Australasian Marketing Journal 21 (2013) 6674

Contents lists available at SciVerse ScienceDirect

Australasian Marketing Journal

journal homepage: www.elsevier.com/locate/amj

The relationship between electronic word-of-mouth motivations and message characteristics: The senders perspective
Kenneth B. Yap , Budi Soetarto, Jillian C. Sweeney
UWA Business School, The University of Western Australia, Australia

a r t i c l e

i n f o

a b s t r a c t
This study investigates how particular motivations are associated with different eWOM message characteristics. This is examined from the senders perspective in both positive and negative eWOM contexts. Responses from a sample of 201 consumers who had posted an online message about a nancial service in the last 12 months were collected through an online survey. Results showed that cognitive and affective characteristics of messages were linked to different motivations to engage in eWOM, which further differed across positive and negative messages. Managers should encourage consumers to share more positive factual information and sort online reviews based on the subject matter, rather than just the positivity of a message. 2012 Australian and New Zealand Marketing Academy. Published by Elsevier Ltd. All rights reserved.

Article history: Received 30 September 2011 Revised 15 August 2012 Accepted 12 September 2012 Available online 9 November 2012 Keywords: Word-of-mouth Cognitive Affective Motivation Financial services Electronic word-of-mouth

1. Introduction It is well-documented that word-of-mouth (WOM) can inuence consumers decisions (e.g. Day, 1971; Harrison-Walker, 2001; East et al., 2008). The persuasiveness of a WOM message may depend on, among other things, the way a sender words a message through logical and emotional appeals or characteristics (Mazzarol et al., 2007; Sweeney et al., 2012). However, what remains unknown is what drives the sender to design their message with such characteristics. The present study attempted to address this research gap in an online environment. The task of identifying and assessing WOM content has in the past been challenging as WOM has often been privately communicated, such that managers are neither privy to what is being said, nor how it is being said. As WOM communication is becoming increasingly transparent in online discussion forums, social networking sites, consumer review sites and blogs (Riegner, 2007), it is now possible to identify and examine individual electronic word-of-mouth (eWOM) messages and gain richer insight into how customers feelings and experiences about a service are represented to others. The present study made use of this new source of WOM.

Much of the research on WOM examines the receivers perspective and little has addressed the generation of WOM (HarrisonWalker, 2001). To advance our knowledge in this area, a suitable point of departure is the investigation of WOM from the senders perspective, in particular their motivation to initiate WOM communication and the characteristics of the WOM message. Specically, the purpose of this study is to examine the relationship between motivations and message characteristics across both positive and negative eWOM. 2. Literature review 2.1. The power of word-of-mouth (WOM) Westbrook (1987, p. 261) dened WOM as informal communication directed toward other consumers about the ownership, usage, or characteristics of particular goods and services and/or their sellers. Researchers have found WOM is more effective than advertising and promotional activities in inuencing consumer decision making, including changing attitudes (e.g. Katz and Lazarsfeld, 1955; Day, 1971) or increased patronage (e.g. Arndt, 1967; Holmes and Lett, 1977). However, little is known about the characteristics that make a persuasive WOM message as researchers have, almost without exception, measured WOM in terms of its frequency and the number of people who receive it (e.g. Westbrook, 1987; Bowman and Naryandas, 2001). Such an aggregated approach to measuring WOM, in that individual messages are often not examined for its content and wording, has been

Corresponding author. Address: UWA Business School (M263), The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia. Tel.: +61 864885876; fax: +61 964881004. E-mail addresses: kyap@biz.uwa.edu.au (K.B. Yap), bsoetarto@gmail.com (B. Soetarto), jsweeney@biz.uwa.edu.au (J.C. Sweeney).

1441-3582/$ - see front matter 2012 Australian and New Zealand Marketing Academy. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ausmj.2012.09.001

K.B. Yap et al. / Australasian Marketing Journal 21 (2013) 6674

67

helpful in studying overall WOM activity, but limits our understanding of the richness and subtleties in individual WOM messages (Mazzarol et al., 2007). Consequently, a more disaggregated measure of WOM in which the individual WOM message is the unit of analysis, was used in the present study. 2.2. The cognitive and affective characteristics of WOM messages Anderson (1998) recognised the importance of examining WOM content and suggested positive WOM can vary in its vividness, pleasantness and novelty, as well as the extent to which it conveys a senders experiences. A number of authors have emphasised the importance of the message characteristics on the message persuasiveness, including both rational and emotional aspects (Allsop et al., 2007; Mason and Davis, 2007). Semin (2000, p.597) stated: . . . talk does not simply involve producing words. It requires choosing words from a lexicon to create sentences that are also linguistically structured. . . with the purpose of communicating an intention to someone else. It is a production with a social end. A number of authors support Mason and Davis (2007, p.505) assertion regarding communication . . . it is more than words; it really is how you say it. For example, the importance of words, content, body language, and expressiveness has been emphasised in WOM messages (Dichter, 1966; East et al., 2008). In this study we specifically explore the cognitive and affective characteristics of eWOM since these are core communication dimensions (e.g. Allsop et al., 2007; Mason and Davis, 2007; Sweeney et al., 2012). In this study, we dene cognitive characteristics as the rational component of a message that typically refers to product attributes including performance, response to problems, and price-value perceptions (Sweeney et al., 2012). Affective characteristics refer to the messages depth, intensity and vividness and reect the language used and the degree of storytelling or depth of information involved in the message (Mazzarol et al., 2007; Sweeney et al., 2012). eWOM messages can be described in terms of cognitive and/or affective characteristics. Examples of this in the nancial services context are shown in Table 1. The link between cognitive and affective WOM message characteristics and subsequent attitudes and behaviour of the receiver has been demonstrated in several studies (e.g. Karmarkar and Tormala, 2010; Sweeney et al., 2012). For example, receivers have higher service and value expectations following the receipt of a message high in cognitive content in particular, as well as affective content (Sweeney et al., 2012). Recipients of eWOM are also more likely to be persuaded in the case of higher informational quality and message clarity (Karmarkar and Tormala, 2010). Such research provides the impetus to investigate what generates cognitive and

affective characteristics, which is a research gap the present study aims to address. 2.3. Motivation to engage in electronic WOM (eWOM) Past studies have suggested a range of motives for engaging in WOM (e.g., Dichter, 1966; Sundaram et al., 1998). Hennig-Thurau et al. (2004) adapted Dichter (1966) and Sundaram et al.s (1998) motivations to an online context, proposed (and empiricallytested) several reasons why a consumer might engage in eWOM communication, six of which are of interest in the present study, namely: 1. Positive self-enhancement reects a consumers need to share their consumption experience to augment their own image as intelligent shoppers. 2. Social benets occur when a consumer transmits a WOM message for identication and social integration purposes. 3. Advice seeking concerns the need to acquire tips and support from others to better understand and use a product or service. 4. Concern for other consumers relates to genuine offers to help other consumers make better purchase decisions. 5. Helping the company relates to a consumers desire to help a company as a result of a particularly pleasing consumption experience. 6. Venting negative feelings relates to a dissatisfying consumption experience that results in the consumer wanting to release frustration and anxiety through negative WOM. 2.4. Link between motivations and message characteristics Research on WOM communication (Wetzer et al., 2007), written communication (Karmarkar and Tormala, 2010; Schellekens et al., 2010) and messages posted online (Schau and Gilly, 2003; Kozinets et al., 2010) suggests that communication motives may be linked to the wording of a message. Wetzer et al. (2007) argued that the specic goal that a consumer strives for in communication may be reected in the content of the message. For example a consumer will talk differently if they want to take revenge from when they seek social connections, and indeed this difference is largely due to varied underlying emotions (e.g. anger and disappointment). This is neither to say that consumers are aware that their communication goals are facilitated by the way they use the language, nor that they are manipulative in the way they talk. However, we do argue that consumers use language as a means to structure and represent reality in a particular way, in order to inuence the belief processes of the recipient (Semin, 2000). Often

Table 1 Examples of cognitive and affective characteristics of eWOM messages. Valence Positive Cognitive Not only was the promotional interest rate attractive, having assessed other nancial products, the comparative rate for this product was extremely competitive. Affective Believe me, I cannot recommend this bank enough. I had soooo many pathetic experiences with other banks; it makes this bank really stand out. This is what ALL banks should be like. Cognitive and affective You are not wrong about this investment company. I think the fund managers there are geniuses and make others look pretty bad. There is a wide selection of investment options for different risk proles or investment objectives. I recommend this company without any hesitation Let me tell you: this company is always one of the rst ones to raise their interest rates and has had at least four interest rate rises above the one by the Reserve Bank. It is so disappointing dealing with them because the advertising portrays them as community-friendly. I denitely would encourage you to do your research. . .this bank may not be as friendly as you think.

Negative

Do not deal with this bank. . .it has too many fees. It has fees for application, maintenance, and discharging the loan. Every direct debit transaction incurred a fee

I feel ripped off! The bastards raised their interest rates faster that you can say BAM!! I hate most banks, but this one is the worst of the lot! If it were not for helping my friends business, I would not give them a single cent!

68

K.B. Yap et al. / Australasian Marketing Journal 21 (2013) 6674

individuals colour their messages through their own opinions and subconscious motivations whether in the context of face-to-face or online communication (e.g. Douglas and Sutton, 2003; Mason and Davis, 2007). While researchers have investigated how motivational goals affect language abstraction (e.g. Douglas and Sutton, 2003) and how different motivations impact communication behaviours (e.g. Hennig-Thurau et al., 2004), no research to our knowledge has investigated how an individual with a particular motive might word an eWOM message in terms of cognitive and affective characteristics. Thus, the research purpose is to examine the relationship between motivations to engage in eWOM communication and the cognitive and affective characteristics of such a message.

may also be in keeping with an impression of sophistication, suggesting: The greater the positive self-enhancement motivation, the greater the: (H1a) cognitive and (H1b) affective characteristics of the eWOM message.

3.2. Social benets Hennig-Thurau et al. (2004) associated this motivation with a persons need for identication and social integration in a community. This motive can manifest in different ways. For example, a consumer may initiate eWOM communication to signify his/her presence in the community. Alternatively, a consumer may post a message to appease gatekeepers who decide who gets accepted or ignored in the community (McWilliam, 2000). Since concerns about presentation are similar to those of positive selfenhancement, we anticipated that the more an eWOM message was used to gain social acceptance, the greater would be its cognitive and affective characteristics. Beuchot and Bullen (2005) found that someone aiming to make interpersonal connections is likely to disclose details in their online communication, while Luminet et al. (2004) found people share an emotional negative experience by explaining the circumstances of the negative event and describing their own reactions to the event, suggesting: The greater the social benets motivation, the greater the: (H2a) cognitive and (H2b) affective characteristics of the eWOM message.

3. Conceptual development The proposed conceptual model suggests a senders motivation to engage in eWOM communication inuences the degree of cognitive and affective characteristics of the message (Fig. 1, Part 2). We use six of Hennig-Thurau et al.s (2004) eWOM motivations to examine the link between motivations and cognitive and affective message characteristics.

3.1. Positive self-enhancement A desired outcome commonly associated with positive selfenhancement is to present oneself as an intelligent or discerning shopper (Sundaram et al., 1998; Hennig-Thurau et al., 2004). It is reasonable to suggest someone with this motive would craft a message containing information that may signal connoisseurship (Hennig-Thurau et al., 2004) and intimate a rst-hand account from an expert point-of-view (Schindler and Bickart, 2005). We also anticipated that such a message would be worded in emphatic and persuasive language. Past studies found people who wish to improve their credibility as experts tend to use words that express certainty and condence in judgement (very, sure, denitely, and condently) (e.g. McEwen and Greenberg, 1970; Karmarkar and Tormala, 2010). A well-articulated or vividly-worded message

3.3. Advice seeking Consumers motivated by advice seeking seek to maximise personal utility by prompting advice or information from others to better understand and use a product or service (Sundaram et al., 1998; Hennig-Thurau et al., 2004). We expected someone who initiates eWOM communication to seek advice is likely to offer significant detail in their message. A person with this motive initiates eWOM communication either by giving their current views on the service or explaining their predicament, both of which imply the message is likely to contain product-specic information to contextualise their request for help. In addition, it is likely the advice-seeking message will contain emotive language as a means of relating to others and gaining empathy before asking for help (Luminet et al., 2004; Wetzer et al., 2007). Wetzer et al. (2007) suggested expressions of regret, disappointment, and uncertainty may be evident in negative WOM communication that is designed to solicit advice, suggesting: The greater the advice seeking motivation, the greater the: (H3a) cognitive and (H3b) affective characteristics of the eWOM message. Hypotheses 13 were tested on positively-valenced and negatively-valenced eWOM messages separately. 3.4. Concern for other consumers The motivation Hennig-Thurau et al. (2004) labelled concern for other consumers implies an element of altruism, which is a voluntary act beneting others without expecting something in return (Piliavin and Charng, 1990; Sundaram et al., 1998). Studies have suggested altruists recognise the best way to be helpful in WOM communication is to be informative and functional (Sen and Lerman, 2007; Bronner and de Hoog, 2011). This suggests a

Fig. 1. Conceptual framework.

K.B. Yap et al. / Australasian Marketing Journal 21 (2013) 6674

69

message resulting from a concern for other consumers will have greater cognitive characteristics. We also suggest the more a person is concerned for others, the greater will be the affective characteristics of the eWOM message. Jeffries (1998) found altruistic behaviour tends to be more intense when the altruist perceives him/herself to be a defender of justice and may articulate a passionately-worded message to represent the cause. It may also be that the tone and language used has to be sufciently moving so the altruists deeds are not in vain. However, prior to stipulating our hypotheses with respect to concern for others, some revision of this motivation is needed. Hennig-Thurau et al. (2004) conceptualised concern for others in terms of both helping and warning other consumers. Consumers with a favourable service experience display concern by helping others make the right decision. Conversely, consumers with an unfavourable service experience are concerned about warning other consumers about a service provider. Since Herche and Engelland (1996) cautioned that a construct is not likely to be unidimensional if it has both positively- and negatively-worded items, this motivation was treated as two constructs (helping other consumers and warning other consumers), suggesting: For positive eWOM messages, the greater the helping other consumers motivation, the greater the: (H4a) cognitive and (H4b) affective characteristics of the eWOM message.1 For negative eWOM messages, the greater the warning other consumers motivation, the greater the: (H4c) cognitive and (H4d) affective characteristics of the eWOM message. As WOM is sought for its credibility and independence from marketer interests, it could be that someone who initiates eWOM communication with altruistic motives wants to be helpful through some level of persuasion, without discrediting their intentions by appearing to be emotionally invested in the service provider (i.e. playing it cool). The aim here is to avoid coloring the message with emotion or vivid language, in an effort to appear particularly objective (Pollach, 2007). Kozinets et al. (2010) observed that some people who go online to evaluate a product may emphasise product attributes to demonstrate their trustworthiness, suggesting:(H4e) For positive eWOM, the greater the helping other consumers motivation, the greater the cognitive characteristics of the message compared to its affective characteristics. (H4f) For negative eWOM, the greater the warning other consumers motivation, the greater the cognitive characteristics of the message compared to its affective characteristics of the message. 3.5. Helping the company The motivation to help a company comes from a positive consumption experience and the individuals goal is to reward the company by referring it to others. In doing so, the sender is likely to recap details of their experience and provide sufcient factual information to substantiate the recommendation. Isen et al. (1985) and Mano (1997) found consumers who are happy tend to increase their cognitive deliberation and thoroughness, which in turn, suggests their eWOM message has greater cognitive characteristics. Such an eWOM message is also likely to be worded emotively to encapsulate the consumers post-consumption reactions of delight and pleasure (Mano, 1997; Schellekens et al., 2010). In an attempt to assist the company in a meaningful way, the sender is likely to convey a strong sense of conviction, suggesting:

For positive eWOM messages, the greater the helping the company motivation, the greater the: (H5a) cognitive and (H5b) affective characteristics of the eWOM message.

3.6. Venting negative feelings Conversely, people who are unhappy with a consumption experience may use negative eWOM to convince others to boycott the offending organisation, intending to seek vengeance and punish the organisation (Sundaram et al., 1998; Ward and Ostrom, 2006). To gain compliance from others, the communicator is likely to include clear descriptions or detailed examples, thus enhancing the logical appeal of their argument (Kowalski, 1996; Schindler and Bickart, 2005). Such a message is thus likely to have signicant cognitive content. There is also evidence to suggest someone looking to punish an organisation or vent negative feelings is likely to word their eWOM message with strong emotional language. Wetzer et al. (2007) and McColl-Kennedy et al. (2009) found negative WOM messages with venting and vengeance motives are likely to contain expressions of anger, frustration, and irritation, suggesting: For negative eWOM messages, the greater the venting negative feelings motivation, the greater the: (H6a) cognitive and (H6b) affective characteristics of the eWOM message. The highly-emotive nature of venting negative feelings suggests an additional hypothesis. This motivation is likely to manifest in an eWOM message that has greater affective characteristics than cognitive characteristics. When enraged, a consumers need for revenge and desire for catharsis is strong and better served by an emotional message displaying their feelings, rather than a message that is logical in appeal (Wetzer et al., 2007; McColl-Kennedy et al., 2009), suggesting: H6c) For negative eWOM, the greater the venting negative feelings motivation, the greater the affective characteristics of the message compared to its cognitive characteristics. 4. Research methodology These hypotheses were investigated within a nancial services context. Allsop et al. (2007) found WOM occurs frequently in the nancial services area, with 70% of respondents seeking (and 44% providing) information and advice on nancial services. Further, WOM is also particularly effective in credence services such as the study context, where quality is difcult to evaluate, even after the service encounter (Zeithaml, 1981). The data used to test the hypotheses were collected through an online survey using a national online consumer panel service. Online consumer panels have been commonly used for market research since the emergence of the Internet (Poynter, 2006; Macdonald and Uncles, 2007). A questionnaire was developed to measure the constructs of interest and tested for face validity with three academic marketing experts and subsequently pre-tested on a sample of undergraduate students. Subsequently, an email invitation was sent to 7,654 online panel members across Australia, 2293 of whom (30%) responded to the request. The respondents were asked if they had posted an eWOM message in an online forum. In all, 304 who had made such a post were invited to continue the survey. Those who either made the post over a year ago or those who could not recall the content of the post were ltered out of the sample. The nal sample size was 201. Respondents were asked to recall their most recent online posting about a nancial service and report details of the message concerning content and delivery aspects, as well as valence (whether

1 In the case of H1-H3 the relative weight of each motivation on cognitive and affective characteristics was also examined. However, we did not nd supporting literature to form specic hypotheses on relative weights.

70

K.B. Yap et al. / Australasian Marketing Journal 21 (2013) 6674

the message was more positive or negative). Similar recall approaches have been used by Christophe and Rime (1997) in a study of social sharing of emotions, Bogomolova et al. (2009) in a study of service quality ratings, and Sweeney et al. (2012) in a study of WOM message characteristics. Further, East and Uncles (2008) have argued retrospective surveys are a useful way of studying WOM. The nal sample comprised almost equal numbers of males and females; while 22% were under 25 years, 42% were aged 2544 and 36% were aged 45 years and older. Half used the internet daily, while the remainder used it at least once a week. Median internet usage was 15 h a week. Over 60% of posts had been made in the last three months and 80% in the last six months, which is considerably more recent than the interval between incidents and survey reporting described by Christophe and Rime (1997). The motivation to post the message was measured by adapting Hennig-Thurau et al.s (2004) original motivation measures of eWOM, while the cognitive and affective characteristics of the eWOM message were measured using the WOM content scales developed by Sweeney et al. (2012).2 All of the items, which can be seen in Table 2 below, were measured on ve-point Likert scales (1 = strongly disagree, 5 = strongly agree). Further, we asked respondents about the message valence on a 5-point scale (1 = positive, 5 = negative).

strength of delivery is likely to be communicated through rich language using the written word, rather than through intonation, body language, verbal and visual cues that are part of face-to-face communication. Based on Fornell and Larckers (1981) criterion, discriminant validity was established between the cognitive and affective dimensions as the minimum AVE was 0.53 (affective characteristics) and the squared-correlation between the two constructs was 0.30. The measurement properties of each construct in the conceptual model were examined through one-factor congeneric models (Fornell and Larcker, 1981; Anderson and Gerbing, 1988). Goodness-of-Fit indices, item loadings, construct reliability, and average variance extracted scores can be seen in Table 2. The loadings for each variable on its respective construct was statistically signicant and construct reliabilities were all above the suggested 0.70 minimum level (Hair et al., 2006). Convergent validity was demonstrated as the AVE for each construct exceeded 0.50 (Fornell and Larcker, 1981). Discriminant validity between all of the constructs pairs was also examined through Fornell and Larckers (1981) test. As the minimum AVE was 0.61 and the maximum squared-correlation between constructs was 0.48 (Personal self-enhancement and Helping the company), discriminant validity was supported.3 5.2. Hypotheses Testing The sample was split into positive and negative eWOM subsamples, according to the valence of the message,4 resulting in a sample of 92 positive messages (all or mostly positive) and 109 negative messages (all or mostly negative, and mixed positive and negative). For subsequent analyses, each construct was operationalised through the average value of all the corresponding items listed in Table 2 (Rodrguez-Pinto et al., 2007). Correlation and multiple regression analysis were used in both sub-samples to test the hypotheses relating to relationships between motivations, message characteristics, and consumer outcomes. Steigers Z-test was used to examine the relative strength of the relationship between cognitive and affective message characteristics and other constructs as hypothesised in H4e, H4f and H6c. Regression analyses were conducted, in which each of the two message characteristics were regressed on motivations to engage in eWOM (see Fig. 1). The analysis was also investigated through correlational analyses. For positive eWOM messages, the motivations of positive selfenhancement (rcognitive = 0.44, raffective = 0.38), social benets (0.39, 0.37), advice seeking (0.34, 0.25) and helping the company (0.54, 0.38) were signicantly associated with cognitive and affective message characteristics (all p < 0.01), supporting H1a, H1b, H2a, H2b, H3a, H3b, as well as H5a and H5b (see Table 3). The motivation to help other consumers was associated with cognitive characteristics (0.41, p < 0.01), but not with affective characteristics (0.12, n.s.), thus supporting H4a but not H4b. In the case of positive eWOM messages, the adjusted R2 when regressing cognitive characteristics on the full set of motivations was 0.36, but only 0.16 in the case of affective characteristics. Thus the level of cognitive characteristics was better explained by motivations than the level of affective characteristics. The results of a Steigers Z-test suggested the association of helping other consumers with cognitive characteristics was greater that with affective characteristics (0.41 vs. 0.12, p < 0.01), thus supporting H4e. For negative eWOM messages, the motivations of positive selfenhancement, social benets and advice seeking were not associAvailable from authors on request. We grouped positive and negative eWOM givers as follows: positive included positive or mostly positive eWOM messages, while negative included negative and positive negative or mostly negative eWOM messages.
4 3

5. Findings 5.1. Construct reliability and validity Common method bias was tested using Lindell and Whitneys (2001) approach which involves estimating a proxy for CMV though and re-estimating all correlations between model variables. Since none of the original correlations were signicantly different from their CMV-adjusted values, common method bias was not likely to confound the results (see Appendix A). The psychometric properties of the scales were assessed using exploratory and conrmatory factor analyses. Exploratory factor analyses of the six sets of motivation items replicated the original factor structure with the exception of concern for other consumers. In line with expectations, the concern for other consumers motivation comprised both helping other consumers and warning other consumers representing positive and negative experiences, respectively. These factors had higher reliabilities (Cronbach alphas of 0.68 and 0.81) than the original four item scale (0.57 in the present study and 0.58 in Hennig-Thurau et al.s 2004 study). To demonstrate discriminant validity, Fornell and Larcker (1981) recommended that the AVE (Average Variance Extracted) for each construct had to be higher than the squared-correlation between them. Based on this criterion, discriminant validity was established for the two factors (AVEhelping = 0.73, AVEwarn = 0.82, squared-correlation = 0.02). Two factors emerged from an exploratory factor analysis of Sweeney et al.s (2012) scale of WOM message characteristics: a 4-item factor representing the cognitive dimension (construct reliability = 0.90) and an 8-item factor representing the affective dimension (combining content richness and strength of delivery; construct reliability = 0.90). The combined construct rather than separate constructs had face validity in the online setting, as
2 Given the online context, we expected that Sweeney et al.s (2012) content richness and strength of delivery message characteristics derived from face-to-face WOM research would likely represent the same affective dimension. We reasoned that with only one dimension of communication available in the online forum context, which is the written word, and not having access to other cues such as voice inection and body language, that these two characteristics would be indistinguishable.

K.B. Yap et al. / Australasian Marketing Journal 21 (2013) 6674 Table 2 Results of conrmatory factor analysis. Factor Items Goodness-of-t indices (one factor models) GFI = 0.99 CFI = 0.99 Item loading Construct reliability AVE Positive eWOMa Mean 0.84 0.84 0.78 0.64 GFI = 0.92 CFI = 0.82 0.84 0.83 0.77 GFI = 0.92 CFI = 0.88 0.90 0.86 GFI = 0.88 CFI = 0.50 0.90 0.81 GFI = 0.99 CFI = 0.98 GFI = 0.90 CFI = 0.55 0.95 0.87 0.88 0.82 0.90 0.82 N/A N/A 4.04 0.87 0.84 0.73 4.13 0.87 N/A N/A 0.88 0.78 3.75 0.95 3.77 0.92 0.86 0.67 3.74 0.80 3.50 0.85 0.86 0.61 3.78 SD 0.82 Negative eWOMb Mean 3.10 SD 1.09

71

Personal self-enhancement

This way I can express my joy about a good buy I can tell others about a great experience I feel good when I can tell others about my buying success My contributions show others that I am a clever customer I believe a chat among like-minded people is a nice thing It is fun to communicate this way I meet nice people this way I hope to receive advice from others to help solve my problems I expect to receive tips or supports from other users I want to help others with my own positive experiences I want to give others the opportunity to buy the right product I want to warn others of bad products I want to save others from having the same negative experiences as me In my opinion, good companies should be supported I am so satised with a company and its product that I want to help the company to be successful I like to get anger off my chest I want to take vengeance upon the company The company harmed me, and now I will harm the company My contributions help me to shake off frustration about bad buys Specic Clear Informative Reliable Delivered in a strong way Delivered powerfully Delivered in an important manner Intense Delivered using strong words Reinforcing Elaborate Explicit

Social Benets

Advice seeking

Help other consumersc

Warn other consumersd

Helping the companyc

0.84

0.72

4.09

0.87

N/A

N/A

Venting negative feelingd

GFI = 0.87 CFI = 0.82

0.81 0.81 0.79 0.76

0.87

0.63

N/A

N/A

3.16

1.02

Cognitive Characteristics (I believe the message I posted was. . .) Affective characteristics (I believe the message I posted was. . .)

GFI = 0.90 CFI = 0.88

0.87 0.85 0.82 0.79 0.84 0.83 0.80 0.75 0.73 0.70 0.60 0.52

0.90

0.69

4.10

0.66

4.07

0.62

GFI = 0.89 CFI = 0.88

0.90

0.53

3.29

0.67

3.49

0.71

Items are measured on a 5-point Likert scale where 1 = strongly disagree, 5 = strongly agree. N/A denotes not applicable. a N = 92. b N = 109. c Indicates that only positive eWOM is applicable. d Indicates that only negative eWOM is applicable.

ated with either cognitive or affective message characteristics. The exception to this is the association between the motivation of advice seeking and cognitive characteristics 0.34 (p < 0.01) (H3a); thus H1a, H1b, H2a, H2b and H3b were not supported in the case of negative eWOM. Both cognitive and affective message characteristics were positively associated with the motivation to warn other consumers (0.44 and 0.33, respectively, p < 0.01 in both cases) and to vent negative feelings (0.20, p < 0.05 and 0.42, p < 0.01, respectively), thus supporting H4c, H4d, H6a and H6b. In contrast to the positive eWOM case, the adjusted R2 in negative eWOM was similar when regressing cognitive and affective charac-

teristics on motivations (0.20 and 0.18, respectively). Thus the level of both cognitive and affective characteristics was equally explained by motivations in the case of negative eWOM; however the explanation was not as high as for cognitive characteristics in the case of positive eWOM. The hypothesis that a negative eWOM message motivated by wanting to warn other consumers would have greater cognitive characteristics than affective characteristics was also tested through Steigers Z-test. Results were in the expected direction (0.44 vs. 0.33), but the difference was not signicant, hence H4f was not supported. The venting-affective link (0.42) was signicantly greater than the venting-cognitive link

72

K.B. Yap et al. / Australasian Marketing Journal 21 (2013) 6674

Table 3 Correlations between Motivations and Cognitive and Affective Characteristics of an eWOM Message. Positive eWOM Cognitive characteristics Personal Self-enhancement Social Benets Advice Seeking Help Other Consumers Warn Other Consumers Helping the Company Venting Negative Feeling Adj. R2 for regression equation N Note: * p < 0.05. ** p < 0.01. 0.44** 0.39** 0.34** 0.41** N/A 0.54** N/A 0.36 92 (H1a) (H2a) (H3a) (H4a) (H5a) Affective characteristics 0.38** (H1b) 0.37** (H2b) 0.25* (H3b) 0.12 (H4b) N/A 0.38** (H5b) N/A 0.16 92 Negative eWOM Cognitive characteristics 0.15 (H1a) 0.17 (H2a) 0.34** (H3a) N/A 0.44** (H4c) N/A 0.20* (H6a) 0.20 109 Affective characteristics -0.14 (H1b) 0.11 (H2b) 0.10 (H3b) N/A 0.33** (H4d) N/A 0.42** (H6b) 0.18 109 Comparative hypotheses (Steigers Z)

0.41 > 0.12** (H4e) 0.44 > 0.33 (H4f) 0.20 < 0.42** (H6c)

(0.20) (p < 0.01), thus supporting H6c. The results of the correlation and regression analyses are displayed in Table 3, while a summary of hypotheses testing is given in Table 4.

6. Discussion and implications 6.1. Discussion and theoretical implications The present study sought to identify how motivations to engage in eWOM inuence the characteristics of the eWOM message, in particular cognitive and affective message components following the centrality of cognitive and affective elements in communication (e.g., Allsop et al., 2007; Mason and Davis, 2007). The ndings suggest that for positive eWOM messages, personal selfenhancement, social benets, and advice seeking are positively linked to both cognitive and affective characteristics of the message. Efforts to make personal or social gain from eWOM communication seem to evoke clear factual information that is presented in a persuasive and emotive manner. However, these social motives largely do not appear to be associated with the extent of cognitive and affective characteristics in a negative eWOM message. However, as expected, messages motivated by warning others in the negative eWOM case, do generate cognitive and

affective characteristics, particularly the former; although the difference was not signicant. In contrast, venting evoked signicantly stronger affective than cognitive message characteristics. It seems people are more circumspect in the case of warning than venting, placing more emphasis on cognitive content (Bronner and de Hoog, 2011). Helping other consumers was, as expected significantly more linked to cognitive than affective message characteristics in positive eWOM messages, providing additional support for the notion that those altruistically-motivated to share their positive or negative service experience with others are likely to focus on factual content in the message content (Sen and Lerman, 2007; Bronner and de Hoog, 2011).

6.2. Managerial implications The proposed framework emphasises the importance of understanding the links between motives for initiating eWOM communication and the cognitive and affective communication characteristics of eWOM messages (Fig. 1). Managers should be aware that some of these motivations may give rise to eWOM messages with sufcient cognitive and affective characteristics to impact their business. The impact of cognitive and affective elements on receiver expectations of, for example, service quality

Table 4 Summary of Hypotheses Testing. # Hypothesis Result Partial support eWOM) Partial support eWOM) Partial support eWOM) Partial support eWOM) Supported Partial support eWOM) Supported Not supported Supported Supported Supported Not supported Supported Supported Supported Supported Supported (only for positive (only for positive (only for positive (only for positive

Motivations and message characteristics H1a Self-enhancement ? cognitive characteristics H1b H2a H2b H3a H3b H4a H4b H4c H4d H4e H4f H5a H5b H6a H6b H6c Self-enhancement ? affective characteristics Social benets ? cognitive characteristics Social benets ? affective characteristics Advice seeking ? greater cognitive characteristics Advice seeking ? affective characteristics For positive eWOM, helping other consumers ? cognitive characteristics For positive eWOM, helping other consumers ? affective characteristics For negative eWOM, warning other consumers ? cognitive characteristics For negative eWOM, warning other consumers ? greater affective characteristics For positive eWOM, helping other consumers has a greater effect on cognitive characteristics than affective For negative eWOM, warning other consumers has a greater on cognitive characteristics than affective For positive eWOM, helping the company ? cognitive characteristics For positive eWOM, helping the company ? affective characteristics For negative eWOM, venting negative feelings ? cognitive characteristics For negative eWOM, venting negative feelings ? affective characteristics For negative eWOM, positive effect of motivation of venting negative feelings has a greater effect on cognitive characteristics than affective

(only for positive

K.B. Yap et al. / Australasian Marketing Journal 21 (2013) 6674

73

and value was established in Sweeney et al.s (2012) study. Therefore, the framework in the present study should help managers to better understand the nature of eWOM communication and assess the extent to which these motivations inuence message characteristics and ultimately their organisation. The rapid rise and transparency of social networking sites offers researchers and managers a signicant opportunity to track eWOM concerning their organisation. A discussion forum or social networking page initiated by the organisation should enable managers to not only access eWOM communication but also to promote social interaction among consumers in the hopes of creating socially-driven motivations to initiate eWOM. These sites should enable consumers to interact with each other socially, ask questions, provide tips and buying advice on a product or service and write product reviews. The ndings re-afrm the value of sites designed for this purpose. At the very least, managers may use online sentiment monitoring tools (e.g. Buzz Monitoring) and assess the cognitive and affective elements of discussions surrounding their brands. Our expectation is that as people become part of online communities, particularly with the increase in social networking; group norms and personal social agendas as reected in our motivation set will play a greater role in determining how eWOM messages are worded (e.g. Trusov et al., 2009; Kozinets et al., 2010; Higgins, 2011). As a result, it is increasingly important for managers of social networking campaigns to understand the social norms of communication in each online community and support a culture of open and constructive feedback. In addition, given the greater impact of cognitive message elements on the receivers perceptions on the organisation (Sweeney et al., 2012), we recommend managers sort reviews by its factual content or subject matter to improve the value and userfriendliness of the site. An example of this is tripadvisor.com where hotel reviews are indexed by factual topics (e.g. room service, great view, subway station, etc.), instead of the valence of reviews.

ten. Sun et al. (2006) argued written communication offers better logical order than oral communication. As online communities facilitate a greater ow of information (Brown et al., 2007), the relative effectiveness of eWOM messages warrants more attention in future research. Appendix A Lindell and Whitney (2001) argue that in a typical survey in which the same rater responded to the items in a single survey form at the same point in time, the collected data may be susceptible to CMV. They argue that instead of a marker variable approach to CMV that the second-smallest positive correlation among the manifest variables can be used as proxy for common method variance (CMV), which in this study was 0.009. A method factor is assumed to have a constant correlation with all of the measured items. Under this assumption, a CMV-adjusted correlation between the variables under investigation, rA, can be computed by partialling rM from the uncorrected correlation, rU. The rA coefcient can be calculated as:

rA rU r M =1 rM :
Therefore, in the present study, rA = (rU 0.009)/(10.009). A Zvalue was calculated to assess if a signicant difference existed between the two correlation coefcients (i.e. rU and rA). References
Albers-Miller, N.D., Stafford, M.R., 1999. International services advertising: an examination of variation in appeal use for experiential and utilitarian services. The Journal of Services Marketing 13 (4/5), 390406. Allsop, D.T., Bassett, B.R., Hoskins, J.A., 2007. Word-of-mouth research: principles and applications. Journal of Advertising Research 17 (4), 398411. Anderson, E.W., 1998. Customer satisfaction and word of mouth. Journal of Service Research 1 (1), 517. Anderson, J.C., Gerbing, D.W., 1988. Structural equation modeling in practice: a review and recommended two-step approach. Psychological Bulletin 103 (3), 411423. Arndt, J., 1967. Role of product-related conversations in the diffusions of a new product. Journal of Marketing Research 4 (3), 291295. Beuchot, A., Bullen, M., 2005. Interaction and interpersonality in online discussion forums. Distance Education 26 (1), 6787. Bowman, D., Naryandas, D., 2001. Managing customer-initiated contacts with manufacturers: the impact on share of category requirements and word-ofmouth behaviour. Journal of Marketing Research 38 (3), 281297. Bogomolova, S., Romaniuk, J., Sharp, A., 2009. Quantifying the extent of temporal decay in service quality ratings. International Journal of Market Research 51 (1), 7191. Bronner, F., de Hoog, R., 2011. Vacationers and eWOM: who posts, and why, where, and what? Journal of Travel Research 50 (1), 1526. Brown, J., Broderick, A.J., Lee, N., 2007. Word of mouth communication within online communities: conceptualizing the online social network. Journal of Interactive Marketing 21 (3), 220. Christophe, V., Rime, B., 1997. Exposure to the social sharing of emotion: emotional impact, listener responses and secondary sharing. European Journal of Social Psychology 27 (1), 3754. Day, G.S., 1971. Attitude change, media and word of mouth. Journal of Advertising Research 11 (6), 3140. Dichter, E., 1966. How word-of-mouth advertising works. Harvard Business Review 44 (6), 147166. Douglas, K.M., Sutton, R.M., 2003. Effects of communication goals and expectancies on language abstraction. Journal of Personality and Social Psychology 84 (4), 682696. East, R., Hammond, K., Lomax, W., 2008. Measuring the impact of positive and negative word of mouth on brand purchase probability. International Journal of Research in Marketing 25 (3), 215224. East, R., Uncles, M.D., 2008. In praise of retrospective surveys. Journal of Marketing Management 24 (910), 929944. Fornell, C., Larcker, D.F., 1981. Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research 18 (1), 3950. Hair, J.F.J., Black, W.C., Babin, B.J., Anderson, R.E., Tatham, R.L., 2006. Multivariate Data Analysis, sixth ed. Prentice Hall, Upper Saddle River, New Jersey. Harrison-Walker, L.J., 2001. The measurement of word-of-mouth communication and investigation of service quality and customer commitment as potential antecedents. Journal of Service Research 4 (1), 6075.

7. Limitations and future research The study has some limitations that suggest opportunities for future research. First, the study was restricted to nancial services and since this is a service where quality can be difcult to evaluate, it is uncertain whether this model can be generalised to other industries in which eWOM activity is also common (Allsop et al., 2007). For instance, Albers-Miller and Staffords (1999) research suggests the affective characteristics of a WOM message would be more inuential in experiential services (e.g. travel and tourism) than in utilitarian services, due to its hedonic nature. Second, the testing of the model in the present study was conned to the perceptions of eWOM senders of their message. Clearly, the perspective investigated here is just one half of the WOM communication dyad and that the other half the perceptions of eWOM receivers of a message with particular cognitive and affective characteristics needs to be evaluated in future research. Since receivers must also be motivated to read an eWOM message, there is an opportunity to match their motivations, perceptions of the message, as well as potential reactions to the message. Research in this area will inform managers of the potency in the wording of an eWOM message. The present study modied Hennig-Thurau et al.s (2004) concern for others motivation by dividing it into helping other consumers (for positive eWOM) and warning other consumers (for negative eWOM). As eWOM can either be positive or negative, it would be useful for future studies to examine eWOM motives that are uniquely positive or negative, as well as valence-neutral motives. Finally, the amount of information consumers receive from a WOM message may depend on whether it is spoken or writ-

74

K.B. Yap et al. / Australasian Marketing Journal 21 (2013) 6674 McWilliam, G., 2000. Building stronger brands through online communities. Sloan Management Review 41 (3), 4354. Piliavin, J.A., Charng, H.W., 1990. Altruism: a review of recent theory and research. Annual Review of Sociology 16 (1990), 2765. Pollach, I., 2007. Electronic world of mouth: a genre analysis of product reviews on consumer opinion web sites. In: Gopalan, S., Taher, N. (Eds.), Viral Marketing: Concepts and Cases. Ifchai University Press, Hyderabad, pp. 117140. Poynter, R., 2006. How Has the Internet Changed Research? Paper presented at the 2006 ESOMAR Congress, 17-20 September, London, United Kingdom. Riegner, C., 2007. Word of mouth on the web: the impact of Web 2.0 on consumer purchase decisions. Journal of Advertising Research 47 (4), 436447. Rodrguez-Pinto, J., Gutirrez-Cilln, J., Rodrguez-Escudero, A.I., 2007. Order and scale of market entry, rm resources, and performance. European Journal of Marketing 41 (5/6), 590607. Schau, H.J., Gilly, M.C., 2003. We are what we post? Self-presentation in personal web space. Journal of Consumer Research 30 (3), 385404. Schellekens, G.A.C., Verlegh, P.W.J., Smidts, A., 2010. Language abstraction in word of mouth. Journal of Consumer Research 37 (2), 207223. Schindler, R.M., Bickart, B., 2005. Published word of mouth: referable, consumergenerated information on the internet. In: Haugtvedt, C.P., Machleit, K.A., Yalch, R.F. (Eds.), Online Consumer Psychology: Understanding and Inuencing Consumer Behaviour in the Virtual World. Lawrence Erlbaum Associates, Hillsdale, New Jersey, pp. 3561. Semin, G.R., 2000. Agenda 2000 communication: language as an implementational device for cognition. European Journal of Social Psychology 30 (5), 595612. Sen, S., Lerman, D., 2007. Why are you telling me this? An examination into negative consumer reviews on the web. Journal of Interactive Marketing 21 (4), 7694. Sun, T., Youn, S., Wu, G., Kuntaraporn, M., 2006. Online word-of-mouth (or mouse): an exploration of its antecedents and consequences. Journal of ComputerMediated Communication 11 (4), 11041127. Sundaram, D.S., Mitra, K., Webster, C., 1998. Word-of-mouth communications: a motivational analysis. Advances in Consumer Research 25 (1), 527531. Sweeney, J.C., Soutar, G.N., Mazzarol, T., 2012. Word of mouth: measuring the power of individual messages. European Journal of Marketing 46 (1/2). Trusov, M., Bucklin, R.E., Koen, P., 2009. Effects of word-of-mouth versus traditional marketing: ndings from an internet social networking site. Journal of Marketing 73 (5), 90102. Ward, J.C., Ostrom, A.L., 2006. Complaining to the masses: the role of protest framing in customer-created complaint web sites. Journal of Consumer Research 33 (2), 220230. Westbrook, R.A., 1987. Product/consumption-based affective responses and postpurchase processes. Journal of Marketing Research 24 (3), 258270. Wetzer, I.M., Zeelenberg, M., Pieters, R., 2007. Never eat in that restaurant, I did!: exploring why people engage in negative word-of-mouth communication. Psychology and Marketing 24 (6), 661680. Zeithaml, V.A., 1981. How consumer evaluation processes differ between goods and services. In: Donnelly, J.H., George, W.R. (Eds.), Marketing of Services. American Marketing Association, Chicago, Illinois, pp. 186190.

Hennig-Thurau, T., Gwinner, K.P., Walsh, G., Gremler, D.D., 2004. Electronic wordof-mouth via consumer opinion platforms: what motivates consumers to articulate themselves on the internet. Journal of Interactive Marketing 18 (1), 3852. Herche, J., Engelland, B., 1996. Reversed-polarity items and scale unidimensionality. Journal of the Academy of Marketing Science 24 (4), 366374. Higgins, K., 2011. Word of Mouth the Key to Sales. The Sunday Mail (Queensland), 5th February, available at http://www.couriermail.com.au/ipad/word-ofmouth-the-key-to-sales/story-fn6ck45n-1226000644162 (Accessed 28 March 2011). Holmes, J.H., Lett, J.D., 1977. Product sampling and word of mouth. Journal of Advertising Research 17 (5), 3540. Isen, A.M., Johnson, M.M.S., Mertz, E., Robinson, G.F., 1985. The inuence of positive affect on the unusualness of word associations. Journal of Personality and Social Psychology 48 (6), 14131426. Jeffries, V., 1998. Virtue and the altruistic personality. Sociological Perspectives 41 (1), 151166. Karmarkar, U.R., Tormala, Z.L., 2010. Believe me, I have no idea what Im talking about: the effects of source certainty on consumer involvement and persuasion. Journal of Consumer Research 36 (6), 10331049. Katz, E., Lazarsfeld, P.F., 1955. Personal Inuence: The Part Played by People in the Flow of Mass Communication. Free Press, Glencoe, Illinois. Kowalski, R.M., 1996. Complaints and complaining: functions, antecedents, and consequences. Psychological Bulletin 119 (2), 179196. Kozinets, R.V., de Valck, K., Wojnicki, A.C., Wilner, S.J.S., 2010. Networked narratives: understanding word-of-mouth marketing in online communities. Journal of Marketing 74 (2), 7189. Lindell, M.K., Whitney, D.J., 2001. Accounting for common method variance in cross-sectional research designs. Journal of Applied Psychology 86 (1), 114121. Luminet, O., Bouts, P., Delie, F., Manstead, A.S.R., Rime, B., 2004. Social sharing of emotion following exposure to a negatively valenced situation. Cognitive and Emotion 14 (5), 661688. Macdonald, E.K., Uncles, M., 2007. Consumer savvy: conceptualisation and measurement. Journal of Marketing Management 23 (5/6), 497517. Mano, H., 1997. Affect and persuasion: the inuence of pleasantness and arousal on attitude formation and message elaboration. Psychology and Marketing 14 (4), 315335. Mason, P.R., Davis, B.H., 2007. More than the words: using stance-shift analysis to identify crucial opinions and attitudes in online focus groups. Journal of Advertising Research 47 (4), 496506. Mazzarol, T., Sweeney, J.C., Soutar, G.N., 2007. Conceptualizing word-of-mouth activity, triggers and conditions: an exploratory study. European Journal of Marketing 41 (11/12), 14751494. McColl-Kennedy, J.R., Patterson, P.G., Smith, A.K., Brady, M.K., 2009. Customer rage episodes: emotions, expressions and behaviours. Journal of Retailing 85 (2), 222237. McEwen, W.J., Greenberg, B.S., 1970. The effects of message intensity on receiver evaluations of source, message and topic. The Journal of Communication 20 (4), 340350.