Do Online Gossipers Promote Brands?

Abstract

Online gossip has been recognized as small talk on social networking sites (SNSs) that influences consumer behavior, but little attention has been paid to its role. This study makes three theoretical predictions: (a) propensity to gossip online leads to greater information value, entertainment value, and friendship value; (b) upon exposure to a high-involvement product, online gossipers are more willing to spread such information through electronic word-of-mouth (eWOM) in search of prestige or fame as a knowledge expert; and (c) this tendency will be more pronounced when they are connected with strong ties (rather than weak ties) and belong to a large network (rather than a small network). An experimental survey was conducted with a scenario method. In total, 818 general consumers participated in the survey. A multivariate analysis of variance (ANOVA) provides empirical support for prediction (1). With regard to predictions (2) and (3), a series of three-way and two-way between-subjective ANOVAs were performed. When a high-involvement product is promoted, gossipers, rather than nongossipers, are more willing to participate in eWOM on an SNS. Furthermore, a significant interaction effect indicates that online gossipers' willingness to particiapte in eWOM would be more pronounced if they belonged to a large network rather than a small network. However, when a low-involvement product is promoted, no interaction effect is found between online gossip and network size. In closing, theoretical and managerial implications are discussed, while important limitations are recognized.

Introduction

Results of long-term studies from different regions of the world confirm that the content and frequency of gossip are universal—we devote much of our daily conversation to gossip.¹ Gossip may involve putting someone else down to feel better by comparison, or may simply be a way to connect with someone else and share insecurities. The end result is often a healthy relief of social and professional anxiety.¹ In an increasingly digital era, gossipers are moving from offline to online. Social networking sites (SNSs) are an ideal platform to feed people's curiosity, including possession or consumption of famous brands. For example, discussing who owns or why to purchase specific brands can draw attention from SNS members who may express their opinions about a product or service.²

This leads to our fundamental research question: Do online gossipers promote brands? We seek to answer this by addressing the impact of gossip on electronic word-of-mouth (eWOM). Typically, online gossipers show off their information or knowledge on SNSs, because they want to be recognized as unique or authoritative. Prior research indicates that gossip is an efficient way to remind group members of the importance of the group's norms and values.³ Gossip may play an even more interesting role in an SNS, given the complexity of members' online networking behavior. However, the existing literature offers scant evidence in this area. The purpose of this research is to fill this gap.

Theoretical Framework

Online gossip and eWOM

Gossip is small talk about a mutual friend or acquaintance that has curious, scandalous, novel, humorous, and unexpected elements.⁴ In the literature of evolutionary psychology, gossip has been conceptualized as one of the verbal (or conversational) strategies people use to influence others in some way, and there are probably as many verbal strategies as there are social interactions.⁵ In this light, this study regards online gossip as an attention-grabbing, cybercommunication strategy about social and personal topics that people use to influence others on the Internet.

Gossiping about brands somehow resembles celebrity gossip, which usually grabs quick attention, because it engages the listeners with scandalous, hilarious, outrageous, or horrific content.⁴ Even to someone with no interest in the lives of celebrities, telling a funny and interesting story about a big name can always get people to pay attention. In this sense, celebrity gossip may be especially relevant to marketing, since many brands employ celebrity endorsement by casting famous actors, actresses, singers, or athletes as spokespersons or brand characters.⁶

In contrast to gossip, WOM can be defined as informal conversation by which opinions on products and brands are developed, expressed, and spread.⁷ WOM has been extended to the computer-mediated environment, hence, the term eWOM.⁸ In this study, our primary interest lies in the question as to whether gossiping propensity (i.e., general verbal strategy) has significant effects on eWOM behavior (i.e., brand or product referral).

Social functions of online gossip

Gossip has been recognized as small talk with a social purpose. Rosnow⁹ defines gossiping as “an instrumental transaction in which A and B trade small talk about C for something in return” (p. 158). Such information trading could provide status, fun, money, social control, or any material or psychological stimulus capable of fulfilling preconditioned needs, wishes, and expectations. Gossip, therefore, serves as a strong driver for establishing friendships,¹⁰ exchanging knowledge,¹¹ and providing mutual entertainment.¹² Prior research suggests that gossip can remind group members of the importance of the group's norms and values.³ Social exchange theory explains four major social functions of gossip—information, entertainment, friendship, and influence.¹³ In other words, gossip facilitates information exchange, provides recreational pastime, and brings groups together through the sharing of norms, thereby creating groups' influence faction. This leads to the following hypothesis:

H1: Propensity to gossip online influences online social functions such that these functions are more consolidated among members with high propensity to gossip, compared with members with low gossiping propensity.

Outcome of online gossip

After consolidating online social functions—information value, entertainment value, friendship value, and normative pressure—on an SNS, members are more likely to engage in gossiping behavior. Our primary interest lies in a relationshp between online gossip and eWOM. Online gossipers are more likely to transmit mutually benefitial information to peers and friends, while eWOM is related more to brand or product information. Brown et al.⁸ found that online community members recognize the social value of product-related gossip, since knowledge of such unofficial information makes them feel like an expert. They demonstrate their knowledge and insider status by reporting it to other network members. Brown et al.⁸ pointed out that long-term relationships established with particular online communities play an especially important role, because members need to feel part of a scene. This was particularly the case, where they had a higher degree of involvement in a product category. Similar findings have been reported in Algesheimer et al.¹⁴ who found that the strength of consumers' relationship with a brand community determines the level of community engagement in “helping other members, participating in joint activities, and otherwise acting volitionally in ways that the community endorses and that enhance its value for themselves and others” (p. 21). Furthermore, along with evidence on tie strength, prior research indicates that the online community size affects the strength of value perceptions. Dholakia et al.¹⁵ found that social influence variables, group norms, and social identity are perceived as value perceptions to a much greater extent among individuals belonging to small group-based virtual communities, compared with those belonging to large network-based virtual communities. Specifically, their findings indicate that members of large network-based communities seek a reputation to establish trust and status and to foster social interactions, since they usually don't know each other initially, and their motives are self-referent. This reputation-seeking behavior seems consistent with propensity to gossip, as gossip is viewed as an agent of struggle for status and prestige among individuals and cliques.¹⁶ Gossip is often used for social comparison between in-group and out-group, in favor of the former by providing knowledge that would strengthen the preference for feeling positive about one's in-group.¹⁷ In contrast, small-group communities see little need for such struggle or social comparison, and are less prone to engage in gossip. For a high-involvement product, we therefore propose this main effect:

H2a: When the effects of tie strength and network size are simultaneously considered, gossipers, rather than nongossipers, would be more willing to participate in eWOM on an SNS.

Concerning the interaction between propensity to gossip, tie strength, and network size, we hypothesize as below:

H2b: Gossipers' willingess to participate in eWOM would be more pronounced if they were connected with strong ties rather than weak ties.

H2c: Gossipers' willingness to particiapte in eWOM would be more pronounced if they belonged to a large network rather than a small network.

In contrast, gossipers may be less motivated when the degree of involvement in a product category is low. Because the product is less important, they would not feel like an expert, and reporting this gossip would draw little attention from other network members. In this case, tie strength and network size are the key determinants of eWOM intention. Small-group members may pay more attention because they are likely to be close friends, and therefore connected with strong ties, due to more intimate communication. As a result, for a low-involvement product, we postulate the following hypotheses:

H3a: When the effects of tie strength and network size are simultaneously considered, there is no significant difference between gossipers and nongossipers in terms of eWOM intention on an SNS.

H3b: Regardless of propensity to gossip, members connected with strong ties, rather than weak ties, would be more willing to participate in eWOM on an SNS.

H3c: Regardless of propensity to gossip, members belonging to a small network rather than a large network would be more willing to participate in eWOM on an SNS.

Methods

Stimulus materials

Pretests conducted with 69 business-major university students served two purposes: (a) to assess whether the desired level of product involvement was induced by the sales promotion, and (b) to assess what type of sales promotion could induce realistic and actionable interests. All subjects used one or more SNS. We asked what kind of sales promotion they may refer to friends on an SNS. Discount information was the most frequent response. Next, we asked the subjects to rate five products (beer, sports shoes, mobile phone, laptop computer, and travel) using a product-involvement scale suggested by Mittal.¹⁸ Analysis of variance (ANOVA) suggests significant differences among the products, with beer the lowest and laptop computer the highest.

Measures

Information value, entertainment value, friendship value were measured by the purposive value (three items), entertainment value (four items), and interpersonal connectivity (three items) scales, respectively, suggested by Dholakia et al.¹⁵ A three-item scale was adapted from Verhoef et al.¹⁹ to measure eWOM intention. Propensity to gossip was measured by a six-item social value subscale of the attitude toward the gossip scale developed by Litman and Pezzo.²⁰ Tie strength was measured by a four-item scale adapted from Mittal et al.²¹ For these constructs, seven-point scales anchored by strongly disagree/strongly agree were used. For each construct, mean values were calculated and used for the statistical analysis. The network size was measured by the number of friends the respondent has registered and regularly communicates with on the SNS. The additional scales used in the study are shown in Appendix.

Procedure

In total, 818 general consumers participated in our study. Approximately, 76 percent of the respondents have been using an SNS for more than 3 years. A Web-based survey site was created and the respondents were sent a survey invitation by an e-mail, along with a link. They were randomly assigned to two scenarios, according to low and high levels of involvement. The low-involvement scenario was as follows:

Imagine that while viewing your SNS account, you find an ad for a popular beer brand. This brand is currently promoting a new beer and offers a 40 percent discount coupon. To obtain this coupon, you click the link and register your name and e-mail address. This promotion lasts 1 week, and the coupon can be redeemed at any store. Based on this scenario, please rate the following questions according to a seven-point scale (1=completely disagree; 7=completely agree).

After reading this, the respondents completed a structured questionnaire.

Results

Manipulation and realism check

Levels of involvement were compared for beer and laptop computer with the Mittal's¹⁸ scale. A t-test indicates that the difference was statistically significant at p<0.01. A realism check, performed by a scale suggested by Dabholkar and Bagozzi,²² suggests that the scenarios were sufficiently realistic (M=4.51).

Reliability of the dependent variables

We conducted a confirmatory factor analysis for information value, entertainment value, friendship value, normative pressure, and eWOM intention. Loadings, t-values, and goodness-of-fit indexes appear in Table 1. Table 2 summarizes Cronbach's alpha, composite reliability, and average variance extracted (AVE), which are all satisfactory according to Hair et al.'s²³ criteria. In the same table, discriminant validity was established by comparing construct-to-construct correlations with the squared roots of AVE.

Table 1.

Confirmatory Factor Analysis

		Loadings	t-Value
Information value	I think the information I found may be helpful in solving someone's problem.	0.85
Information value	The information I found may provide specific benefits to others.	0.85	26.72
Information value	Posting comments in this SNS may help the other users generate ideas.	0.78	23.90
Entertainment value	SNS really entertains me.	0.84
Entertainment value	Exchanging information in this SNS is really enjoyable.	0.79	26.61
Entertainment value	Distributing information in this SNS is a fun way to kill time.	0.87	31.09
Entertainment value	Using this SNS relaxes me.	0.87	31.91
Friendship value	This SNS has something to do with or say to others.	0.84
Friendship value	This SNS helps me to stay in touch with others.	0.82	27.37
Friendship value	I use this SNS because it reduces the feeling of loneliness.	0.86	29.61
Normative pressure	In order to be accepted on this SNS, I feel like I must behave as other users expect me to behave.	0.86
Normative pressure	On this SNS, my actions are often influenced by how other users want me to behave.	0.89	31.62
Normative pressure	I feel pressured to use this SNS as much as the other users do.	0.80	26.80
eWOM intention	I would write about this campaign in my SNS so that my contacts would get to know this promotion.	0.91
eWOM intention	I would pass this information to my fiends through my SNS.	0.94	44.45
eWOM intention	I would maximize the diffusion of this campaign on my SNS to make sure all my contacts would know.	0.91	41.69

Standardized coefficients. All t-values are statistically significant at p<0.001.

Goodness-of-fitness indexes: χ²₉₄=472.27; CFI=0.97; TLI=0.96; IFI=0.97; RMSEA=0.065.

eWOM, electronic word-of-mouth; SNS, social networking site.

Table 2.

Discriminant Validity

Constructs	M	SD	α	CR	AVE	1	2	3	4	5
1. Information value	4.17	1.66	0.86	0.86	0.68	0.82
2. Entertainment value	4.46	1.57	0.90	0.93	0.71	0.65	0.84
3. Friendship value	5.16	1.48	0.88	0.88	0.71	0.56	0.77	0.84
4. Normative pressure	3.00	1.69	0.88	0.88	0.72	0.50	0.44	0.20	0.85
5. eWOM intention	3.72	1.91	0.94	0.94	0.85	0.39	0.38	0.27	0.38	0.92

The analysis was performed with maximum likelihood method. Diagonal elements in bold are the square root of AVE between the constructs and their indicators. Off-diagonal elements are correlations between the constructs.

M, mean; SD, standard deviation; α, Cronbach's alpha; CR, composite reliability; AVE, average variance extracted.

Hypotheses testing for H1

Data were analyzed on the basis of a one-way between-subjects multivariate analysis of variance (MANOVA). The independent variable was propensity to gossip. A median split was performed on this variable such that half the participants were coded as “gossipers” when the score exceeded 3.30 (high propensity to gossip; M=4.58). The remaining half was coded as “nongossipers” (low propensity to gossip; M=2.25; F(1, 816)=1769.23, p<0.00l). Dependent variables were information value, entertainment value, friendship value, and normative pressure.

Before performing MANOVA, we tested basic assumptions. MANOVA is particularly sensitive to outliers; thus, we tested univariate and multivariate outliers²⁴ separately for each cell of the design. Our analyses revealed no serious outliers. Given the large sample size and the robustness of MANOVA, violations of multivariate normality are not expected to be severe.²³

Another assumption underlying MANOVA is equality of variance–covariance matrices. To test this, the Box's M test was used for homogeneity of dispersion matrices. The score was statistically significant, yet violation of this assumption has minimal impact if the groups are of approximately equal size.²³ In our study, sample sizes for gossipers (395) and nongossipers (423) are roughly matched.

H1 posits that gossipers would perceive online social functions more strongly than nongossipers. MANOVA results reveal a significant difference between gossipers and nongossipers [Pillai's criterion=0.17, F(4, 813)=41.82, p<0.001]. These overall differences in MANOVA justify further investigation of the direction and significance of each dependent variable using univariate tests of ANOVA. Results indicate that all variables exhibit significant differences between gossipers and nongossipers: information value: F(1, 814)=20.10, p<0.001; entertainment value: F(1, 814)=45.55, p<0.001; friendship value: F(1, 814)=10.24, p<0.01; and normative pressure: F(1, 814)=154.90, p<0.001. The mean values in Table 3 indicate that gossipers uniformly exhibit higher scores for each value. Thus, H1 is supported.

Table 3.

Means and Standard Deviations for H1

	Information value		Entertainment value		Friendship value		Normative pressure
Constructs	M	SD	M	SD	M	SD	M	SD
Nongossipers	3.91	1.52	4.13	1.45	5.04	1.41	2.40	1.30
Gossipers	4.39	1.38	4.81	1.22	5.29	1.20	3.66	1.47

Hypotheses testing for H2a–H3c

Data were analyzed on the basis of three-way (H2a and H3a) and two-way (H2b-c and H3b-c) between-subjects ANOVA. Median splits on responses to these scales divided participants into two groups. The dependent variable was eWOM intention. Tables 4 and 5 summarize means and standard deviations for three-way and two-way ANOVA, respectively. Table 6 shows two-way ANOVA results for both low- and high-involvement product scenarios.

Table 4.

Means and Standard Deviations of eWOM Intention (Three-Way ANOVA)

	Low involvement (beer)				High involvement (laptop computer)
	Gossipers		Nongossipers		Gossipers		Nongossipers
Constructs	M	SD	M	SD	M	SD	M	SD
Weak ties
Large network size	3.46	1.19	3.33	1.71	3.69	1.70	3.19	1.85
Small network size	3.17	1.48	2.95	1.71	3.63	1.42	3.06	1.64
Strong ties
Large network size	4.53	1.87	4.28	1.96	4.93	1.66	3.58	1.84
Small network size	4.42	1.80	4.33	1.73	4.03	1.68	4.36	1.79

ANOVA, analysis of variance.

Table 5.

Means and Standard Deviations of eWOM Intention (Two-Way ANOVA)

	Low involvement (beer)				High involvement (laptop computer)
	Gossipers		Nongossipers		Gossipers		Nongossipers
Independent variables	M	SD	M	SD	M	SD	M	SD
Tie strength
Weak	3.28	1.37	3.13	1.72	3.65	1.53	3.11	1.72
Strong	4.48	1.83	4.30	1.86	4.50	1.72	3.89	1.85
Network size
Small	3.83	1.76	3.36	1.82	3.82	1.56	3.39	1.77
Large	4.23	1.77	3.75	1.88	4.46	1.77	3.37	1.85

Table 6.

Summary of Two-Way ANOVA Results

High involvement
		High involvement (laptop computer)
Hypotheses	Source	df	MS	F
H2b	Gossipers (G)	1	32.32	11.16**
	Tie strength (T)	1	63.98	22.10***
	G x T	1	0.14	0.83
	Error	405	2.90
H2c	Gossipers (G)	1	58.53	19.50***
	Network size (N)	1	9.99	3.33
	G x N	1	11.18	4.01*
	Error	405	3.00

Low involvement
		Low involvement (beer)
Hypotheses	Source	df	MS	F
H3b	Gossipers (G)	1	2.75	0.92
	Tie strength (T)	1	134.35	44.98***
	G x T	1	0.02	0.01
	Error	405	2.99
H3c	Gossipers (G)	1	22.74	6.93**
	Network size (N)	1	15.77	4.81*
	G x N	1	0.01	0.00
	Error	405	3.28

^***p<0.001, ^**p<0.01, ^*p<0.05.

MS, mean squares.

The difference in the amount of variance of one group versus another was examined using the Levene's test for equality of variance for two scenarios. This assumption was met for all cases.

H2a-c relate to the high-involvement product. H2a postulates that, when the effects of tie strength and network size are simultaneously considered, gossipers are more prone to participate in eWOM, compared with nongossipers. To address this hypothesis, a three-way ANOVA was performed. Our results show that the main effect of propensity to gossip was statistically significant [F(1, 401)=8.86, p<0.001] in the predicted direction—gossipers exhibit stronger eWOM intention than nongossipers. Thus, H2a was supported.

H2b contemplates that, when gossipers are connected with strong ties rather than weak ties, they tend to show more willingness to participate in eWOM on SNSs. Nonetheless, in our two-way ANOVA, the interaction effects between propensity to gossip and tie strength were not statistically significant. Thus, H2b was not supported.

In H2c, we predict that gossipers belonging to a large network, as opposed to a small network, are more eager to spread the word via eWOM on SNSs. Our two-way ANOVA indicates that the interaction effects between propensity to gossip and network size were significant at p<0.05. As Figure 1 shows, the positive effect of propensity to gossip on eWOM intention is greater among large network members than small network members. This supports H2c.

FIG. 1.

Interaction effect of propensity to gossip and network size (high-involvement product).

H3a–c were formulated for low involvement. Under this condition, H3a hypothesizes that, when the effects of tie strength and network size are simultaneously considered, no statistically significant difference in eWOM intention between gossipers and nongossipers. Our three-way ANOVA indicate that only the main effect of tie strength was statistically significant and propensity to gossip had no influence on eWOM intention [F(1, 401)=0.89]. Thus, H3a was supported by our data.

Next, H3b predicts that, regardless of propensity to gossip, members with strong ties are more willing to participate in eWOM than members with weak ties. Our two-way ANOVA results were consistent with our prediction—members with strong ties indeed demonstrate stronger eWOM intention [strong ties=4.40 vs. weak ties=3.20, F(1, 405)=44.98, p<0.001], while no interaction effect was observed between propensity to gossip and tie strength. Thus, H3b was supported.

Finally, H3c postulates that, regardless of propensity to gossip, members in a small network, rather than those in a large network, are more prone to participate in eWOM. However, our two-way ANOVA results contradict this—eWOM intention was actually greater for members in a large network [large network=3.99 vs. small network=3.60, F(1, 405)=4.81, p<0.05]. There was no interaction effect between propensity to gossip and the network size. Thus, H3c was not supported. Our hypotheses testing results are summarized in Table 7.

Table 7.

Summary of Hypotheses Testing Results

Hypotheses		Results
H1	Propensity to gossip influences online social functions such that these functions are more consolidated among members with high propensity to gossip, compared with members with low gossping propensity.	Supported
H2a	When the effects of tie strength and network size are simultaneously considered, gossipers, rather than nongossipers, would be more willing to participate in eWOM on an SNS.	Supported
H2b	Gossipers' willingess to participate in eWOM would be more pronounced if they are connected with strong ties rather than weak ties.	Rejected
H2c	Gossipers' willingness to particiapte in eWOM would be more pronounced if they belong to a large network rather than a small network.	Supported
H3a	When the effects of tie strength and network size are simultaneously considered, there is no significant difference between gossipers and nongossipers in terms of eWOM intention on an SNS.	Supported
H3b	Regardless of propensity to gossip, members connected with strong ties, rather than weak ties, would be more willing to participate in eWOM on an SNS.	Supported
H3c	Regardless of propensity to gossip, members belonging to a small network rather than a large network would be more willing to paticipate in eWOM on an SNS.	Rejected

Conclusions

Overall results are consistent with theoretical predictions. Gossipers, compared with nongossipers, perceived information value, entertainment value, and friendship value more strongly, while feeling more pressured by group norms. They are interested in prestige or fame by being recognized as an expert on a particular topic or area of knowledge by their fellow SNS members. When they find interesting information to serve this purpose, they are willing to spread it. This tendency is stronger when they belong to a large network, but the closeness of the relationship among members is no longer relevant. We tested this thesis with a high-involvement product, laptop computer, compared with a low-involvement product, beer. Results are consistent in that online gossipers on SNSs tend to act as we predicted. In contrast, propensity to gossip loses its role when the product is uninteresting or unattractive. Such trivial information can be spread only among those connected with strong ties, rather than weak ties.

One interesting implication from this research is the use of celebrities in SNS ads as a spokesperson or endorser. Members may engage in tabloid gossip about that celebrity, and turn their attention to the advertised item. Theoretically, the literature has long addressed celebrity endorsement in a context of categorization theory, scheme theory, and image transfer, thus the role of celebrity as a gossip inducer adds to the current knowledge. Managerially, the use of celebrity in SNS marketing can be demonstrated, testing different levels of product involvement. If our research results hold true, the higher the product involvement, the more likely SNS members will be to gossip about the advertised brand.

Limitations and Future Research

This research should recognize two important limitations. First and foremost, this survey screened the SNS users with self-reported questions. However, the respondents claimed their usage behavior, since it is socially more desirable. Second, this study uses the scenario method with two types of products: beer and laptop computer. While the levels of product involvement were rigorously tested, the typical characteristics of each product category might have affected the survey results. Third, this scenario assumes that a sales promotion (discount coupon) could be gossiped about as an anecdotal story, while SNS members indulge in more personal gossip. This assumption may receive criticism about its artifact. Future research should overcome these limitations to improve the study's generalizability.

Appendices

Appendix

Additional Scales Used in This Study

(1) Propensity to gossip舒based on Litman and Pezzo²⁰ (Cronbach's α=0.88)*

 Other people's life stories interest me.

 I love to know what is going on in people's lives.

 I like to share interesting news I have heard.

 It is more fun to talk about other people than serious topics.

 Gossiping is a great way to pass time.

 Gossip is a good icebreaker.

(2) Tie strength—adapted from Mittal et al.²¹ (Cronbach's α=0.89)*

 I am very close to this social networking site (SNS).

 I am tied to this SNS very strongly.

 I strongly feel that I am a part of this SNS.

 My relation with this SNS will continue in the future.

(3) Realism check—adopted from Dabholkar and Bagozzi²² (Cronbach's α=0.78)*

 The situation described was realistic.

 You had no difficulty imagining yourself in the situation.

*Measured on a scale of 1-7 (completely disagree/completely agree).

Acknowledgments

This research was funded by a grant from the Spanish Ministry of Science and Innovation (National Plan for Research, Development and Innovation ECO2011-30105).

Author Disclosure Statement

No competing financial interests exist.