Academy of Marketing Studies Journal (Print ISSN: 1095-6298; Online ISSN: 1528-2678)

Original Articles: 2017 Vol: 21 Issue: 2

Positive Versus Negative Word-of-mouth: Effects on Receivers

Martin C. William, Eastern Washington University

Abstract

Despite extensive research which has examined the impact of word-of-mouth (WOM) valence, results are often conflicting with regard as to whether receivers are biased toward positive or negative WOM. In this study, the impact of WOM valence on receivers’ use of actual WOM in their decision making is directly examined. Results indicate that positive WOM has a significantly greater effect on receivers’ attitude toward the focal product than does negative WOM. Receivers perceive senders of WOM concerning utilitarian products to be more trustworthy when their WOM is positive rather than negative, though no differences in the perceived altruism of senders is found. Senders of positive WOM are also viewed as having greater experience and evidence of their claims than are senders of negative WOM. Implications for marketing practitioners, in addition to directions for future research, are provided.

Keywords

Word-Of-Mouth (WOM), Receivers.

Introduction

Recent research is beginning to place particular emphasis on investigating how receivers of WOM make use of it in their decision making. Receivers have been found to be selective in determining how much value they place on information gathered from WOM incidents and whether they will make use of it (Martin & Lueg, 2013; Sen & Lerman, 2007; Sweeney, Soutar & Mazzarol, 2008 & 2012). As such, identifying the factors that impact receivers’ use of WOM and their conditions and boundaries is key to understanding how WOM functions and how marketers’ strategies should be tailored around WOM.

A key feature of WOM impacting receivers is valence, which refers to whether the focal product is endorsed or eschewed by the sender. However, conflicting results abound as to whether positive or negative WOM information has a greater impact on receivers. Some researchers have found evidence of a negativity bias whereby negative WOM has a greater effect on receivers than does positive WOM (Arndt, 1967; Herr, Kardes & Kim, 1991; Yang & Mai, 2010), others have observed a positivity bias (Gershoff, Mukherjee & Mukhopadhyay, 2003; Kim, Sung & Kang, 2014; Skowronski & Carlston, 1989; Sweeney et al., 2012; Wu, 2013; Xue & Zhou, 2010) and still others have concluded that the type of bias present is contextual and not universal (Ahluwalia, 2002; Laczniak, DeCarlo & Motley, 1996; Kim & Gupta, 2012; Zhang, Ye, Law & Li, 2010). This research aims to shed new light on the effects of WOM valence on receivers in two ways. First, rather than examine the impact of valence on a review’s perceived helpfulness (e.g. Gilly, Graham, Wolfinbarger & Yale, 1998; Kim & Gupta, 2012; Wu, 2013) or brand attitudes (e.g. Herr et al., 1991), this study examines the impact of WOM valence on a framework of receivers’ use of WOM in their decision-making processes (Martin and Lueg, 2013; Martin, 2014). This allows for more precise examination of the underpinnings of the effects of WOM valence. Specifically, by building on this framework of WOM usage, the impact of WOM valence on both the antecedents and effects of receivers’ use of the WOM can be surveyed. Second, this study investigates WOM incidents actually experienced by receivers, which may lend greater external validity to the results than the hypothetical scenarios frequently used in WOM research (e.g. Herr et al., 1991; Kim & Gupta, 2012).

Literature Review and Hypotheses

Humans are more attentive to negative information and this information tends to be more heavily weighted in their evaluations of objects than does positive information (Ito et al., 1998). Some researchers assert that a negativity bias is present among receivers of WOM, arguing that negative WOM is more influential due to its reduced occurrence compared to positive WOM (Herr et al., 1991; Yang & Mai, 2010). Additional support for this explanation is based on the finding that individuals are generally more attentive to negative than positive information; the threat of a potential loss is typically viewed as more influential than the hope of a potential gain (Kahneman & Tversky, 1984).

Others, arguing for a positivity bias, claim that positive WOM has greater accessibility and diagnosticity than negative WOM (Showronski & Carlson, 1989). Information which is extremely positive appears to have a greater impact on consumers’ product evaluations than does extremely negative information (Gershoff et al., 2003). Further, as there is indeed robust evidence that positive WOM occurs more frequently in the marketplace than does negative WOM (East, Hammon & Lomax, 2007; Naylor & Kleiser, 2000), consumers appear to often be persuaded to carry a positive attitude toward products in general. In accordance with social judgment theory, the greater potential impact of negative WOM compared to positive WOM may be more than offset by consumers’ preexisting positive attitude toward the focal product (Peterson & Wilson, 1992). As such, if consumers already believe a product to be favorable, negative WOM may be less successful in changing their views than positive WOM (Sweeney et al., 2014).

Recent research regarding the mechanisms underlying consumers’ utilization of WOM in their decision making may provide new insight into which type of bias, if any, valence may have on receivers. This research has examined how receivers’ perceptions of WOM senders influences the receivers’ use of that WOM in product evaluation, perceived risk and purchase intentions (Sweeney et al., 2008; Martin & Lueg, 2013; Martin, 2014). Receivers’ perceptions of senders’ trustworthiness and altruism have a strong impact on whether receivers will use the provided information in determining a course of action (Martin & Lueg, 2013; Martin, 2014). According to attribution theory (Eagly & Chaiken, 1993; Kelley, 1973), if a sender is believed to be providing reliable information or genuinely helping the receiver to experience optimum benefits, receivers will respond in a manner consistent with the recommendation being offered to them. However, if receivers believe a sender to be engaging in WOM merely in an attempt to exact revenge on the offending firm, a common motivation of negative WOM (Sundaram, Mitra & Webster, 1998), they are presumably less likely to act on the information provided. It has been suggested that receivers may be apt to believe that senders of negative WOM are acting vindictively or had an unfavorable encounter with the problem due to a lack of ability on their part (Kim et al. 2014; Skowronsky & Carlston, 1989). To the extent that senders of negative WOM are believed to be acting out of a desire for spite directed toward the focal product, receivers seem apt to discount the value of this WOM.

Taken together, it appears that receivers of positive WOM may believe senders to be more trustworthy and altruistic than senders of negative WOM. In this context, trustworthiness refers to the degree of authenticity of the sender’s information (Pornpitakpan, 2004) and altruism is “the extent to which the speaker is believed to be acting in the interests of the listener” (Martin, 2014).

H1 Receivers believe senders of positive WOM to be more (a) trustworthy and (b) altruistic than senders of negative WOM.

Along similar lines, it can further be hypothesized that the extent to which receivers use the information provided by senders, referred to as WOM usage (Martin & Lueg, 2013), will have a greater impact on receivers’ attitude toward the focal product when WOM is positive rather than negative.

H2 The relationship between WOM usage and attitude toward the recommended product is stronger when WOM is positive rather than negative.

While there may indeed be a generalizable positivity bias as suggested by H1, this bias may not hold in all instances. Boundary conditions for this hypothesized positivity bias may well exist. One of these may be the type of value provided by the cited product. Research has indicated that positive WOM is viewed as being more useful than negative WOM for hedonic products, while the inverse is true of utilitarian products (Sen & Lerman, 2007). Part of the reasoning for this finding lies in the affect-confirmation hypothesis; when consumers receive WOM regarding a hedonic product, they begin anticipating the potential positive, emotional benefits of the product and disregard information (i.e. negative WOM) that is counter to their positive mood (Adaval, 2001; Sen & Lerman, 2007). In contrast, negative WOM involving utilitarian products may be more diagnostic in assessing the utility of the product. “Negative experiences with tangible attributes can directly impact the utility that the consumer will likely derive from the product. Because the goal of utilitarian consumption is to maximize utility, such negativity will likely be weighted rather heavily when evaluating a utilitarian product” (Sen & Lerman, 2007).

Therefore, it seems likely that for utilitarian products, the information provided by senders of positive WOM will be viewed as less reliable, useful and helpful than that offered by senders of negative WOM, while the inverse may be true of hedonic products. In sum, it appears that there is a positivity bias in senders’ perceived trustworthiness and altruism may be attenuated for utilitarian products but strengthened for hedonic products.

H3 For utilitarian products, receivers believe senders of positive WOM to be less (a) trustworthy and (b) altruistic than senders of negative WOM.

H4 For hedonic products, receivers believe senders of positive WOM to be more (a) trustworthy and (b) altruistic than senders of negative WOM.

Martin and Lueg (2013) demonstrated that senders’ experience, referring to their firsthand knowledge of the product being recommended, as well as senders’ evidence, confirmation of the validity of their claims, had a positive effect on WOM usage. While no hypotheses were made concerning either of these two variables in the current study, they were investigated for exploratory purposes.

Methods

Responses were collected using the student referral method (Babin, Hardest & Suter, 2003) at a mid-sized university in the Pacific Northwest. Students in marketing courses were offered extra credit for participating in the survey as well as by recruiting up to four other people, at least two of whom were required to be over the age of forty. This resulted in the collection of usable data from 581 respondents. Fifty-one percent of respondents were female and mean age was 31.

Respondents were asked to think back to the last positive or negative WOM incident they experienced. Source trustworthiness was measured with a four-item semantic differential response scale using seven points (Pornpitakpan, 2004). Source altruism was measured using the four-item Likert anchored by “Strongly Disagree” (1) and “Strongly Agree” (7) developed by Martin (2014). Source experience was measured using a three-item Likert scale anchored by “Strongly Disagree” (1) and “Strongly Agree” (7) developed by Braunsberger and Munch (1998) and revised by Martin and Lueg (2013). Source evidence and WOM usage were measured with three-item and six-item, respectively, Likert scales anchored by “Strongly Disagree” (1) and “Strongly Agree” (7) (Martin & Lueg, 2013). Attitude toward the recommended product was measured using the four-item semantic differential response scale using seven points (Iyer, 1988).

In order to determine whether the cited product in the WOM incident was utilitarian or hedonic, responses to the utilitarian/hedonic value scale (Voss, Spangenberg & Grohman, 2003) were compared. Products receiving a higher rating on the utilitarian scale were coded as utilitarian products; the same approach was taken for hedonic products. Respondents who scored equally on both measures were disregarded from further analysis. Using this approach, 332 of the focal products in the WOM encounters primarily elicited utilitarian value and 249 primarily elicited hedonic values.

An exploratory factor analysis of the six constructs of interest was then conducted using principal axis factoring. This analysis indicated that seven factors had an eigenvalue greater than one. These factors were then rotated using the Promax rotational method as the factors were assumed to be correlated with one another. This indicated that the items of each scale loaded strongly (78 or higher) on their respective factor with no significant cross-loadings present. These factors cumulatively accounted for 81.95% of the variance in their indicators. Coefficient alpha was greater than .90 for each measure, indicating that the measures possess strong reliability.

All of the measures were then subjected concurrently to a confirmatory factor analysis. This was accomplished using maximum likelihood estimation of the covariance matrix. The fit of the model was very good (χ2 =787.10, df=284, p<0.001; SRMR=0.033; RMSEA=0.055; NNFI=0.96; CFI=0.97). The standardized construct correlation matrix is shown in Table 1 in addition to the construct reliability, mean and standard deviation for each measure. Standardized loading estimates were above .70 for all items, construct reliability was above .90 for all measures and in each pair of measures, the average variance extracted for each measure was greater than the squared correlation coefficient for the two measures. In sum, the measures exhibit unidimensionality, convergent validity and discriminant validity (Hair, Black, Babin, Anderson &Tatham, 2006).

Table 1
Standardized Construct Correlation Matrix
  Altruism Trustworthiness Experience Evidence WOM Usage Attitude toward Product
Trustworthiness 0.69          
Experience 0.30 0.32        
Evidence 0.48 0.49 0.58      
WOM Usage 0.50 0.51 0.45 0.60    
Attitude toward Product 0.03 0.10 0.15 0.24 0.14  
Construct Reliability 0.91 0.91 0.94 0.94 0.96 0.94
Mean 5.25 5.59 5.47 5.26 5.30 5.04
Std. Deviation 1.43 1.20 1.35 1.42 1.38 1.61

Results

Regression analyses indicate that WOM usage has a significant impact on participants’ attitude toward the recommended product, though the strength of this impact depends on WOM valence and product type. The results are shown in Table 2. For both positive and negative WOM, WOM usage has a stronger impact on attitude toward the product when the recommended product is utilitarian as opposed to hedonic. However, the impact of WOM usage is far weaker when the WOM is negative for utilitarian products and non-existent for hedonic products. These results indicate that the relationship between receivers’ intent to use the WOM in their purchase decision and their attitude toward the product is far stronger when the WOM is positive than negative, a definite positivity bias. Further, the relationship between WOM usage and attitude toward the product is stronger for utilitarian than hedonic products.

Table 2
Impact of Wom Usage on Attitude toward the Product
WOM Valence Product Type Standardized β p-value
Positive WOM Utilitarian 0.608 <0.001
Hedonic 0.497 <0.001
Negative WOM Utilitarian -0.157 0.036
Hedonic -0.058 0.510

Respondents’ attitude toward the product was then examined across both positive and negative WOM as well as product type. These results are shown in Figure 1. There is a moderately significant interaction effect between WOM valence and product type (F=3.75, p=.053). Attitude toward the product is highest for utilitarian products receiving positive WOM (6.01) but lowest for utilitarian products receiving negative WOM (4.08). While receivers’ attitudes toward the focal product are diminished when the WOM is negative (4.36) as opposed to positive (5.92), the impact of WOM is significantly weaker for hedonic products than for utilitarian products. These findings are particularly intriguing given that the impact of WOM usage on receivers’ attitude toward the product is much stronger when WOM is positive rather than negative.

Figure 1: Attitude Toward yhe Product across Wom Valence and Product Type

To test H1, a MANOVA was conducted wherein senders’ trustworthiness and altruism were examined across positive and negative WOM incidents. This analysis indicates that WOM valence has a significant impact on the two dependent variables (Wilks’ Lambda=.988, F=3.57, df=2, p=0.028). Trustworthiness is significantly higher when WOM is positive rather than negative (meanpositive =5.72, meannegative =5.48, p=0.015), in support of H1a. There is no significant difference, however, in altruism across the two valences (meanpositive =5.31, meannegative =5.22, p=0.470), providing no support for H1b.

H2 was tested via a regression analysis wherein attitude toward the product was used as the dependent variable and WOM usage, WOM valence and an interaction term of WOM usage and valence were used as predictor variables. This model is significant (F=109.40, p<0.001, R2 =0.365) and indicates that a significant interaction effect between WOM usage and valence is present (p<0.001). When WOM is positive, WOM usage has a moderately strong, positive effect on receivers’ attitude toward the cited product (βpositive =0.564, p<0.001). When WOM is negative, WOM usage has a weak, though significant, negative effect on attitude toward the cited product (βnegative =-0.118, p=0.038). These results support H2.

Trustworthiness was then compared across the two types of WOM valence for utilitarian products.

The results indicate that trustworthiness is significantly higher for positive WOM than negative (F=4.52, df=2, p=0.034; meanpositive =5.84, meannegative =5.56), counter to H3a. Altruism is not significantly impacted by valence (F=1.57, df=2, p=0.212; meanpositive =5.50, meannegative =5.33), providing no support for H3b. Similar analyses were conducted for hedonic products. These indicate that WOM valence has no impact on trustworthiness (F=1.68, df=2, p=0.196; meanpositive =5.57, meannegative =5.37) nor altruism (F<0.01, df=2, p=0.960; meanpositive =5.05, meannegative =5.06), providing no support for H4.

Finally, the perceived experience and evidence of the WOM senders was examined across both positive and negative WOM via a MANOVA analysis. This indicated that WOM valence has a significant impact on these two variables (Wilks’ Lambda =0.932, F=20.99, df=2, p<0.001). Both the receivers’ experience (meanpositive =5.77, meannegative =5.21, p<0.001) and their evidence (meanpositive =5.64, meannegative =4.95, p<0.001) were higher for positive WOM.

Discussion

Research examining the effect that WOM valence has on receivers has repeatedly provided contradictory results. In this study, a unique approach is taken in that the impact of WOM valence on a framework of receivers’ intention to use the information provided by the sender is investigated rather than the extent to which receivers simply find the information to be helpful. This distinction is noteworthy as WOM usage has been shown to have a substantial impact on both receivers’ attitude toward the cited product and their intentions to purchase it (Martin & Lueg, 2013; Martin, 2014), whereas the same is not necessarily true of helpfulness. A review may be viewed as potentially helpful, but this does not mean that it will be a significant factor in the receiver’s decision. Further, this study utilizes actual WOM encountered by receivers as opposed to the hypothetical scenarios often employed in this area, improving the external validity of the findings.

Overall, the results of this study provide support for a positivity bias amongst receivers of WOM. The impact of receivers’ using negative WOM in their purchase decision does indeed diminish their attitude toward the cited product, though their use of positive WOM has a much stronger, positive effect on their attitude toward the product. This suggests that positive WOM is, at least generally, more influential in consumers’ decision making than is negative WOM.

At least a part of the mechanism underlying the observed positivity bias appears to be the impact of WOM valence on senders’ perceived trustworthiness. Consistent with other research (Kim et al., 2014), this study indicates that receivers view positive WOM sources as being more trustworthy than negative WOM sources, an important finding as trustworthiness is one of the key factors influencing whether a receiver will use the sender’s information in their purchase decision (Martin & Lueg, 2013; Martin, 2014). This means that receivers believe senders’ claims to be more genuine and accurate when their claims are positive rather than negative. Perhaps receivers believe that the poor attitude toward the product among senders of negative WOM is a result of their lacking the ability to use the product correctly or an otherwise aberrant situation outside the marketer’s control, leading receivers to discount the senders’ information.

Further, when WOM is positive, the perceived trustworthiness of senders is higher for utilitarian than hedonic products. This finding is counter to that suggested by prior research (Sen and Lerman 2007) and provides additional evidence of a positivity bias toward WOM on the part of receivers. A possible explanation for this effect is that consumers may believe that due to the relatively objective nature of evaluations of utilitarian products, it is easier to determine whether a sender’s claims are genuine. Assuming that the claims are, in fact, accurate, trustworthiness would logically be greater for utilitarian products. By comparison, hedonic products are more dependent on both the sender’s and the receiver’s potentially disparate personalities and values, making it comparatively difficult to determine the accuracy and reliability of the senders’ information.

Unlike the sources’ perceived trustworthiness, senders’ altruism was unaffected by WOM valence or product type. Receivers appear to believe that senders of both positive and negative WOM are equally concerned about the receivers’ welfare. A receiver may not be providing very accurate information (low in trustworthiness), but may well be spreading WOM out of a genuine desire to help the receiver (high in altruism). Further, given that altruism is a common motivation among WOM senders for both positive and negative WOM (Sundaram et al., 1998), receivers may implicitly believe that senders are trying to be sincerely helpful in both instances. And whereas the trustworthiness of senders can often be verified in some way by receivers, the same cannot be said of altruism. In the absence of evidence to the contrary, such as the sender potentially receiving a benefit from persuading the receiver (e.g. Martin, 2014), receivers may be willing to give senders the ‘benefit of the doubt’ when it comes to judging their motivations for the WOM.

Also, it seems that the observed positivity bias in WOM is also partially due to differences in receivers’ views of the senders’ experience concerning the focal product, as well as evidence of their claims. Receivers view senders of positive WOM as having both greater experiences with the product being discussed than senders of negative WOM as well as more evidence to support their claims. Potentially, receivers may believe that senders of negative WOM are more apt to be spreading their information on the basis of what they have heard from others rather than their own encounters with the product.

Thus, it seems that firms may not need be as concerned with negative WOM as previously suggested in the literature. Positive WOM seems to be more impactful in creating favorable views of products among consumers than negative WOM is in diminishing their opinions of the same. Implementing programs that facilitate the sharing of information among customers would be beneficial for firms, even though this will likely mean that some amount of negative WOM will be distributed in the process. Social media, in particular, may be a very effective means of doing so as it has proven itself to be a very powerful source of WOM (Breazeale, 2009). While some marketers might be overly concerned with negative WOM resulting from social media, rather than attempting to censor negative WOM from social media, a preferable strategy would seem to be to encourage as much feedback as possible from consumers. This would seem to be especially effective for marketers of hedonic products, where negative WOM does not have a significant impact on consumers’ attitude toward the product, but even for utilitarian products, the beneficial impact of positive WOM on attitude toward the product is far greater than the adverse impact of negative WOM.

Limitations and Future Research

It may be that the observed bias that receivers have toward positive WOM is due to the generally favorable outlook that consumers have toward most products. If the majority of information, whether from WOM or other sources, suggests that a product is inferior, the positivity bias observed here may well dissipate.

The method used in this research does not directly compare product categories. Results might vary if positive and negative WOM were directly compared for specific product categories (e.g. Clothing, Electronics).

Results of this study indicate that receivers believe senders of both positive and negative WOM to be behaving in an equally altruistic manner, though they believe the senders of positive WOM to be more trustworthy. Future research should seek to determine the underlying causes of this effect.

Further, most research, including this study, examines WOM valence as a dichotomy: positive or negative. WOM may be viewed by receivers as being generally positive but frequently also possesses some negative statements regarding a product and the same could be said of negative WOM. As such, future research should potentially investigate WOM valence as either a continuous variable or attempt to disentangle the multiple statements that can easily exist in a single WOM encounter.

References