The E ﬀ ect of Natural Celebrity–Brand Association and Para-Social Interaction in Advertising Endorsement for Sustainable Marketing

: Celebrity endorsement has been regarded as an e ﬀ ective strategy for enterprises to implement sustainable marketing strategies. However, frequent use of celebrity ads renders consumers skeptical of the ads’ and celebrity’s proﬁt-making intent, which may adversely a ﬀ ect the sustainable marketing of the brand. This has given rise to “natural celebrity-brand association” that features celebrities using the brand in real-life settings, which is usually presented on social media rather than mass media. Using a boot-strapping approach, the study contrasts the e ﬀ ects of natural and commercial endorsements (i.e., natural vs. advertising, natural vs. product placement) on consumer brand responses. Results showed that natural celebrity–brand association exerts superior e ﬀ ects. Further, the mediating variables, para-social interaction (PSI) and its drivers (celebrity attractiveness, consumers’ perceived homophily with the celebrity), which reﬂect consumer’s emotional connection with a celebrity, also exert signiﬁcantly stronger e ﬀ ects in the natural (vs. commercial) endorsement context. This research provides insights for advertisers and marketers in exploring new patterns of brand presentation forms in advertisements and gaining competitive advantage in sustainable marketing. show that attractiveness exerts a signiﬁcant mediating e ﬀ ect in the contrast between advertising and real-life context on all of the brand response measures: Brand attitude ( β = 0.05, P M = 0.31, SE = 0.02, 95% CI (0.01, 0.11)), self–brand connection ( β = 0.04, P M = 0.19, SE = 0.02, 95% CI (0.01, 0.10)), purchase intention ( β = 0.05, P M = 0.27, SE = 0.02, 95% CI (0.01, 0.11)), and perceived brand quality ( β = 0.04, P M = 0.21, SE = 0.02, 95% CI (0.01, 0.10)). Attractiveness also exerts a signiﬁcant mediating e ﬀ ect on the contrast between product placement and real-life context on these variables: Brand attitude ( β = 0.12, P M = 0.44, SE = 0.04, 95% CI (0.05, 0.22)), self–brand connection ( β = 0.10, P M = 0.22, SE = 0.04, 95% CI (0.04, 0.20)), purchase intention ( β = 0.09, P M = 0.28, SE = 0.04, 95% CI (0.03, 0.19]), and perceived brand quality ( β = 0.13, P M = 0.35, SE = 0.05, 95% CI (0.05, 0.24)). Regarding the mediating e ﬀ ect of homophily, results show that in the contrast between advertising and real-life context, it exerts signiﬁcant e ﬀ ects on all of the brand response measures (brand attitude: β = 0.05, P M = 0.31, SE = 0.02, 95% CI (0.01, 0.09), self–brand connection: β = 0.08, P M = 0.35, SE = 0.03, 95% CI (0.02, 0.14], purchase intention: β = 0.04, P M = 0.24, SE = 0.02, 95% CI (0.01, 0.08), and perceived brand quality: β = 0.05, P M = 0.26, SE = 0.02, 95% CI (0.01, 0.10)). Homophily also exerts signiﬁcant mediating e ﬀ ect in the contrast between product placement and real-life context on these variables: Brand attitude ( β = 0.13, P M = 0.47, SE = 0.04, 95% CI (0.05, 0.22)), self–brand connection: ( β = 0.21, P M = 0.47, SE = 0.07, 95% CI (0.09, 0.35)), purchase intention: ( β = 0.13, P M = 0.40, SE = 0.05, 95% CI (0.06, 0.23)), and perceived brand quality: ( β = 0.13, P M = 0.34, SE = 0.04, 95% CI (0.06, 0.23)). Turning to consumer–celebrity PSI, results show that in the contrast between advertising and real-life context, PSI exerts signiﬁcant mediating e ﬀ ect on brand attitude ( β = 0.09, P M = 0.57, SE = 0.03, 95% CI (0.04, 0.16)), self–brand connection ( β = 0.12, P M = 0.54, SE = 0.03, 95% CI (0.06, 0.19)),


Introduction
Celebrity endorsement is a popular sustainable marketing strategy that can increase brand awareness, brand image, consumer attitude, and purchase intention. Nevertheless, frequent exposure to celebrity ads may heighten the consumer's awareness of their underlying commercial purposes, leading to a higher level of skepticism [1] or even disgust [2]. To counter the potential negative effects of celebrity ads, marketers have devised alternative strategies such as product placement to provide a more neutral context to disguise the underlying commercial purposes of celebrity endorsement.
Recently, the need to conceal the purpose of celebrity advertising has given rise to a new format dubbed "natural celebrity endorsement". Rather than featuring the celebrities in studio poses that feature the celebrity and advertised product in their best light, natural celebrity endorsement features celebrities in real-life settings as they hold, use, or otherwise engage with the advertised brand. While some of these settings may be planned, many others are consumers' real encounters with the celebrity eating/wearing/using the brand. These activities are facilitated by fans, who follow closely the celebrity's whereabouts and then spread the photos across social media [3]. Regardless of the planned/unplanned nature of the encounter that consumers may have difficulties telling apart, natural endorsement is perceived as more authentic, thus increasing the endorsement's acceptability [4]. Recently, the brands Everlane, Christian Louboutin, Cartier, and Alexander McQueen have seen markedly increased internet searches and sales traceable to natural celebrity endorsement [5]. While scholars [6] are beginning to test the impact of this style of celebrity endorsement, studies in this area remain lacking.
In China, celebrity endorsement is a highly popular strategy, with over 50% of ads featuring one or more celebrities. This is well above the 25% in the U.S [7]. On the one hand, the popularity of this strategy speaks to the sustainable marketing strength of celebrity endorsement in the Chinese market. On the other hand, the extensive use of celebrity advertising could lead to over-exposure and consumer fatigue. In the worst-case scenario, celebrities with limited knowledge of the endorsement (e.g., a non-user of a product) may end up endorsing under-qualified products and services, causing damages to both the endorser and consumers [8]. Related past events have led to a social call for more truthfulness and transparency in celebrity endorsement. Regulators have now mandated the amended advertising law that a celebrity needs to have experience with a product or service before he/she can endorse it [8].
As celebrity endorsement as a genre is rebuilding its credibility and relationship with consumers in China, natural celebrity endorsement provides one such means.
This study aims to assess the impact and drivers of natural celebrity endorsement. To clarify these issues, the study will compare natural celebrity endorsement with paid advertising and product placement, both of which carry salient commercial purposes. We posit para-social interaction (PSI) as a major explanatory factor of natural endorsement effects. PSI refers to the one-way interaction between consumer and media figures. When a consumer engages in PSI with a celebrity, he/she extends emotional energy, interest, financial, and time investment to the celebrity and in return, he/she receives a sense of intimacy and friendship as well as psychological satisfaction [9,10]. Not surprisingly, PSI has been shown to be a crucial factor in traditional celebrity advertising [11,12]. The effect of the celebrity's perceived attractiveness, a known driver in traditional celebrity ads, will also be examined in a natural endorsement context. This paper examines the following issues: (1) Compare the levels of consumer-celebrity PSI and the degrees of its drivers, when consumers are exposed to different endorsement contexts; (2) compare the level of positive brand responses when consumers are exposed to different endorsement contexts; and (3) test the role of PSI and its drivers in mediating the effects of endorsement contexts on brand responses. Results of the study will enhance our understanding of the impact of diverse endorsement contexts and their underlying processes. This, in turn, would facilitate marketers to delineate the salient drivers, when they engage in advertising planning involving different endorsement contexts.

Contexts of Celebrity-Brand Association and Consumer Brand Attitudes
Consumers and advertisers have been quite familiar with the endorsement process within traditional advertising contexts, including formal and explicit endorsement in formal advertisements where the celebrity endorses the brand/product through verbal endorsement, and implicit endorsement where the celebrity uses or just appears with the brand/product [13][14][15]. Endorsement effects in traditional contexts with profit tendencies have been paid much attention by academia [7,[16][17][18]. In addition to commercial contexts, which have always been the mainstream form of conveying persuasive messages, the unpaid celebrity endorsement, wherein associations between celebrity endorsers and brands are displayed in real life scenes, have started to gain some attention [19,20].
Advertisers have noticed that paid endorsement with explicit commercial purposes in normal advertising may cause consumers to dislike the selling behaviors thus enhancing consumers' negative attitudes towards a brand [2]. Even product placement in reality shows or TV series, which seemed more natural and fluent in conveying the symbolic meaning of a brand to an audience, might still irk audiences. For example, Homer [21] found that consumers had negative impressions of placement advertisements. Such an evaluation might be because placement advertisements were usually repetitive and in obvious places [22]. This might interfere with audiences in fluently viewing normal media content, which in turn contributes to consumers' negative attitudes towards the brand in the placement advertisements [23].
Celebrity-brand association with no obvious profit orientation may help increase consumers' favorable impression of a brand/product as consumers are increasingly regarded as beacons guiding sustainable marketing means [24]. For example, the endorsement would be more effective if the audience felt that the celebrity endorser had really used the brand prior to endorsing the brand [20]. Consumers would have more positive attitudes towards the brand if they felt that the celebrity endorsed the product because of recognition of a brand rather than for commercial profit [16].
In the social media era, consumers can readily gain access to aspects of the celebrity's private life [25]. Many consumers have smartphones and are connected on social media. They can take a photograph of a celebrity on the street, and then upload and viral it on social media. This practice allows consumers to learn firsthand what a celebrity actually eats, wears, and uses in the course of his/her daily life. It thus blurs the line between a celebrity-user and a celebrity-endorser [26].
Information related to products appearing together with celebrities spreads quickly through social media enabling natural celebritybrand associations to gain attention of potential users [27,28]. The consumer-generated brand messages disseminated through ways of electronic word-of-mouth (eWoM) communication is considered as a more convincing source of brand information than market-generated advertising. Escalas [29] studied the difference between the narrative produced by participants involved in the brand story of the real world against the narrative in commercial advertising, and found that the former had a more positive influence on consumers' SBC than the latter. Marwick and Boyd [30] found that natural celebrity-brand practice has begun to become popular on social media. Jin and Phua [28] pointed out that there should be more positive relationships between consumers and brands in the context of natural celebrity-brand associations than through commercial endorsement as the former seemed more credible. Van Laer et al. [31] indicated that consumers with relatively high degrees of emotional attachment to a celebrity tended to hold higher expectations of the celebrity to use the brand in the real world. Russell and Rasolofoarison [6] tested the effect of natural celebrity-brand association on consumer's perceived credibility of the celebrity and draw the conclusion that celebrity was regarded more credible when appearing together with a brand/product in real-life scene compared with commercial endorsement situations.
Taken together, these studies suggest that the different endorsement contexts may yield different consumer responses, with natural celebrity endorsement having an advantage over commercial celebrity endorsement on its persuasive brand effects. This study proposes the following hypotheses, extending the investigation into the impact of endorsement contexts on brand-related responses (e.g., purchase intention) salient to marketers: Hypothesis 1 (H1). Compared to consumers exposed to advertising or product placement, consumers exposed to a natural endorsement context will hold significantly more positive brand responses, including self-brand connection (H1a), brand quality (H1b), brand attitude (H1c) and purchase intention (H1d).

Contexts of Celebrity-Brand Association and Consumer Perceived Relations with Celebrity
Para-social interaction (PSI) is a one-way, non-reciprocal relationship that consumers form with publicly renowned persons. In the advertising context, this imaginary interpersonal relationship refers to the consumers' felt emotional intimacy with the celebrity and it is real in the consumers' mind. Studies on the effects of word-of-mouth communication have shown that consumers are more prone to adopt brands or products recommended by individuals they know. Thus, PSI enhances celebrity endorsement effects [3,12,32], especially in accordance with the growing choice initiatives of consumers in deciding their consumption behavior.
Studies on PSI have not only examined its consequences, but also its antecedents. It has been proved that the closer the media content was to reality in audiences' perceptions, the more frequent audiences expose to the media, which led to stronger audience perception of PSI with media figures [33,34].
Social media platforms such as WeChat and micro-blogs in China and Facebook and Twitter globally have relaxed people's safeguard for privacy and many would disclose their ideas and feelings to people they do not know well, and even strangers, on these platforms. Celebrities in particular are using social media platforms to actively interact with stranger-fans. They provide content that is not only entertaining, but that discuss their daily life, political views, personal thoughts, and feelings. Through these enhanced interactions and self-disclosure, consumers are getting the celebrity's true voice rather than a media depiction of the celebrity [35]. This allows consumers to perceive a closer PSI with the celebrity.
Scholars [36][37][38][39] pointed out that the similarity between audiences and media figures, especially similarity in terms of attitudes and appearance, all had a positive effect on the strength of PSI. As celebrities were often regarded as a reference group aiding the construction of the self [29], once consumers found that their own personality characteristics were similar to the brand traits used by some celebrity, they would feel that their self is confirmed by the reference group [40]. The perceived similarity or homophily of consumers on the celebrity reflected through using certain brand/product might arouse consumer's willingness to get close to the celebrity to gain the self-perception and self-identity [41]. It has been shown that people tended to get along with "like-minded" people, and the sense of group identity with such people drove individuals to form positive attitudes and a willingness to communicate with group members [42].
A situation in which a celebrity uses a brand/product in real life, with no obvious profit motive, may be more effective at building consumer trust on the brand/product [6,7]. To some extent, this function of increasing trust of natural celebrity-brand association is similar to the function of PSI in reducing audiences' sense of uncertainty [43,44]. In the context of PSI-the longer the awareness of, and the deeper degree of familiarity an audience has with a media figure, uncertainty is reduced and the audience would perceive higher degrees of intimacy with the media figure, increasing the intensity of PSI [43,44]. People may trust the persuasion messages of celebrities with whom they have PSI in the same way as trusting the suggestion of a friend. What is more, consumers' trust on brand/product in real-life situation may also arouse consumer's closeness and resonance with the brand, helping form a kind of brand association, which is the basis of SBC [40]. Consuming certain brands/products used by celebrity in normal life, especially the brands/products with strong sign-value, helps consumers to express their personality and distinguish themselves from others [29,45,46]. That is how the PSI facilitated consumer-celebrity relation to be converted to consumer-brand connection in endorsement process.
A quantitative review of the celebrity endorsement literature identified the celebrity's perceived attractiveness as one of the most salient celebrity traits that exerts endorsement effects [7]. An attractive/ likeable celebrity draws viewer attention, lowers the viewer's defensive barriers, reduces his/her likelihood to counter-argue an ad message [47], and allows the endorsed brand to take on positive emotions [15]. Indeed, most studies have found that there is a positive correlation between celebrity attraction and brand evaluations [17], thereby providing evidence that this affective trait complements the celebrity's cognitive traits (e.g., source credibility) to enhance endorsement effectiveness.
Although the literature has not specifically examined the relationship between endorsement context and perceived attractiveness, it has noted that some consumers do not like celebrity advertising when they feel the celebrity is doing it for the money [12]. Thus, showcasing a celebrity in a natural context without obvious profit orientation may bring about more favorable consumer impressions.
Taken together, consumers may have good impressions and sense of similarity/homophily with a celebrity if they are exposed to the celebrity-brand association in a real-life context. Hence, we propose the following hypotheses: Hypothesis 2 (H2). Celebrity attractiveness (H2a) and consumer perceived homophily with celebrity (H2b) are drivers for consumer-celebrity PSI.

Hypothesis 3 (H3).
Consumer's PSI with celebrity (H3a) as well as its influencing factors, namely celebrity attractiveness (H3b) and consumers' perceived homophily with celebrity (H3c), are at higher levels when consumers are exposed to the celebrity-brand association in real-life context than in commercial endorsement contexts (advertisement or product placement).
As discussed earlier in the theoretical framework, the endorsement literature has provided evidence that consumer-celebrity PSI can exert persuasive brand responses [32]. The current study builds on past studies and hypothesized that compared to commercial endorsement contexts, a natural context can exert direct effects and enhance brand responses (H1), PSI, and its drivers (H2 and H3). Putting these strands of research together, we further posit that PSI and its drivers can mediate the relationship between endorsement contexts and brand responses. Thus, Hypothesis 4 (H4). Consumer's PSI with celebrity (H4a) and its drivers, including celebrity attractiveness (H4b) and consumers' perceived homophily (H4c) with celebrity, mediate the relationship between contexts of celebrity-brand association and consumer's brand-related responses.
According to the hypotheses, the theoretical model of this paper is designed as Figure 1.
Sustainability 2020, 12, x FOR PEER REVIEW 5 of 20 consumers are exposed to the celebrity-brand association in real-life context than in commercial endorsement contexts (advertisement or product placement).
As discussed earlier in the theoretical framework, the endorsement literature has provided evidence that consumer-celebrity PSI can exert persuasive brand responses [32]. The current study builds on past studies and hypothesized that compared to commercial endorsement contexts, a natural context can exert direct effects and enhance brand responses (H1), PSI, and its drivers (H2 and H3). Putting these strands of research together, we further posit that PSI and its drivers can mediate the relationship between endorsement contexts and brand responses. Thus,

Hypothesis 1 (H4). Consumer's PSI with celebrity (H4a) and its drivers, including celebrity attractiveness (H4b) and consumers' perceived homophily (H4c) with celebrity, mediate the relationship between contexts of celebrity-brand association and consumer's brand-related responses.
According to the hypotheses, the theoretical model of this paper is designed as Figure 1.

Materials and Methods
An empirical study realized through an online questionnaire was designed to examine the hypotheses. We created three different celebrity-brand associations to compare the effects of commercial endorsements and natural celebrity-brand association on consumer's responses to the celebrity and the brand.

Product Selection
The pretest aims to select a product that carried both high utilitarian and high symbolic values in consideration of reflecting the celebrity appearance, self-brand connection, and brand quality. Fifty respondents (29 females, 21 males; age 16 to 69) joined the pretest. They were given a list of four products (mobile phone, bottled mineral water, facial cleanser, and facecloth). Using a sentence completion task adapted from Mittal and Lee [48], the respondents were instructed to select one (or more) products from the list to complete the sentences, thus indicating the respective products' utilitarian and symbolic values. Multiple responses were allowed. Results in Table 1 showed that bottled mineral water had the highest utilitarian (133 out of 150 possible mentions) and symbolic (97 out of 150 possible mentions) values, followed by mobile phone (utilitarian: 127; symbolic: 81). Thus, bottled mineral water was selected as the stimulus. Please refer to Annex A for details of the pretest results.

Materials and Methods
An empirical study realized through an online questionnaire was designed to examine the hypotheses. We created three different celebrity-brand associations to compare the effects of commercial endorsements and natural celebrity-brand association on consumer's responses to the celebrity and the brand.

Product Selection
The pretest aims to select a product that carried both high utilitarian and high symbolic values in consideration of reflecting the celebrity appearance, self-brand connection, and brand quality. Fifty respondents (29 females, 21 males; age 16 to 69) joined the pretest. They were given a list of four products (mobile phone, bottled mineral water, facial cleanser, and facecloth). Using a sentence completion task adapted from Mittal and Lee [48], the respondents were instructed to select one (or more) products from the list to complete the sentences, thus indicating the respective products' utilitarian and symbolic values. Multiple responses were allowed. Results in Table 1 showed that bottled mineral water had the highest utilitarian (133 out of 150 possible mentions) and symbolic (97 out of 150 possible mentions) values, followed by mobile phone (utilitarian: 127; symbolic: 81). Thus, bottled mineral water was selected as the stimulus. Please refer to Appendix A for details of the pretest results.  0.765 (37) I will give priority to Whistler when purchasing mineral water. 0.703 Note: Alpha refers to Cronbach's alpha reliability estimate; SFL refers to standardized factor loading.
Whistler was a brand of bottled mineral water produced in Canada that was not available in China at the time of the study. Using a real brand provided an externally valid context but its unavailability in the Chinese market minimized prior consumer-brand association.

Celebrity Selection
Then, the pretest aimed to select a celebrity that satisfied the characteristics of both good looks and reputation. We began with the top 30 celebrities listed in the 2017 Chinese Celebrity Income Rank [49]. The income ranking allowed us to include celebrities highly involved in endorsement, as endorsement often makes up a significant part of a celebrity's income. All but one (athlete) of the celebrities on the list worked in the entertainment industry. We sent the list to 40 respondents (20 males, 20 females; age 22 to 51) and asked them to select one celebrity among the 30 perceived to be both attractive and with good reputation. The respondents worked independently. The results showed that Liu Tao (LT) had the highest votes (11 votes), followed by Jacky Cheung (9 votes). The other votes were quite diverse. Thus, LT, who had the seventh highest income according to the list, was selected for study 1.
LT's popularity and constant presence on TV advertisements, magazines, and social media [50] make her a good candidate for the research. At the beginning of the data collection, the selected celebrity LT' popular TV series Ode To Joy 2 had just finished broadcasting during prime-time on Zhejiang Satellite TV, and the new series Ode To Joy 3 will begin its production in 2020. LT's image is easily and readily recognizable among consumers.

Empirical Manipulations
The study was carried out online. We chose the platform WJX to recruit respondents and distribute the questionnaires. WJX is the largest non-proprietary online survey platform in China, with more than 2.6 million people on its panel. Its client list includes international corporations such as Walmart, Starbucks, BMW, and McKinsey & Company that have used this platform to conduct surveys. All WJX panel members can access the link containing the questionnaire of this study. To avoid repeats, members with the same IP address or same computer/mobile phone equipment could access the questionnaire once only.
At the beginning of the study, the respondents were informed that their participation in the academic research would be voluntary and anonymous, and that they would be exposed to a photo and some questions (for questionnaire details, please refer to the Appendix B). Then, the respondents were asked to choose randomly a letter, A, B, or C, without knowing that the letters correspond to the three conditions. To ensure a balance among the three conditions, the maximum number of respondents in each condition is set at 200.
In the proper study, each respondent was exposed to one picture only. In condition A, the respondents were exposed to a picture that appeared that LT was endorsing Whistler water in an advertisement. In condition B, the respondents were exposed to a picture that appeared that Whistler water was placed in a TV series with LT as the actress. In condition C, the respondents were exposed to a picture showing LT holding a bottle of Whistler water on the street. This picture had been digitally modified and it looked like that LT was unconsciously captured by passers-by in her daily life. The pictures in the three conditions were presented as either "taken from a forthcoming advertisement for a new bottled mineral water" (ad endorsement condition), "taken from LT's Ode To Joy 3: The TV series to be shown next year" (product placement condition), or "taken by passers-by who saw LT on the street" (real-life condition). The respondents were asked to spend a few minutes observing the pictures.
Afterwards, the respondents performed an unrelated task to clear their short-term memory. Following this task, they were asked to choose from several options the brand they saw in the previous pictures. The options included Whistler and some other brands. They were also asked to identify the context they saw in the previous pictures (i.e., ad, placement, real-life). The respondents (1) who were not regular users of bottled mineral water, (2) who were aware of the Whistler brand, (3) who failed to identify the brand name (i.e., Whistler) correctly, and (4) who could not correctly identify the context they were exposed to were thanked and dismissed.
The remaining respondents continued to the next session. First, before they answered the brand response questions, they were told that "Whistler, originated from a foreign country, is water that is natural and free of pollutants. It also provides people with rich beneficial minerals. Whistler will soon be available in China. We would like to know your perceptions concerning this brand." Second, before they answered questions related to their perceptions of the selected celebrity, the respondents were told that "Liu Tao (LT) is an actress in China and has starred in many films and television shows. Please answer the following questions."

Respondents
A total of 345 respondents, including 149 females and 196 males, passed the four screening criteria/manipulation checks and completed the questionnaire successfully. They were quite evenly distributed across the three conditions (110 in A, 119 in B, 116 in C). Two-thirds of them (n = 230) were young people between 18-30 and three-quarters of them (n = 259) were well educated, with a bachelor's degree. This younger, well-educated respondent pool is consistent with the segment of bottled mineral water users in China that according to ASKCI Consulting (Retrieved from https:// www.askci.com/news/chanye/20190528/1045131146931_2.shtml) consists of upscale consumers looking for healthier consumption choices. We ran a series of chi-square tests and the results showed no significant differences across the three conditions in terms of the respondents' gender (x 2 = 12.66, p > 0.05), age (x 2 = 11.23, p > 0.05), and educational level (x 2 = 6.05, p > 0.05).
The respondents were debriefed at the end of the study. There was a disclaimer in the debriefing reminding the respondents that the study was a piece of academic research. The depiction of LT and Whistler in the study was unrelated to either party. Each respondent received 5 RMB (about USD 80 cents) through the WJX system as incentive for participation.

Measures
The measures were adapted from existing scales. PSI was adapted from Rubin et al. [34]. The factors influencing PSI, namely attractiveness and homophily, were separately adapted from Ohanian [51] and McCroskey, McCroskey, and Richmond [52]. Consumer's brand attitude and brand quality were separately adapted from Torres, Sierra, and Heiser [53] and Grier and Deshpandé [54]. Consumers' SBC and purchase were separately adapted from Escalas [29] and Sirgy [55]. The detailed measurement items and reliability are shown in Table 1.

Mediation Analysis Approach
The mediation analysis was carried out using bootstrapping, a method that has become popular in testing indirect effects, given its high statistical power [56]. Bootstrapping is a non-parametric test based on resampling with replacement that is done many times (e.g., 5000). If the analysis shows that the confidence level does not contain zero (i.e., the indirect effect is different from zero), the researcher can determine that the mediation between independent and dependent variables exists [57]. The current analysis was conducted on SPSS using the PROCESS procedure. We followed the recommendations of Cheung and Lau [58] and set the number of samples for bias correction at 5000 and the confidence interval at 95%.

Consumer Attitudes in Different Endorsement Contexts (H1)
A MANOVA (multivariate analysis of variance) was utilized to test whether there are significant differences in the mean scores for all variables by three contexts of celebrity-brand association. The results in Table 2 revealed significant differences on consumer brand-related responses (self-brand connection F(2, 342) = 10.4, p = 0.000; brand quality F(2, 342) = 8.94, p = 0.000; brand attitude F(2, 342) = 7.60, p = 0.001; purchase intention F(2, 342) = 7.33, p = 0.001) across the three contexts of celebrity-brand association. The Tukey post-hoc comparisons revealed significant differences between the advertisement vs. real-life context and product placement vs. real-life context in terms of all the four brand responses. Consumers who were exposed to the celebrity-brand association in real life context had the highest level of brand-related responses, namely self-brand connection, perceived brand quality, brand attitude, as well as purchase intention. The results supported H1.

Drivers for Consumer-Celebrity PSI (H2)
Linear regression analysis of SPSS is used to test the effects of attractiveness and homophily between on PSI. There is no presence of collinearity (Tolerance = 0.832 > 0.10, VIF = 1.202 < 10) of the two variables of attractiveness and homophily. Accordingly, there are significant effects of both attractiveness (β = 0.361, p < 0.001, R 2 = 0.614) and homophily (β = 0.518, p < 0.001, R 2 = 0.614) on PSI. Thus, H2 is supported. Someone who perceived a celebrity to be more attractive would attribute more perceived PSI with the celebrity. Meanwhile, someone who perceived more homophily with a celebrity would attribute more perceived PSI with the celebrity.

Consumer Perceived Relations with Celebrity in Different Endorsement Contexts (H3)
The results in Table 2 showed significant differences on consumers' perceived celebrity attributes (attractiveness F(2, 342) = 5.01, p = 0.007, consumers' perceived homophily with celebrity F(2, 342) = 7.40, p = 0.001, and para-social interaction F(2, 342) = 10.3, p = 0.000) across the three contexts of celebrity-brand association. The Tukey post-hoc comparisons showed significant differences between the advertisement vs. real-life context and product placement vs. real-life context. Consumers who were exposed to the celebrity-brand association in a real-life context perceived the highest level of celebrity attractiveness, homophily, and para-social interaction with the celebrity, in support of H3.

Mediating Effects of PSI and Its Drivers (H4)
The mediation analyses were performed using SPSS Process Macro Model 4 [57] to identify consumer's PSI as well as its drivers (attractiveness and homophily) in mediating the relationships between celebrity-brand association contexts and brand responses. Three dummies were created to represent the contrasting between contexts. The results are listed in Table 3.
Results show that attractiveness exerts a significant mediating effect in the contrast between advertising and real-life context on all of the brand response measures: Brand attitude (β = 0.05, P M = 0.31, SE = 0.02, 95% CI (0.01, 0.11)), self-brand connection (β= 0.04, P M = 0.19, SE = 0.02, 95% CI (0.01, 0.10)), purchase intention (β = 0.05, P M = 0.27, SE = 0.02, 95% CI (0.01, 0.11)), and perceived brand quality (β = 0.04, P M = 0.21, SE = 0.02, 95% CI (0.01, 0.10)). Attractiveness also exerts a significant mediating effect on the contrast between product placement and real-life context on these variables:  Together, the results revealed significant mediating effect (95% confidence intervals did not include zero) in two contrasts; namely, advertising vs. real-life, and product placement vs. real-life. The remaining context (i.e., advertising vs. product placement) showed no significant mediating effect. Accordingly, the mediating process only worked when comparing real-life context and commercial endorsement contexts. The results provided partial support for H4. Figure 2a-c graphically illustrates the mediation effects identified in Table 3. Together, the results revealed significant mediating effect (95% confidence intervals did not include zero) in two contrasts; namely, advertising vs. real-life, and product placement vs. real-life. The remaining context (i.e., advertising vs. product placement) showed no significant mediating effect. Accordingly, the mediating process only worked when comparing real-life context and commercial endorsement contexts. The results provided partial support for H4. Figure 2a-c graphically illustrates the mediation effects identified in Table 3.  For the mediating effects that are statistically significant, Table 3 also reported the effect size, as indicated by percent mediation (PM) (Preacher and Kelly 2011). As shown, it is widespread (PM = 0.19 to 0.72), with para-social interaction being the strongest mediator, accounting for about half of the total effect and more (PM = 0.46 to 0.72). The effect size for attractiveness (PM = 0.19 to 0.44) and homophily (PM = 0.24 to 0.47) is considerably smaller, accounting for less than half of the effect. These findings provide support for our postulate that para-social interaction is a major explanator for natural endorsement effects. For the mediating effects that are statistically significant, Table 3 also reported the effect size, as indicated by percent mediation (P M ) (Preacher and Kelly 2011). As shown, it is widespread (P M = 0.19 to 0.72), with para-social interaction being the strongest mediator, accounting for about half of the total effect and more (P M = 0.46 to 0.72). The effect size for attractiveness (P M = 0.19 to 0.44) and homophily (P M = 0.24 to 0.47) is considerably smaller, accounting for less than half of the effect. These findings provide support for our postulate that para-social interaction is a major explanator for natural endorsement effects.

Discussion and Conclusions
This study conceptualized and examined empirically the effect of three celebrity-brand association contexts (i.e., advertising, product placement, and real-life) on the consumer's brand responses. Results revealed that, compared to commercial endorsement contexts (i.e., advertising and product placement), a natural endorsement context (i.e., real-life) produced better brand responses, including perceived brand quality, self-brand connection, brand attitude, and purchase intention. Further, the results showed that the effect induced by the various contexts was mediated by the consumer's para-social interaction with the celebrity. Thus, a "real" context that can enhance the consumer's para-social interaction with the celebrity can improve the consumer's brand-related responses.
Consumers in the digital age have diverse platforms to learn about a celebrity beyond the official channels. The information and images they are able to access via unofficial channels, including fans pages and consumer blogs, may alert them of differences between the celebrity's media-constructed image and the "real" person, including his/her brand usage and consumption choices in everyday life. The "real" everyday context that generated superior effect in this study may alert consumers to the "real" person in the celebrity, thus closing the gap between the celebrity and consumer and improving the consumer's brand responses.

Natural Celebrity-Brand Association in Closing Consumer-Brand Relationship
Building on the previous point, where consumers are able to hold stronger relationships with the celebrity in the real-life context indicate that this context could give a sense of reality to consumers. When consumers saw the celebrity using a brand/product without any persuasive purpose, people would feel that the celebrity was down-to-earth. The results correspond to the argument of Biel and Bridgwater [33] and Rubin, Perse, and Powell [34] that the sense of reality shortens the psychological distance between consumers and celebrity, which in turn improves consumer's PSI with celebrity. According to Van Laer et al. [31], if the celebrity meets the expectation of consumer of using a brand/product in real world, consumers' willingness of interacting with the celebrity was bound to be affected.
Celebrity was often reference group for normal people as mentioned by Escalas [29]. Imitating the celebrity and trying to look similar to the celebrity was a common way for consumers to construct and confirm the self through consumption as Litter and Melanthiou [41] and Swann, Stein-Seroussi, and Giesler [40] have proved in their studies. When consumers saw the brand/product they would consume was also accepted and used by the celebrity in real life, consumers might have thought that their taste of the brand/product or even way of life were similar and consistent with the celebrity, which made consumers feel that the celebrity was a like-minded person. According to the arguments of Cohen [37], Gooch [38], and Himelboim, McCreery, and Smith [39], the more the audiences recognize that the media figure is similar to them, the more communication attempts the audiences have with the media figure. Natural celebrity-brand association, which helped created sense of homophily between consumer and celebrity, enhanced consumer-celebrity PSI.
With the feeling of ease and similarity in exposing to the natural celebrity-brand association, consumers are less defensive and resentful of the brand/product that is displayed and indirectly promoted by a celebrity. The decrease in disgust may help build consumer trust on the celebrity's motivation in using a brand/product, wherein the results of the study correspond to the findings of Click et al. [27], Marwick [30], and Silvera and Benedikte [20]. The trust reduces consumer's stranger-ness of the celebrity according to the viewpoint of Yin [44] and reduces uncertainty of the brand according to the argument of Swann, Stein-Seroussi, and Geisler [40], helping form consumer-celebrity PSI and consumer-brand SBC. The inherent logical consistency of the function of reducing uncertainty between natural celebrity-brand association and consumer-celebrity PSI caused the former to have the effect of increasing the closeness of the consumer-celebrity relationship.
In addition, the result that consumers perceived the celebrity more attractive when exposed to the natural celebrity-brand rather than the other two commercial endorsement forms corresponded to the assumption that the change in forms of celebrity-brand association might affect consumer's perceived attractiveness and overall liking of the celebrity, which the result supports the assumptions of Russel and Rasolofoarison [6].

Natural Celebrity-Brand Association in Enhancing Persuasion Effects
Commonly, celebrity endorsements occur in advertisements, with the obvious intent of promoting a product. Whether explicit endorsement such as traditional advertisements in mass media, or implicit endorsement such as product placement in TV shows and entertainment programs, according to Avery and Ferraro [13], Hackley and Tiwsakul [14], these are commercial activities with profit-oriented nature. Audiences are smart enough to distinguish between selling behavior and real-life scenarios, and they know that the products and settings they have seen in advertisements, TV shows, and films have been elaborately set up by marketers and advertisers in order to attract their attention and convince them to spend money.
The result of this study shows that consumers hold better attitudes towards the brand used by celebrity in real-life settings rather than in commercial advertisements contexts. It corresponds to the findings of Seno and Luka [15] and Homer [21] that the normal endorsement activities may be considered merely as money-making behaviors of celebrities and bring unsatisfactory endorsement effect. If consumers are unfamiliar with a celebrity endorser and happen to see a brand endorsed by the celebrity, they will regard the celebrity as a figure living in another world with a large gap in financial and social status between them, according to Bressoud, Lehu, and Russell [22]. The result also indicates that if consumers see a picture showing a celebrity using a brand/product in his/her real-life scene, especially when the picture looks like a snap of the celebrity who is unaware of the camera, consumers' disgust on profit-orientation may not be easily aroused; this result corresponds to the assumptions of Russell and Rasolofoarison [6] that natural celebrity-brand association more easily causes consumers to have a good impression of the brand. As celebrities are usually reference groups for normal people, according to Escalas [29], consumers will increase their trust in brand/product usability if they think their idols indeed use a brand/product in daily life rather than merely applauding for the brand/product in advertisements.
Celebrity-brand associations in real-life context significantly helped strengthen consumer-brand SBC and increased consumers' positive evaluation of the concerned brands, as shown in the data analysis portion of this research. The result supported the finding of Escalas [29], that consumers who are involved in a brand story in the real world have stronger SBC. They also support the assumptions of Jin and Phua [28] and Russell and Rasolofoarison [6] that a natural celebrity-brand association has a more positive effect on consumers' brand-related responses and perceived celebrity credibility than a market-oriented endorsement.

Theoretical Contributions
Whereas the overall positive findings of this study yielded useful information, this study provides only a starting point in advertisers' understating of the effects of natural endorsement. Aside from PSI, attractiveness and homophily, there are other factors effective in traditional advertising but their effectiveness in a natural context remains to be tested. Alternatively, since natural endorsement comprises a new communication platform, there may be new variables effective in this context that have not been studied. Further, the growing popularity of natural endorsement signals changes in an information-rich consumption environment. The consumer's changing views towards celebrity advertising in particular and their skepticism towards paid advertising in general present important research topics. Meanwhile, this study examined the effects of natural endorsement in the context of a particular celebrity and product in China. The authors welcome replication and extension of this study to further specify the effectiveness of natural celebrity endorsement.
In addition, this study emphasizes the consumer-celebrity relation perspective in endorsement advertising research. In most studies in the field of endorsement, celebrity endorsers were regarded as having the lead role in deciding the success of the endorsement, while consumers were seen as merely waiting for the persuasion and suggestion of the celebrity endorser. Although the consumer perspective has received more academic attention in recent years, especially with the boom of social media, which offers greater possibilities of interaction between consumer and celebrity, there remains a lack of research that integrates the consumer-celebrity relation factor into the field of celebrity endorsement. The verified high level of consumer-celebrity PSI in the natural endorsement context in this study provides insights for exploring the factors that may influence the relationship between consumer and celebrity, which emphasizes the consumer-celebrity relation perspective in endorsement field.

Implications for Practice
A major question for advertisers is how to make consumers less likely to develop antipathy towards the obvious commercialism of an endorsement. A common practice is to let the celebrity really use the product before endorsing it. With the increasing numbers of reality shows and other similar entertainment programs on television, audiences were becoming aware of the fact that the use of a product by a celebrity in a seemingly real-life situation in front of the camera was often a deliberate set-up by marketers or advertisers with a commercial purpose. To create a more natural celebrity-brand association and win the favor of consumers, market practitioners have now begun to try a variety of techniques to arrange for celebrities to appear with their brands inadvertently. For example, the "candid" airport photograph was quite a popular way of promoting brands, and it continues to gain consumer trust based on the perceived real-life usage of the brand by a celebrity. The "street snap" was a culture that has been derived from western countries [59], where "street" referred to almost any ordinary public place that normal people are familiar with, and "snap" meant a snapshot that was taken quickly, flexibly, and suddenly. Street snaps could be photographs of both celebrities and normal people, because the character of the street "offers causal encounters, the possibilities of engagement, the adoption or relinquishing of a personality" [60] (p. 12). As a culture that was based on taking and publishing casual photographs, street snaps indicate infatuation of audiences with visual images in the age of the consumption economy. Audience infatuation and desire to imitate celebrities drive marketers to look for opportunities to use street snaps as a means of promoting brands and products. For example, on Taobao (the largest online retail business district in China), there are many goods on sale in the style of products used by celebrities in airport snaps, which help enterprises to profit indirectly from these natural celebrity-brand connections.

Limitations
For the purpose of enhancing internal validity, this study tested the proposed model on a general range of consumers without distinguishing between the different levels of celebrity involvement that existed among segmented consumer groups. For example, some consumers are fanatical fans, and some are mild enthusiasts of the same celebrity. Thus, the level of PSI with the celebrity is only a little higher than the scale midpoint. Consumer segmentation in terms of their levels of involvement with a celebrity should ideally be taken into consideration in any follow-up studies.
In addition, the current study was conducted in the Chinese market, using Chinese consumer responses to a local celebrity, which limits the scope of external validity of this study. Future comparative studies that examine the proposed variable relationships in cross-cultural context are required to deliver new academic and managerial insights for global markets. Also, diverse samples in larger size should be employed in future replicate studies to ensure the reliability and validity of the study. Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflict of interest. Note. % refers to the percentage of respondents who chose the product among the 50 people. Raw No. refers to the real number of respondents who chose the product.

Questionnaire
Please spend a few minutes to complete this questionnaire. There is no right or wrong answer. Your participation is voluntary. The resources generated by this questionnaire will only be used in my research papers. There will be no content that shows your personal information in the dissertation. Everything will be kept confidential. Thank you very much for your participation.
Please choose one option to do the following questions. (Options: 1. A; 2. B; 3. C) Condition A: Ad Commercial Endorsement "As you know, celebrities such as LT often endorse products in advertising. The following photo is from a forthcoming commercial. Please take a while to observe it." (Readers who are interested in the pictures please approach me for them) Note. % refers to the percentage of respondents who chose the product amon No. refers to the real number of respondents who chose the product.

Questionnaire
Please spend a few minutes to complete this questionnaire. There is no Your participation is voluntary. The resources generated by this questionn my research papers. There will be no content that shows your perso dissertation. Everything will be kept confidential. Thank you very much for Please choose one option to do the following questions. (Options: 1.A;

Condition A: Ad Commercial Endorsement
"As you know, celebrities such as LT often endorse products in advertis is from a forthcoming commercial. Please take a while to observe it." (Read the pictures please approach me for them) Condition B: Product Placement "As you know, celebrities such as LT are often seen with product plac following photo is of LT's new TV series 'Ode To Joy 3′ which will begin Picture: Liu Tao's poster holding a bottle of water with Whistler brand logo Condition B: Product Placement "As you know, celebrities such as LT are often seen with product placements in TV series. The following photo is of LT's new TV series 'Ode To Joy 3' which will begin to be produced in 2019. Please take a while to observe it." Sustainability 2020, 12, x FOR PEER REVIEW

Condition C: Real-Life Condition
"As you know, celebrities such LT are often caught on camera during their following photo of LT was taken by passers-by who spotted her on the street. Please observe it." (1) You may have noticed that, there is a bottle of water in the photo of last questio brand name of the bottled water in the photo in the previous question? (Op 2.Volvic; 3. Whistler; 4.Lurisia; 5.Zajecicka) (2) How the bottle of water is promoted in the photo you have observed in prev (Options: 1.photo of a forthcoming commercial ad; 2. shot of forthcoming TV ser LT in real life/others) (3) Have you heard of the Whistler before? (4) Whistler, originated from a foreign country, is natural and non-polluting water, w people with rich beneficial minerals. Whistler will be soon be available in China like to know your options about it. (5) Whistler seems suits me well. (6) Whistler seems like a brand I can see myself drinking. (7) Whistler could be a brand representing me. (8) Whistler seems to convey some of my personal information. (9) Whistler seems reflect the way that I want to present myself to others. Condition C: Real-Life Condition "As you know, celebrities such LT are often caught on camera during their daily life. The following photo of LT was taken by passers-by who spotted her on the street. Please take a while to observe it." Sustainability 2020, 12, x FOR PEER REVIEW

Condition C: Real-Life Condition
"As you know, celebrities such LT are often caught on camera during their daily following photo of LT was taken by passers-by who spotted her on the street. Please take observe it." (1) You may have noticed that, there is a bottle of water in the photo of last question. W brand name of the bottled water in the photo in the previous question? (Options 2.Volvic; 3. Whistler; 4.Lurisia; 5.Zajecicka) (2) How the bottle of water is promoted in the photo you have observed in previous (Options: 1.photo of a forthcoming commercial ad; 2. shot of forthcoming TV series; 3 LT in real life/others) (3) Have you heard of the Whistler before? (4) Whistler, originated from a foreign country, is natural and non-polluting water, which people with rich beneficial minerals. Whistler will be soon be available in China and like to know your options about it. (5) Whistler seems suits me well. (6) Whistler seems like a brand I can see myself drinking. (7) Whistler could be a brand representing me. (8) Whistler seems to convey some of my personal information. (9) Whistler seems reflect the way that I want to present myself to others. (1) You may have noticed that, there is a bottle of water in the photo of last question. What is the brand name of the bottled water in the photo in the previous question? (Options: 1. Sirma; 2. Volvic; 3. Whistler; 4. Lurisia; 5. Zajecicka) (2) How the bottle of water is promoted in the photo you have observed in previous questions (Options: 1. photo of a forthcoming commercial ad; 2. shot of forthcoming TV series; 3. photo of LT in real life/others) (3) Have you heard of the Whistler before? (4) Whistler, originated from a foreign country, is natural and non-polluting water, which provides people with rich beneficial minerals. Whistler will be soon be available in China and we would like to know your options about it. (5) Whistler seems suits me well. (6) Whistler seems like a brand I can see myself drinking. (7) Whistler could be a brand representing me. (8) Whistler seems to convey some of my personal information. (9) Whistler seems reflect the way that I want to present myself to others. (10) I think Whistler could help me become the type of person I want to be. (11) I think Whistler is in high quality. (12) I think Whistler can bring me positive energy.