The influence of persuasion knowledge, third-person
perception, and affect on coping behavior in the
Instagram stories feature
r
Ie
0
Tser-Yieth Chen1
Graduate Institute of International Business, National Taipei University Kang-Ting Wang
Department of Business Administration, National Taiwan University
Abstract: The purpose of this study was to introduce a novel augmented coping behavior model within the Instagram Stories feature. We employed a 3x3x3 research design that used 450 samples to manipulate the influence of three types of antecedents on the coping behaviors of people in Taiwan. The empirical results demonstrate that third-person perception and affect influence coping behavior through the interactive character of the Stories feature, and our results can be used to enhance the utilization of an elaboration likelihood model for the Instagram Stories feature. Moreover, the results show that persuasion knowledge does not influence coping behavior and that coping behavior can influence the persuasive effects of a narrative and electronic word-of-mouth. Hence, we explored and clarified the connections between “what audiences think,” “how audiences react,” and “how the Stories feature works”. The results provide useful managerial implications for businesses to use to enhance their marketing operations for the effective dissemination of official messages through the Stories feature.
Keywords: Coping behavior, stories feature, elaboration likelihood model,
Corresponding author: Tser-Yieth Chen, Professor, Graduate Institute of International Business, National Taipei University, No.151, University Road, Sanhsia District, New Taipei City, Taiwan ROC. E-mail: chenty@mail.ntpu.edu.tw.
third-person perception, persuasion knowledge g " " r 3 m Ie tI g 0 a Ien m s r 5 0 Ie Ie g x " " r Ie n r 4 5 g e r I n Ie r x 0
1. Introduction
The Instagram Stories feature (SF) is a new form of communication for Instagram users that was launched in August 2016. Amancio (2017) described it as a storytelling device that allows users to post about an event or thought instantly. A SF post remains on the platform for 24 hours. The desire to use this feature is influenced by the pleasure, convenience, relative advantage, and observability that users associate with it. According to the statistics portal Statista (2018b), the number of active users on Instagram, a photo- and video-based application, soared from 90 million in January 2013 to 1 billion in June 2018. This huge growth that Instagram has seen has altered the modes of interaction between audiences and brands. In particular, in June 2018, the Instagram SF was adopted by 400 million users worldwide each day. In contrast, the usage rate was 100 million per day in October 2016 at the launch of the feature (Statista, 2018a). Hence, the importance of this topic becomes clearer with the increasingly significant role of the Instagram SF in daily life. There is a growing interest in the effects that the stories posted by brands have on
audiences. Visual images, including dynamic images (e.g., videos) and static images (e.g., photographs), have become prevalent methods of communication (Newton, 2017). Thus, individuals are likely to share their own stories, express their feelings, or share other messages through visual images. With the emergence of social media platforms (SMPs), such as Facebook, Instagram, Snapchat, and Twitter, Romney and Johnson (2018) asserted that social interaction has become progressively dependent upon visual storytelling. In addition, an increasing number of individuals tend to receive information through social media (Casaló, Flavian, and Sergio, 2018), and thus, with the dominance of SMPs, visual storytelling has become increasingly vital.
Given the nature of this “perceived interactivity” in the Instagram SF, we argue that the SF should shed additional light on emotional responses and interpersonal relationships and that third-person perception (TPP) and affect (AFF) will play a critical role in influencing CB and this is the main focuses of our research. This study considers the implications of the perceived interactivity in the Instagram SF, and thus, provides a new perspective for coping behavior (CB) research (Romney and Johnson, 2018). It is appropriate to examine the proposed augmented coping behavior model and is our research motivation.
The aim of this study was to investigate the antecedents and consequences of CB. We sought to demonstrate the relationships between what audiences think (antecedent), how audiences react (CB), and how the SF works (consequences). The antecedent variables for CB included persuasion knowledge (PK), TPP, and AFF. The consequential variables incorporated narrative persuasion effect (NPE) and electronic word of mouth (e-WOM). In addition, this study investigated the factors that “prompt interactivity” in the SF, offering an alternate perspective on CB study (Romney and Johnson, 2018). Based on the essence of what prompts interactivity in the SF, we have determined that the SF highlights interpersonal relationships and emotional responses, and thus, TPP and AFF should play an imposing role in directing CB. The previous relevant literature has explored the causal relationship between PK, TPP, and CB or between e-WOM and CB, but AFF has not yet been included as an antecedent variable for CB nor has NPE yet
been regarded as a consequence variable for CB. As for the research gap of this study, the previous relevant literature has explored the causal relationship between PK, TPP, and CB or between e-WOM and CB. AFF had not been included as an antecedent variable of CB, and NPE had not been regarded as a consequence variable of CB. Therefore, there has been a gap in the research of the existing literature and helping to fill this gap is a main contribution of this study.
In our study, we have used that the elaboration likelihood model (ELM) proposed by Petty and Cacioppo (1986) to interpret the findings for the drivers of CB in the Instagram SF. In the ELM, when consumers positively seek for product/brand information and to shape attitude/behavior, the process is referred to as the central path, where consumers have the will or capability to handle the product-related message. However, when the consumer deals with the product/brand information with low engagement and low desire to shape an attitude/behavior, the process is referred to as the peripheral path, where the consumer does not have the will or ability to deal with the product-related message. The ELM indicates that different extents of self-relevance determine different information processing paths (Cyr, Head, Lim, and Stibe, 2018). Processing information with a high or low level of engagement relies on personal motivation, which is affected by personal relevance (Kruglanski and VanLange, 2012). The motivation that people have is more likely to be strengthened when a self-related stimulus resonates with them, and they also intend to engage in an activity or an interaction more actively once their motivation has been supported. Such a high level of engagement generates an enduring belief and attitude through a personal evaluation. People’s attitudes are employed when dealing with information (Boninger, Krosnick, and Berent, 1995). When people tackle messages with high motivation, they are more likely to react to a message, such as a SF posted by a brand, based on a personal evaluation caused by high engagement.
In our study, we have extended the works of Ham (2017), Ham and Nelson (2016), and Ham, Nelson and Das (2015), proposing three routes for the
persuasion process: PK, TPP, and AFF, to organize the cognitive learning factor, the cognitive processing factor, and the cognitive appraisal factor, respectively. Among these studies, Ham et al. (2015) developed reliable scales for PK and CB, Ham and Nelson (2016) indicated that PK and TPP would positively influence CB, and Ham (2017) proposed that PK would positively influence two types of CB (approach and avoidance CB).
We proposed PK as the cognitive learning factor as individuals are active participants in cognitive learning contexts when they encounter many cognitive elements in the PK (Boerman, Willemsen, and Van Der Aa, 2017). Furthermore, TPP and AFF were deemed as a cognitive processing factor and cognitive appraisal factor, respectively. In addition, Ham (2017) proposed that CB is a dependent variable for exploring two driving factors, the cognitive processing factor and cognitive appraisal factor, by using the persuasion knowledge model (PKM). First, the cognitive processing factor is a procedure that handles all of the messages we receive from the environment. There are many messages, and an individual’s brain works continuously (Tutaj and Reijmersdal, 2012). Perceived personalization and self-defense are two typical main cognitive processing mechanisms, and TPP can be deemed as a parallel term to perceived personalization (Ham and Nelson, 2016). Jang and Kim (2018) defined TPP as the assessment of individuals of the influence of phenomena on themselves and others while judging others who are dissimilar based on the self-enhancement theory. Second, the cognitive appraisal factor is an evaluation of an emotional situation wherein individuals will assess how an event will impact them, explain the different aspects of an event, and then reach a reaction based on that explanation (Kirmini and Campbell, 2004). AFF is one of the main emotional situations for conducting cognitive appraisals (Scherer, Shorr, and Johnstone, 2001). Yu, Hu and Cheng (2015) indicated that AFF is a feature of the connection between the individual and the environment rather than being comprised of only personal characteristics and elucidated the affect infusion model (AIM) to illustrate the influence of mood on the individual’s ability to tackle information (Kim, Han, Park, and Park, 2016).
In our study, we further employed e-WOM and NPE as two main outcomes to online advertising efforts. This was inspired by the work of Lyons, Huebner and Hills (2016). They employed CB as a mediator between life satisfaction and predictors referring to personality characteristics and environmental events. Thus, there are significant relationships among these two significant predictors (personality characteristics and environmental events), one mediator (CB), and the outcome (life satisfaction). As for the discussions from Dessart (2018), Boerman et al. (2017), Wang, Yeh, Chen and Tsydypov (2016), and Hassan, Nadzim and Shiratuddin (2015), we have found that e-WOM and NPE were employed to represent the effects of advertising on social media. Thus, we have employed e-WOM and NPE as two main outcomes in our study.
CB has been defined as the mental and behavioral responses to events or problems (Okafor, Lucier-Greer, and Mancini, 2016; Snyder, 1999). This study has categorized CB into three sub-items: share, see more, and ignore (including blocks and reports) in the Instagram SF. Our study then expands the PKM to the Instagram platform setting by introducing the work of Ham (2017) and Ham and Nelson (2016). In summary, our study has been the first to encompass TPP and AFF in an investigation of audience responses to the Instagram SF.
As for the antecedents, PK refers to a perspective on marketing-related messages (Boerman et al., 2017). Based on the PKM by Friestad and Wright (1994), individuals intend to respond to information from agents, for example salespeople, based on existing knowledge (Ham et al., 2015). The PKM indicates that targets, such as consumers who obtain information, intend to cope with the information from agents through three types of knowledge (e.g., topic, agent, and PK), which we will consider in the Instagram SF (Friestad and Wright, 1994; Hwang and Zhang, 2018). Jang and Kim (2018) considered TPP as the judgements of individuals about the influence of circumstances on themselves and on others. In addition, the responses to messages depend on the evaluations of individuals of the influence that the messages have on themselves and others (Ham and Nelson, 2016). Moreover, AFF is a sensory understanding of an impression or image concerning a product or service experience (Kim et al.,
2016). These factors also play a role in the Instagram SF.
Berger (2014) indicated that e-WOM is the social sharing of messages between two or more online users. Social media have shifted the constraints and limitations of interpersonal communications, and they have developed various modes of interaction (Daugherty and Hoffman, 2014). With the persisting dependence of consumers on social media such as Instagram, Twitter, and Facebook, e-WOM has become more critical to the facilitation of message seeking, for example for a product message (Chu and Kim, 2018). In our study, NPE refers to the effect that a message has on peoples’ beliefs, attitudes, and behaviors through the Instagram SF (Cho, Shen, and Wilson, 2014). Please note that narrative thinking promotes emotions that are positively connected with brands and images in individuals (Escalas, 2004).
2. Literature review and hypothesis development
Previous research has investigated CB to observe how people respond to situations and to link further individual responses to different topics through observation (Lyons et al., 2016; Fransen, Verlegh, Kirmani, and Smit, 2015; Smit, Van Noort, and Voorveld, 2014). Smit et al. (2014) investigated CB to explore the issue of online behavioral advertising (OBA), cookies, and privacy. They classified CBs into two types: coping by approach (Raman and Pashupati, 2004) and by avoidance (McDonald and Cranor, 2010). Their empirical results showed that one third of participants decided to avoid OBA by withdrawing from using websites and cookies, and their engagement of privacy protection was not caused by an abundant knowledge of OBA or cookies but out of a concern for lacking an understanding of them (Gibs and Bruich, 2010; Hajli, 2014). Furthermore, Fransen et al. (2015) illustrated the ACE typology, which includes three negative types of CB that audiences use as resistance in confronting advertisements: avoiding, contesting, and empowering (Knowles and Linn, 2004). They also indicated strategies for neutralizing these negative reactions to advertisements individually. It is effective for brands to alleviate the resistance of individuals who are observing advertisements and to promote persuasion through
communicative intentions.
In this study, CB is viewed as a mediator. We adopted “see more” and “share” as the approach and used “ignore” as the avoidance for the Instagram SF. The main point of this study was to explore the CBs of audiences to their own cognitions (PK, TPP, and AFF) and to investigate the subsequent outcomes (e-WOM and NPE) in the Instagram SF. The cognitive learning, cognitive processing, and cognitive appraisal factors were used to investigate the effects of the antecedents on CB. As there are three steps in the standard pattern of reflective behavior: stimulus, reaction, and outcome, this study considered PK, TPP, and AFF as stimulus was, CB as reaction, and e-WOM and NPE as outcome. Lyons et al. (2016) used CB as a mediator between life satisfaction and predictors such as personality characteristics and environmental events. In their study, CB was classified into approach coping and avoidance coping. Note that there are a variety of significant relationships and different individual pathways among predictors (personality and environmental variables), mediators (approach coping and avoidance coping) and outcomes (life satisfaction)(Lyons et al., 2016). In our study, PK and TPP are types of environmental events, and AFF is a type of personal characteristic, and the outcome in our study is e-WOM and NPE, which are types of life satisfaction. Thus, individual behavioral responses can be determined as a result of the survey or mediator of the research (Fransen et al., 2015; Lyons et al., 2016; Smit et al., 2014).
In order to explore useful insights into different effects on CB using cognitive learning, cognitive processing, and cognitive appraisal factors, our study considered “what audiences think” as classified into three parts. First, PK was defined as to the interaction between audiences and events/objects in the Instagram SF (Boerman et al., 2017). This was based on cognitive learning and PKM, which indicate that people intend to cope with an event, such as receiving information from an agent, concerning existing knowledge (Friestad and Wright, 1994). Second, TPP was defined as the interaction between audiences and other people in the Instagram SF (Ham and Nelson, 2016). This was based on the social penetration theory (SPT) proposed by Altman and Taylor (1973), which
states that interpersonal relationships are dependent on intimacy levels from the cognitive process (Jang and Kim, 2018). Third, AFF was defined as the inner situations of the audience of the Instagram SF (Chen and Ng, 2016). This was based on the affect infusion model (AIM), which states that cognitive appraisal is influenced by mood and emotion (Forgas, 1995). Collectively, PK, TPP, and AFF show “what audiences think” while yielding a complete cognitive structure through the application of PKM, SPT, and AIM, respectively.
The structure of this study is presented in Figure 1. Figure 1 shows the processing of the investigated audiences of posts using the SF placed by brands and the effects that these posts had on Instagram audiences. First, the PK, TPP, and AFF that shaped judgement were accepted in order to obtain a deep understanding of the perceptions and cognitions of the Instagram audience. Second, based on PK, TPP, and AFF, the audience responses in the SF were investigated with the designed operation. Third, e-WOM and NPE were examined after the audience reacted to the SF.
Here we list all of the definitions of the variables in order to present their meanings as used in this study clearly. PK is a perspective or evaluation of marketing-related messages (Boerman et al., 2017). TPP is the judgement of individuals regarding the influence of circumstances on themselves and others (Jang and Kim, 2018). AFF is the positive or negative sensory feelings from an impression or image concerning a product or service experience that encompasses both mood and emotion (Forgas, 1995; Kim et al., 2016). CB is the individual mental and behavioral responses to events or problems (Okafor et al., 2016; Snyder, 1999). e-WOM is the personal network sharing of positive or negative statements between two or more online users (Berger, 2014; Hennig-Thurau, Gwinner, Walsh, and Gremler, 2004). NPE is the effect of a message on a person’s beliefs, attitudes, and behaviors through stories (Cho et al., 2014). Table 1 lists the abbreviations and definitions of the studied variables. 2.1 Relationship between PK and coping behavior
Figure 1
The research framework of the study model
Table 1
Abbreviations and definitions of the studied variables
Variables Abbreviations Definitions
persuasion knowledge PK PK is a perspective or an evaluation to marketing-related messages. third-person perception
TPP TPP is the individuals’ judgements about the influence of circumstance on themselves and others.
affect AFF AFF is a positive or negative sensory feeling of an impression or an image concerned as a product or a service experience that encompasses both mood and emotion.
coping behavior CB CB is the individual mental and behavioral responses to events or problems.
e-word-of-mouth e-WOM e-WOM is personal network sharing of positive or negative statements between two or more online users.
narrative
persuasion effect
NPE NPE is an effect of a message treatment to peoples’ beliefs, attitudes, and behaviors through stories.
remember the messages; they decide how to cope with the persuasion attempts (Ham, et al., 2015). This is the type of cognitive learning mechanism that individuals encounters from persuasion knowledge providers in the Instagram SF. It is remarked by many cognitive elements, such as vocabulary, image, voice, and expressions of PK (Hamby and Brinberg, 2018). Cognitive learning indicates that people are active participants in the learning process. They proactively present assumptions, validate them, and solve problems through on persuasion knowledge in the situation (Matthes and Naderer, 2016). In order to meet the requirements and limitations in the Instagram SF, individuals also reorganize their experience and readjust or change their cognitive structures of the environment (Brandon, Beike, and Cole, 2017).
Based on Boerman et al. (2017), people consume such messages to catch marketers’ purposes and persuasive intentions and to germinate viewpoints, i.e., PK. Individuals are desired to cope with an event, such as an advertisement, with prior knowledge in the PKM (Friestad and Wright, 1994). This could incorporate the knowledge linked to an activity or a product, the event maker, e.g., a brand, and the awareness of prior persuasion intention (Ham and Nelson, 2016). Therefore, knowledge owns an impact on the realization of an event or an object. Responses to messages intend to be based on evaluation that includes prior knowledge in the Instagram SF.
Knowledge of a subject embodies the judgement of the subject and its judgement is the effective predictor of this behavior. The usage of an judgement is vital for information processing, decision-making, and behavior (Boninger et
al., 1995). Since personal judgements are based on current knowledge, the
judgements symbolize the personal evaluations of an event or a subject and that personal judgements could affect responses (Hamby and Brinberg, 2018). Also, interactions among the groups and information systems produce specific symbolic meanings that satisfy self-valued performance (Hsu, Fan, Chen, and Wang, 2013). Therefore, reactions and behaviors are linked to perceptions and experiences in the Instagram SF. The following hypotheses H1 was proposed:
2.2 Relationship between TPP and coping behavior
TPP posits that individuals believe they are less affected than others by a message or an event than others, which also applied to the Instagram SF (Jang and Kim, 2018). From self-enhancement theory, individuals are likely to refuse persuasion attempts considered unwanted and undersirable in order to imporove their positive self-esteem by judging others are dissimilar to them (Jung and Park, 2018). That is a kind of cognitive processing mechanism that individuals are in charge of messages received from the SF environment and that is their brains handle a few of tasks constinuously (Brinol, Rucker, and Petty, 2015). Since that information is encoded by individuals to give meaning and can be compared with their own interests, especially in the perceived personalization and self defense mechanism (Ham, 2017).
Individuals would like to develop misestimated expectations about the Instagram SF media effects of a message on themselves and others (Sun, Pan, and Shen, 2008). From self-enhancement motivation, individuals would like to contradict personal susceptibility to messages considered negative impacts and accept messages considered positive impacts (Gunther and Mundy, 1993; Perloff, 1989). They perceive themselves to boast stronger ability than others to refuse negative effects from Instagram SF media (Park and Salmon, 2005; Perloff, 1989). This would influence their subsequent behavioral responses. Therefore, TPP-based perspectives have an effect on individuals’ judgements of SF messages and the perception of its effects on others represents the subjective evaluation of the subject (Chung, Munno, and Moritz, 2015). Evaluating perceived judgement can be employed to predict the behavior. Individuals intend to adopt an evaluation to process information, make a decision, and behave (Boninger et al., 1995). In summary, personal judgements are based on the perception of its effects on others in the Instagram SF. Judgements, which are composed of conceptual disparity, result in diverse cognitive, attitudinal, and behavioral consequences (Chung et al., 2015; Peiser and Peter, 2000). Hence, reactions and behaviors are connected with perceptions and experiences in the Instagram SF. The following hypotheses H2 was proposed.
Hypothesis 2. TPP would positive influence coping behavior.
2.3 Relationship between AFF and coping behavior
AFF refers to a positive or negative sense of an image or an impression about the targeted object or message (Forgas, 1995). It is the individuals’ evaluation of an emotional stiuation when they assess their cognition in the Instagram SF environment (Ham, 2017). That type of cognitive appraisal mechanism allow a person to evaluate how the event will impact on them on their specific explanation to the SF post (Wachyudy and Sumiyana, 2018). Since that different emotions, mood, and sentiments will influence individuals to assess the Instagram SF post (Seo, Li, Choi, and Yoon, 2018). Under these, affect infusion model can play an important role since the mood of the individuals will significantly influence the tackle of the SF post (Yu et al., 2015).
Based on the AIM, information judging is influenced by mood and emotion (Forgas, 1995). Different affective conditions provide message recipients slightly distinct information so that their decision-making could be altered (Raghunathan and Pham, 1999), including job satisfaction (Weiss, Nicholas, and Daus, 1999), shopping behavior (Mittal and Ross, 1998), attitude change or persuasion (Petty, DeSteno, and Rucker, 2001), etc. Affective states have a more important role than cognitive factors in ubiquitous media systems (Zhang, 2013). Therefore, affect influences peoples’ judgements that messages should be processed, especially in increasingly intricate and unexpected situations likes the Instagram SF (Wachyudy and Sumiyana, 2018). It is brought out at first and make a swift response as individuals encounter a SF post (Lerner, Valdesolo, and Kassam, 2015). Similarly, the feelings about the subject mean the evaluation of the subject in the Instagram SF. The adoption of an evaluation is employed to decision-making, information processing, and subsequent behavior (Boninger et
al., 1995). Judgements symbolize the personal evaluations about issues and
personal evaluation would affect reactions or responses in the Instagram SF (Kim et al., 2016). The following hypotheses, H3 was therefore proposed.
2.4 Relationship between coping behavior and e-WOM
Berger (2014) denoted e-WOM as the social sharing of messages (e.g., products or brands) between two or more cyber users. Based on the self-perception theory (Bem, 1967), users evaluate their psychological situation with their external behaviors. Therefore, behavior affects judgements of a subject, and this is deemed as attitude (Peter and Olson, 1987). After showing attitudes, mental states are linked to satisfaction and dissatisfaction, that has been examined as a reaction of their fulfilment (Oliver, 1997). Satisfaction provides a pleasurable level of fulfilment concerning a subject (brand or story in the Instagram SF) by way of perspectives on a product, service, or activity.
The positive relationship between satisfaction and WOM is investigated (Sweeney and Swait, 2008; Brown, Barry, Dacin, and Gunst, 2005; Oliver and Swan, 1989). Based on Casaló et al. (2018), an increasing number of users obtain message through social media (e.g., Instagram). Opinions or statements about subjects, such as products or services, are shared by Instagram SF in a variety of channel through the Internet. The current research classified e-WOM into two sub-items: brand and story. Shaikh, Karjaluoto and Hakkinen (2018) indicated that brand satisfaction intended to influence brand e-WOM, and story satisfaction intended to influence story e-WOM. In summary, the CB (block or share) is the users’ attitudes about the SF post. Their satisfaction levels are produced by their psychological situation; therefore, satisfaction is shaped by feelings and evaluation. Satisfaction levels are determined in the statements/comments made by users in the Instagram SF. Thus, H4 was proposed:
Hypothesis 4. Coping behavior would have a positive influence on e-WOM.
2.5 Relationship between coping behavior and NPE
In the current study, CB was shown by four types: sharing, seeing more, blocking, and reporting in the Instagram SF. NPE was divided into reminiscence, purchase behavior, psychological, and awareness. Based on Prendergast, Ko and Yuen (2010), purchase intention could be higher for users who obtain shared
comments, experiences or interests, particularly positive messages, from a related forum. Social media, like Instagram, users can offer and share their expression about an SF post. Users’ shared values have a positive impact on trusting belief (Wu, Chen, and Chung, 2010). Hajli (2014) indicated the significant influence of trusting belief on purchase intention. Trust enhances a relationship between shared opinions or suggestions and purchase intention.
In the extension of the theory of planned behavior, Pavlou and Fygenson (2006) illustrated that purchase intention promoted product purchase. Thus, the “share” in the Instagram SF is related to the purchase behavior. The psychological effect means an individual belief to a subject, such as favourable sense and amusing content (Lee, Lee, and Park, 2009). The “see more” in the Instagram SF is similar to the “learn more” in the Facebook used by Chen, Yeh and Chang (2018). It supplies a linkage for individuals to immediately seek more messages to subjects or products. Based on the study of Ashcroft and Hoey (2001), the reminiscence effect means the frequency of the recall of an advertisement (Lee et al., 2009). Please note that people who are attracted by an object, like an advertisement, are more likely to seek related messages (Ashcroft and Hoey, 2001). The “see more” provides an opportunity in the Instagram SF. Mehta and Purvis (2006) indicated a relationship between liking and recall. The “see more” could augment or decline audience attention and interest about the watched Instagram SF. Hamilton (2015) argued that retrieval of memories is influenced by the methods in which they are reserved. The reminiscence effect elucidates the impact of memory retrieval on the frequency of recall. Thus, the “see more” in the Instagram SF is linked to the reminiscence effect. The awareness effect means attention (Ashcroft and Hoey, 2001). Audience attraction determines the awareness effect. The usage of the “block” or “report” in the Instagram SF refers to the end of story viewing, i.e., the halt of the information flow and the decrease of exposure to the story. Exposure to an advertisement in a social network has been proved to be positively related to awareness (Gibs and Bruich, 2010). Also, consumers can recognize cognitive information that influences stickiness so that marketing managers can employ to maximize the
volume of information scent and message framing (Chen, Hsu, and Wu, 2019). As a result, the “block” and the “report” are linked to the awareness effect. In short, the dimension of CB enjoys a relationship with the NPE in the Instagram SF. Thus, H5 was proposed:
Hypothesis 5. Coping behavior would positive influence NPE.
3. Methodology
3.1 Measurement and scale
The questionnaire were designed as follows: firstly, based on the research of Pauwels, Aksehirli and Lackman (2016), e-WOM was divided into two types: brand (EW1) and story (EW2). Six items related to e-WOM were developed based on previous researches (Yen and Tang, 2019; Hwang and Zhang, 2018; Boerman et al., 2017). Secondly, NPE were divided into four dimensions (Pozharliev, Verbeke, and Bagozzi, 2017; Lee et al., 2009): awareness (NPE1), reminiscence (NPE2), psychological (NPE3), and purchase behavior (NPE4). The 12 items on NPE were developed in accordance with Lee et al. (2009) and Gupta, Singh and Sinha (2017). Thirdly, CB was divided into three dimensions based on Ham and Nelson (2016): share (CB1), see more (CB2), and ignore (CB3). The 9 items on coping behavior were developed from Ham and Nelson (2016). Fourthly, PK was divided into three dimensions (Brandon et al., 2017; Friestad and Wright, 1994): agent (PK1), topic (PK2), and episodic (PK3). The nine items on PK were developed on the basis of Mo, Liu and Liu (2018); Brandon et al. (2017); and Hartmann, Apaolaza and Eisend (2016). Fifthly, the present study extended the out-group genres of TPP (Chung et al., 2015) and divided it into three types: impact on self (TPP1), friends (TPP2), and family members (TPP3). Nine items were designed based on Ham and Nelson (2016). Lastly, in line with Kim et al. (2016), AFF was divided into three types: primitive (AFF1), descriptive (AFF2), and evaluative (AFF3). These were developed to nine items. All of the items were evaluated with 6-point Likert scales.
3.2 Research design
As for the sampling method, we employed a designed sample structure of 450 samples of males and females numbering 207 (45.1%) and 243 (54.9%), respectively. As for age, those under 24 years old totaled 173 (38.4%), 25–44 years old totaled 237 (52.7%), and over 45 years old totaled 40 (8.9%). The geographical distribution of the population of the Instagram users among the different regions of Taiwan was: northern Taiwan totaled 214 (47.5%), central Taiwan totaled 86 (19.1%), southern Taiwan totaled 140 (31.0%), and eastern Taiwan and the outer islands totaled 10 (2.3%). Table 1 shows the sample structure. In detail, we designed the dispatch of our 450 samples as follows. For males, 80 participants were under 24 years old, 109 were 25–44 years old, and 18 were over 45 years old for a total of 207 males. For females, 93 participants were under 24 years old, 128 were 25–44 years old, and 22 were over 45 years old for a total of 243 females (Table 2).
We determined what ratio of the population that had engaged in Instagram use based on gender, age, and region from a survey of Taiwan’s online users, and we then determined the sample structure and how many samples for each cell we needed in order to prove the representativeness of the samples. The number of Instagram users in Taiwan is estimated to be 6.46 million people, accounting for 27.1% of the total population of Taiwan in 2019. Among the Instagram users, the gender, age, and region volumes and ratios were obtained from NapoleonCat (2019). Among this population, 2.98 million males accounted for 46.1% and 3.48 million females accounted for 53.9%. As for age, we divided the Instagram users into three different age groups: there were 2.48 million (38.4%) under 24 years old, 3.4 million (52.5%) 25–44 years old, and 0.58 million (9.1%) over 45 years-old. A total of 91% of global Instagram users were aged under 44 years old. In other words, the number of older people using Instagram was less than one-tenth of users. As for the regions of Taiwan, Taiwan was divided into northern Taiwan, central Taiwan, southern Taiwan, and eastern Taiwan and the outer islands. Instagram users in the northern region accounted for 3.07 million (47.5%), users in the central region were 1.23 million (19.1%), users in the
southern region were 2.01 million (31.0%), and users in the eastern region and the outer islands were 0.15 million (2.3%).
To confirm the representativeness of the samples, quota sampling, which requires that the statistical features of the samples meet those of the population, was adopted for our study. The samples were classified by age and gender, the ordinary demographics that were deemed as segmentation variables. All samples were collected via SurveyCake, who boosted their statistical assistance through multiple online surveys in March 2019.
The research design was developed so that all of the respondents would participate in a three-factorial between-subjects experiment with 3 levels of PK (agent, topic, and episodic) × 3 levels of TPP (impact on self, family, and friends) × 3 levels of AFF (primitive, descriptive, and evaluative affects). Among the 450 samples in the sample structure, they were each randomly assigned to view one of 27 versions (3x3x3) of a targeted newsfeed to meet the representative sample. When the participants had finished watching, they provided the dependent variable measure: CB. PK was investigated by the respondents watching three types of SF posts: agent-, topic-, and episodic-based. The agent-based SF post consisted of the obvious and visible images of a brand, product, product description, and hash tags. The topic-based SF post showed mainly an activity with slogans, theme photographs, brief statements, and hashtags. The episodic-based SF post showed reminiscence themes with slogans, life philosophies, and hashtags. Each SF post was presented for about 15 seconds and in each type of SF post was embedded one type of TPP and AFF.
Our design facilitated the discernment of the effects of PK on CB. The mobile device has become the main instrument for interactions on social networks. This study emphasized the effects of PK, TPP, and AFF on initial audience responses rather than general actions. Hence, the CB in our study was narrowly defined and distinct from normal behavior.
As for the experimental procedure, an online experiment was employed to explore the effects of the Instagram SF. Prior studies have employed online experiments (Chen et al., 2018; Deters and Mehl, 2013). The CBs of the
participants were observed while they watched of the SF post on Instagram. Each audience member was randomly confronted with one of 27 specifically designed SF posts; therefore, they were each randomly assigned to an SF post until the end of the experiment. After watching the SF post, they were asked to choose an option: “see more,” “share,” or “ignore.”
Table 2
Sample structure of our study
Items
Under 24 years-old 25-44 years-old Over 45 years-old Total
(%)
Male Female Male Female Male Female
North Taiwan 38 44 51 61 10 10 214 (47.5%)
Central Taiwan 15 18 21 25 3 4 86 (19.1%)
South Taiwan 25 29 34 39 6 7 140 (31.0%)
East Taiwan and
Islands 2 2 2 3 0 1 10 (2.3%) Sub-total 80 (17.8) 93 (20.7) 108 (24.0) 128 (28.4) 19 (4.2) 22 (4.9) 207 (M, 46.1%) / 243 (F, 53.9%) Total 173 (38.4%) 236 (52.5) 41 (9.1) 450 (100.0%)
Note. Total male is 207 (46.1%) and total female is 243 (53.9%).
4. Empirical results
In the pre-test survey, 24 people, including 10 males and 14 females, were invited to participate. The demographic characteristics of the participants were classified into several groups with respect to gender, age, and region of residence. The proportion of women to men in this survey was 60% to 40%. For the age distribution, the number of 18 to 24-year-old respondents accounted for up to 91.7% of the samples. As for region of residence, close to 95.8% of the individuals in this per-test survey lived in northern Taiwan. In addition, a Cronbach’s α and the items for a total correlation was employed to examine the reliability of the pre-test survey. Table 3 shows that the Cronbach’s α value for each individual construct in the survey was 0.818, 0.821, 0.943, 0.708, 0.959,
and 0.918 for PK, TPP, AFF, CB, EW, and NPE, respectively, and all were greater than 0.7, indicating that each construct has met a standard level of internal consistency and reliability.
As also shown in Table 3, of the 450 valid questionnaires collected in March 2019, males accounted for 43.2% and females, the remainder. A total of 5.5% of the participants were under 18 years old, 34.6% were 18–24 years old, 36.1% were 25–34 years old, 17.3% were 35–44 years old, 3.8% were 45–54 years old, and 2.6% were over 55 years old. Table 4 shows the means, standard deviations, and Pearson correlation matrices of the variables.
We further conducted a one-way ANOVA (analysis of variance) to examine whether the sample characteristics would affect the results (e-WOM and NPE) as it can indicate if a sample is representative or not. We adopted age, gender, and region as the characteristics that may affect results. The results for e-WOM indicated p-values for gender, age, and region that were 0.004, 0.296, and 0.394, respectively. As for NPE, the p-values for gender, age, and region were 0.257, 0.070, and 0.837, respectively. These results show that age and region in Taiwan are insignificant in their effect on e-WOM and NPE (p-value > 0.05).
4.1 Common method variance analysis
Based on Podsakoff, MacKenzie, Lee and Podsakoff (2003), common method variance (CMV), spurious correlations between two variables evaluated by the same method, is a potential issue in self-reported surveys in behavioral research. The present study avoided the risk of a spurious relationship between the independent and dependent variables by employing a cover photo at the interval between them. Using a Harman’s one-factor test with an unrotated method, eight factors with eigenvalues were obtained that were greater than 1.0 rather than a single factor for all the items. Table 5 demonstrates that these eight factors accounted for 72.21% of the total variance. None were significant since the percentage of explained variance regarding the first factor, 49.176, did not exceed 0.5 (Malhotra, Kim, and Patil, 2006). Hence, CMV was not a serious issue.
Table 3
Demographics characteristics of 450 samples
Table 4
Means, standard deviations, and Pearson correlation matrices of the variables
Variables Mean Standard
Deviation A B C D E F A. persuasion knowledge 3.42 1.01 1.000 B. third-person perception 3.25 0.98 0.757 1.000 C. affect 3.57 1.10 0.740 0.679 1.000 D. coping behavior 3.18 0.97 0.666 0.669 0.753 1.000 E. e-WOM 3.20 1.07 0.667 0.700 0.628 0.700 1.000 F. narrative persuasion effect 3.31 1.08 0.778 0.760 0.845 0.843 0.810 1.000
Characteristics Category Count Percentage (%) Cumulative percentage (%) Gender Male Female 195 255 43.2 56.8 43.2 100.0
Age Under 18 years old
18 ~ 24 years old 25 ~ 34 years old 35 ~ 44 years old 45 ~ 54 years old Above 55 years old
25 155 163 78 17 12 5.5 34.6 36.1 17.3 3.8 2.7 5.5 40.1 76.3 93.6 97.3 100.0 Region Northern Taiwan
Central Taiwan Southern Taiwan Eastern Taiwan and islands district 195 103 146 6 43.5 22.8 32.4 1.3 43.5 66.3 98.7 100.0 Total -- 450 100.0 100.0
Table 5
Common method variance analysis of this study
Number of factors Factor loading Percentage of explained variance
Accumulation percentage of explained variance 1 25.571 49.176 49.176 2 2.692 5.177 54.353 3 2.208 4.247 58.600 4 1.836 3.530 62.130 5 1.726 3.320 65.450 6 1.336 2.569 68.019 7 1.104 2.124 70.142 8 1.076 2.069 72.211
4.2 Reliability and validity analysis
This study showed that the Cronbach’s α-values for PK, TPP, AFF, CB, e-WOM, and NPE (0.826, 0.848, 0.935, 0.725, 0.925, and 0.931, respectively) all exceeded 0.5, which shows that the internal consistency of the factors has been held (Nunnally, 1978). In addition, as argued by Fornell and Larcker (1981), the value of CR should be at least 0.6, which is considered as the primary standard for an acceptable fit of the data. The higher the value of CR, the less difficultly the factor has in measuring this variable. The CR values for PK, TPP, AFF, CB, e-WOM, and NPE (0.782, 0.812, 0.915, 0.713, 0.910, and 0.907, respectively) wholly exceeded the fundamental level. CR was calculated with the following formula: CR = (sum of standardized loading)2 / [(sum of standardized loading)2 + (sum of measurement error)]. In order to interpret the proportion of precisely elucidated variance within an examined measurement, the values of the average variance extracted (AVE) from each variable (0.544, 0.591, 0.781, 0.500, 0.834, and 0.709) were computed by the formula as follows: AVE = (sum of square standardized loadings)2 / [(sum of square standardized loading)2 + (sum of measurement error)]. Each value of AVE for each construct was at least 0.5, which is a value proposed by Fornell and Larcker (1981). The discriminant validity among the measures can be held. Finally, the loading (λ) values were
0.768, 0.799, and 0.788 for PK; 0.892, 0.742, and 0.772 for TPP; 0.875, 0.925, and 0.931 for AFF; 0.915, 0.887, and 0.277 for CB; 0.918 and 0.938 for e-WOM; and 0.880, 0.860, 0.883, and 0.894 for NPE. The construct validity of the measures still stood. Table 6 lists the results of the reliability and validity analysis.
Table 6
Results of reliability and validity analysis
Variables Items Cronbach’s α Loading CR AVE Persuasion Knowledge (PK) 0.826 0.782 0.544 PK1 0.768 PK2 0.799 PK3 0.788 Third-person Perception (TPP) 0.848 0.812 0.591 TPP1 0.892 TPP2 0.742 TPP3 0.772 Affect (AFF) 0.935 0.915 0.781 AFF1 0.875 AFF2 0.925 AFF3 0.931 Coping Behavior (CB) 0.725 0.713 0.500 CB1 0.915 CB2 0.887 CB3 0.277 e-WOM (EW) 0.926 0.910 0.834 EW1 0.918 EW2 0.938
Narrative Persuasion Effect (NPE) 0.931 0.907 0.709 NPE1 0.880 NPE2 0.860 NPE3 0.883 NPE4 0.894
4.3.1 The overall model evaluation
The empirical study employed structural equations modeling (SEM)(Joreskog and Sorbom, 1993). For this SEM model, the examination of overall model fit is divided into two kinds of measurement: absolute measurements and incremental measurements. Absolute measurements are used to confirm the degree of predicting covariance or related matrix. The indicators include χ2 / degree of freedom (df), goodness of fit index (GFI), root mean
square of residual (RMSR), root mean square error of approximation (RMSEA), and incremental fit index (IFI). Such fitness criteria were employed to inspect the validity of the model.
Based on Bagozzi and Yi (1988), the proportion of χ2 and df is supposed to
be between 2 and 5. The lower the proportion, the fitter the model (Bianchi and Bivona, 2002). The χ2/df of our model in this study was calculated at 4.323,
which reached the acceptable interval. Some scholars indicate that the value of GFI and AGFI should be above 0.9 (Bentler and Bonett, 1980). Some indicate that each outcome of GFI and AGFI is supposed to be at least 0.8 (MacCallum and Hong, 1997; Doll, Xia, and Torkzadeh, 1994). As for this model, the outcomes of GFI and the AGFI, which were 0.878 and 0.823 separately, did cater to the recommended value. With reference to the viewpoint of Bagozzi and Yi (1988), the advised value seemed too strict to fit the model. The more the items measured, the harder the appropriateness of the model.
The RMSEA and the RMSR indicate the symptom about the fit of the model with unknown but optimally chosen parameter values for the population covariance matrix. The estimation of the RMSEA is acceptable as it is lower than 0.05 (McDonald and Ho, 2002). Additional level of the RMSEA, illustrated by Browne and Cudeck (1993), is below 0.1. As for to the RMSR, Hu and Bentler (1999) indicate that the value of the RMSR should be lower than 0.08. The results of the RMSR and the RMSEA were 0.051 and 0.086, respectively.
As for incremental measurements, Bollen (2014) advises that the CFI is supposed to be referred so as to make the variance more stable. The fit of the
model become more ideal once the value of the CFI is close to one, which indicates that it can effectively optimize the level of centrality. As for the NFI and IFI, which are used to compare the value of χ2 between the proposed model and the independent model. All of them should be higher than 0.9 (Bentler and Bonett, 1980). The consequences about the CFI, the IFI, and the NFI were all exceed the threshold for the acceptable fit of the model of this study.
4.3.2 Results of research hypothesis
Table 7 shows the coefficient and significant relationship and offers a reasonably evidence for the study. Persuasion knowledge (H1: β1= 0.147, t-value
= 0.889), third-person perception (H2: β2= 0.383, t-value = 2.738) and affect (H3:
β3= 0.452, t-value = 8.027) had positive significant effects on coping behavior.
Eventually, coping behavior had positive significant effects on e-WOM (H4: β4=
0.866, t-value = 23.281) and narrative persuasion effect (H5: β5= 0.999, t-value =
28.243).
We have also provided the bootstrapping samples as coefficient of determination (R2), effect size (f2), and collinearity detection (VIF) in Table 7. First, R2 measures variance, which is explained in each of the endogenous constructs and is therefore a measure of the model’s explanatory power (Ramli, Latan, and Solovida, 2019). R2 is also referred to as in-sample predictive power. R2 ranges from 0 to 1, with higher values indicating a greater explanatory power. As a guideline, R2 values of 0.02, 0.13, and 0.26 can be considered small, medium, or large, respectively (Ashrafi, Ravasan, Trkman, and Afshari, 2019). Acceptable R2 values are based on the context, and in some disciplines, an R2 value as low as 0.10 is considered satisfactory. Based on the empirical results in Table 7, the modeled constructs explain a mediate amount of 48.6% (e-WOM) and 71.0% (NPE) variance, followed by CB. The modeled constructs for CB explain a mediate amount of 44.3% (PK), 44.7% (TPP), and 56.7% (AFF) variance. They all lie at satisfactory levels above 0.26.
Second, Cohen (1988) proposed effect size (f2) to assess how the removal of certain predictor constructs affect an endogenous construct’s R2 value. The effect
size for each structural path is estimated by the change in R2 that would occur if the structural path was omitted from the model (Risher and Hair, 2017). Thus, calculated effect sizes were examined for each of the structural paths, and they were somewhat redundant to the size of the path coefficients (Hair, Risher, Sarstedt, and Ringle, 2019). These were calculated as follows: f2 = [(R2 of structural path included) – (R2 of structural path excluded)] / [1/(R2 of structural path included)] (Ashrafi et al., 2019). As a rule of thumb, f2 values higher than 0.02, 0.15, and 0.35 show that the model has small, medium, or large effect sizes, respectively (Cohen, 1988). Based on the empirical results in Table 7, the achieved effect sizes of 0.015, 0.133, and 0.328 (for CB as the dependent variable) represent weak and moderate redundancy and effect size. Therefore, the results show that PK, TPP, and AFF successfully explained the effects that CB has on how well an SF post works.
Third, the variance inflation factor (VIF) was employed to assess the collinearity of the formative indicators. VIF values above 5 present critical collinearity issues among predictor constructs. However, collinearity issues can occur at lower VIF values of 3–5. Ideally, VIF values should be close to 3 and lower (Hair et al., 2019). VIF must be examined to ensure that it does not bias the regression results. If collinearity is a problem, a frequently used option is to create higher-order models that can be supported by theory (Ramli et al., 2019). Based on the empirical results in Table 7, there was no collinearity problem as VIF is smaller than 3, and the CB variable is paired to two dependent variables (e-WOM and NPE).
Table 7
Results of research hypothesis
Hypothesis Path Coefficient T-value R2 VIF f2
H1: Persuasion Knowledge → Coping Behavior β1= 0.147 0.889 0.443 2.9 0.015
H2: Third-person Perception → Coping Behavior β2= 0.383 2.738*** 0.447 2.5 0.133
H3: Affect → Coping Behavior β3= 0.452 8.027*** 0.567 2.4 0.328
H4: Coping Behavior → e-WOM β4= 0.866 23.281*** 0.486 1.0 1.504
H5: Coping Behavior → Narrative persuasion effect β5= 0.999 28.243*** 0.710 1.0 3.878
Notes. χ2/df = 4.323; GFI = 0.878; AGFI = 0.823; RMSR = 0.051; RMSEA = 0.086; CFI = 0.952; IFI = 0.952; NFI = 0.939
Based on one-tailed test: for t-value greater than 1.96 or smaller than -1.96 (*); for t-value greater than 2.33 or smaller than -2.33(**): for t-value greater than 2.58 or smaller than -2.58(***).
4.3.3 Mediating effect analysis
The mediation effect is usually based on the three-stage analysis proposed by Baron and Kenny (1986). Zhao, Lynch and Chen (2010) then amend the Baron and Kenny (1986) approach. In addition, Efron (1979) proposes the bootstrap method which the number of existing samples is amplified by repeated sampling, so that the distribution of their number of times is closer to the way of population distribution. With the computer application of structural equation model, the bootstrap method becomes the latest famous method to explore the mediation effect. Then, the bootstrap method is conducted in a virtual manner, and the confidence interval can be calculated when repeated extractions 1000 times (Efron and Tibshirani, 1993). As for SEM model, when we detect and analyze the mediation effect, we can use the bootstrap method to obtain the confidence interval of the indirect effect, which can be called the mediation effect if the 95% confidence interval does not contain zero and reaches a significant level. If the direct effect in the 95% confidence interval contains 0, it indicates that the direct effect is not significant, with a full mediation effect. If the indirect effect and the direct effect in the 95% confidence interval are not included in 0, and all reach a significant level, then the total effect in the 95% confidence interval does not contain 0, reaching a significant level, and it can be considered as a partial mediation effect (Mackinnon, 2008; Nitzl, Roldán, and Cepeda, 2016; Preacher and Hayes, 2008).
Six types of the serial mediation effect were examined through the bootstrap method from the PK (TPP, AFF) to EWOM (NPE) of the study. They were a mediation effect of CB between PK and EWOM (case 1), a mediation effect of CB between PK and NPE (case 2), a mediation of CB between TPP and EWOM (case 3), a mediation effect of CB between TPP and NPE (case 4), a mediation of CB between AFF and EWOM (case 5), a mediation effect of CB between AFF and NPE (case 6), respectively. The results are listed in Table 8.
We can find that they owned total mediation effects in the case 1, case 2, case 3, and case 5. For example, case 1 showed that the confidence interval
(0.507~0.763) of the indirect effect (0.625; 0.625 = 0.817 x 0.765) did not contain 0 and had a significant effect (p < 0.05), indicating that CB had a mediation effect between PK and NPE. The confidence interval (-0.029~0.288) of the direct effect from PK to NPE (0.127) contained 0, up to an insignificant effect, and the confidence interval (0.676~0.815) of the total effect (0.752), which included the direct effect and indirect effect (0.752 = 0.625 + 0.127) did not contain 0, thus achieving significant results. These outcomes showed that the CB was a total mediation effect between PK and NPE. Similar results can be obtained in case 2, 3, and 5.
Additionally, we can find that partial mediation effects were present in case 2 and case 6. For example, case 2 showed that the confidence interval (0.517~0.749) of the indirect effect (0.631; 0.631 = 0.827 x 0.763) did not contain 0 and had a significant effect (p < 0.05), indicating that CB had a mediation effect between PK and NPE. The confidence interval (0.122~0.390) of the direct effect from PK to NPE (0.250) did not contain 0, up to a significant effect, and the confidence interval (0.824~0.920) of the total effect (0.881), which included the direct effect and the indirect effect (0.881 = 0.631 + 0.250) did not contain 0, thus achieving significant results. These outcomes showed that the coping behavior (CB) is a partial mediation effect between PK and NPE. Similar results can be obtained in case 6.
4.3.4 Moderating Effects
In the one-way ANOVA, p-value presented different genders of the audiences had different e-WOM. Therefore, the gender was employed as a moderator variable to examine the moderating effect. Chi-square difference test is adopted to perform the overall model comparison between groups (Bollen and Long, 1993). Empirical results illustrated that χ2 within these two sub-groups were 347.608 and 460.678 (df = 127; Chi-square difference=113.07), and their degrees of freedom difference smaller than 1. Thus, gender promised to a moderator variable for 113.07>χ!.!",!! .
Table 8
Empirical result of serial mediation effects
Effects Contents Estimate p-value Confidence interval
Case 1: PK à CB à EWOM
Indirect effect PK à CB à EWOM 0.625 p < .05 0.507 ~ 0.763
Direct effect PK à CB CB à EWOM PK à EWOM 0.817 0.765 0.127 p < .05 p < .05 p > .05 (p = .119) 0.747 ~ 0.869 0.607 ~ 0.904 -0.029 ~ 0.288
Total effect PK à EWOM 0.752 p < .05 0.676 – 0.815
Case 2: PK à CB à NPE
Indirect effect PK à CB à NPE 0.631 p < .05 0.517 ~ 0.749
Direct effect PK à CB CB à NPE PK à NPE 0.827 0.763 0.250 p < .05 p < .05 p < .05 0.763 ~ 0.875 0.625 ~ 0.881 0.122 ~ 0.390
Total effect PK à NPE 0.881 p < .05 0.824 ~ 0.920
Case 3: TPP à CB à EWOM
Indirect effect TPP à CB à EWOM 0.595 p < .05 0.453 ~ 0.771
Direct effect TPP à CB CB à EWOM TPP à EWOM 0.855 0.695 0.203 p < .05 p < .05 p > .05 (p = .053) 0.809 ~ 0.898 0.521 ~ 0.885 -0.006 ~ 0.389
Total effect TPP à EWOM 0.797 p < .05 0.730 – 0.849
Case 4: TPP à CB à NPE
Indirect effect TPP à CB à NPE 0.738 p < .05 0.623 ~ 0.871
Direct effect TPP à CB CB à NPE TPP à NPE 0.857 0.861 0.127 p < .05 p < .05 p > .05 (p = .074) 0.811 ~ 0.896 0.725 ~ 0.986 -0.010 ~ 0.261
Total effect TPP à NPE 0.864 p < .05 0.819 ~ 0.902
Case 5: AFF à CB à EWOM
Indirect effect AFF à CB à EWOM 0.893 p < .05 0.721 ~ 1.105
Direct effect AFF à CB
CB à EWOM AFF à EWOM 0.894 0.998 0.010 p < .05 p < .05 p > .05 (p = .065) 0.804 ~ 0.892 0.860 ~ 1.259 -0.446 ~0.026
Total effect AFF à EWOM 0.903 p < .05 0.600 – 0.731
Case 6: AFF à CB à NPE
Indirect effect AFF à CB à NPE 0.590 p < .05 0.447 ~ 0.730
Direct effect AFF à CB
CB à NPE AFF à NPE 0.857 0.688 0.321 p < .05 p < .05 p < .05 0.813 ~ 0.894 0.528 ~ 0.833 0.171 ~ 0.487
Total effect AFF à NPE 0.910 p < .05 0.877 ~ 0.940
Note. PK is persuasive knowledge; TPP is third-person perception; AFF is affection;
significance in both groups. The p-values of other paths from the AFF to CB, from the CB to e-WOM and from the CB to NPE were all significant. As to the path from the TPP to CB, the female group revealed greater significant p-value than the male group. Furthermore, the estimate of the path from the TPP to CB in the male group (0.325) was higher than that of in the female group (0.497). The estimate of the path from the AFF to CB in the male group (0.593) is greater than that of in the female group (0.454).
The main paths of these two groups were different. In the male samples, the primary path was from the AFF to NPE through CB (AFF-CB-NPE). A different path was shown in the female samples. The primary path was from the TPP to NPE through CB (TPP-CB-NPE).
5. Conclusions
The empirical results in our augmented CB model support four of our hypotheses. Namely, that TPP and AFF shaped CB. However, PK provided a non-significant result (H1). This study found that participants responded to an SF post made by a brand through TPP and AFF rather than PK, indicating that the responses of individuals are determined through self-related assessment and self-emotional status rather than an understanding of the SF post. Consistent with the ELM, self-related criteria (e.g., TPP and AFF) trigger the motivation of individuals and further enhance their engagement. This approach led the participants to deal with the information through the central path, thus influencing CB through the Instagram SF. While PK represents comprehension of a message sent by a brand, such knowledge does not appertain to self-related characteristics but to object-oriented understanding and reaches a limited degree of motivation. This message-processing mechanism leads individuals to tackle information through the peripheral path, and this has little impact on CB. Therefore, TPP and AFF lead to different results for CB as compared to PK.
An understanding of the SF post made by brands did not have an effect on subsequent reactions. Chen (2018) suggested that an awareness of a commercial effort had little influence on the responses of Instagram audiences since they
Table 9
Model comparison between groups in t-value
Path Male Group Female Group
Estimate t-value p Estimate t-value p
PK CB 0.059 0.362 0.717 0.048 0.263 0.792
TPP CB 0.325 2.618 0.009** 0.497 3.136 0.002**
AFF CB 0.593 6.274 *** 0.454 6.653 ***
CB EWOM 0.860 15.429 *** 0.863 17.304 ***
CB NPE 0.999 24.526 *** 0.996 18.745 ***
Note. χ2/df= 2.737 (male); 3.627 (female).
were accustomed to the ubiquitousness of promotion-oriented information. The subtle innate character of the messages affected audience receptivity. The level of comprehension of the SF post was not viewed as a significant matter. Kruglanski and VanLange (2012) contended that personal relevance contributes to motivation.
The evaluations by individuals of the influence messages have on themselves and others affect their responses to the SF posts made by brands. Ham and Nelson (2016) agreed that personal rather than social CB is shaped by TPP. This observation has been supported by social penetration theory. The Instagram SF is a platform where users can present private information to those with whom they have reached a certain level of intimacy. In particular, users’ responses to the SF posts shown by brands tend to be based on their judgements of a post’s influence on themselves and others. The sensory awareness of the usage of the SF by brands shapes the audience’s CB. The AIM indicates the involvement of emotion and mood in tackling messages. Entertainment is viewed as one of the main reasons for the use of the Instagram SF. It can be assumed that audiences’ responses to posts in the SF are influenced by feelings. In other words, individuals would prefer to make emotional or affective responses in their appraisals of confrontational situations or events. Decisions are often reached through a combination of external information and internal cognition. In our
study, the audience’s CB urrounding the brands’ usage of the SF shaped the related descriptions that were dispatched on the Instagram website. Each type of CB shows a different type of evaluation of a subject.
As for the academic contributions of this study, the previous relevant literature has explored the causal relationship between PK, TPP, and CB or between e-WOM and CB. AFF had not been included as an antecedent variable of CB, and NPE had not been regarded as a consequence variable of CB. Therefore, a research gap in the existing literature had formed. This provides the main academic contribution of this study. We then proposed three routes for the persuasion procedure―PK, TPP, and AFF―to highlight the drivers of CB in the Instagram SF; that is, the cognitive learning factor, the cognitive processing factor, and the cognitive appraisal factor, respectively. The empirical outcomes of our augmented CB model support most of the hypotheses. TPP and AFF influence CB in the Instagram SF (H2a and H3a). However, PK displays the opposite outcome (H1a). Based on the ELM, high and low engagement with information is influenced by motivation (Petty and Cacioppo, 1986). These results support the ELM in that the peripheral paths (TPP and AFF) obviously tend to influence CB in the Instagram SF; however, the central path (PK) does not influence CB. Thus, the higher the number of respondent reactions, the more likely it is for positive or negative comments to go viral. The audience’s CB influences psychological status regarding the Instagram SF. The following details reveal additional information.
As for practical contributions and managerial implications, the audience’s CB with the SF posts made by brands also has an influence on purchase behavior. When individuals share an SF post, the shared values have a positive effect on trust (Wu et al., 2010), which is related to purchase intention (Hajli, 2014). With an extension of the theory of planned behavior (Pavlou and Fygenson, 2006), purchase behavior has been found to be affected by purchase desire intensity. Using the “see more” option is related to psychological and reminiscence effects and has been explained as an interest in an issue or object (Ashcroft and Hoey, 2001). Liking is related to recall (Mehta and Purvis, 2006). Impressive memory
has a considerable effect on memory retrieval (Hamilton, 2015). CB, which halts or expands the flow of information, is related to the awareness effect. The levels of information flow represent the levels of exposure, which have a positive relationship with awareness (Gibs and Bruich, 2010). In addition, does “how they cope” have a positive influence on “the effects of the SF post” or not? The consequential variables can include NPE and e-WOM. The relationship between the CB regarding the Instagram SF, e-WOM, and NPE is worth exploring in the future.
Further findings of this study illustrate that CB has a significant influence on e-WOM and NPE in the Instagram SF. Thus, the greater the number of respondent responses, the more likely it is for positive or negative comments to go viral. The audience’s CB influences their psychological status concerning to the Instagram SF. As for the impact of CB on e-WOM, an audience’s CB refers to a personal evaluation regarding an issue or message based on self-perception theory (Bem, 1967). Namely, the attitudes of individuals toward an event result from their actions and behaviors, developing a psychological status encompassing satisfaction or dissatisfaction. As noted by Shaikh et al. (2018), the satisfaction of individuals exerts an effect on e-WOM. As for the impact of CB on NPE, the sharing of these SF posts has an effect on purchase behavior. When users share a SF post, the shared values generate a positive effect on trust (Wu et al., 2010), which is linked to purchase intention (Hajli, 2014). Based on an extension of the theory of planned behavior, purchase behavior has been found to be affected by purchase desire intensity. The use of the “see more” option in the Instagram SF has been linked to psychological and reminiscence effects. It has been employed to indicate an interest in an event or object. Preference is related to recall. Impressive memory has a considerable effect on memory retrieval (Hamilton, 2015). CB, which halts or expands the flow of information, has a connection with the awareness effect in the SF. The degree of information flow indicates the degree of exposure, and that enjoys a positive causal relationship with awareness in due course.
