Tourist–Host Identity Risk: Scale Development and Consequences

Risk perception has been an important construct in understanding tourists’ trip planning. Surprisingly, while tourists’ perceived tourist–host identity risk (THIR) plays an essential role in this process, its effects have been overlooked. Against this backdrop, this study develops a THIR scale and investigates how it shapes the intention of mainland Chinese tourists to revisit Hong Kong, and also investigates the spread of positive word-of-mouth (WoM) by conducting multiple online surveys with mainlanders. The results show that THIR is a uni-dimensional construct which is robustly measured by 12 items. It undermines revisit intention and the spread of positive WoM because of lower emotional solidarity followed by anticipation of unpleasant experience. This negative serial mediation is weaker for people who have a strong Chinese identity and perceive high THIR of an alternative destination (Macao). Meaningful implications are provided for destination marketers.


Introduction
Recent years have seen the rise of social and ethnic tensions around the globe. Many incidents and events, including Brexit, COVID-19 pandemic, and Russia-Ukraine war have stimulated debates related to ethnic diversity and even racism (Adam et al. 2022; UN News 2020) as those events (re) construct identity boundaries between ingroup and outgroup members of societies. One striking example is the continuous identity crisis in Hong Kong (HK), featured by the 2019-20 protests. For many, "being Hongkonger and being Chinese, long complementary, suddenly came to feel exclusive" (Fisher 2019). Before the protests, over 80% of tourist arrivals in HK were mainland Chinese, contributing to over 5% of HK's GDP. However, since the HK protests, tourist arrivals have reduced by more than 50%, largely due to the drop in mainland Chinese tourists before the pandemic (HKTB 2020). It is acknowledged that a tourist destination has a higher level of risk when there is a clash of identities, including genders, nationalities, religions, etc. (Hughes 2002). Here, the avoidance of HK by the mainland Chinese is strongly associated with their reflective reactions toward the clear identity boundaries created during the protests. To recover its tourism industry, it is thus important for HK destination marketers to understand the identity-related risks involved in the trip planning of mainland Chinese tourists.
So far, the extensive travel-risk literature has not documented the risk arising from identity tension. This study introduces the concept of tourist-host identity risk (THIR). As THIR has not been conceptualized previously, tourism studies regarding tourist-host interactions shed light on our pioneering work. Hosts exhibit their discrimination against tourists in the forms of unfriendly attitudes, an unwillingness to help, and mockery (Tse and Tung 2022). Hosts also demean tourists whose behavior is considered uncivilized (Chen, Hsu, and Li 2018;Lowe and Tsang 2017;Yan and Cheng 2021). These insightful studies, however, have not thoroughly revealed the hosts' rivalrous attitudes and actions against the tourists, which were beyond their research objectives. Nonetheless, they supplement the conceptualization of THIR (see the next section) and development of its measurement scale in this study.
Without proving its predictive ability, a construct's value cannot be completely realized. As risk aversion is intuitive (O'Donoghue and Somerville 2018), this study focuses on the psychological mechanism that articulates "how" THIR predicts avoidance behaviors including lower intention to revisiting the destination or to spreading positive word-ofmouth (WoM). This study proposes a serial psychological process characterized with emotional solidarity (the affective bond between tourist and host (Woosnam and Aleshinloye 2013)) and anticipated experience based on the theoretical perspective of emotion as feedback (Baumeister et al. 2007). The findings offer a deeper understanding of emotion and cognitive responses in the risk evaluation process, and these possible negative consequences of THIR help convince destination marketers to managing potential tourists' THIR.
This study situates itself in the identity crisis in HK and explores the intense tensions between defining ingroup and outgroup members for both mainland tourists and HK hosts. We propose mainlanders' Chinese identity moderates the above serial psychological process based on social identity, which is a social psychological concept that explains intergroup relations and behaviors (Hogg, Abrams, and Brewer 2017). In addition to this dispositional moderator (Chinese identity is a personal trait), this study proposes a situational moderator, namely THIR of an alternative destination (THIR-AD), given its relevance to the context of this study. Specifically, we examine if the THIR of Macao (a sibling city of HK and the second special administrative region of China after HK) interferes with our proposed serial psychological process. These two moderators indicate "when" the psychological process will be salient.
In sum, the objectives of this study are threefold. The first objective is to develop the THIR scale. The second is to verify the proposed psychological mechanism that bridges THIR and revisit intention/spread of positive WoM. The third objective is to examine whether the mechanism is moderated by Chinese identity and THIR-AD, as well as how they work.

Tourist-Host Relationship and Risk in Tourism
Tourist-host contact is plausibly inevitable in a trip and a crucial factor that shapes tourist experience and residents' pro-tourism attitude (Fan 2020). The prevalence and importance of this social phenomenon have given rise to extensive research that delves into the formation processes and influences, through the lens of the tourists and residents. For examples, tourists' contact evaluation rests upon guest-host, guest-community, and guest-guest interactions (Lin et al. 2019). The contact quality influences tourists' perceived cultural distance (Fan et al. 2017). Residents' support for tourism is reliant on the health risk arisen from their contact with tourists (Joo et al. 2021). Although some residents' behavior against the tourists is offensive, others are willing to offer helps (Tse and Tung 2022). Lin and Fu (2021) recently synthesized tourist-host relationship studies in three major streams including tourist-initiated gaze, host-initiated gaze, and tourist-host interactions. The tourist-initiated gaze research synthesizes studies about tourists' power and authority over their hosts as well as tourists' interaction with each other. The host-initiated gaze research encompasses research related to the hosts' perception of tourists' behavior and resultant impacts. The tourist-host interactions concern the impact of their interactions. While previous studies have focused on enhancing the host-tourist experience, recent literature has started to understand tourist-host identity tension and has called for further research (e.g., Zhang, Wang, and Rickly 2021); this recent trend provides useful information to develop a THIR scale.
Risk is perceived as known uncertainties associated with negative outcomes (Williams and Baláž 2015), and has been regarded as an influential element to understand consumer behavior as purchasing decisions always involve risk and its associated consequences (Mitchell 1999). Within the tourism literature, perceived risk as a tourist's subjective assessment of the uncertainty of the results of tourism activities is often considered to be a major factor that hinders travel intention (Ritchie and Jiang 2019). For many, the tourism industry is fundamentally related to various risks, including natural disasters, health threats, political instability, criminality, terrorism, financial loss, and many others (Karl, Muskat, and Ritchie 2020;Williams and Baláž 2015). These risks vary in terms of their probability of occurrence and severity of outcome (Ritchie and Jiang 2019). The tourism literature has documented a long list of risks and the term perceived risk has become popular, especially in the recent surge of Covid-19 tourism research. Still, researchers argue that risk study in tourism is developing and descriptive due to the lack of theoretical foundations (Williams and Baláž 2015;Yang, Zhang, and Rickly 2021). Arguably, this is mainly because people's subjective assessment of risk may vary according to products (or services) and most importantly its contexts (Dayour, Park, and Kimbu 2019). As risk is specific and contextual, it is vital to situate risk discussions within specific contexts to enhance its theoretical sensitivities. Here, we contextualize risk within the recent rise in the identity debate globally through the case of HK.

Tourist-Host Identity Risk and Consequences
When concerning about destination, many believe that tourists tend to avoid places with higher likelihoods of risk while favoring less risky destinations. An emerging trend of research shows that this might not be the case. For example, backpackers tend to choose risky destination to fulfill their needs as a backpacker and enjoy an adventurous destination experience (e.g., Dayour, Park, and Kimbu 2019). While some destinations might be safe for many, they might be considered risky for some. LGTB tourists, for example, often emphasize that their sexual orientation and their perceived tourist-host identity tensions are a priority when choosing destinations (e.g., Usai, Cai, and Wassler 2022). It is clear that many studies reveal that tourists' risk perceptions are strongly associated with the perceived identity relationship between tourists and hosts (e.g., Hughes 2002;Zhang, Wang, and Rickly 2021). However, previous studies concerning risk and destination selection have often aimed to provide a general list of various risks associated with destinations (e.g., financial, political, psychological, and others) without an indepth investigation. Hence, the psychological reasoning and identity implications beyond these general dimensions have been overlooked and therefore hinder our understanding of the effect of risk perceptions on behavioral consequences. As social identity plays a central role in intergroup tensions and subsequent behavior (Ellemers, Kortekaas, and Ouwerkerk 1999;Hogg, Abrams, and Brewer 2017), this study deploys social identity theory to develop a tourist-host identity risk perception scale (THIR). In doing so, the study moves the risk perception discussion from destination attributes to the social identity positions between groups.
At the theoretical level, social identity is defined as ". . . that part of an individual's self-concept which derives from his [or her] knowledge of his [or her] membership of a social group (or groups) together with the value and emotional significance attached to that membership" (Tajfel 1978, 63). The evaluative process of ingroup and outgroup identity guides individuals' emotions and behavior toward ingroup and outgroup members. For many, as social group processes fundamentally structure our perception of safety versus vulnerability, we often tend to perceive ingroup members as less risky than potential identity threats arising from outgroup members (Cruwys, Stevens, and Greenaway 2020). Indeed, group members "react protectively if they feel their cherished and distinctive social identity's existence is being threatened" (Hogg, Abrams, and Brewer 2017, 574). The intergroup tension expands their divergence, which in turn reinforces their group identity in relation to the outgroup. The group's protective reaction manifests the perceived risk when interacting with the outgroup members. This identity-based tension is formed not only by personal experience, but also indirectly through exposure to news about the tension (Curran and Gibson 2020), for example, news about a tourist's compatriot being unfairly treated by the host at the destination. Within the tourism literature, previous studies often assume that tourists are outgroup members and therefore hold different social identities from the hosts; but in fact, identity boundaries are subjective and contingent (Zhang, Wang, and Rickly 2021). In the case of HK, Chinese tourists could be considered as ingroup members as well as outgroup members for the host communities. It is this uniqueness of the HK case that offers rich insights into intergroup (potentially intragroup) identity tension and its associated perceived risk. The development of Tourist-host Identity Risk (THIR) is therefore fundamentally about tourists' risk perceptions that arise from intergroup identity evaluations.
High risk perception is generally followed by risk-avoidance behavior across domains in daily life (O'Donoghue and Somerville 2018). The exception is when the person has disinhibitory psychopathology (e.g., substance use disorder) (Maner and Schmidt 2006), which is beyond the scope of tourism. Likewise, the negative relationships between THIR and two common but practically important consequences, revisit intention and positive word-of-mouth, appear intuitive. Thus, this research focuses on the underlying mechanism of these relationships.
Recent research findings highlight emotional solidarity as a key indicator that shapes the responses of tourists and hosts. Emotional solidarity represents "the affective bonds individuals experience with one another, which are characterized by perceived emotional closeness and degree of contact" (Woosnam and Aleshinloye 2013, 495). It is deeply planted in both contact parties, such as tourist and host in the context of this study. Frequent interaction nurtures hosts' emotional solidarity (Joo et al. 2018). These hosts then tend to support tourism development, recognize the contribution brought by tourists (Joo et al. 2021;Woosnam 2012), and perceive a better quality of life (Seo et al. 2021). Similarly, tourists' emotional solidarity is nurtured by their frequent interaction and shared beliefs with the hosts (Woosnam and Aleshinloye 2013). Their high emotional solidarity is followed by positive destination image (Woosnam, Stylidis, and Ivkov 2020), satisfaction, and thereafter, destination loyalty (Ribeiro et al. 2018). Following this rationale, the positive influences of emotional solidarity on revisit intention and positive word-of-mouth are plausible.
On the other hand, prior research shows that tourists who share beliefs and interact pleasantly with hosts are more likely to affectively bond with them (Woosnam and Aleshinloye 2013). While the identity risk was built on the tension and possible unpleasant experience, tourists' perception of THIR likely inhibits the construction of their emotional solidarity, lending support to the negative path from THIR to emotional solidarity and hence the behavorial consequences.
Prior research has shown that emotional solidarity induces positive behavioral consequences. We take a step forward by examining the inner mechanism from the theoretical perspective of emotion as feedback (Baumeister et al. 2007). This perspective argues that the direct impact of emotion on behavior, albeit parsimonious, is not counted in the cognitive processing of prior experience. The direct impact works when the emotion is spontaneously elicited, which then drives the immediate action. For example, fear triggers avoidance. However, if the emotion is consciously built on the processing of information and previous experience, like evaluation of identity risk, the emotion will guide future action through the anticipated outcome of the action (Jenkins and Molesworth 2018). In other words, the counterfactual thinking is triggered so that avoidance is the resultant consequence. In this sense, tourists' low level of emotional bond with the hosts may discourage revisit behavior or the spread of positive WoM, because an unpleasant travel experience is anticipated if they revisit the destination and if their friends/ relatives do the same. Taking this postulation together with the above discussions regarding the negative relationship between THIR and emotional solidarity, we propose the following serial mediation hypotheses: Hypothesis 1: Emotional solidarity and anticipated experience serially mediate the negative predicting relationship between tourist-host identity risk and revisit intention (H1a)/positive WoM (H1b)

The Moderating Role of Chinese Identity
Grounded in the "one country, two systems" constitutional principle, HK is governed as a special administrative region of China. Although mainlanders and Hongkongers share the same nationality and politically belong to one social group-Chinese, a huge cultural distance (in terms of civilization, hygiene standards, and privacy protection) exists between them (Fan et al. 2017), as was evident in 19/20 protests. The cultural distance fuels their tension to a critical point that some Hongkongers resist being grouped with their mainland compatriots (Chow, Fu, and Ng 2020). Such identity-based categorization was escalated by the arguments between HK and mainland users on the web (Wong 2021). From the viewpoint of mainland users, HK is part of China and enjoying economic benefits given by it, so Hongkongers deservedly appreciate it and recognize their own Chinese identity (Luo and Zhai 2017). Such viewpoint manifests in mainland Chinese's strong sense of belonging to China (i.e., a strong national identity according to Triandafyllidou (1998)) and blame directed at Hongkongers' ungrateful behavior. Based on this reasoning, the Chinese identity possibly moderates the relationship between THIR and its consequences, specifically emotional solidarity, which is about their affective bond with the hosts.
To understand intergroup relations and behavior, social identity theory is useful, which argues that people divide individuals into groups and label a person as an ingroup member if s/he shares common traits, opinions, and behaviors, whereas others are labeled as outgroup members (Hogg, Abrams, and Brewer 2017;Wang 2017). Here, the blurry boundary-making between ingroup and outgroup members raises two competing results. On the one hand, the intensive identity crisis in HK indicates Chinese and HK identities have become mutually exclusive (Fisher 2019). This social categorization indicates mainlanders with a strong Chinese identity may now tend to treat HK hosts as outgroup members and vice versa. Thus, their high perception of THIR is likely to further undermine their affective bond with the hosts.
On the other hand, social identity theory suggests ingroup favoritism, which indicates that members tend to have a positive evaluation of ingroup members and tolerate any conflicting behaviors among some sub-group members (Trepte 2017). Here, given the ethnic and political identity of being Chinese, mainland Chinese tourists might also categorize Hongkongers as ingroup members, and their emotional affiliation with the Chinese identity will certainly increase their tolerance level to tensions and risks. Hence, the negative impact of THIR on emotional solidarity would be buffered by the Chinese identity.
While the blurry boundary-making raises competing moderating effects of Chinese identity, we believe that the ingroup favoritism might be active when it targets mainlanders. According to Hu et al. (2016), mainlanders' contemporary wisdom features a compassionate relationship, feedback and adjustments, and striving for common good, which indicates ingroup members' endeavor for the group's positive identity and active ingroup favoritism. A sentiment analysis of China's social media reports that some mainlanders support Hongkongers in their request for democratic rights but disagree with their way of asking for them (i.e., Occupy Central) (Luo and Zhai 2017). Also, recent identity studies related to Chinese tourists tend to show an increase in emotional affiliation of Chinese identity and ingroup favoritism (Zhang, Wang, and Rickly 2021). So, the Chinese identity is likely to buffer the negative effect of THIR on emotional solidarity, and hence the subsequent consequences: Hypothesis 2: Chinese identity moderates the serial mediation effect of emotional solidarity and anticipated experience between THIR and revisit intention (H2a)/positive WoM (H2b). Specifically, when Chinese identity is strong (vs. weak), the negative predicting relationship between THIR and emotional solidarity is weaker (vs. stronger), and hence anticipated experience and revisit intention (H2a)/positive WoM (H2b) are higher (lower).

The Moderating Role of Tourist-Host Identity Risk of Alternative Destination
The "one country, two systems" constitutional principle is being applied to HK and its sibling city, Macao, which is also governed as a special administrative region of China. Unlike HK, which was an enclave of Great Britain, Macao was colonized by Portugal and returned to China about two years later than HK. Over the past two decades, Macao's economy has experienced a phenomenal growth owing to the large number of mainland Chinese tourists spending money in the city (Bavik, Correia, and Kozak 2021). Negative opinions of their mainland compatriots are rarely heard from Macao residents, while mainland Chinese tourists generally appreciate their interaction with the Macao hosts (Yan and Cheng 2021). Such a phenomenon, which is contrary to that in HK, sets the stage for studies focusing on comparing tourists' attitude between these sibling destinations. Macao naturally becomes a reference point when mainlanders are evaluating the risk of visiting HK.
Reference point theory suggests that evaluation of a stimulus is dependent not only on the traits of the stimulus itself but also on the reference point considered in the evaluation process (Aytaç 2018). Following this reasoning, the lower emotional solidarity (emotional suffering) in consequence of THIR of HK (a loss) may worsen if the risk evaluation involves comparison with THIR of Macao (a better alternative because of the lower risk). However, the reference-point literature also counter-argued that the available alternatives, even though better, may not sway the decision because people perceive ownership with the stimulus and any switching of choice as a huge loss that cannot be compensated by the gain derived from the alternative (Samuelson and Zeckhauser 1988). This phenomenon is called status quo bias, which however is not applicable to this study because the current state (or status quo) was generally neutral (Kahneman, Knetsch, and Thaler 1991), rather than a loss like THIR in this study. Indeed, it is hard to imagine that someone would stick to a loss if a better option were available. So, the reference point theory should better support the moderating role of THIR of an alternative destination (Macao) than status quo bias, and hence we hypothesize that: Hypothesis 3: THIR of alternative destination (Macao) moderates the serial mediation relationship from THIR of the focal destination (HK) to emotional solidarity, anticipated experience and (3a) revisit intention/(3b) positive WoM. Specifically, when THIR of alternative destination is high (vs. low), the negative predicting relationship between THIR of the focal destination and emotional solidarity is weaker (vs. stronger), and hence anticipated experience and (3a) revisit intention/(3b) positive WoM are higher (lower).

Methodology and Results
Our literature review indicates that no study thus far has made an attempt to develop and validate a scale that measures THIR. Moreover, a major objective of this study is to reveal the factors that constitute the THIR perceived by mainlanders about Hongkongers. To discover the factors, this study followed the standard procedures of scale development including (1) item generation and content validation, (2) item purification, (3) scale validation and application, and (4) scale replication and robustness (Churchill 1979). The proposed hypotheses are tested in phases 3 and 4.

Item Generation and Content Validation
The items were generated from two major sources, which were the literature and in-depth interviews with 31 mainlanders. The interviewees were recruited through the personal contacts of an author, while the interviewees' profiles were deliberately diversified in terms of gender (male = 51.6%), age (30 and below = 38.7%, and 31%-50% = 51.6%, over 50 = 9.7%), and education level (high school = 6.5%, junior college = 19.4%, Bachelor = 41.9%, and Postgraduate = 32.2%), for the purpose of maximizing representativeness of responses. The interviews were conducted between January 2021 and July 2021. The process generated 20 items (in Chinese), which can be found in the Appendix.
The generated items were then assessed by five independent researchers who are mainland Chinese. They were explained the definition of THIR and then asked to rate the items in terms of their representativeness of the concept (using a 4-point scale ranging from 1: not at all representative, 2: minimally representative, 3: moderately representative, and 4: highly representative) (Taheri et al. 2018). If an item was rated 2 or below, the researcher was asked to provide the reason why. The process ended up with 13 items, which were considered to have face validity (see Appendix for these items and the reasons for excluding other items). The retained items were empirically validated in the phase of item purification.

Item Purification (Study 1)
A pilot survey study (Study 1) was conducted to purify the items. Data were collected using the sampling service of WJX.com, which has been widely used and deemed suitable for academic research that targets Mainland Chinese respondents Wang et al. 2021). Qualified respondents of this online survey were mainlanders aged 18 or above who had experience of visiting HK, and three corresponding screening questions were asked at the beginning to ensure this. The 13 items were operationalized using a 7-point Likert scale (1: strongly disagree; 7: strongly agree). Attention check questions (e.g., "please choose 'Agree' for this item") were added for quality control of data. The survey started in early August 2021 and collected 70 responses. No respondent expressed difficulties in understanding the survey questions.
Exploratory Factor Analysis (EFA) was performed to explore the factor structure. The results show a single factor with all factor loadings exceeding .4 (the recommended threshold) and all communalities, except one item, namely "locals always speak English with me despite knowing I am from Mainland China" exceeding .4 (the recommended threshold) (Field 2014). The total variance explained is 58.225% while the Bartlett's Test of Sphericity (χ 2 (78) = 595.016, p = .000) and Kaiser-Meyer-Olkin (KMO = .916) results are satisfactory. If this marginal item is removed, the Cronbach's alpha would increase from .937 to .940 (both greater than .7). Considering the satisfactory factor-loading of this marginal item and the small sample size which threatens the validity of EFA results, we decided to further assess it in the main study. In sum, the reliability and validity of the measurement of tourist-host identity risk are deemed adequate.

Scale Validation and Application (Study 2)
Design, measure, and respondents. The research design of this main study (Study 2) resembled the pilot study. THIR was measured by the same 13 items. This study also examines H1 and H2 so that the applicability of the THIR scale can be assessed. H3 was examined in an additional study because it involves the THIR of an alternative destination, which will be operationalized by adapting our THIR scale to a different host (i.e., not HK). Without validating the THIR scale, adapting it would be vulnerable.
Revisit intention was measured by three items adapted from Chaulagain, Wiitala, and Fu (2019), while anticipated experience was measured by four 7-point bipolar scale items (undesirable-desirable, unpleasant-pleasant, unfavorablefavorable, and unenjoyable-enjoyable) (Kim, Njite, and Hancer 2013). As these two constructs were related to future visits and our data were collected during the COVID-19 pandemic, their measurement items state that responses should be based on ignorance of COVID-19, so that the pandemic effect on the visit can be ruled out. Positive word-of-mouth was measured by three items adapted from Šegota, Chen, and Golja (2022) Both measures used a 7-point Likert scale. Emotional solidarity was operationalized with six 5-point Likert-scale items adapted from Ribeiro et al. (2018) Chinese identity was measured by ten 7-point Likert-scale items adapted from Ellemers, Kortekaas, and Ouwerkerk (1999). The covariate, frequency of visit, together with demographic information including gender, age, education, and monthly household income, were asked about at the end of survey. These measurement items and questions were translated into Chinese by one of the authors and translated back into English by an independent person, in order to ensure the semantic equivalence of the Chinese instrument. As the THIR scale was originally developed in Chinese, it was translated into English and translated back into Chinese before being reported in this paper.
WJX started data collection from mid-August 2021 and collected 530 responses by the end of the month. Among these responses, 31 outliers were eliminated because at least one variable had a z-value outside the range of −4 and 4. Therefore, 499 responses were retained for subsequent analysis. The respondents tended to be female (62.7%), aged 25-34 (74.4%), holding a Bachelor's degree or above (97.6%), and with a monthly household income amounting to 10,000-29,999 Yuan (60.8%).
Exploration of factor structure. We randomly split the data into two sets to validate the scale of THIR. Exploratory Factor Analysis (EFA) was performed to explore the factor structure using the first dataset. The results guided the subsequent confirmation of factor structure using the second dataset. Field (2014) recommended a sample size of over 300 for EFA. Hence, the first dataset contains 301 samples, whereas the second dataset contains 198 responses. EFA generated a factor structure similar to that in Study 1. A single factor with all factor loadings greater than .4 was generated. Like Study 1, only the item "locals always speak English with me despite knowing I am from Mainland China" has a communality lower than .4, while removing it increases Cronbach's alpha from .924 to .929 (both >.7). The consistent results of this study and Study 1 provide a strong data-driven reason to remove this marginal item. Actually, this item will not be valid if English is not a common language of another host city (e.g., Macao in the next study). Taken together, we decided to delete this item from the scale. The retained 12 items were analyzed by EFA again. A single factor with satisfactory results was again derived (factor loadings and communalities greater than .4; KMO = .938; Bartlett's Test of Sphericity: χ 2 (66) = 2,170.474, p = .000; 56.693% of variance is explained).
Confirmation of factor structure. The 12-item factor structure was assessed by the second dataset. Recent research suggested the use of Confirmatory Composite Analysis (CCA), rather than Confirmatory Factor Analysis (CFA) at the stage of scale development because of its better performance in content coverage and construct validity (Hair, Howard, and Nitzl 2020). CCA is enabled by Partial Least Square Structural Equation Modeling (PLS-SEM), which is suitable for small sample sizes (Hair et al. 2017) such as this dataset which contains 198 responses. So, this study adopts CCA and follows the procedures articulated in Hair, Howard, and Nitzl (2020).
A single-factor measurement model of THIR was specified. Results show that all outer loadings exceed .4 and are statistically significant. Internal consistency is demonstrated in the Cronbach's alpha (.922), Composite reliability (.933), and rho_A (.925), which are all above .7. Convergent validity is evidenced by the average variance extracted (AVE) (.538) which is above .5. Discriminant validity is also attained given that three criteria were met: (1) outer loadings are greater than absolute value of cross loadings; (2) squareroot of AVE is greater than absolute value of correlations between corresponding construct and other constructs; (3) Heterotrait-Monotrait Ratio (HTMT) values are lower than .85 while HTMT inference criterion (not including 1 in the range of confidence interval bias corrected) is supported. Nomological validity was assessed by correlating THIR with intention to revisit HK. As expected, a significant negative correlation was revealed (r = −.573, t = 10.176, p < .001).
Taking the above results together, the scale should be deemed valid. Its applicability is assessed by testing the hypotheses in the conceptual model.

Hypotheses testing (H1 and H2).
Hypotheses testing is based on the full dataset (499 responses). While the largest number of paths pointing to a construct in the model is 4, the minimum sample size should be 40 (4 × 10). The sample size is adequate. The measurement model testing results show that the AVE of social identity is lower than .5. Hence, the items with low outer loadings were excluded one by one until the AVE threshold was passed. This repetitive process eliminated four items (see Appendix). After that, all construct measures attained internal consistency, convergent validity, and discriminant validity. Tables 1 and 2 exhibit the corresponding indicators. Common method bias (CMB) was assessed following Harman's single-factor approach. The factor analysis generates five factors and the first factor accounts for 38.522% of variance (<50%), so that CMB was not detected. Like Tsai and Fong (2021), we made a further assessment using a more rigorous approach called Unmeasured Latent Marker Construct (ULMC). Results show that most method factor loadings are not statistically significant while the ratio of average indicators' substantive variance to average method variance is 80:1, which is greater than 42:1 in Liang et al. (2007). The results provide additional evidence on the absence of CMB.
The structural model test reveals that VIF values are less than 5, indicating that collinearity is not an issue. H1a and H1b are supported as the negative indirect effects from THIR through emotional solidarity (ES) and anticipated experience (AE) to revisit intention (RI) and positive WoM (PWoM) respectively are statistically significant (see Table 3). Specifically, THIR negatively predicts ES (coefficient = −.620, t = 17.750, p = .000), which positively predicts AE (coefficient = .373, t = 6.419, p = .000), while AE positively predicts RI (coefficient = .386, t = 7.889, p = .000) and PWoM (coefficient = .260, t = 6.260, p = .000) respectively. The R 2 values are .577 for RI, .610 for PWoM, .366 for AE, and .451 for ES. The blindfolding procedure demonstrates the predictive relevance of the structural model as all Q 2 values are above 0. The PLS-predict procedure reports a satisfactory predictive power as all Q 2 predict values are above zero and a majority of Linear Model (LM) values are greater than PLS-SEM values, irrespective of using the Root Mean Squared Error (RMSE) approach or Mean Absolute Error (MAE) approach (Shmueli et al. 2019). The results testify the applicability of THIR.
A two-stage approach was deployed to examine the moderating effect of Chinese identity on the serial mediations (H2a and H2b). Latent scores of constructs were generated (stage 1) and utilized to examine the structural model with the interaction variable, which was derived from multiplication of latent scores of independent variable and moderator (stage 2). H2a and H2b are marginally supported as their p-values are below .1 (Table 3). Figure 2 illustrates strong (vs. weak) Chinese identity marginally weakens (strengthens) the negative influence of THIR on emotional solidarity.
Discussion of findings. The scale validation process supports the uni-dimensional structure of the THIR scale which contains 12 items. While the host in this study was HK, we do not know if the scale works for a different host. So, we conducted an additional study to validate whether the scale is robust against Macao, which is a sibling city of HK and is also governed under China's "one country, two systems" principle. In this regard, the additional study enables the examination of another proposed moderator, namely THIR of alternative destination (Macao). Meanwhile, the replicability of results can be assessed.

Scale Replication and Robustness (Study 3)
Design, measures and respondents. The research design of this additional study (Study 3) followed Study 2. Given the results of Study 2, where the THIR of HK (the focal destination) was operationalized by 12 items (see Table 4), the THIR of Macao (the alternative destination) was measured by adapting the same 12 items to Macao as the host. Hence, the respondents should have experience of visiting Macao, and this additional screening question was asked together with other three questions (see Section 3.2) at the beginning.
The measures of other constructs and socio-demographic variables followed Study 2. The structure of survey and order of questions resemble Study 2 so as to maximize measurement invariance. The only exception is that the additional questions related to Macao were asked right after the questions related to HK. WJX started data collection from early September 2021 and collected 522 responses in half of a month. Among these responses, 54 outliers were eliminated and hence 468 responses were left for subsequent analysis. The respondents tended to be male (54.5%), aged 25-34 (72%), holding a Bachelor's degree or above (95.5%), and with a monthly household income amounting to 10,000-29,999 Yuan (60.3%).

Assessment of measurement model. The PLS-SEM results
indicate that the measures of THIR of both HK and Macao are reliable and valid as the results lend support to their internal consistency, convergent validity, and discriminant validity (see Tables 4 and 5). Owing to its low AVE (<.5), the measure of Chinese identity needs to exclude four items, which is consistent with the findings in Study 2. After that, no item needs to be excluded. CMB was not detected as Harman's single-factor test generated seven factors and the first factor accounts for 31.427% of variance (<50%). A consistent conclusion was drawn using the ULMC approach, which found that most method factor loadings were not statistically significant while the ratio of average indicators' substantive variance to average method variance is 59:1, which is greater than 42:1 in Liang et al. (2007).
Hypotheses testing (H1-H3). The structural model test does not report any collinearity issues as the VIF values are below 5. H1a and H1b are supported as the negative indirect effects from THIR through emotional solidarity (ES) and anticipated experience (AE) to revisit intention (RI) and positive     WoM (PWoM) respectively are statistically significant (see Table 6). Specifically, THIR negatively predicts ES (coefficient = −.750, t = 17.463, p = .000) which positively predicts AE (coefficient = .491, t = 11.364, p = .000), whilst AE positively predicts RI (coefficient = .414, t = 7.895, p = .000) and PWoM (coefficient = .377, t = 8.632, p = .000) respectively. The R 2 values are .541 for RI, .613 for PWoM, .452 for AE, and .444 for ES. The blindfolding procedure indicates good predictive relevance as all Q 2 values are above 0. The PLSpredict procedure indicates satisfactory predictive power as all Q 2 predict values are above 0 whilst a majority of LM values are greater than PLS-SEM values, irrespective of using the RMSE approach or MAE approach (Shmueli et al. 2019).
Deploying the two-stage approach, this study found significant moderating effects of Chinese identity and THIR of alternative destination (i.e., Macao) on the serial relationship, so that H2-H3 are supported (Table 6). As shown in Figures 3 and 4, strong (vs. weak) Chinese identity and THIR of alternative destination weaken (strengthen) the negative influence of THIR of focal destination on emotional solidarity.
Discussion of findings. The findings provide additional evidence for the validity of a uni-dimensional structure of the THIR scale, which comprises 12 items. The robustness of scale is demonstrated in the measurement model results with HK and Macao as hosts. Like Study 2, the emotional solidarity and anticipated experience serially mediate the negative relationship from the THIR of focal destination to revisit intention and positive WoM. Study 2 found a marginally significant buffering effect of Chinese identity on the serial mediation, whereas this study found the effect significant. This study also found that the THIR of alternative destination is another buffering factor.

General Discussion
This study has three major objectives which are achieved by the findings drawn from in-depth interviews and three surveys. First, the scale development process generated a unidimensional measure of THIR which contained 12 items. This measure was demonstrated replicable and robust against different destinations (HK and Macao). The results of outer loadings consistently indicate that being treated unfairly, unfriendly, and badly were the three major constituent components of THIR. Hence, the risk was primarily emotionladen which echoes the "risk as feeling" argument of Slovic et al. (2004).
Regarding the second objective, the findings show that THIR is likely to shatter revisit intention and spread of positive WoM, because of tourists' lower emotional solidarity with the host which possibly induces a mental projection of unpleasant experience on future visits. This psychological process indicates the risk-induced emotion needs cognitive processing to transform it into behavioral outcomes. The cognitive process involves a counterfactual thinking (the ifthen rule), which is an anticipation of experience that only exists mentally. This serial process is consistent with the theoretical argument of emotion as feedback (Baumeister et al. 2007).
The third objective concerns whether the serial psychological process is contingent on Chinese identity and the THIR of an alternative destination. The moderating role of Chinese identity plausibly suggests the existence of ingroup favoritism. Therefore, conflicting opinions are sometimes acceptable and welcomed from their perspective. In this sense, mainlanders who strongly recognize their Chinese identity can better tolerate conflict with HK hosts so that their emotional solidarity as a result of identity risk was less negative. As such, these mainlanders do not determine the hosts to be outgroup members, which then undermines the emotional solidarity, as suggested by the social identity theory. Regarding the moderating role of THIR of an alternative destination, the findings show that mainlanders' emotional solidarity was weakened by the THIR of HK, especially when the THIR of Macao was low. The finding is consistent with the reference point theory that the emotional suffering of loss (i.e., risk) is enhanced if a better alternative (i.e., lower risk) is available. While both moderating variables change emotional solidarity, the findings also show that anticipated experience, positive WoM, and revisit intention are influenced accordingly.

Theoretical Implications
The findings contribute to the literature in four major aspects. First, the proposed THIR expands the variety of travel risks identified in previous research. Recent years have seen political episodes that have resulted in aggravation of relations among countries. As THIR focused on identity-based risk perceptions, our study moves risk discussions from destination attributes to social identities between groups. In doing so, THIR scale contains important statements that reveal the psychological reasoning associated with risk perceptions, especially for destinations experiencing tourist-host identity tensions. The THIR is salient in trip planning and this research stream is likely to expand. Our scale paves the ground and provides an imminent instrument for future research on this topic.
Second, the serial psychological process explains how THIR shapes tourist behavioral outcomes. More importantly, as previous research showed that emotional solidarity influences tourist behavioral responses (Ribeiro et al. 2018;Woosnam, Stylidis, and Ivkov 2020), this study delves into this influence by revealing anticipated experience as a mediator, based on the "emotion as feedback" hypothesis (Baumeister et al. 2007). Future research can further extend our findings by identifying other possible mediators.   Third, while the mainland tourists have clearly played a significant role in the identity crisis in HK, their voice remains under-explored. In prioritizing the voice of tourists to understand identity issues through risk perception, the current study advances our knowledge on identity, tourism and risk. Additionally, the complex and competing understandings of the moderating effects of the Chinese identity have been empirically tested. Our study shows that ingroup favoritism is more suitable in explaining the moderating effect of Chinese identity on the THIR-emotional solidarity path.
Fourth, reference point theory effectively explains phenomena across domains including tourism. Our findings provide additional evidence by extending its explanatory power to the moderating effect of tourist-host identity risk of an alternative destination. To the best of our knowledge, although there is a plethora of studies focusing on the risk effect on travel behavior, the risk of an alternative destination as a situational moderator has not been discussed. Future research can include multiple referent destinations as multiple reference points may result in unexpected findings (Boles and Messick 1995).

Practical Implications
The HK destination marketers should be alerted to the threat of THIR. Based on the 12 items of our scale, mainland tourists are concerned that they will be treated unfairly, badly, unfriendly and that they will even be cheated. They also worry that their spoken language will enable the hosts to identify them, which then jeopardizes their personal safety. To minimize the negative impact of THIR, the HK destination marketers need to emphasize the importance of professional and sincere hospitality to the front-line service providers and inform them of the 12 measurement items and the possible consequences of THIR, such as the damage to HK's positive destination image, which has been established for years. On the other hand, when promoting the destination to mainland Chinese tourists, the HK marketers need to customize risk-relieving contents according to the 12 measurement items. For example, a promotional video can feature a HK resident speaking Putonghua and helping mainland Chinese tourists.
When promoting its destination to mainlanders, HK marketers should avoid doing so together with their Macao counterparts. Otherwise, the reference point effect of Macao will be triggered in the target audience's mind. For example, in the destination marketing exhibition in mainland China, HK exhibitors should avoid co-presence with their Macao counterparts or should have their booths distanced from Macao booths. As mainlanders who strongly recognize their Chinese identity can better adapt to the tension with HK hosts, this market segment (i.e., strong Chinese identity) should be prioritized and the channels that can effectively reach this segment should be sourced for marketing purposes.

Limitations and Future Studies
The implications should be considered with several limitations. First, this study was conducted with a focus on the constitutional relationships among mainland China, HK, and Macao. While THIR provides insights into the identity-based risk perceptions and its implications could be applied in destinations with identity tensions, future research should note that modifications of the scale items are needed to fit it into context. For example, given the reported cultural differences between Catalans and their Spanish compatriots (Elwes 2021), future research can examine Spanish visitors' THIR against Catalonia by adapting our scale to their context. THIR should not be limited by the country border. The rise of anti-Russian hate because of the Russia-Ukraine war (Adam et al. 2022) may make THIR salient among Russians who plan to travel to Europe. Future research on THIR can target Russian respondents. Second, although we conceptually posit the directional paths between constructs in the serial mediation, the cross-sectional survey design of this study is not sufficient to infer their causal relationships (Fong 2022). Future research should use other research designs such as experiment, or a longitudinal approach to examine causality. Longitudinal approach can also be used to investigate how THIR and its effects on the consequences vary with time. Third, the survey respondents tended to be young and had completed a Bachelor's degree; the implications may not work for the older population or for mainlanders with a lower educational level. This sample limitation needs to be addressed in future studies. Finally, people can easily access the news and information about a place. The information enables them to estimate the THIR even they have never visited the destination. For these potential first-time visitors, their THIR effects on the consequences may differ from their counterparts who have visited the destination before. Future research can make a comparison of these two types of tourists.