performs ad-hoc analyses of tweets by day, hour, tweeters, hash tags and @mentions (Analytics for Twitter, 2013). Endogenous tweets were collected via the twitter accounts @ING and #ING where customers have conversations that directly convey their experiences concerning the services and products of their primary bank. Exogenous tweets were collected via @INGnl_webcare, where tweets are posted as a response to tweets by their bank. The webcare team responds to questions, complaints and customer problems regarding the products and services online and announces actions (such as promotional actions) via this account (Twitter, 2013).

(33)

Design

The experiment used a 3x3x3 mixed design. Respondents were divided in order to participate in three separate experiments where each respondent was in each experiment assigned to one particular condition, either positive, negative or neutral. In this research three manipulations were performed: firstly, respondents were shown a scenario about an

experience with their primary bank and were asked about the chances of them posting a tweet about this scenario (endogenous WOM) to test the ‘agent’ relationship. Secondly, they were shown a tweet describing an action of the bank to measure whether respondents would retweet this post and to measure the chances of them posting a positive, negative and neutral message about this post (exogenous WOM) to test the ‘agent’ relationship and finally they were asked how this influenced their BA. These manipulations measure whether and which sentiment (less) loyal customers spread, relating to endogenous or exogenous WOM. Furthermore, the second manipulation measures the relationship between exogenous WOM and BA. Thirdly, they were shown a tweet regarding a random customer’s experience

(endogenous WOM) to measure whether respondents would retweet this post and to measure how this influenced their BA to test the ‘observer’ relationship. All examples contained positive, negative or neutral sentiment.

The proposed group was 30 respondents per condition; because each respondent answered the questionnaire tailored to one of the conditions, the research required a minimum of 90 respondents. A total number of 213 valid responses were collected; the respondents were randomly assigned to the different questionnaires. In experiment 1, 69 respondents were exposed to a positive scenario, 77 to a negative scenario and 67 to a neutral scenario. In experiment 2, 65 respondents were exposed to a tweet regarding a positive action of the company, 76 to a tweet regarding a negative action of the company and 72 to a tweet

regarding a neutral action of the company. In experiment 3, 64 respondents were exposed to a

(34)

positive tweet, 90 to a negative tweet and 59 to a neutral tweet.

These experiments were conducted in order to test which form of WOM (less) loyal customers spread, and how CL affects the impact of different forms of WOM on BA. The order of the questions of the experimental tweets in the survey was randomized to control for order effects.

Pretests

For the experiments, twenty fellow university students and relatives were asked to test the different examples of tweets. Respondents were exposed to affect-laden tweets of the two forms of WOM. The test was derived from methods as seen in Mitchell’s research, who had respondents selecting between the positive, negative and neutral experimental photographs (1986). Respondents were requested to rate each tweet as positive, negative or neutral.

Initially, nine tweets were pretested for endogenous WOM, consisting of three tweets for each sentiment. For exogenous WOM respondents were asked to evaluate posted tweets by their primary bank, consisting of two tweets for each sentiment. The criteria for selecting the tweets were that the content should contain understandable information and that the tweet should be clearly posted by either a customer or the bank. The questions and results from the pretest of the experimental tweets can be found in appendix I and II.

Ten fellow university students, relatives and acquaintances conducted another pretest to test the questionnaire. These testers were asked whether the questions were clear,

understandable, unambiguous and attractive, and whether the data was valid and useful in generating reliable and correct answers. To encourage a high level of response for completing the questionnaire, the introduction assures respondents that the questionnaire will take a maximum of 5 minutes to complete. Besides that, it explains that all answers will be

processed anonymously and that all data will be used for research proposes only. Those who

(35)

were interested in the outcomes and the resulting implications for existing research could sign up for a summary of the results to be sent to them via e-mail. To increase the validity of the research, the most negative answer option was the leftmost fill option and the most positive answer option was the rightmost fill option of the Likert scale. To reduce the chances of missing values, each question was outfitted with a force option, so that respondents could not continue to the next page before answering the preceding one. Furthermore, the beginning of the survey described what Twitter is; respondents who did not have an account were asked to imagine having one when answering the questionnaire.

3.2 Data collection

The self-administered questionnaire focuses on respondents aged between 18 and 44 who are customers of a bank because most Twitter users range between these ages according to research by Honey and Herring (2009). The questionnaire was distributed to 398

respondents via e-mail and Facebook: the invitation contained a URL link to the online survey. After a week a reminder was sent to thank the people who filled out the questionnaire, while those who had not yet done so received a request. The data collection process took three weeks during June 2013.

3.3 Procedure

The online survey was divided into five sections where the first section consisted of questions relating to CL and BA. Sections 2 to 4 contained the randomized manipulations and consisted of questions, which were mainly related to the two forms of WOM. The sentiment of the examples depended on randomization. Section 5 consisted of questions relating to demographic information.

The second section consisted of a question about the ‘agent’ relationship between CL and endogenous WOM intention. The question measured the chances of respondents posting a

(36)

tweet about a described scenario of an experience with their primary bank. This question was based on the idea that the scenario is recognizable to the respondent. The third section focused on the ‘agent’ relationship between CL and exogenous WOM intention and the ‘observer’ relationship between exogenous WOM and BA. Respondents were shown a posted tweet by the bank. Next, they were asked about their attitude towards the tweet and the chances of their retweeting the message. Exogenous WOM was measured by asking respondents the chances of them sending a tweet with a certain sentiment in response to the tweet by the bank. The last question of the third section concerned BA. The fourth section focused on the ‘observer’ relationship between endogenous WOM and BA. This section started with a posted tweet by a customer of the bank about their experiences. The next questions asked respondents for their attitude towards the tweet, retweet chance and BA. In all the answers, higher numbers represented stronger agreement with the item.

In the last section, respondents were asked to answer several questions to determine demographic variables. These variables included questions about gender, age, level of education, daily use of the Internet and social media, and whether the respondent had a Twitter account. The following section will discuss the different scales.

3.4 Measures

The questionnaire was based on scales from previous research, where some scales were adjusted to match the characteristics of the variables in this survey in order to measure the relevant variables. To test all hypotheses an online survey with multiple-item

measurement scales was developed (see appendix III). Most questions were based on Likert item scales, namely 5 or 7 scale points that allow for “finer discriminations” (Fornell, 1992, p.13).

(37)

Customer loyalty

In this research, CL is the independent variable in the agent relationship and the moderating variable in the observer relationship. This research utilizes the five-item scale for CL, adapted from Zeithaml et al. (1996), who operationalized the variable by the indicators ‘recommendation’ and ‘purchase intention’. The first three items are related to

recommendation and the last two measure behavioral purchase intentions. All items are based on a 7-point likelihood scale (1 = not at all likely, 7 = extremely likely).

Brand attitude

BA is the dependent variable in the observer relationship and measures the attitude toward the bank company and was measured along seven bipolar scale endpoint evaluative scales. The first scale to measure BA is based on the four 5-point scales of Olson et al. (2000) (‘Good-bad; Dislike very much-like very much; Pleasant-unpleasant; Poor quality–high quality’). Gardner (1985) used the first three bipolar evaluative scales as a measure of attitude toward the brand as well. The second scale is based on research conducted by Yoo and

Donthu (2001) and consists of five items (‘Nice-awful; Attractive-unattractive; Desirable-undesirable; Very bad-very good’; Extremely likable-extremely unlikable’). The last two items are similar to two items of the first scale, so these are excluded from the survey.

Endogenous and exogenous WOM intention

The two forms of WOM are the dependent variables in the agent relationship and constitute the independent variables in the observer relationship. The following scales are designed to make endogenous and exogenous WOM measurable.

(38)

Chance of posting a tweet

To measure endogenous WOM intention, respondents were asked to indicate the chances of them posting a tweet about the given scenario, based on a 5-point Likert item scale (1= very unlikely, 5 = very likely).

Exogenous WOM intention was measured by asking respondents whether they would send a positive, negative and neutral tweet in response to a tweet by the bank. The three items are based on a 5-point Likert item scale (1 = very unlikely, 5 = very likely).

Other measured variables Attitude toward the tweet

This research will examine the following variable, which is not included in the

conceptual model, but is related to the observer relationship. After a tweet is shown regarding a tweet by a bank and a customer’s experience, a question is asked to measure whether (less) loyal customers have a different attitude toward the tweet.

The attitude toward the tweet was measured by using bipolar items, based on two scales. These scales are nominally used to measure attitudes towards advertisements, but in this study it served to measure attitudes towards tweets. Respondents’ attitude toward the tweet was measured by six five-point bipolar scales. The first scale was based on research conducted by Mitchell et al. (1981) and Gardner (1985) who used four evaluative scales to measure attitude toward advertisements (Good-bad; Dislike-like; Not irritating-irritating; Uninteresting-interesting). The second scale is based on work by MacKenzie and Lutz (1989) and consists of three items. The first item (Good-bad) has overlap with the first scale, so only the last two items are used (Pleasant-Unpleasant; Favorable-unfavorable).

(39)

Retweeting the tweet

This variable, which is not included in the conceptual model, is also related to the observer relationship. This variable asked respondents to estimate the chances of them

retweeting (repeating a tweet from someone else) a tweet related to a posted tweet by the bank and a tweet related to a given customer’s experience to test which sentiment of tweets would be retweeted. All items are based on a 5-point Likert item scale (1 = very unlikely, 5 = very likely).

After the results were gathered and assembled as a dataset, all hypotheses were tested. Before testing, all variables were assessed on their quality and reliability. First, missing data was excluded and a reliability analysis was conducted using Cronbach’s alpha, which should be larger than .70. After scale means were computed, the hypotheses were tested by

performing linear regression analyses on significant results with a significant level with α = 0.05. The next chapter describes all results of the data analysis from the survey and explains the results of the tests.

Chapter 4: Results

This chapter reports the results of the data analysis of the conducted online survey. First, the demographic statistics are described. Then the results of the hypotheses tests are explained.

4.1 Demographics

A total number of 398 respondents were asked to complete the questionnaire; 276 of them completed the online survey, yielding a response rate of 60.3%. This response

(40)

constitutes a big enough sample size to serve as representative distribution of the population. Out of these responses 63 were incomplete; 37 respondents answered only the first question and 26 respondents did not answer the demographic questions. Some only answered questions related to the first and/or second experimental condition. The final sample consisted of 213 valid responses, which were included in the analysis to test the hypotheses.

The collected data showed that 62 percent of the respondents were female and 38 percent were male. The age of the respondents ranged from 18 to 68; 76% were aged between 21 – 26 years. The majority of the respondents (51.2%) indicated a university or otherwise scientific education as their highest level of education, 22.1% secondary school, 20.2%

Higher Vocational Education, 5.6% Intermediate Vocational Education and 0.9% PhD Degree. The primary bank of most respondents was Rabobank (57.7%); 21.6% ING, 14.1% ABN AMRO, 0.9% SNS bank and 5.6% had another primary bank (ASN, Regiobank, Argenta, Triodos, AXA, KBC or Dexia).

Use of the Internet and social media

The last three questions of the survey asked whether respondents have a Twitter

account and how many hours they spend daily on the Internet and social media. Merely 39.4% of the respondents have a Twitter account. The number of hours respondents spent per day on the Internet ranges from 0.5 to 9 hours, the average being 4.19 hours a day. The number of hours respondents spent per day on social media ranges from 0 to 10 hours, with an average of 1.6 hours per day.

Assessing variables on quality

Before testing the hypotheses, variables should be tested on their quality and reliability. To ensure the reliability of the analysis, scale reliability is computed for all variables. Cronbach’s alpha, α, is the most common measure of scale reliability and is a statistical measure of the

The influence of customer loyalty on word of mouth : an agent and an observer perspective.

Master thesis