A roadmap for driving customer word-of-mouth

Timothy Lee Keiningham (Department of Marketing, Peter J. Tobin College of Business, St John’s University, New York, New York, USA)
Roland T. Rust (Department of Marketing, Robert H. Smith School of Business, University of Maryland, College Park, Maryland, USA)
Bart Lariviere (Center for Service Intelligence, Department Innovation, Entrepreneurship and Service Management, Ghent University, Ghent, Belgium)
Lerzan Aksoy (Gabelli School of Business, Fordham University, Bronx, New York, USA)
Luke Williams (Qualtrics, Provo, Utah, USA)

Journal of Service Management

ISSN: 1757-5818

Publication date: 8 January 2018

Abstract

Purpose

Managers seeking to manage customer word-of-mouth (WOM) behavior need to understand how different attitudinal drivers (e.g. satisfaction, positive and negative emotion, commitment, and self-brand connection) relate to a range of WOM behaviors. They also need to know how the effects of these drivers are moderated by customer characteristics (e.g. gender, age, income, country). The paper aims to discuss these issues.

Design/methodology/approach

To investigate these issues a built a large-scale multi-national database was created that includes attitudinal drivers, customer characteristics, and a full range of WOM behaviors, involving both the sending and receiving of both positive and negative WOM, with both strong and weak ties. The combination of sending-receiving, positive-negative and strong ties-weak ties results in a typology of eight distinct WOM behaviors. The investigation explores the drivers of those behaviors, and their moderators, using a hierarchical Bayes model in which all WOM behaviors are simultaneously modeled.

Findings

Among the many important findings uncovered are: the most effective way to drive all positive WOM behaviors is through maximizing affective commitment and positive emotions; minimizing negative emotions and ensuring that customers are satisfied lowers all negative WOM behaviors; all other attitudinal drivers have lower or even mixed effects on the different WOM behaviors; and customer characteristics can have a surprisingly large impact on how attitudes affect different WOM behaviors.

Practical implications

These findings have important managerial implications for promotion (which attitudes should be stimulated to produce the desired WOM behavior) and segmentation (how should marketing efforts change, based on segments defined by customer characteristics).

Originality/value

This research points to the myriad of factors that enhance positive and reduce negative word-of-mouth, and the importance of accounting for customer heterogeneity in assessing the likely impact of attitudinal drivers on word-of-mouth behaviors.

Keywords

Citation

Keiningham, T., Rust, R., Lariviere, B., Aksoy, L. and Williams, L. (2018), "A roadmap for driving customer word-of-mouth", Journal of Service Management, Vol. 29 No. 1, pp. 2-38. https://doi.org/10.1108/JOSM-03-2017-0077

Download as .RIS

Publisher

:

Emerald Publishing Limited

Copyright © 2018, Emerald Publishing Limited


Introduction

Word-of-mouth (WOM) represents one of the most influential sources of information transfer by consumers. WOM affects our behavior as consumers by creating awareness, changing or confirming our opinions, creating interest in purchasing products/brands (Risselada et al., 2014; Van den Bulte and Wuyts, 2009) in addition to accelerating new purchases/adoption (Hennig-Thurau et al., 2015; Kumar et al., 2016; Libai et al., 2013) and encouraging repeat purchase (Iyengar et al., 2015). It can often be more persuasive than traditional media channels like advertising (Godes and Mayzlin, 2004; Herr et al., 1991), have longer carryover effects (Trusov et al., 2009), and provide a way to complement advertising (Hogan et al., 2004). Consequently, the potential of WOM to shape consumer decisions has resulted in a great deal of research in marketing (Libai et al., 2010).

Although face-to-face transmission of WOM information was the traditional way to communicate WOM, digital platforms have greatly expanded the channels and opportunities available for consumers to interact (Blazevic et al., 2013). We are now able to exchange information through online forums, chat rooms, social media, and blogs and microblogs, in addition to communicating face-to-face (Berger and Iyengar, 2013; Hewett et al., 2016). These new channels are highly influential on our decisions. For example, in a global survey Nielsen Research finds that 78 percent of consumers believe WOM from others and 61 percent indicate online WOM to be the second most trustworthy source of information (Pratt, 2008). Manifestations of customer engagement such as WOM shared on social media and other platforms represents an opportunity for firms to gain added consumer insights to complement more traditional sources of insights (MSI, 2014). Furthermore, consumer sharing of information (such as through social tagging) can enhance a firm’s brand equity and financial performance (Nam and Kannan, 2014) especially for brands with weaker brand equity (Ho-Dac et al., 2013). The power of WOM lies in the potential of consumers to shape the marketplace by being both producers and consumers of WOM, and engaging through multiple channels in which both positive and negative information can be transmitted.

As a result, it is not surprising that managers are keenly interested in understanding what drives WOM. The recommendations presented to managers by consultants in the trade press, however, tend to provide overly simplistic solutions that focus on internal company operations instead of customer attitudes that drive customers’ WOM behaviors (e.g. Sutter, 2015; Jankowski, 2013). Moreover, the recommendations almost always treat WOM behaviors as equivalent regardless of the strength relationship between WOM participants (i.e. tie strength), or the channel used to transmit/receive WOM. Managers need answers to the following questions (see Figure 1):

RQ1.

What are the dimensions of WOM behavior?

RQ2.

What is the impact of consumer attitudes on WOM behavior across different channels?

RQ3.

How do customer characteristics impact the relationship between consumer attitudes and WOM behavior across different channels?

To date, the scientific literature has not provided managers with the necessary insight to answer these questions. Although the extant literature has investigated a variety of issues relevant to WOM behavior, and the variety of new channels available to consumers, there are still critical aspects about our understanding of the nature of WOM that remain unclear (Barreto, 2014; King et al., 2014). Much of the research to date investigates WOM behavior in a siloed approach by either examining WOM to strong ties such as friends and family (Anderson, 1998; Bowman and Narayandas, 2001; Brown et al., 2005) or weak ties/online WOM (Bruce et al., 2012; Chen et al., 2011; Chevalier and Mayzlin, 2006; Godes and Mayzlin, 2004). Given that consumers today have a variety of options for how to acquire and disseminate product information, managers need insight into how a customer engages in WOM behaviors between individuals/groups with differing tie strengths and across a combination of channels. Since WOM behaviors are found to vary with the communication channels used (Berger and Iyengar, 2013), Schweidel and Moe (2014) emphasize the need to jointly model the sentiment expressed in social media posts and the venue format to which content is posted as two interrelated processes.

WOM transfer of information includes both senders (e.g. review writers) and receivers (e.g. review readers) (Moore, 2015). Moreover, the same consumer can be both a producer and consumer of WOM. Despite this, the extant literature has largely examined consumption (or receiving) of WOM, and production (or giving) of WOM by different individuals. The few notable exceptions include East et al. (2015) in which giving and receiving positive and negative WOM is examined albeit without inclusion of tie strength differences. Yang et al. (2012) also considered both WOM generation and consumption by the same individual, although a distinction between positive and negative WOM, different WOM platforms, and different tie strengths was not made explicit. In a similar vein, Ranaweera and Jayawardhena (2014) studied both giving and receiving positive and negative WOM albeit only to close friends or relatives during social interactions.

This research argues that the choices of how to engage in WOM, where to engage and with whom to engage are not independent, and these factors need to be considered jointly when investigating WOM. Similarly, this research argues that what leads consumers to engage in WOM exhibits varying strength and direction of impact when WOM valence, directionality and tie strength are considered jointly. This would have important practical implications for what tools managers should invest in when building relationships with customers and when developing WOM strategies. Baker et al. (2016) find this to be the case in their examination of valence, channel, and social tie strength on purchase intentions and retransmission likelihood. Specifically, they find that certain interactions result in mixed WOM conversations (i.e. both positive and negative WOM) that may not be desired. However, more detail regarding how to best generate WOM to achieve management goals is needed by investigating attitudinal drivers, especially across countries.

Finally, this research argues that understanding the role that customer characteristics play on WOM is likely to yield increased insight into how attitudinal drivers influence WOM. This would enable managers to segment their customer base and allocate resources in a much more targeted way for purposes of increasing positive WOM and managing negative WOM.

This research is designed to contribute to the literature in the following ways:

  1. It proposes and validates a more holistic conceptualization of WOM (including incidence and volume) that captures its multidimensional nature. More precisely, it simultaneously considers three important WOM facets that are grounded in the Sender Message Channel Receiver (SMCR) and Transactional Model of Communication (TMC) frameworks adapted from the communications literature: valence of WOM; direction of WOM communication (giving vs receiving); and tie strength through multiple channels.

  2. It examines the impact that various attitudinal drivers – specifically (dis)satisfaction, emotions, commitment and self-brand connection – have on WOM given a conceptualization of WOM that is determined jointly by factors such as valence, directionality and tie strength.

  3. It investigates whether the impact of these attitudinal drivers on WOM varies with various customer characteristics, specifically: age, gender, income and country of origin.

WOM behavior

WOM is widely accepted to be an important influencer of consumer choice. While this is likely true for most firm offerings, researchers have argued for decades that this is particularly true for services. Kaynak (1986, p. 105) observes, “There is sufficient evidence in the marketing literature that in the selection on any supplier of any service […] the consumer tends to rely heavily on word-of-mouth advertising.” As a result, given is importance to the discipline, numerous studies of word-of-mouth have appeared in the service literature; in fact, the lead article of the first issue of the Journal of Service Research explored the relationship between customer satisfaction and WOM (Anderson, 1998).

Defining WOM domain based on the SMCR-TMC frameworks

Models from the communications literature have frequently been used as a means of analyzing effective marketing-related communications. This research draws upon two streams of communication theory, the SMCR model (Berlo, 1960; Shannon and Weaver, 1949) and the TMC model (Barnlund, 2008), to provide a framework for customer WOM. According to Berlo (1960), the communication process is comprised of the Communication Source-encoder, the Message, the Channel, and the Communication Receiver-decoder (SMCR). This model helps to conceptualize WOM by proposing that consumers communicate about something, to someone, through a particular channel. Berlo (1960), Barnlund (2008) proposed the TMC. The TMC holds that messages are transmitted amongst people involved in an interaction and individuals can simultaneously influence one another (emphasizing both giving and receiving WOM by the same individual). While SMCR provides a useful framework to identify the essential components of WOM, the latter model better describes the interactive nature of WOM communication. The former frameworks help us to better understand the WOM domain by acknowledging three important dimensions: customers both give and receive information about something (i.e. the direction of communication); which can be both positive and negative information (i.e. the valence of WOM); and through a particular channel in which both strong ties (e.g. face-to-face recommendation) and weak ties (e.g. writing a negative experience on an online forum) are represented (i.e. tie strength during the WOM encounter).

Prior research on WOM however has largely ignored using the multiple facets of WOM proposed by the communications literature in the same study context. Schweidel and Moe (2014) caution against studying single venues or ignoring differences across venues in aggregated WOM data because this can lead to misleading insights. Therefore there is a need to adopt a holistic view of WOM and its drivers. Building on communication theory – using the SMCR-TMC as theoretical anchor – this study proposes and empirically examines a multidimensional instrument for WOM. This new measurement tool for WOM provides incremental validity (see Netemeyer et al., 2003) by proposing a more wide-ranging approach to measuring the targeted WOM construct.

Attitudinal drivers of WOM behaviors

WOM is a social process in which we are motivated by a variety of things such as a sense of obligation, a desire to help others/altruism, and/or a feeling of pleasure from telling others about products (Mazzarol et al., 2007; Sundaram et al., 1998). We can be driven to provide WOM because of high involvement (Dichter, 1966), to justify our decisions (generate approval), achieve social status (Gatignon and Robertson, 1986), increase our self-esteem (Wilcox and Stephen, 2013), or raise self-enhancement and visibility (Hennig-Thurau et al., 2004; Lovett et al., 2013). Zhang et al. (2014) find that males and females can differ in negative WOM transmission due to differences in image-impairment concerns.

Self-enhancement plays a role in whether we share positive or negative WOM (De Angelis et al., 2012). Studies find that positive WOM is much more widespread than negative WOM (East et al., 2007; Godes and Mayzlin, 2004; Naylor and Kleiser, 2000) and that the effect may depend on the timing of the WOM relative to the consumption experience (Chen and Lurie, 2013). Most people prefer to be associated with positive information and so they tend to share information with positive content (Berger and Milkman, 2012) and they tend to talk more when there is a moderate amount of controversy (Chen and Berger, 2013). There may however be situations when we prefer to share negative information, such as when we are concerned for others (Schlosser, 2005), want to vent (Wetzer et al., 2007) or if the information communicates expertise about us. As this makes clear, WOM is a form of self-presentation to others.

The WOM literature provides some insight into the drivers of positive or negative WOM. How satisfied (or dissatisfied) we are with a product directly influences what we share (Baker et al., 2016; Bowman and Narayandas, 2001; Brown et al., 2005; Lovett et al., 2013; Swan and Oliver, 1989; Wangenheim and Bayón, 2007), especially at the extremes (Anderson, 1998; Oliver, 2010). Emotions also play a role in heightened communication. Emotion-based measures can be stronger predictors of WOM compared to traditional cognitive measures (De Matos and Rossi, 2008; Dick and Basu, 1994; Martin et al., 2008; Söderlund and Rosengren, 2007) because sharing can help us deal with that emotional state and facilitate empathy. Information high in intensity and arousal (Berger and Milkman, 2012) and that evokes emotions (positive and negative) such as disgust, delight, interest, surprise, joy or contempt is more likely to be shared (Heath et al., 2001). Our level of commitment toward the product can also play an important role in what we say about it. The bond we experience toward a product can involve affective commitment (i.e. want to maintain the relationship) and/or calculative commitment (i.e. have to maintain the relationship) (Allen and Meyer, 1990; Garbarino and Johnson, 1999; Morgan and Hunt, 1994; Keiningham et al., 2015). Customer commitment is found to have one of the strongest effects in WOM transmission (De Matos and Rossi, 2008).

We also communicate more about product/brands that have an impact on our perception of “self.” Specifically, brands can be crucial to the creation of our self-concept (e.g. Belk, 1988; Levy, 1959; McCracken, 1986; Richins, 1994) and the formation of self-brand connections (Escalas and Bettman, 2005). Chung and Darke (2006) find that higher positive self-brand connection is associated with an increased likelihood to provide WOM.

As this makes evident, a number of researchers have investigated the relationship between attitudinal drivers and WOM. However, this review also demonstrates that a vast majority of these studies examines a single or small number of drivers of WOM incidence and none takes a holistic view of giving and receiving positive/negative WOM across channels and tie strength. For example, if consumers who have high self-brand connection are more likely to give positive WOM, does this also mean they are more likely to receive positive WOM about that brand? Are they also likely to engage in giving and receiving negative WOM. Given the opportunity, do they tend to engage in this WOM with strong or weak ties? As alluded to earlier in RQ2, this research explores how various attitudinal drivers impact WOM; specifically, this investigation uses a conceptualization of WOM that considers valence, directionality and tie strength jointly.

Impact of customer characteristics on the attitudinal drivers – WOM behavior relationship

Given that many behaviors associated with consumer loyalty have been shown to be impacted by customer characteristics (e.g. Cooil et al., 2007; Mittal and Kamakura, 2001), it is natural to expect them to play a role in customers’ WOM transmission and reception. The research to date tends to support this. Specifically, research into the drivers of WOM has found that some relationships can vary based on customer characteristics such as country (specifically, developing vs developed countries), age, gender, and income.

With regard to the impact of country development on WOM, the research to date is very limited. Nonetheless, there is early evidence that WOM behavior differs between the developing and developed world. For example, Mason (2008, p. 207) argues, “[WOM] is important because, in South Africa and possibly in other under-developed and developing countries, word-of-mouth is critical to marketing to less sophisticated or literate markets.” Furthermore, Kaynak and Jain (1993, p. 68) observe that as developing countries advance, “word-of-mouth communication becomes most significant. This is because rural customers are less mobile and depend on opinion leaders in their purchase decisions.” Empirical investigations comparing WOM between developing and developed countries are few. The limited number of existing studies, however, tends to support differences. For example, Cui et al. (2008) find that Chinese consumers are less likely to complain or engage in negative WOM behavior than are Canadian consumers. On the other hand, Cheung et al. (2009) find that Chinese consumers are more likely to engage in WOM behaviors and actively seek information from other consumers than US consumers.

With regard to age effects, East and Lomax (2010) and Yang et al. (2012) find that younger consumers tend to have greater WOM volume than older consumers. Additionally, Moliner-Velázquez et al. (2015) find that younger consumers are more likely to use online channels whereas older consumers tend to use conventional WOM; satisfaction with service recovery, however, positively impacts older consumers’ use of online WOM.

With regard to gender, Yang et al. (2012) find that males tend to generate and consume more WOM. East and Lomax (2010) find that the relationship to WOM volume is much less clear. Of the 20 studies they examined, in half of these studies, men had greater WOM volume, in the other half women did. In their examination of the data, they find (pp. 4-5):

Men tend to give more positive advice about technical and financial matters while women are more concerned with social and domestic matters. On restaurants; men give more advice about favourite restaurants […] while women give more advice in the cases of ethnic […] restaurants […]. Even on luxury goods, which might be seen as the sphere of women, the ratio is less than two to one.

With regard to income, Yang et al. (2012) and Lucid Marketing (2006) find that higher income consumers are more likely to generate and consume WOM.

As these studies make clear, there is good reason to believe that customer characteristics will play a role in WOM transmission and reception. Yang et al. (2012) find, however, that multiple customer characteristics tend to work together to define segments with high/low WOM generation and consumption. Moreover, given the findings of East and Lomax (2010) that subject matter influences differences in WOM volume for men and women, there is evidence that attitudes and gender jointly impact WOM. Therefore, this research investigates the extent to which customer characteristics (specifically age, gender, income, and country of origin) moderate the influence that attitudinal drivers have on WOM.

Data

To be able to relate attitudinal drivers to WOM behavior, this investigation uses a comprehensive multi-national cross-sectional survey data set that includes information about positive and negative WOM giving and receiving incidence and volume across a variety of channels including friends and family, online, forums, blogs, newspapers, magazines and TV. WOM drivers suggested by the literature include satisfaction, positive and negative emotions, affective and calculative commitment, self-brand connection, and consumer demographics were collected from 15,147 unique respondents spanning 10 countries and 793 different brands. The ten countries in the sample are Australia, Brazil, Canada, China, France, India, Russia, Spain, UK and USA, of which Brazil, Russia, India and China (referred to as “BRIC countries” in the remainder of this paper) are grouped as they are considered to be the world’s most important developing countries (e.g. O’Neill, 2011; Strizhakova and Coulter, 2013). To help ensure cross-national equivalence, the questionnaire was translated into the appropriate language for each non-English-speaking country, and then back translated into English to check for consistency (Wind and Douglas, 1982). To help ensure that the data reflected the general population of each country, samples were designed to be generally nationally representative in terms of age, income, and gender.

Next, the two-digit North American Industry Classification System was used to classify the 793 brands into 8 broad industry categories, weighted toward service, with 7 of the 8 investigated categories being part of the service sector: utilities, manufacturing, pharmacy, retail, transportation and warehousing, information, accommodation and food services, and finance and insurance. You et al. (2015) find evidence that WOM volume and valence elasticities can vary across industries.

Respondents were asked about their product/services usage of different industry categories and then given the opportunity to fill out the survey for up to two different product/services categories depending on whether or not they made purchases in the category. Out of all respondents, 7,315 (48 percent) evaluated only one brand, whereas 7,832 respondents (52 percent) evaluated two brands that represent different product/services categories. WOM behavior was measured via the following question. “How many times in the last year, have you [given/received WOM in a particular way]?” The one-year timeframe is adopted from Yang et al. (2012). A battery of 22 items was generated by considering the three identified dimensions of the WOM domain that are anchored in the communication literature. More precisely, the items aimed at understanding the nature of WOM include questions to capture components of both the SMCR and TMC frameworks including: whether WOM was given vs received; whether the WOM was positive or negative; and which channel was used such as online, blogs, forums, TV, newspapers, magazines, or friends and family indicating tie strength as well (Table II, discussed later, provides a complete listing of the 22 items). In addition, when generating WOM items, this investigation also distinguished between “when asked” and “spontaneously” to disentangle whether someone gave/received the WOM information with or without active solicitation. This is similar to the approach used by Wien and Olsen (2014) who differentiated between solicited and unsolicited WOM. Table I provides details on characteristics of the sample composition.

Methodology and findings

What are the different WOM behaviors?

The most precise and efficient measures are those with established construct validity; they are manifestations of constructs in an articulated theory that is well supported by empirical data (Clark and Watson, 1995, p. 310). Therefore, the WOM instrument that is proposed and tested in this study is grounded in the theoretical frameworks of SMCR and TMC from the communications literature, and formed by three dimensions: WOM valence, WOM direction of communication, and tie strength.

Dimensionality, reliability, convergent, and discriminant validity

This investigation used factor analyses to gain insight into the different dimensions of WOM volume across multiple channels by exploring the underlying structure of the 22 WOM items described earlier. All WOM items were first factor analyzed by using a principal components analysis with varimax rotation and then subjected to confirmatory factor analysis. Based on the eigenvalue criterion, six factors with eigenvalues greater than one were extracted (CFI=0.871; RMSEA=0.103; standardized RMR=0.047). To establish dimensionality, reliability and validity, Netemeyer et al. (2003) recommend conducting factor analyses that restrict a solution to an a priori theoretically derived number of factors and comparing with solutions not restricted to an a priori number. Following this approach, the non-restricted factor solution was compared with the theory-driven eight factor solution (CFI=0.902; RMSEA=0.093; standardized RMR=0.040) that was grounded in the SMCR and TMC frameworks, and formed by WOM direction (2: give and receive) × tie strength (2: weak and strong) × WOM valence (2: positive and negative). The goodness-of fit criteria of the theoretical model outperformed the non-restricted solution and showed an acceptable fit between the measurement model and the data (CFI>0.90, RMSEA<0.10, standardized RMR<0.05; see Byrne, 1998; Netemeyer et al., 2003). In addition, an examination of the AIC, which adjusts fit to compare models with differing numbers of estimated parameters (Byrne, 1998) confirmed the improvement in fit of the theoretical model over the non-restricted model. The data also showed convergent validity because all item loadings were significant, all factor reliabilities (composite reliabilities (CR)) were greater than 0.70 (lowest CR was 0.82), and all average variance extracted (AVE) exceeded 0.50 (lowest AVE was 0.67). Finally, the AVEs for each construct exceeded the squared correlation between the factors, providing evidence for discriminant validity (Fornell and Larcker, 1981). For the remainder of this paper, the respective items of the eight WOM factors are summated to represent the eight WOM dependents under investigation.

The eight types of WOM volume and their corresponding items’ CV and AVE are presented in Table II. In addition, a factorial analysis of variance revealed that it is the combination of all three WOM factors (valence, direction of communication, and tie strength), rather than each factor in isolation that determines the volume of WOM; specifically, all two-way (e.g. tie strength × direction of communication), and three-way interactions (valence × tie strength × direction of communication) were significant (lowest F-statistic=6.70), whereas only the main effect of the direction of communication dimension was insignificant (F=0.42, significant at 0.516).

Table III provides further insights into the WOM activity for the eight types of WOM behavior. The descriptive statistics reveal that for both tie strengths (i.e. strong and weak) customers are more likely on average to engage in positive WOM. Giving positive WOM to strong ties (59 percent) is the most popular type of WOM incidence followed by receiving positive WOM from both strong and weak ties (34 and 32 percent, respectively). However, when customers engage in WOM, the frequency (i.e. WOM volume) is much higher for weak ties as compared with strong ties. For instance, when customers provide positive feedback about a company/brand to their friends or family, they recommended it 6.4 times on average, whereas spreading positive WOM via weak channels was almost double in frequency (11.0 times). Finally, with respect to strong ties, giving WOM occurs more often than receiving WOM, whereas for weak ties, individuals do substantially receive more information than they generate.

Nomological validity and accounting for customer characteristics (moderating effects)

Evidence of nomological validity is provided by a construct’s possession of distinct causes, consequential effects, and/or modifying conditions, as well as quantitative differences in the degree to which a construct is related to causes or consequences (Netemeyer et al., 2003). This study investigates whether the eight different types of WOM behavior have distinct drivers (i.e. causes), by considering significance levels, and effect size (β). More precisely, customer satisfaction, positive and negative emotions, affective and calculative commitment, and self-brand connection are explored as potential drivers. Items for these constructs were borrowed from prior literature (see web Appendix 1: for more information on the measurement items, reliability and validity).

In addition, to fine-tune the influence of the drivers on WOM behavior, it is crucial to account for customer heterogeneity. Managers who want to optimize WOM can benefit from insights beyond population-averaged effects (Larivière et al., 2016) such that a more tailored WOM strategy based on customer differences is plausible. This study assesses whether the impact of the six drivers investigated on the eight WOM behaviors is influenced by customer characteristics, specifically: gender, age, income, and whether the customer lives in a developing country (operationalized here as one of the BRIC countries).

A hierarchical model for understanding WOM incidence and volume

To optimize individual WOM behavior, the impact of key customer constructs such as customer satisfaction, positive and negative emotions, affective and calculative commitment, and self-brand connection on various WOM behaviors must be modeled. As this investigation aims to account for heterogeneity at the customer level to develop a more customized WOM strategy, it proposes a hierarchical Bayes model (Rossi and Allenby, 2003). In this study, both the incidence (WOM Incidenceijk) and volume (WOM Volumeijk) are estimated jointly (Asparouhov and Muthén, 2010) for all eight WOM behaviors in the following structure:WOM Incidenceijk=1 if W⩾0, 0 otherwise, where:

(1) W = X i j β i + C i j φ + ε i j
(2) Log ( WOM Volume i j k ) = X i j γ i + C i j η + γ i j
where W is the response for individual i with respect to brand j for WOM dimension k, Xij the vector of the antecedent constructs, Cij the vector of industry dummies that serve as control variables and εij and γij are the error terms, with intercorrelation ρ.

The customer-specific response parameter vectors βi and γi are specified as:

(3) β i = Z i α + δ i
(4) γ i = Z i θ + ϕ i
where Zi is the vector of customer characteristics including gender, age, income and a dummy for the BRIC countries, α the coefficient vector representing the moderating effect of Zi on the impact that the antecedent constructs have on WOM incidence, θ the coefficient vector representing the moderating effect of Zi on the impact that the antecedent constructs have on WOM volume, and δi and ϕi are the error terms.

The aforementioned model was estimated with Bayesian inference using Markov chain Monte Carlo (MCMC) techniques. As Yuan and MacKinnon (2009) note, compared with conventional frequentist analysis, the Bayesian approach does not impose restrictive normality assumptions on sampling distributions of estimates, making statistical inference straightforward and exact. In addition, Goldstein (1995) recommends Bayesian inference to estimate random parameters. For this analysis, three independent MCMC chains with different starting points (as suggested by Gelman and Rubin, 1992) and 100,000 iterations each were run, of which the first half was considered as the “burn-in” phase, and the remaining half used to estimate the posterior distribution for the parameters, resulting in a distribution based on 150,000 data points. The eight WOM types were estimated in a multivariate model, and accounted for correlated errors between these dependent variables across both levels. To assess the convergence of the MCMC algorithm, autocorrelation and trace plots of the residual variance for the parameter estimates were inspected. More details about the error covariance matrix and convergence inspection can be found in web Appendix 2. In addition, web Appendix 2 provides details regarding tests for multicollinearity (Chatterjee et al., 2000; Hair et al., 2010) and common method bias (Podsakoff et al., 2003), neither of which were found to be significant sources of concern regarding the analyses.

Creating a roadmap for driving WOM

Table IV details the drivers of WOM behavior and provides strong evidence for nomological validity of the proposed WOM measurement model since the drivers of the eight types of WOM behavior are different in terms of significance level and magnitude (i.e. size of the β). In addition, the antecedent causes of WOM incidence vs WOM volume also differ, which reveals that the underlying mechanism for two different forms of WOM operationalization differs.

The results presented in Table IV provide researchers with clear relationships between the drivers of WOM across industries and customer characteristics. Specifically, Table IV summarizes the parameter estimates of Equations (3) and (4). For example, the impact that customer satisfaction has on the volume of positive WOM given to strong ties (WOMF1) is as follows:

β SAT WOMF1 = 0.101 ( intercept ) 0.044 gender _ male ( 1 = male , 0 = female ) + 0.051 age _ group18 _ 34 ( 1 = if yes ; 0 = if older ) 0.030 income _ high ( 1 = top half within each country , 0 = bottom half ) + 0.162 country _ Brazil _ China _ India _ Russia ( 1 = BRIC countries , 0 = otherwise ) .

As this example illustrates, the impact of satisfaction on WOM is heterogeneous and dependent on customer characteristics. More precisely, the intercept of 0.101 represents the positive impact that customer satisfaction has on the volume of positive WOM to strong ties (WOMF1) for female customers, aged 35+ with lower incomes, who do not live in the developing countries (i.e. the reference group; when all independent variables are set to zero). In addition, the satisfaction effect is greatest (0.314=0.101+0.051+0.162) for younger (+0.051) females with lower incomes who live in the developing countries (+0.162) and lowest (0.027=0.101−0.044−0.030) for male (−0.044) customers, aged 35+ with higher incomes (−0.30) who do not live in the developing countries. As such the positive effect of satisfaction on this particular WOM behavior ranges from 0.027 to 0.314, which is, respectively, 5.0 times lower and 2.3 times higher than the population-averaged, fixed effect of 0.136.

Clearly, this level of detail allows researchers and managers to estimate the impact on WOM incidence and volume of efforts to improve specific attitudinal drivers across industries, countries, and customer groups. Despite this benefit, however, most managers would find it difficult to use the parameter estimates in Table IV to develop a coherent strategy to leverage attitudinal drivers of WOM across countries and different customer segments. Therefore, while the level of detail offers precision, most managers would consider the model overly complex and difficult to use to guide strategic decision-making. As a result, presenting the findings “as is” would likely have little managerial impact; when confronted with complex models, managers typically ignore them and rely on overly simplistic models (Little, 1970, 2004).

To make it easier to see the big picture and to help guide strategic decision-making without sacrificing the specificity of this analysis, this research uses simple yet powerful visualizations of the core information contained in the analysis. The logical place to begin is with a general understanding as to whether the attitudinal drivers result in positive or negative WOM behaviors for the firm. Figure 2 presents the positive or negative WOM behaviors (from a firm’s point of view) associated with the attitudinal drivers investigated. An examination of the chart reveals that all attitudinal drivers significantly impact most of the WOM behaviors investigated. Not all behaviors, however, go in the direction that would most be expected based on the positivity/negativity generally associated with the attitudinal drivers. For example, customer satisfaction is associated with an increase in giving of positive WOM to strong ties (both incidence and volume), but a decrease in all other positive WOM behaviors. By contrast, negative emotions are associated with a decrease in giving of positive WOM to strong ties (both incidence and volume), but an increase in all other positive WOM behaviors. When comparing the relationship between the attitudinal drivers and positive/negative WOM behaviors, however, the most striking finding is that most attitudinal drivers are associated with increased positive and negative WOM.

Moreover, most attitudinal drivers are not associated with reductions in negative WOM behaviors. In fact, customer satisfaction is the only attitudinal driver that is associated with reduced negative WOM across all conditions investigated. The only other attitudinal driver associated with reductions in negative WOM is affective commitment: specifically, reductions in giving negative WOM to strong ties (both incidence and volume) and giving negative WOM to weak ties (volume only).

While knowing the direction of association between attitudinal drivers and corresponding WOM behaviors is important, it is still necessary to know the impact on WOM. Figure 3 presents a visual representation of the relative effect on WOM behaviors associated with the attitudinal drivers under investigation. Impact is represented by the size of a circle which is determined by parameter estimate’s absolute value divided by the largest coefficient (in absolute value) for incidence or volume, respectively. Note, the size of circle reflects the attitudinal variables only (i.e. absent the effect industry and customer characteristics), and circle sizes are relative to the maximum absolute value across all coefficients in the respective category (i.e. volume or incidence).

An examination of Figure 3 immediately reveals large differences in the relative impact on WOM behaviors associated with the different attitudinal drivers. The largest impact results from negative emotions on giving negative WOM to strong ties for both incidence and volume. Moreover, negative emotions have the largest impact across all negative WOM behaviors investigated.

The largest drivers of positive WOM behaviors vary somewhat depending upon the specific behavior. The largest impact on any positive WOM behavior results from affective commitment on the volume and incidence of positive WOM given to strong ties. Across all other positive WOM behaviors, however, the largest impact is associated with positive emotions. Furthermore, positive emotions have the second largest impact on WOM incidence and volume of positive WOM given to strong ties.

Finally, only one attitudinal driver reduced the volume and incidence of negative WOM across all behaviors investigated: customer satisfaction. In fact, affective commitment was the only driver associated with a reduction in negative WOM behaviors, and this was only for giving negative WOM to strong ties (both incidence and volume) and giving negative WOM to weak ties (incidence only).

Therefore, when creating a roadmap to drive WOM, it is important to understand that while all roads (i.e. attitudinal drivers) may lead to greater WOM, some roads are superhighways while others are narrow passageways. Moreover, some paths have both pluses and minuses (from the firm’s perspective) in driving WOM. If managers’ goal, however, is simply leverage specific attitudinal drivers to maximize positive WOM and minimize negative WOM across all behaviors investigated, the path is relatively straightforward. Specifically, the most effective way to drive positive WOM is through affective commitment and positive emotions. The most effective way to minimize negative WOM is by minimizing negative emotions, and ensuring that customers are satisfied.

Of course, other factors impact WOM behaviors. Figure 4 presents the increase/decrease in WOM behaviors associated with an industry relative to the finance and insurance industry (the reference category). For example, WOM for firms in manufacturing tend to be higher for both positive and negative WOM. And WOM behaviors tend to be worse (from a firm’s perspective) for giving/receiving positive WOM to/from weak ties, and giving/receiving negative WOM to/from strong ties for the retail and information industries.

Moreover, customer characteristics moderate the effect of the drivers of WOM incidence and volume. Figures 5 and 6 presents the increase/decrease in WOM behaviors associated with gender, age, income, and country for each of the attitudinal drivers under investigation; the authors wish to remind the reader that the arrows and colors in the figure would be reversed were the analysis to have reversed the coding (e.g. for Gender, Female=1 instead of Male=1).

Taken together, Figures 4-6 make clear that managers seeking to enhance firm-beneficial WOM behaviors should account for the impact of industry and customer characteristics. Whether this alters the choice of attitudinal driver expected to generate the maximum impact, or confirms the initial decision (based on Figure 3), it will provide managers with insight as to whether the effort is likely to achieve desired WOM outcomes.

Discussion and contribution

This study contributes to what is known about WOM by undertaking one of the most comprehensive explorations of its attitudinal drivers. It utilizes the various proposed components of the SMCR model (Berlo, 1960) and the TMC (Barnlund, 2008) to provide a holistic view of WOM (Research Objective 1). Further, it examines the attitudinal drivers of WOM behavior (Research Objective 2), and it explores the moderating impact of customer characteristics (Research Objective 3).

Far more important than the scope of the investigation, however, are the theoretical and managerial implications uncovered. One of the first things to acknowledge is that the results are far more complex than the commonly espoused, but non-robust management maxims often associated with WOM (e.g. a dissatisfied customer will tell nine or more people TARP, 1986), but the more granular details provide researchers and managers with much richer insight into what drives WOM.

Research implications

This investigation provides important insights for researchers. First, this study validates a multidimensional conceptualization of WOM that consists of eight WOM behaviors formed by WOM direction (give vs receive) × WOM valence (positive vs negative) × tie strength across various communication channels (weak vs strong ties). In addition, in testing nomological validity, this research finds evidence that customer satisfaction, emotions, commitment, and self-brand connections have different impacts (significance and effect size) across the eight different WOM behaviors identified. As a result, this study highlights the need to adopt a more holistic multifaceted approach to conceptualizing and measuring WOM in order to better understand the nuances that exist in WOM behaviors. This turns out to be especially important when examining attitudinal drivers of WOM behaviors.

Second, the need to study WOM in a multifaceted way is further demonstrated by the fact that the same driver can have both a positive and negative influence on WOM behavior. For instance, calculative commitment enhances both the volume of positive and negative WOM behavior via weak ties. This finding may be explained by the “rational” bond that calculative commitment represents, which includes both the benefits and costs of being connected with a brand/company (Larivière et al., 2014). In addition, the drivers of WOM also differ between WOM incidence and volume, which reveals the need to distinguish between these two outcomes.

Third, the findings highlight the dangers of treating customers as homogeneous in terms of how attitudinal drivers impact WOM behavior. The impact of WOM drivers such as satisfaction, emotions, and commitment on WOM behavior is found to vary by demographic characteristics. Therefore, it is important for WOM researchers to understand whether customer segments exist before conclusions are made from the data, as ignoring this could lead to misleading insights from aggregated results.

Fourth, this research points to the difficulty in generalizing the antecedents of WOM behaviors across industries. The findings of this investigation clearly demonstrate that industry plays an important role in the impact of key drivers on WOM behaviors. Therefore, investigations of WOM should incorporate multiple industries to identify and distinguish between generalizable and industry-specific effects.

Finally, the findings of this research point to the potential for publication bias in the scientific literature (e.g. Easterbrook et al., 1991; Rothstein et al., 2006) regarding the drivers of WOM. To be clear, this study has not proven such bias exists. Rather because several of the results run counter to most published findings on the topic, it is possible that researchers reporting similar findings have abandoned studies (e.g. because of perceived error on their part, an inability to explain unexpected or counter-intuitive findings, etc.), or because investigations reporting such findings are unable to make it through the peer review process. For example, the scientific literature overwhelmingly reports that customer satisfaction results in beneficial (to the firm) WOM behaviors (see De Matos and Rossi, 2008). While the findings of this investigation in general support the idea that customer satisfaction results in firm-beneficial WOM behaviors, this was largely through the reduced likelihood of negative WOM. However, customer satisfaction was also associated with negative (to the firm) WOM behaviors: specifically, a lower likelihood of receiving of positive WOM (both strong and weak ties) and of giving positive WOM to weak ties.

Managerial implications

The results of this study provide important implications for managers seeking to enhance customers’ WOM behaviors. At a basic level, managers must recognize that WOM is a multidimensional construct that spans communication channels, tie strength, valence, and transmission and/or reception. This breadth fundamentally challenges the simplistic “X Steps to Build Word-of-mouth” lists that are common in the trade press (e.g. Jankowski, 2013; Sutter, 2015). Consultants tend to treat WOM across channels and tie strength as being interchangeable, and therefore all WOM as being equal (with the exception of the popularly espoused yet tactically difficult task of “target influencers” to enhance WOM). As a result, the proposed tactics are likely to have limited impact across all channels, and can even result in negative outcomes (from the perspective of managers). Moreover, rarely do these “top x” lists focus on the attitudinal drivers of WOM, and instead focus on tactics (e.g. “Make it easy for customers to leave reviews” Sutter, 2015) or truisms (e.g. “Always be honest” Jankowski, 2013).

Even when these WOM how-to lists do include occasional attitudinal drivers, the prescriptions tend to be overly simplistic. For example, Grauer (2014) argues, “Keep current customers extremely satisfied, and it will encourage them to spread the word.” This line of thinking is common among managers; specifically managers (and researchers) often focus on customer satisfaction based on the following chain of effects: satisfaction → recommend intention → recommend behavior (e.g. Keiningham et al., 2007); this in part is why satisfaction is the most widely used gauge of customers’ perceptions of firm-customer encounters (Aksoy, 2013). While acknowledging that this chain of effects is indeed real, these findings clearly indicate that efforts to improve positive WOM through satisfaction measurement and management are unlikely to be the most productive approach as its impact tends to be small relative to two other attitudinal drivers: affective commitment and positive emotions. Instead, the primary benefit of satisfaction is that satisfied customers give and receive less negative WOM via both strong and weak ties.

Sutter (2015) encourages managers to “Connect to the human emotion” to enhance WOM. Again, managers and researchers would likely find this assertion to be self-evident. This investigation however finds that emotions stimulate both favorable and unfavorable WOM behaviors in tandem, although this research finds evidence (specifically, different effect sizes) that the net impact of positive emotions is always favorable and the net impact of negative emotions is unfavorable from a managerial point of view. Moreover, emotions have the strongest impact on WOM for 14 out of the 16 WOM dependents examined (the exception being the incidence and volume of giving positive WOM to strong ties). Therefore, on a general level, these findings support the call to “connect to the human emotion.” In terms of “how best to connect” to drive WOM, however, this research offers unexpected insight.

Typically managers seeking to drive positive WOM through positive emotions have tended to follow pleasure-arousal theory (PAT) (Reisenzein, 1994) to impact emotional intensity (although it must be acknowledged that most managers do this intuitively). This occurs because circumplex models of emotions (Plutchik and Conte, 1997), PAT, and customer satisfaction blend intuitively (Oliver, 2010, Figure 12.4, p. 322). Much of the business press has focused on management efforts to achieve high pleasure and high arousal to stimulate positive WOM. In fact, Sutter’s (2015) example for leveraging emotions was what he called the “largest and most elaborate staged event ever.” These findings, however, indicate that managers may not need to focus on high arousal to engender WOM. In this investigation, emotions split into two groups, positive and negative, and not along the arousal dimension (i.e. high vs low arousal). As positive emotions in general drive positive WOM (and negative emotions drive negative WOM), managers appear to be able to generate WOM without requiring high arousal (although it should be noted that prior research finds that high arousal appears necessary for “viral” online sharing Berger, 2013). As a result, these findings offer evidence that managers may be able to focus on delivering consistently positive emotional experiences, which can be standardized, as opposed to focusing on generating high levels of arousal, which have the potential for raising the expectations of customers.

It is important to note that managers’ ultimate goal in measuring and managing customer satisfaction and consumption emotions is to engender customer commitment to their firms or brands to enhance repeat purchase behavior. The importance of building affective commitment becomes clear when comparing it to customer satisfaction. Unlike satisfaction, affective commitment enhances all positive WOM behaviors and lowers the giving of negative WOM behaviors. Most importantly, in terms of the roadmap for driving WOM discussed earlier, affective commitment is the strongest driver of incidence and volume of giving positive WOM to family and friends (i.e. strong ties).

While managers tend to treat high satisfaction levels and affective commitment as interchangeable, these findings indicate that this may have adverse consequences on WOM. While consistently satisfactory experiences are important in building affective bonds, satisfaction and affective commitment are distinct constructs. Unlike satisfaction, nurturing affective commitment requires demonstrating reciprocity and fostering personal involvement with the company (Gustafsson et al., 2005, p. 211).

Finally, managers need to be cautious about relying on simple population averages to estimate the impact of any driver on WOM. The findings demonstrate that these effects can be more/less positive/negative depending on customer characteristics. As these findings demonstrate, the deviation from the average can be quite large. As noted earlier, satisfaction’s impact on WOM is 2.3 times higher than the average for lower income women living in BRIC countries, and five times lower for higher income men living in developed countries. As this example makes clear, it is important for managers to consider different customer characteristics in combination rather than in isolation to better understand how the impact of the various antecedent drivers impact WOM behaviors. This requires that managers use sophisticated and robust analytic approaches to assess the impact of an attitudinal driver that takes into account customer characteristics. Clearly this is not as simple as the rules of thumb that permeate the management press with regard to the relationship between attitudinal drivers and WOM. Nevertheless, by leveraging this added complexity marketing managers can meaningfully identify and target specific customer segments with relevant messaging and actions designed to address specific WOM drivers that will offer the greatest beneficial impact to their firms.

Limitations and directions for future research

Although this study is comprehensive and integrative on the topic of WOM, it is not without limitations.

First, Harrison-Walker (2001) and Baker et al. (2016) propose neutral as a WOM valence category in addition to positive and negative WOM. Although the neutral classification is not used as frequently as positive and negative WOM in research, it nonetheless must be acknowledged that this study has not included this category in this research. Perhaps future research can explore extending the findings in the context of neutral WOM.

Second, as in some previous recent research on WOM (e.g. Baker et al., 2016), this study makes use of survey and self-reported WOM data. Although this data collection approach was necessary to study attitudinal drivers and the receiving/consumption of WOM (in line with Yang et al., 2012), this investigation did not have information about the content of the WOM communication. As Godes and Mayzlin (2004) suggests, WOM incidence and volume may not necessarily provide a comprehensive picture of WOM even if one knows whether the communication was positive or negative. This is because some positive (negative) reviews can be much more flattering (demeaning) and impactful than others due to match of linguistic style (Ludwig et al., 2013), vividness, and depth and intensity of the message (Mazzarol et al., 2007), or it could contain indifferent (neutral) and/or both negative and positive content in the same message (Tang et al., 2014). The way the WOM is communicated may also change based on whether the product is utilitarian (focus on explaining actions) or hedonic (focus on explaining reactions) (Moore, 2015). For negative WOM, consumer reactions to trust and liking of the communicator of WOM messages could also depend on whether the communicator softens pronouncements of bad news (dispreferred markers) (Hamilton et al., 2014). As such this investigation does not include a measure of “dispersion” or linguistic content in the WOM data. It would be beneficial if future research could explore the impact of the degree of the positivity or negativity of the WOM on the relationships explored in this study.

Third, another limitation of the study is that it uses cross-sectional data. In other words, it is not possible in this investigation to consider dynamic effects of WOM, such as how WOM impact changes over time or over the course of the product life cycle (Bruce et al., 2012) or the temporal evolution of WOM (Godes and Silva, 2011; Schweidel and Moe, 2014).

Fourth, the broad nature of the study also constrained the number of variables that could be included and investigated. As such the list of drivers is not an exhaustive list of all possible antecedents covered in the WOM literature to date. For instance, this investigation was not able to include variables in this model such as service quality, perceived value, trust and quality used by other researchers (De Matos and Rossi, 2008). Nonetheless, a diligent attempt was made to include as many of the important attitudinal driver variables as possible.

Fifth, while efforts were made per the recommendations of Podsakoff et al. (2003) to reduce the potential for common method bias (see Appendix 3 for more details), it should be noted that all dependent and independent variables were provided by respondents in a single survey instrument.

Finally, this research does not distinguish between online and offline WOM with respect to strong ties (i.e. WOMF1, 2, 3, and 4).

Despite these limitations, these findings offer compelling new insights into the nature of giving and receiving WOM across countries, industries, modes of communication.

Figures

Conceptual framework

Figure 1

Conceptual framework

Positive or negative WOM behaviors associated with attitudinal drivers

Figure 2

Positive or negative WOM behaviors associated with attitudinal drivers

Relative effect on WOM behaviors associated with attitudinal drivers

Figure 3

Relative effect on WOM behaviors associated with attitudinal drivers

Effect of industry on attitudinal drivers of WOM behaviorsa

Figure 4

Effect of industry on attitudinal drivers of WOM behaviorsa

Effect of customer characteristics on attitudinal drivers of positive WOM behaviors

Figure 5

Effect of customer characteristics on attitudinal drivers of positive WOM behaviors

Effect of customer characteristics on attitudinal drivers of positive WOM behaviors

Figure 6

Effect of customer characteristics on attitudinal drivers of positive WOM behaviors

Autocorrelation plot for the effect of affective commitment on giving positive word-of-mouth to strong ties (from 100,000 iterations, chain 1)

Figure A1

Autocorrelation plot for the effect of affective commitment on giving positive word-of-mouth to strong ties (from 100,000 iterations, chain 1)

Trace plot of the impact of affective commitment on giving positive word-of-mouth to strong ties (from 100,000 iterations)

Figure A2

Trace plot of the impact of affective commitment on giving positive word-of-mouth to strong ties (from 100,000 iterations)

Composition of the sample

Customer demographics Demographics
Gender (% male) 51.2%
Age 18-24 25-34 35-44 45-54 55+
(% of sample) 8.5 25.2 26.5 22.5 17.3
Countries Australia Brazil Canada China France India Russia Spain UK USA
(% of sample) 10.8 9.0 10.6 10.2 9.1 3.7 3.7 10.6 12.9 19.4
Industries Utilities Manufacturing Pharmacy Retail Transportation and warehousing Information Accommodation and food services Finance and insurance
(% of sample) 5.3 24.4 1.9 9.3 1.7 24.1 2.0 31.3

Dimensions of word-of-mouth behavior

Constructs AVE CR
Give/Positive/strong ties (Family and Friends) (WOMF1) 0.75 0.85
1. Recommended (company/brand) to family and friends (spontaneously)
2. Recommended (company/brand) to family and friends (when asked)
Receive/Positive/strong ties (Family and Friends) (WOMF2) 0.76 0.87
1. Family and friends recommended (company/brand) to me (spontaneously)
2. Family and friends recommended [company/brand] to me (When asked)
Give/Negative/Strong ties (Family and Friends) (WOMF3) 0.69 0.82
1. Given negative feedback about (company/brand) to family and friends (spontaneously)
2. Given negative feedback about (company/brand) to family and friends (when asked)
Receive/Negative/Strong ties (Family and Friends) (WOMF4) 0.72 0.83
1. Family and friends gave negative feedback about (company/brand) to me (spontaneously)
2. Family and friends gave negative feedback about (company/brand) to me (when asked)
Give/positive/weak ties (online and blog) (WOMF5) 0.73 0.89
1. Given a positive review about (company/brand) online (spontaneously)
2. Given a positive review about (company/brand) online (when asked)
3. Posted positive comment about (company/brand) on forums/blogs (spontaneously)
4. Posted positive comment about (company/brand) on forums/blogs (when asked)
Receive/positive/weak ties (online and blog and newspaper, TV, magazines) (WOMF6) 0.69 0.87
1. Read/saw positive review about (company/brand) online (spontaneously)
2. Read/saw positive comment about (company/brand) on forums/blogs (spontaneously)
3. Read/saw positive review about (company/brand) somewhere else (newspapers, magazines, TV, etc.) (spontaneously)
Give/negative/weak ties (online and blog) (WOMF7) 0.71 0.91
1. Given a negative review about (company/brand) online (spontaneously)
2. Given a negative review about (company/brand) online (when asked)
3. Posted negative comment about (company/brand) on forums/blogs (spontaneously)
4. Posted negative comment about (company/brand) on forums/blogs (when asked)
Receive/negative/weak ties (online and blog and newspaper, TV, magazines) (WOMF8) 0.67 0.89
1. Read/saw negative review about (company/brand) online (spontaneously)
2. Read/saw negative comment about (company/brand) on forums/blogs (spontaneously)
3. Read/saw negative review about (company/brand) somewhere else (newspapers, magazines, TV, etc.) (spontaneously)

Notes: AVE, average variance extracted; CR, composite reliability

Word-of-mouth incidence and volume

WOM dimension WOM behavior (incidence) (%)a WOM behavior (volume)b
WOMF1: Give PWOM to strong ties 59 6.4 times
WOMF2: Receive PWOM from strong ties 34 5.5 times
WOMF3: Give NWOM to strong ties 23 5.9 times
WOMF4: Receive NWOM from strong ties 19 5.3 times
WOMF5: Give PWOM to weak ties 21 11.0 times
WOMF6: Receive PWOM from weak ties 32 13.1 times
WOMF7: Give NWOM to weak ties 9 9.7 times
WOMF8: Receive NWOM from weak ties 20 9.6 times

Notes: PWOM, positive customer word-of-mouth behavior; NWOM, negative customer word-of-mouth behavior. aThe % reflects the proportion of customers who engaged in this WOM behavior (for one company within 1 year); bon average, how many times has the customer engaged in this WOM behavior in case of incidence (for one company within one year)

Examining nomological validity and testing for customer heterogeneity: drivers and moderating effects on WOM

WOMF1 WOMF2 WOMF3 WOMF4 WOMF5 WOMF6 WOMF7 WOMF8
Give/positive/strong ties Receive/positive/strong ties Give/negative/strong ties Receive/negative/strong ties Give/positive/weak ties Receive/positive/weak ties Give/negative/weak ties Receive/negative/weak ties
Incid. Vol. Incid. Vol. Incid. Vol. Incid. Vol. Incid. Vol. Incid. Vol. Incid. Vol. Incid. Vol.
Explanatory variables β β β β β β β β β β β β β β β β
WOM driversa
Customer satisfaction 0.030* 0.136* −0.041* −0.053* −0.112* −0.227* −0.084* −0.136* −0.033* −0.012 −0.028* −0.023* −0.046* −0.036* −0.060* −0.074*
Positive emotions 0.113* 0.237* 0.104* 0.212* −0.004 −0.001 0.020 0.035* 0.079* 0.146* 0.096* 0.209* 0.015 0.013* 0.046* 0.066*
Negative emotions −0.054* −0.039* 0.033* 0.073* 0.195* 0.355* 0.149* 0.268* 0.049* 0.143* 0.027* 0.080* 0.177* 0.251* 0.101* 0.226*
Affective commitment 0.133* 0.293* 0.039* 0.083* −0.028* −0.045* −0.014 −0.019 0.039* 0.081* 0.077* 0.170* −0.035 −0.017* 0.011 0.024*
Calculative commitment 0.020* 0.048* 0.035* 0.082* 0.022* 0.032* 0.023* 0.035* 0.030* 0.064* −0.008 −0.005 0.028 0.032* 0.007 0.008
Self-brand connection −0.007 −0.035* 0.039* 0.080* 0.017 0.017* 0.024* 0.032* 0.058* 0.121* 0.018* 0.046* 0.074* 0.073* 0.024* 0.043*
Control variables: industry effects
Industry utilities −0.104 −0.156* −0.024 −0.099 −0.056 −0.060 −0.033 −0.038 0.005 −0.084 −0.052 −0.146* −0.003 0.005 0.060 −0.034
Industry manufacturing 0.177* 0.375* 0.259* 0.453* 0.097* 0.167* 0.136* 0.210* 0.123* 0.173* 0.265* 0.490* 0.031 0.059* 0.176* 0.309*
Industry pharmacy −0.059 −0.023 0.132 0.098 −0.001 0.083 0.084 0.090 0.085 −0.008 −0.062 −0.187 0.003 −0.005 −0.043 −0.010
Industry retail 0.087 0.189* 0.239* 0.385* 0.166* 0.251* 0.103* 0.179* −0.107* −0.182* −0.122* −0.225* −0.131 −0.016 −0.102 −0.087*
Industry transportation and warehousing 0.042 0.019 0.151 0.176 0.040 0.021 −0.332* −0.093 −0.007 −0.170 −0.018 −0.174 −0.082 −0.096 −0.016 0.070
Industry information −0.004 0.015 0.037 0.045 0.065* 0.106* 0.093* 0.177* −0.112* −0.129* −0.093* −0.189* −0.101* −0.035 −0.049 −0.022
Industry accommodation and food services 0.046 0.087 0.272* 0.348* 0.097 0.044 −0.057 −0.006 0.181 0.105 0.029 −0.023 0.101 0.064 −0.018 −0.035
Moderators of the customer satisfaction effect
Intercept −0.016 0.101* −0.073* −0.067* −0.124* −0.174* −0.082* −0.126* −0.098* −0.054* −0.063* −0.059* −0.086* −0.024* −0.058* −0.093*
Gender_male (1=male, 0=female) −0.039* −0.044* −0.020* −0.038* 0.018 −0.012 −0.007 −0.013 0.070* 0.034* −0.007 −0.008 −0.011 −0.005 0.009 0.005
Age_group18_34 0.019 0.051* 0.013 0.017 0.043* 0.019 0.024 0.023* −0.062* −0.050* −0.032* −0.050* 0.004 −0.017* −0.013 −0.018*
Income_high (1=top half within each country, 0=bottom half) −0.024* −0.030* −0.034* −0.046* −0.011 −0.003 −0.007 −0.012 −0.047* −0.051* 0.015* 0.003 −0.022* −0.012 −0.011 0.003
Country_Brazil_China_India_Russia 0.071* 0.162* 0.048 0.087* 0.041* 0.073* 0.050* 0.070* 0.057* 0.005 0.042* 0.097* 0.015 −0.009 0.063* 0.095*
Moderators of the positive emotions effect
Intercept 0.132* 0.250* 0.086* 0.138* −0.085* −0.051* −0.047* −0.032* 0.096* 0.058* 0.088* 0.110* −0.038* −0.013 0.009 −0.013
Gender_male (1=male, 0=female) −0.001 −0.002 −0.009 −0.020 0.031* 0.031* 0.014* 0.026* −0.014 0.018 0.013* 0.030* 0.011 0.008 −0.002 0.029*
Age_group18_34 0.028* 0.033* 0.028* 0.074* 0.021* 0.015 0.028* 0.036* 0.027 0.061* 0.036 0.074* 0.017 0.006 0.042* 0.071*
Income_high (1=top half within each country, 0=bottom half) −0.017 −0.014 0.006 0.009 −0.001 0.021* 0.014 0.020* 0.040* 0.045* 0.021* 0.039 0.030* 0.009 0.032* 0.028
Country_Brazil_China_India_Russia −0.039* −0.077* 0.063* 0.100* 0.026 −0.016 0.048* 0.041* −0.013 0.078* 0.004 0.049* 0.039* 0.020 −0.001 0.022
Moderators of the negative emotions effect
Intercept −0.120* −0.178* −0.014 −0.064* 0.178* 0.347* 0.094* 0.164* −0.002 −0.008 −0.041* −0.063* 0.162* 0.078* 0.035* 0.061*
Gender_male (1=male, 0=female) 0.023* 0.021 0.005 0.009 −0.022* −0.041* −0.008 −0.011 0.006 −0.006 0.025* 0.014 0.016 0.063* 0.021* 0.073*
Age_group18_34 0.021 0.021* 0.000 0.019 −0.011 0.024 0.000 0.015 0.021* 0.032* 0.006 0.017 −0.007 0.054* −0.003 0.004
Income_high (1=top half within each country, 0=bottom half) −0.014* −0.024* −0.017* −0.005 0.026* 0.040* 0.003 0.014 −0.011 −0.015* −0.003 −0.004 0.009 −0.008 −0.007 −0.005
Country_Brazil_China_India_Russia 0.063* 0.141* 0.026* 0.036* 0.005 0.092* 0.074* 0.204* 0.046* 0.035* 0.031* 0.066* 0.028* 0.135* 0.071* 0.193*
Moderators of the affective commitment effect
Intercept 0.082* 0.160* 0.029* 0.035* 0.003 −0.046* −0.001 0.002 0.062* 0.053* −0.053 0.083* 0.068* −0.016* −0.021* 0.012
Gender_male (1=male, 0=female) 0.080* 0.101* 0.030* 0.058* −0.028* 0.013 −0.035* −0.033* −0.074* −0.025 0.015* 0.045* 0.051 −0.005 −0.014 −0.018
Age_group18_34 −0.022 −0.051* −0.030 −0.062* −0.032 −0.013 −0.033* −0.047* −0.013 −0.055* −0.021 −0.017 −0.069* −0.045* −0.019 −0.028
Income_high (1=top half within each country, 0=bottom half) 0.029* 0.054* 0.008 0.016 0.032* −0.024 −0.042 −0.028 0.002 0.021 −0.052* −0.034 −0.010 0.008 −0.003 −0.026*
Country_Brazil_China_India_Russia −0.014 −0.008 −0.092* −0.085* −0.029* −0.045* −0.016 −0.046* −0.012 0.003 0.021 0.088* −0.012 −0.035 0.025 0.013
Moderators of the calculative commitment effect
Intercept 0.030* 0.063* −0.016 0.004 0.037* −0.002 0.007 −0.007 −0.012 0.007 −0.028* −0.029* −0.025* −0.009 −0.029* −0.021*
Gender_male (1=male, 0=female) −0.029 −0.052* 0.001 0.008 −0.009 −0.004 0.019* 0.004 0.005 −0.002 −0.049* −0.056* 0.010 0.003 −0.003 −0.024*
Age_group18_34 −0.003 0.038* 0.054* 0.078* 0.037 0.059* 0.011 0.029 0.036* 0.047* 0.040* 0.040* 0.090* 0.055* 0.030 0.032
Income_high (1=top half within each country, 0=bottom half) −0.036 −0.061* −0.012 −0.037 −0.041* −0.011 0.022* 0.019 0.020 0.023 −0.027* −0.039* 0.011 0.004 −0.012 0.011
Country_Brazil_China_India_Russia 0.005 0.009 0.036 0.060* 0.001 0.037* −0.025 −0.012 −0.034* −0.043 0.015 0.006 −0.017 0.007 −0.007 −0.028*
Moderators of the self-brand connection effect
Intercept −0.003 −0.011 0.040* 0.044* 0.010 0.019 0.030* 0.007 0.050* 0.048* −0.014 0.019* 0.159* 0.028* 0.038* 0.017
Gender_male (1=male, 0=female) −0.031* −0.022* −0.001 −0.002 0.001 −0.015 0.029 0.035* 0.024* 0.012 0.037* 0.028* −0.047* 0.004 0.028* 0.036*
Age_group18_34 −0.022* −0.056* −0.022* −0.023* −0.033* −0.037* 0.006 0.018 0.048* 0.078* 0.003 0.010 0.019 0.051* 0.011 0.026
Income_high (1=top half within each country, 0=bottom half) 0.064* 0.073* 0.047* 0.071* 0.005 0.012 0.017 0.004 0.004 −0.016 0.056* 0.059* −0.018 −0.010 0.011 0.010
Country_Brazil_China_India_Russia 0.045* 0.068* 0.094* 0.158* 0.054* 0.069* 0.006 0.051* 0.103* 0.228* 0.046* 0.083* 0.051* 0.085* −0.001 0.048*

Notes: aTo facilitate the interpretation of the moderating influence of customer heterogeneity on the drivers, this table also reports the fixed slopes for these variables, which are obtained from a model in which the parameter estimates for the antecedent constructs are not allowed to be customer-specific (i.e. fixed).*Significant, indicating that the 95 percent credibility interval of the posterior density does not contain 0

Attitudinal drivers

Mean
Constructs (SD) AVE CR
Customer satisfaction (Mittal et al. (1999) 7.5 (1.93) 1 1
Overall satisfaction
Commitment (Gustafsson et al., 2005)
Affective commitment 6.3 (2.27) 0.85 0.94
 1. I take pleasure in being a customer of (company/brand)
 2. (company/brand) is the provider that takes the best care of its customers
 3. I get back what I put into my relationship with (company/brand)
Calculative commitment 5.7 (2.24) 0.73 0.84
 1. It pays off economically to be a customer of (company/brand)a
 2. I would suffer economically if the relationship were broken
 3. The economic benefit of dealing with (company/brand) is more than the costs
Self-brand connection (Escalas and Bettman, 2005) 5.0 (2.57) 0.86 0.98
 1. This brand reflects who I am
 2. I can identify with this brand
 3. I feel a personal connection to this brand
 4. This brand conveys to other people who I am
 5. This brand helps me become the type of person I want to be
 6. I consider this brand to be “me” (it reflects who I consider myself to be or the way that I want to present myself to others)
 7. This brand stands for values that are very consistent with “me”
 8. If this brand were a person, I would have a very close connection with that person
Emotions (Oliver, 1993, 2010)
Positive emotions 5.4 (2.68) 0.82 0.98
 Fulfilled, Peaceful, Delighted, Thrilled, Happy, Loved, Desired, Warm-Hearted, Pride, Important, Self-respect
Negative emotions 2.2 (1.9) 0.79 0.97
 Angry, Irritated, Regret, Afraid, Nervous, Sad, Helpless, Miserable, Embarrassed, Humiliated

Notes: AVE, average variance extracted; CR, composite reliability. aThis item was removed due to cross-loadings (confirmatory factor analysis)

Error covariance matrix for εij and γij

εij WOMF1 εij WOMF2 εij WOMF3 εij WOMF4 εij WOMF5 εij WOMF6 εij WOMF7 εij WOMF8 γij WOMF1 γij WOMF2 γij WOMF3 γij WOMF4 γij WOMF5 γij WOMF6 γij WOMF7
εij WOMF2 0.021*
εij WOMF3 0.011 0.015*
εij WOMF4 0.013* 0.031* 0.030*
εij WOMF5 0.011* 0.015* 0.032* 0.013*
εij WOMF6 0.022* 0.034* 0.020* 0.030* 0.022*
εij WOMF7 0.019* −0.003 0.024* 0.024* 0.021* 0.004
εij WOMF8 0.032* 0.030* 0.037* 0.053* 0.033* 0.188* 0.047*
γij WOMF1 1.738* 0.158* 0.105* 0.045* 0.082* 0.190* 0.058 0.057*
γij WOMF2 0.398* 1.534* 0.005 0.087* 0.076 0.042* 0.089 0.039 1.163*
γij WOMF3 0.060 0.030 1.341* 0.139* 0.037* 0.035* 0.175 0.076* 0.350* 0.250*
γij WOMF4 0.066* 0.105* 0.286* 1.306* 0.020 0.044* 0.100* 0.116* 0.342* 0.458* 0.832*
γij WOMF5 0.089* 0.068* 0.020 0.010 1.309* 0.039* 0.164 0.039* 0.533* 0.476* 0.217* 0.237*
γij WOMF6 0.040* 0.039* 0.014 0.127* 0.237* 1.527* 0.060* 0.541* 0.702* 0.545* 0.244* 0.385* 0.868*
γij WOMF7 0.002 0.014* 0.063* 0.048* 0.115* 0.011* 0.723* 0.035* 0.120* 0.162* 0.451* 0.335* 0.505* 0.240*
γij WOMF8 −0.002 0.030 0.040 0.221* 0.096* 0.363* 0.132* 1.352* 0.277* 0.298* 0.485* 0.648* 0.511* 1.460* 0.385*

Note: *Significant, indicating that the 95 percent credibility interval of the posterior density does not contain 0

Appendix 1. Measurement of WOM drivers

Customer satisfaction was measured via an overall satisfaction measure (1=completely dissatisfied, 10=completely satisfied) (Mittal et al., 1999), items for affective commitment and calculative commitment were adopted from research by Gustafsson et al. (2005) and items for positive and negative emotions were adopted from Oliver (1993, 2010). There were eight primary emotional states anger, fear, sadness, shame, contentment, happiness, love, and pride each measured by multiple items. Respondents evaluated the extent to which they experienced the emotion as it relates to the brand. Items for self-brand connection were adopted from Escalas and Bettman (2005). All items were translated and back translated before being administered in each country. The investigation assessed all antecedent constructs for the underlying factor structure by means of a factor analysis. In line with prior literature, items loaded on constructs as intended, resulting in the following six factors: customer satisfaction, affective commitment, calculative commitment, and self-brand connection. Emotions split into two groups representing positive and negative emotions. In addition, discriminant validity between the WOM dimensions and the antecedent constructs was observed since the AVEs for each construct exceeded the squared correlation between the factors (Fornell and Larcker, 1981). Table AI provides details on the items of the antecedent constructs in addition to reliability and AVE statistics.

Appendix 2. Influence of customer characteristics on the impact that the drivers have on various WOM behaviors

As is clear from Table IV, demographic variables moderate the impact that all drivers have on WOM behaviors. Nevertheless, the meaning of what a favorable moderating effect of customer characteristics entails depends upon whether or not the WOM behaviors are wanted (i.e. positive WOM behaviors) or unwanted (i.e. negative WOM behaviors). More precisely, a moderated and more positive effect of a driver is favorable when it impacts a positive WOM behavior, whereas a moderated and more negative effect of a driver is favorable when its outcome is a negative WOM behavior. The next paragraphs discuss the influence of customer characteristics in favoring the impact that the drivers have on the various WOM behaviors.

Customer characteristics impacting the drivers of giving positive WOM to strong ties (WOMF1)

First, for female customers, customers younger than 35 years, customers having lower incomes and customers living in developing countries, the findings reveal that satisfaction translates into more WOMF1. Second, for younger customers, and customers living in industrialized countries, the impact of positive emotions on WOMF1 is stronger. Third, the positive influence of affective commitment on WOMF1 is higher for older individuals, males and individuals with higher incomes. Fourth, the positive effect of calculative commitment on WOMF1 is strongest for young females with lower incomes, but this effect is only observed for the volume of WOMF1. Fifth, although no positive population-averaged effect was found for the impact of self-brand connection on WOMF1, a significant positive relationship for females, aged 35+, having higher incomes and living in developing countries was observed. Finally, although the population-averaged effect of negative emotions on WOMF1 was found to be significantly negative (−0.054 and −0.039 for WOMF1 incidence and volume, respectively), this destructive effect becomes nonexistent for young males with low incomes living in the developing countries (−0.013 and 0.005 for WOMF1 incidence and volume, respectively).

Customer characteristics impacting the drivers of receiving positive WOM from strong ties (WOMF2)

The moderating impact of customer characteristics on WOMF2 is similar but less pronounced than that of WOMF1. In contrast with WOMF1, age has no moderating influence on the impact that satisfaction and negative emotions have on WOMF2; gender doesn’t impact the influence of negative emotions, calculative commitment, and self-brand connection on WOMF2; income doesn’t impact the influence of both types of commitment on WOMF2; but the impact of affective commitment on WOMF2, however, is found to be weaker for developing countries.

Customer characteristics impacting the drivers of giving positive WOM to weak ties (WOMF5)

The findings reveal that the moderating influence of customer characteristics on the relationship between antecedent constructs and positive WOM is different for WOM directed toward strong (WOMF1) vs weak (WOMF5) ties. First, the impact of satisfaction on WOMF5 is more favorable for males, aged 35+, with lower incomes, and living in the developing countries. Interestingly, for this segment the impact of satisfaction on WOMF5 becomes positive (0.029), whereas the population-averaged finding was found to be significantly negative (−0.033). Second, positive emotions lead to more WOMF5 for younger individuals living in developing countries and having higher incomes, whereas for negative emotions a stronger positive influence on WOMF5 for younger individuals living in developing countries who have lower incomes was observed. Third, the positive influence of affective commitment on WOMF5 is stronger for females age 35+. Fourth, the positive impact of calculative commitment on WOMF5 is stronger for younger customers living in industrialized (developed) countries. Finally, self-brand connection results in more WOMF5 for young male customers who live in developing countries.

Customer characteristics impacting the drivers of receiving positive WOM from weak ties (WOMF6)

As mentioned earlier, the moderating influence of customer characteristics on giving and receiving of positive WOM to strong ties is quite similar. In contrast, with respect to weak ties, more differences in the moderating role that customer characteristics has on giving (WOMF5) vs receiving (WOMF6) positive WOM was observed. This finding might be caused by the fact that the mediums to receive (WOMF6) this type of WOM are much broader (also including newspapers, magazines, and TV; see Table II) than the mediums to give positive WOM via weak ties (WOMF5). First, the impact of satisfaction on WOMF6 is more positive for customers aged 35+, who live in developing countries and have higher incomes. As a result, the negative population-averaged effect becomes favorable for this particular segment of customers (i.e. close to zero for WOMF6 incidence, and positive for WOMF6 volume). Second, for male customers from developing countries, the impact of both positive and negative emotions on WOMF6 is more pronounced. In addition, the positive influence of positive emotions on WOMF6 is stronger for younger individuals with higher incomes. Third, the positive effects of affective commitment and self-brand connection on WOMF6 are stronger for male customers living in developing countries. The impact of affective commitment is also stronger for individuals with high incomes, whereas for self-brand connection, its impact is stronger for lower incomes. Finally, the impact of calculative commitment on WOMF6 becomes significantly positive for young female customers with lower incomes.

Customer characteristics impacting the drivers of giving negative WOM to strong ties (WOMF3)

For negative WOM behaviors, negative effects of the drivers of WOM are preferred, as well as moderating effects that (i) make the positive effects on negative WOM behaviors less positive, or (ii) make the negative effects on negative WOM behaviors more negative such that customers are less likely to engage in these unfavorable behaviors. First, satisfaction is likely to reduce WOMF3 to a larger extent for older individuals living in industrialized countries. Second, for females, aged 35+ with lower incomes, positive emotions lower WOMF3, whereas no such favorable effect was found for the entire sample (i.e. non-significant population-averaged effect). Third, the diminishing impact of affective commitment on WOMF3 is more pronounced for male customers with lower incomes living in developing countries. Fourth, in contrast with the population-averaged effect, calculative commitment is likely to reduce WOMF3 for the segment of customers aged 35+ having higher incomes and living in industrialized countries. In a similar vein, self-brand connection can translate into less WOMF3 for younger individuals living in the industrialized countries. Finally, customers with negative emotions report WOMF3, but this negative effect is less salient for male customers with low incomes who live in industrialized countries.

Customer characteristics impacting the drivers of receiving negative WOM from strong ties (WOMF4)

In line with WOMF3, the negative impact of satisfaction on WOMF4 is more pronounced for older individuals living in industrialized countries. For all other WOM drivers, however, there is evidence of another moderating influence of customer characteristics on giving (WOMF3) vs receiving (WOMF4) negative WOM via strong ties. First, positive emotions translate into less WOMF4 for female customers, aged 35+ with low incomes living in industrialized countries. Second, the harmful impact of negative emotions in augmenting WOMF4 is less pronounced in industrialized countries. Third, for younger male customers living in developing countries, affective commitment is likely to lower WOMF4. Fourth, for female customers with lower incomes, calculative commitment is likely to reduce WOMF4. Finally, the positive impact of self-brand connection on WOMF4 is found to be weaker for female customers living in industrialized countries.

Customer characteristics impacting the drivers of giving negative WOM to weak ties (WOMF7)

First, satisfied customers are less inclined to give negative WOM to weak ties WOMF7), and this favorable effect is more pronounced for younger individuals with higher incomes. Second, positive emotions are likely to reduce WOMF7 for customers with lower incomes living in industrialized countries. Third, the extent to which negative emotions stimulate WOMF7 is less pronounced for female customers, aged 35+ living in industrialized countries. Fourth, the negative impact of affective commitment on WOMF7 is stronger for younger customers, whereas calculative commitment is found to reduce WOMF7 for older customers. Finally, the positive impact of self-band connection on WOMF7 is less strong for male customers, aged 35+, living in industrialized countries.

Customer characteristics impacting the drivers of receiving negative WOM from weak ties (WOMF8)

First, the diminishing impact of satisfaction on WOMF8 is stronger for younger customers living in industrialized countries. Second, the positive population-averaged effect of positive emotions on WOMF8 fades out for female customers, aged 35+ with lower incomes. Third, the unfavorable positive relationship between negative emotions and WOMF8 weakens for female customers living in industrialized countries. Fourth, in contrast with the population-averaged findings, affective commitment is likely to reduce WOMF8 for customers with higher incomes, whereas calculative commitment is likely to reduce WOMF8 for male customers who live in developing countries. Finally, the positive impact of self-band connection on WOMF8 is less strong for female customers living in industrialized countries.

Appendix 3. Model estimation details

Common method bias

Since respondents provided information on a range of WOM behaviors and attitudinal drivers (satisfaction, positive and negative emotion, commitment and self-brand connection), there is potential for common method bias. In line with the recommendations of Podsakoff et al. (2003), the potential for common method bias was reduced by: using measures based on existing scales for the attitudinal drivers and a careful construction of the items for the WOM behaviors following the procedure outlined by Netemeyer et al. (2003); proximally separating measures of predictors and criterion variables; and protecting the respondents’ anonymity. Additionally, a Harmon’s single-factor test using exploratory factor analysis was conducted to check whether a single factor emerged or one general factor accounted for the majority of the covariance among the measures. The results showed nine factors, by which the first factor accounted for 33.2% of the variance and all factors together explained 78.8% of the variance. None of these factors accounted thus for the majority of the covariance among the items, as a result of which the common method bias was not a serious threat to the analyses (Podsakoff et al., 2003). (The use of marker variables (Williams et al., 2010) was not possible as the data set did not include any variables that met the necessary conditions).

Testing the need for using hierarchical models

Since different company evaluations are nested within customers, random-effects models for each WOM dependent were estimated, as suggested by Raudenbush and Bryk (2002). For each dependent variable, the random intercept variance was found to be significant, and the intra-class correlations (ICC) varied from low to moderate. For the negative WOM dimensions (WOMF3, WOMF4, WOMF7 and WOMF8), the ICC ranged between 8.8 and 21.8%, whereas substantially higher ICC were observed for the positive WOM dimensions (WOMF1, WOMF2, WOMF5 and WOMF6), ranging between 43.3 and 56.1%. Due to the significant random intercept variances and the moderate to low ICC, hierarchical models are chosen for the analyses in the manuscript.

Collinearity

To check for multicollinearity, OLS regressions were run to generate variance inflation factors (VIF). The VIF values of all variables were below the suggested cutoff of 5 (Hair et al., 2010). In addition, stepwise variable selection techniques (VST) were applied for each WOM dimension and compared the sign and magnitude of the selected variables with the estimates and significance levels of the parameters in the corresponding full models that included all independent variables simultaneously. The VST and full models for each WOM dependent revealed similar findings in terms of parameter significance, sign and magnitude for each key variable. Based on the VIF values and VST analyses, multicollinearity is not a significant problem in the data set (Chatterjee et al., 2000; Hair et al., 2010).

Error covariance matrix

The WOM dependents are estimated in a multivariate model, and do account for correlated errors between these dependent variables: εij and γij. As recommended in the literature, the investigation used the Inverse-Wishart for the covariance matrix (Asparouhov and Muthen, 2010). Table AII reports that covariance matrix of the error terms following the joint estimation of Equations 1 and 2.

In addition, the error terms δi and ϕi (Equations 3 and 4) were estimated. Inspection of the error estimates reveals that all modeled residual variances were significant, ranging from 0.001 to 0.072.

Convergence assessment

To assess the convergence of the MCMC algorithm, autocorrelation and trace plots of the residual variance were inspected for all key variables.

The basic idea is that although a chain may look different at early iterations because of different starting points, when the MCMC algorithm is converged, the chains should mix together and become indistinguishable from each other as they converge to the same posterior distribution. For this analysis, three MCMC chains with 100,000 iterations each are estimated, of which the first half is considered as the “burn-in” phase, and the remaining half is used to estimate the posterior distribution for the parameters, resulting in a distribution based on 150,000 points. Autocorrelations denote the degree of correlatedness of parameter values across iterations of different lags (i.e. intervals in the chain). If the chains are mixing satisfactorily then the autocorrelations in the one-step apart iterates will fade to zero as the lag increases (e.g. at lag 10 or 20) (Congdon, 2006). Therefore, a small autocorrelation value is desirable to obtain approximately independent draws from the posterior, and a value of 0.1 or lower has been suggested (Muthén, 2010). As an example, Figure A1 plots the autocorrelation plots for the effect of affective commitment on giving positive WOM to strong ties (WOMF1) for MCMC chain 1.

Finally, the inspection of the residual variance of the estimated parameters is assessed. Trace plots of the different chains for the residual variance that show a stable process with no upward or downward trend (i.e. a thick horizontal line) indicate model convergence (Muthén, 2010). As an example, Figure A2 plots the trace plot for the impact of affective commitment on giving positive WOM to strong ties (WOMF1).

References

Aksoy, L. (2013), “How do you measure what you can’t define? The current state of loyalty measurement and management”, Journal of Service Management, Vol. 24 No. 4, pp. 356-381.

Allen, N. and Meyer, J. (1990), “The measurement and antecedents of affective, continuance, and normative commitment to the organization”, Journal of Occupational Psychology, Vol. 63 No. 1, pp. 1-18.

Anderson, E. (1998), “Customer satisfaction and word-of-mouth”, Journal of Service Research, Vol. 1 No. 1, pp. 1-14.

Asparouhov, T. and Muthén, B. (2010), “Bayesian analysis using mplus: technical implementation (Technical appendix)”, Muthén & Muthén, Los Angeles, CA, available at: http://statmodel2.com/download/Bayes3.pdf (accessed December 13, 2017).

Baker, A., Donthu, N. and Kumar, V. (2016), “Investigating how word-of-mouth conversations about brands influence purchase and retransmission intentions”, Journal of Marketing Research, Vol. 53 No. 2, pp. 225-239.

Barnlund, D. (2008), “A transactional model of communication”, in Mortensen, C.D. (Ed.), Communication Theory, 2nd ed., New Brunswick, NJ, pp. 47-57.

Barreto, A. (2014), “The word-of-mouth phenomenon in the social media era”, International Journal of Market Research, Vol. 56 No. 5, pp. 631-654.

Belk, R. (1988), “Possessions and the extended self”, Journal of Consumer Research, Vol. 15 No. 2, pp. 139-168.

Berger, J. (2013), Contagious: Why Things Catch On, Simon & Schuster, New York, NY.

Berger, J. and Iyengar, R. (2013), “Communication channels and word of mouth: how the medium shapes the message”, Journal of Consumer Research, Vol. 40 No. 3, pp. 567-579.

Berger, J. and Milkman, K.L. (2012), “What makes online content viral?”, Journal of Marketing Research, Vol. 49 No. 2, pp. 192-205.

Berlo, D. (1960), The Process of Communication: An Introduction to Theory and Practice, Holt, Rinehart and Winston, New York, NY.

Blazevic, V., Hammedi, W., Garnefeld, I., Rust, R. and Keiningham, T.L. (2013), “Beyond traditional word of mouth”, Journal of Service Management, Vol. 24 No. 3, pp. 294-313.

Bowman, D. and Narayandas, D. (2001), “Managing customer-initiated contacts with manufacturers: the impact on share of category requirements and word-of-mouth behavior”, Journal of Marketing Research, Vol. 38 No. 3, pp. 281-297.

Brown, T.J., Barry, T.E., Dacin, P.A. and Gunst, R.F. (2005), “Spreading the word: investigating antecedents of consumers’ positive word-of-mouth intentions and behaviors in a retailing context”, Journal of the Academy of Marketing Science, Vol. 33 No. 2, pp. 123-138.

Bruce, N.I., Foutz, N.Z. and Kolsarici, C. (2012), “Dynamic effectiveness of advertising and word-of-mouth in sequential distribution of new products”, Journal of Marketing Research, Vol. 49 No. 4, pp. 469-486.

Byrne, B.M. (1998), Structural Equation Modeling with LISREL, PRELIS, and SIMPLIS: Basic Concepts, Applications, and Programming, Lawrence Erlbaum, Mahwah, NJ.

Chatterjee, S., Hadi, A.S. and Price, B. (2000), Regression Analysis by Example (3rd Ed.), John Wiley and Sons, Inc., New York, NY.

Chen, Y., Wang, Q. and Xie, J. (2011), “Online social interactions: a natural experiment on word-of-mouth versus observational learning”, Journal of Marketing Research, Vol. 48 No. 2, pp. 238-354.

Chen, Z. and Berger, J. (2013), “When, why, and how controversy causes conversation”, Journal of Consumer Research, Vol. 40 No. 3, pp. 580-593.

Chen, Z. and Lurie, N. (2013), “Temporal contiguity and negativity bias in the impact of online word of mouth”, Journal of Marketing Research, Vol. 50 No. 4, p. 463.

Cheung, M.Y., Luo, C., Sia, C.L. and Chen, H. (2009), “Credibility of electronic word-of-mouth: informational and normative determinants of on-line consumer recommendations”, International Journal of Electronic Commerce, Vol. 13 No. 4, pp. 9-38.

Chevalier, J. and Mayzlin, D. (2006), “The effect of word-of-mouth on sales: online book reviews”, Journal of Marketing Research, Vol. 43 No. 3, pp. 345-354.

Chung, C.M.Y. and Darke, P.R. (2006), “The consumer as advocate: self relevance, culture and word-of-mouth”, Marketing Letters, Vol. 17 No. 4, pp. 269-279.

Clark, L.A. and Watson, D. (1995), “Constructing validity: basic issues in scale development”, Psychological Assessment, Vol. 7 No. 3, pp. 309-319.

Congdon, P. (2006), Bayesian Statistical Modeling, 2nd ed., John Wiley & Sons, West Sussex.

Cooil, B., Keiningham, T.L., Aksoy, L. and Hsu, M. (2007), “A longitudinal analysis of customer satisfaction and share of wallet: investigating the moderating effect of customer characteristics”, Journal of Marketing, Vol. 71 No. 1, pp. 67-83.

Cui, G., Chan, T.S. and Joy, A. (2008), “Consumers’ attitudes toward marketing: a cross-cultural study of China and Canada”, Journal of International Consumer Marketing, Vol. 20 Nos 3-4, pp. 81-93.

De Angelis, M., Bonezzi, A., Peluso, A.M., Rucker, D.D. and Costabile, M. (2012), “On braggarts and gossips: a self-enhancement account of word-of-mouth generation and transmission”, Journal of Marketing Research, Vol. 49 No. 4, pp. 551-563.

De Matos, C.A. and Rossi, C.A.V. (2008), “Word-of-mouth communications in marketing: a meta-analytic review of the antecedents and moderators”, Journal of the Academy of Marketing Science, Vol. 36 No. 4, pp. 578-596.

Dichter, E. (1966), “How word-of-mouth advertising works”, Harvard Business Review, Vol. 44 No. 6, pp. 147-166.

Dick, A.S. and Basu, K. (1994), “Customer loyalty: toward an integrated conceptual framework”, Journal of the Academy of Marketing Science, Vol. 22 No. 2, pp. 99-113.

East, R. and Lomax, W. (2010), “Demographic bases of word of mouth,” paper presented at Proceedings of the Australian and New Zealand Marketing Academy (ANZMAC) Conference, University of Canterbury, College of Business and Economics, Department of Management, Christchurch, available at: www.anzmac.org/conference_archive/2010/pdf/anzmac10Final00007.pdf (accessed July 29, 2016).

East, R., Hammond, K. and Wright, M. (2007), “The relative incidence of positive and negative word-of-mouth: a multi-category study”, International Journal of Research in Marketing, Vol. 24 No. 2, pp. 175-184.

East, R., Uncles, M., Romaniuk, J. and Dall’Olmo, F. (2015), “Factors associated with the production of word of mouth”, International Journal of Market Research, Vol. 57 No. 3, pp. 439-458.

Easterbrook, P.J., Gopalan, R., Berlin, J.A. and Matthews, D.R. (1991), “Publication bias in clinical research”, The Lancet, Vol. 337 No. 8746, pp. 867-872.

Escalas, J. and Bettman, J. (2005), “Self-construal, reference groups and brand meaning”, Journal of Consumer Research, Vol. 32 No. 3, pp. 378-389.

Fornell, C. and Larcker, D.F. (1981), “Evaluating structural equation models with unobservable variables and measurement error”, Journal of Marketing Research, Vol. 18 No. 1, pp. 39-50.

Garbarino, E. and Johnson, M.S. (1999), “The different roles of satisfaction, trust, and commitment in customer relationships”, Journal of Marketing, Vol. 63 No. 2, pp. 70-87.

Gatignon, H. and Robertson, T.S. (1986), “An exchange theory model of interpersonal communication”, in Lutz, R.J. (Ed.), NA – Advances in Consumer Research, Vol. 13, Association for Consumer Research, Provo, UT, pp. 534-538.

Gelman, A. and Rubin, D.B. (1992), “Inference from iterative simulation using multiple sequences”, Statistical Science, Vol. 7 No. 4, pp. 457-472.

Godes, D. and Mayzlin, D. (2004), “Using online conversation to study word-of-mouth communication”, Marketing Science, Vol. 23 No. 4, pp. 545-560.

Godes, D. and Silva, J.C. (2011), “Sequential and temporal dynamics of online opinion”, Marketing Science, Vol. 31 No. 3, pp. 448-473.

Goldstein, M. (1995), “Bayes linear methods I – adjusting beliefs: concepts and properties”, Part 1 of online tutorial, available at: www.jstatsoft.org/article/view/v005i02/blm1.pdf (accessed July 28, 2016).

Grauer, Y. (2014), “Three highly effective word-of-mouth marketing tactics”, Vertical Response Blog, available at: www.verticalresponse.com/blog/3-highly-effective-word-of-mouth-marketing-tactics/ (accessed July 18, 2016).

Gustafsson, A., Johnson, M.D. and Roos, I. (2005), “The effects of customer satisfaction, relationship commitment dimensions, and triggers on customer retention”, Journal of Marketing, Vol. 69 No. 4, pp. 210-218.

Hair, J.F., Black, W.C., Babin, B.J. and Anderson, R.E. (2010), Multivariate Data Analysis, 7th ed., Pearson Prentice Hall, Upper Saddle River, NJ.

Hamilton, R., Vohs, K. and McGill, A. (2014), “We’ll be honest, this won’t be the best article you’ll ever read: the use of dispreferred markers in word-of-mouth communication”, Journal of Consumer Research, Vol. 41 No. 1, pp. 197-212.

Harrison-Walker, J.L. (2001), “The measurement of word-of-mouth communication and an investigation of service quality and customer commitment as potential antecedents”, Journal of Service Research, Vol. 4 No. 1, pp. 60-75.

Heath, C., Bell, C. and Sternberg, E. (2001), “Emotional selection in memes: the case of urban legends”, Journal of Personality and Social Psychology, Vol. 81 No. 6, pp. 1028-1041.

Hennig-Thurau, T., Gwinner, K.P., Walsh, G. and Gremler, D.D. (2004), “Electronic word-of-mouth via consumer-opinion platforms: what motivates consumers to articulate themselves on the internet”, Journal of Interactive Marketing, Vol. 18 No. 1, pp. 38-52.

Hennig-Thurau, T., Wiertz, C. and Feldhaus, F. (2015), “Does Twitter matter? The impact of microblogging word of mouth on consumers’ adoption of new movies”, Journal of Academy of Marketing Science, Vol. 43 No. 3, pp. 375-394.

Herr, P.M., Kardes, F.R. and Kim, J. (1991), “Effects of word-of-mouth and product attribute information on persuasion: an accessibility-diagnosticity perspective”, Journal of Consumer Research, Vol. 17 No. 4, pp. 454-462.

Hewett, K., Rand, W., Rust, R.T. and van Heerde, H.J. (2016), “Brand buzz in the echoverse”, Journal of Marketing, Vol. 80 No. 3, pp. 1-24.

Ho-Dac, N., Carson, S.J. and Moore, W.L. (2013), “The effects of positive and negative online customer reviews: do brand strength and category maturity matter?”, Journal of Marketing, Vol. 77 No. 6, pp. 37-53.

Hogan, J.E., Lemon, K.N. and Libai, B. (2004), “Quantifying the ripple: word-of-mouth and advertising effectiveness”, Journal of Advertising Research, Vol. 44 No. 3, pp. 271-280.

Iyengar, R., Van den Bulte, C. and Lee, J.Y. (2015), “Social contagion in new product trial and repeat”, Marketing Science, Vol. 34 No. 3, pp. 408-429.

Jankowski, P. (2013), “4 tactics to build your word of mouth”, Forbes, March 13, available at: www.forbes.com/sites/pauljankowski/2013/03/13/4-tactics-to-build-your-word-of-mouth/#1e2722d72745 (accessed July 15, 2016).

Kaynak, E. (1986), Service Industries in Developing Countries, Routledge, New York, NY.

Kaynak, E. and Jain, S.C. (1993), Market Evolution in Developing Countries: The Unfolding of the Indian Market, The Hayworth Press, Binghamton, NY.

Keiningham, T.L., Cooil, B., Aksoy, L., Andreassen, T.W. and Weiner, J. (2007), “The value of different customer satisfaction and loyalty metrics in predicting customer retention, recommendation, and share-of-wallet”, Managing Service Quality, Vol. 17 No. 4, pp. 361-384.

Keiningham, T.L., Frennea, C.M., Aksoy, L., Buoye, A. and Mittal, V. (2015), “A five-component customer commitment model: implications for repurchase intentions in goods and services industries”, Journal of Service Research, Vol. 18 No. 4, pp. 433-450.

King, R.A., Racherla, P. and Bush, V.D. (2014), “What we know and don’t know about online word-of-mouth: a review and synthesis of the literature”, Journal of Interactive Marketing, Vol. 28 No. 3, pp. 167-183.

Kumar, A., Bezawada, R., Rishika, R., Janakiraman, R. and Kannan, P.K. (2016), “From social to sale: the effects of firm-generated content in social media on customer behavior”, Journal of Marketing, Vol. 80 No. 1, pp. 7-25.

Larivière, B., Keiningham, T.L., Cooil, B., Aksoy, L. and Malthouse, E.C. (2014), “A longitudinal examination of customer commitment and loyalty”, Journal of Service Management, Vol. 25 No. 1, pp. 75-100.

Larivière, B., Keiningham, T.L., Cooil, B., Aksoy, L., Yalcin, A., Morgeson, F.V. III and Mithas, S. (2016), “Modeling heterogeneity in the satisfaction, loyalty intention, and shareholder value linkage: a cross-industry analysis at the customer and firm levels”, Journal of Marketing Research, Vol. 53 No. 1, pp. 91-109.

Levy, S. (1959), “Symbols for sale”, Harvard Business Review, Vol. 37 No. 4, pp. 117-124.

Libai, B., Bolton, R., Bügel, M.S., de Ruyter, K., Götz, O., Risselada, H. and Stephen, A.T. (2010), “Customer-to-customer interactions: broadening the scope of word-of-mouth research”, Journal of Service Research, Vol. 13 No. 3, pp. 267-282.

Libai, B., Muller, E. and Peres, R. (2013), “Decomposing the value of word-of-mouth seeding programs: acceleration versus expansion”, Journal of Marketing Research, Vol. 50 No. 2, pp. 163-175.

Little, J.D.C. (1970), “Models and managers: the concept of a decision calculus”, Management Science, Vol. 16 No. 8, pp. B466-B485.

Little, J.D.C. (2004), “Comments on ‘models and managers: the concept of a decision calculus’: managerial models for practice”, Management Science, Vol. 50 No. S12, pp. 1854-1860.

Lovett, M.J., Peres, R. and Shachar, R. (2013), “On brands and word of mouth”, Journal of Marketing Research, Vol. 50 No. 4, pp. 427-444.

Lucid Marketing (2006), “US adults word of mouth communication study”, LLC white paper, March, Lucid Marketing, Allentown, NJ, available at: https://greenleaf.org/wp-content/uploads/2013/11/WOM_Communications_Study_Report.pdf (accessed July 30, 2016).

Ludwig, S., de Ruyter, K., Friedman, M., Brüggen, E.C., Wetzels, M. and Pfann, G. (2013), “More than words: the influence of affective content and linguistic style matches in online reviews on conversion rates”, Journal of Marketing, Vol. 77 No. 1, pp. 87-103.

McCracken, G. (1986), “Culture and consumption: a theoretical account of the structure and movement of the cultural meaning of consumer goods”, Journal of Consumer Research, Vol. 13 No. 1, pp. 71-84.

Martin, D., O’Neill, M., Hubbard, S. and Palmer, A. (2008), “The role of emotion in explaining consumer satisfaction and future behavioural intention”, Journal of Services Marketing, Vol. 22 No. 3, pp. 224-236.

Mason, R.B. (2008), “Word of mouth as a promotional tool for turbulent markets”, Journal of Marketing Communications, Vol. 14 No. 3, pp. 207-224.

Mazzarol, T., Sweeney, J.C. and Soutar, G.N. (2007), “Conceptualizing word-of-mouth activity, triggers and conditions: an exploratory study”, European Journal of Marketing, Vol. 41 Nos 11/12, pp. 1475-1494.

Mittal, V. and Kamakura, W.A. (2001), “Satisfaction, repurchase intent, and repurchase behavior: investigating the moderating effect of customer characteristics”, Journal of Marketing Research, Vol. 38 No. 1, pp. 131-142.

Mittal, V., Kumar, P. and Tsiros, M. (1999) “Attribute-level performance, satisfaction, and behavioral intentions over time: a consumption-system approach”, Journal of Marketing, Vol. 63 No. 2, pp. 88-101.

Moliner-Velázquez, B., Ruiz-Molina, M.E. and Fayos-Gardó, T. (2015), “Satisfaction with service recovery: moderating effect of age in word-of-mouth”, The Journal of Consumer Marketing, Vol. 32 No. 6, pp. 470-484.

Moore, S. (2015), “Attitude predictability and helpfulness in online reviews: the role of explained actions and reactions”, Journal of Consumer Research, Vol. 42 No. 1, pp. 30-44.

Morgan, R.M. and Hunt, S.D. (1994), “The commitment-trust theory of relationship marketing”, Journal of Marketing, Vol. 58 No. 3, pp. 20-38.

MSI (2014), “2014-2016 research priorities”, Marketing Science Institute, Cambridge, MA, available at: www.msi.org/research/2014-2016-research-priorities/ (accessed December 1, 2014).

Muthén, B.O. (2010), “Bayesian analysis in mplus: a brief introduction”, available at: www.statmodel.com/download/IntroBayesVersion%203.pdf (accessed December 13, 2017).

Nam, H. and Kannan, P.K. (2014), “The informational value of social tagging networks”, Journal of Marketing, Vol. 78 No. 4, pp. 21-40.

Naylor, G. and Kleiser, S.B. (2000), “Negative versus positive word-of-mouth: an exception to the rule”, Journal of Satisfaction, Dissatisfaction and Complaining Behavior, Vol. 13, pp. 26-36.

Netemeyer, R.G., Bearden, W. and Sharma, S. (2003), Scaling Procedures. Issues and Applications, Sage Publications, Thousand Oaks, CA.

O’Neill, J. (2011), The Growth Map: Economic Opportunity in the BRICs and Beyond, Penguin Group, New York, NY.

Oliver, R.L. (1993), “Cognitive, affective, and attribute bases of the satisfaction response”, Journal of Consumer Research, Vol. 20 No. 3, pp. 418-430.

Oliver, R.L. (2010), Satisfaction: A Behavioral Perspective on the Consumer, 2nd ed., M.E. Sharpe, Armonk, NY.

Plutchik, R. and Conte, H.R. (1997), Circumplex Models of Personality and Emotions, American Psychological Association, Washington, DC.

Podsakoff, P.M., MacKenzie, S.B., Lee, J.-Y. and Podsakoff, N.P. (2003), “Common method biases in behavioral research: a critical review of the literature and recommended remedies”, Journal of Applied Psychology, Vol. 88 No. 5, pp. 879-903.

Pratt, M. (2008), “How important is word-of-mouth advertising”, available at: http://blog.madiganpratt.com/2008/01/07/how-important-is-word-of-mouth-advertising/ (accessed December 13, 2017).

Ranaweera, C. and Jayawardhena, C. (2014), “Talk up or criticize? Customer responses to WOM about competitors during social interactions”, Journal of Business Research, Vol. 67 No. 12, pp. 2645-2656.

Raudenbush, S.W. and Bryk, A.S. (2002), Hierarchical Linear Models: Applications and Data Analysis Methods, 2nd ed., Sage Publications, Thousand Oaks, CA.

Reisenzein, R. (1994), “Pleasure-arousal theory and the intensity of emotions”, Journal of Personality and Social Psychology, Vol. 67 No. 3, pp. 525-539.

Richins, M. (1994), “Valuing things: the public and private meanings of possessions”, Journal of Consumer Research, Vol. 21 No. 3, pp. 504-521.

Risselada, H., Verhoef, P.C. and Bijmolt, T.H.A. (2014), “Dynamic effects of social influence and direct marketing on the adoption of high-technology products”, Journal of Marketing, Vol. 78 No. 2, pp. 52-68.

Rossi, P.E. and Allenby, G.M. (2003), “Bayesian statistics and marketing”, Marketing Science, Vol. 22 No. 3, pp. 304-328.

Rothstein, H.R., Sutton, A.J. and Borenstein, M. (Eds) (2006), Publication Bias in Meta-Analysis: Prevention, Assessment and Adjustments, John Wiley & Sons, Hoboken, NJ.

Schlosser, A.E. (2005), “Can including pros and cons increase the helpfulness and persuasiveness of online reviews? The interactive effects of ratings and arguments”, Journal of Consumer Psychology, Vol. 21 No. 3, pp. 226-239.

Schweidel, D.A. and Moe, W.W. (2014), “Listening in on social media: a joint model of sentiment and venue format choice”, Journal of Marketing Research, Vol. 51 No. 4, pp. 387-402.

Shannon, C.E. and Weaver, W. (1949), The Mathematical Theory of Information, The University of Illinois Press, Champaign, IL.

Söderlund, M. and Rosengren, S. (2007), “Receiving word-of-mouth from the service customer: an emotion-based effectiveness assessment”, Journal of Retailing and Consumer Services, Vol. 14 No. 1, pp. 123-136.

Strizhakova, Y. and Coulter, R.A. (2013), “The ‘green’ side of materialism in emerging BRIC and developed markets: the moderating role of global cultural identity”, International Journal of Research in Marketing, Vol. 30 No. 1, pp. 69-82.

Sundaram, D.S., Mitra, K. and Webster, C. (1998), “Word-of-mouth communications: a motivational analysis”, in Alba, J.W. and Wesley, J. (Eds), Advances in Consumer Research, Vol. 25, Association for Consumer Research, Hutchinson, Provo, UT, pp. 527-531.

Sutter, B. (2015), “5 ways to maximize word-of-mouth marketing”, available at: www.entrepreneur.com/article/251576 (accessed July 15, 2016).

Swan, J.E. and Oliver, R.E. (1989), “Postpurchase communications by consumers”, Journal of Retailing, Vol. 65 No. 4, pp. 516-533.

Tang, T., Fang, E. and Wang, F. (2014), “Is neutral really neutral? The effects of neutral user-generated content on product sales”, Journal of Marketing, Vol. 78 No. 4, pp. 41-58.

TARP (1986), Consumer Complaint Handling in America: An Update Study, Technical Assistance Research Programs, Washington, DC.

Trusov, M., Bucklin, R.E. and Pauwels, K. (2009), “Effects of word-of-mouth versus traditional marketing: findings from an internet social networking site”, Journal of Marketing, Vol. 73 No. 5, pp. 90-102.

Van den Bulte, C. and Wuyts, S. (2009), “Leveraging customer networks”, in Kleindorfer, P.R., Wind, Y.R. and Gunther, R.E. (Eds), The Network Challenge: Strategy, Profit, and Risk in an Interlinked World, Wharton School Publishing, Upper Saddle River, NJ, pp. 243-258.

Wangenheim, F.V. and Bayón, T. (2007), “The chain from customer satisfaction via word-of-mouth referrals to new customer acquisition”, Journal of the Academy of Marketing Science, Vol. 35 No. 2, pp. 233-249.

Wetzer, I.M., Zeelenberg, M. and Pieters, R. (2007), “Never eat in that restaurant, I Did!’: exploring why people engage in negative word-of-mouth communication”, Psychology & Marketing, Vol. 24 No. 8, pp. 661-680.

Wien, A.H. and Olsen, S.O. (2014), “Understanding the relationship between individualism and word of mouth: a self-enhancement explanation”, Psychology & Marketing, Vol. 31 No. 6, pp. 416-425.

Wilcox, K. and Stephen, A.T. (2013), “Are close friends the enemy? Online social networks, self-esteem, and self-control”, Journal of Consumer Research, Vol. 40 No. 1, pp. 90-103.

Williams, L.J., Hartman, N. and Cavazotte, F. (2010), “Method variance and marker variables: a review and comprehensive CFA marker technique”, Organizational Research Methods, Vol. 13 No. 3, pp. 477-514.

Wind, Y. and Douglas, S.P. (1982), “Comparative consumer research: the next frontier”, Management Decision, Vol. 20 No. 4, pp. 24-35.

Yang, S., Hu, M., Winer, R., Assael, H. and Chen, X. (2012), “An empirical study of word-of-mouth consumption”, Marketing Science, Vol. 31 No. 6, pp. 952-963.

You, Y., Vadakkepatt, G. and Joshi, A.M. (2015), “A meta-analysis of electronic word-of-mouth elasticity”, Journal of Marketing, Vol. 79 No. 2, pp. 19-39.

Yuan, Y. and MacKinnon, D.P. (2009), “Bayesian mediation analysis”, Psychological Methods, Vol. 14 No. 4, pp. 301-322.

Zhang, Y., Feick, L. and Mittal, V. (2014), “How males and females differ in their likelihood of transmitting negative word of mouth”, Journal of Consumer Research, Vol. 40 No. 6, pp. 1097-1108.

Further reading

Allsop, D., Bassett, B. and Hoskins, J. (2007), “Word-of-mouth research: principles and applications”, Journal of Advertising Research, Vol. 47 No. 4, pp. 398-411, available at: www.versatilesolutionsbb.com/web_documents/word_of_mouth_research_jar_dec-07.pdf (accessed July 29, 2016).

Beuchel, E. and Berger, J. (2012), “Facebook therapy? Why do people share self-relevant content online?”, in Gürhan-Canli, Z., Otnes, C. and Zhu, R. (Eds), NA-Advances in Consumer Research, Vol. 40, Association for Consumer Research, Duluth, MN, pp. 203-208.

Blakeshaw, P. (2008), Happy Customers Tell Three Friends, Angry Customers Tell 3,000, Doubleday, New York, NY.

Hayes, B.E. (2008), “Customer loyalty 2.0”, Quirk’s Marketing Research Review, Vol. 22 No. 10, pp. 54, 56-58, 60, 62.

Hofstede, G. (1980), “Motivation, leadership, and organization: do American theories apply abroad?”, Organizational Dynamics, Vol. 9 No. 1, pp. 42-63.

Hult, G.T.M., Morgeson III, F.V., Morgan, N.A., Mithas, S. and Fornell, C. (2016), “Do managers know what their customers think and why?”, Journal of the Academy of Marketing Science, Vol. 45 No. 1, pp. 37-54, available at: http://link.springer.com/article/10.1007/s11747-016-0487-4

Lehmann, D.R., McAlister, L. and Staelin, R. (2011), “Sophistication in research in marketing”, Journal of Marketing, Vol. 75 No. 4, pp. 155-165.

Muthén, L.K. and Muthén, B.O. (2010), Mplus: User’s Guide, 6th ed., Muthén & Muthén, Los Angeles, CA.

Paridon, T.J., Carraher, S. and Carraher, S.C. (2006), “The income effect in personal shopping value, consumer self-confidence, and information sharing (word of mouth communication) research”, Academy of Marketing Studies Journal, Vol. 10 No. 2, pp. 107-124.

Richins, M. (1983), “Negative word-of-mouth by dissatisfied consumers: a pilot study”, The Journal of Marketing, Vol. 47 No. 1, pp. 68-78.

Wan, L.C. (2013), “Culture’s impact on consumer complaining responses to embarrassing service failure”, Journal of Business Research, Vol. 66 No. 3, pp. 298-305.

Supplementary materials

JOSM_29_1.pdf (5.9 MB)

Acknowledgements

The authors would like to thank Katrien Verleye, Dries Benoit and the members of the Center for Service Intelligence (Ghent University) for their insightful comments on earlier versions of the paper. This work was carried out using the STEVIN Supercomputer Infrastructure at Ghent University, funded by Ghent University, the Flemish Supercomputer Center (VSC), the Hercules Foundation and the Flemish Government – department EWI.

Corresponding author

Timothy Lee Keiningham is the corresponding author and can be contacted at: tlkeiningham@yahoo.com

About the authors

Timothy Lee Keiningham is the J. Donald Kennedy Endowed Chair in E-Commerce and an Associate Professor of Marketing at the Tobin College of Business, St John’s University. He received the AMA’s Christopher Lovelock Career Contributions to the Services Discipline Award for teaching, research, and service that has had the greatest long-term impact on the development of the services discipline. He has won best article awards from the Journal of Marketing (twice), Journal of Service Research, Journal of Service Management (twice), and Journal of Service Theory and Practice (twice), and the Citations of Excellence “Top 50” Award from Emerald Management Reviews.

Roland T. Rust is a Distinguished University Professor, David Bruce Smith Chair in Marketing, and Executive Director of the Center for Excellence in Service at the Smith School of Business, University of Maryland, the Visiting Chair in Marketing Research at Erasmus University, and an International Research Fellow at Oxford. He has received the top academic honors from AMA, EMAC, and the INFORMS Society for Marketing Science, the Converse Award, as well as top career honors in service marketing, marketing research, marketing strategy, advertising, and statistics, and honorary doctorates from the University of Neuchatel and the Norwegian School of Economics, and was featured in the AMA’s Marketing Legends video series. He has won four best article awards from the Journal of Marketing, as well as the Berry/AMA Book Award. He served as an editor-in-chief JM, and founded the Frontiers in Service Conference and Journal of Service Research, and is currently an editor-in-chief of IJRM.

Bart Lariviere is an Associate Professor of Service Management, and the Founder and Executive Director of the Center for Service Intelligence at Ghent University. He received the AMA’s Emerging Scholar Award for significant contributions to the services discipline. He is co-founder of BAQMaR (the Belgian community for marketing academicians and practitioners) and the Let’s Talk About Service (LTAS) Conference. He has won several best paper awards, including the Finalist Best Paper Award for the Journal of Service Research (twice), the Best Paper Award for the Journal of Service Management (twice), the Best Practitioner Paper Presentation Award at the Frontiers in Service Conference (three times). He also received the Best PhD Tutor Award from Ghent University’s Faculty of Economics and Business Administration.

Lerzan Aksoy is the Associate Dean of Undergraduate Studies and a Professor of Marketing at the Gabelli School of Business, Fordham University. She received the Ten Outstanding Young Persons (TOYP) Award for Scientific Leadership by the Junior Chamber International in Turkey. She has won best article awards from the Journal of Marketing, Journal of Service Management (three times), and Journal of Service Theory and Practice (two times), and the Citations of Excellence “Top 50” Award from Emerald Management Reviews. She has authored/edited five books. Her most recent book is the NY Times bestseller The Wallet Allocation Rule.

Luke Williams is the Head of CX at Qualtrics. He has significant experience in research design, analytics design, interpretive diagnostics, consumer insights, customer experience research and strategy. He is a co-author of the NY Times and USA Today bestselling book, The Wallet Allocation Rule, and the Nielsen Bookscan bestselling book, Why Loyalty Matters. He has coauthored numerous academic and trade publications including a Harvard Business School Case Study.