The Inﬂuence of Word-Of-Mouth Referral on Consumers’ Purchase Intention: Experimental Evidence from WeChat

: Based on the perspective of dual-system information processing, this article explores how word-of-mouth (WOM) referral affects the purchase intentions of consumers with different types of self-construal. Two experimental studies, using WeChat as a research setting, tested the interactive inﬂuence of type of WOM referral (economic vs. public welfare) and type of self-construal (independent vs. interdependent) on consumers’ purchase intention toward the referred product and its intermediary mechanism. The results showed that economic (vs. public welfare) WOM referrals, which emphasize a product’s economic beneﬁts, increased the purchase intention of consumers with an independent (vs. interdependent) self-construal, who typically take an individualistic perspective. In contrast, public welfare (vs. economic) WOM referrals, which emphasize the social beneﬁts connected to a company for the community at large, increased the purchase intention of consumers with an interdependent (vs. independent) self-construal, who take a more collectivistic perspective. The results also shed light on the underlying mechanism by showing that different types of self-construal activate different information processing systems in consumers, which mediate consumers’ reaction to different types of WOM referrals. Speciﬁcally, economic WOM referrals can inspire consumers with an independent self-construal to engage in more emotional processing, thereby generating higher purchase intentions; in contrast, public welfare WOM referrals can stimulate consumers with an interdependent self-construal to get involved in more cognitive processing, thereby generating higher purchase intentions. These ﬁndings contribute to WOM literature and provide practical implications for companies adopting societal marketing strategy and implementing sustainable promotional plans.


Introduction
The Internet and mobile Internet in China are developing rapidly, and the online instant messaging tools are convenient and to use frequently; therefore, the scale of (mobile) Internet users is huge. According to the 45th statistical report released by the China Internet Network Information Center (CNNIC), as of March 2020, the number of Internet users in China has reached 904 million. Internet penetration has reached terrific levels also in the rest of the world: for example, in 2020, the number of Internet users reached 332 million in North America and 727 million in Europe (Statista.com, 2020-a global provider of market and consumer data based in New York, USA). In the (mobile) Internet era, QQ, Weibo, WeChat, and other tools have become important media for people to transmit information. Most people are used to sharing their consumption experiences such as food, shopping, travel, and entertainment online. For example, consumers may shop on Tmall.com (an e-commerce platform which belongs to Alibaba company), upload pictures of the goods, and then share their own experiences after receiving the goods. More and more consumers are getting used to posting their own experiences and perception of a certain food, hotel, or tourist attraction in the WeChat Moments. For instance, on WeChat,

WOM Referral and WeChat
WOM referral is a non-commercial act of communication about products, brands, or organization [9]. Past research has investigated the antecedents of WOM referral in online settings by showing, for example, that consumers are more likely to spread information about their consumption experiences in the presence of higher levels of satisfaction, affective commitment [10], and perceived quality [11]. In addition to such transactional determinants, past research has also examined other relational and motivational antecedents. For example, regarding the relational factors, existing evidence indicates that consumers are more willing to spread information about their consumption experiences on the Internet when they perceive an opportunity to share such information with close others and believe that such others could be a trustworthy source of information in future occasions [12]. As for the motivational factors, existing evidence suggests that consumers are more likely to spread information about their consumption experiences when they perceive that doing so will allow them to signal their competence or increase their self-image [13] Many studies believe that consumption experiences spreading through WOM can significantly affect recipients' purchase decisions [14,15] or corporate image associations [16]. While past research has examined several aspects of this phenomenon, such as the influences of rewards in recommendation plans [17][18][19], the role of the recommender's professionalism and expertise [20], the effects of loyalty and satisfaction [21], and the role of recommender-recipient relationship strength in determining recipients' purchasing decision and behavior [2,22], little is known about the effects of different WOM recommendation contents. An innovative aspect of the present research is that it contributes to this underdeveloped stream of literature by dividing WOM referrals into two categories based on content. One is economic WOM referral, while the other is public welfare WOM referral. Economic WOM referrals involve product-related aspects and emphasize practical benefits for consumers; in contrast, public welfare WOM referrals involve company-related aspects and emphasize more general benefits for the society at large. Thus, the research empirically investigates the relative influence of these two forms of WOM referral on consumers' purchase intention.
The volume of WOM referrals shared via the Internet has been growing significantly since the advent of social networking sites. In particular, WeChat is among the most popular ones worldwide and the largest messaging app in China. Due to its high transparency, strong sociality, good interaction, and high degree of personalization [23,24], WeChat has become a commonly used channel for information communication, acquisition, recommendation, and sharing between people. So far, Tencent has announced that the combined monthly active users of WeChat have exceeded 1.1 billion, and the total number of WeChat users continues to grow. The "2016 WeChat Key Statistics", published by the official account Curiosity China, shows that WeChat is used frequently, 94% of WeChat users use this platform every day; and WeChat users have many friends, 55.1% have more than 100 friends, 28% have more than 200 friends.
Over the past few years, scholars have conducted more and more studies on WeChat, which regarded, for example, the influencing factors of WeChat adoption from the perspective of communication science [25,26], WeChat communication mechanism [27,28], social impact [29], and functional applications [30,31]. There are also some studies focusing on WeChat marketing [32]. From a marketing perspective, existing studies have found that WeChat marketing is an appropriate strategy for promoting cities [33], and that this platform has brought unlimited business opportunities to the online-to-offline model [34]. Developing friendship with new customers on WeChat can help establish business relationships [35], improve WeChat users' motivation and attitudes [36], and increase perceived usefulness and perceived pleasure [37]. WOM recommendations shared via WeChat can arouse positive emotions with favorable consequences for the recommended products and brands [38]. However, existing marketing research still ignores the potential impact of different types of WOM referral shared via WeChat on consumers' purchase intention. Therefore, this research aims to explore this phenomenon.

Consumer's Self-Construal
Self-construal refers to the tendency of individuals to put themselves in a reference system for self-knowledge when they self-recognize [39]. In social culture, self-construal reflects how individuals perceive the relationship between themselves and others [34] and, particularly, whether they perceive themselves as connected with or separated from others. Markus and Kitayama [39] believe that people may differ in their views of the relationship between themselves and others as a function of their cultural background: People in individualistic cultures regard themselves as independent of others, while those in a collectivistic culture tend to perceive themselves as belonging to a community of interconnected members. This distinguishes two different types of self-construal: one is independent self-construal, which is typical of individualistic cultures (e.g., Western and European countries) and emphasizes the separation of self from others; the other is interdependent self-construal, which is typical of collectivistic cultures (e.g., East countries Sustainability 2021, 13, 645 4 of 18 such as China, Japan and South Korea) and emphasizes the connection between the self and others. Under normal circumstances, the level of individual self-construal is relatively stable. This means that self-construal can be seen as a personal trait formed under longterm social and cultural influence. Yet, in addition to this view, self-construal can also be considered as a malleable construct insofar as its level can significantly vary across situations. Essentially, independent self-construal and interdependent self-construal coexist in the same individual [40,41]; however, because their respective levels can fluctuate as a consequence of situational influences, one of the two types may become dominant in certain situations, generating significant behavioral consequences. For example, when an individualistic perspective ("I") becomes salient in the environment, consumers may become more individualistic, thus more impulsive, emotional, and inclined to consume more hedonic products (e.g., beautifully packaged specialty drinks). Conversely, when a collectivistic perspective ("We") is aroused, consumers may become collectivistic, thus more rational and more inclined to consume utilitarian products (e.g., mineral water to quench thirst) [42]. The existing literature on consumer self-construal has mainly examined how selfconstrual can affect cognitive style [43][44][45], social comparison [46][47][48][49], interpersonal communication [50][51][52][53], individual autonomy [54][55][56], and self-regulation [57][58][59], among other factors. However, there is little research on the relationship between consumer selfconstrual and the type of information received. To contribute to this stream of literature, the present research investigates how the type of self-construal interacts with the type of WOM referral in determining consumers' intention to purchase the recommended product. As an underlying mechanism, the research proposed and tested a dual-system information processing account.

Dual System Information Processing Theory
Metcalfe and Mischel [7] proposed a dual-system response processing theory, according to which individuals can mainly use one of two processing systems to process incoming information, namely: the cold system and the hot system. The cold system is a cognitive processing system, which is also known as the "know system". It is mainly responsible for processing, analyzing, and thinking about information in a complex, fragmented, and cross-temporal manner. The processing speed is low. The hot system is an emotional processing system, also known as the "go system", which is mainly responsible for simple and intuitive emotional response and fast processing of information.
Many studies have found that cognitive and emotional processing systems affect consumer attitudes [60], product evaluation [61], purchase intention [62], decision-making [63,64], and entrepreneur image evaluation [65]. Prior research evidence suggests that people use different information processing systems for different types of information. Complex, professional, data-based, and rhetorical information will cause people to allocate more cognitive resources for analysis, thinking, and processing [66][67][68][69]. Therefore, the higher the complexity of the information, the more likely it is that people will activate the cognitive system and thus invest a greater amount of cognitive resources in thinking and understanding [70]. In contrast, simple, emotional information will likely activate the "go system", which will lead people to produce more emotional processing [71].
Therefore, based on dual-system information processing theory and a literature review, this research proposes the following research framework (see Figure 1).

WOM Referral, Self-Construal, and Consumers' Purchase Intention
Consumers' purchase intention, which preludes their actual purchases, is typically affected by both product-and consumer-related characteristics. Further, purchase intention is also affected by information consumers receive in the form of WOM referrals through the Internet. As discussed above, the present research divides WOM referrals into two categories according to its content: one is economic WOM referral, which involves product-related factors and emphasizes some practical benefit for the individual consumer; while the other is public welfare WOM referral, which involves company-related factors and emphasizes a general benefit for the society at large. Different types of WOM referral may have different effects on consumers with different types of self-construal. More specifically, as self-construal can also vary situationally, it is possible that when consumers are in different self-construal situations, they have different perceptions of the two types of WOM referral and exhibit different reactions in terms of purchase intention.
In marketing practice, companies often use different tactics to persuade consumers to buy their products. Essentially, companies can employ economic tactics, whereby they adopt, for example, low-price or discount promotions [1,2], or public welfare tactics, whereby they emphasize their social responsibility and sustainability [3,72]. However, not all consumers have the same response to economic versus public welfare information stimuli. In this research, we propose that consumers with an independent self-construal are more impulsive, perceptual, emotional, and centered on individual interests; thus, they would be more susceptible to economic information stimulation and tend to consume impulsively and emotionally. For example, increasing the amount of beautifully packaged specialty beverages is more attractive to independent self-construal consumers. In contrast, consumers with an interdependent self-construal are more sensible, weigh the overall interests, and are more concerned about others' interests. Affected by public welfare information, such consumers tend to consume more rationally by preferring utilitarian products or products offered by socially responsible companies. For example, they are more willing to buy mineral water from companies that promise to donate a certain amount of money for public welfare [42]. Therefore, this research proposes the following hypotheses: Hypothesis H1. Consumers with an independent self-construal exhibit a higher purchase intention for the referred product in the presence of an economic WOM referral as compared to a public welfare WOM referral.

WOM Referral, Self-Construal, and Consumers' Purchase Intention
Consumers' purchase intention, which preludes their actual purchases, is typically affected by both product-and consumer-related characteristics. Further, purchase intention is also affected by information consumers receive in the form of WOM referrals through the Internet. As discussed above, the present research divides WOM referrals into two categories according to its content: one is economic WOM referral, which involves productrelated factors and emphasizes some practical benefit for the individual consumer; while the other is public welfare WOM referral, which involves company-related factors and emphasizes a general benefit for the society at large. Different types of WOM referral may have different effects on consumers with different types of self-construal. More specifically, as self-construal can also vary situationally, it is possible that when consumers are in different self-construal situations, they have different perceptions of the two types of WOM referral and exhibit different reactions in terms of purchase intention.
In marketing practice, companies often use different tactics to persuade consumers to buy their products. Essentially, companies can employ economic tactics, whereby they adopt, for example, low-price or discount promotions [1,2], or public welfare tactics, whereby they emphasize their social responsibility and sustainability [3,72]. However, not all consumers have the same response to economic versus public welfare information stimuli. In this research, we propose that consumers with an independent self-construal are more impulsive, perceptual, emotional, and centered on individual interests; thus, they would be more susceptible to economic information stimulation and tend to consume impulsively and emotionally. For example, increasing the amount of beautifully packaged specialty beverages is more attractive to independent self-construal consumers. In contrast, consumers with an interdependent self-construal are more sensible, weigh the overall interests, and are more concerned about others' interests. Affected by public welfare information, such consumers tend to consume more rationally by preferring utilitarian products or products offered by socially responsible companies. For example, they are more willing to buy mineral water from companies that promise to donate a certain amount of money for public welfare [42]. Therefore, this research proposes the following hypotheses: Hypothesis H1. Consumers with an independent self-construal exhibit a higher purchase intention for the referred product in the presence of an economic WOM referral as compared to a public welfare WOM referral.

Hypothesis H2.
Consumers with an interdependent self-construal exhibit a higher purchase intention for the referred product in the presence of a public welfare WOM referral as compared to an economic WOM referral.

The Mediating Role of Consumer Information Processing Style
Existing research shows that people usually use two different systems when processing information: a cognitive system and an emotional system [73], which are also known as the cold system and hot system, respectively [7]. When faced with complex, professional, databased, and rhetorical advertising information, people are more inclined to allocate more cognitive resources for analysis, thinking, and processing [66][67][68][69]; when faced with simple and emotional information, people will tend to engage in more emotional processing [70]. In terms of individual difference in self-construal, this research suggests that people with an independent self-construal are relatively more self-oriented and tend to process information more emotionally, whereas those with an interdependent self-construal are others-oriented and thus more process information more rationally.
Economic WOM referrals contain more perceptual and emotional components and thus are more likely to cause consumers to produce emotional processing [71]. Furthermore, consumers with an independent self-construal are more willing to pay attention to intuitive and emotional information stimuli. Therefore, when consumers are in a context that activates an independent self-construal mode, economic WOM referrals, as opposed to public welfare ones, should be more able to influence their purchase intention. Indeed, in such a context, economic WOM referral, as compared to public welfare WOM referral, should be more able to cause consumers to perform more emotional processing and then to affect their intention to purchase the referred products. In contrast, public welfare WOM referrals are based on rationality and responsibility; thus, they have a higher degree of complexity, which is more likely to cause consumers to activate the cognitive processing system [69,71]. When consumers invest more cognitive resources for information processing, the cognitive processing style is more impactful than the emotional processing one [74]. Furthermore, consumers with an interdependent self-construal are more willing to pay attention to complex and data-based information stimuli. Therefore, when consumers are in a context that activates an interdependent self-construal mode, public welfare WOM referrals, as opposed to economic ones, should be more likely to affect their purchase intention. That is, when consumers are in an interdependent self-construal context, public welfare WOM referrals, as compared to economic ones, will cause consumers to perform more cognitive processing and then to affect their intention to purchase the referred products. Taken together, this research proposes the following hypotheses: Hypothesis H3. For consumers with an independent self-construal, economic (vs. public welfare) WOM referrals can lead to more emotional processing, which in turn increases consumers' intention to purchase the referred products. That is, for these consumers, emotional processing mediates the relationship between the type of WOM referral and purchase intention.
Hypothesis H4. For consumers with an interdependent self-construal, public welfare (vs. economic) WOM referrals can lead to more cognitive processing, which in turn increases consumers' intention to purchase the referred products. That is, for these consumers, cognitive processing mediates the relationship between the type of WOM referral and purchase intention.

Pretest
Before conducting the main studies, we conducted a pre-test with two main purposes. One was to define the content, wording, form, and layout of the questionnaires, while the other purpose was to check whether the WOM referral examples used in the main studies' scenarios can be clearly identified by consumers.
To examine the impact of WOM referral on consumers' purchase intention, we collected relevant examples of WOM referrals; thus, we identified and filtered those consistent with the two proposed types of WOM referral (economic vs. public welfare). These materials mainly come from WeChat Moments. The authors collected 300 WOM referrals published by WeChat friends from January to December 2019 from the WeChat Moments of 20 Zhejiang Normal University undergraduates (upon their consensus). According to the proposed WOM referral classification, information materials were identified and screened. Then, six examples of WOM referrals were selected in total: three for the economic category and three for the public welfare category. To minimize confounding influences, the name and avatar of the information poster, product or brand, and company name were replaced with fictitious names, and the number of likes and comments in the circle of friends, as well as the release time were deleted. Therefore, for the pretest, we designed a set of six WOM referrals of which three were economic WOM referrals (i.e., "To buy food brand A, it cost only 70% of the original price"; "Tea brand B is promoting now, it has 50% price discount"; "Drink brand C is 30% price off now") and three were public welfare WOM referrals (i.e., "Food brand A is a warm brand, 5% percent of sale's income for each consumer's purchasing will be donated to a charity agency"; "Tea brand B is a generous brand, 8% of each sale's income will be donated to local poor areas"; "Drink brand C is a kind brand, 10% income for each sale will be given to support education in underdeveloped areas").
Next, we asked two questions assessed on a binary scale (true/false) to test whether the pretest's participants recognized the economic WOM referrals and the public welfare WOM referrals (i.e., "I think the above WOM referral reflects economic benefits contained in the product (brand)"; "I think the above WOM referral reflects the social benefits contained in the product (brand)"). This short questionnaire was administered to a sample of 86 randomly selected undergraduate students. Before answering the questions, the respondents were asked to indicate whether they had used WeChat. Then, they were asked to classify the six WOM referrals according to whether each WOM referral communicated either an economic benefit or a social benefit. At the end of the pretest, the participants' involvement in the pretest and their background information including gender, age, and major were collected. After eliminating two participants who had never used WeChat, we collected a final sample of 84 questionnaires (45.7% males, 54.3% females). The analysis found that the participants were able to classify the six WOM referrals correctly in 98.5% of cases. Thus, the six WOM referrals could be employed as prototype materials in the main studies to manipulate the type of WOM referral.

Study 1:
The Interaction between WOM Referral and Self-Construal 4.2.1. Experimental Design and Procedure Study 1 adopted a between-subjects experimental design that manipulated the type of WOM referral (economic vs. public welfare). One hundred sixty-eight MBA students (87 males, 81 females) participated in what was presented as a study on students' skills of using WeChat. Participants were randomly assigned to one of the two experimental conditions. They were first asked to read the assigned WeChat WOM referral, which was derived from the pretested materials. The scenario description of the economic WOM referral was as follows: Student A is your WeChat friend. Recently, student A has obtained a good consumption experience by purchasing X brand herbal tea, so student A published a post in Moments to share this consumption experience. A said in the Moments dynamics: "Recently, the summer is hot, and the heat is pressing. In addition, staying up late makes it even more upset and irritable. Since I drank the X brand herbal tea, it relieves the heat and makes me feel refreshing. Buy now and enjoy 15% discount." The description of the public welfare WOM referral was as follows: Student A is your WeChat friend. Recently, student A has obtained a good consumption experience by consuming X brand herbal tea, so student A published a post in Moments to share this consumption experience. A said in the Moments dynamics: "Recently, the summer is hot, and the heat is pressing. In addition, staying up late makes it even more upset and irritable. Since I drank the X brand herbal tea, it relieves the heat and makes me feel refreshing. And for every bottle purchased, 10 cents will be donated to Project Hope." Next, participants were asked to answer three questions that measured their purchase intentions (e.g., "I may buy the brand X herbal tea described by student A"). Answers to these questions were assessed on a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree; Cronbach's α = 0.95). Then, participants completed the self-construal measurement questionnaire adapted from Singelis [75], which included 24 items assessed on a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree). Twelve of these items measured the independent self-construal (α = 0.99), while the other 12 items measured the interdependent self-construal (α = 0.99) (see the Appendix A for a full list of the items).
In the third part, participants were asked to rate the extent to which the assigned WOM referral incorporated either economic or public welfare aspects. To this end, they answered two items using a seven-point Likert scale (i.e., "I think the above WeChat WOM referral reflects the economic benefits contained in the product (brand)"; "I think the above WeChat WOM referral reflects the social benefits contained in the product (brand)"; 1 = strongly disagree, 7 = strongly agree). These two questions served as manipulation check items in the analysis. Finally, participants reported their personal information, including gender, age, and major.

Results
To check the effectiveness of the WOM referral manipulation, two distinct one-way Analyses of Variance (ANOVAs) were conducted on the two manipulation check items. The results confirmed that participants in the economic WOM referral condition rated the assigned WOM referral as more associated with an economic benefit (M = 6.45, SD = 0.65) compared to those in the public welfare WOM referral condition (M =1.68, SD = 0.64), F = 2294.72, p < 0.001. In contrast, participants in the public welfare WOM referral condition rated the assigned WOM referral as more associated with public welfare (M = 6.02, SD = 0.74) compared to those in the economic WOM referral condition (M = 1.88, SD = 0.86), F = 1121.44, p < 0.001.
In line with prior studies on self-construal [76], the scores on the 12 items assessing independent self-construal were averaged to obtain an aggregate measure of this construct, whereas the 12 items assessing interdependent self-construal were averaged to obtain an aggregate measure of this construct. Then, the aggregated measure of interdependent selfconstrual was subtracted from that of independent self-construal. The resulting difference score indicated the dominance of one form of self-construal over the other. Specifically, positive values on this difference score reflected relative independent self-construal, which is typical of respondents with an individualistic orientation, while negative values reflected relative interdependent self-construal, which is typical of collectivistic respondents. Thus, this difference score served as the moderating variable in the statistical analysis. Furthermore, the scores obtained on the three items assessing respondents' purchase intention were averaged to constitute an aggregate measure of this construct.
Next, a regression-based moderation analysis was conducted using Hayes' SPSS PROCESS macro (Model 1; [77]) in which respondents' purchase intention was the dependent variable. This dependent variable was expressed as a function of the type of WOM referral (coded as −1 for public welfare WOM referral and 1 for economic WOM referral), the moderating variable regarding relative independent self-construal, and their interaction. The obtained results showed a significantly positive effect of WOM referral on purchase intention (b = 0.68, p < 0.001), which suggested that economic WOM referral increased respondents' purchase intention. Furthermore, there was a significantly negative relationship between relative independent self-construal and purchase intention (b = −0.10, p < 0.001), suggesting that respondents with a relative interdependent selfconstrual exhibited a greater purchase intention. More interestingly, the results revealed a significantly positive effect of the interaction term (type of WOM referral × relative independent self-construal) on the dependent variable (b = 0.17, p < 0.001). To delve deeper into this interaction effect, the analysis estimated the conditional effect of WOM referral Sustainability 2021, 13, 645 9 of 18 on the dependent variable at higher (M + 1SD) and lower (M − 1SD) levels of the relative independent self-construal variable. In the presence of a higher level of independent self-construal, economic (vs. public welfare) WOM referral was associated with a greater purchase intention (b = 1.48, p < 0.001), thus supporting H1. In the presence of a lower level of independent self-construal (corresponding to an interdependent self-construal), public welfare (vs. economic) WOM referral was associated with a greater purchase intention, although this effect was marginally significant (b = −0.13, p = 0.06), thus supporting H2.

Study 2: The Mediating Role of Information Processing Style
While in Study 1 we assessed self-construal as an individual difference construct, in Study 2 we treated self-construal as a situationally malleable construct. Thus, we manipulated this construct to induce respondents to take either an independent or interdependent self-construal perspective. Furthermore, we tested whether information processing mode plays a mediating role in the process of WOM referral affecting purchase intention.

Experimental Design and Procedure
WOM referral and self-construal were manipulated according to a 2 (type of WOM referral: economic vs. public welfare) × 2 (type of self-construal: independent vs. interdependent) between-subjects experimental design. One hundred seventy-six undergraduate students (76 males, 100 females) participated in the experiment; thus, they were randomly assigned to one of the four experimental conditions.
Participants were preliminarily told that they were partaking in a study on the skills of using WeChat and were asked to answer questions carefully. Therefore, they were exposed to the same WOM referral manipulation as in Study 1. Next, based on the story activation method proposed by Lee, Aaker, and Gardner [58], taking the running competition as a background, participants were asked to read some materials designed to manipulate self-construal (independent vs. interdependent). The event associated with independent self-construal read as follows: You are participating in a running competition and have entered the finals, but your ability is only enough to make it to the finals and it is difficult to achieve further results. But you think: "In any case, I have to prove the value for myself". The event associated with interdependent self-construal read as follows: Your team is participating in a running competition and has entered the finals, but your ability is only enough to make it to the finals and it is difficult to achieve further results. But you think: "In any case, I have to prove the value for my team". Afterward, following the same procedure as in Study 1, participants read the assigned WeChat scenario and answered the same three-item measure of purchase intention (α = 0.96). To check the effectiveness of the self-construal manipulation, participants were asked to rate their independent self-construal and their interdependent self-construal using two different items (i.e., "To what extent do you think you are independent of others?" and "To what extent do you think you are interdependent with your team?", respectively), which were assessed on a seven-point rating scale (1 = not at all, 7 = very much). Next, the participants were asked to recall their feelings when reading the WOM referral and complete a six-item scale assessing the cognitive processing and emotional processing styles. These items were adapted from Kim et al. [70] and McQuarrie and Mick [66] and assessed on a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree). Three items dealt with by the cognitive system (e.g., "This message sent by student A in Moments of Friends gave me a lot of ideas"; Cronbach's α = 0.96), while the other three items dealt with the emotional system (e.g., "This message sent by student A in the circle of friends has caused me a lot of emotions such as moving, happiness, etc."; Cronbach's α = 0.97) (see the Appendix A).
Thereafter, as in Study 1, participants were asked to rate the extent to which the WOM referral they read was associated with an economic benefit or public welfare. Finally, they provided personal information including gender and profession.

Results
Two distinct one-way ANOVAs checked the effectiveness of the implemented manipulations. Regarding the self-construal manipulation, participants in the independent self-construal condition reported higher levels of independent self-construal (M = 5.31, SD = 1.03) compared to those in the interdependent self-construal condition (M = 2.89, SD = 1.16), F = 217.19, p < 0.001. Conversely, the former reported lower levels of interdependent self-construal (M = 3.15, SD = 1.03) compared to the latter (M = 5.02, SD = 1.16), F = 129.00, p < 0.001. Therefore, participants' self-construal was successfully manipulated. Similarly, regarding the WOM referral manipulation, participants in the economic WOM referral condition rated the assigned WOM referral as more associated with an economic benefit (M = 6.13, SD = 0.75) compared to those in the public welfare WOM referral condition (M = 1.78, SD = 0.67), F = 1629.70, p < 0.001. In contrast, participants in the latter condition rated the assigned WOM referral as more associated with public welfare (M = 5.97, SD = 0.79) compared to those in the economic WOM referral condition (M = 1.99, SD = 0.79), F = 1102.12, p < 0.001. Thus, the type of WOM referral was successfully manipulated.
The three items regarding respondents' purchase intention were combined as in Study 1 to obtain an aggregate measure of this construct. Then, a two-way ANOVA expressed purchase intention as a function of type of WOM referral (economic vs. public welfare), type of self-construal (independent vs. interdependent), and their interaction. The results revealed a significant main effect of WOM referral (F = 87.34, p < 0.001) and a significant main effect of self-construal (F = 45.48, p < 0.001). However, these main effects were qualified by a significant interaction between the type of WOM referral and the type of self-construal activated in participants (F = 370.74, p < 0.001), which is summarized in Figure 2. with by the cognitive system (e.g., "This message sent by student A in Moments of Friends gave me a lot of ideas"; Cronbach's α = 0.96), while the other three items dealt with the emotional system (e.g., "This message sent by student A in the circle of friends has caused me a lot of emotions such as moving, happiness, etc."; Cronbach's α = 0.97) (see the Appendix A).
Thereafter, as in Study 1, participants were asked to rate the extent to which the WOM referral they read was associated with an economic benefit or public welfare. Finally, they provided personal information including gender and profession.

Results
Two distinct one-way ANOVAs checked the effectiveness of the implemented manipulations. Regarding the self-construal manipulation, participants in the independent self-construal condition reported higher levels of independent self-construal (M = 5.31, SD = 1.03) compared to those in the interdependent self-construal condition (M = 2.89, SD = 1.16), F = 217.19, p < 0.001. Conversely, the former reported lower levels of interdependent self-construal (M = 3.15, SD = 1.03) compared to the latter (M = 5.02, SD = 1.16), F = 129.00, p < 0.001. Therefore, participants' self-construal was successfully manipulated. Similarly, regarding the WOM referral manipulation, participants in the economic WOM referral condition rated the assigned WOM referral as more associated with an economic benefit (M = 6.13, SD = 0.75) compared to those in the public welfare WOM referral condition (M = 1.78, SD = 0.67), F = 1629.70, p < 0.001. In contrast, participants in the latter condition rated the assigned WOM referral as more associated with public welfare (M = 5.97, SD = 0.79) compared to those in the economic WOM referral condition (M = 1.99, SD = 0.79), F = 1102.12, p < 0.001. Thus, the type of WOM referral was successfully manipulated.
The three items regarding respondents' purchase intention were combined as in Study 1 to obtain an aggregate measure of this construct. Then, a two-way ANOVA expressed purchase intention as a function of type of WOM referral (economic vs. public welfare), type of self-construal (independent vs. interdependent), and their interaction. The results revealed a significant main effect of WOM referral (F = 87.34, p < 0.001) and a significant main effect of self-construal (F = 45.48, p < 0.001). However, these main effects were qualified by a significant interaction between the type of WOM referral and the type of self-construal activated in participants (F = 370.74, p < 0.001), which is summarized in Figure 2. Contrasts revealed that within the independent self-construal condition, participants who read the economic WOM referral exhibited a greater purchase intention (M = 6.19,  Contrasts revealed that within the independent self-construal condition, participants who read the economic WOM referral exhibited a greater purchase intention (M = 6.19, SD = 0.45) compared to those who read the public welfare WOM referral (M = 2.83, SD = 1.22), t = 20.34, p < 0.001, thus supporting H1. Conversely, within the interdependent self-construal condition, participants who read the public welfare WOM referral reported a greater purchase intention (M = 5.88, SD = 0.67) compared to those who read the economic WOM referral (M = 4.72, SD = 0.51), t = − 6.97, p < 0.001, thus supporting H2.
In a subsequent step of the analysis, a moderated-mediation model was estimated to test H3 and H4. This model was estimated by implementing Model 7 of the SPSS PROCESS macro [77,78], with a bootstrap mediation effect with a 95% confidence interval (CI). H3 essentially predicts that, under the independent self-construal condition, the emotional processing style mediates the relationship between the type of WOM referral and purchase intention. In contrast, H4 predicts that, under the interdependent self-construal condition, the cognitive processing style mediates the relationship between the type of WOM referral and purchase intention. Thus, in the estimated model, the type of WOM referral (coded as −1 for public welfare WOM referral and 1 for economic WOM referral) served as the independent variable, whereas the type of self-construal (coded as −1 for the interdependent self-construal condition and 1 for the independent self-construal condition) was the moderator. The scores regarding the three individual items assessing the emotional processing style as well as those regarding the three items assessing the cognitive processing style were averaged to obtain two aggregate measures of the respective constructs. Then, these two measures assessing the emotional and cognitive processing styles, respectively, served as the mediators. Respondents' purchase intention served as the dependent variable. Therefore, this dependent variable was expressed as a function of the two mediators. Each mediator was expressed as a function of the type of WOM referral, the type of self-construal, and their interaction.
The analysis revealed a significantly positive main effect of the type of WOM referral (b = 1.82, p < 0.001) and a significantly positive main effect of the type of self-construal (b = 0.39, p < 0.001) on emotional processing style. However, these main effects were qualified by a significant effect on this mediator exerted by the interaction between the type of WOM referral and the type of self-construal (b = 0.43, p < 0.001). As predicted, estimates of the effects of WOM referral on emotional processing style conditional to the type of self-construal activated in respondents showed that, within the independent self-construal condition, the presence of an economic (vs. public welfare) WOM referral increased respondents' emotional processing (b = 2.25, p < 0.001). However, within the interdependent self-construal condition, there was a similar effect of economic (vs. public welfare) WOM referral on respondents' emotional processing style (b = 1.40, p < 0.001) (see Figure 3). Although the direction of this effect was inconsistent with the theoretical reasoning behind H3 and H4, its magnitude is smaller than that of the effect observed in the independent self-construal condition. This means that the presence of an economic (vs. public welfare) WOM referral increases emotional processing to a greater extent among respondents in the independent self-construal condition than among those in the interdependent self-construal condition.
Regarding the cognitive processing style, the analysis revealed a significantly negative main effect of the type of WOM referral (b = −1.66, p < 0.001) and a significantly negative main effect of the type of self-construal (b = −0.44, p < 0.001) on this mediator. Importantly, these main effects on respondents' cognitive processing style were qualified by a significant effect exerted by the interaction between the type of WOM referral and the type of selfconstrual (b = 0.30, p < 0.001). Estimates of the conditional effects of WOM referral showed that, within the independent self-construal condition, the presence of a public welfare (vs. economic) WOM referral increased respondents' cognitive processing (b = −1.36, p < 0.001). A similar effect also emerged among respondents in the interdependent self-construal condition. However, within the latter condition, such an effect was larger in magnitude (b = −1.95, p < 0.001), suggesting that the presence of a public welfare (vs. economic) WOM referral increased cognitive processing to a greater extent among respondents in the interdependent self-construal condition than among those in the independent self-construal condition (see Figure 3). Sustainability 2021, 13, x FOR PEER REVIEW 12 of 19 respondents in the interdependent self-construal condition than among those in the independent self-construal condition (see Figure 3). Next, the analysis estimated the relationship between each of the two mediators and purchase intention, controlling for the effect on this dependent variable exerted by WOM referral. The results showed that both emotional processing (b = 0.76, p < 0.001) and cognitive processing (b = 1.28, p < 0.001) increased respondents' purchase intention. Importantly, the analysis revealed two moderated mediation effects: one was an indirect effect of the interaction term (type of WOM referral × type of self-construal) on purchase intention via the emotional processing style, which was positive and significant (Index of moderated mediation: b = 0.65, 95% CI = 0.49, 0.83); the other was a significantly positive indirect effect of the same interaction term on purchase intention via the cognitive processing style (Index of moderated mediation: b = 0.76, 95% CI = 0.49, 1.04). Estimates of the conditional indirect effects clarified the nature of these two moderated mediation effects. Such estimates revealed that an economic (vs. public welfare) WOM referral led to a more emotional processing and thus to a greater purchase intention among respondents in both self-construal conditions. However, the indirect effect of economic (vs. public welfare) WOM referral on purchase intention via the emotional processing style was greater among respondents in the independent self-construal condition (b = 1.72, 95% CI = 1.35, 2.12) than among those in the interdependent self-construal condition (b = 1.07, 95% CI = Next, the analysis estimated the relationship between each of the two mediators and purchase intention, controlling for the effect on this dependent variable exerted by WOM referral. The results showed that both emotional processing (b = 0.76, p < 0.001) and cognitive processing (b = 1.28, p < 0.001) increased respondents' purchase intention. Importantly, the analysis revealed two moderated mediation effects: one was an indirect effect of the interaction term (type of WOM referral × type of self-construal) on purchase intention via the emotional processing style, which was positive and significant (Index of moderated mediation: b = 0.65, 95% CI = 0.49, 0.83); the other was a significantly positive indirect effect of the same interaction term on purchase intention via the cognitive processing style (Index of moderated mediation: b = 0.76, 95% CI = 0.49, 1.04). Estimates of the conditional indirect effects clarified the nature of these two moderated mediation effects. Such estimates revealed that an economic (vs. public welfare) WOM referral led to a more emotional processing and thus to a greater purchase intention among respondents in both self-construal conditions. However, the indirect effect of economic (vs. public welfare) WOM referral on purchase intention via the emotional processing style was greater among respondents in the independent self-construal condition (b = 1.72, 95% CI = 1.35, 2.12) than among those in the interdependent self-construal condition (b = 1.07, 95% CI = 0.82, 1.37). Indeed, the difference between these two indirect effects reached statistical significance (Contrast = 0.65, 95% CI = 0.49, 0.83). Regarding the mediating role of the cognitive processing style, the analysis revealed that a public welfare (vs. economic) WOM referral increased cognitive processing and thus purchase intention among respondents in both self-construal conditions. However, the indirect effect of a public welfare (vs. economic) WOM referral on purchase intention via the cognitive processing style was greater among respondents in the interdependent self-construal condition (b = −2.51, 95% CI = −2.90, −2.13) than among those in the independent self-construal condition (b = −1.74, 95% CI = −2.02, −1.48), with a difference between these two indirect effects that reached significance (Contrast = 0.76, 95% CI = 0.48, 1.04).
Overall, albeit not perfectly in line with H3 and H4, these results are consistent with the proposed theorizing as they showed that, for consumers with an independent selfconstrual, an economic (vs. public welfare) WOM referral is more effective at increasing such consumers' purchase intention because of a greater ability of this type of WOM referral to increase their emotional processing. In contract, for consumers with an interdependent self-construal, a public welfare (vs. economic) WOM referral is more effective due to its greater ability to increase cognitive processing.

Discussion
This research examines how different types of WOM referral disseminated in the WeChat platform can affect the purchase intentions of consumers with different types of self-construal. Based on the dual-system information processing theory and across two experimental studies that employed different measurement methods, we showed that the type of WOM referral (economic vs. public welfare) and the type of self-construal (independent vs. interdependent) interact to affect consumers' purchase intention. Specifically, consistent with H1 and H2, our two studies showed that consumers with an independent (vs. interdependent) self-construal exhibit a greater purchase intention for the referred product in the presence of an economic (vs. public welfare) WOM referral. In contrast, consumers with an interdependent (vs. independent) self-construal exhibit a greater purchase intention in the presence of a public welfare (vs. economic) WOM referral. Furthermore, Study 2 results showed that regardless of their nature (economic vs. public welfare), WOM referrals can activate both processing modes although to a different extent. Albeit not perfectly in line with H3 and H4, these results seem logical and realistic, as WOM messages typically comprise a multitude of cues, some of which are emotional while others are more rational. However, economic (vs. public welfare) WOM referrals mainly activate the emotional processing mode and, through this, increase consumers' purchase intention for the referred product. As predicted, this effect was found to be significantly stronger when an independent (vs. interdependent) self-construal is salient among consumers. Indeed, it is plausible that when consumers take an independent self-construal perspective, thus becoming more self-oriented, the presence of economic WOM referrals, which emphasize a product's economic benefits, generates a sense of excitement and emotional pleasure in consumers for the personal benefits they might get. This reasoning finds support in prior marketing literature [79] and might explain the effect we observed for this type of WOM referrals.
Conversely, public welfare (vs. economic) WOM referrals mainly activate the cognitive processing mode, which in turn increases consumers' purchase intention for the referred product. This effect was found to be significantly stronger when an interdependent (vs. independent) self-construal is salient. This effect was consistent with the idea that when consumers take an interdependent self-construal perspective and thus become more others-oriented, the presence of public welfare WOM referrals, which emphasize a company's good deeds for the society at large, induces such consumers to rationalize the potential societal consequences of such deeds. This pattern of results is compatible with our theoretical reasoning and provides several implications.

Theoretical Implications
Our findings carry three main theoretical implications. First, they enrich the literature on WOM recommendation. Prior literature mainly studies the influencing factors of WOM referral and its impact on consumer attitudes, WOM re-spreading, purchase intention, and purchase behavior. Our research explores WOM in the WeChat context and divides WOM referrals into two categories, namely economic and public welfare WOM referral, with the latter being particularly relevant to those companies that want to promote their sustainability and social responsibility. In doing so, our research enriches and expands the theoretical literature on the content and role of WOM referral and brand sustainability.
Second, our findings elucidate the moderating role of consumers' self-construal in the relationship between the type of WOM referral and purchase intention. According to self-construal theory, consumer's self-construal can take either the form of an independent self-construal or that of an interdependent self-construal. Study 1 assessed self-construal as an individual characteristic, while Study 2 manipulated self-construal as a situational factor. Despite such a difference in the assessment method, both studies provided consistent evidence for the moderating role of self-construal, thus conferring robustness to our findings and expanding the current state of knowledge on the role of self-construal in WOM and, particularly, in determining the effectiveness of WOM referrals.
Third, our findings reveal an intermediary mechanism by which the interaction between the type of WOM referral and the type of self-construal affects consumers' purchase intention. Our research builds on the theory of dual-system information processing to propose the emotional and cognitive processing styles as relevant mediators of WOM referral's effects. By showing the differential mediating roles of these two variables, our research deepens the theoretical understanding of the influence of WOM referral on consumer psychology and behavior.

Managerial Implications
From a managerial perspective, our findings offer practical insights for companies. Since companies, influencers, or other brand promoters often use WeChat to communicate with consumers, it is important for these subjects to consider the psychological characteristics of their audiences. Essentially, our findings suggest that messages that underline the economic benefits of a product or brand are more appealing to consumers with an independent self-construal, while messages that emphasize a company's sustainability or social responsibility are more effective among consumers with an interdependent self-construal. Interestingly, Study 2 results showed that self-construal can be seen as a malleable construct, which can be manipulated situationally. This means that companies interested in promoting the economic benefits related to their products (e.g., price discounts) could increase the effectiveness of their communications by making an independent self-construal more salient in the environment. For example, this may be possible by associating their core messages with other communication elements that arouse a sense of uniqueness and differentiation in consumers. On the other hand, companies interested in promoting their efforts in the areas of sustainability and social responsibility (e.g., charitable donations) could increase the effectiveness of their communications by increasing the salience of an interdependent self-construal. For example, they could combine their messages with other marketing elements that arouse a sense of social connection, belongingness, and community in consumers.

Limitations and Future Research Directions
Our research has some limitations that offer opportunities for future studies. First, it only considered university students, who are relatively young (aged 20-40). Therefore, our findings are not easily generalizable to other consumer segments. Future studies could test our hypotheses on consumers belonging to other age groups. Similarly, our findings derive from two studies conducted in China using WeChat. Future studies could replicate them in other cultural backgrounds (e.g., U.S.) and using other social networking sites (e.g., Facebook). Second, our second study manipulated self-construal using a scenario in a laboratory setting. Future studies could employ other manipulations of this construct by using field research methods to increase the external validity of our findings. Furthermore, future studies could explore the role of other potential moderators, such as crisis event scenarios, corporate (brand) awareness, or other consumer characteristics. Third, our research neglected the potential interplay between chronic and primed responses and considered survey-based measures. Future studies could employ more sophisticated measurement methods based, for example, on neuroscientific techniques [80], and data management approaches based, for example, on data partitioning [81].

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.