The hidden treasures of online consumer reviews
The interrelationships between utilitarian and hedonic value (non)fulfillment, postconsumption emotions and customer retention in a context of online consumer reviews
Maaike Agricola 10858946
Final version of Master Thesis in Marketing MSc. in Business Administration – Marketing Track
Amsterdam Business School, part of the University of Amsterdam Supervised by dhr. dr. F.B. Situmeang and dhr. dr. U. Konus
2 Abstract
The current study investigates the relationship between utilitarian and hedonic value (non)fulfillment deriving from an online consumer review on the one hand, and customer retention on the other. The mediating influence of postconsumption emotions of satisfaction, anger, delight and dissatisfaction is examined. Also, ‘lack of consensus’ among reviewers is studied as a potential moderator of the relationship between postconsumption emotions and customer retention. It is expected that utilitarian value (non)fulfillment is related to postconsumption emotions of satisfaction (anger) and subsequently results in a high (very low) degree of customer retention. Hedonic value (non)fulfillment is expected to result in postconsumption emotions of delight (dissatisfaction) and subsequently results in a very high (low) degree of customer retention. For postconsumption emotions high (low) in arousal, the relationship with customer retention is expected to be moderated positively (negatively) by a lack of consensus among reviewers. These expectations have been tested by analyzing the content of 6433 online consumer reviews coming from the gaming industry. Regression analysis points out that direct effects are found, but mediation and moderation do not occur in all cases. Several alternative explanations are mentioned. This is the first time that a study investigates in the content of online consumer reviews, applies constructs of utilitarian and hedonic value (non)fulfillment, and tracks customer retention within the same individual writing a review. The most important take away for both theory and practice is that the content of online consumer reviews contains valuable information and that there is a potential for time, money and effort to be saved in collecting this information otherwise. Also, implications for product development are provided. Future research should demonstrate the generalizability of the application of utilitarian and hedonic value (non)fulfillment principles on online consumer reviews in other setting than the gaming industry.
3
Statement of Originality
This document is written by Maaike Agricola, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its
references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
4 Table of Content
1. Introduction 5
2. Conceptual framework 10
2.1. Content of online consumer reviews: utilitarian and hedonic
fulfillment of value 10
2.2. Linking utilitarian and hedonic value fulfillment to
postconsumption emotions 11
2.3. Linking postconsumption emotions to behavior: customer
retention 17
2.4. Differences in the strength of the relationships between
postconsumption emotions and customer retention 20 2.5. ‘Lack of consensus’ as a potential moderator 21
3. Methodology 25 3.1. Participants/sample 25 3.2. Measurement 26 3.3. Data analysis 28 4. Results 30 4.1. Descriptive statistics 30 4.2. Hypothesis testing 31
4.2.1. Value fulfillment and postconsumption emotions of
satisfaction and delight 31
4.2.2. Value nonfulfillment and postconsumption emotions of
anger and dissatisfaction 32
4.2.3. Postconsumption emotions and customer retention 33
4.2.4. Mediation effects 35
4.2.5. Moderation effects 38
5. Discussion 43
5.1. Discussion of hypothesis 44
5.2. Strengths, limitations and avenues for future research 51
5.3. Theoretical and practical implications 52
6. Conclusion 54
7. References 56
5 1. Introduction
When considering a certain purchase, checking online consumer reviews is a common behavior. As users of the product or service describe their opinions and experiences, it enables future consumers to facilitate their decision making processes (Mudambi & Schuff, 2010). Due to the enormous rise of the Internet (Gereffi, 2001) consumers are now able to obtain even more information. A shift of power has taken place from the companies towards the consumers (Archol & Kotler, 2012). Consumers in the Internet era are provided with access to more information than before (Peterson & Merino, 2003). Internet facilitates learning about products and services, comparing offerings between companies, and accessing third-party evaluations of product- or service performance, such us online consumer reviews. This results in consumers becoming better informed and thus increasing in power (Archol & Kotler, 2012). The Internet does not only provide consumers with more information, it delivers the possibility of creating own content as well (Daugherty, Eastin & Bright, 2008). Online consumer reviews
(hereafter also referred to as ‘reviews’) are an example of such consumer generated content.
Online consumer reviews contain the ability to influence the opinion of other consumers (Kaplan & Haenlein, 2008) and cause so called electronic word-of-mouth (eWOM, Gruen, Osmonbekov & Czaplewski, 2006). This eWOM can, when positive in nature, influence other consumers to proceed to purchase as it generates trust about the quality of a certain product or service (Crandall, Cosley, Huttenlocher, Kleinberg & Suri, 2008). In turn, this can influence product performance. Thus, online consumer reviews have a certain impact on product performance (Duan, Gu & Whinston, 2008). Returning to the former example of reviews influencing product performance, sales might increase and
6
therefore the product performance will enhance. As a high product performance is a desired outcome for organizations (Montoya-Weiss & Calantone, 1994), it is essential to obtain proper and detailed insights in the relationship between online consumer
reviews and product performance. This is necessary in order to improve product performance through the use of online consumer reviews.
As a result of the importance of the relationship between reviews and product performance, online consumer reviews have been widely studied in the academic literature (Chevalier & Mayzlin, 2006; Yang & Peterson, 2004). In order to obtain the aforementioned detailed insights, elaboration is needed when it comes to online
consumer reviews. In most current research, the score of an online consumer review is directive. However, by only elaborating on the score of a review, it is hard to compare scores between different reviews, as scores as such seem to be quite superficial in nature. The scores of one review might not correspond to that of the other, and double standards are used. Also, only a score does not indicate emotions that are paired with that score, it only indicates the polarity. When elaborating more one the content of a review, a deeper understanding of the review is at hand. Yet this has not been covered adequately in the existing literature. In order to tighten this gap, the current study focuses on the content of online consumer reviews instead of just the simplistic score. Also, the influence of eWOM generated by online consumer reviews on behavior has been extensively studied (Hung & Li, 2007; Park & Lee, 2009; Lee, 2009; Lee & Joun, 2009). Only, these studies mostly take into account what the effect of eWOM is on behavioral changes of other consumers. Instead, it would be interesting to measure the consequences of person X writing an online consumer review on the behavior of that same person X later on. In this way, the behavior of this person can be compared over two points of time. For instance, the purchasing behavior could be measured or
7
predicted after the writing of an online consumer review. Thus, conclusions can be drawn for customer retention. In the current scientific literature, to date this has not been executed in a real life setting. The current study aims to fill this gap by tracking whether the same person writes another review about a new product of the same developing company. From these data, customer retention can be predicted.
Combining both gaps, this study will elaborate more on the content of online consumer reviews (instead of only the score of a review) and its effects on behavior within the same individual (instead of measuring the effect of an online consumer review on other individuals behavior) in order to predict customer retention. Customer retention might lead to, amongst others, increased sales. In turn, these will contribute to increased product performance (Montoya-Weiss & Calantone, 1994).
The content of online consumer reviews will be measured in terms of both utilitarian and hedonic value. Within this distinction (Holbrook & Hirschman, 1982), utilitarian value indicates value allocation on the basis of task completion, whereas hedonic value indicates value allocation on the basis of fun and playfulness of the product. A self-made dictionary will be used to determine the fulfillment of each dimension of value and its polarity. By doing so, a more in depth analysis of online consumer reviews can be conducted. Subsequently, this content analysis can derive several emotional states from the content of a review. According to a study of Chitturi, Raghunathan & Mahajan (2008), the relationship between utilitarian and hedonic value and retention is subject to several mediators. These include the already widely studied construct of customer satisfaction (Chitturi et al., 2008, Yang & Peterson, 2004), but Chitturi et al. (2008) also take anger, delight and customer dissatisfaction into
8
In the current study, the same extensive amount of potential mediators is considered, but the study takes place in a real life setting. This can lead to different results and new insights. Thus, it can be of added value for academic research. In their article, Chitturi et al. (2008) state in the suggestions for future research that potential mediators and moderators of the relationships studied in their research should be examined.
Therefore, a potential moderator (‘lack of consensus’) on the relationship between the aforementioned postconsumption emotions (satisfaction, anger, delight and
dissatisfaction) on the one hand, and customer retention on the other hand, is suggested. The research questions are as follows:
RQ 1: What is the relationship between utilitarian- and hedonic (non)fulfillment of value (in an online consumer review) on the one hand, and customer retention on the other hand?
RQ 2: Are these relationships mediated by postconsumption emotions of satisfaction, anger, delight and dissatisfaction?
RQ 3: Are the relationships between postconsumption emotions and customer retention moderated by a lack of consensus among a group of reviewers?
By taking a closer look at the relationship of the value of online consumer reviews and customer retention and possible mediating factors, this study has the potential to contribute to both science and business. For scientists, a better understanding of the manifestation of utilitarian and hedonic value (non)fulfillment, customer retention, and the mediating role of satisfaction, anger, delight, dissatisfaction, and their
interrelationships can be gained. Also, a more detailed understanding of the
postconsumption emotions – customer retention relationship is at hand as this study also takes ‘lack of consensus’ into consideration as a moderator on this relationship.
9
Insights can be gained about the exact manifestation of customer retention. For
managers or/and practitioners, the outcomes of this study might be a starting point for influencing customer retention on the basis of online consumer reviews. The study might enable them to make effective use of online consumer reviews, as new
information might arise as a result of the more in-depth analysis conducted in this study. For instance, dependent whether the outcomes indicate a positive or negative
relationship between certain reviews and retention, companies can choose to disclose the reviews in public or keep them for themselves. In case of disclosure, customers might become more conscious of their own opinion of a certain product’s value as a consequence of being faced with other opinions. Subsequently, they might stick more to it (Coyles & Gokey, 2005). When this concerns a positive opinion, it might contribute positively to customer retention and in turn, this may lead to higher levels of overall customer loyalty (Rust, Lemon & Zeithaml, 2004). Furthermore, if the online consumer reviews contain significant emotions, this can enable companies to facilititate the process of tracking opinions of customers about their products. Also, as this study builds upon the work of Chitturi et al. (2008), a contribution is that this will be the first time that the value-retention relationship is investigated in a real life setting. It will be the first time that principles of utilitarian and hedonic value fulfillment are applied to a setting of online consumer reviews.
First of all, a detailed overview of the current literature on this subject will be provided. This conceptual framework will clarify and elaborate upon the primary constructs used in this study and its interrelationships. This will help the hypotheses being developed and justified. The methodology section will serve as a guideline for the study to be replicated, elaborating on the dictionaries used in the content analysis. Also, this section describes the characteristics of the sample used in this study and the statistical steps
10
that will be taken in order to test the hypotheses. The outcomes of these will be shown in the results section. Finally, the discussion section will cover a critical analysis of the results, elaborate on both expected and unexpected outcomes, and consider the significance of the study. Moreover, scientific and managerial implications will be presented. Strengths and limitations are mentioned and directions for further research are suggested. The conclusion provides a brief summary of the study.
2. Conceptual framework
2.1 Content of online consumer reviews: utilitarian and hedonic fulfillment of value
Woodruff (1997, p. 141) defines customer value as ‘a customer's perceived preference for, and evaluation of, those product attributes, attribute performances, and
consequences arising from use that facilitates (or blocks) achieving the customer's goals and purposes in use situations’. Customer value can be evaluated in a preconsumption stage and in a postconsumption stage. In this study, the emphasis is on the latter. In this way, the extent to which the customer value is fulfilled or unfulfilled can be assessed. Holbrook and Hirschman (1982) and Babin, Darden and Griffin (1994) consider customer value to be discerned into two dimensions. The first dimension is utilitarian value and refers to ‘the functional, instrumental and practical benefits of consumption offerings (Batra & Ahtola, 1990, p. 159), resulting primary from task completion
(Holbrook & Hirschman, 1982). The hedonic dimension of value refers to the aesthetic, experiential and enjoyment-related benefits (Chitturi, Raghunatan & Mahajan, 2007). For a better understanding a look is taken at the fulfillment of these dimensions of value. Utilitarian value fulfillment occurs when, for instance, one purchases a laptop and the battery has a long lifespan and the laptop is equipped with a high level core processor.
11
Hedonic value fulfillment takes place when the laptop is perceived by the buyer as, for instance, stylish in terms of design or color. Previous research has extensively examined the division of utilitarian and hedonic benefits (e.g., Chitturi et al., 2007, Hanzaee & Khonsari, 2011, Yang & Peterson, 2004). Mostly, this research has been focused on the relative weight that consumers attach to these dimensions in preconsumption decisions (Chitturi et al., 2008, p. 49). However, no full conformity has been reached on this topic up to now. Moreover, utilitarian and hedonic value (non)fulfillment have, to date, not been applied to postconsumption stages. Therefore, this study will take the content of online consumer reviews into account. Online consumer reviews are described as a new type of word-of-mouth information (Chen & Xie, 2008), written to either recommend or discourage others from buying the product (Sen & Lerman, 2007). As they are written after use of a certain product or service, they are considered to be established in the postconsumption stage. They concern after use evaluations, instead of pre-usage information, such as the basis of preconsumption decisions or information about product features.
2.2. Linking utilitarian and hedonic value fulfillment to postconsumption emotions
As stated previously, reviews have the aim to either recommend or discourage others from buying a certain product or service (Sen & Lerman, 2007). Due to this aim, it is presumable that the content of a review contains a lot of emotions to describe an opinion (Schindler & Bickart, 2012). These emotions can be classified as
postconsumption emotions, as they are obtained after having used the product or service. Through the use of a content analysis, several patterns can be examined in documents (Stemler, 2001). More specifically, a sentiment analysis (a subtype of
12
content analysis) can be used to reveal patterns of certain postconsumption emotions in the content of an online consumer review. To underpin this academically, this article builds upon a framework concerning the relationship between the (non)fulfillment of utilitarian or hedonic value on the one hand and customer loyalty on the other hand, mediated by postconsumption emotions (Chernev, 2004; Chitturi et al., 2007, Chitturi et al., 2008). This study aims to transfer this framework into the context of online
consumer reviews. Figure 1 shows the framework.
Design Benefits Postconsumption emotions Customer loyalty
Figure 1. Replicated conceptual framework connecting design benefits with customer
loyalty through postconsumption emotions (see: Chitturi, Raghunathan & Mahajan, 2007, p. 703, 705 & Chitturi, Raghunathan & Mahajan, 2008, p. 49).
As this study will not focus on utilitarian or hedonic benefits, but elaborate more on value and its (non)fulfillment, from now on the latter terminology will be used. As these benefits seem to be an antecedent of the more final stage of value fulfillment (Lai, 1995), it is assumed that this framework can be applied to the concepts of utilitarian and
hedonic value fulfillment as well.
What is striking in this framework and therefore needs more explanation, is that consumption experience can be both positive and negative. Consumption experiences are established through the so called expectations discrepancy (Oliver, 1997). This
13
implies that a consumer has a certain expectation of the product performance prior to consumption. After consumption, a consumer has a judgment about the actual product performance. When this is not the same as the expected product performance, a
discrepancy appears. This can be both in a positive direction (when the actual product performance is perceived as better than the expected product performance) and in a negative direction (when the actual performance is perceived worse than the expected product performance).
Also notable is the appearance of different postcompsumption emotions attached to different types of value fulfillment. Chernev (2004) assigns this to the difference in experience associated with the two types of value fulfillment. Higgins (1997, 2001) explains these different emotions by the involving different types of goals. As seen in the framework, ‘prevention emotions’ and ‘promotion emotions’ are antecedents to
respectively satisfaction and anger, and delight and dissatisfaction. The former are associated with utilitarian value (non)fulfillment, whereas the latter are associated with hedonic value (non)fulfillment. Higgins (2001) states that those types of emotions are resulting from different goals, namely ‘prevention goals’ and ‘promotion goals’.
Prevention goals. Prevention goals are defined as ‘goals that are ought to be met’
(Higgins, 1997, p. 1281). The achievement of these goals in a product consumption related context reduces the probability of a painful experience. Consumers, in case of achieving prevention goals, pass through emotions of confidence and security (Chitturi et al., 2008). Prevention goals have been linked frequently to utilitarian benefits
(Chernev, 2004; Chitturi et al., 2007; Higgins, 1997, 2001) and are thus expected to be linked to the utilitarian (non)fulfillment of value as well. In the already used example of the purchase of a laptop, prevention goals are for instance achieved by the store
14
prevention goals have a ‘must-meet’ nature, they have to be met in order to avoid painful experiences. Looking again to the laptop example, the laptop needs to possess a certain amount of memory, a software so that one can actually use it, and so on. If these ‘needs’ would not have been met, the laptop could not be used properly, which can be called a painful experience for the consumer.
Promotion goals. On the contrary, promotion goals are ‘goals that a person aspires
to meet’. Achieving those increases the probability of a pleasurable experience. Emotions deriving from the achievement of promotion goals are cheerfulness and excitement (Chitturi et al., 2007; Chitturi et al., 2008). Promotion goals have been linked often to hedonic benefits (Chernev, 2004; Chitturi et al., 2007; Higgins, 1997, 2001) and are thus expected to be linked to the hedonic (non)fulfillment of value as well. In case of the laptop, this could be the stylish design of the laptop that upgrades the social status of the customer. Just like prevention goals have a must-meet nature, Chitturi et al. (2007) have assigned an ‘aspire-to-meet’ nature to promotion goals. These goals are achieved in order to increase pleasure. In case of the laptop, the stylish design is not a prevention goal. Because when the stylish design would be missing, no painful experiences would be the consequence as the laptop still serves its functional goals. It is a promotional goal, because a more stylish laptop, on top of it functioning well, enhances the pleasure one gets from using the product. Thus, a condition for a promotional goal to be achieved is that the prevention goal(s) are achieved as well. Otherwise, a painful experience is still possible to occur (Chitturi et al., 2008). The laptop could then be stylish in terms of design, but when it does not function as it should, it could not be used properly.
Fulfillment of utilitarian and hedonic value. It has been shown that in a preconsumption
stage, customers feel more confident and secure with products that contribute to the fulfillment of utilitarian value. They feel more cheerful and excited with products that
15
contribute to the fulfillment of hedonic value. That is why consumers associate the achieving of prevention goals with utilitarian value fulfillment and the achieving of promotion goals with hedonic value fulfillment (Chitturi et al., 2007, in Chitturi et al., 2008). Now the preconsumption emotions of confidence, security, cheerfulness and excitement are being linked to a postconsumption stage. Products or services that contribute to utilitarian value fulfillment will raise low-arousal emotions; confidence and security. In turn, this leads to satisfaction in a postconsumption stage, also
characterized by its low level of arousal. Adversely, products or services that contribute to the fulfillment of hedonic value will raise high-arousal emotions; cheerfulness and excitement. Those preconsumption feelings lead to delight as a postconsumption
emotion. Delight is an emotion that is high in arousal (Hunt, 1997, in Chitturi et al., 2008, Roseman, 1991). It is now clear that the fulfillment of either utilitarian or hedonic value has different postconsumption emotions as a consequence. The first hypothesis is derived from the aforementioned:
H1a: Utilitarian value fulfillment (derived from an online consumer review) is positively related to postconsumption feelings of satisfaction.
H1b: Hedonic value fulfillment (derived from an online consumer review) is positively related to postconsumption feelings of delight.
Non-fulfillment of utilitarian and hedonic value. With the previous hypothesis concerning
the fulfillment of utilitarian and hedonic values, the question arises whether certain patterns appear in the context when utilitarian and hedonic values are not fulfilled. As stated before, negative postconsumption emotions arise when a discrepancy appears between the expected product performance and the actual product performance. In this case, the expected product performance was higher than the actual perceived product
16
performance (Oliver, 1997). Returning to the nature of utilitarian value fulfillment, this is considered a ‘must-meet’ nature. Consumers consider the achievement of utilitarian value fulfillment a necessity (Chitturi et al., 2007). When a product or service in turn does not meet these needs, consumers are likely to experience more intense
postconsumption emotions. These emotions are high in level of arousal. Anger is proposed by several studies (Chernev, 2004; Chitturi et al., 2007; Chitturi et al., 2008; Higgins; 2001) to be the consequence of the non-fulfillment of utilitarian value. In case of the non-fulfillment of utilitarian value, consumers are more tended to attribute negative outcomes to others than themselves, a phenomenon called ‘external
justification’ (Aronson, 1995). This is because of the assumption that a consumer has accurately selected the product or service before purchasing it. By putting this effort, failing will be attributed to external sources. Roseman (1991, in Chitturi et al., 2008) states that negative outcomes attributed to others are likely to lead to anger. Therefore, hypothesis 2a is developed:
H2a: Non-fulfillment of utilitarian value (derived from an online consumer review) is positively related to postconsumption feelings of anger.
On the contrary, hedonic value fulfillment is associated with an ‘aspire-to-meet’ nature. In this sense, consumers consider the achievement of hedonic value fulfillment as a luxury. When this so called ‘luxury’ condition is not met, emotions of sadness and disappointment arise. This instead of anger, as the absence of the luxury condition does not lead to painful experiences (Chitturi et al., 2007). Also, non-fulfillment of hedonic value does not exclude the non-fulfillment of utilitarian value either. In such a case, when the utilitarian value is fulfilled, the emotions deriving from the non-fulfillment of hedonic values seem not to be as intense. Previous studies (Chernev, 2004; Chitturi et
17
al., 2007; Chitturi et al., 2008) show that preconsumption emotions of sadness and disappointment transfer into postconsumption emotions of dissatisfaction.
Dissatisfaction is not as intense as anger, and is paired with relative low levels of arousal (Oliver, 1994). On the basis of the aforementioned information, hypothesis 2b is
established:
H2b: Non-fulfillment of hedonic value (derived from an online consumer review) is positively related to postconsumption feelings of dissatisfaction.
2.3. Linking postconsumption emotions to behavior: customer retention
The aforementioned demonstrates the establishment of the relationship between different kinds of value (non)fulfillment and postconsumption emotions. To complete the patterns proposed in the framework of Chitturi et al. (2008), the relationship between postconsumption emotions and behavior needs to be investigated. This link seems logical as emotions are the key mechanism in guiding human actions (Thoits, 1989). Also, Frijda (1987, in Chitturi et al., 2008) states that different emotions lead to different action tendencies. Based on that statement, it is assumed that the effect of postconsumption emotions deriving from either fulfillment or non-fulfillment differs. Analyzing the consequences of postconsumption emotions helps predict
postconsumption behavior (Chitturi et al., 2008). In this study, postconsumption behavior is reduced to customer retention. Customer retention is defined as ´the future propensity of a customer to stay with their service provider´ (Ranaweera & Prabhu, 2003) and therefore strongly contains the postconsumption component. The
relationship between the postconsumption emotions used in this study (satisfaction, anger, delight and dissatisfaction) and customer retention has been studied extensively
18
(Soscia, 2007). First, a look is taken at the postconsumption emotions associated with the fulfillment of value.
Satisfaction has been widely studied to be an antecedent of customer retention (Gustafsson, Johnson & Roos 2005; Rust & Zahorik; 1993; Ranaweera & Prabhu, 2003; Andreassen & Lindestad, 1998), but only a few studies have truly examined the nature of this relationship (Bloemer & Poiesz, 1989). One of the possible explanations is the
involvement of quality. As satisfaction as a postconsumption emotion derives from utilitarian value fulfillment, it is assumed that the needs of the product or service in terms of functionality are met. This indicates a certain amount of quality (Steenkamp, 1989). Quality, in turn, is desired and because a customer doesn’t want to get rid of quality, it leads to higher levels of customer retention. Hennig-Thurau & Klee (1997) take a closer look and find this ‘quality perceived by the customer’ as a moderator in the satisfaction-retention relationship. The quality perceived by the customer is related to higher levels of commitment and trust, which in turn results in higher satisfaction. Satisfied customers are also more willing to tolerate increases in price and thus keep using the product (Anderson, 1996).
Although to a lesser extent, customer delight has also been positively linked to customer retention (Alexander, 2010; Friedman, 2000). Jones and Sasser (1995) explain this relationship by viewing delight as an ultimate form of customer satisfaction. As satisfaction results in retention, an ultimate form of satisfaction will naturally do too. Rust, Stewart, Miller and Pielack (1996) agree on the idea of satisfaction being a component of delight. More in detail, they consider delight as an extension of
satisfaction. For a customer being delighted, a condition that has to be met is that he or she is 100% satisfied. This component of satisfaction is associated with higher switching costs to another manufacturer or service provider and thus results in retention. Another
19
explanation of the positive delight-retention relationship is delight viewed as a positive consequence of exceeding the customers’ expectations (Keiningham & Vavra, 2001). This has been shown to happen in product evaluation (Oliver, 1997). This exceeding of expectations hence leads to higher levels of customer retention. Summarizing the aforementioned, the following hypothesis is proposed:
H3a: Postconsumption emotions of satisfaction and delight are positively related to customer retention.
Adversely, the non-fulfillment of value is associated with postconsumption emotions of dissatisfaction and anger. These emotions also have been linked to retention, for which several explanations are possible. First, the relationship between dissatisfaction and retention is elaborated upon.
In general, consensus exists about the relationship between dissatisfaction and retention to be negative (Pathman, Konrad, Williams, Scheckler, Linzer & Douglas, 2002). Dissatisfaction needs to be avoided, as it is associated with future plans for leaving. Similarly, Rhodes and Nevill (2004) conclude in their study that dissatisfaction with a product or service is related to a lack of motivation to stay. Finally, the retention rate lowers. Garner & Garner (2011) consider dissatisfaction to be the contrary of satisfaction. For that reason, it is expected to be negatively related to retention, as they find satisfaction to be positively related to retention.
Anger seems to result in lower levels of retention as well (Westbrook, 1987). Bougie, Pieters and Zeelenberg (2003) conclude that anger is a consequence from the aforementioned dissatisfaction. However, it is the anger that in turn results in switching behavior and thus contributes to low levels of retention. This is supported by a study of Roos (1999), where anger seems to be a strong determinant for switching to another
20
product or service as well. Pick and Eisend (2014) explain this relationship by the desire of the customer to decrease the anger and rebuilding their emotional balance. By
sticking to a certain product or service, the feeling of anger remains. By switching to another provider, the emotional balance is restored. Building upon this idea, generally the reducing of a painful experience of anger (caused by the non-fulfillment of utilitarian values, as stated in Chitturi et al., 2007; 2008) will cause a consumer to switch. Taken together, the following hypothesis is proposed:
H3b: Postconsumption emotions of anger and dissatisfaction are negatively related to customer retention.
2.4. Differences in the strength of the relationships between postconsumption emotions and customer retention
Up to now, the differences resulting from either the fulfillment or nonfulfillment of value have been discussed. It is also supposed that utilitarian value (non)fulfillment results in different postconsumption emotions than does hedonic value (non)fulfillment. Assumed is that different kinds of postconsumption emotions have a different relationship with the concept of customer retention. The latter is now taken into closer consideration. As mentioned earlier, emotions differ in the extent of arousal they evoke (Reisenzein, 1994). In this study, delight and anger are emotions high in arousal,
whereas satisfaction and dissatisfaction are paired with relatively low levels of arousal. Looking at several psychological studies, the phenomenon of arousal seems to play a substantial role in the causation of certain behavorial patterns (Oatley & Jenkins, 1996; Bandura, 1997; Groeppel-Klein, 2005). Subsequently, different levels of arousal evoke different behavioral actions and intensity of those actions. In general, greater arousal
21
leads to an intensification of subsequent behavior (Stauss, Schmidt & Schoeler, 2005, p. 237). Therefore, the question arises whether the relationships between the
postconsumption emotions are equally strong or not.
From this line of reasoning, the relationships of the postconsumption emotions high in arousal (delight and anger) with customer retention should be stronger than the relationships of the postconsumption emotions low in arousal (satisfaction and
dissatisfaction) with customer retention. Results from Chitturi et al. (2008) support this assumption. Therefore, the fourth hypothesis is developed:
H4a: Postconsumption emotions of delight are more strongly positively related to customer retention than postconsumption emotions of satisfaction.
H4b: Postconsumption emotions of anger are more strongly negatively related to customer retention than postconsumption emotions of dissatisfaction.
2.5. ‘Lack of consensus’ as a potential moderator
In the discussion section of Chitturi et al. (2008), future research implications point out that the relationships in their study can be explored even more in detail. With that as a starting point, the fifth hypothesis on this study zooms in on the relationship between postconsumption emotions and customer retention. In past research, several
moderators have been proposed to influence this relationship (a.o. Walsh, Evanschitzky & Wunderlich, 2004; White & Yanamandram, 2007; Leisen Pollack, 2013; Chebat, 2002; Stauss et al., 2005). Customer retention is an important and desirable outcome for managers and organizations. Therefore it is necessary to obtain detailed insights into the establishment of this construct.
22
Although in the behavior within one person over time is predicted in this study and therefore has the main focus, the influence of others on one’s behavior cannot be ignored. Social influence theory (Kelman, 1958; 1961) states that people are influenced by other people around them in the establishment of their attitude. In case of different or opposite opinions about a topic/issue, social influence theory predicts that people might conform to the opinion of other people. This happens even if this person’s own opinion is not congruent with that of the other people involved (Wood, 2000). This also applies for online consumer reviews (Kaplan & Haenlein, 2008).
However, the question raises under which circumstances this assumption is valid. Hornsey, Majkut, Terry & McKimmie (2003) find a different effect when an issue has a high social importance or deep personal relevance for someone. Then, that person is more quickly tended not to conform or even counter conform to the majority. In conditions in which an issue has high social importance or deep personal relevance, it seems to be an intense emotion. As mentioned previously, intense emotions are the ones paired with high levels of arousal (Stauss et al., 2005). Therefore, with regard to this study, it is expected that different processes take place in case a lack of consensus (thus, different or opposite opinions about a topic/issue) exists.
Regarding the current study, it is expected that a delighted or angry person (assumed this person experiences high levels of arousal) tends not to conform or counter conform to other people’s opinion when writing an online consumer review. While being confronted with an opposite opinion, one is expected to become more aware of his or her own opinion. This results in sticking to their opinion even more (Coyles & Gokey, 2005). For delight, the positive opinion will strengthen, as for anger, the negative opinion will strengthen. Logically, the tendency to stick to the product (customer retention) will also respectively increase or decrease. This (non)retention
23
may serve as a proof of ones opinion to other people which strengthens their statement. Thus, the relationship between these postconsumption emotions and customer
retention strengthens. It either becomes even more positive when someone is delighted, or becomes more negative when someone is angry. It is expected that a ‘lack of
consensus’ positively moderates the relationship between emotions high in arousal (delight and anger) and customer retention. On the basis of this, hypothesis 5a is proposed:
H5a: For postconsumption emotions high in arousal, a lack of consensus among a group of reviewers will positively moderate the relationship between postconsumption emotions and customer retention.
For emotions relatively less intense (and thus low in arousal), such as satisfaction and dissatisfaction, the social influence theory of Kelman (1958, 1961) seems to apply. In a case of different or opposite opinions about an issue (and thus a lack of consensus), people experiencing low arousal emotions will tend to conform to other peoples opposite opinions (Hornsey et al., 2003), more quickly than people experiencing high arousal emotions. Having a different opinion results in a state of cognitive dissonance. This is explained as ‘discomfort experienced by an individual who holds two or more contradictory beliefs, ideas, or values at the same time, or is confronted by new
information that conflicts with existing beliefs, ideas, or values’ (Festinger, 1962, p. 4). The new information conflicting with existing beliefs can be applied to this case. Studies concerning social influence and persuasion research have shown two general motives for reducing this cognitive dissonance (Wood, 2000). The informative motive
encompasses the idea of being better informed when exposed to opinions of others, as there is more information available. It helps to act effectively. The normative motive
24
means that conforming helps one to build and maintain relationships and managing the self-concept (Cialdini & Trost, 1999).
Back to this study, satisfied or dissatisfied people are expected to switch their respectively positive or negative opinions as an attempt to reduce cognitive dissonance. As a result, the level of customer retention will decrease (as a satisfied person becomes dissatisfied) or increase (as a dissatisfied person becomes satisfied). A lack of consensus will thus weaken the pre-existing relationship between postconsumption emotions low in arousal and customer retention, as stated in hypothesis 5b:
H5b: For postconsumption emotions low in arousal, a lack of consensus among a group of reviewers will negatively moderate the relationship between postconsumption emotions and customer retention.
To summarize, all the hypothesis and interrelationships among the different constructs investigated in this study are shown in the framework below:
25 3. Methodology
In the following section, an outline is provided of how the study has been executed. There will be elaborated on the dataset used, the measurement of constructs and the statistical data analysis.
3.1. Participants/sample
This study made use of a database analysis, with a purpose to investigate the relationships between different types of value (non)fulfillment, postconsumption emotions, and customer retention. Also, lack of consensus is elaborated upon to be moderating the relationship between postconsumption emotions and customer retention. The database used is obtained by a crawler software that gathers online consumer review data from metacritic.com. Metacritic is a data gathering platform, on which online consumer reviews and expert reviews about the gaming industry are collected. There has been chosen for this platform because the largest amounts of data on online consumer reviews can be obtained here. The participants are free to express their opinion about creative products that they have consumed and provide a score that can summarize their opinion. Even though that this website has an extensive amount of consumer review data, they do not disclose the consumers’ demographic data due to privacy purposes. The current dataset exists from 6433 reviews. The reviews have been written about three platforms: the XBOX, the XBOX 360 and the XBOX 1. Even though that this study focuses on the video games industry, the results seem to be generalizable to other contexts. Mudambi and Schuff (2010) conducted a study on the usefulness of consumer reviews and showed that experience goods (such as videogames) often have a high review depth and contribute to perceived helpfulness of a review. Also, both
26
study are investigated in other studies in which the reviews consider different products (i.e. Smith, Menon & Sivakumar, 2005; Sen & Lerman 2007; Racherla & Friske, 2012).
3.2. Measurement
Utilitarian and hedonic value (non)fulfillment
To decide whether a review contains either utilitarian, hedonic or neutral content, a function is imported in Microsoft Excel. This function will perform a content analysis. It searches for a utilitarian or hedonic noun in the review. In order to do so, a list of 13 utilitarian and 15 hedonic nouns specific for the video games context (see Appendix 8.1) is imported as a dictionary (obtained from IGN.com, 16th of May 2015). The function will
(reverse-) search a dot that comes before the utilitarian or hedonic noun. The position of the dot plus 1 will be the starting point of the sentence in which the utilitarian or
hedonic noun is located. Starting from that point, the function will then try to find the next dot. This next dot will be the end point of the sentence in which the utilitarian or hedonic noun is located. After knowing the beginning and end point of the sentence, the function will search for an emotional word (adjective) within that sentence. Then, the program will record whether it is a negative (indicating non-fulfillment), positive (indicating fulfillment) or neutral utilitarian or hedonic message. This is done with the help of a dictionary of 126 positive and 193 negative adjectives, obtained from the General Inquirer (obtained from www.wjh.harvard.edu/~inquirer/, 18th of May, 2015).
Postconsumption emotions
The four postconsumption emotions in this study are satisfaction, dissatisfaction, delight and anger. After performing the steps to detect a negative, positive or neutral utilitarian or hedonic message out of the reviews, these postconsumption emotions will be
27
detected. For each review, the function will record whether it contains words indicating anger, satisfaction, delight and dissatisfaction. In order to do so, lists of words indicating the four postconsumption emotions are imported as a dictionary (the 75 words
indicating anger are obtained from dailywritingtips.com, the 42 words indicating satisfaction, the 46 words indicating delight and the 42 words indicating dissatisfaction are obtained from thesaurus.com).
Customer retention
Another function will be set up to determine the degree of customer retention. The total of reviews will be sorted on user ID. All reviews written by the same individual will be clustered, based on date of writing. A function will be run that identifies the developer and publisher of the game the individual has written a review about. Then, the next review of the same individual is searched. It is checked whether this review is about the same developer or publisher of the first game. If yes, the function stops searching. This indicates customer retention as the customer in this case has reviewed a game from the same developer or publisher at least twice. It is assumed that in order to write a review, the customer must have purchased the newer game. Thus, it indicates retention.
If the second review is not from the same developer or publisher, the function keeps on searching until it finds a review about the same developer or publisher. If it does not find this, it will return as ‘false’. This indicates that the individual has not shown customer retention by writing another review.
After this, for each individual reviewer, Boolean values for true and false are assigned, in which ‘true’ indicates a loyal customer (and thus customer retention) and ‘false’
indicates a disloyal customer (and thus no customer retention). Then, the percentage of loyal customers per game is calculated. This is done for both loyalty towards the
28
Lack of consensus
Whether there exists a lack of consensus among the reviews is determined by computing the variance of postconsumption emotions in the total of reviews in Microsoft Access. A high variance in postconsumption emotions indicates a lack of consensus.
3.3. Data analysis.
First of all, the data will be prepared for analysis in both Microsoft Excel, AMOS (Arbuckle, 2006) and SPSS Statistics 20 (IBM SPSS Statistics for Windows 20, 2011). This is done by cleaning the data and excluding the missing values. Subsequently, there is made use of several mediation analyses. The independent variables in this case were ‘utilitarian value (non)fulfillment’ and ‘hedonic value (non)fulfillment’. The dependent variable is customer retention. The potential mediator variables are respectively ‘satisfaction’ and ‘anger’, and ‘delight’ and ‘dissatisfaction’.
At first, it is checked whether the relationship between utilitarian value fulfillment and customer retention is mediated by satisfaction. This is done by following the Baron & Kenny procedure of mediation (Baron & Kenny, 1986). They state that for mediation to exist, several conditions have to be met. The first condition is that there should be a significant relationship between the independent variable, utilitarian value fulfillment, and the dependent variable, customer retention. This is tested by performing a linear regression in which utilitarian value fulfillment is the independent variable and
customer retention is the dependent variable. The second condition is that there should be a significant relationship between the independent variable, utilitarian value
29
a linear regression in which utilitarian value fulfillment is the independent variable and satisfaction is the dependent variable. The third condition is that there should be a significant relationship between the potential mediator variable, satisfaction, and the dependent variable, customer retention. This is tested by performing a linear regression in which satisfaction is the independent variable and customer retention is the
dependent variable. If all those relationships are significant, the analysis is continued. If not, mediation does not occur.
The next step is to check whether the relationship between the independent variable, utilitarian value fulfillment, and the dependent variable, customer retention, will become insignificant or at least less significant when adding the potential mediator, satisfaction, in the regression model. This is tested by performing a multiple linear regression in which utilitarian value fulfillment is the independent variable and customer retention is the dependent variable. Also, a separate block in the regression analysis is included in which the potential mediator, satisfaction, is put. If the
relationship between utilitarian value fulfillment and customer retention becomes insignificant or less significant, respectively total or partial mediation occurs. In case of partial mediation, its significance needs to be tested by conducting a Sobel-test (Sobel, 1982). This procedure is repeated for every postconsumption emotion.
To test whether ‘lack of consensus’ is a moderator of the relationship between
postconsumption emotions and customer retention, a moderator analysis is performed. This is done by the plug-in ‘Process’ in SPSS (Hayes, 2015). Here, the independent variable is one of the postconsumption emotions, in the first case ‘satisfaction’. The dependent variable is the first type of loyalty, ‘loyalty towards the developer’. The moderator is ‘lack of consensus’, the variance of postconsumption emotions in the total
30
of reviews. Model 1 in Process is used. Subsequently, the same analysis is conducted, but now ‘loyalty towards the publisher’ is the dependent variable. Hereafter, for the
remaining postconsumption emotions a moderator analysis is executed the same way.
4. Results
4.1. Descriptive statistics
The means, standard deviations, and Pearson-correlations between variables are shown in table 1. As expected, it can be seen that the two types of loyalty; loyalty towards the developer and loyalty towards the publisher, show a positive, high correlation (r = 0.66; p <0.01) with each other. What is striking is that the correlation between satisfaction and dissatisfaction is positively and high (r = 0.71, p <0.01), while those constructs were expected to correlate negatively or not at all. However, this is a normal phenomenon in the videogames industry, as tastes may differ among reviewers. Someone assessing a videogame with a grade of 5, 6, or 7, might be satisfied about some elements and
dissatisfied about others. This may result in causing the somewhat aberrant correlation. Table 1 Means (M), standard deviations (SD) and correlations between variables
(N = 6433). Variables M SD 1 2 3 4 5 6 7 8 1. Utilitarian V(N)F 0.95 1.06 - 2. Hedonic V(N)F 1.52 1.08 0.50** - 3. Anger 0.22 0.41 0.14* * 0.19* * - 4. Delight 0.04 0.17 0.11* * 0.15** 0.06** - 5. Dissatisfac tion 0.14 0.32 0.14** 0.15** 0.04** 0.10** - 6. Satisfactio n 0.25 0.44 0.18** 0.24** 0.10** 0.11** 0.71** -
31 7. Loyalty –
Dev 0.14 0.32 0.07** 0.03** -0.00 0.02 -0.01 -0.01 - 8. Loyalty –
Pub 0.24 0.39 0.06** 0.04** 0.00 0.02 -0.01 -0.01 0.66** -
Note: Utilitarian V(N)F = Utilitarian Value (Non) Fulfillment, Hedonic V(N)F = Hedonic
Value (Non) Fulfillment, Loyalty – Dev = Loyalty towards developer, Loyalty – Pub = Loyalty towards publisher, ** p < 0.01.
4.2. Hypothesis testing
4.2.1. Value fulfillment and postconsumption emotions of satisfaction and delight.
To test hypothesis 1a and 1b, a simple linear regression was performed to test the ability of utilitarian and hedonic value fulfillment to predict levels of respectively satisfaction and delight. See table 2.
Table 2 Simple linear regression with satisfaction and delight as dependent variables (N = 6433).
Satisfaction
B SE Β t
Utilitarian value fulfillment 0.183 0.001 0.970** 317.72
R² 0.940
Delight
B SE Β t
Hedonic value fulfillment 0.043 0.000 0.831** 120.030
R² 0.691
Note: RMSEA = 0.042, CFI = 0.938, * p < 0.01.
The model, in which utilitarian value fulfillment was entered as the independent variable and satisfaction as the dependent, was statistically significant F (1, 6433) = 100946.27, p < 0.01 and explained 94% of the variance in customer satisfaction. The predictor utilitarian value fulfillment was statistically significant (β = 0.970, p < 0.01). Hypothesis 1a is confirmed.
32
variable and delight as the dependent, was statistically significant F (1, 6433)
=14406.95, p < 0.01 and explained 69% of the variance in customer satisfaction. The predictor hedonic value fulfillment was statistically significant (β = 0.831, p < 0.01). Hypothesis 1b is confirmed.
4.2.2. Value nonfulfillment and postconsumption emotions of anger and dissatisfaction
To test hypothesis 2a and 2b, a simple linear regression was performed to test the ability of utilitarian and hedonic value nonfulfillment to predict levels of respectively anger and dissatisfaction. See table 3.
Table 3 Simple linear regression with anger and dissatisfaction as dependent variables (N = 6433).
Anger
B SE Β t
Utilitarian value non-fulfillment 0.874 0.003 0.962** 283.44
R² 0.926
Dissatisfaction
B SE Β t
Hedonic value non-fulfillment 0.746 0.005 0.879** 147.612
R² 0.691
Note: RMSEA = 0.042, CFI = 0.938, * p < 0.01.
The model, in which utilitarian value nonfulfillment was entered as the independent variable and anger as the dependent, was statistically significant F (1, 6433) = 80339.71, p < 0.01 and explained 93% of the variance in customer satisfaction. The predictor utilitarian value nonfulfillment was statistically significant (β = 0.962, p < 0.01). Hypothesis 2a is confirmed.
The model, in which hedonic value nonfulfillment was entered as the
independent variable and dissatisfaction as the dependent, was statistically significant F (1, 6433) = 21789.41, p < 0.01 and explained 77% of the variance in customer
33
satisfaction. The predictor hedonic value fulfillment was statistically significant (β = 0.879, p < 0.01). Hypothesis 2b is confirmed.
4.2.3. Postconsumption emotions and customer retention.
To test hypothesis 3 and 4, a simple linear regression was performed to test the ability of satisfaction, delight, anger and dissatisfaction to predict loyalty towards the developer and loyalty towards the publisher. See table 4.
Table 4 Simple linear regression with loyalty towards the developer and loyalty towards the publisher as dependent variables (N = 6433).
Loyalty towards the developer
B SE β t
Satisfaction 0.379 0.025 0.166** 14.953
Delight 4.024 0.066 0.299** 61.054
Anger -0.422 0.022 -0.192** -19.180
Dissatisfaction -0.378 0.024 -0.091** -16.000
Loyalty towards the publisher
B SE Β t
Satisfaction 0.329 0.024 0.145** 13.456
Delight 3.879 0.064 0.290** 60.980
Anger -0.407 0.021 -0.186** -19.196
Dissatisfaction -0.334 0.023 -0.079** -14.311
Note: RMSEA = 0.042, CFI = 0.938, * p < 0.01.
The relationship between satisfaction and loyalty towards the developer (β = 0.166, p < 0.01) is significant and positive. The relationship between satisfaction and loyalty towards the publisher (β = 0.145, p < 0.01) is also significant and positive. The
relationship between delight and loyalty towards the developer (β = 0.299, p < 0.01) is significant and positive. The relationship between delight and loyalty towards the publisher (β = 0.290, p < 0.01) is also significant and positive. Thus, hypothesis 3a is confirmed.
34
< 0.01) is significant and negative. The relationship between anger and loyalty towards the publisher (β = -0.186, p < 0.01) is also significant and negative. The relationship between dissatisfaction and loyalty towards the developer (β = -0.091, p < 0.01) is significant and negative. The relationship between dissatisfaction and loyalty towards the publisher (β = -0.079, p < 0.01) is also significant and negative. Therefore, hypothesis 3b confirmed.
In hypothesis 4a and 4b, it was expected that a. the relationship between delight and retention would be stronger positive than the relationship between satisfaction and loyalty and that b. the relationship between anger and loyalty would be stronger
negative than the relationship between dissatisfaction and loyalty. As seen in table 4, the β-values for delight with both loyalty towards the developer (β = 0.299, p < 0.01) and loyalty towards the publisher (β = 0.290, p < 0.01) are higher than those of satisfaction with both loyalty towards the developer (β = 0.166, p < 0.01) and loyalty towards the publisher (β = 0.145, p < 0.01). This indicates that postconsumption emotions of delight are more strongly positively related to customer retention than postconsumption emotions of satisfaction. Hypothesis 4a is confirmed.
Again looking at table 4, the β-values for anger with both loyalty towards the developer (β = -0.192, p < 0.01) and loyalty towards the publisher (β = -0.186, p < 0.01) are higher than those of dissatisfaction with both loyalty towards the developer (β = -0.091, p < 0.01) and loyalty towards the publisher (β = -0.079, p < 0.01). This indicates that postconsumption emotions of anger are more strongly negatively related to
customer retention than postconsumption emotions of dissatisfaction. Hypothesis 4b is confirmed.
35
After completing the regression analysis, several mediation analysis were performed. The procedure of Baron and Kenny (1986), as described in the Methodolody, has been followed. For the results concerning the relationship between utilitarian/hedonic value (non)fulfillment and loyalty towards the developer, see table 5. For the results
concerning the relationship between utilitarian/hedonic value (non)fulfillment and loyalty towards the publisher, see table 6.
Table 5: Regression analysis of the four steps of the procedure of Baron and Kenny (1986), in which the mediating effect of the postconsumption emotions is tested on the relationship between utilitarian/hedonic value
(non)fulfillment and loyalty towards the developer (N = 6433).
Variable B SE β p
Loyalty towards the developer
Step 1 Utilitarian value
fulfillment 0.307 0.005 0.686 0.00**
Step 2 Utilitarian value
fulfillment satisfaction
0.183 0.001 0.970 0.00**
Step 3 Satisfaction 0.379 0.025 0.166 0.00**
Step 4 Utilitarian value
fulfillment & satisfaction: Utilitarian value fulfillment Satisfaction 0.511 -0.131 0.006 0.034 1.016 -0.049 0.00** 0.00** Δ β of utilitarian value fulfillment after controlling for satisfaction 0.330
Variable B SE β p
Loyalty towards the developer
Step 1 Utilitarian value
nonfulfillment -0.498 0.020 -0.250 0.00**
Step 2 Utilitarian value
nonfulfillment anger 0.874 0.003 0.962 0.00**
Step 3 Anger -0.422 0.022 -0.192 0.00**
Step 4 Utilitarian value
nonfulfillment & anger: Utilitarian value nonfulfillment Anger 1.174 1.212 0.029 0.032 0.502 0.470 0.00** 0.00** Δ β of utilitarian value nonfulfillment after controlling for anger 0.752
36 Loyalty towards the developer
Step 1 Hedonic value
fulfillment 0.128 0.003 0.182 0.00**
Step 2 Hedonic value
fulfillment delight 0.043 0.000 0.831 0.00**
Step 3 Delight 4.024 0.066 0.299 0.00**
Step 4 Hedonic value
fulfillment & delight: Hedonic value fulfillment Delight 1.016 -6.091 0.006 0.123 1.234 -0.386 0.00** 0.00** Δ β of hedonic value fulfillment after controlling for delight 1.052
Variable B SE β p
Loyalty towards the developer
Step 1 Hedonic value
nonfulfillment -0.009 0.021 -0.458 0.998
Step 2 Hedonic value non
fulfillment dissatisfaction
0.746 0.005 0.879 0.00**
Step 3 Dissatisfaction -0.387 0.024 -0.091 0.00**
Step 4 Hedonic value
nonfulfillment & dissatisfaction: Hedonic value nonfulfillment Dissatisfaction -0.844 5.301 0.049 0.058 -0.200 1.065 0.00** 0.00** Δ β of hedonic value nonfulfillment after controlling for dissatisfaction 0.258 Note: ** = p < 0.01.
Table 6: Regression analysis of the four steps of the procedure of Baron and Kenny (1986), in which the mediating effect of the postconsumption emotions is tested on the relationship between utilitarian/hedonic value
(non)fulfillment and loyalty towards the publisher (N = 6433).
Variable B SE β p
Loyalty towards the publisher
Step 1 Utilitarian value
fulfillment 0.307 0.005 0.718 0.00**
Step 2 Utilitarian value
fulfillment satisfaction
0.183 0.001 0.970 0.00**
Step 3 Satisfaction 0.329 0.024 0.145 0.00**
Step 4 Utilitarian value
fulfillment & satisfaction:
37 Utilitarian value fulfillment Satisfaction 0.514 -0.124 0.006 0.033 1.015 -0.046 0.00** 0.00** Δ β of utilitarian value fulfillment after controlling for satisfaction 0.297
Variable B SE β p
Loyalty towards the publisher
Step 1 Utilitarian value
nonfulfillment -0.480 0.019 -0.242 0.00**
Step 2 Utilitarian value
nonfulfillment anger 0.874 0.003 0.962 0.00**
Step 3 Anger -0.407 0.021 -0.186 0.00**
Step 4 Utilitarian value
nonfulfillment & anger: Utilitarian value nonfulfillment Anger 1.174 1.231 0.028 0.031 0.499 0.475 0.00** 0.00** Δ β of utilitarian value nonfulfillment after controlling for anger 0.717
Variable B SE β p
Loyalty towards the publisher
Step 1 Hedonic value
fulfillment
0.124 0.003 0.178 0.00**
Step 2 Hedonic value
fulfillment delight 0.043 0.000 0.831 0.00**
Step 3 Delight 3.879 0.064 0.290 0.00**
Step 4 Hedonic value
fulfillment & delight: Hedonic value fulfillment Delight 1.017 -5.951 0.006 0.121 1.229 -0.375 0.00** 0.00** Δ β of hedonic value fulfillment after controlling for delight 1.051
Variable B SE β p
Loyalty towards the publisher
Step 1 Hedonic value
nonfulfillment 0.000 0.020 0.000 0.998
Step 2 Hedonic value non
fulfillment dissatisfaction
0.746 0.005 0.879 0.00**
Step 3 Dissatisfaction -0.334 0.023 -0.079 0.00**
Step 4 Hedonic value
nonfulfillment & dissatisfaction: Hedonic value nonfulfillment Dissatisfaction -0.830 5.327 0.049 0.057 -0.195 1.065 0.00** 0.00** Δ β of hedonic value nonfulfillment after controlling for dissatisfaction 0.195 Note: ** = p < 0.01.
38
Taking a look at the tables above, no mediation effects have occurred. In most cases, the first three conditions of the procedure of Baron and Kenny (1986) were met; only the last and fourth condition did not prove mediation, as none of the relationships between the independent variable and the dependent variable became insignificant or less significant. In two cases, the first condition (that there should be a significant
relationship between the independent variable and the dependent variable) was not even met. This occurred in the relationship between hedonic value nonfulfillment and both loyalty towards the developer and loyalty towards the publisher, with
dissatisfaction as a potential mediator. In the discussion section of this study, there will be further elaborated upon these results.
4.2.5. Moderation effects
After completing the mediation analysis, a moderation analysis was performed. This is done in order to investigate the influence of a lack of consensus among reviewers on the relationship between postconsumption emotions and customer retention. Per
independent variable (satisfaction, anger, delight and dissatisfaction), the moderating influence of lack of consensus on the dependent variable (customer retention, existing from loyalty towards the developer and loyalty towards the publisher), is examined. This is done with help of the SPSS plug-in Process. This is executed as explained in the Methodology section. For the results, see table 5 and 6.
Loyalty towards the developer
Satisfaction. By adding ‘lack of consensus’ as a moderator, the model is able to explain
1,65% of the variance in loyalty towards the developer. It could be shown that for a level of confidence of 95% there is a significant (p =0.0368, p <0.05) negative moderating
39
effect of satisfaction (β = -0.0051, p <0.05) on the relationship between satisfaction and loyalty towards the developer. The interaction leads to an increase of R² by 0.0017. This moderating effect is significant for low levels of lack of consensus (p < 0.05). Low levels of lack of consensus have a significant moderating effect on the relationship between satisfaction and loyalty towards the developer.
Anger. By adding ‘lack of consensus’ as a moderator, the model is not able to explain an
extra significant amount of the variance in loyalty towards the developer.
Delight. By adding ‘lack of consensus’ as a moderator, the model is not able to explain an
extra significant amount of the variance in loyalty towards the developer.
Dissatisfaction. By adding ‘lack of consensus’ as a moderator, the model is not able to
explain an extra significant amount of the variance in loyalty towards the developer. Table 7 Analysis of the moderating effect of lack of consensus on the relationship
between the postconsumption emotions and loyalty towards the developer. Model summary R R-sq F df1 df2 P Satisfaction 0.1285 0.0165 13.9374 3.00 2492.00 0.00 Anger 0.1196 0.0143 12.0584 3.00 2492.00 0.00 Delight 0.1273 0.0162 13.6864 3.00 2492.00 0.00 Dissatisfaction 0.1194 0.0143 12.0082 3.00 2492.00 0.00
R-sq increase due to interaction(s):
R2-chng F df1 df2 P
Satisfaction 0.0017** 4.3662 1.00 2492.00 0.0368
Anger 0.0002 0.5405 1.00 2492.00 0.4623
Delight 0.0008 2.0151 1.00 2492.00 0.1559
Dissatisfaction 0.0002 0.4104 1.00 2492.00 0.5218
Conditional effects of X on Y at values of the moderator:
Lack of consensus Effect SE t P Satisfaction 0.0000 0.0414 0.0184 2.2534 0.0243 6.6236 0.0075 0.0166 0.4533 0.6504 13.8884 -0.0295 0.0285 -1.0352 0.3007 Anger 0.0000 -0.0117 0.0199 -0.5878 0.5567 6.6236 -0.0010 0.0171 -0.0610 0.9514 13.8884 0.0107 0.0257 0.4140 0.6789
40 Delight 0.0000 0.0395 0.0474 0.8326 0.4052 6.6236 0.0736 0.0405 1.8179 0.0692 13.8884 0.1110 0.0480 2.3144 0.0207 Dissatisfaction 0.0000 0.0133 0.0267 0.4986 0.6181 6.6236 -0.0020 0.0243 -0.0821 0.9345 13.8884 -0.0188 0.0420 -0.4472 0.6548 Note: * = p < 0.05, ** = p < 0.01.
Table 8 Interaction variables.
Loyalty towards the developer Variable Interaction variables Β SE t Satisfaction 0.0414 0.0184 2.2534** Lack of consensus -0.0029 0.0007 -3.9343** Satisfaction x lack of consensus -0.0051 0.0024 0.0368* Anger -0.0117 0.0199 -0.5878 Lack of consensus -0.0040 0.0007 -5.4256** Anger x lack of consensus 0.0016 0.0022 0.7352 Delight 0.0395 0.0474 0.8326 Lack of consensus -0.0040 0.0007 -6.1184** Delight x lack of consensus 0.0052 0.0036 1.4195 Dissatisfaction 0.0133 0.0267 0.4986 Lack of consensus -0.0035 0.0007 -5.0207** Dissatisfaction x lack of consensus -0.0023 0.0036 -0.6400 Note: * = p < 0.05, ** = p < 0.01.
Loyalty towards the publisher
Satisfaction. By adding ‘lack of consensus’ as a moderator, the model is able to explain
1,90% of the variance in loyalty towards the publisher. It could be shown that for a level of confidence of 95% there is a significant (p =0.0217, p <0.05) negative moderating effect of satisfaction (β = -0.0063, p <0.05) on the relationship between satisfaction and loyalty towards the publisher. The interaction leads to an increase of R² by 0.0021. This moderating effect shows a tendency to be significant for high levels of lack of consensus
