Trust Transfer in Sharing Accommodation: The Moderating Role of Privacy Concerns

: Sharing accommodation (SA) has gained rapid growth in the last decade. To offer better service to users, the platform and hosts have to extensively collect and utilize conﬁdential user data and information. With the extensive collection and utilization of personal user information, there are potential problems of data abuse and leakage, which makes users’ privacy concerns an important and unavoidable issue for repeated purchases and the sustainable development of SA. Privacy concerns are thus not only an important antecedent of purchase behaviors, but also an important conditional variable that will have impacts on the formation of trust and user purchase behaviors. However, the moderating effect of privacy concerns on trust formation has rarely been examined in the SA literature. To ﬁll this knowledge gap, drawing on trust transfer theory and trust literature, this study builds a theoretical model to examine the relationships of three types of institution-based trust (i.e., trust in the SA platform, trust in the user community, and trust in the host community) and their effects on continuous use intention. Moreover, this study explores the moderating effect of privacy concerns on institution-based trust transfer in the context of SA. We then collected data through a questionnaire survey from experienced users of two reputable SA platforms in China, and empirically tested the research model with 470 valid responses. The results show that trust in the user community positively affects trust in the SA platform and trust in the host community; trust in the SA platform and trust in the host community positively affect users’ continuous use intention. Meanwhile, privacy concerns negatively moderate the relationship between trust in the user community and trust in the SA platform, as well as the relationship between trust in the user community and trust in the host community. The ﬁndings conﬁrm the moderating role of privacy concerns in the trust transfer process, complementing existing research on trust transfer theory and trust.


Introduction
The development of emerging technologies such as the Internet and platform-based enterprises has given birth to the sharing economy, which has developed into a huge commercial economy in recent years. In the last decade, the sharing economy has developed rapidly around the world, reaching billions of dollars in scale and penetrating many industries, including accommodation, catering, transportation, and manufacturing. With the emergence of the sharing economy, shared tourism resources have significantly changed the way tourists travel, making sharing accommodation (SA) an important form of accommodation for tourists. SA uses a two-way transaction platform based on the Internet to match the SA need between hosts and tourists, meeting tourists' non-standardized needs for accommodation and benefiting the platforms, hosts, and tourists in the sharing process [1]. In recent years, SA has given full play to its platform advantages and become an important driving force for the digital transformation of the economy while meeting people's daily accommodation needs.
The rapid development of SA has resulted in the accumulation of a large number of experienced users. The continuous use behaviors of existing users are important to maintain the sustainable development of SA. However, its rapid development has also brought serious problems and challenges such as information privacy leakage and personal safety injuries. For example, in May 2019, a guest of Airbnb in Qingdao was secretly photographed by the host, and the candid videos and photos were sold online [2]. In March 2021, Urban Report reported that a couple was secretly photographed by the host for eight hours during their homestay, and the candid videos were also sold online [3]. Worries on privacy leakage will affect users' perceptions of the SA and their willingness to continuously use, which will ultimately jeopardize the healthy development of the industry. Therefore, how to promote the continuous participation of existing users under the condition of privacy concerns has become an important problem that urgently needs to be solved [4]. This issue has also attracted great attention in the latest literature on SA.
Trust is considered to be one of the important factors to promote the continuous use intention of users [5]. Trust helps to maintain a long-term connection between the trusting and the trusted. When transactions are opportunistic and uncertain, trust can promote risk-taking behavior [6]. Existing research pointed out that in the platformbased economy, individual trust is not enough to facilitate user engagement behavior, and in such circumstances institution-based trust plays a more important role [7,8]. For example, Tussyadiah et al. [5] pointed out that trust in the platform, as the institutionbased trust, can significantly promote user participation in SA. Pavlou et al. [7] believed that trust in the community of sellers, as a type of institution-based trust, can enhance users' willingness to trade. In addition, some studies have emphasized the importance of trust in the user community to users' continuous use intention [9]. However, very few studies have comprehensively considered the transfer relationship of these institutionbased trusts and their effects on continuous use intentions. Trust transfer, as an effective mechanism to build and enhance trust, has been explored in different research contexts [10]. Based on prior research of institution-based trust [7,8], there are at least three types of institution-based trust in the context of SA in terms of the trusting objects (e.g., platform and users), named trust in the SA platform, trust in the host community and trust in the user community. According to Pavlou et al. [7] and Lu et al. [8], trust in the SA platform and trust in the host community are institution-based trusts [7]. Similarly, trust in the user community, which represents a general opinion about the user community, is also a type of institution-based trust. In online context, users tend to rely on their peer users' actions and recommendations [11] to form their perceptions and judgments towards their transactional parties (e.g., platform or hosts). Thus, trust in the user community, as a type of perception towards the peer users, should help to engender trust towards the other two trusting objects, i.e., the platform and the host community. Drawing on trust transfer theory, this study proposes trust in the user community as a key antecedent of both trust in the platform and trust in the host community, and examines how trust is transferred from the user community to both the platform and the host community.
In addition, most studies ignore the boundary conditions under which trust transfer occurs. This will impede us to gain a deep understanding of how trusts are developed and take effect in the SA. Van der Heijden et al. [12] also pointed out that when the trust reaches a certain level of evaluation, it no longer promotes people's willingness to buy online. In recent years, scholars have called for further exploration of the boundary conditions where trust functions. Existing research mainly emphasizes the important role of institutional mechanisms as contextual variables for trust to be transferred and to take effect. For instance, Chen et al. [13] explored the moderating effects of institutional effectiveness and perceived website quality on trust transfer. However, the institutional mechanism mainly considers the impact at the macro level, ignoring users' situational awareness at the internal psychological level. Trust transfer is closely related to the risk context, and it does not occur independently of users' inherent perception of risk, such as privacy concerns. Privacy concerns and trust are two negatively correlated constructs, but they are not antithetical. Individuals can perceive little trust and little privacy concern or have high trust but still perceive high privacy concern at the same time [14]. This has given rise to the call to examine the moderation role of privacy concerns in online contexts [15]. Despite the importance of privacy concerns, the literature rarely addresses its impact on trust transfer. This research gap is not helpful to gain an insightful understanding of whether and how trust transfer occurs in different risk scenarios in the context of SA.
Given the above research gap, this study first considers the transfer relationship among trust in the SA platform, trust in the host community, and trust in the user community based upon trust transfer theory. Second, this study applies and extends the privacy calculus model, taking privacy concerns as a contextual variable of trust transfer and exploring the extent to which it affects how trust in the SA platform and trust in the host community are transferred from trust in the user community. Finally, this study constructs a theoretical model including institution-based trusts, privacy concerns, and continuous use intention and employs empirical research methods to test this model. Through empirical testing, this study finds that trust in the user community positively affects trust in the SA platform and trust in the host community, trust in the SA platform and trust in the host community positively affect continuous use intention, and privacy concerns negatively moderate the impact of trust in the user community on trust in the SA platform and trust in the host community.
The main contributions of this study to related research are as follows: first, this paper comprehensively considers and studies the relationships of three types of institution-based trust in the SA and expands the research on trust transfer. Second, this paper reveals the effects of trust in the SA platform and trust in the host community on continuous use intention, which suggests the importance of institution-based trust on user behavioral intentions. Third, from the perspective of users, this paper examines the role of privacy concerns as a contextual variable of trust transfer and expands the research on the boundary conditions of trust transfer by revealing the moderating effect of privacy concerns in the SA. Fourth, this paper studies the interaction of trust and risk beliefs, thus extending the privacy calculus model by uncovering the moderating effect of privacy concerns on trust transfer.

Institutional Trust and Sharing Accommodation
Trust refers to the principal's subjective beliefs that the trustee can meet the principal's expectations [16]. During the transaction process, trust helps to reduce the risk and uncertainty between the transaction parties. Existing research shows that in an environment with high risk and uncertainty, a lack of trust will lead to the failure of both parties' transactions [17]. In the SA, interpersonal trust is difficult to form because almost every time users do not trade with a specific host but with an unknown housing provider in the host community. In this case, institution-based trust becomes particularly important. Institution-based trust is a special trust-generating mechanism that is associated with formal social structures and is not limited to specific transactions and specific parties [8]. Existing research has confirmed the importance of institution-based trust in various online contexts [7,8,18].
Research on institution-based trust is widely used in online environments such as e-commerce and the sharing economy. For example, Pavlou et al. [7] proposed in the online auction market that trust in the intermediary market and trust in the community of sellers are two institution-based trusts that can significantly promote transactions. Lu et al. [8] also paid attention to institution-based trust at the market level in the field of e-commerce and confirmed that trust in the marketplace positively affects transaction intention. Mao et al. [18] studied the influence of trust in the platform in the context of online accommodation, and revealed the importance of trust in the SA platform on travelers' continuous use intention. Therefore, the importance of institution-based trust has been confirmed in different studies. However, most of the existing research only focuses on trust in the SA platform and fails to consider the impacts of other types of institution-based trusts. In the platform-based SA, there are at least two other institution-based trusts: trust in the host community, and trust in the user community. Trust in the host community comes from the relationship between users and the host community. We define it as the subjective belief of users that a host from the host community will perform transaction obligations as agreed upon [18]. In a given market, users can form an overall view of the community of sellers in that market by judging whether multiple individual sellers can be trusted [7]. Trust in the community of sellers is thus a specific type of institution-based trust, and can influence user behavior in e-commerce [19]. The same logic can be applied to the platform-based SA context. When users participate in SA, each transaction object is almost a different unknown individual, and it is thus difficult to form interpersonal trust toward a specific host, which seriously hinders the formation of transactions. In this case, trust in the host community, as institution-based trust, will influence continuous use intention through users' general expectations [18].
Furthermore, in the sharing economy, users rely heavily on the collective comments and acts of other users to make decisions [20]. Similar to trust in the host community, trust in the user community is another type of institution-based trust that is important but has been overlooked in the literature. We define trust in the user community as the degree to which users are willing to rely on the comments, actions, and decisions of their peer users on the same platform [21]. In e-commerce, Benlian et al. [22] pointed out that peer users' comments will elicit users' emotional responses, which help them form trusting beliefs with online sellers. Söllner et al. [21] revealed that users' trust in the community of information system has a positive impact on their trust towards the information system. Similarly, we expect that users tend to rely on actions and recommendations of their peer users to form perceptions and make decisions in the SA. Existing surveys show that users with a high level of trust in their peers can form their trusting beliefs towards transactional objects without the need to access other information such as personal experience [11]. For example, Xiao et al. [9] pointed out that trust in the user community positively affects trust in the focal merchant. Therefore, trust in the user community can be an important source to generate trust in the platform and trust in the host community, and it is of great significance in the SA and deserves further exploration.

Trust Transfer
Trust transfer theory argues that user's trust can be transferred from an entity to an unknown entity that is associated with the entity [23]. The trust transfer process generally involves three actors, namely, the principal, the trustee, and the third party. The principal is the person who makes evaluations on whether to trust others; the trustee is the person who is evaluated by the principal based on her trustworthiness; the third party acts as a middleman. The basic logic among them is that when the principal trusts the third party and there is a close connection between the trustee and the third party, the trust of the principal towards the third party can be transferred to the trustee. Therefore, in the SA, the trustees (e.g., platforms and hosts) will adopt various strategies to promote the generation of trust from the principals (i.e., users). Trust transfer can build trust in the network environment, which has been confirmed in the research on trust. For example, Chen et al. [13] pointed out that trust in the SA platform positively affects trust in sellers in the e-commerce. Zhao et al. [24] proposed that trust in the specific seller positively affects brand trust in the context of mobile commerce. However, the existing research on the antecedents of trust mainly paid attention to the technological and institutional factors, and few studies explore the antecedents of institution-based trust from the perspective of trust transfer. Based on previous studies, we argue that institution-based trust can be transferred within the relationship networks including the SA users the peer user community, the host community, and the platform. Therefore, based on trust transfer theory, this research will explore how trust in the user community can enhance both trust in the SA platform and trust in the host community.
The literature also points out that it is very important to explore the boundary conditions of trust transfer. Chen et al. [13] explored the moderating effects of institutional effectiveness and perceived website quality on trust transfer and pointed out that insti- tutional effectiveness negatively regulated the relationship between trust in the platform and trust in sellers, while perceived website quality positively moderated the relationship between trust in the platform and trust in sellers. Xiao et al. [9] proposed in the O2O context that the effectiveness of institutional mechanisms strengthens the relationship between trust in the platform, trust in the focal merchant and trust in the user community.
It can be seen that trust transfer and its boundary conditions are still the current research hotspots. Nevertheless, the existing research on boundary conditions of trust mainly focuses on the influence of institutional mechanisms, and few studies are focusing on the role of users' negative perceptions, thus ignoring the important role of users as decision-making individuals when making online transactions. In the context of SA, users' subjective consciousness determines their behaviors. The privacy calculus model argues that users' decisions are influenced by a combination of their benefit beliefs (such as trust) and risk beliefs [25]. Therefore, perceptions related to risk should be an important boundary condition for trust transfer. Privacy concerns are a relatively important type of risk belief in the SA context. From the perspective of the user, this study treats privacy concerns as an important boundary condition of trust transfer and explores how it affects the impacts of trust in the user community on both trust in the SA platform and trust in the host community.

Privacy Concerns and Sharing Accommodation
The right to privacy is legally defined as an individual's right to control the flow of personal information [26]. Online privacy concerns reflect how user information is collected and used in an online environment [27] Privacy concerns are one of the most widely used variables in the study of information systems. In e-commerce, Dinev et al. [25] proposed that privacy concerns refer to users' concerns about opportunistic behaviors related to personal information submitted to the Internet. Choi et al. [28] further argued that privacy concerns are users' concerns about the possible negative effects of the misuse of the information. Moreover, some scholars have developed the measurement of privacy concerns. For instance, Smith et al. [29] developed the concern for information privacy (CFIP) scale. The CFIP identifies four dimensions of privacy concerns: collection, error, secondary use, and unauthorized access, and it is used to measure the individual's concern about an organization's information privacy practices. Malhotra et al. [30] proposed the Internet users' information privacy concerns (IUIPC) scale for the Internet context. The IUIPC consists of three dimensions: control, cognition, and collection. In the SA, users need to submit their information and data to the platform, and the platform relies on users' information to match service providers with users. The extensive collection of user information and the threat of privacy breaches have exacerbated users' concerns about privacy, which have also gradually increased users' awareness of privacy concerns. In this research, we define privacy concerns as users' concerns about negative consequences such as misuse of information due to the inability or unwillingness of platforms and hosts to protect their personal information.
Research on privacy concerns mainly focuses on its antecedents and its direct effects on behaviors such as information disclosure and repurchase. Some scholars have studied the moderating effect of privacy concerns. For example, Smith et al. [31] proposed a model named APCO (Antecedents → Privacy Concerns → Outcomes). Antecedent variables of privacy concerns may include privacy experiences, privacy awareness, personality differences, demographic differences, and culture. In the sharing economy, Lutz et al. [32] found that users' online privacy concerns are affected by privacy guarantees. Some scholars also pointed out that the interaction between personal connection and perceived audience size would affect privacy concerns in C2C [33].
The impact of privacy concerns on user behavior is found to be inconsistent. Some studies have shown that privacy concerns could weaken Internet users' willingness to provide personal information [25]. But, other studies (e.g., Joinson et al. [34]) found that although there is high concern about information privacy, users still choose to submit information to obtain services. This has triggered scholars to study the moderating effect of privacy concerns. For instance, in the online environment, privacy concerns were found to weaken the effect of Internet trust on online purchasing attitudes [35]. However, existing research mostly focuses on the direct role of privacy concerns, ignoring their interaction with trust. In the SA, users would generate concerns about the personal information which may be misused, secondary used, or unprotected by related parties (such as platforms and hosts). Privacy concerns should interact with trust and, in turn, will affect how trust is transferred [35]. In the SA, users are more concerned about privacy violations from platforms and hosts compared with other aspects of personal privacy violations (such as hacking and tracking) because platforms and hosts hold a large amount of data and use them frequently [36] Therefore, users' privacy concerns will moderate the effect of trust in the user community on both trust in the SA platform and trust in the host community. Based on the above discussion, this study explores the moderating effect of privacy concerns in the trust transfer process.

Research Model and Hypothesis Development
Based on existing research on institutional trust and privacy concerns, this study constructs a conceptual model (see Figure 1) to examine the impact of trust in the user community on trust in the SA platform and trust in the host community, which in turn affects users' continuous use intention. In addition, the conceptual model also introduces privacy concerns as a moderating variable to examine the impact of privacy concerns in the trust transfer process. The model also incorporates control variables known to affect continuous use intention, including trust disposition, age, gender, and income. Below, we will explain each assumption in detail.
scholars also pointed out that the interaction between personal connection and perceived audience size would affect privacy concerns in C2C [33].
The impact of privacy concerns on user behavior is found to be inconsistent. Some studies have shown that privacy concerns could weaken Internet users' willingness to provide personal information [25]. But, other studies (e.g., Joinson et al. [34]) found that although there is high concern about information privacy, users still choose to submit information to obtain services. This has triggered scholars to study the moderating effect of privacy concerns. For instance, in the online environment, privacy concerns were found to weaken the effect of Internet trust on online purchasing attitudes [35]. However, existing research mostly focuses on the direct role of privacy concerns, ignoring their interaction with trust. In the SA, users would generate concerns about the personal information which may be misused, secondary used, or unprotected by related parties (such as platforms and hosts). Privacy concerns should interact with trust and, in turn, will affect how trust is transferred [35]. In the SA, users are more concerned about privacy violations from platforms and hosts compared with other aspects of personal privacy violations (such as hacking and tracking) because platforms and hosts hold a large amount of data and use them frequently [36] Therefore, users' privacy concerns will moderate the effect of trust in the user community on both trust in the SA platform and trust in the host community. Based on the above discussion, this study explores the moderating effect of privacy concerns in the trust transfer process.

Research Model and Hypothesis Development
Based on existing research on institutional trust and privacy concerns, this study constructs a conceptual model (see Figure 1) to examine the impact of trust in the user community on trust in the SA platform and trust in the host community, which in turn affects users' continuous use intention. In addition, the conceptual model also introduces privacy concerns as a moderating variable to examine the impact of privacy concerns in the trust transfer process. The model also incorporates control variables known to affect continuous use intention, including trust disposition, age, gender, and income. Below, we will explain each assumption in detail.

Trust Transfer
In an online marketplace, users not only seek information but also share their knowledge and experiences with others on the platform [37]. According to trust transfer theory, an individual's trust can be transited from a familiar entity to an unfamiliar entity because of the association between these two entities [23]. Users rely heavily on the reviews, recommendations, and behaviors of other users to infer the trustworthiness of unknown entities [38]. Therefore, we argue that trust in the user community can be transferred to users' trust in the SA platform. If users believe that the user community is

Trust Transfer
In an online marketplace, users not only seek information but also share their knowledge and experiences with others on the platform [37]. According to trust transfer theory, an individual's trust can be transited from a familiar entity to an unfamiliar entity because of the association between these two entities [23]. Users rely heavily on the reviews, recommendations, and behaviors of other users to infer the trustworthiness of unknown entities [38]. Therefore, we argue that trust in the user community can be transferred to users' trust in the SA platform. If users believe that the user community is credible, they will think the SA platform with which these users engage is also trustworthy. That is, users' trust in the user community can be transferred to users' trust in the SA platform. In addition, some studies have proposed that trust in the user community positively affects trust in the focal merchants [9]. The host community will try their best to obtain positive comments and recommendations from the user community; therefore, the more users trust the user community, the greater they will trust the host community. Accordingly, we have the following two hypotheses: H1. Trust in the user community positively affects trust in the SA platform.
H2. Trust in the user community positively affects trust in the host community.

Institutional Trust and Continuous Use Intention
Tussyadiah et al. [5] pointed out that trust can exist between users and platforms as well as users and hosts in the SA. Trust in the SA platform refers to the degree of users' recognition of the platform's integrity, benevolence, and capability [39]. Here, we use continuous use intention as an outcome variable for trust. Continuous use intention reflects the willingness of users to purchase products or use services again [40], which can be used as a proxy variable for users' continued use behavior [41]. According to the extended value framework, trust directly affects users' purchase intention [42]. Specifically, participating in the SA has various potential risks, and the platform will implement effective institutional mechanisms to limit the opportunistic behavior of the hosts [43]. Therefore, trust in the platform can help mitigate risks and uncertainties in the transaction process [19]. If users have a high level of trust in the SA platform, they are more likely to transact on the platform. Previous studies have also found that trust in the platform positively affects users' continuous use intention [18,44]. For instance, Lee et al. [45] proposed that trust in the platform is a significant predictor of users' intention to participate in Uber. Mao et al. [18] found that trust-in-platform has a positive effect on behavioral intention. When users trust a platform, they are more willing to transact on that platform. Accordingly, we make the following hypothesis: H3. Trust in the SA platform positively affects users' continuous use intention.
In the context full of uncertainty and risk, trusting the transactional parties is the premise of participating in online transactions. In online marketplaces, community trust, as a type of trust towards the whole seller community, is the major determinant of interactions between individuals and community members [7]. Community trust helps to overcome social uncertainty and the risks posed by unknown members of the community, thereby facilitating transactions between users and community members [7]. Existing research also shows that user trust in sellers can enhance transaction willingness in the e-commerce market [19]. Therefore, in the SA context, we argue that trust in the host community, as a type of institutional trust, plays a key positive role in reducing uncertainty and thus can facilitate users' repeated transactions. Accordingly, we make the following hypothesis: H4. Trust in the host community positively affects users' continuous use intention.

Moderating Effects of Privacy Concerns
Privacy concerns refer to users' concerns about negative consequences such as misuse of information due to the inability or unwillingness of platforms and hosts to protect their personal information. Generally, users need to offer their personal information and data to the platform and hosts to complete their online transactions. Due to the difference in personal characteristics and experiences, people have various levels of concern for their personal information [46]. When users have high levels of privacy concerns, the connection between the two trusting subjects may not be sufficient to facilitate the trust transfer. In addition, existing studies have revealed the moderating role of privacy concerns. Tan et al. [47] found that in social networks, for users with high levels of privacy concerns, perceived ease of use has a weaker effect on their behavioral intentions. Some scholars have proposed that privacy concerns will significantly enhance the relationship between trust and users' continued use of mobile health apps [48]. Gan et al. [49] proposed that users' privacy concerns about mobile map apps significantly moderate the impact of perceived product trust on continuous use intention.
As mentioned earlier, users' trust in the user community can be transferred to trust in the SA platform and trust in the host community because users perceive the connection between the user community and other parties (e.g., the platform and hosts). This link may be weakened when users' privacy concerns are heightened. Both platforms and hosts may use personal information in unexpected ways, or the information is not secured by them. Users with high levels of concern about information privacy will seek assurances from other parties [46] to facilitate trust transfer among these parties. This means that trust in the user community will not be sufficient to engender trust in the SA platform and trust in the host community. Conversely, users with low levels of concern about information privacy do not need to seek assurance from other parties except the user community. In other words, trust in the user community by itself will help to develop both trust in the SA platform and trust in the host community. Therefore, we argue that privacy concerns act as a moderating factor in the relationship between trust in the user community and trust in the SA platform as well as the relationship between trust in the user community and trust in the host community. Accordingly, we make the following two hypotheses: H5. Privacy concerns negatively moderate the relationship between trust in the user community and trust in the SA platform.
H6. Privacy concerns negatively moderate the relationship between trust in the user community and trust in the host community.

Measurement Development
To ensure the reliability and validity of the scale, all measurement items are derived from mature scales of existing research (see Appendix A). We adopted three items from Yang et al. [50] to measure trust in the SA platform. The measurement items of continuous use intention and trust disposition were all from Gefen and Straub [51]. Three measurement items of trust in the host community were from Pavlou and Gefen [7], and three items of trust in the user community were from Söllner et al. [21]. Three items from Dinev and Hart [25] were taken to measure privacy concerns. At the same time, all items were adapted and modified in the SA context. All measurement items were evaluated on standard Likert-type scales which ranged from 1 = "strongly disagree," to 7 = "strongly agree." After the preliminary questionnaire was developed, we invited two experts in the field of information systems to give feedback on the questionnaire structure, sentence expression, etc. We then revised the words of some items according to their opinions. Next, 112 experienced users were invited to conduct a pretest over our questionnaire. We then finalized the questionnaire based on the pretest. The measurement scales of each construct are shown in Appendix A.

Sample and Data Collection
We select the experienced users of the two reputable SA platforms (Meituan Homestay and Tujia) in China to collect data. According to a recent report on the SA industry [52] Formal surveys still use online questionnaires for data collection. The survey was distributed on sojump.com (accessed on 12 May 2021), a professional data collection website in China. Respondents were asked to fill out a questionnaire based on their recent SA experience on the above two platforms. To ensure respondents had experience with the SA, the survey included a screening question asking them if they had any SA experience recently.
Out of the 693 questionnaires received, 470 valid samples were obtained. Table 1 shows the demographic information of the valid sample. Among them, female users account for 40%, and male users account for 60%; most of them are aged 18-30, accounting for 74.3%; the education level is mainly junior college and undergraduate, accounting for 85.3%. The Research Report on the Development of China's Homestay Industry (2019) [53] shows that young users aged 29 and under account for 64% and the users aged 39 and under account for 91.6%. Therefore, the users of SA are mainly born after 1995 and 2000, and the results of our questionnaire have certain generalizability.

Data Analysis
The partial least squares (PLS) model was adopted to analyze the collected data. There are several reasons to choose PLS. First, PLS is a particularly effective method for testing causal models based on empirical data, and thus, it is well suited for this study. Second, PLS can simultaneously estimate the measurement model, structural model, and moderating effects. Therefore, we use SmartPLS 3.0 for data analysis.
To assess whether there is serious common method bias, this paper adopts the following two methods: Harman's single-factor test [54], and latent variable analysis. First, we performed a non-rotated factor analysis on all measurement items in SPSS to check whether the first principal component accounted for more than 50% of the total variance. The variance explained by the largest factor was 41.98% for principal component factoring, which is under the critical value of the 50% cutoff. Second, the correlation matrix in Table 2 does not present any highly correlated factors, whereas evidence of common method bias should have resulted in extremely high correlations (r > 0.90) [55]. These two tests indicate that our data does not suffer from high common method bias.

Measurement Model
This study first assessed the reliability and validity of the measurement model. Reliability is used to describe the internal consistency of each factor. Composite reliability (CR) and Cronbach's alpha are important indicators to evaluate reliability. In the exploratory study, Cronbach's alpha is required to be above 0.6, and the combined reliability is required to be above 0.7. As shown in Table 3, the combined reliability and Cronbach's alpha of all constructs are above the minimum criterion of 0.7, indicating good internal consistency of the measured variables. The validity test is divided into two parts: convergent validity and discriminant validity. Convergent validity can be assessed by examining the standardized factor loadings and the average variance extracted (AVE) of each measurement index on its corresponding latent variable. As shown in Table 3, the factor loadings of all items are greater than 0.7, and the AVE scores are greater than 0.5, indicating that this measurement scale has satisfactory convergent validity. Discriminant validity refers to whether there is a low correlation and significant difference between latent variables. This study used three methods to assess discriminant validity. First, as shown in Table 2, the square root of the AVE for each construct was greater than the correlations between that construct and others. Second, Henseler et al. [56] proposed that the heterotrait-monotrait ratio of correlations (HTMT) can better test discriminant validity. As shown in Table 4, the HTMT values between all constructs are less than the 0.85 criterion. Finally, the study continues to use SmartPLS for cross-loading analysis. The results are shown in Table 5, and the loading terms of each variable are higher than the cross-loadings on other variables. In conclusion, these results suggest that our study has good discriminant validity.

Structural Model
Hypothesis testing of the structural equation model was performed by using Smart-PLS 3.0, and the bootstrapping method with 2000 repetitions was used to estimate the significance of all paths. The results are shown in Figure 2. The results show that 33.0% of the variance explained in trust in the SA platform, 41.2% in trust in the host community, and 55.0% in continuous use intention are explained by the model.

Structural Model
Hypothesis testing of the structural equation model was performed by using SmartPLS 3.0, and the bootstrapping method with 2000 repetitions was used to estimate the significance of all paths. The results are shown in Figure 2. The results show that 33.0% of the variance explained in trust in the SA platform, 41.2% in trust in the host community, and 55.0% in continuous use intention are explained by the model. As shown in Figure 2, trust in the user community has a significant positive impact on trust in the SA platform (β = 0.615, p < 0.001) and trust in the host community (β = 0.543, p < 0.001), which supports H1 and H2. Both trust in the SA platform (β = 0.336, p < 0.001) and trust in the host community (β = 0.373, p < 0.001) are found to have a positive impact on continuous use intention. Hence, H3 and H4 are supported. Among the control variables, trust disposition (β = 0.145, p < 0.01) positively affects users' continuous use intention.

Structural Model
Hypothesis testing of the structural equation model was performed by using SmartPLS 3.0, and the bootstrapping method with 2000 repetitions was used to estimate the significance of all paths. The results are shown in Figure 2. The results show that 33.0% of the variance explained in trust in the SA platform, 41.2% in trust in the host community, and 55.0% in continuous use intention are explained by the model. As shown in Figure 2, trust in the user community has a significant positive impact on trust in the SA platform (β = 0.615, p < 0.001) and trust in the host community (β = 0.543, p < 0.001), which supports H1 and H2. Both trust in the SA platform (β = 0.336, p < 0.001) and trust in the host community (β = 0.373, p < 0.001) are found to have a positive impact on continuous use intention. Hence, H3 and H4 are supported. Among the control variables, trust disposition (β = 0.145, p < 0.01) positively affects users' continuous use Accept,

Structural Model
Hypothesis testing of the structural equation model was performed by using SmartPLS 3.0, and the bootstrapping method with 2000 repetitions was used to estimate the significance of all paths. The results are shown in Figure 2. The results show that 33.0% of the variance explained in trust in the SA platform, 41.2% in trust in the host community, and 55.0% in continuous use intention are explained by the model. As shown in Figure 2, trust in the user community has a significant positive impact on trust in the SA platform (β = 0.615, p < 0.001) and trust in the host community (β = 0.543, p < 0.001), which supports H1 and H2. Both trust in the SA platform (β = 0.336, p < 0.001) and trust in the host community (β = 0.373, p < 0.001) are found to have a positive impact on continuous use intention. Hence, H3 and H4 are supported. Among the control variables, trust disposition (β = 0.145, p < 0.01) positively affects users' continuous use Reject. As shown in Figure 2, trust in the user community has a significant positive impact on trust in the SA platform (β = 0.615, p < 0.001) and trust in the host community (β = 0.543, p < 0.001), which supports H1 and H2. Both trust in the SA platform (β = 0.336, p < 0.001) and trust in the host community (β = 0.373, p < 0.001) are found to have a positive impact on continuous use intention. Hence, H3 and H4 are supported. Among the control variables, trust disposition (β = 0.145, p < 0.01) positively affects users' continuous use intention.
To test the moderating effect, this study uses the product indicator method proposed by Chin et al. [57]. The method suggests that interaction terms can be created by multiplying the indicators of the predictor and moderator constructs. The results show that privacy concerns significantly and negatively moderate the relationship between trust in the user community and trust in the SA platform (β = −0.089, p < 0.001). Hence, H5 is supported. Figure 3 shows a graph of the two-way interaction. Simple slope analysis suggests that at low levels of privacy concerns (Mean − 1SD), trust in the SA platform increases rapidly when trust in the user community increases. However, at high levels of privacy concerns (Mean + 1SD), trust in the SA platform increases slightly as trust in the user community increases. This result further supports H5. Hypothesis 6 states that the positive effect of trust in the user community on continuous use intention can be weakened by privacy concerns. As expected, privacy concerns significantly and negatively moderate the relationship between trust in the user community and trust in the host community (β = −0.074, p < 0.001), supporting H6. The result is shown in Figure 4. As hypothesized, the simple slopes of trust in the host community are significantly positive at low levels of privacy concerns (Mean − 1SD) and become slightly positive at high levels of privacy concerns (Mean + 1SD), further supporting hypothesis H6.

Conclusion and Discussion
Based on previous research on institution-based trust and privacy concerns, this study reveals, firstly, the importance of three types of institution-based trust in the context of SA, namely, trust in the user community, trust in the SA platform and trust in the host community. Secondly, according to trust transfer theory, we propose that trust in the user

Conclusion and Discussion
Based on previous research on institution-based trust and privacy concerns, this study reveals, firstly, the importance of three types of institution-based trust in the context of SA, namely, trust in the user community, trust in the SA platform and trust in the host

Conclusions and Discussion
Based on previous research on institution-based trust and privacy concerns, this study reveals, firstly, the importance of three types of institution-based trust in the context of SA, namely, trust in the user community, trust in the SA platform and trust in the host community. Secondly, according to trust transfer theory, we propose that trust in the user community will positively affect both trust in the SA platform and trust in the host community. Thirdly, we explore the impacts of trust in the SA platform and trust in the host community on continuous use intention. Finally, this study takes privacy concerns as a contextual factor of trust transfer and uncovers the moderating effect of privacy concerns on the relationship between trust in the user community and trust in the SA platform, as well as the relationship between trust in the user community and trust in the host community. Through empirical research, this study has the following three major findings.
First, trust in the user community has a significant positive impact on both trust in the SA platform and trust in the host community. This finding discloses the relationships among the three types of institution-based trust in the SA from the perspective of trust transfer and confirms the importance of trust in the user community. In the SA, experienced users on the platform will provide their own opinions and evaluations based on their own experiences. Users' beliefs in the peer user community should matter in the online context such as SA. Because the user community is generally associated with the platform (e.g., Airbnb) that users make transactions in and the host community that users have transactions with, trust can then be transferred from the user community to the other parties, i.e., the platform and the hosts, associated with the user community.
Second, trust in the SA platform and trust in the host community positively affect users' continuous use intention. This finding is in line with the research of Mao et al. [18], in which trust in the platform and the hosts are found to reduce the risk in the transaction process, thereby promoting transaction willingness. When users participate in the SA, they rarely transact with the same service providers. The interpersonal trust formed in user-seller interactions is thus ineffective in promoting continuous transaction willingness. Trust in the SA platform and trust in the host community, as institutional trust, then play a more important role in driving repeated transactions. This finding also confirms the importance of institution-based trust in the sharing economy [7].
Third, privacy concerns weaken the relationship between trust in the user community and trust in the SA platform and the relationship between trust in the user community and trust in the host community. Drawing on the privacy calculus model, this study reveals the boundary condition under which trust is transferred. Different users hold different information privacy concerns when making SA transactions. Users with strong privacy concerns will generally seek other sources of trust or protective mechanisms to strengthen their trust toward both the platform and the hosts.

Research Implications
This research has three major theoretical contributions to the existing research. First, this study reveals the positive impacts of both trust in the SA platform and trust in the host community on continuous use intention, thus confirming the importance of institutionbased trust in online contexts such as SA. Previous studies have emphasized and suggested the importance of institutional trust in the field of e-commerce. For example, McKnight et al. [58] argued that in the absence of prior interactions such as e-commerce, institutional trust can enable unfamiliar members of society to cooperate and share with others. Based on our findings, this study finds that another type of institution-based trust, trust in the host community, can contribute to continuous use intention in addition to the effect of trust in the SA platform. Our study thus deepens the understanding of institutional trust by extending the institution-based trust to the SA context and illustrating the positive impact of trust in the host community on continuous use intention.
Second, this study expands the research on institution-based trust [59,60] based on the finding that trust in the user community is an important influencing factor for both trust in the SA platform and trust in the host community. Although the importance of institutionbased trust has been demonstrated and tested in the sharing economy literature [61], most of these studies focused only on trust in the SA platform, and trust in the host community and trust in the user community are two other types of institution-based trust that have received very little attention. In the SA context, we find that trust in the host community also has a positive and significant impact on users' continuous use intention. At the same time, trust in the user community is an important source of both trust in the SA platform and trust in the host community. Therefore, this study comprehensively considers the three types of institution-based trust into one single model and examines the relationships among them, making a good complement to the existing research [7,62] on institutional trust.
Third, this research extends the research on trust transfer [63] by examining the moderating effect of privacy concerns. Previous research [13], mostly on the boundary conditions of trust transfer, has been conducted in the e-commerce context and most of them [9] have only considered institutional factors as situational variables. The literature rarely examines the boundary condition of trust transfer from the individual user perspective. Users' internal psychological state is also an influential situational factor when users participate in the sharing economy such as the SA. From the perspective of individual users, this study proposes privacy concerns as a situational factor under which trust is transferred and explores how privacy concerns affect the relationship between trust in the user community and trust in the SA platform and the relationship between trust in the user community and trust in the host community. Our study not only extends the research on privacy concerns and the privacy calculus model [20] to the context of SA but also makes contributions to the existing research on trust and trust transfer [18].

Practical Implications
First, the platform should pay attention to the important role of the user community and encourage them to offer their feedback, ratings, and comments concerning their past transactions. Trust in the user community is found to contribute to both trust in the SA platform and trust in the host community. Platforms should then find effective ways such as encouraging interactions among users and improving the effectiveness of user rating systems to enhance their trust toward the user community.
Second, for SA service providers and platform firms, our study finds that trust in the SA platform and trust in the host community can promote users' continued willingness to participate. In SA, it is almost impossible for users to deal with the same host every time, so it is difficult to form interpersonal trust between individuals. Compared with interpersonal trust between individuals, institution-based trusts play more important role in promoting user behaviors. Platforms should then focus on building user trust from the institutional level. They should first enhance users' trust in the SA platform by continuously investing in establishing effective institutional mechanisms (such as feedback mechanisms, escrow services, and provider certification) [59], improving service quality, and building a good platform image. In addition, platforms can consider adding more trust-enhancing signals to their websites, such as hosts' trust levels and rating scores, to offer more information to engender trust toward the host community.
Finally, our results show that the impacts of trust in the user community on both trust in the SA platform and trust in the host community are different under different conditions of users' privacy concerns. Platforms should then give special attention to privacy concerns and their impacts in the context of SA. For users who are highly concerned about their privacy, the impact of trust in the user community on trust in the SA platform or on trust in the host community is weakened, as these users seek more assurance from other parties than the user community to handle their concerns about personal data. Platforms can alleviate users' information privacy concerns by improving their compliance with the related laws and regulations and allowing authorities to monitor and supervise their practices and procedures regarding data collection, storage, processing and distribution. Platforms can also demonstrate their credibility by adopting relevant cutting-edge cyber security technologies, choosing reputable payment systems, and formulating privacy protection and compensation policies.

Limitations and Future Research Directions
This study still has some limitations. First, the survey data comes from experienced users of typical platforms in the SA industry, and the variables and meanings constructed may be different from users in other industries. Therefore, further testing is needed to generalize the results from a single industry to other sharing economy industries. Future research can further advance our investigation by including users in other industries, such as sharing mobility. Second, this study considers the moderating role of privacy concerns at the individual user level and does not consider the impact of other types of risk perceptions. Future research can then consider the impact of other risk perceptions on institution-based trust. Finally, with the survey method adopted in this paper, the respondents may be affected by the environment and other factors, which may cause deviations in the research results. In the future, various research methods can be considered to verify the measurement model.
Author Contributions: B.L. developed conceptual models and was responsible for data analysis and paper revision. Z.W. was responsible for data collection, model analysis and writing the first draft of the paper. All authors have read and agreed to the published version of the manuscript. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

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

TIP2
The platform is trustworthy.

TIP3
The platform knows how to provide an excellent online booking accommodation service.

Continuous
Use Intention CUI1 I will continue to use the platform to book rooms.

Gefen & Straub, 2004 [43] CUI2
In the short term, I will continue to use the platform to book rooms. CUI3 In the long term, I will continue to use the platform to book rooms. Other users of the platform are truthful in dealing with one another

TIUC3
The information provided by other users in the comment area is worth considering.