Asymmetric Persuasive Effects of Gain- and Loss-related Messages in Electronic Word of Mouth

This study examines the aggregate effect of electronic word mouth (eWOM) communications containing multiple messages of different type on brand attitude. It focuses on the moderating role of individuals’ regulatory focus and message proportion in influencing the extent to which consumers respond to gainand loss-related messages. We develop some hypotheses regarding the interplay between the constructs, and test them through two web-based experimental studies on online product reviews. In study 1, we examine the persuasiveness of four different reviews composed of several combinations of gainand loss-related messages. In study 2, we modify the proportion of positive and negative messages to examine how the impact of eWOM is affected by disproportionate message structure. The results reveal that different combinations of message types lead to different evaluation of the focal brand. Furthermore, subjects with different regulatory focus exhibit different attitudes toward the focal brand when exposed to the same message combination. In addition, the moderating effects of regulatory focus appear to be altered by eWOM message proportion. Theoretical and managerial implications of this study are discussed.


Introduction
The persuasive impact of eWOM on various aspects of consumer behavior has been well recognized in the literature.Studies have shown that eWOM can affect consumers' attitude toward a brand (Lee, Park, & Han, 2008;Lee, Rodgers, & Kim, 2009;Wu & Wang, 2011), product evaluation (Zhang, Craciun, & Shin, 2010;Kim & Gupta, 2012;Dou et al., 2012), purchase intention (Wang, Yu, & Wei, 2012;Jimenez & Mendoza, 2013;Fang & Yu, 2017), and brand choice (Senecal & Nantel, 2004;Cheung, Lee, & Rabjohn, 2008;East, Hammond, & Lomax, 2008).Recent research has indicated the growing number of consumers who perceive eWOM as more reliable than firm-generated communication tools such as prints ads, personal selling, or TV commercials (Trusov, Bucklin, & Pauwels, 2009;Lee & Youn, 2009).The primary distinction between eWOM and traditional marketing communications is that the former may contain positive as well as negative messages about a product or service (Hennig-Thurau et al., 2004).Thus, while firm-generated communications are expected to influence consumers in favorable ways, eWOM can cause unfavorable impacts on consumers' attitudes or decisions.Accordingly, how individuals will be ultimately affected by each message that constitutes an eWOM has been a primary concern among academicians and practitioners (King, Racherla, & Bush, 2014).
However, despite a large body of research on this topic, there is an ongoing conflict about the intensity of eWOM messages.On one hand, a research stream contends that positive messages are more influential than negative messages, a phenomenon well-known as the positivity bias (East et al., 2008;Fang & Yu, 2017).On the other hand, other researchers assert the negativity bias, that is, the impact of negative messages is greater than positive messages (Skowronski & Carlston, 1989;Park & Lee, 2009).With this regard, there is ample work intended to reconcile the conflicting ideas and figure out potential conditions under which positivity or negativity biases are likely to occur.Recent studies have revealed that factors such as product characteristics (Park & Lee, 2009;Pan & Zhang, 2011), recipient characteristics (Sen & Lerman, 2007;Jones, Aiken, & Boush, 2009), provider characteristics (Shin, Song, & Biswas, 2013;Hornik et al., 2015), and message characteristics (Park & Kim, 2008;Melián-González et al., 2013) are likely to moderate the extent to which consumers evaluate and accept a particular message.Focusing on recipient characteristics, few researchers have recently found that regulatory focus theory (Higgins, 1997) can be used to better explain under what condition positive messages overwhelm negative messages, and vice versa.Drawing on the theory, positivity (negativity) bias is postulated to occur when promotion (prevention) focus consumer is exposed to positive (negative) messages.
The explanation based on regulatory focus theory seems to be plausible, given the ample evidence that individuals' goal orientations are associated with their attentions and responses to a particular persuasion (Aaker & Lee, 2001;Lee & Aaker, 2004;Keller, 2006;Wang & Lee, 2006).Specifically, when the valence of a message is congruent with consumers' regulatory foci, its effect would appear to be more salient because consumers are more inclined to pay attention and behave in a determined way recommended by the message.However, we argue that the existing literature grounded on the theory have some limitations in the assessment of eWOM effects.First, most studies have focused on a single message that is either positive or negative.This is impractical because in general consumers encounter with eWOM constituted by both positive and negative messages.Therefore, the ultimate effect of eWOM on any outcome variable must be assessed as the net effect of all messages contained.Few studies including Doh and Hwang (2009) and Melián-González et al. (2013) indeed examined the aggregate effects of multiple messages but they did not consider consumers' regulatory focus.Second, existing studies did not account for the fact that for a given message valence, its congruency with consumers' regulatory foci would depend on the type of the message.For example, a positive message can either be the one that conveys the presence of product advantages (gain) or the absence of product disadvantages (non-loss).The impacts of gain and non-loss messages should be different for promotion and prevention focus consumers (Idson, Liberman, & Higgins, 2000, Liberman, Idson, & Higgins, 2005).This is also the case for any negative message that can be either the one telling the presence of product disadvantages (loss) or the absence of product advantages (non-gain).
These gaps in the literature remain some open questions: (1) How would consumers' responses to eWOM containing, for example, gain and non-loss messages be different from their responses to that containing gain and loss messages?(2) How can the differences be explained by using regulatory focus theory?(3) How would the outcome change if one of the message types outnumbers the other?In this study, we aimed to address these questions by examining the aggregate effects of online product reviews which are composed of various messages of different types, where brand attitude is considered as the outcome variable.For this purpose, we developed testable relevant hypotheses based on the literature and subsequently conducted two web-based experiments in which we exposed subjects to ten reviews concerning a product in a single board and then asked their attitudes toward the focal product.Individual's regulatory focus was measured by using a scale frequently used in the previous studies.In study 1, we tested the persuasiveness of four different eWOM designed as some combinations of gain-related (gain or non-gain) and loss-related (loss or non-loss) messages.For each eWOM, we balanced the proportion of message valences so that it has five positive and negative messages.In study 2, we extended study 1 by modifying the proportion of positive and negative messages to examine how the results would change when a type of message outnumbers the other.
The main contribution of the present research is that it expands the existing literature on eWOM by taking into account the net effect of several message types: gain, non-gain, loss, and non-loss messages.In that sense, the focus is beyond on positivity and negativity nature of a message as in past studies since it includes the examination of message intensity in terms of individual's goal orientation.The analysis of two studies resulted in some important insights.First, we found that different combinations of message types lead to different responses to eWOM.Second, subjects' responses to eWOM are moderated buy their goal orientations.Third, the moderating effect of regulatory focus appears to be altered by message proportion.
The remainders of this paper are organized as follows.In the next section we discuss the theoretical backgrounds underpinning our expectation the role of one's goal orientation in influencing message intensity.Subsequently, we illustrate the analytical framework used to examine the effect of eWOM.We then describe the experimentation designs and present the results of study 1 and study 2. Following these sections, we discuss the theoretical and managerial implications of our findings.Finally, we conclude with some limitations and directions for future research.

Positivity and Negativity Biases
Research studies have shown the asymmetric effects of positive and negative messages.However, the results are inconclusive because some researchers found positive messages are stronger than negative messages (i.e., the positivity bias), whereas some others found the other way around (i.e., the negativity bias).For example, it has been shown that people tend to utilize positive WOM, rather than negative WOM, as a main source in the adoption of new product (Keaveney, 1995).Further, East et al. (2008) pointed out that because, in many cases, positive messages outnumber negative messages, the impact of the former appears to be greater than the latter.In the context of eWOM, a recent study by Fang and Yu (2017) suggested that positive messages have a greater effect on purchase intention compared to its negative counterparts.By contrast, an early study on traditional WOM by Arndt (1967) revealed that negative messages have greater effect on consumer decision to buy a new product.In the context of product evaluation, Skowronski and Carlston (1989) suggested that consumers put more weight to negative rather than positive information in forming their judgments.More recently, Liao et al. (2015) provided evidence that negative eWOM has a stronger effect in generating information richness than positive eWOM.Other studies supporting negativity bias include Homer and Yoon (1992), Park and Lee (2009) and Richins (1983).

Regulatory Focus Theory
Regulatory Focus theory (Higgins, 1997) suggests two motivational orientations that influence individual's behavioral intention or decision making: promotion and prevention focus.According to the theory, promotionand prevention-focused individuals are influenced by different strategic means; that is, the former tend to employ an approach strategy and the latter tend to employ an avoidance strategy.Specifically, promotion focused individuals would pay more attention to the presence or the absence of gain (an approach strategy), whereas prevention focused consumers are more concerned with the absence or presence of loss (an avoidance strategy) (Aaker & Lee, 2001;Higgins, 1997;Tuan Pham & Chang, 2010).Therefore, promotion focused consumers are likely to be sensitive to positive outcomes, whereas prevention focused consumers are likely to be sensitive to negative outcomes.
With regard to individual's response to particular information, the theory implies that the fluency of processing and the likelihood of acceptance of the information should depend on the congruency between its content and the goal of her/his own (Higgins, 2000).That is, information concerning gain (loss) is more likely to be processed and perceived to be more persuasive by promotion-(prevention-) focused individuals.A study by Aaker and Lee (2001) suggested that individuals demonstrate greater recall and more favorable attitude toward information that is compatible with regulatory focus, providing a support for this contention.In the context of eWOM, Kim and Lee (2015) found that promotion-focused subjects rated the usefulness of a positive product review higher than did prevention-focused subjects.The result was reversed when subjects were exposed to a negative product review.Similarly, Zhang, Craciun, and Shin (2010) pointed out that consumers with promotion (prevention) goals tend to perceive positive (negative) reviews to be more persuasive than negative (positive) ones when making product evaluation.In another study, however, the moderating effect of individual regulatory focus was insignificant when message credibility was treated as the outcome variable (Lee and Koo 2012).This is in contradiction with the work by Lee and Yi (2010) who partially found significant moderating effect of regulatory focus on the credibility of negative product reviews, although they did not find the same result for positive reviews.

Message Intensity
Based on regulatory focus and message valence, we can categorize eWOM messages into four types: gain, non-loss, non-gain, and loss (see Table 1).The first two are positive messages presenting the presence (absence) of product advantages (disadvantages) and the last two are negative messages presenting the absence (presence) of product advantages (disadvantages).While most studies on eWOM have focused on the persuasiveness of messages of different valence, it can be expected that for the same valence, the outcomes would vary depending on whether they are gain-related or loss-related messages.In fact, Lee and Aaker (2004) confirmed that gain-framed (i.e., gain and non-gain) messages appeared to be more persuasive when presented to promotion focused subjects than prevention focused subjects, and vice versa.Further, drawing on the principle of loss aversion (Kahneman & Tversky, 1979), Liberman, et al. (2005) predicted that losses would be perceived as more intensely negative than non-gains, and that non-losses would be perceived as more positive than gains.Their results revealed that this is the case for the former relation, but not for the latter relation.This finding suggests that loss (gain) messages would have greater negative (positive) effect than non-gain (non-loss) messages.However, as the study did not account for individuals' goal orientation, how the results would be different for promotion and prevention focused individuals remains unexplored.regulatory focus would be more salient for low-involvement individuals because they lack ability to make objective evaluation about the encountered information.Thus, the magnitude of its moderating effect should depend on the extent of involvement.Finally, the literature has shown that message acceptance determines the extent to which individuals process a message and behave in accordance with its content (Xu et al., 2010;Iyengar et al., 2015;Gupta & Harris, 2010).In our context, this implies that the persuasiveness of product reviews would be larger for individuals who are more inclined to accept the messages.

Hypotheses
As suggested by regulatory focus theory, individuals attach different weights to gains and losses according their goal orientations (Halamish et al., 2008;Zhang et al., 2010).This implies that the persuasive effect of a message depends on whether it conveys gain-related or loss-related information (Higgins, 2000).Promotion (prevention) focused individuals will pay more attention to and are more likely to be influenced by gain-related (loss-related) messages.Thus, the impacts would be enhanced when there is congruency between message type and regulatory focus.For example, the fit between information type and individuals' goal orientation should result in better recall and attitude toward the content of the information (Aaker & Lee, 2001).This argument should also apply in the context of eWOM, where consumers perceive its messages as more persuasive if the valence of the messages is congruent with their regulatory focus.Hence, H1: Individual's regulatory focus moderates the impacts of eWOM containing different type of messages on brand attitude.
Furthermore, as pointed out by Idson et al. (2000) and Liberman et al. (2005), individuals perceive that the pleasure from gains is greater than the pleasure from non-losses.While these studies built the premise for a single message, it is plausible to expect the same to hold for the case when multiple messages are exposed simultaneously.However, the results should vary depending on individual's regulatory focus.For promotion focused consumers, an eWOM containing gain and non-gain messages should result in more favorable attitudes than that containing non-loss and non-gain messages, provided that they attach greater weight to gain-related messages (Lee & Aaker, 2004).By contrast, for prevention focused consumers, the latter should lead to more favorable attitudes than do the former because they are more influenced by non-loss messages than by gain messages.In other words, the positivity of gain messages is less influential than that of non-loss messages for these consumers.Using the same reasoning, we expect to observe the same results when gain and non-loss messages are combined by loss messages.Hence, H2: Promotion (prevention) focused consumers will have more (less) favorable brand attitude when receiving gain/non-gain eWOM than they will when receiving non-loss/non-gain eWOM.
H3: Promotion (prevention) focused consumers will have more (less) favorable brand attitude when receiving gain/loss eWOM than they will when receiving non-loss/loss eWOM.
Another important finding from the study by Liberman et al. (2005) is that the pain from losses is greater than the pain from non-gains, which is consistent with the principle of loss aversion (Kahneman & Tversky, 1979;Tversky, 1994).In line with our previous predictions, we expect the argument to hold when loss and non-gain messages are combined with gain and non-loss messages.Specifically, it can be expected that eWOM containing loss and gain messages should result in unfavorable attitudes compared to those containing non-gain and gain messages for prevention focused consumers, provided that the negativity of loss messages are more salient for these consumers.By contrast, promotion focused consumers are likely to find the former to be less negative than the latter because they are less influenced loss-related messages.Applying the same reasoning to the case when the negative messages are combined with non-loss messages, we expect the same results for (loss + non-loss) and (non-gain + non-loss) combinations.Hence, H4: Promotion (prevention) focused consumers will have more (less) favorable brand attitude when receiving loss/gain eWOM than they will when receiving non-gain/gain eWOM.
H5: Promotion (prevention) focused consumers will have more (less) favorable brand attitude when receiving loss/non-loss eWOM than they will when receiving non-gain/non-loss eWOM.
Additionally, being composed of both positive and negative messages, eWOM persuasiveness may be affected by the ratio of message valences.Intuitively, if positive messages outnumber negative messages, then the positivity of the eWOM will become more intense, and vice versa.A study by Doh and Hwang (2009) revealed that a higher ratio of positive messages in eWOM results in more favorable attitudes and higher purchase intentions.However, they results also suggested that when all messages contained are positive, the credibility of the eWOM turns to diminish.In the context of this study, message proportions may enhance the role of message valence in consumer evaluation about a brand, altering the moderating effect of regulatory focus.For example, an eWOM containing the same number of gain and loss messages should result in favorable (unfavorable) brand attitude for promotion (prevention) focused consumers.However, when positive messages outnumbers negative messages, prevention focused consumers are likely to have favorable brand attitude, as promotion focused consumers do.Hence, H6: When messages of either valence outnumber the others, the effect of message valence on brand attitude will overwhelm the effect of regulatory focus.

Outline
In study 1, we conducted a web-based experiment to examine the moderating effect of regulatory focus and brand attitude differences as stated in H1 through H5.We designed four online reviews concerning a toothpaste product, each of which contains 5 positive and 5 negative statements (see Appendix A), representing message combinations depicted in Figure 1.To measure brand attitude, the dependent variable, we used four items rated in a five-point scale: low/high quality, bad/good, unfavorable/favorable, and negative/positive (Swaminathan et al., 2007).Individual's regulatory focus was measured by a scale used in Lockwood et al. (2002) and Lee and Koo (2012).The scale is composed of eight items representing the extent to which an individual is inclined to pursue gains and avoid losses.The sum of the latter items was subtracted from that of the former items, and then the median of the differences was used to split the subjects into promotion-and prevention-focused consumers.Further, we measured product involvement by asking how the subjects perceive the toothpaste category in a seven-point scale: unimportant/important, unattractive/attractive, and uninteresting/interesting (Zaichkowsky, 1985).In addition, message acceptance was measured by the scale proposed by (Lee & Koo, 2012;Cheung et al., 2009;Zhang & Watts, 2008).

Sample and Procedure
The experiment was conducted by an online research company targeted at randomly chosen 200 subjects (100 men).The ages of the subjects range from 15 to 69 years old, and the average was 40.28 years old.Table 2 shows the experimental design of this study.There are four groups receiving 10 product reviews containing different message combinations.We assigned each subject to one of the groups, and thus, all groups have 50 subjects.At the beginning of the experiment, we presented the product reviews in a single board to the subjects and asked them to read the reviews carefully.The order of the reviews shown to each subject was randomized to rule out the primacy and recency effects.After completing this task, we asked the subject evaluate the toothpaste brand to measure their attitudes.Finally, we asked them to answer the questionnaire on involvement, message acceptance, and regulatory focus.

Manipulation Check
To assure that the gain-and loss-related messages are perceived as different message types, we conducted a manipulation check targeted at 30 undergraduate and graduate students (13 males).First, we showed them the product reviews containing 5 gain, 5 non-gain, 5 loss, and 5 non-loss messages, and then asked them whether each message was telling the presence (or absence) of the advantages (or disadvantages) of the product.Non-gain gro wever, the ave n, leading to t ficant and with mers.Finally, w used consumer of the latter mers was insig t that the nega promotion-focu ss attention to might not be nsumers.

Outline
In study 2, we modified the proportion of message valences shown in a single board so that either positive or negative messages outnumber the others.In particular, we designed some message combinations containing 7 positive (negative) messages and 3 negative (positive) messages.This resulted in eight experimental groups which were exposed to different message type and proportion (see Table 5).We expected positivity (negativity) biases become more salient when positive (negative) messages were dominant in the product reviews, diminishing the moderating roles of regulatory focus.

Sample and Procedure
As in study 1, we conducted a web-based experiment to a randomly chosen 400 sample (200 men).Each subject was assigned to one of the eight experimental groups such that each group was composed of 50 subjects.The procedure was completely identical to that used in study 1.That is, we presented the product reviews to the subjects and then asked their attitude toward the brand under consideration along with involvement and message acceptance variables.However, to enhance external validity, in study 2 we selected a brand from body soap category as the experimental object rather than toothpaste category used in study 1.

Manipulation Check
To assure that the reviews containing 7 positive (negative) messages are perceived as positive (negative) reviews, we conducted a manipulation check targeted at 56 undergraduate students enrolling a marketing course in a large state university in Western Japan.The participant were exposed to one of the eight message combinations and then asked to evaluate the reviews whether it sounded positive or negative in aggregate in a 5-point Likert scale (1=negative, 5=positive).We subsequently conducted pairwise comparisons between two combinations of the same message type but with different proportion, and confirmed that positive dominant reviews are perceived as more positive than negative dominant reviews, and vice versa.For example, the test between the first group (7 gain : 3 non-gain) and the fifth group (3 gain : 7 non-gain) resulted in the rejection of the null hypothesis that there no difference between the two (t(54) = 4.82, p < 0.01).

Result and Discussion
We applied ANCOVA to subsamples with different message ratio to examine the influence of disproportionate message structure.Table 6 shows the results of the test.First, when positive messages outnumber negative messages, the interaction effect between message combination and regulatory focus turned to be insignificant (F(3,188) = 0.62, p = 0.60), indicating that the moderating role of regulatory focus was weakened for product reviews dominated by positive messages.Further, when negative messages accounted for a larger portion in the product reviews, the interaction effect was significant (F(3,188) = 2.97, p = 0.03); however, the magnitude of the F-value was smaller than that when positive and negative messages are of the same ratio as in study 1 (2.97 vs. 3.24).Because the F-value can be interpreted as the degree of deviance from the null hypothesis, we concluded that the moderating effect of regulatory focus is also weakened when negative messages outnumber positive messages.
ijbm.ccsen message valence is more influential.The consensus on this issue has not been reached because both conflicting ideas have empirical supports from the literature.The application of regulatory focus theory provides an alternative explanation concerning certain conditions under which positivity and negativity biases tend to occur (Kim & Lee, 2015;Zhang et al., 2010).This study extended previous findings by examining multiple messages of different type contained in a product review simultaneously.Consistent with previous findings, we confirmed that consumers' responses to eWOM vary depending on their goal orientations.Further, our results suggested that positivity bias is likely to occur when promotion-focused consumers receive eWOM containing gain messages.Likewise, negativity bias is likely to occur when prevention-focused consumers are exposed to eWOM containing loss messages.However, the extent of the biases appeared to be lower when promotion-focused consumers receive non-loss messages or when prevention-focused consumers receive non-gain messages.Our study also revealed that the moderating role of regulatory focus can be altered when eWOM contains disproportionate message valences.Specifically, when eWOM is dominated by gain messages, positivity bias can occur for prevention-focused consumers.Similarly, when loss messages are dominant, negativity bias may occur among promotion-focused recipients.Thus, our research contributes to the literature by elucidating how the interplay among message type, regulatory focus, and message ratio can give rise to positivity and negativity biases.

Managerial Implications
In addition, our findings may also be useful for marketers to anticipate the sales impacts of eWOM.Specifically, we argue that the understanding of how various message types would affect brand attitudes should be helpful for firms to better predict future sales by including product reviews as one of the predictors.For example, when the proportion of positive and negative messages is approximately equal, gain messages will result in more favorable brand attitude than non-loss messages, regardless of the type of other messages combined.Thus, marketers may expect an increase in the sales of their products if many consumers tell the others about the advantages of their products.Accordingly, our analysis suggest that marketers should design their marketing communications to improve consumers' understanding of the value of their products so that the consumers will help them spread positive eWOM, particularly gain messages, about the products.
More importantly, marketers should concern with the negative impacts of loss messages, rather than non-gain messages, because they may inflict appalling damage on their brand image, which eventually reduces consumers' purchase intention.Thus, if many consumers send loss messages about a product, this would result in a considerable decrease in the future sales.For this reason, some researchers emphasize the importance of managing negative online reviews to minimize the damage they would make (Lee et al., 2008).Although in many situations it would be difficult to restrict the number of such reviews, larger online providers like Amazon.com has been successful in reducing the number of harmful loss messages by providing guidelines that prohibit "profanity, obscenities, or spiteful remarks" for consumers who are willing to write a review.Further, if a manager is able to decide the order in which the reviews are displayed, she may place gain messages in the first order to be easily visible, and loss messages after the others, as suggested by the primacy effect (Lee & Koo, 2012).

Conclusion
This study investigated the aggregate effect of eWOM communication on brand attitude by taking into account the role of message valence, individual's regulatory focus, and message ratio.Through two web-based experimental studies, we examined how consumers' responses varied depending on message combination and regulatory focus, and the results supported the interaction effect between these variables.Further, this study provided evidence that the positivity (negativity) of gain (loss) messages is greater than that of non-loss (non-gain) messages in the context of eWOM.Finally, we confirmed the influence of message ratio in altering the moderating effect of regulatory focus.However, despite the substantial contribution it made to the body of knowledge, we note some limitations of this study.First, we only considered two grocery products whose attributes are relatively easier to evaluate prior to direct inspection.Thus, different results are likely to be derived if the analysis is conducted on experience goods such as automobiles or cosmetics for which the impact of eWOM on consumer decision tends to be greater.Second, we did not examine the potential effect of the order by which the messages are presented to the subjects.As suggested by previous studies, successive opposing messages can influence the final judgment or evaluation of a product (Haugtvedt & Wegener, 1994;Brunel & Nelson, 2003).Future research could address these issues by manipulating the message order and examining a wider range of products to improve its generalizability.

Table 2 .
Experimental design

Table 5 .
Experimental design of study 2