Understanding Consumers’ Purchase Intentions in Social Commerce through Social Capital: Evidence from SEM and fsQCA

Social commerce has become a fiercely competitive environment. Understanding consumers’ purchase intentions can help social commerce platforms retain and attract more consumers. Social capital is one of the primary resources that plays a critical role in facilitating consumers’ purchase intentions in social commerce. Here, complex relationships between different dimensions of social capital are further clarified and its impact on consumers’ purchase intentions are discussed. Based on a survey of 302 social commerce users, this study utilizes an SEM and fsQCA approach to validate the effect of social capital on consumers’ purchase intentions. The SEM results reveal that the effect of structural social capital on consumers’ purchase intentions is fully mediated by relational and cognitive social capital. The fsQCA results confirm the significance of consumers’ social capital as determinants and provide the configurations that can lead to high purchase intentions. Though previous studies have discussed the factors influencing consumers’ purchase intentions, this study takes the first step toward enhancing the understanding of the configurations that link dimensions of social capital to consumers’ purchase intentions in s-commerce using fsQCA approach.


Introduction
Nowadays, online transaction has shifted from electronic commerce (e-commerce) to social commerce (s-commerce), which emphasized social interaction and interpersonal communication [1][2][3]. Traditional websites also added social functions to encourage content generation and improve consumers' experience. Online business process in s-commerce was facilitated by social media, especially social networks sites [4]. Consumers created user-generated content, shared their purchasing, and spread product-related information with their friends in social network sites [5][6][7]. Usually, consumers' purchase decisions in s-commerce relied on peer consumers' shopping experience, eWOM, and recommendations [8,9]. Social features were foundations of s-commerce because online purchase was facilitated by the social relationships among peer consumers [10]. Therefore, social capital was one of the primary resources that played a critical role in facilitating consumers' purchase intentions in an s-commerce context [11,12].
Social capital reflects valuable resources embedded in individuals' relational networks [13]. It consists of three dimensions, that is, structural social capital (SSC), relational social capital (RSC), and cognitive social capital (CSC). Social capital can influence knowledge sharing and purchasing [14,15]. However, the conclusions regarding the effect of different dimensions of social capital on the outcome variables were not consistent. For example, CSC, manifested by a shared language, was positively related to knowledge sharing in some studies [16], while other studies argued that a shared language had no effect on knowledge sharing [12]. What is more, some studies considered the different dimensions as parallel elements [17][18][19][20], while other studies proposed a causal relationship between the different dimensions [21][22][23]. Hsu and Hung (2013) put forward the inter-action model of three dimensions of social capital [24]. Because of inconsistent research results, specific research on the relationship between different social capital dimensions is needed. The effect of social capital dimensions on the dependent variables also needs to be made explicit.
Therefore, this study aimed to further clarify the complex relationships between different dimensions of social capital and make clear its impact on consumers' purchase intentions utilizing the method of structural equation modeling (SEM) and fuzzy-set qualitative comparative analysis (fsQCA). Ragin (2008) proposed the fsQCA approach, which can provide combinations of conditions to improve results [25]. The integration of SEM and fsQCA method has been applied in the information system studies [26,27]. Although the factors influencing consumers' purchase intentions in s-commerce have been discussed, this study enhanced the understanding of the configurations that link dimensions of social capital to consumers' purchase intentions in s-commerce using fsQCA approach. This research provided a new methodology to use in s-commerce research, and utilizing this novel analytical technique was critical to our research.

S-Commerce
Facilitated by information and communication technologies (ICTs), s-commerce enabled consumers to create user-generated content, share their purchasing, and spread product-related information with their friends in social network sites [2,5,28,29]. The socialtechnical features of s-commerce make it a favorable platform to share information and knowledge [30]. This consumer-generated content can benefit online purchase decisionmaking [31,32], trust building [10], relationship quality and brand loyalty [4], and brand co-creation [8,11].
One stream of previous studies focused on information sharing and eWOM behavior in s-commerce [33][34][35]. These studies have shown that extrinsic motivation, especially external motivation and identified motivation [34], utilitarian motivation, hedonic motivation and social motivation [36], community commitment and trust [10,37], passion [38], and perceived social distance [35], were significant predictors of information sharing and eWOM behavior in s-commerce. Recent research showed that the effect of information sharing activities on intention to buy was mediated by trust in platform and perceived risk [2]. Furthermore, effect of perceived participation risk on post comments in s-commerce was positively moderated by social identity [39].
Another stream of previous studies has explored the factors influencing consumers' purchasing in s-commerce [31,40,41]. Interpersonal contact and relationships in social networks can generate commercial opportunities, such as improving sales [30,42]. For example, it was found that guanxi elements, including ganqing, renqing, xinren [43], communication and social interaction [1,32,40], consumer experience [44], social desire, and commercial desire [41], were positively related to purchase intention. In addition, trust is a critical predictor of purchase intention as well [45][46][47]. Furthermore, it was found that the impact of trust on purchase intention was mediated only by positive valence, and the mediation effect of negative valence and social media content were insignificant [48]. Purpose of sociability, policies of sociability, and people of sociability were positively related to trust in product recommendations [49], which can in turn increase the likelihood of purchase in s-commerce [44]. Specifically, it was found that perceived risk was negatively related to purchase intention, and identity positively moderated the negative effect of perceived risk on purchase intention [39]. It was also found that design quality and website features were significant predictors of purchase behavior in s-commerce [50,51].

Social Capital
Social capital reflects valuable resources embedded in individuals' relational networks [13]. It is embedded in social relationships and can facilitate resource exchanges. Diversified and informed social connection was the access to favorable resources [52,53].These resources can benefit a series of activities ranging from technology transfer [54], employee performances and satisfaction [55,56], contract duration [57], information disclosure [58], eWOM and online purchase [23,59,60], information sharing and knowledge exchange [18,19,61], and popularity of user-generated content [22], to crowdfunding [62].
Social capital consists three dimensions, that is structural, relational, and cognitive dimension [13]. SSC refers to the structure of actors' relations, which can be manifested as social interaction. It represents the pattern and frequency of connections between individuals or organizations [17]. Social network ties can provide the access to the information and resources that are embedded in social relationships [63,64], and actors in the critical position of the social network, such as occupying the position of structural hole, have priority to access such resources [65]. RSC reflects the quality of exchanged relationship, such as trust [66] and norm of reciprocity [67]. RSC can be cultivated through the history of interactions between individuals [68]. CSC reflects the resources that enable collective goals and shared rules among parties, which are manifested as shared language [13].
The first stream of studies considered the different dimensions of social capital as parallel elements [17][18][19][20]. SSC was found to be a positive predictor of quantity of knowledge sharing [67,69]. Similarly, it was found that the effect of SSC on tacit knowledge creation was significant [19]. However, a recent research showed the effect of SSC on perceived knowledge quality was insignificant [18]. Variables related to RSC were found to have different influence on outcome variables as well. For example, trust was found to be positive predictor of knowledge quality in virtual communities but to have no influence on quantity of knowledge sharing [67]. Trust and identification were found to have positive effect on perceived knowledge quality, whereas the effect of reciprocity on perceived knowledge quality was insignificant [18]. Variables related to CSC were found to have different influence on outcome variables as well. For example, the effect of CSC on member's perception of knowledge quality [18] and tacit knowledge sharing [19] were positive. Shared language was positive driver of knowledge quality in virtual communities but had no influence on quantity of knowledge sharing [67]. In another study, shared language was found to have positive effect on both quantity and quality of knowledge sharing [17].
The second stream of studies proposed the causal relationship between the different dimensions of social capital [21][22][23]. First, SSC was found to be positive predictors of RSC and CSC [12,22,56]. Specifically, it was found that tie strength was positively related to trust and shared language, which is in turn positively related to eWOM sharing [12]. Familiarity was positively related to trust and perceived similarity, which was in turn positively related to sense of belonging [70]. Structural linkage was positively related to relational linkage and cognitive linkage [71]. However, there existed exceptions, for example, SSC, manifested by social interaction, was found to have no effect on RSC, manifested by trust [16]. Moreover, CSC was found to be a positive predictor of RSC [21,22,72]. Specifically, shared vision and shared language were positively related to trust [12,16]. The positive effect of CSC on RSC reached a consensus.
The potential third stream of studies was put forward by Hsu and Hung (2013) [24], in which the interaction model of three dimensions of social capital was proposed. They found that the effect of paired dimensions, including the interaction effect of SSC and RSC, the interaction effect of SSC and CSC, and interaction effect of RSC and CSC on process performance are significant but in different ways. However, only the interaction effects of SSC and RSC on product performance were observed.

RSC: Trust in Peers and Reciprocity
RSC reflects the quality of exchanged relationship, including trust [66] and norm of reciprocity [67]. Trust is often cultivated through frequent interactions between individuals. In an online marketing context, trust in peers is the essential driving factor of consumers' purchase intentions [73]. Reciprocity refers to perceived fairness regarding the time and effort spent in resource exchange [67]. If there exists strong sense of reciprocity, that is, if individuals consider resource exchange to be reciprocal, they are more willing to engage more [17]. In an online marketing context, reciprocity can represent the quality of relationships, which can influence individuals' behaviors [74]. First, RSC can be viewed as a kind of governance mechanism, reducing uncertainty and perceived risk in embedded relationships [67,74,75]. Specifically, RSC can reduce the opportunism and contradictions in social relationships [76] and enhance commitment to social relationships [77]. In addition, when there exists high level of RSC between two individuals, they are more willing to engage in resources exchange [12,78] and then enhance consumers' intention to buy [73]. Thirdly, RSC is positively linked to consumers' attitudes [79] and consumers' satisfaction [80], which can in turn positively influence consumers' intention to buy.

CSC: Shared Language
CSC reflects the resources that enable collective goals and shared rules among parties, manifested as shared language [13]. In an online marketing context, shared language is one of the essential driving factor of consumers' purchase intentions. First, shared language can not only improve the common understanding [19] and shared perceptions among individuals but also reduce each individual's cognitive barriers [56]. When individuals shared their shopping experience and recommendation products among peers with strong cognitive capital, they could easily understand each other without ambiguity, which in turn facilitated their subsequent purchase decisions. Second, communication efficiency can be improved using shared language, specifically shared vocabulary [12,17]; thus, the individuals can experience a better resource exchange process. Sometimes, consumers may be confronted with the problem of information overload when they search for eWOM in s-commerce platform [81]. When cognitive capital is strong, the problem of information overload can be relieved and consumers can absorb useful information efficiently and effectively. Thirdly, CSC is positively related to peers' perception of knowledge quality [18]. When CSC is strong, consumers' perceptions of recommendation quality are high. Therefore, they are willing to take high-quality recommendations, and purchase, in the s-commerce platform.

Structural Social Capital: Social Interaction
SSC reflects the structure of actors' relations, which was manifested as social interaction. Social interaction can facilitate information transfer and recourse exchange [19,57]. According to social embeddedness theory, actor's position in the social network results in his/her ability to acquire resources [82]. Social interaction is one of the essential driving factor of consumers' purchase intentions [83]. On one hand, SSC is directly related to consumers' purchase intentions. Due to the fact that actors in the central position in the social network can utilize personal contacts to gain specific resources and curial information [84], consumers that participate more in social interaction and social exchange can get sufficient production recommendations more easily. These product recommendations can in turn help consumers to make better purchase decisions, and increase their purchase intentions in s-commerce. One the other hand, the effect of SSC on consumers' purchase intentions is partially mediated by RSC and CSC. SSC represents the frequency of exchanges and communications among consumers. High level of SSC enhanced the possibility that more consumers will generate, share, and utilize the product recommendation from peers [68]. Consumers with high level of social interaction will have frequent mutual communication, product-related information exchange, and shopping experience sharing. These interpersonal interactions can cultivate reciprocity [21], trust [12,23], and shared language [16,21], which in turn increase consumers' purchase intentions.

Hypothesis 3A (H3A):
The effect of SSC on consumers' purchase intentions is partially mediated by RSC.

Hypothesis 3B (H3B):
The effect of SSC on consumers' purchase intentions is partially mediated by CSC.
The conceptual model is presented in Figure 1.
possibility that more consumers will generate, share, and utilize the product recommendation from peers [68]. Consumers with high level of social interaction will have frequent mutual communication, product-related information exchange, and shopping experience sharing. These interpersonal interactions can cultivate reciprocity [21], trust [12,23], and shared language [16,21], which in turn increase consumers' purchase intentions.

Hypothesis 3A (H3A): The effect of SSC on consumers' purchase intentions is partially mediated by RSC.
Hypothesis 3B (H3B): The effect of SSC on consumers' purchase intentions is partially mediated by CSC.
The conceptual model is presented in Figure 1.

Data Collection
WeChat is a popular mobile instant message app in China. Specially, WeChat users can share their daily life with friends in the WeChat moment and make online transactions in WeChat store. WeChat store is a typical s-commerce platform, in which online purchase is facilitated by the social interaction and recommendation of WeChat friends. This study employed a sampling service by a professional company. A survey link was provided to the potential users of the WeChat store, and then the participants gave feedback through the online survey directly and anonymously. The data were collected in 2018. Finally, a total of 302 valid questionnaires was collected. Majority of the respondents were female (51.99%), aged from 20 to 30 years old (59.27%), and had a university degree (50.99%). 72.19% of the respondents had used WeChat for more than three years. The maximum variance inflation factor (VIF) was 2.28, indicating there was no multicollinearity. Harman's single factor test showed that the first factor explained 38.11% of the variance. Then, common method factor was examined in the PLS model (see Appendix A). The average squared values of the method factor loadings (0.381) were smaller than the average squared loadings of substantive constructs (0.625), suggesting that common method biases are not an issue in this study. Moreover, a t-test of early and late questionnaires showed that the t-values are nonsignificant, so non-response bias is not an issue either.

Measurements
All the items used in this study were adapted from prior studies. Measures of social interaction and shared language were adapted from Chiu et al. (2006) [67]. Measures of trust in peers were adapted from Chang and Chuang (2011) [17]. Measures of reciprocity were adapted from Pai and Tsai (2016) [85]. Measures of purchase intentions were

Data Collection
WeChat is a popular mobile instant message app in China. Specially, WeChat users can share their daily life with friends in the WeChat moment and make online transactions in WeChat store. WeChat store is a typical s-commerce platform, in which online purchase is facilitated by the social interaction and recommendation of WeChat friends. This study employed a sampling service by a professional company. A survey link was provided to the potential users of the WeChat store, and then the participants gave feedback through the online survey directly and anonymously. The data were collected in 2018. Finally, a total of 302 valid questionnaires was collected. Majority of the respondents were female (51.99%), aged from 20 to 30 years old (59.27%), and had a university degree (50.99%). 72.19% of the respondents had used WeChat for more than three years. The maximum variance inflation factor (VIF) was 2.28, indicating there was no multicollinearity. Harman's single factor test showed that the first factor explained 38.11% of the variance. Then, common method factor was examined in the PLS model (see Appendix A). The average squared values of the method factor loadings (0.381) were smaller than the average squared loadings of substantive constructs (0.625), suggesting that common method biases are not an issue in this study. Moreover, a t-test of early and late questionnaires showed that the t-values are nonsignificant, so non-response bias is not an issue either.

Measurements
All the items used in this study were adapted from prior studies. Measures of social interaction and shared language were adapted from Chiu et al. (2006) [67]. Measures of trust in peers were adapted from Chang and Chuang (2011) [17]. Measures of reciprocity were adapted from Pai and Tsai (2016) [85]. Measures of purchase intentions were adapted from Liang et al. (2011) [86]. A five-point Likert scale, ranging from "strongly disagree-1" to "strongly agree-5", was used to measure all the items. Measurements was shown in Supplementary Materials.

Analytical Methods
This study utilized SEM and fsQCA method to analysis the data. The fsQCA method put forward a way to find out the different configurations of causal conditions that may result in the same outcome [25]. Traditional statistical methods focused on the net effect of independent variables on dependent variables and the results showed that the different causal path can result in the same outcome, whereas the fsQCA method took a step forward to clarify the configurations of conditions that may result in an outcome, more importantly, the presence and absence of the outcome, respectively, may be caused by different reasons [87,88].

Calibration
The fsQCA method requires the calibration of all condition and outcome variables [25]. Calibration relies on theoretical and substantial knowledge to generate a fuzzy-set score that relates to the degree of membership in a set [88]. In this study, to produce the fuzzy-set scores, calibration involves the use of "the direct method" [25]. Since the variables were measured using a five-point Likert scale, the original value of 5 was set as full membership, the original value of 3 was set as cross over point, and the original value of 1 was set as full non-membership.

Results of Measurement Model
Reliability, convergent validity, and discriminant validity were examined [89]. Composite reliability (CR) and average variance extracted (AVE) were assessed to make sure the reliability. As shown in Table 1, Cronbach's alphas of purchase intention being lower than 0.7 is a shortcoming in this study. However, several published papers also suffered from this issue, in which the alpha scores below 0.7 are acceptable. For example, in Kim et al. (2020)'s study, the reliability coefficients were 0.684 for bonding social capital [90]. CR values ranging from 0.825 to 0.865 exceeded the cutoff values of 0.7, and AVE ranging from 0.612 to 0.638 exceeded the cutoff values of 0.5, demonstrating good construct reliability. Convergent validity was examined by checking the item loadings on their respective constructs. All the loadings exceeded the required values of 0.7 (See Table 1). To examine discriminant validity of the measurement model, we compared the square root of the AVE values with inter-construct correlation coefficients. As shown in Table 2, the square root of the AVEs for each construct ranging from 0.782 to 0.799 was greater than those for all the inter-construct correlations ranging from 0.335 to 0.667. Furthermore, HTMT ratio shown in Table 3 falls below the threshold of 0.9, with two exceptions [91].
First, the SEM results demonstrated that SSC was positively related to both RSC and CSC, and CSC was positively related to RSC. These findings were consistent with previous studies, such as Tsai and Ghoshal (1998)'s [66] and Wang et al. (2016)'s [12] studies, suggesting that the three dimensions of social capital were correlated with each other, and this causal relationship still exists in a s-commerce context. SSC was a critical predictor of RSC and CSC, and CSC was a critical predictor of RSC.
Moreover, of all three social capital dimensions, only RSC had a positive direct impact on purchase intention. Trust in peers and reciprocity influenced the consumers' purchase intentions in s-commerce. Consumers are willing to accept the advice and recommendations form the reliable and reciprocal peers in s-commerce. This result agreed with previous studies, for example, Luo et al.(2020) found that trust was positively related to transaction intention in e-commerce [23]. JTAER 2021, 16  First, the SEM results demonstrated that SSC was positively related to both RSC and CSC, and CSC was positively related to RSC. These findings were consistent with previous studies, such as Tsai and Ghoshal (1998)'s [66] and Wang et al. (2016)'s [12] studies, suggesting that the three dimensions of social capital were correlated with each other, and this causal relationship still exists in a s-commerce context. SSC was a critical predictor of RSC and CSC, and CSC was a critical predictor of RSC.
Moreover, of all three social capital dimensions, only RSC had a positive direct impact on purchase intention. Trust in peers and reciprocity influenced the consumers' purchase intentions in s-commerce. Consumers are willing to accept the advice and recommendations form the reliable and reciprocal peers in s-commerce. This result agreed with previous studies, for example, Luo et al.(2020) found that trust was positively related to transaction intention in e-commerce [23].
Contrary to expectations, the direct effect of CSC on purchase intention was insignificant. One explanation was that the likelihood of consumers' purchase intentions in scommerce depends on trust and reciprocal relationships regardless of the shared language between them. Purchasing online may be perceived to be a risky behavior, resulting in shared language being a less critical factor directly predicting purchase intention in scommerce. Similarly, Wang et al.(2016) found that shared language had no significant effect on eWOM on social networking sites [12].
Thirdly, the results revealed that the effect of SSC on purchase intentions was fully mediated through RSC and CSC. In other words, social interaction had no direct influence on consumers' purchase intentions in s-commerce. A possible explanation is that only the history of the social interaction cultivated a good relationships among peers and can influence the consumers' purchase intentions. Similarly,  found that the effect of SSC on loyalty to seller was fully meditated by RSC and CSC [21].

Results from fsQCA
First, this study conducted the necessary conditions analysis. If the consistency score exceeded the cutoff of 0.9, a condition or a combination of conditions were considered as "necessary" [88]. Table 4 shows the results of necessary conditions analysis for the presence of purchase intentions. It is found that social interaction, reciprocity, and shared language are necessary conditions for the purchase intentions in s-commerce. Contrary to expectations, the direct effect of CSC on purchase intention was insignificant. One explanation was that the likelihood of consumers' purchase intentions in s-commerce depends on trust and reciprocal relationships regardless of the shared language between them. Purchasing online may be perceived to be a risky behavior, resulting in shared language being a less critical factor directly predicting purchase intention in s-commerce. Similarly, Wang et al.(2016) found that shared language had no significant effect on eWOM on social networking sites [12].
Thirdly, the results revealed that the effect of SSC on purchase intentions was fully mediated through RSC and CSC. In other words, social interaction had no direct influence on consumers' purchase intentions in s-commerce. A possible explanation is that only the history of the social interaction cultivated a good relationships among peers and can influence the consumers' purchase intentions. Similarly,  found that the effect of SSC on loyalty to seller was fully meditated by RSC and CSC [21].

Results from fsQCA
First, this study conducted the necessary conditions analysis. If the consistency score exceeded the cutoff of 0.9, a condition or a combination of conditions were considered as "necessary" [88]. Table 4 shows the results of necessary conditions analysis for the presence of purchase intentions. It is found that social interaction, reciprocity, and shared language are necessary conditions for the purchase intentions in s-commerce. Next, as shown in Table 5, the first configuration indicates that the absence of reciprocity and the presence of shared language led to the presence of purchase intentions. The second configuration indicates that the presence of trust in peers and reciprocity led to the presence of purchase intentions. The third configuration indicates that the presence of social interaction and trust in peers led to the presence of purchase intentions. The first configuration indicates that the absence of reciprocity and the presence of shared language are the sufficient conditions for high purchase intentions. When reciprocity is absent, shared language is very critical for generating purchase intentions. Reciprocity reflects perceived fairness of resource exchange [67]. When reciprocity is absent, resource exchange may be perceived as non-reciprocal. Thus, shared language plays an important role in improving mutual understanding among members. This mutual understanding can dispel misgivings when adopting recommendations in s-commerce, thus improving purchase intentions. This finding is in line with previous studies highlighting the role of shared language in enhancing mutual understanding [21].
The second configuration indicates that the presence of trust in peers and reciprocity are the sufficient conditions for purchase intentions. When consumers maintain reciprocal and trustful relationships, they will be more willing to purchase. Some of previous studies emphasized the foundation role of SSC (e.g., social interaction) linked to RSC [56]. However, this study found another explanation, that is, the presence or absence of social interaction is insignificant, and reciprocal and trustful relationships can lead to purchase intentions.
The third configuration indicates that the presence of social interaction and trust in peers is a sufficient condition for purchase intentions. The effect of trust on outcome variables was inconsistent in previous studies [17,67]. This study found another explanation, that is, the presence of social interaction and trust in peers is a sufficient condition leading to purchase intentions. Frequent communication with peers and trustful relationship between peers together led to increased purchasing in s-commerce.
In conclusion, the SEM results revealed that shared language had no direct effect on consumers' purchase intentions, whereas shared language was positively related to RSC, which can in turn improve consumers' purchase intentions. The fsQCA findings further confirmed the critical role of shared language, especially when reciprocity is absent, leading to consumers' purchase intentions. In addition, the SEM results revealed that RSC was positively related to consumers' purchase intentions and the fsQCA findings further confirmed the important role of RSC as well, that is, that presence of trust in peers and reciprocity are the sufficient conditions for purchase intentions. Furthermore, the SEM results revealed that the effect of SSC on purchase intentions was fully mediated through RSC and CSC and the fsQCA findings further confirmed that the presence of social interaction and trust in peers is a sufficient condition for purchase intentions.

Theoretical Implications
This study has two important theoretical contributions. First, this study shed light on some insightful implications for social capital theory. Previous studies regarding social capital had some inconsistent research results [12,16]. To make further explicit the effect of social capital in cultivating purchase intentions, this study explored the effect of different dimensions of social capital on purchase intentions utilizing SEM and fsQCA method. The results demonstrated the heterogeneous role of social capital dimensions, that is, social interaction, trust in peers, reciprocity, and shared language, in promoting purchase intentions. Specifically, the SEM results presented the effect of causal relationship of social capital on purchase intentions. The fsQCA results suggested three configurations of causal conditions for high purchase intentions. Although the factors influencing consumers' purchase intentions in s-commerce have been discussed, this study enhanced the understanding of the configurations that link dimensions of social capital to consumers' purchase intentions in s-commerce using fsQCA approach.
Moreover, the current study utilizing SEM and fsQCA methods provides innovative analytic techniques to investigate the influence of different dimensions of social capital on purchase intentions in s-commerce. To make clear the complex relationships between different dimensions of social capital, the study explored the configurations of different dimensions of social capital on purchase intentions. In contrast to similar studies, which only considered the different dimensions of social capital either as parallel elements or causal relationship, this study showed the configurations of sufficient conditions for purchase intentions. SEM is a traditional method to explore the effect of social capital on purchase intentions. The fsQCA approach allows the current study to explore the complex configurations of the social capital dimensions that lead to purchase intentions. Therefore, this study contributes to the methods used in s-commerce research.

Managerial Implications
This study also has some managerial contributions. The results indicated that scommerce platform managers should encourage peers develop strong social capital. Especially, the SEM results revealed that of all three social capital dimensions, only RSC had a positive direct impact on purchase intentions, and the effect of SSC on purchase intentions was fully mediated through RSC and CSC. The fsQCA findings suggested that the absence of reciprocity and the presence of shared language, the presence of trust in peers and reciprocity, and the presence of social interaction and trust in peers are all sufficient conditions for purchase intentions.
Therefore, only enhancing one dimension of social capital is not sufficient for leading to purchase intentions. Structural, relational, and cognitive social capital should be maintained carefully. First, the SEM results revealed that shared language had no direct effect on consumers' purchase intentions, whereas shared language was positively related to RSC, which can in turn improve consumers' purchase intentions. The fsQCA findings further confirmed the critical role of shared language, especially when reciprocity is absent, leading to consumers' purchase intentions. S-commerce managers should establish shared values, shared vision, shared narration, and shared language in the s-commerce community. For example, abbreviations with special meaning are encouraged for use by the consumers in the s-commerce community. By doing so, consumers can gain cohesion power, reduce social distance, and gain a sense of belonging to the s-commerce community.
Moreover, the SEM results revealed that RSC was positively related to consumers' purchase intentions and the fsQCA findings further confirmed the important role of RSC as well, that is, the presence of trust in peers and reciprocity are the sufficient conditions for purchase intentions. S-commerce managers are encouraged to establish rules and regulations to guarantee the transaction environment in s-commerce. Especially, the scommerce platform should eradicate fake products, false advertising, and misleading recommendations. Opportunism should be punished by the s-commerce platform. S-commerce managers should encourage consumers to be honest and genuine to build a reciprocal atmosphere as well.
Thirdly, the SEM results revealed that the effect of SSC on purchase intentions was fully mediated through RSC and CSC, and the fsQCA findings further confirmed that the presence of social interaction and trust in peers are the sufficient conditions for purchase intentions. Interpersonal interaction and communication should be encouraged to improve SSC. For example, s-commerce platform can encourage active consumers through spiritual rewards, such as membership rank and badges, and monetary rewards, such as coupons and credits.

Limitations
There are two limitations. First, only three different dimensions of social capital were considered as antecedent variables. Other components, such as social support, may be other explanations for solutions. Future work should consider more factors that can enrich the configurations. Moreover, this study only explored the factors influencing the presence of purchase intentions. Further work could investigate the configurations leading to the absence of purchase intentions, that is, which factors might prevent consumers from purchasing in s-commerce. Reasons for purchasing or not purchasing in s-commerce may vary.