How Do Virtual Social Media Influencers Affect Millennial and Gen Z Female Consumers’ Purchase Intention? A Comparison of Fast Fashion Versus Luxury Fashion Brands

With the emergence and increasing prevalence of social media, the effectiveness of social media influencers (SMIs) is apparent in today’s fast-paced life. Virtual social media influencers (VSMIs), computer-generated influencers with a real human appearance, have gained popularity. The purpose of this study is to examine important VSMI’s characteristics that affect brand image and trust which lead to purchase intention (fast fashion vs. luxury fashion brand). Social exchange theory and source credibility model were adopted to develop a theoretical framework for the present study. Two Instagram ads were developed that included the post’s image and corresponding caption: VSMI endorsing H&M and Gucci. Each participant was randomly presented with one of the two Instagram ads. A total of 163 participants completed the survey (H&M: 93, Gucci: 73). The majority of the sample were female Caucasian American students (68.7%) between the ages of 18 and 40 years old. The results showed VSMIs’ characteristics significantly affect brand image and trust leading young adult consumers’ purchase intentions both fast fashion and luxury fashion brands. For both brands, authenticity significantly influenced brand image and trust and homophily influenced brand trust. For the fast fashion brand, attractiveness influenced brand image and trust, whereas wishful identification influenced brand image only for the luxury fashion brand. Both brand image and trust influenced purchase intention. Findings suggested that both fast fashion and luxury fashion brands would benefit from the VSMIs that present genuine messages Instagram followers resonate with. Fast fashion brands should focus on delivering the message in an appealing manner, while luxury fashion brands collaborating with VSMIs should focus on creating an aspirational message that followers can relate with. By strategically developing social media content according to the type of fashion brand, VSMIs can positively influence brand image and trust motivating followers’ intention to purchase fashion brands.

While partnering with a human social media influencer (HSMI) can be beneficial to create consistency with a brand's message, (Kadekova & Holiencinova, 2018), it can often be difficult for brands to find the appropriate influencer that shares their identity and image (Santora, 2021). Brands face another problem when SMIs connected to their campaign is involved in a scandal. Since SMIs are human, they are susceptible to indiscretion (Adut, 2008) leading to inconsistency in their online image (Khamis et al., 2017). The brand partnering with the influencer becomes vulnerable to criticism (Kintu & Ben-Slimane, 2020). With this in mind, VSMIs are beneficial by the fact that they are not human in nature and can post content that is more thoughtfully curated to share an intended message (Baklanov, 2020). Although VSMIs have gained popularity, VSMIs are unknown and little empirical studies have investigated the impact of VSMIs on brand responses (Tayenaka, 2020). The intersection between influencer marketing and virtual reality could be very influential given the growing investments in each sector. While most empirical research has focused study on HSMIs, scant research has examined similar areas relating to VSMIs even despite the increasing deployment.
The current study develops a conceptual framework by applying the social exchange theory (Homans, 1974) and the source credibility model (Hovland et al., 1953). The purpose of this study is to examine whether the influencer's credibility (i.e., expertise, authenticity) and/or attractiveness (i.e., attractiveness, wishful identification, and homophily) can lead to positive brand image and brand trust. Specifically, this study focuses on comparing the effect of the influencer's credibility and attractiveness on brand image and trust of VSMIs on Instagram for fast fashion and luxury fashion brands. Finally, this study investigates the effect of brand image and trust on purchase intention.
The findings of this study will advance the theoretical understanding regarding the impactful characteristics of VSMIs on Instagram. This research will be the first attempt to empirically investigates social media marketing strategies according to brand type using VSMIs.
The results will provide online fashion brand managers with practical implications for how to curate social media content by partnering with virtual influencers on Instagram.

DEFINITION OF TERMS
Social media influencer -Self-generated, regular people who post content in specific areas such as beauty, fashion, food, and travel on social media .
Virtual social media influencer -Computer-generated influencers with a real human appearance and curated fictional narrative (Arsenyan & Mirowska, 2021;Robinson, 2020) Source credibility -The judgements made about another person regarding the believability of the communicator (O' Keefe, 1990, p. 130-131;Wilson & Sherrell, 1993) Expertise -An individual's perception about another as a source of valid assertions (Hovland et al., 1953) Authenticity -The intrinsic motivation from an individual's true self (Jun & Yi, 2020) Attractiveness -Human preferences about the physical appearance of other people (Ellis et al.,

2019)
Wishful identification -The aspiration to be like or act like another person (Hoffner & Buchanan, 2005) Homophily -The perceived similarity to another person (Simons et al., 1970) Brand image -A set of cognitive, sensory, and/or emotional associations that consumers interpret about a brand, product, or service (Cho et al., 2015) Brand trust -A consumers' willingness to rely on the brand's ability to do or perform as promised (Chaudhuri & Holbrook, 2001;Moorman et al., 1992;Morgan & Hunt, 1994) Purchase intention -A deliberate plan to purchase a brand (Spears & Singh, 2004)

CHAPTER 2. LITERATURE REVIEW
The present chapter provides a review of literature that explains the theoretical framework and primary constructs examined in the study. This chapter begins by explaining the social exchange theory and source credibility model which were adopted to develop a conceptual framework. The second section discusses hypothesized relationships among the variables: source credibility, source attractiveness, brand image, brand trust, and purchase intention.

Social Exchange Theory
The social exchange theory explains the behavioral psychology of interaction (Homans, 1974). According to Homans (1974), all interpersonal social behavior can be viewed as an exchange of activity, either tangible (e.g., money) or intangible (e.g., social services and relationships) benefits which may result in a consequence such as trust and commitment (Chia et al., 2021;Cook & Rice, 2006). The consequence from the interaction between individuals is either costly or rewarding (Thibaut & Kelley, 1959). Every interaction or exchange involves an investment of resources (i.e. time or effort) (Chia et al., 2021;Homans, 1974). People seek to earn a reward in response from their investment and are incentivized to return to rewarding situations (Homans, 1974;Salam et al., 1998;Shiau & Luo, 2012). With an impersonal relationship, the person engaging in an interaction with the other is aware that there are many other people they could seek out to achieve the same reward. With a personal relationship, the person engaging in the interaction knows they have limited options in who they could seek to receive the same reward. In the context of social media, users interact frequently expecting to obtain potential benefits through the exchange (Ferm & Thaichon, 2021;Shiau & Luo, 2012) and develop rewarding social relationships (Jaing et al., 2021;Liu et al., 2016). Social media influencers share content with their followers. Followers return appreciation by engaging with their content. Sharing, liking, commenting, or subscribing to the influencers content offers a reward (Kim & Kim, 2021;O'Donell, 2018).

Source Credibility Model
Credibility is a complex and multi-dimensional concept (Qureshi et al., 2021) that can be defined as "reliability, accuracy, fairness, and objectivity, as well as various combinations of these concepts" (Hilligoss & Rieh, 2008, p. 1468. Source credibility is defined as the "judgements made by a perceiver concerning the believability of a communicator" (O' Keefe, 1990, p. 130-131;Wilson & Sherrell, 1993). Since the effectiveness of communication is often dependent upon who delivers the message (Hovland, Janis, & Kelley, 1953), the source credibility model examines the attitude of the audience towards the speaker (Hovland & Weiss, 1951). The audience's perception of a speaker's message is dependent on the established credibility of the speaker (Umeogu, 2012). Many studies have found if the source is presumed to be credible, the message presented by the source is also believed to be credible (Fragale & Heath, 2004;Hovland & Weiss, 1951;Qureshi et al., 2021;Rieh & Danielson, 2007;Riquelme & Gonzalez-Cantergiani, 2016). For example, credibility is essential to the success of the celebrity endorsement campaign Silvera & Austad, 2004). The speaker has the capability of creating, managing, and cultivating their credibility (Corman et al., 2006;Umeogu, 2012).

Source Credibility, Brand Image, and Brand Trust
The majority of researchers define source credibility in terms of two primary dimensions, expertise and authenticity (Hovland et al., 1953;Kim & Kim, 2021;Metzger et al., 2003;Qureshi et al., 2021). Expertise is "the extent to which a communicator is perceived to be a source of valid assertions" (Hovland et al., 1953, p. 21) and refers to the "authoritativeness, competence, expertness, qualification of the speaker" (Ohanian, 1990, p. 42). The speaker displays expertise through the quality and quantity of information shared, their degree of ability, education, and professional achievement as well as the validity of their judgements (Giffin, 1967;Kim & Kim, 2021).
Brand image is defined as the set of associations that consumers interpret about a brand, product, and service cognitively, sensory, and emotionally (Cho et al., 2015). This meaning and overall impressions of the brand is created through product attributes, use of the product, and meanings that consumers associate with the brand (Cho et al., 2015;Keller, 2008). Brand image can be transferred from the influencer to the endorsed brand. For example, the transfer of brand image from a celebrity athlete to the endorsed brand was more effective when similar associations between the brand and athlete were found in the context of the advertisement (Arai et al., 2013;Rai et al., 2021).
Brand trust can be defined as the consumers' willingness to rely on the brand's ability to do or perform as promised (Chaudhuri & Holbrook, 2001;Moorman et al., 1992;Morgan & Hunt, 1994). Trust gives customers confidence in a brand's product by reducing uncertainty (Chen & Cheng, 2019;Jun & Yi, 2021;Khan & Zaman, 2021). Source credibility has been found to have a direct effect on persuasion and behavioral response (Pornpitakpan, 2004).
Several studies have shown the impact of source credibility on consumers and its persuasiveness.
For example, an influencer's expertise significantly affects Instagram followers' trust in their product review (Djafarova & Rushworth, 2017;. Thus, the following hypothesis is formulated: H1. The expertise of a VSMI will positively influence (a) brand image and (b) brand trust.
Brand authenticity and brand image are related to each other; however, they are distinct constructs (Bruhn et al., 2012). Research has found authenticity in a brand positively impacts the overall brand image (Yildiz & Ulker-Demirel, 2017). Since brands are identified by their name, symbols, and distinct features, SMIs can serve as human brands since they possess the same characteristics (Ki et al., 2020;Moulard et al., 2015;Thomson, 2006). Human brands are well known people with unique brandable features (Ki et al., 2020). Positive thoughts and feelings towards human brands (or HSMIs) can be transferred to the brand endorsed by the influencer (Ki et al., 2020;Thomson, 2006). When consumers perceive influencers are authentic and they create the content without any compensations, consumers are confident with the content created by the influencers.
Authenticity allows consumers to ensure quality of products and services (Rodrigues et al., 2021). Likewise, if consumers believe the influencer created the content without external compensations, they trust the influencer. Researchers empirically supported that the influencer's authenticity takes an important role in building brand trust (Eggers et al., 2014;Hernandez-Fernandez & Lewis, 2019;Jun & Yi, 2020;Kim & Kim, 2021). Accordingly, the influencer's authenticity will be likely to affect brand image and brand trust. These findings lead to the following hypothesis: H2. The authenticity of a VSMI will positively influence (a) brand image and (b) brand trust.

Source Attractiveness, Brand Image, and Brand Trust
The source attractiveness model suggests that familiarity, similarity, likability, and attractiveness are important factors of source characteristics (McGuire, 1989). Attractive individuals are popular and likely to be perceived as interesting, sociable, strong, and responsive (Dion et al., 1972). The importance of attractiveness has also been emphasized in the context of social media. The positive attitude toward an attractive influencer results in favorable evaluation of the message conveyed by the influencer. For example,  suggested that attractiveness of Instagram influencers enhance a luxury fashion brand's credibility. Chu and Kamal (2008) found that blog readers' product evaluations are influenced by bloggers' attractiveness. Rashinda & Weerasiri (2016) showed that the attractiveness of a celebrity endorser positively influences brand image. Similarly, Till and Busler (2000) found that attractive celebrities positively impact brand satisfaction and brand attitude. Thus, the following hypothesis is proposed: H3. The attractiveness of a VSMI will positively influence (a) brand image and (b) brand trust.

Wishful Identification, Homophily, Brand Image, and Brand Trust
Identification stems from perceived similarity, and wishful identification is the aspiration to like or act like the other person (Hoffner & Buchanan, 2005). Homophily refers to the communicators' similarity perceived by the receiver. As the perceived similarity tends to reduce the receiver's uncertainty (Simons et al., 1970), the receiver interprets that the communicators' beliefs, interests, and attitude are similar to those of the receiver (McGuire, 1985). The theory of homophily (Lazarsfeld & Merton, 1954) suggests that individuals like to choose a homophilous source because they feel comfortable. Normally, homophilous sources are more persuasive than the expertise of the sources (Pornpitakpan, 2004). A recent study found that influencers' homophily is important in building trust among followers (Kim & Kim, 2021). Hence, these findings lead to the following two hypotheses: H4. The wishful identification of a VSMI will positively influence (a) brand image and (b) brand trust.
H5. The homophily of a VSMI will positively influence (a) brand image and (b) brand trust.

Brand Image, Brand Trust, and Purchase Intention
Brand image taps into consumers' perception of a brand that is associated with specific attributes of the brand (Cretu & Brodie, 2007;Keller, 1993;Padgett & Allen, 1997). Brand trust refers to consumers' belief that the brand is consistent, honest, and responsible (Chaudhuri & Holbrook, 2001;Doney & Cannon, 1997). Purchase intention is defined as "an individual's conscious plan to make an effort to purchase a brand" (Spears & Singh, 2004, p.56). Purchase intention is used to measure consumer's actions and is typically a strong indicator of actual purchase (Abdullah et al., 2020;Lee et al., 2015). Purchase intention helps increase sales and maximize profits for companies (Hosein, 2012). Previous research showed that brand image and trust influence purchase intention (Takaya, 2017;Then & Johan, 2021). Godey et al. (2016) found that social media marketing efforts can create a positive connection to brand image and enhance consumer's purchase intention. Thus, following hypotheses are proposed: H6. Brand image will positively influence purchase intention.

H7.
Brand trust will positively influence purchase intention.

CHAPTER 3. METHOD
Chapter 3 explains the procedures and methods used for data collection and data analysis.
An experimental design was used to examine relationships between the variables: expertise and brand image (H1a), expertise and brand trust (H1b), authenticity and brand image (H2a), authenticity and brand trust (H2b), attractiveness and brand image (H3a), attractiveness and brand trust (H3b), wishful identification and brand image (H4a), wishful identification and brand trust (H4b), homophily and brand image (H5a), homophily and brand trust (H5b), brand image and purchase intention (H6), and brand trust and purchase intention (H7). The following section discusses sampling, stimuli development, survey instruments, data collection procedure, and data analyses.

Study Design & Manipulation
This study employed an online experimental between-subjects. The type of fashion brand shown in the VSMI's Instagram post was manipulated to show either a fast fashion or luxury fashion brand. Fast fashion brands such as H&M and Zara focus on emerging trends to provide mainstream consumers with the latest fashion products at an affordable price (Choi et al., 2010).
On the other hand, luxury fashion brands such as Chanel and Gucci focus on superior quality, a heritage of craftsmanship, premium price, unique design, and global reputation (Nueno & Quelch, 1998). Instagram posts illustrating the two requirements were presented to the survey participants: a) Virtual social media influencer endorsing a fast fashion brand b) Virtual social media influencer endorsing a luxury fashion brand

Stimulus Materials
The participants were asked to view a social media advertisement developed from VSMI's Instagram profiles. The influencers selected were relatively obscure and not typically internationally recognizable. Each participant was shown one of the two Instagram posts as outlined in Appendix A. including the post's image and corresponding caption. The VSMI was portrayed endorsing either a fast fashion brand or luxury fashion brand. Participants were given time to freely view the post just as they would view a post from an influencer they typically follow. The text and image were manipulated to adjust for the experimental factors. While the images were obtained from real VSMI's profiles to ensure quality and credibility, the influencers were given fictious names to omit their identity.

Pre-Test
A pretest (n = 32) was conducted to examine if the fashion brands and VSMIs served as a good fit to the intended category. Participants were presented with the names for each fashion brand and asked to indicate how they perceived the brand type. Then, participants were randomly assigned to one of the two stimulus images and ask to report whether the influencer appeared to be virtually created. All scale items were measured using a 5-point Likert type scale ranging from strongly disagree (1) to strongly agree (5). The pretest was conducted from a convenience sample of students at least 18 years of age attending a major Mid-Southern university.

Sample and Data Collection
The online survey was administered using Qualtrics software to collect data. After obtaining approval for the use of human subjects by the Institutional Review Board (IRB), nonstudents and college students were recruited to improve external validity. A randomly selected sample of survey participants was recruited from Amazon Mechanical Turk, a professional marketing research firm and student sample was recruited from a Mid-Southern university. This study targeted consumers who were familiar with social media influencers on Instagram. The survey participants were females over 18 years old living in the U.S. and following at least one fashion social media influencer on Instagram. Participants were informed the aim of the study was to investigate individuals' opinions about the social media marketing efforts of fashion brands in general.
In the first section, the definition of a SMI was provided to help the participants understand the context of the study. Then, screening questions were asked to verify participants met the specified requirements and ensure survey participants were currently following at least one social media influencer on Instagram. Participants were asked about their experience with SMIs by indicating the username of a fashion SMI they follow. Participants were also asked if they had purchased a fashion product from a SMI in the past, a fashion brand promoted by the opinions about an influencer's traits in terms of expertise, authenticity, attractiveness, wishful identification, and homophily, and their thoughts about the fashion brand endorsed by the influencer. In the third section, questions were included to ask demographic information (i.e., age, gender, ethnicity, educational attainment, and annual household income).

Instruments
A self-administered survey was distributed online to test the proposed model (see Appendix D). Eight constructs were measured in this study: expertise, authenticity, attractiveness, wishful identification, homophily, brand image, brand trust, and purchase intention. Reliable and valid scale items were adapted from existing literature to ensure content validity. Scale items were modified to relate to the topic of VSMIs' traits influencing the purchase of fashion products. All scale items except demographic information were measured using a 5-point Likert type scale ranging from strongly disagree (1) to strongly agree (5).
SMIs' expertise was measured with six scale items adopted from three studies (Ohanian, 1990;Peetz, 2012;Wiedmann & Mettenheim, 2021). Authenticity was measured with five scale items adopted from Kim and Kim (2021) and Ohanian (1990). Attractiveness was measured with five scale items adopted from Peetz (2012) and Wiedmann and Mettenheim (2021). Wishful identification and homophily were measured with four scale items for each construct adopted from Hoffner and Buchanan (2005)     • If I were to buy a fashion product, I would consider buying it from the brand endorsed by the influencer. • The likelihood of my fashion product purchase from the brand endorsed by the influencer is high. • My willingness to buy a fashion product from the brand endorsed by the influencer is high. • The probability that I would consider buying a fashion product from the brand endorsed by the influencer is high.

Data Analyses
Data collected from the online survey were analyzed using Statistical Package for the Social Sciences (SPSS) 27.0. First, descriptive statistics was performed to report means, standard deviations, and frequencies of item scores. Second, T-tests were conducted to assess the effectiveness of the manipulation of the stimuli. Third, exploratory factor analysis (EFA) was performed to evaluate the dimensionality of each variable (i.e., expertise, authenticity, attractiveness, wishful identification, homophily, brand image, brand trust, purchase intention).
Third, a Cronbach's alpha coefficient for each of the eight measures was calculated to assess reliability and test discriminant validity of the constructs. Fourth, Pearson's correlation coefficient, means, standard deviations, and average variances were calculated to investigate the relationship between the variables proposed in the study. Lastly, regression analysis was employed to test the seven hypotheses proposed in the conceptual model (see Figure 2.1.).

CHAPTER 4. RESULTS
Chapter 4 presents the samples description and results from exploratory factor analysis, reliability, and regression analysis. This chapter discusses the correlations between the variables and results of regression analysis.

Sample Characteristics
The sample is comprised of females living in the U.S. and over 18 years old who follow

SMI and shopping experiences
Respondents were required to follow at least one fashion social media influencer in order to participate. Of the 166 participants, 59.04% indicated they had purchased a fashion product promoted by a social media influencer they were currently following. When shopping for fashion products, 55.4% of participants indicated online stores are their preferred transactional channel followed by brick-and-mortar stores (12.0%) and department stores (10.2%). The complete description of experiences is summarized in Table 4.2.

Manipulation checks
Of the participants, 87.5% of respondent agreed (4 or higher on the scale) H&M was a fast fashion brands, and 96.9% agreed Gucci was a luxury fashion brand. Additionally, the data confirmed the VSMI appeared to be virtually created for the fast fashion stimuli (M = 4.63, SD = .60) and luxury fashion stimuli (M = 4.56, SD = .61).

EFA and Reliability of the Model Constructs
Exploratory factor analysis (EFA) using principal axis factoring was conducted with varimax rotation using SPSS version 27.0 to determine the multi-item measurement scale's underlying dimension. Scale items measuring expertise, authenticity, attractiveness, wishful identification, homophily, brand image, brand trust, and purchase intention were adapted from previous studies and formatted to the context of social media influencers. Following the Kaiser criterion, factors with an eigenvalue greater than one were retained (Huck, 2012). Items with a factor loading greater than or equal to .50 (Nunnally & Bernstein, 1994) on one factor and factor loadings of less than or equal to .30 on the other factor were retained on one factor. Findings show that each variable had a single factor dimension with high factor loadings (.60-.94). The internal consistency of each item was measured with Cronbach's alpha value and composite reliability (CR) greater than .70 (Cronbach, 1951). Each construct demonstrated satisfactory internal consistency with a Cronbach's alpha value of =.81 to .94. The EFA and reliability test results are summarized in the following section and displayed in Table 4.4.

Expertise
All six expertise items were retained based on factor loadings. The items captured VSMI's perceived expertise. The six-item expertise factor had an eigenvalue of 3.83 and explained 63.81 % of the variance for the items. The Cronbach's alpha for the six items was .89.
The composite reliability for the six items was .91.

Authenticity
All five authenticity items were retained based on factor loadings. The items captured VSMI's perceived authenticity. The five-item authenticity factor had an eigenvalue of 3.84 and explained 76.70 % of the variance for the items. The Cronbach's alpha for the five items was .92. The composite reliability for the five items was .94.

Attractiveness
All five attractiveness items were retained based on factor loadings. The items captured VSMI's perceived attractiveness. The five-item attractiveness factor had an eigenvalue of 3.31 and explained 66.11 % of the variance for the items. The Cronbach's alpha for the five items was .86. The composite reliability for the five items was .91.

Wishful identification
All four wishful identification items were retained based on factor loadings. The items captured the participants wishful identification to the VSMI. The four-item wishful identification factor had an eigenvalue of 3.20 and explained 80.09 % of the variance for the items. The Cronbach's alpha for the four items was .92. The composite reliability for the four items was .94.

Homophily
All four homophily items were retained based on factor loadings. The items captured participants homophily to VSMI's. The four-item homophily factor had an eigenvalue of 3.36 and explained 84.04 % of the variance for the items. The Cronbach's alpha for the four items was .94. The composite reliability for the four items was .96.

Brand image
Both brand image items were retained based on factor loadings. The items captured participants perceived image of the brand. The two-item brand image factor had an eigenvalue of 1.68 and explained 84.06 % of the variance for the items. The Cronbach's alpha for the two items was .81. The composite reliability for the two items was .92.

Brand trust
All three brand trust items were retained based on factor loadings. The items captured participants perceived trust of the brand. The three-item expertise factor had an eigenvalue of 2.36 and explained 78.59 % of the variance for the items. The Cronbach's alpha for the three items was .86. The composite reliability for the three items was .91.

Purchase intention
All four purchase intention items were retained based on factor loadings. The items captured participants purchase intention. The four-item purchase intention factor had an eigenvalue of 3.18 and explained 79.54 % of the variance for the items. The Cronbach's alpha for the four items was .91. The composite reliability for the four items was .94. The influencer has a good understanding of the product and brand. .74 The influencer is an expert.
.82 The influencer is knowledgeable.
.87 The influencer is qualified to offer.
.78 The influencer is skilled.
.80 The influencer has experience with the product and brand.
.85 The influencer is honest.
.86 The influencer is reliable.
.89 The influencer is sincere.
.87 The influencer is trustworthy.
.84 The influencer is charismatic.
.60 The influencer is good-looking.
.90 The physical makeup of the influencer is admirable.
.81 The influencer is beautiful.
. If I were to buy a fashion product, I would consider buying it from the brand endorsed by the influencer. .86 The likelihood of my fashion product purchase from the brand endorsed by the influencer is high. .91 My willingness to buy a fashion product from the brand endorsed by the influencer is high. .91 The probability that I would consider buying a fashion product from the brand endorsed by the influencer is high. .88 Percentage of variance explained = 79.5; Cronbach's α = .91; Eigenvalue = .91; Composite reliability = .94

Correlations between the Variables and Discriminant Validity
Pearson's correlation coefficient was conducted to assess discriminant validity among variables (Kline, 1998 (Kline, 1998), confirming the discriminant validity of constructs.
Attractiveness did not significantly influence brand image or brand trust, rejecting H3a and H3b.

CHAPTER 5. DISCUSSION AND CONCLUSIONS
This chapter summarizes the research results and discusses the findings presented in Chapter 4. The conclusions along with theoretical and practical implications are presented in this chapter.

Research Summary
The virtual environment is growing especially across social media platforms. Increasing investments are being made in virtual reality and social media platforms making VSMIs an intersection of interest for fashion brands. Despite the growing investment, few empirical studies have examined VSMIs effectiveness and potential benefits. As a result, the purpose of this study was to examine how VSMI's credibility and attractiveness affect brand image and brand trust, therefore influencing customer's purchase intention. The effects were compared between fast fashion and luxury fashion brands from the perceptions of millennial and Gen Z female consumers' who historically followed fashion social media influencers. Theoretically, this study advances literature related to VSMIs characteristics generating purchase intention and creates a framework for future research in this area. The findings can provide fashion brand social media mangers with practical implications for how to curate social media content by partnering with VSMIs on Instagram.
Participants were recruited from a Mid-Southern university in the U.S. and from Amazon Mechanical Turk to complete an online survey. Reliable and valid scale items were adapted from existing literature to measure the seven variables. A total of 166 responses were used for data analysis. The majority of the sample was female Caucasian students (68.7%) between 18 and 24 years old (57.5%) Shopping online was the preferred shopping channel (55.42%) and most participants had purchased fashion products purchased by a SMI (59.04%) Descriptive statistics were conducted to examine a mean and standard deviation of the of the eight variables (expertise, authenticity, attractiveness, wishful identification, homophily, brand image, brand trust, and purchase intention). Manipulations checks were performed to validate the stimuli. EFA with varimax rotation and reliability tests were assessed to ensure internal consistency and a single factor of each measure. Pearson's correlation coefficients were examined to assess discriminant validity among the eight variables. Simple linear regression analysis was conducted to test the eight hypotheses. The results statistically supported four of the twelve hypotheses for luxury fashion brands and six of the twelve hypotheses for fast fashion brands. While previous studies focused on source attributes for human SMIs, this study examines similar qualities related to VSMIs. The results showed VSMIs can effectively generate customer's purchase intention. VSMIs can provide fashion brands with an innovative and costeffective methods to engage with their social media audience.

Theoretical Implications
This study contributes to social media marketing literature by applying the social exchange theory and source credibility model to VSMIs. As a result, this study provides beneficial insights to the strategic usage of VSMIs. Through the social exchange theory which views all interaction as a rewarding or costly exchange (Homans, 1974), this study brings a new perspective to its application in social media marketing's emerging virtual environment. This study investigated the role of VSMI's credibility and attractiveness in forming brand image and brand trust, then subsequently examined the effect on millennial and Gen Z customer's purchase intention.
This study provides empirical evidence to advance the source credibility (Hovland et al., 1953;O'Keefe, 1990) and source attractiveness models (McGuire, 1989). This study found the effectiveness of the VSMI's credibility and attractiveness varied across fast fashion and luxury fashion brands. While this is the first known attempt to assess VSMIs according to brand type, other studies have applied source credibility and source attractiveness models for human SMIs (Schouten et al., 2020. For example, Kim & Kim (2021) found authenticity and homophily influenced brand trust and Wiedmann & VonMettenheim (2021) found expertise did not significantly influence brand trust for luxury brands which aligned with the findings of this study. Additionally, luxury fashion brands from this study supported Kim & Kim's (2021) finding that attractiveness did not influence brand trust. Fast fashion brands, however, did find

Practical Implications
The findings from this study provides practical implications for fashion brand social media managers when collaborating with VSMIs. The results showed VSMIs' characteristics significantly affected brand image and trust while confirming influence of young adult consumers' purchase intention for both fast fashion and luxury fashion brands. VSMIs can serve as an effective alternative to human SMIs offering an interesting new perspective (Arsenyan, 2021) and reducing the potential risk from human indiscretion (Robinson, 2020).
Since the significant VSMIs' characteristics differed according to the brand type, it would be beneficial for social media brand managers collaborating with VSMIs to focus on appealing to specific aspects according to brand type. For example, luxury fashion brands should focus on creating an aspirational message with VSMIs given wishful identification was found to be an influential factor creating brand trust. Fast fashion brands should seek to deliver a visually appealing message since attractiveness was found to influence brand image and brand trust. Both fast fashion and luxury fashion brands, would benefit from displaying a genuine message that followers relate with since authenticity influenced brand image and brand trust and homophily influenced brand trust. Authenticity was a concern for VSMIs (Kadekova, 2018), however, this study showed it had a significant and influential effect. In conclusion, by strategically developing social media content according to the type of fashion brand, VSMIs can positively influence brand image and trust, therefore, motivating followers' intention to purchase fashion brands.

CHAPTER 6. LIMITATIONS AND FUTURE RESEARCH
This study was conducted in the U.S.; therefore, the cultural environment might have This research showed that VSMI's influential characteristics differed according to the type of brand (luxury and fast fashion). Future research could improve validity by applying this framework to other fast fashion and luxury fashion brands. Additionally, a comparison approach could also be assumed to explore the similarities and difference between human and virtual influencers.

Appendix D: Online Survey Questionnaire
Part I. Please think of all your experiences and opinions about following social media influencers for a few seconds before beginning the questionnaire.
Social media influencers are self-generated, regular people who post content in specific areas such as beauty, fashion, food, and travel on social media  Scenario: You are scrolling through Instagram browsing posts about fashion products such as clothes, shoes, and handbags. You find the below post of a social media influencer wearing fashion products you are interested in buying.
(Participants were randomly assigned to browse one of the two images) The following statements describe your opinions about the fashion influencer in the Instagram post you viewed. Please select the answer that best matches your agreement with each statement.
Strongly Disagree (1) Disagree (2) Neither Agree not Disagree (3) Agree (4) Strongly Agree (5) The influencer has a good understanding of the product and brand. Please select the answer that best describes your thoughts and opinions about the fashion brand promoted by the influencer in the Instagram post you viewed for each question.

I feel the fashion brand endorsed by the influencer is:
Strongly Disagree (1) Disagree (2) Neither Agree not Disagree (3) Agree ( Please select the answer that best describes your thoughts and opinions about the fashion brand promoted by the influencer in the Instagram post you viewed for each question.