How Generation X and Millennials Perceive Influencers’ Recommendations: Perceived Trustworthiness, Product Involvement, and Perceived Risk

Abstract

Previous literature has found underlying differences in purchasing behaviors, consumption habits, and Internet and social media usage between Generation X and Millennials. The activities and how users engage with consumer advice made by popular social media personalities can differ according to their age. Recent studies have shown that trust in the message transmitted by influencers is a critical factor in explaining the impact of consumer recommendations on their followers. However, so far there is little evidence of the possible variation according to the generational cohort to which they belong. This paper attempts to fill this gap by reviewing theoretical contributions on the relationships between perceived trustworthiness, perceived risk, product involvement, and purchase intention. Next, we proposed an exploratory model that analyzes the differences through partial least squares structural equation modeling (PLS-SEM) with multigroup analysis. The resulting hypotheses were tested on a sample of 116 Millennial and 135 Generation X influencer followers. The results confirmed moderating effects of the generational cohort on message credibility and purchase intention, as well as on Millennials’ risk perception. Additionally, social norm and gender were analyzed, and heterogeneity was found according to the level of social norm of the followers.

1. Introduction

Currently, the Internet and social media are part of the daily lives of nearly two-thirds of the world’s population: “As of April 2022, there were five billion Internet users worldwide, accounting for 63% of the world’s population. Of this total, 4.65 billion, or more than 93%, were social network users” [1]. Facebook, YouTube, Twitter, Instagram, Twitch, and TikTok offer an extremely flexible and almost unlimited digital environment in which people communicate and share their interests, lives, and passions [2,3]. The exponential increase in the number of users of social media has changed the current marketing trends [4] and has led to the emergence of the so-called influencer marketing [5], a strategic alternative to communicate and interact differently with potential customers using the popularity of these characters to promote products [5]. Influencers are external and independent agents that generate interactions with a growing social mass of followers on different platforms, posting videos, photos, tweets, comments, and multimedia content [6]. People who follow the activity of influencers frequently observe their publications, identify with them, interact, and consider them opinion leaders in specific fields, such as fashion [4]. This closeness increases trust and engagement with the content they generate [7].

The crisis resulting from COVID-19 has meant that the retail sector has had to rapidly implement digital technologies aimed at finding new solutions to deliver and promote products [8]. An increasing number of retailers with online stores are turning to introduce messages sponsored by influencers or streamers [9]. This type of media offers potential benefits but presents significant challenges, for example, choosing the right person to promote a product, or determining the group of potential buyers that will be reached by the sponsored message [10]. Current developments in influencer marketing have demonstrated the effectiveness of their endorsements [5,11], especially in certain age ranges [12,13]. Previous findings have shown high acceptance rates among youth and adolescents [14,15]. Dabija et al. [16] point out the importance of knowledge and understanding of the values, needs, demands, expectations, and preferences of different generations for marketers. In this regard, it appears that members of Generation X (born roughly between 1961 and 1981), do not fully trust online environments, watch product reviews more intensely, and are more reticent. In contrast, their successors, the Millennials (born approximately from 1980 onwards) are digital natives, are very comfortable with new technologies, and are always aware of the messages conveyed on social media [16,17,18]. In this context, the way in which different generations make their purchasing decisions based on the recommendations made by influencers gains relevance. Multiple factors that contribute to explaining the effectiveness of influencer marketing have been previously explored, the most prominent being credibility, personal attractiveness, or parasocial interaction [5,10]. However, very little is known about the reception of their recommendations according to the population group to which their followers belong. In addition, it seems likely that intergenerational cultural differences may play a role in the perception of risks associated with product recommendations [19], as well as different levels of involvement with certain product categories, e.g., green, tech, or fashion products [20,21,22].

Although influencer marketing is a growing field of research, new empirical contributions on the determinants of its effectiveness are needed [10]. In a context in which more and more companies are using influencer marketing to promote their products on social media, how followers perceive the advertising message conveyed, as well as the possible risks associated with it, and the degree of involvement of the audience with the sponsored product is a challenge for marketing specialists [23,24]. Previous research on influencer marketing has pointed out how the relationships between the influencer and the message or the influencer and the audience [5,7,10] have usually been considered, overlooking those linked to the identification of the consumer with the sponsored product. In this sense, Belanche, Casaló, Flavián, and Ibáñez-Sánchez [23] pointed out how consumer-product adjustments represent an underlying mechanism that activates the attitudes and behavioral intentions of the audience towards influencer endorsement. Following this reasoning, introducing the concept of product involvement [25], unexplored in this field, would be particularly valuable, since it can contribute to explaining how the audience evaluates commercial messages and considers the derived risks depending on the importance they attach to the endorsed product category (e.g., fashion, eco-friendly products, etc.) [25]. At the same time, it could facilitate the introduction of possible moderating effects depending on the particular characteristics of the generational cohort to which the message recipients belong (e.g., Generation X, Millennials), as well as their preferences [26,27].

The work presented here bridges the previously exposed gap and aims to explore the links between perceived trustworthiness of the message transmitted, product involvement, and perceived risk on the purchase intention of influencers’ followers in two different population groups (Generation X and Millennials). From this point, two questions guide the research: what are the inter-relationships between the above variables and purchase intention derived from sponsorships, and is the generational cohort positioned as a significant moderating variable? To address both questions, we propose a new exploratory theoretical model that deepens the described connections (message–implication–risk/moderation), based on previous findings founded on different psychosocial behavioral theories: generational theory [26,27], source credibility [28], expectations [29], involvement with the product [25], and risk [30]. We tested the model in a context with high rates of social media participation: the Internet penetration rate in Spain stands at 94%; the number of social media users at 87.1% [31]; and recommendations made on fashion products were used, as it is one of the most represented product categories in influencer marketing studies [4,32].

2. Literature Review and Hypotheses

The so-called generational theory indicates that people of a similar age have shared similar formative experiences and cultural experiences, generating important commonalities in their sociodemographic traits [26,27]. For this reason, there is a high degree of differentiation compared to people of other age groups [33,34]. This approach is often used as a segmentation tool [17]. Most research agrees to discern up to four cohorts since 1925: Silent Generation (1925–1945), Baby Boomers (1940/5 to 1960/5), Generation X (1960/5–1980), Generation Y or Millennials (1977/1980–2000/5); although there is no widespread agreement to define exactly the thresholds of separation between one generation and another [17,33,34].

The existence of intergenerational differences has been recognized in previous literature in different research domains, e.g., mobile commerce (m-shopping) [35], e-consumption [36,37], fashion, and influencers [38]. This paper focuses on shedding light on the possible differences in receiving influencer-sponsored product recommendations between Generation X and Millennials. This is followed by a brief literature review of the proposed variables and relationships, ending with the introduction of the generational cohort as a moderating variable.

2.1. Relationship between Perceived Trustworthiness of the Message Transmitted and Purchase Intention

Trust has been positioned as a key variable in different disciplines: philosophy, psychology, sociology, politics, economics, organizational studies, and marketing [39,40]. It is a multidimensional and complex construct. Zur et al. [41] highlighted two dimensions of trust: cognitive and affective. The first represents the degree of security or the willingness that exists to rely on the reliability and competence of another partner, and the second refers to the security that a person places in another based on the feelings generated by the level of concern or care he or she shows. Pop, Săplăcan, Dabija, and Alt [39] cited Golembiewski and McConkie [42] to warn that “there is no other variable which so thoroughly influences interpersonal and intergroup behavior”. Shamim and Islam [40] used Moorman et al. [43] to define trust from a marketing perspective as “the inclination to rely on a trusted exchange partner”, and to point out how, from the perspective of signaling theory, trust has a significant influence on attitudes and purchase intention. In virtual environments, the trustworthiness of an information source is an essential aspect that determines how consumers respond to consumer recommendations [28]. Credibility theory and marketing research point out that it is founded on the reliability, attractiveness, and expertise of the communicator, and depends on the quality of the arguments and the persuasive force of the sender [44,45]. When the concept of trust reaches the realm of influencer marketing, it is exposed how followers and audiences of influencers tend to trust the content they generate [45,46]. Recent empirical research has highlighted that an influencer’s credibility is primarily based on their expertise and authenticity [47]. More recently, Shamim and Islam [40] found that message credibility plays a determining role in the development of trust towards influencers. In this research, we define perceived trust in the message conveyed by influencers according to the reasoning stated by Shamim and Islam [40] and Pop, Săplăcan, Dabija, and Alt [39] as the anticipation of a message that is safe, reliable. and cares about the product needs of its audience.

For example, Sokolova and Kefi [48], in a study on fashion Instagrammers and YouTubers, found a strong relationship with purchase intention derived from their promotions. In turn, Shamim and Islam [40] indicated how consumers trust their online opinion leaders, and that credibility has a strong positive impact on persuasion. Masuda et al. [49], through a survey of Korean consumers who had observed advertisements sponsored by influencers, observed the same relationship, accentuated by the parasocial relationships they generate. Based on the above, the first hypothesis is proposed:

Hypothesis 1 (H1).

Perceived trustworthiness of the message transmitted has a positive effect on the purchase intentions generated by influencers.

2.2. Relationship between the Trustworthiness of the Message and Perceived Risk in the Recommendations

From the expectations theory approach, consumers tend to gather information before making their decisions. Anticipating the product experience helps to verify that it will deliver what is expected, reducing perceived risk and increasing trust [29]. In this regard, reviews from other consumers and endorsements from influencers are considered credible sources of information [50]. However, behind a sponsored message there is usually some benefit [51]. Followers of influencers are increasingly aware of the existence of some underlying interest for the sender, and this in a way may anticipate some risk derived from their recommendations [19]. In this regard, the construct of perceived risk is one of the most prominent in studies of purchase behavior [52,53]. Perceived risk can be defined as the anticipation of possible loss linked to product purchase [30]. In this paper, the variable risk is defined in a non-restrictive and general sense, to refer to any type of negative utility associated with following a recommendation [54]. The relationship between perceived trustworthiness in influencers’ messages and the perceived risk derived from their recommendations has been little studied. Previous literature, with rare exceptions [19,55], assumes that, on the one hand, influencers are assumed to be highly trustworthy, as they are considered knowledgeable in specific domains such as fashion [4,10,11,14], and on the other, that they risk their own reputation in case they promote products that do not align with what their followers expect: “influencers must strengthen their credibility by acting ethically, or they could lose followers’ trust for not doing so. In order to avoid unethical perceptions in their audiences, influencers should disclose and justify their sponsorships” [56]. In addition, the recent review on the perceived risk variable by Wolff et al. [57] encourages us to pose the second hypothesis in a negative sense:

Hypothesis 2 (H2).

The trustworthiness of the messages transmitted by influencers has a negative effect on the perceived risk generated by their recommendations.

2.3. Perceived Risk in Recommendations as a Mitigating Factor in Purchase Intention

According to the above, influencers are considered experts in certain domains [11], opinion leaders who usually sponsor brands and products consistent with their image, and people close to their followers [58]. Indeed, the trust and commitment they generate in their audience are reasons why influencer marketing triumphs [5,7]. However, one might ask whether this fact completely eliminates the existence of risk, and renders the construct irrelevant, i.e., whether their followers are so committed that they do not consider any kind of loss caused by following any of their recommendations. In their research on influencers, Deshbhag and Mohan [19] found that perceived risk significantly influenced the purchase intentions of 250 Indian consumers of consumer goods. Previously, Biswas et al. [59] tested for differential effects concerning perceived risk depending on whether the endorsements come from a celebrity or an expert. In this research, in line with the precepts of the theory of planned behavior, it is assumed that behavior is determined by intention, and this, in addition to the antecedents mentioned in that framework (attitudes, subjective norm, and perceived control) [60], can be affected by other variables such as perceived risk. This has been shown by a multitude of research in different fields [61,62,63]. According to the above, the third hypothesis is proposed:

Hypothesis 3 (H3).

The perception of risk derived from recommendations made by influencers has a negative effect on the purchase intention derived from their recommendations.

2.4. Relationships between Fashion Involvement, Perceived Trustworthiness of the Message, Perceived Risk, and Purchase Intention

The recent work by Razzaq et al. [64] takes an approach to the complex concept of fashion involvement and defines it as: “an individual’s apparent peculiar significance or importance toward fashion clothing. Fashion involvement highlights the worth and application of clothing in the lives of consumers and portrays their consumption behavior” [65]. The origin of the construct is linked to the development of scales to capture the meaning that people attach to products, according to their inherent needs, values, and interests [25], i.e., the definition of Razzaq, Ansari, Razzaq, and Awan [64] attempts to particularize involvement in a specific product category. However, the complete review of the concept developed by Naderi [66] highlights the individualized existence of the variable, due to the considerable symbolic value that fashion items have for the user. Clothing allows expressing social identity, communicating messages, and turning fashion shopping into a social act [67]. In this sense, work on consumer behavior has highlighted that involvement affects purchase decision making [67,68]. Likewise, O’Cass [67] highlights that fashion involvement can be understood through attachment or the relationships between the consumer and fashion apparel, and defines fashion involvement as the personal degree to which a consumer feels that fashion is an attractive and central activity in his or her life. In this research, we adopt the definitions made by Razzaq, Ansari, Razzaq, and Awan [64] and O’Cass [67].

Regarding the relationship between fashion involvement and message trustworthiness of social media influencers, Parmar et al. [69] showed in an experimental study that the presence of a celebrity in advertising increases purchase impulse, and that endorsement by popular personalities is more effective with high-involvement products. People highly involved with fashion seek information to stay updated through their opinion leaders [4]. Consequently, involvement with the product could influence the credibility of recommendations. Thus, Reyes-Menendez et al. [70] found this relationship in their work on electronic word-of-mouth. On the other hand, Razzaq, Ansari, Razzaq, and Awan [64] also showed how buyers involved with fashion always want to be up to date on the latest trends and turn to so-called “fashion innovators” who help create awareness among their followers. This leads us to propose that the message of influencers could be considered more credible, as they are people who are seen by their followers as introducing new changes and trends. Based on the above, the hypothesis is put forward:

Hypothesis 4 (H4).

Fashion involvement has a positive effect on the trustworthiness of the message delivered by influencers.

Regarding the relationship between fashion involvement products and perceived risk, Hong [71], in his work on the online merchant selection process, expounds that involvement can be divided into long-term and situational: the first refers to the continued concern for a product, for its value to the person, usually derived from their positive experiences or their values; and the second refers to a momentary involvement based on the consumer’s desires to obtain the benefits of the product to satisfy a specific need. In addition, his research found that situational involvement had positive impacts on risk components. These types of relationships have previously been shown in exploratory consumer studies. For example, Bruwer and Campusano [72] argued that people with high levels of involvement spend more time informing themselves to obtain valuable product information and mitigate risks, showing a higher likelihood of experiencing risks as a function of their involvement with the product. Similarly, Bruwer and Cohen [73] revealed significant relationships between involvement and risk perception activation. Accordingly, it is hypothesized:

Hypothesis 5 (H5).

Fashion involvement has a positive effect on the perceived risk derived from recommendations made by influencers.

On the other hand, Mou et al. [74] point out that when involvement with a given product is high, consumers are more likely to evaluate it positively. This is because the consumer feels a higher emotional level with the product, which is transferred to the purchase intention: “the involvement theory states that a high level of product involvement could change consumers’ attitudes toward products and thus promote the generation of consumers’ purchase intentions and behaviors”. This form of direct relationship has been previously found, for example, on wine consumption [75], organic products [76], and online purchases of mobile phones [77].

Finally, regarding the relationship between fashion involvement and purchase intention, Naderi’s [66] summary traces articles that have extensively linked the two constructs, specifically to cover objectives related to the clarification of purchase intention, for example, the work of [78] in the retail domain. Therefore, the following hypothesis is posed:

Hypothesis 6 (H6).

Fashion involvement has a positive effect on purchase intention derived from recommendations by influencers.

2.5. Moderating Variables: Generational Cohort, Social Norm, and Gender

Moderating variables are those that systematically modify the direction or strength of the relationships between exogenous and endogenous variables such as those reviewed above [79,80]. Thus, moderation is used when faced with situations in which the relationship between variables (or constructs) is not expected to be constant, but could depend on a third variable [81]. In this paper, in line with the second research question, we explore the possible effects of three moderating variables: generational cohort, gender, and social norm.

Concerning the generational cohort, marketers usually take into account the particular characteristics of the set of users, potential customers, or buyers [80]. Some variables have traditionally been shown to modulate consumer interest, attitudes, and behaviors, such as age or gender, and have been used to segment and analyze their effects by reinforcing or attenuating other variables such as purchase intention [82,83,84]. Generational theory serves as a basis for grouping people based on their year of birth, as it has been recognized that certain cohorts have developed specific common characteristics based on similar sociocultural experiences [33]. In this sense, market segmentation by generational cohort is considered more efficient than if it is conducted simply by age [80]. Specifically, in the case of Generation X and Millennials, it has been identified that the former is much more skeptical and pragmatic, and that they tend to be more insecure people, which leads them to seek alternative information and try to establish whether their choices are correct; in contrast, Millennials are digital natives, they feel much more confident with technology, and for them social networks are part of their daily lives [17,33]. Under the premises of generational theory, these differences lead us to propose that they will receive influencers’ recommendations differently, which suggests proposing a first moderating effect by generational cohort. On the other hand, the moderating role of gender linked to generational cohort has also been observed on several occasions [85,86]. In addition, previous literature has pointed to differences in fashion shopping behaviors between men and women, for example, in evaluation from personal interaction processes [87], leading us to posit gender as a second moderating variable.

In this regard, previous literature has shown how generational cohort membership, as well as gender, can exert significant moderating effects on the different relationships proposed in consumer behavior research. For example, Jackson et al. [88] observed how the two variables moderate the perception of mall attributes and purchase value. More recently, Amin et al. [89] showed differences in tourists’ purchase satisfaction on repurchase intention, loyalty, and desire between Generation X and Y. This type of moderation has also been found in the realm of influencer endorsements on emerging platforms such as TikTok [90].

Finally, one would expect members of different generations to exhibit different levels of social norm associated with their generational cohort. The concept of social norm is analogous to the subjective norms of the theory of planned behavior [60,91], i.e., it refers to the perceived social pressure to act in a certain way, and to satisfy the desires of their environment and their referents by performing behaviors in line with what is expected. As explained by Hui-Wen Chuah, Sujanto, Sulistiawan, and Cheng-Xi Aw [91], this means that there is a set of rules and standards that are understood by society and that guide or restrict social behavior; transgression of such norms usually results in social sanction in the form of demonstrative disapproval or even discrimination. In this paper, we use social norm as a moderating variable according to the importance that the individual attaches to the opinion of his or her environment [92]. Based on this approach, we propose the last hypothesis:

Hypothesis 7 (H7).

The extent of perceived trustworthiness, perceived risk, fashion involvement, and purchase intention differ by (a) generational cohort, (b) gender, and (c) degree of social norm.

Based on the background reviewed and the set of hypotheses, the model is summarized in Figure 1.

3. Research Methodology

This paper is focused on developing an exploratory–descriptive approach to the effects of influencer purchase recommendations. Some of the variables used, perceived risk and involvement, have been little analyzed in influencer marketing research (previous section). Information was collected through a non-probabilistic online questionnaire by convenience and snowballing, encouraging respondents to distribute the message through their social media and among their closest circle of contacts. This technique is indicated to address emerging issues in social sciences [93]. The model proposed is specified from reflective measures with a partial least squares approach (PLS-SEM), appropriate for delving into new constructs when the theoretical basis is not settled [94], together with multigroup analysis (MGA) [81]. The questionnaire was designed from the theoretical basis exposed.

3.1. Data Collection and Sample Design

A pre-test was prepared beforehand and distributed on Facebook until 25 responses were reached. Once analyzed, some of the questions were reformulated and the final order was established (the pre-test responses were discarded and are not part of the study). The data were collected in November 2021, and the questionnaire remained hosted on the SurveyMonkey platform for 15 days, through a link and a generated QR code to get the answers..

Two populations are the object of analysis: Spanish buyers of a fashion product recommended by Generation X influencers (between 41 and 60 years old; born between 1981 and 1961) and Millennials (between 26 and 40 years old; born between 1995 and 1981), establishing the age ranges according to [33]. To verify compliance with the objectives, along with the age requirement, two filter questions were asked: “Do you follow any influencer on any social media?”; “In the last year have you made any purchase of a fashion product that you have seen recommended by an influencer?”. The questionnaire was distributed through Twitter, Facebook, and Instagram (mainly). At the beginning of the questionnaire, the objectives of the study, voluntary participation, guarantee of anonymity, and confidentiality were presented, and it was specified that by initiating the test, informed consent was given. A total of 527 responses were obtained; however, 276 were discarded for being outside the age range, for not having made any purchase of a sponsored product, for not following influencers, or for leaving the questionnaire incomplete, although the main reason for discarding was the second (more than 60%).

Table 1. Sociodemographic profile of the sample.

Variable	Categories	Male (n = 68)	Female (n = 183)	Total (%)
		27.0%	72.90%
Age	26–30 (M *)	26	43	69 (27.5%)
	31–35 (M)	10	23	33 (13.1%)
	36–40 (M)	4	10	14 (5.6%)
	41–45 (X *)	6	22	28 (11.2%)
	46–50 (X)	6	27	33 (13.1%)
	51–55 (X)	7	31	38 (15.1%)
	56–60 (X)	7	20	27 (10.8%)
	60 (X)	2	7	9 (3.6%)
Education	Primary level or lower	0	2	2 (0.8%)
	Middle school	3	19	22 (8.8%)
	High school	23	49	72 (28.7%)
	University	25	82	107 42.6%)
	Master	15	26	41 (16.3%)
	PhD	2	5	7 (2.8%)
Employment	Entrepreneur/self-employed	14	20	34 (13.5%)
	Employed	28	68	96 (38.2%)
	Officer	10	40	50 (19.9%)
	Student	11	15	26 (10.4%)
	Housework	1	21	22 (8.8%)
	Retired	2	5	7 (2.8%)
	Unemployed	2	14	16 (6.4%)
Available family income (EUR/month)	EUR < 1000	13	27	40 (15.9%)
	EUR 1001–2000	23	71	94 (37.5%)
	EUR 2001–3000	15	42	57 (22.7%)
	EUR 3001–4000	8	24	32 (12.7%)
	EUR 4001–5000	5	9	14 (5.6%)
	EUR 5.001	4	10	14 (5.6%)
Social Norm	High	38	89	127 (50,6%)
	Medium–low	30	94	124 (49,4%)

* M = Millennials; X = Generation X.

shows the sociodemographic profile of the sample, and a clear predominance of women can be observed (72.9% versus 27.1%), something quite frequent in research on influencers and fashion [4,95]. The distribution fits the profile of one of the most well-known influencers on fashion in Spain, María Pombo [96]. The educational level of the respondents is high: practically 72% have university studies or higher. Approximately 46% of the sample are representatives of the Millennial generation (16% men/30% women); 54% belong to Generation X (11% men/43% women). In terms of employment status, the distribution is homogeneous, with only the employed category standing out (38.2%). As for family disposable income, approximately 77% reported an income of less than EUR 3000. Finally, 50.6% showed a high social norm, compared to 49.4% with medium–low. In our analysis, the social norm variable is included together with the sociodemographic data as it is considered a moderating variable. It was measured with five items on a Likert scale of 1 to 7 adapted from [92]: 1. “My public image is very important to me”, 2. “I always look for ways to improve my relationships with the people I care about”, 3. “I strive to seek acceptance from my friends”, 4. “I identify with my social group in society”, 5. “I care what others think about the way I dress”. It was then transformed into a dichotomous variable following the proposals of Rasoolimanesh et al. (2021): scores between 0 and 25—medium–low social norm; and between 25 and 35—high social norm.

3.2. Measurement Instrument and Scales

Four blocks were added to the sociodemographic block, one for each construct: perceived trustworthiness, perceived risk, fashion involvement, and purchase intention. The items used to measure each of them are based on scales applied in previous literature and are shown in

Table 2. Measurement model and factor loadings.

Constructs	Items	Factor Loading	Average (Sd. Dev)	Adapted from:
Perceived Risk (PR)	PR1. I think it is risky to buy products recommended by influencers	0.933	4.116 (1.800)	[97,98]
	PR2. I am concerned about the result I will get if I buy a product sponsored by an influencer	0.868	3.980 (1.806)
	PR3. I am concerned about the overall risk I take by buying products recommended by influencers	0.807	3.773 (1.851)
Perceived Trustworthiness (PT)	PT1. I believe fashion influencers’ recommendations are honest	0.918	2.669 (1.609)	[40,44,99]
	PT2. I consider the recommendations of fashion influencers to be trustworthy	0.942	2.725 (1.634)
	PT3. I think fashion influencers’ recommendations are truthful	0.944	2.813 (1.600)
Fashion Involvement(FI)	FI1. I usually have one or more outfits of the very newest style	0.706	4.275 (2.251)	[25,64,100]
	FI2. I keep my wardrobe up-to-date with the changing fashions	0.825	3.028 (2.034)
	FI3. Fashionable, attractive styling is very important to me	0.869	3.131 (1.928)
	FI4. I am very involved with fashion. Fashion items are part of my way of life	0.855	3.104 (1.903)
Purchase Intention (PI)	PI1. I intend to buy fashion products recommended by influencers	0.925	2.104 (1.566)	[4,60]
	PI2. In the future, I will try to buy products sponsored by influencers	0.948	2.016 (1.453)
	PI3. I will effort to buy fashion products recommended by influencers	0.906	1.78 (1.432)

* All loadings have p < 0.001.

. The responses are expressed on a 7-point Likert scale (1—not at all agree; 7—strongly agree).

4. Results

To determine the possible existence of common method bias (CMB) in the data, a full collinearity test was performed. The variance inflation factors (VIF) obtained for all latent variables did not exceed in any case the cut-off value of 5 proposed by Kock [101]. Therefore, it is concluded that the empirical analysis was free of possible biases.

4.1. Measurement Model

For the estimation of the theoretical model through partial least squares (PLS), the SmartPLS 3.3.7 software was used. Table 2 shows the factor loadings of the items used in the measurement model, together with the mean and standard deviation. All factor loadings exceed the value of 0.7 proposed by Hair Jr et al. [102]. In turn, bootstrapping performed with 5000 iterations confirmed the statistical validity of the loadings through the Student’s t-test, with all p-values < 0.001 [103].

As shown in

Table 3. Internal consistency, convergent and discriminant validity of the measurement model.

					Fornell–Larcker Criterion
	Cronbach’s Alpha	Rho_A	Composite Reliability	AVE	PR	PT	FI	PI
Perceived Risk (PR)	0.855	1.088	0.904	0.759	0.871
Perceived Trustworthiness (PT)	0.928	0.929	0.954	0.874	−0.177	0.935
Fashion Involvement (FI)	0.832	0.848	0.888	0.666	−0.057	0.495	0.816
Purchase Intention (PI)	0.918	0.929	0.948	0.858	−0.239	0.543	0.360	0.926

, Cronbach’s alpha, Rho coefficients, and composite reliability (-CR-) of the constructs were calculated. In all cases the values exceeded 0.7 [102], thus being considered an acceptable level of reliability. The AVE criterion was also greater than the value of 0.5, thus accepting convergent validity (Hair Jr. et al., 2021). As for discriminant validity, it was verified through the Fornell–Larcker criterion: the values of the main diagonal exceed the rest of the values of the matrix [104], and the square roots of the shared variance between each construct and its measures (AVE), reflected in the main diagonal in bold, exceed the correlations of each construct with any other (rest of the matrix) [102].

4.2. Measurement Model

Once the measurement model was validated, we proceeded to estimate the structural model, examine the standardized path coefficients, and check compliance with the proposed hypotheses.

Table 4. Structural model. Path coefficients and hypotheses results.

Hypothesis	Independent Variable		Dependent Variable	Path Coefficient (p-Value)	Result
H1	PT		PI	0.453 (0.000) ***	Supported
H2	PT		PR	−0.197 (0.005) **	Supported
H3	RP		PI	−0.152 (0.003) **	Supported
H4	FI		PT	0.495 (0.000) ***	Supported
H5	FI		PR	0.041 (0.305)	Not supported
H6	FI		PI	0.127 (0.015) **	Supported

PT: perceived trustworthiness; PI: purchase intention; PR: perceived risk; FI: fashion involvement; *** p < 0.001; ** p < 0.05; * p < 0.1.

shows the results obtained for each of them according to the corresponding path coefficient βi, together with the respective p-value of the Student’s t-test obtained by bootstrapping to confirm the statistical significance of each parameter.

According to the results obtained, all the hypotheses posed concerning the relationships between the variables in the model can be accepted except H5. Consequently, perceived trustworthiness positively influences purchase intention, while it implies a reduction in the perceived risk of influencer followers (H1 and H2). Likewise, perceived risk negatively affects purchase intention, as proposed in (H3). Finally, fashion involvement contributes to an increase in both perceived trustworthiness (H4) and purchase intention (H6); however, it cannot be affirmed that it significantly affects risk perception, which leads to the rejection of hypothesis (H5) (the p-value associated with the path coefficient between these variables greatly exceeds 0.05).

As for the model’s goodness of fit,

Table 5. Model prediction and goodness of fit.

	R2	Adjusted R2
PI	0.328	0.320
PT	0.246	0.242
PR	0.032	0.025

shows the values of the R² and adjusted R² coefficients obtained. Both purchase intention and message credibility present values well above the 0.1 cut-off value established by Falk and Miller [105] as a determinant for considering that a model has sufficient predictive capacity. This is not the case, however, for perceived risk, which is only explained in the model by perceived trustworthiness and fashion involvement, with only the effect of the former being significant.

Finally, the standardized root mean square residual (SRMR), as an index of the average of standardized residuals between the observed and hypothetical covariance, obtained a value of 0.061, below the threshold value of 0.10 proposed by Ringle et al. [106] to consider that the model has a good fit.

4.3. PLS-MGA Multigroup Analysis

Once the structural model was estimated and the first six hypotheses were analyzed, the next step was to study the moderating variables. For this purpose, a bootstrap-based multigroup analysis (PLS-MGA) was performed for each of the three proposed variables, which must be divided into two groups for each case. Thus, gender already has such a split between males and females, just as for the generation variable, Millennials having been grouped as those respondents aged between 26 and 40 years (n = 116) and Generation X as those aged between 41 and 60 years (n = 135). Likewise, as previously mentioned, the social norm variable was subdivided according to the total score in the five items that compose it: low social norm group (0–25 points) formed by 124 respondents and high social norm group (26–35 points) formed by 127 people.

Before the application of PLS-MGA, the requirement of measurement invariance was verified through the MICOM procedure [103]. In this case, the first two requirements (configurational invariance and compositional invariance) are met but not the third one, equality of means and variances, which implies a partial invariance that, in any case, allows the comparison of groups in the three moderator variables.

Table 6. Multi-group analysis by generation and social norm.

Variable	Path Coefficients		Path Coefficient Difference	Henseler’s MGA	Welch–Satterthwaite Test
Generation	Millennials	Gen X	β1-β2	p-valor	Student’s t-test	p-valor
PT-PI	0.365 ***	0.541 ***	−0.176	0.070 *	1.470	0.072 *
RP-PI	−0.238 **	−0.076	−0.162	0.078 *	1.369	0.087 *
Social Norm	High	Medium–Low	β1-β2	p-valor	Student’s t-test	p-valor
PT-RP	−0.392 ***	−0.060	−0.332	0.034 **	1.934	0.055 *

*** p < 0.001; ** p < 0.05; * p < 0.1.

presents only those differences between groups that were significant for each variable and through the two approaches provided by the procedure (MGA and Welch–Satterthwaite test). From the results, it can be seen that the gender variable does not show differences in the relationships between the variables of the model in the responses of men and women, which is why it is not represented in the table.

Regarding the Generation variable, perceived trustworthiness has a positive and significant influence on purchase intention for both Millennials and Generation X; however, the coefficient is statistically higher for Generation X respondents. Consequently, influencer followers between the ages of 41 and 60 are more conditioned by message trustworthiness than younger respondents. On the other hand, the impact of risk perception on purchase intention is negative and significant in Millennials, but not statistically significant in Generation X, so the first group is more influenced when purchasing a fashion product by the risk they may perceive in the purchase.

Finally, the social norm groups show only one significant difference. In this case, perceived trustworthiness negatively conditions risk perception in respondents with higher social norm scores, while the medium or low social norm group does not show a statistically significant relationship between these constructs. Thus, those who feel more inclined to satisfy their community and comply with social requirements tend to consider the risk of a product to a lesser extent if they perceive that the message transmitted by the influencer is trustworthy.

5. Discussion and Conclusions

Returning to the two research questions, what the inter-relationships between perceived trustworthiness, fashion involvement, perceived risk, and purchase intention derived from influencer endorsements are, and how the Generational cohort might moderate these relationships, our results reflected that the perceived trustworthiness conveyed by influencers attenuates the perception of risk derived from recommendations and increases purchase intention. Additionally, fashion involvement showed a significant positive effect on the perceived trustworthiness, and also stimulated purchase intention, although its relationship with overall perceived risk could not be demonstrated. Regarding this last variable, the existence of a significant negative incidence for the intention was verified, which contributes to assessing the presence of risk anticipated by the followers. On the other hand, the multigroup analysis revealed that the perceived trustworthiness affects to a greater extent people belonging to Generation X, with statistically significant differences also appearing for risk in the case of Millennials. Finally, no moderating effects of gender were observed, but the social norm was positioned as a relevant construct in moderating the relationship between perceived trustworthiness and risk.

Our first finding reflected significant direct links between perceived trustworthiness as an antecedent of intention (H1). This result was not surprising, given that it is in line with the core body of knowledge on influencer marketing [10]. Among others, Kim and Kim [47], in a sample of 384 American influencer followers, showed how trust becomes a determinant driver between credibility, source attractiveness, loyalty, attitudes, and purchase intent generated by influencers. This same relationship has been found, even in industries linked to health, where the repercussions of a poor product choice could affect to a greater extent (because of its consequences) the intention to follow an influencer’s recommendations [107]. Second, perceived trustworthiness was positioned as a determining factor in reducing the risk of recommendations (H2). The inclusion of this variable in influencer research has hardly been considered, as it seems to be presumed that followers highly engaged with influencers do not experience reticence to their purchase advice [5,10]. Nevertheless, our findings are in line with Deshbhag and Mohan [19], and position trust as a prominent dimension for mitigating risk perception. Like Deshbhag and Mohan [19], it was also verified that risk acts as a mitigant of intention (H3). In this sense, from here we make a call to include the different dimensions of risk, and its attenuators in future research on influencers, as is being encouraged in other fields of research related to consumption [108].

Third, the exploratory analysis confirmed that involvement with the product positively affects perceived trustworthiness and purchase intention (H4, H6), in line with previous research such as Razzaq, Ansari, Razzaq, and Awan [64]; Reyes-Menendez, Saura, and Martinez-Navalon [70]; Parmar, Mann, and Ghuman [69]; and Chetioui, Benlafqih, and Lebdaoui [4]. This seems to indicate that followers of fashion influencers would have a high degree of knowledge about this product category and would see much more trustworthiness in the messages of influencers whom they consider their opinion leaders on fashion, encouraging their purchase intention. In contrast, it was surprising that the relationship between fashion involvement and perceived risk was not verified in any sense (H5); a possible explanation would be through the study of Dholakia [109], where he explains how risk could be considered an implicit measure of involvement, and to verify this type of relationship would require contemplating the different types of risk, as well as the two dimensions of involvement [71,108,109]. An example of the latter can be found in the work of Han and Kim [110], where high or low involvement with the product impacts risk according to their typology “high-involvement products induced a higher sense of financial, privacy, security, product, time, and social/psychological risk on Taobao than low- involvement products”.

Regarding the proposed heterogeneities (H7), it was striking that gender did not moderate any of the relationships in the model. There have been more than a few papers that have positioned the gender variable moderating linkage in the fashion domain, e.g., Wu and Chalip [111] or Bhatia [112]. In addition, it has also been found in recent research on information quality and source credibility, more specifically Dedeoglu [113]. It seems to be in this case that the Spanish context could have influenced the result, given that it is very advanced in terms of equality between women and men [114]. However, Dedeoglu [113] also pointed out that more empirical evidence is needed on the role of gender as a moderating variable in the field of social networks. Regarding the two generational cohorts studied, the results of the multigroup analysis seem to fit with preconceived ideas from generational theory [17,33]. Specifically, members of Generation X observe as more credible the messages of influencers, and this would indicate that being theoretically more insecure people, they seek alternative information through influencers whom they consider reliable sources in their field of action [17]. On the other hand, Millennials are more intensely affected by perceived risk, probably influenced by lower purchasing power [17,33]. The last moderator, social norm, was shown to be an important moderating factor in the relationship between message credibility and perceived risk, i.e., those who attach more importance to their social environment reduce perceived risks from influencers’ message credibility, in line with previous research that gives a key moderating role to social norm in risk perception [115].

5.1. Theoretical and Managerial Implications

Despite its potential drawbacks, e.g., the use of bots to artificially modify interactions, audience manipulation, and other unethical behaviors [116,117], the so-called influencer marketing has become a strategic communication tool for all types of companies to reach their potential customers naturally and effectively [9,116,118]. This research started with the objective of studying the effectiveness of influencers’ recommendations based on the inter-relationships between perceived trustworthiness, a variable extensively used in influencer research [10,40,48,56], and two constructs that need new empirical contributions in that field: risk perception and fashion involvement [5,7,19]. Furthermore, under the support of generational theory [26,27], possible moderating effects introduced by identity characteristics specific to Generation X and Millennials [17,33], including gender, and two levels of social norm were tested. From a theoretical point of view, this represents an advance, and adds to the open debate on the effectiveness of influencer marketing, by proposing an exploratory model that can be integrated into consolidated theoretical frameworks that will optimize the analysis of purchasing behavior, such as the theory of planned behavior [60].

On the other hand, the proposed exploratory model has a strong audience focus, targeting how the influencers’ message (trust) impacts intention, including risk perception, and the particular characteristics of followers for the recommended product category (fashion involvement). In addition, it considers generational differences between members of Generation X and Millennials. Consequently, the results offer several contributions to support influencers, marketers, and managers in the fashion field. First, companies using or considering influencer marketing campaigns should approach this possibility based on the level of involvement their product has with the influencer’s audience. Our study has shown that when it comes to products with high involvement, a double effect on intention is achieved (indirectly from the perceived trustworthiness of the message and directly on the purchase intention). Therefore, efforts should be made to find the right product, those with a particular characteristic linked to audience involvement (ecological, technological, etc.). Secondly, given that involvement has been shown to exert a positive influence on the trustworthiness of the message conveyed by the influencer, efforts should make sure that they choose well those products they wish to promote, since if they are not congruent with their image, or belong to categories of low involvement, they could deteriorate their credibility [23]. Thirdly, marketing managers should choose the right influencer based on segmentation by the generational cohort of their audience, as this will allow a better evaluation of the impact of the advertising messages transmitted [33,34].

5.2. Limitations and Future Research

By exploring the incidence of two frequently overlooked variables in the influencer marketing arena, this work adds to its knowledge base. Nevertheless, it is subject to limitations, some of which represent new and promising avenues of work. Firstly, although of particular interest for the Spanish textile trade (according to characteristics of Spanish fashion influencers’ followers [96]), the exploratory nature of the model, together with the technique used to collect the sample (convenience sampling), implies that this is a case study in each context and for a given product. This prevents generalization of the results, but sheds light on the direction of the proposed relationships, opening the way to future experimental designs with random probability sampling and a control group. Second, the scales used in some constructs, e.g., perceived risk and fashion involvement, do not fully capture their dimensions (types of risk, long-term or situational involvement) [71,108]. Future work should expand the questionnaire using extended measures to capture all the dimensions linked to risk and short- and long-term involvement with the product. Thirdly, the generational cohort theory that supports the inclusion of moderating variables should be verified in other geographical contexts, given that cultural differences in other countries may influence by introducing nuances in each generation with a possible impact on how Millennials and Generation X receive influencers’ recommendations. Finally, new avenues of research emerge if the relationships described here between the recommendations made by influencers and the risks perceived by the consumer are assessed. Behind a sponsored recommendation is a brand promoting a product. The economic interests of influencers can generate distrust, an aspect that has not been fully reflected in this work, and that can reduce purchase intentions [47].