The role of perceived religiosity in strengthening celebrity endorser’s credibility dimensions

Abstract This present study investigates the moderating effect of celebrity endorsers’ perceived religiosity in the relationship between dimensions in source credibility theory and consumers’ purchase intention. A self-administered survey was conducted using mock-up stimuli to 505 respondents in major universities located in Jakarta, Indonesia. Subsequently, exploratory and confirmatory factor analysis was employed, followed by regression analysis using AMOS 24.0 to test the hypotheses. The results revealed that endorsers’ perceived religiosity significantly strengthened the influences of trustworthiness and attractiveness dimensions toward consumers’ purchase intention. Conversely, perceived religiosity did not significantly strengthen the effect of expertise on consumers’ purchase intention. Theoretically, this present study fulfills the knowledge gap of the influence of perceived religiosity construct on the relationship between source credibility dimensions and consumers’ purchase intention. While practically, this present study provides suggestions for marketers to involve the celebrity’s perceived religiosity dimensions to increase their endorsement campaigns’ credibility and effectiveness.


PUBLIC INTEREST STATEMENT
The use of celebrities to endorse products has gained popularity among marketers to advertise their brands. It aims to create a differentiation via emotional attributes as it is now more challenging to be stand out in a market that offers a similar product attribute. Nonetheless, the use of religious construct remains less investigated. Hence it provides an opportunity for marketers to employ that construct in strengthening the messages conveyed. It is an alternative strategy to develop a more stand-out yet emotionally attached endorsement, which is crucial to creating stronger brand equity. This present study has proved a significant improvement in the credibility dimensions of endorsement campaigns by employing a robust religiosity scale. Eventually, it strengthens the consumers' purchase intention. Therefore, it gives insights for marketers to consider the religiosity construct in developing their marketing campaign.

Introduction
The use of celebrities to endorse products has gained popularity among marketers to advertise various products, services, or brands. As it is now increasingly difficult for firms in competitive markets to differentiate their brands via rational attributes, such as quality and reliability (Schimmelpfennig & Hunt, 2019), marketers turn their attention to focus on what kind of person would buy the product rather than on the product itself (Baran & Blasko, 1984;Goodyear, 1996). Celebrity endorsement is a vital marketing communication strategy that creates a competitive differential advantage for a firm's products or services and shapes positive effects in the consumers' minds (Erdogan, 1999). It also creates an emotional attachment with the brand (Srivastava & Sharma, 2016), and provides more brand acknowledgment among their target consumers (Singh & Banerjee, 2018).
The endorsement campaign has been applied in various areas such as entertainment, sports, politics, and religion (Frimpong et al., 2019). Endorsement campaigns could take forms as through social media and internet communication, point of purchase, or other traditional mass advertising such as outdoor or printed media. Religion is considered as an important factor for understanding the consumer's attitudes and behaviors (Bachleda et al., 2014;Minton et al., 2015). Religion influences a wide variety of consumer behaviors (Hirschman, 1981), and consumers tend to express the intensity of their beliefs through consumption choices (Cosgel and Minker, 2004). Researchers have become increasingly interested in the topic of religion along with the increasing use of a religion-based communication approach to effectively fulfill the customer's needs (Bilgin & Nakata, 2016). However, surprisingly, few studies have investigated the role of religious belief in the consumer's behavior field (Abu-Alhaija et al., 2018;Bachleda et al., 2014), as researchers are advised to investigate further regarding the relationship between religiosity constructs and various marketing variables (Tang & Li, 2015). Similarly, the author could not find any study that examined the role of religiosity constructs in the relationship between celebrity endorsement and consumers' purchase intention (PI).
Source credibility theory (SCT) and its dimensions which involve trustworthiness (TR), expertise (EX), and attractiveness (ATR), have become a prominent lens to study the impact of endorser campaigns in pursuing buying behavior (Arai et al., 2014;Bergkvist et al., 2015;Cuomo et al., 2019;Erdogan, 1999;Frimpong et al., 2019;McCracken, 1989;McGuire, 1969;Spry et al., 2011;Von Felbert & Breuer, 2020;Wang & Scheinbaum, 2017). These studies found that SCT variables have a different degree of significance toward purchasing behavior. For instance, the endorser's perceived expertise does not significantly influence brand attitude and PI (Bergkvist et al., 2015;Von Felbert & Breuer, 2020). While Spry et al. (2011) found that endorser's credibility only indirectly impacts brand equity when mediated by brand credibility. Hence, it is argued that attitude towards endorsers represented by TR, EX, and ATR could not replace the overall factors that influence consumers' purchasing behavior (von Felbert and Breuer, 2021). Mueller et al. (2001) argued that religiously-committed people tend to show a higher level of participation in religious activities. People with a higher level of perceived religiosity (PR) are reported to have higher trust among their peers (Furrow et al., 2004). Similarly, Petersen (1994) argued that if the individual is not seen to participate in the religious network, his/her network beliefs are not reinforced or confirmed by others. Thus, the beliefs eventually lose credibility. Fleischmann and Bohn (2016) argued that people who are more willing to involve in religious service convey a more credible personality. Pace (2014) asserted that religiosity is a vital instrument to connect with other people. In particular, this connection is formed through the degree of likeability as consumers will strongly relate attractiveness and people's religiosity (Feldman & Crandall, 2007). In this sense, building a connection between a brand with the target consumers by employing endorser's PR may strengthen the messages conveyed by marketing campaigns. Therefore, this present study attempts to integrate the role of PR in the context of endorser's credibility and its relationship toward consumers' PI.
Heeding the suggestion from Tang and Li (2015), this present study adds PR as a moderator in the relationship between SCT dimensions and PI. The belief into action (BIAC) scale was employed as it is argued to be a valid measure of religious commitment across a wide range of religious beliefs. BIAC also offers a more straightforward assessment of people's religious commitment by referring religious belief into terms of action (Koenig et al., 2015). This present study focuses on investigating two types of relationships. Firstly, it examines the direct influence of TR, EX, ATR toward PI. Secondly, it examines the role of BIAC as a PR construct in moderating the relationship between TR, EX, ATR, and PI.
Hence, this present study aims to address these questions: (1) How do the effects of SCT dimensions (i.e., TR, EX, ATR) toward PI?; and (2) How do the moderation effect of PR in the relationship between TR, EX, ATR, and PI? Theoretically, this study contributes to bridging the knowledge gap of the interplay between endorser's PR,SCT dimensions, and consumers' PI which may impact the overall effectiveness of a celebrity endorsement campaign. Practically, this study provides insight for marketers, especially in strengthening their endorsement campaign by involving a factor that represents celebrity's religiosity.
The later parts of this paper were organized as follows: in the following section, some relevant literature was described with the theories needed to develop the hypotheses. Subsequently, the research methodology was described, covering the measurement, data collection, and analysis steps. Next, the findings were presented, and some highlights were discussed. The last section is the conclusion, theoretical and practical implications, as well as limitations and directions for further studies.

Source credibility theory
The primary focus of extant studies on celebrity endorsement has been determining what endorsers' traits account for the effectiveness of endorsement. As one prominent theory in celebrity endorsement studies, SCT postulates that people or receivers are more likely to be persuaded when the source presents itself as credible (Hovland & Weiss, 1951). O'Keefe (2002) argued that credibility is not an intrinsic characteristic of a source; instead, it emphasizes the credibility of the communicators. The judgments made by receivers regarding the communicators take an essential role regarding the message contents itself. A more credible source of information is preferred and is also more persuasive (McGuire, 1969). SCT involves a complex construct and is generally believed to consist of TR, EX, and ATR (Fogg, 2003;McGuire, 1969;Shimp, 2000;Willemsen et al., 2012). Nevertheless, that three dimensions reflect distinct aspects of credibility and do not necessarily go hand in hand. Therefore, it is possible to measure them separately to examine the individual effect on persuasive outcomes (Fogg, 2003;O'Keefe, 2002).
Trustworthiness refers to people's assessment of whether the communicator will likely be inclined to tell the truth (O'Keefe, 2002). In the context of the producer-consumer relationship, trustworthiness refers to the benevolence and integrity of producers or their endorsers to care and act in the consumer's interest (Xiao & Benbasat, 2007). Trustworthiness is essential since it conveys integrity and believability towards an endorser, which implies a brand is likely to deliver as promised (Arai et al., 2014). When a consumer believes that a source is trustworthy, he/she also assumes that the conveyed message is highly believable (Hovland & Weiss, 1951).
Expertise is defined as the degree of perceived understanding, skills, and knowledge that the endorsers have and directly influences the level of conviction to persuade consumers to purchase the endorsed products (Hovland et al., 1953). Unlike trustworthiness, which refers to a more general term, perceived expertise refers to the degree to which a source is considered capable of making valid assertions (McCracken, 1989) and is considered context-specific (Bergkvist et al., 2015). For example, only golf apparel manufacturers may suitably use Tiger Woods as their endorser. Mayer et al. (1995) supported this argument by arguing that expertise is the ability of a source to have an influence only in a specific domain.
Meanwhile, attractiveness as the third dimension of SCT refers to the degree of likability of endorsers by the targeted consumers; and is essential in measuring the effectiveness of the messages (McGuire, 1969). People are more likely to be influenced by the source when they desire to identify with it (Li & Yin, 2018). They are more likely to identify with people they find likable (Kelman, 1961). Hence, attractive people are likely easier to persuade people to act in the intended behavior.

Perceived religiosity
Religion is one of the universal social institutions that influence people's attitudes, values, and behaviors (Mokhlis, 2009). It is a part of the culture that shapes people's behavior (Kotler, 2000) and plays a significant role in sculpting social behavior (Mansour & Diab, 2016). Religiouslycommitted people regard religious teachings according to their values. Thus, they oblige the religious rules and do not adjust their beliefs according to their interests (Pace, 2014). Koenig (2001) referred religious commitment as how much individuals are involved in their religions. Worthington Jr. et al. (2003) added that a religiously committed person is supposed to adhere to his/her religious values, beliefs, and practices, and more importantly, use them in daily living. Allport (1963) introduced the concept of intrinsic religiosity and extrinsic religiosity into the psychology of religion. Intrinsic religiosity is a dimension of religious involvement in which religion serves as its own end goal. It reflects a person's personal and intimate core value of the self (Pace, 2014). Individuals described by intrinsic religiosity view their religion as the framework for their lives. Thus, they try to consistently live the religion they believe in (Masters, 2013). Extrinsic religiosity refers to social connection. Extrinsic religiosity is a form of involvement in which religion is a means to some other end or goal (e.g., it is an instrumental value) (Gorsuch, 1998). Pace (2014) added that it is an instrument to attain personal goals by connecting with other people.
Existing literature attempted to operationalize the religiosity construct to provide accurate measures of people's religiosity. Allport and Ross (1967) developed the intrinsic and extrinsic religiosity (I/E) scale and further suggested that personal attitude (e.g., my faith involves all my life) reflects someone's intrinsic religiosity. On the other hand, behavioral aspects (e.g., church attendance and involvement in church activities or wearing hijab for Muslim women) are reflections of extrinsic religiosity. Therefore, people with extrinsic religiousness attempt to convey messages toward their social environment to build the perception that s/he is a religious person. People are more likely to perceive people's extrinsic religiosity reflected by their behavior. Additionally, Feldman and Crandall (2007) asserted that the religious image might carry other perceived positive traits, mainly due to the halo effect presence.
However, the existing religiosity measures are criticized since it only reflects the superficial degree of people's religiosity. (Koenig et al., 2015). Similarly, Cutting and Walsh (2008) argued that the existing religiosity scales could not measure the ethical action in society due to the religious experience. It also brings unavoidable ceiling effects in the context of highly religious populations (Koenig et al., 2015). Therefore, BIAC scale was developed and has been incorporated in the studies within different contexts. It is proved to provide valid measures for people's religiosity in diverse religious groups (Hafizi et al., 2016;Koenig et al., 2015;Martins et al., 2020). The basis for content in the BIAC scale is the importance of religion in a person's life, such as how they spend their time, talents, and financial resources related to their religious belief (Koenig et al., 2015). BIAC aims to answer the question: "while people may say they believe, what does that mean in terms of action?" Thus, this scale reflects the actual behavior of "being religious" and could be useful to explain the PR construct in this present study. Ghamisi and Benediktsson (2015) defined consumer behavior as human's specific behaviors that directly participate in the acquisition and use of goods. It includes various decision-making processes that start from marketing stimulus entering consumers' consciousness and going through the psychological process to guide the actual buying behavior. It is considered that intention could be used for predicting behavior. Specifically, in the marketing context, consumers' PI was used to predict buying or usage behavior in practice (Gazley et al., 2015). The term "intention" is defined as the antecedents that occur before purchasing action. It stimulates consumer's behavior to purchase respective products and services (Hawkins & Mothersbaugh, 2010). Menwhile, Halder et al. (2016) regarded PI as a measure of the possibility of a consumer buying certain products (i.e., the consumer's inclination for a product). Therefore, the more consumers intend to purchase, the more likely they will buy the particular products.

SCT dimensions and purchase intention
Consumers tend to reduce the degree of risks associated with their purchases. They tend to seek more information before making any purchase decision (Hirschman, 1981); thus, they rely upon a highly credible source of information. Advertising conveyed by a trustworthy endorser tends to have a higher recall level in consumers' minds (Frimpong et al., 2019). The trustworthy source helps marketers strengthen brand recall and PI towards their target consumers (Spry et al., 2011). Meanwhile, the expertise dimension explains the capability for providing supporting arguments for the audience (Maddux & Rogers, 1980). Communicators with expertise in a particular field are perceived to be sources of valid assertions (Amos et al., 2018). Ohanian (1990) suggested that expert sources influence perceptions of product quality. A marketing campaign that uses an expert celebrity has been more persuasive and generates more intention to purchase the brand. Celebrity endorsers have always been chosen on their attractiveness, either it is their physical appeal or social status (Erdogan, 1999). The attractiveness of the communicators will influence the persuasion consistently (Maddux & Rogers, 1980). This consistent persuasion eventually influences consumers' attitudes regarding particular products and leads to their PI (Yang, 2018). The following hypotheses were proposed: H1 Celebrity endorsers' TR positively influences PI.

Perceived religiosity as a moderator of relationship between SCT dimensions and purchase intention
Existing studies revealed that people perceived as being highly religious, such as often attending religious occasion (e.g., attending church community services for Catholics/Christians and going to the mosque every day for Muslims) are viewed as more trustworthy than non-religious ones (Fleischmann & Bohn, 2016;Tan & Vogel, 2008). Similarly, atheists are viewed to be more distrusted (Gervais, 2011). Chuah et al. (2016) asserted that religious people are also trusted more by people across different levels of religiosity. Consistent with the motivation to express themselves in consumption choice, consumers need to seek a trustworthy source of information. The consumers may evaluate the source's trustworthiness from the endorser's perceived religiosity. Therefore, this study proposed the following hypothesis: H4 Celebrity endorsers' PR moderates (strengthens) the positive influence of TR on PI Religion is considered an important factor that influences one's ethical judgments (Hunt & Vitell, 1993). Religious people tend to have more moral and ethical consideration, and they will likely object when the product's usage bring unethical implications especially if it against their religious belief (Evans, 2011). Reciprocally, consumers also consider the ethical aspect in their consumption regardless of their religiosity level (Arli, 2017). Due to this ethical consideration, religious people tend to evaluate the products before making consumption decisions (Honkanen et al., 2006). Thus, in the endorsement context, they can make valid assertions (McCracken, 1989) regarding a particular product's promised benefit by intensively evaluating its features or ingredients. This study proposed the following hypothesis: H5 Celebrity endorsers' PR moderates (strengthens) the positive influence of EX on PI The attractiveness dimension increases the appeal and influences the buying behavior. It exceeds the physical factors and includes larger dimensions such as intellectual skills, personality properties, and lifestyle (Shimp, 2000). According to Feldman and Crandall (2007), consumers perceived attractive people are more religious, as indicated by a higher religiosity rating for attractive people. This study emphasized the halo effect, which indicates that attractive individuals are perceived as having other unrelated positive traits. Similarly, Zaidi (2015) asserted that people who have a more religious appearance are more attractive than others. The sixth hypothesis was proposed: H6 Celebrity endorsers' PR moderates (strengthens) the positive influence of ATR on PI

Study context
Indonesian marketers have tried to use religious endorsement in communicating their products to the target market either through electronic, outdoor, or social media advertising. They also emphasize the emotional benefits contained in the brand, which are in line with religious teachings (Handriana et al., 2020). Indonesia has a population of 267 million in 2019, of which the three largest religion adherents are Islam (87%), Christian (6.17%), and Catholic (2.62%) (Central Bureau of Statistics, 2020). This study investigates the influence of perceived religiosity on endorsers representing the three largest religions. Therefore, to avoid bias toward one religion, this study deliberately selected respondents representing that three religions.

Preliminary test and stimuli
This study conducted a preliminary test to choose the appropriate endorsers representing Islam, Christian, or Catholic. The preliminary test involved 50 respondents selected from a national university undergraduate student, and they were presented with 30 well-known celebrity figures who extensively endorse brands through social media. Firstly, they were asked to give scores based on the degree of familiarity (1 means very unfamiliar and 5 means very familiar). The three highest scored Islam celebrities were selected as stimuli as well as two and one of Christian and Catholic celebrities respectively. Secondly, they were presented with the six selected celebrities along with the printed mock-up stimuli. The use of mock-up stimuli has advantages, such as repeating the experiment and providing substitutes when the nature stimuli (e.g., the actual celebrity endorsement ads) were unavailable (Lahti, 2015). Nonetheless, mock-up stimuli may also limit the generalizability of the result (Lahti, 2015), as real advertising could take a more complex form and have a different impact on the viewers.The stimuli were presented according to the participants' religion. To avoid the potential effects of particular brands' pre-existing attitudes, this test replaces the existing brand with fictitious ones. The mock-up stimuli displayed juice as a fictitious product as it is a product that does require too much of an evaluation process (such as a car or house) or too little (such as detergents) (Chung & Cho, 2017;Petty et al., 1983). Thus, it minimizes the involvement of complex considerations in choosing the responses. Besides, the mock-up contains some slogans such as "environmentally friendly" as it is closely linked to ethical and spiritual behaviors (Crane, 2001;McGhee & Grant, 2015) (the stimuli materials are presented in appendix A).
Subsequently, the participants were asked to answer the complete questionnaire. The questionnaire was first developed in English and then subsequently translated into Indonesian. The back-translation technique was also applied to check for its linguistic equivalence. The participants were asked to give feedback on the readability of the questionnaire. The wordings were modified according to their suggestions.

Measurement
In total, 29 questions were developed, which consisted of 25 questions regarding the constructs and three questions regarding the respondents' demographic profiles. The questions on the constructs were adapted from validated scales from some previous studies to ascertain content validity (Yang, 2018) and were modified according to this study's context. Five points scale was used for each construct where "1" means "strongly disagree," and "5" means "strongly agree." The questions regarding respondents' demographic profiles covered age, sex, and religion. TR scale consisted of five items, while EX and ATR each contained four items adapted from Ohanian (1990). PI scale contained six items adapted from Liu and Brock (2011). Meanwhile, the PR scale contained five items and was adapted from BIAC validated scale by Koenig et al. (2015).

Data collection
A self-administered survey was conducted in two major universities located in Jakarta, the capital of Indonesia. The targeted respondents were undergraduate students since they are the most active social media users in Indonesia (Simangunsong & Handoko, 2020). Thus, they will likely be familiar with the celebrity endorsers used in the stimuli. To avoid bias responses from only major religions, the respondents were deliberately selected to represent the Islam, Christian, and Catholic adherence. The respondents were approached and briefly explained the background and objectives of the survey. If the respondents agreed to participate in the survey, they were presented with the mock-up and were instructed to keep the most familiar celebrity in mind as they answered the questions. The respondents who did not familiar with any of the celebrities in the mock-up were dropped from the survey. The respondents were also promptly assisted if they faced any confusion related to the questions. After two months, the survey collected 563 responses, and 505 were completely filled and further processed.

Analysis
Following a recommendation from Orcan (2018), exploratory factor analysis (EFA) using SPSS 24.0 was employed since the scales were adapted from other studies conducted in a different context. This procedure allows this present study to examine the structure between factors and their respective indicators. Kaiser-Meyer-Olkin's (KMO) measure of sampling adequacy and Bartlett's test of sphericity tests were employed to assess the data suitability for further factor analysis (Field, 2000). The common-method bias (CMB) was also examined using Harman's single factor test to avoid the systematic error of measurement, leading to a spurious correlation between constructs (Cote & Burkley, 1987).
Subsequently, the covariance based structural equation modelling (CBSEM) technique was employed using AMOS 24.0 since this technique fit the purpose to confirm a theoretically assumed relationship (Marliana & Nurhayati, 2020). Hypothetical constructs were estimated as common factors assumed to cause their indicators (i.e., observed or manifest variables) (Zhang et al., 2020). This study involved large samples (more than 200 respondents); therefore, the central limit theorem could be applied. Hence the data distribution tends to be normal regardless of the shape (Elliott & Woodward, 2007;Field, 2009). Meanwhile, violation of the normality assumption should not cause significant problems, which implies that parametric procedures can be applied even though the data are not normally distributed (Elliott & Woodward, 2007;Pallant, 2007). Therefore, it allows the use of maximum likelihood estimation (MLE) techniques. To assess the reliability and validity of the measurement, the confirmatory factor analysis (CFA) using AMOS 24.0 was employed. This analysis involves convergent and discriminant validity assessments (Straub et al., 2004). Then, MLE regression was employed to assess the hypothesized causal relationships between TR, EX, ATR, and PI. Then the role of PR as a moderator toward the relationship between TR, EX, ATR, and PI was examined. The goodness-of-fit (GoF) indicators were assessed based on the recommendation of Kline (2011), which include the following cut-off value: χ2/df <3, comparative fit index (CFI) >0.90, Tucker-Lewis index (TLI) >0.9, root mean square error of approximation (RMSEA) <0. 06., and standardized root mean square residual (SRMR) <0.08,

Common method bias
Harman's one-factor test was employed to examine the CMB (Podsakoff et al., 2003). All the items were entered into an unrotated exploratory factor analysis. The first component with the largest eigenvalue explained 16.84% of the variance, which does not exceed the recommended cut-off value of 50 percent of all variance (Podsakoff et al., 2003). This result indicated that CMB was not an issue in this study.

Exploratory factor analysis
The varimax rotation principal component analysis (PCA) was employed to extract important factors that represent the constructs. All the items' EFA loading value related to their factors were above the recommended cut-off value for deletion (>0.5) (Karatepe et al., 2005), and no additional factors were present. All the remaining items' communalities value were above the recommended cut-off value of 0.4. Therefore, the remaining items significantly contributed to explaining the variance of their underlying factors (Osborne et al., 2008). Additionally, the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy showed an adequate value of 0.812, while each individual's KMO exceeded the recommended cut-off value of 0.5. Bartlett's test of sphericity also indicated a significant value (p = 0.000). Therefore, all the items and the data were adequate for factor analysis and further processed into hypotheses assessment (Field, 2000). Table 2 displays each item's EFA loading factor, communalities, and KMO values.

Confirmatory factor analysis
CFA was performed to assess the constructs' dimensionality and psychometric properties in this study's context. Standardized loadings, Cronbach's α (CA), composite reliability (CR), and average variance extracted (AVE) were assessed to test reliability and validity. As shown in Table 3, all the standardized loadings values exceeds the recommended cut-off value of 0.5 (Hair et al., 2006). Furthermore, each indicator has an adequate strength correlation with latent variables. Therefore, they adequately reflect the respective construct. Both CA and CR values of all constructs are above 0.7, which exceeds the minimum recommended cut-off value of 0.6 (Hair et al., 2006). The internal consistency level for each construct was acceptable; therefore, each set of items did measure the intended construct. Table 4 also indicates that the AVE values of all four constructs exceed the cutoff value of 0.5. Therefore, convergent validity was assured (Hair et al., 2006). Each set of items were strongly related and did measure the same constructs. Subsequently, the discriminant validity test was conducted by comparing the square-root of AVE with corresponding correlation values with other constructs. As shown in Table 4, the square-root of AVE exceeds the correlation value with other constructs. It implies that the discriminant validity was assured. Therefore, the items relate more strongly to their construct than to other constructs (Hair et al., 2006). The GoF assessment also shows acceptable values with χ2/df = 1.548, CFI = 0.979, TLI = 0.977, RMSEA = 0.039, and SRMR = 0.027.

Regression analysis
As seen in Table 5, TR has a significant and positive influence on PI (B = 0.54,β = 0.492,p < 0.005, r 2 = 0.37). Additionally, ATR also has a significant and positive influence on PI (B = 0.41, β = 0.392, p < 0.025,r 2 = 0.243). Conversely, EX does not significantly influence PI. In the moderation analysis, PR was found to significantly strengthen the positive influence of TR on PI as indicated by increasing value of its estimates and coeffiecient of determination (B = 0.54,β = 0.492,r 2 = 0.415;

Discussions
This study's objectives were to determine the influences of TR, EX, ATR toward PI and investigate the role of PR as a moderator on the relationships between those predictors and the outcome in the context of celebrity endorsement. This study finds that TR has a positive influence on the PI. This result confirms the findings from other studies (e.g., Frimpong et al., 2019;Von Felbert & Breuer, 2020;Wang & Scheinbaum, 2017), which found that endorser's TR was a significant predictor of PI. Consumers will likely purchase the product since they believe that the endorsers convey truthful messages regarding the product's offered value and benefit. Therefore, the consumers believe that they will get the expected benefit from the products they purchase.
This present study also finds that endorsers' ATR has a positive influence on the PI. This study confirms the findings from previous researches (e.g., Cuomo et al., 2019;Frimpong et al., 2019;Yang, 2018). Attractive endorsers will likely persuade people to purchase the endorsed brands. Meanwhile, consumers also tend to identify with people they find likable (Kelman, 1961). In terms of effect size, TR has a higher contribution in influencing PI, which suggests that endorser's TR plays a more vital role in enhancing consumers' PI rather than ATR. This result is in line with finding from Wang and Scheinbaum (2017), which argued that endorsers' TR is the most crucial factor in enhancing consumers' PI through brand credibility.
Conversely, this present study found that endorsers' EX does not significantly influence PI. This finding supports the argument from previous studies (e.g., Bergkvist et al., 2015;Mayer et al., 1995), which argued that expertise should be considered a context-specifics only capable of bringing assertions within a specific domain. This present study deliberately uses stimuli that involve a product within the low-involvement category, aiming to minimize the respondents' complex consideration in responding to the questionnaire. Therefore, the respondents probably perceive no need for celebritys' specific expertise to make valid recommendations regarding the product used in the stimuli.
Next, this present study adds the PR as a moderator on the relationship between endorsers' TR, EX, ATR, and PI. Echoing the arguments from some previous studies (e.g., Fleischmann and Bohn (2016) and Pace (2014)) that religiosity contributes to increasing the people's credibility, this study finds that PR significantly strengthens the influence of TR and ATR dimensions toward PI. Specifically, the predictive power of endorser's TR and ATR toward PI increases along with the presence of PR as indicated by the higher value of the estimates and the coefficient of determinations. This finding also supports Chuah et al. (2016), which suggested that religious people are more trusted. Additionally, the PR increases the endorsers' appeal, which strengthens their influences on consumers' PI. This finding confirms the arguments from Zaidi (2015), which asserted that people with more religious appearance are perceived to be more attractive. Conversely, this study revealed that PR does not strengthen or contribute to EX and PI's relationship. Therefore the respondents do not perceive as "being more religious" contributes to the endorsers' perceived expertise regarding the products' value and benefit.
Despite the positive findings that confirm the role of PR in strengthening the relationship between TR, ATR, and PI, it is worth noted that this present study did not involve real brands nor real celebrity endorsement campaigns. Therefore, even though celebrities were perceived as religious figures, marketers should take other aspects, such as the product-figure congruence, into further consideration. Furthermore, the respondents were deliberately interviewed using mock-up material to help focus their attention toward chosen endorsers. In fact, consumers are exposed to an excessive amount of advertisings daily. Hence, it may obstruct the key messages and hinders the religious image building toward the celebrity endorser.

Conclusions and implications
In conclusion, this present study confirms the findings from previous researches that revealed the importance of SCT dimensions toward consumers' PI in the context of celebrity endorsement. Nevertheless, in the context of this present study, only TR and ATR dimensions significantly influence the consumers' PI. This study also adds religiosity constructs represented by PR as a moderator and is proved to significantly strengthen the influence of the endorsers' TR and ATR towards PI.
Theoretically, this study contributes by offering empirical evidence to the basic tenets of SCT in shaping the communicators' credibility. Additionally, the role of religiosity construct was added in the context of celebrity endorsements. This study complements previous literature on celebrity endorsement by incorporating the PR and has revealed its importance in strengthening the endorser's credibility dimensions, which is critical in developing effective endorsement campaigns.
Practically, this study provides insights for marketers to take religious consideration in choosing the celebrity in their endorsement campaign. As the main finding of this study, PR strengthens the TR and ATR attributes which eventually increase their influence on consumers' PI. Marketers should carefully assess celebrities' religious behavior (e.g., the frequency of attending the religious occasion, going to mosque or church, or their affiliation with religious communities) and the possibility of the target consumers' perception regarding the celebrities' perceived religious behavior. Marketers may examine the celebrities' activities through online and offline media and conduct interviews with the target consumers regarding their possible perception of the celebrities' religiosity. By carefully assessing the contribution of this religiosity perception, marketers can develop a more persuasive and effective marketing campaign using celebrity endorsement.

Limitations and direction for further studies
This present study has several limitations that future research may address. Firstly, the mock-up stimuli only contain a product from a low-involvement category. Hence, further study may use high-involvement categories to examine the possible effects of SCT dimensions and PR in this product category. Also, the use of a fictitious brand could impinge the consumers' perception and may result in a different degree of influences compared to an actual brand. Further study can also compare the SCT dimensions and PR effects across different product categories. Secondly, this present study only incorporated one construct of religiosity. Further studies may investigate the different constructs of religiosity (e.g., intrinsic vs extrinsic religiosity) and compare their effects on consumers' PI. Further studies may also compare the difference influence between religions; hence marketers can develop the endorsement campaign based on the religious group segment they intend to target. Additionally, this study only involves a narrow age group (17-24 years old). Thus, future researchers may wish to examine the effect of the predictors across a wider group of ages.