Examining Generational Differences in Perception of
Digital Marketing Strategies on Facebook
Master Thesis
MSc in Business Administration – Digital Business Track University of Amsterdam
By
Michaela Halásová (11374314) Under supervision of dr. Frederik Situmeang
Abstract
Purpose – The purpose of this study is to expand the literature of social media marketing on social networking sites (SNS), by examining the perceptions of marketing strategies on Facebook and the differences in them in dependence on the generational cohorts of the SNS’s users.
Design/methodology/approach – To collect the data and test the developed hypotheses we used a between-subjects true experimental design with three vignettes. Also, the path analysis in AMOS software was conducted.
Findings – The results confirm the positive effect of trust on willingness to pay. Moreover, trust toward advertisement is lower with higher age. However, the indirect effect of age was not significant and therefore, not confirmed due to the lack of evidence from data. Except of that, the results showed negative relationship between attitude toward advertisement and willingness to pay, which is contradictory to our hypothesis.
Research limitations/implications – The research was conducted on sample in which 63.43% consists of young people at the age of 19 to 27. The bigger variance in age could provide a better picture of the differences of perception of social media advertisement. Moreover, only Nike brand was investigated. The future research could consider to include more generational cohorts and brands, compare the effects among other popular social networking sites (e.g. Instagram, Snapchat) applying the longitudinal research design.
Practical implications – Managers and companies can employ the findings of the study to improve their marketing strategies on social media in order to increase the users’ willingness to pay, especially among younger users from the Millennial generation.
Originality/value – This study is unique and valuable for literature in the field of marketing on social media because it examines three types of social media advertisement, includes around 50% of non-student responses, which makes the sample more representative, and uses the experimental design leading to causal assertions.
Key words: social networking sites (SNS), generations, digital marketing, electronic word of mouth (e-wom), willingness to pay
Statement of Originality
This document is written by Student Michaela Halásová who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Acknowledgement
First of all, I am grateful especially to the supervisor of my thesis, dr. Frederik Situmeang, for his time he dedicated to supervising the thesis, his valuable comments, which helped to find and mitigate shortcomings of this thesis, as well as his willingness to supervise it. I am really glad that he cooperated with me throughout the whole process of writing, from choosing the topic and writing the proposal until the completion of the thesis.
Moreover, I would like to thank my parents for their huge support during my studies and that they made it possible for me to study at University of Amsterdam. Without their help, it would be very difficult to come to this point of college graduation.
Last but not least, I am thankful for supportive nature of my best friend and soulmate, Martin, who encouraged me every time I needed it.
Table of Contents
1. Introduction ... 8 2. Literature Review ... 11 2.1. Social Media ... 11 2.2. Digital Marketing Strategies on Social Media ... 13 Moderator: Age ... 13 Company-Generated Content ... 16 Company-Generated Content shared by another user (e-wom) ... 17 User-Generated Content (consumer-generated advertising – subtype of UGC) ... 19 2.3. Conceptual Model ... 22 3. Methodology ... 23 3.1. Experimental Vignette Methodology (EVM) ... 23 Planning an EVM study ... 24 Implementing an EVM ... 26 Reporting Results of an EVM Study ... 28 3.2 Measurement of Variables ... 29 3.3. Statistical Procedure ... 32 4. Results ... 34 4.1. Descriptive Statistics ... 34 4.2. Correlation Analysis ... 36 4.3. Regression Analysis ... 39 Direct effects of marketing on SNS on willingness to pay ... 39 Indirect effects of marketing on SNS on willingness to pay ... 40 Effects of Covariates ... 445. Discussion and Conclusion ... 46 5.1. Discussion of Results – Key Findings ... 46 5.2. Research Implications ... 47 5.3. Managerial Implications ... 48 5.2. Limitations and Further Research Implications ... 49 References ... 51 Appendix 1: Questionnaire ... 57 Company-Generated Content (English and Slovak version) ... 57 Company-generated content shared by another user (English and Slovak version) ... 58 User-Generated Advertisement (English and Slovak version) ... 59 Appendix 2: Overview of Hypotheses ... 60
List of Tables and Figures
Tables
Table 1 Categorization of social media based on social presence/media richness and self-presentation/self-disclosure
12
Table 2 Generations defined by birth cohort 14
Table 3 Descriptive Statistics - Demography 35
Table 4 Means, Standard Deviations, Correlations, Reliabilities 37
Table 5 Exploratory Factor Analysis 38
Table 6 Direct Effects of Marketing on SNS 40
Table 7 Indirect Effects of Marketing on SNS 44
Table 8 Control Variables 45
Figures
Figure 1 Conceptual Model 22
Figure 2 Steps and decision points how to design, implement, and analyze an experimental vignette study
1. Introduction
Despite several articles were published about advertising on social media before 2010, the interest of digital marketing researchers in this topic started mostly in 2010 (Knoll, 2016). Since then, the topic remains to be studied as the social media networks’ popularity is increasing. While there are many popular social media, the one with 1.86 billion active users worldwide1 (Statista, n.d.) and continuous growth in this number is Facebook, which makes it the most used social network in world. 88% of 18 to 29 old Americans online use Facebook (Greenwood, Perrin & Duggan, 2016), which makes it an important tool for marketers, especially for increasing brand awareness, positioning advertisements and attracting/retaining customers. Moreover, mostly young people use these social media and despite constant research, a little is known about their behaviour on these sites.
Lately, researchers have studied the attitude toward digital marketing among younger users, i.e. college students, Millennials, and teenagers (Sashittal, Sriramachandramurthy, and Hodis, 2012; Tanyel, Stuart, and Griffin, 2013; Akpojivi, Bevan-dye, 2015). Last year, a book by Ryan (2016) about engaging the digital generation was published. These recent literature shows that the connections of generations/age with digital marketing, including social media advertising, is still an actual topic and worth studying.
Researchers have also tried to examine the attitudes toward the advertisement on social networking sites (SNS) in general (e.g. Kelly, Kerr, and Drennan, 2010), the impact of motivation to join SNS on responses to various advertising on the websites (Chi, 2011), or the factors leading to acceptance of marketing on these sites Taylor, Lewin, and Strutton (2011).
However, because this field is relatively new it provides a rich area for further research. Only limited number of studies has been performed about the differences in perceptions toward
advertisement on social networking sites (SNS) across different generations. The need to fill in this gap is emphasized by above-mentioned literature and also by Rosenberg (2008) stating that in the marketing field, it is important to know the audience, including the differences and similarities among various generations.
Furthermore, Knoll (2016) reviewed 51 empirical studies from the field of social media advertising, summarizing the neglected topics, stressing the focusing on “the intended influences of advertising on recipients’ cognitions, affects, attitudes, and behaviours” as one of them.
In conclusion, the goal of this study is to answer the following research question: “How does users from various generational cohorts differ in a way how they perceive different
marketing channels on SNS?”
As written above, the research examining the topic of advertisement on social media has already been done in a larger scale. Nevertheless, this study contributes to expanding the literature in this field and reduce the research gap as mentioned in the review of empirical evidence by Knoll (2016) for following reasons:
1. It focuses on users’ attitude and overall perceptions of advertising on social media 2. It examines the advertisings’ effects using specific instances not only in one but in three
types of advertisement, particularly company-generated content on brand pages, electronic word of mouth, and generated advertisement as a subcategory of user-generated content
3. The study does not rely on student sample only but involves also full-time employees, who form approximately 50% of the whole sample. As a result, the sample is more representative of the population of social media2 users
4. As a research method, a between-subjects true experimental design is used, from which the causal assertions can be stated.
In order to provide the answer to the research question in understandable and comprehensible matter, the study is structured as follows. The next chapter provides an overview of literature in the field of advertisement on SNS as well as recent research concerning the role of generations/age in marketing. Chapter three describes the experimental survey design method adopted in this study as a way of data collection. Subsequently, the results from the data analysis are presented in chapter four. The most fundamental parts, the conclusions, managerial and academic implications of the results, are discussed in chapter five. Finally, in chapter six, the most important limitations and possible further research suggestions are demonstrated.
2. Literature Review
2.1. Social Media
Social media are defined by Kaplan and Haenlein (2010, p. 61) as “a group of internet-based applications that builds on the ideological and technological foundations of Web 2.0, and it allows the creation and exchange of user-generated content.” In order to provide organized overview of social media, we use the classification by the traditional theories, particularly social presence/media richness and self-presentation/self-disclosure as demonstrated by Kaplan & Haenlein (2010).
Social presence theory distinguishes social media by the degree of physical, acoustic and visual which can be achieved between two people communicating through these media. Associated with this theory is the media richness theory. It claims that the primary goal of any communication media is to reduce ambiguity (Daft & Lengel, 1986). According to this theory the social media vary in the amount of information which can be sent in a given time period. Regarding the other dimension, self-presentation, the concept is characterized as a desire to manipulate what other people think about them (Goffman, 1956). Self-presentation in social media is often done by self-disclosure, therefore these two concepts are measured together. Self-disclosure includes creation of personal website and sharing private information about themselves (Kaplan & Haenlein, 2010). Cheung et al. (2015) studied the effect of cost, benefits, and social influence in the social networking sites (SNS). They found the highest significant effect on self-disclosure on SNS is driven by the need to comply with expectations of the fellow members. As a result, people tend to be influenced by the information on SNS from those who share similar interests (Cheung et al., 2015) and they are also more likely to engage in commercial activities on SNS (Ng, 2014).
Another term used to express social media and social-networking sites is user-generated content sites because of its peer production (Gallaugher, 2015), i.e. users’ collaborative creation of the sites’ content. These includes blogs, social networking sites (SNS), collaborative projects, content communities, as well as virtual social words and games. The overview of various social media based on the abovementioned theories and concepts is summarized in the Table 1 below.
Table 1. Categorization of social media based on social presence/media richness and self-presentation/self-disclosure
Social presence/ Media richness
Low Medium High
Self-presentation/ Self-disclosure High Blogs Social networking sites (e.g. Facebook)
Virtual social worlds (e.g. Second Life)
Low Collaborative project (e.g. Wikipedia) Content communities (e.g. YouTube)
Virtual game worlds (e.g. World of
Warcraft)
Source: Kaplan & Haenlein (2010)
Moreover, social media has become a new fast and cheap way of communication not only with friends but also with companies. Strutton et al. (2011) confirmed that people are connected to other people in majority of their discretionary time. This offers new opportunities how to engage the current or perspective customers, which can lead to increase in trust and consequently intention to buy (Hajli, 2014). Social media have a positive effect on brand relationship and can serve as a tool to build and improve customer relationship. (Hudson, 2015). The results of Hudson (2015) also shown that the emotionally attached customers are also more willing to recommend the product online. The positive attachment and perception of the brand on SNS has an impact on purchase intentions (Schivinski and Dabrowski, 2016;
traditional marketing and found that social media has bigger effect on brand attitude and therefore, also on purchase intentions.
2.2. Digital Marketing Strategies on Social Media
This study focuses on the perception of SNS users from various age cohorts toward three types of marketing. Since the role of age as a moderator is the most fundamental part of our research, the definitions and relevant literature demonstrating the differences between these cohorts is discussed first, followed by the literature review of particular marketing strategies with respect to the age.
Moderator: Age
Our reasoning behind the assumption of birth cohorts as a valid moderator, having an effect on the perception of marketing strategies on SNS and willingness to pay is supported by generational cohort theory (Strauss & Howe, 1991). This theory proposes sorting together people from the similar birth cohorts who went through the same life experiences, and the same events, e.g. technological advancements or global economic situation including depressions and recessions (Carpenter et al., 2012). Consequently, similar background suggests that these people share similar values, beliefs, and attitudes, differing from other generational groups. In brief, Strauss and Howe (1991) define a generation as “a cohort-group whose length approximates the span of a phase of life and whose boundaries are fixed by peer personality” (p. 60). Additionally, they specified the peer personality as “a generational persona recognized and determined by (1) common age location; (2) common beliefs and behavior; and (3) perceived membership in a common generation” (p. 64).
Moreover, in this classical work, they introduced five generational cohorts living in 1991 in the United Stated (Keeling, 2003) depicted in Table 2. Even though they based it on the US, the definitions are used worldwide.
Table 2. Generations defined by birth cohort
Name of the Generation Birth Cohort Other Names
GI Generation 1901 - 1924
Silent Generation 1925 - 1942
Boom Generation 1943 - 1960 Baby Boomers
Generation X 1961 - 1981 13th generation
Generation Y 1982 - 2003 Millennials, Generation Next
Source: Strauss and Howe (1991); Keeling (2003); Carpenter et al. (2012)
This study focuses mostly on the two of them who are the most common Facebook users. The first one is Generation X, defined by years 1965 to 1980, from which approximately 84 % of internet users of that age in the United States use Facebook (Greenwood et al., 2016). The next generation is Generation Y, also called Millennials, from which approximately 88% of all millennial Internet users in the United States are Facebook users. This group is defined by birth years 1981 to 1997, who are mostly students finishing their universities and entering the working workforce these years or young employees. As a result, their buying power is very likely to increase significantly as they manage to find their full-time jobs and start families. Moreover, the number of millennials in the United States has overcome the size of the biggest generation so far, the Baby Boomers, which is generation born in years 1945 to 1964, by 15 million (US Census Bureau, 2015).
Regarding their way of living and thinking, there are many differences when compared to the previous generations. They are considered to be tech-savvy, always online, and the most educated ever (Smith, 2012). Millennials are also considered as the first generation which is
completely “digital native”, whereas the previous generations are “digital immigrants” (Kirk et al., 2015). Furthermore, Carpenter et al. (2012) showed that the younger people have positive attitude toward globalizations, which suggest that marketers in does not need to adapt the advertisement to national countries but can remain global.
All these changes challenge the ways how traditional marketing has been done so far. However, if performed correctly, these challenges could provide enormous opportunities. In order to attract the Generation Y effectively, the marketing strategies, also on the SNS such as Facebook, should be adapted.
While the characteristics of Millennials have been studied extensively in their role in workforce (Goldman Sachs, n. d.; Meyers & Sadaghiani, 2010), less was done from marketing point of view. Duffett (2015) examined the impact of advertisements on Facebook on Millennials in Cape Town, South Africa, on their intention-to-purchase and subsequent purchase. The data were analyzed using generalized linear model and positive effect of the Facebook’s advertisement was confirmed.
Eastman and Liu (2012) compared the level of status consumption among generations of Baby Boomers, Generation X, and Generation Y in the United States. The study shown that Generation Y has the highest level of consumption, whereas Baby Boomers demonstrated the lowest level. Controlling for variables such as gender, income, and education they tested that the results are due to generational differences, not demography. Therefore, they suggest segmenting the consumers by generational cohorts, which also supports the need of answering our research question not only for scholars but also for practitioners.
One of a few papers published about Generation Y (Millennials) and their attitude toward advertisement on the Internet is Smith (2012). In this longitudinal study, it is concluded that Millennials respond positively to online coupons, and personalized websites, but hate pop-up ads. This generational cohort find the advertisement more attractive if they can see the value
created by the ad (Rosenberg, 2008). Similarly, in the mobile setting, only around one third (37.2%) of college students representing the population of Generation Y in Apkojivi and Bevan-Dye (2015) found the advertisement useful and value-adding. Furthermore, in the past, Shavitt, Lowney, and Haefner (1998) showed that younger and less educated people tend to have more positive attitude toward advertisement than older consumers.
Company-Generated Content
The attitude toward advertising on social networking sites by consumers has been in a spotlight of various researchers. According to Knoll (2016), the first study was done by Kelly, Kerr, and Drennan (2010) who focused on teenagers’ attitude toward the advertising on the SNS and whether they avoid advertisements and what is their motivation behind this avoidance. The findings show their acceptance of the advertisement on these sites unless they do not have to pay any fee for the SNS usage. One of the reasons why they tried to avoid the advertisement was that despite the companies’ strategies to target the audience based on their personal characteristics and preferences, the advertisement was not relevant for them and seemed not to be trustworthy. Similar findings were shown in Sashittal, Sriramachandramurthy, and Hodis (2012), who found out that college students consider SNS advertisement absenting credibility. Moreover, their motivations are distinct from marketing purposes (Sashittal, Sriramachandramurthy, and Hodis, 2012).
Another study emphasizing credibility as one of the important factors influencing the attitude toward the advertisement is Boateng and Okoe (2015). In their research, the authors found that credibility has a positive influence on the way how people perceive the advertisement on SNS. Moreover, they showed that reputation of the company has a significant influence on the attitude toward social media advertising.
The background of the brand is especially important to younger generations. In comparison with the older generations, Millennials coming to age are facing a more brand-conscious life and therefore, they tend to think more about the background of the company when purchasing products (Smith, 2012). In other words, younger people tend to think more about the credibility of the advertisement and whether the company is trustworthy.
Therefore, the following hypotheses are proposed:
H1
There is a negative relationship between company-generated content and willingness to pay, such that the relationship is weaker when the content is generated by the user.
H2a
There is a positive relationship between company-generated content shared by another user (e-wom) and trust, such that there is a higher level of trust when the content is generated by the user compared to when it is generated by the company, or when it is generated by the user.
H2b There is a positive relationship between trust and willingness to pay.
H2
The positive relationship between company-generated content shared by another user (e-wom) and willingness to pay is mediated by trust.
H3
The positive relationship between company-generated content shared by another user (e-wom) and trust is moderated by generational cohort, such that this relationship is weaker for higher values of age.
Company-Generated Content shared by another user (e-wom)
Customers can evaluate the products and share their opinion with other people either traditionally (offline), face to face or via phone, or less traditionally, via Internet. Hennig-Thuray et al. (2004, p. 39) define e-wom as “any positive or negative statement made by potential, actual, or former customers about a product or company, which is made available to a multitude of people and institutions via the Internet.”
Word of mouth, including the electronic word of mouth is considered more trustworthy and credible than advertising by the company, especially by Millennials, who react to marketing in another way than their parents (Smith, 2012). However, the research suggests that younger generation can more easily detect credible and not credible cues in the reviews (Liao and Fu, 2014). Moreover, consistency of user reviews with the company message seems to have a reinforcing effect and can help to make a right judgment of the product. On the other hand, when the reviews were inconsistent, older people were less likely to change their opinion on low credibility content than the younger users, who trusted the reviews more (Liao and Fu, 2014). The results are consistent with Smith (2012), who found that reviews from other customers are considered to be more credible than other, more traditional ways of marketing. Perception of e-wom credibility by Millennials is also strengthened or weakened depending on the level of source trustworthiness (Shamhuyenhanzva et al., 2016).
The e-wom from friends have an impact not only on perception of credibility but also on purchase considerations. Barreto (2013) compared e-wom from friends with advertisement on Facebook and found lower levels of intentions-to-purchase in the case of advertisement. See-To et al. (2014) confirms the effect of e-wom on purchase intentions. However, they do not measure credibility but trust and value con-creation.
Another study focusing on the differences among generations from marketing point of view is Strutton, Taylor, and Thompson (2011). The authors examined usage of social media and willingness to spread electronic word of mouth (e-wom). They confirmed the higher usage among younger generation, Generation Y, but not higher propensity to share the advertisement on all social media platforms. The only exception was Facebook where it was shown that generational cohorts Y are more likely to pass along e-wom (Srutton, Taylor, and Thompson, 2011).
Therefore, the following hypotheses are proposed:
H4
There is a negative relationship between company-generated content and willingness to pay, such that the relationship is weaker when it is company-generated content shared by another user.
H5a
There is a positive relationship between company-generated content shared by another user (e-wom) and credibility, such that there is a higher level of trust when the content is generated by the user compared to when it is generated by the company, or when it is generated by the user.
H5b There is a positive relationship between credibility and willingness to pay.
H5
The positive relationship between company-generated content shared by another user (e-wom) and willingness to pay is mediated by credibility.
H6
The positive relationship between company-generated content shared by another user (e-wom) and credibility is moderated by generational cohort, such that this relationship is weaker for higher values of age.
User-Generated Content (consumer-generated advertising – subtype of UGC)
Another form of advertisement or marketing on Facebook and other SNS is user-generated content (UGC). UGC can be seen as different forms of media published on a publicly accessible webpage to certain group of people, requiring some level of creativity, and produced mainly by non-professionals3 (OECD, 2007).
“Although UGC has been closely aligned and often confused with eWOM, the two differ depending on whether the content is generated by users or the content is conveyed by users. For example, footage on YouTube that is generated and posted by users is UGC.
However, an Internet user who sends her friends a link to a YouTube site is engaging in eWOM.” (Cheong and Morrison 2008, 3).
More specifically, we focus on consumer-generated advertising (CGA) also called user-generated advertising, which is a sub-category of user-user-generated content. In general, the difference between UGC and CGA is that in CGA users imitates popular or unpopular advertisements using advertising tools to create videos with the goal of CGA is to “inform, persuade, or entertain other users about brands or products” (Campbell et al., 2011).
Chi (2011) studied the impact of the users’ motivation to use social networking sites on their engagement and responses to advertisement on these websites. The author compared virtual brand communities with paid advertisement on Facebook and found out that the post created in virtual brand communities scored higher on all measures: trust, global, attitude, and participation intentions.
Similar positive findings show Lawrence, Fournier, and Brunel (2013) who compared CGA with company-generated advertisement on YouTube, one of the social networking video sites. In their research, CGA scored higher on credibility, trustworthiness, as well as cognitive and emotional engagement. On contrary, CGA can be also viewed as amateur and of lower quality because of scepticism effect suggesting that non-professionals are less capable of high-quality ad production (Pehlivan et al., 2011).
Electronic word of mouth (eWOM) and user generated content (UGC) are two separate marketing strategies, and therefore, we also distinguish them in our research.
H7
There is a negative relationship between company-generated content shared and willingness to pay, such that the relationship is weaker when the content is generated by the user.
H8a
There is a positive relationship between company-generated content shared by another user (e-wom) and attitude toward social media advertising, such that there is a higher level of attitude toward social media advertising when the content is generated by the user compared to when it is generated by the company, or when it is generated by the user.
H8b
There is a positive relationship between attitude toward social media advertising and willingness to pay.
H8
The positive relationship between company-generated content shared by another user (e-wom) and willingness to pay is mediated by attitude toward social media advertising.
H9
The positive relationship between company-generated content shared by another user (e-wom) and attitude toward social media advertising is moderated by generational cohort, such that this relationship is weaker for higher values of age.
2.3. Conceptual Model
In conclusion, based on the recent literature of marketing on social networking sites and differences in their perception by user of various age differences, we formulated the hypotheses as written below the conceptual model (Figure 1).
3. Methodology
This quantitative study is conducted using a between-subjects true experimental design. Moreover, it is a vignette survey design, a common method used not only in digital marketing. Three different versions of Facebook posts depicting Nike shoes are randomly distributed among respondents.
In order to test the developed hypotheses, deductive approach will be used. The survey was translated to Slovak and then distributed via email and Facebook in both, Slovak and English. Pre-test and pilot testing of the questionnaire was done on several individuals representing the population in order to test reliability, relevance, and clarity of the questions and well as overall structure of the questionnaire.
3.1. Experimental Vignette Methodology (EVM)
Even though the definitions of a vignette vary, we use the recent one by Atzmüller and Steiner (2010)4 who define it as “a short, carefully constructed description of a person, object, or situation, representing a systematic combination of characteristics” (p.128).
In designing and implementing the vignette experiment within the survey, we followed and reviewed the best practices of Aguinis and Bradley (2014), as summarized in their article in the steps and discussion points. The authors developed them by analyzing 328 articles from different industries, who used EVM and were published in the period of 1994 to 2013.
The steps and decision points served us as a path which we pursued, are described below. The graphical summary is provided in Figure 2 at the end of this subchapter.
Planning an EVM study
1. Decision Point: Is EVM a Suitable Approach?
The goal of this study is to test the hypotheses composed of three independent variables, one dependent variable, three mediators, and two moderators. We believe, EVM is an appropriate methodology to use, because it helps us with testing the causal relationship by controlling for factors which might have an effect on the independent variables and which will not be possible with correlational study. These include for instance different brands, settings, and design of the marketing strategies on Facebook. By carefully crafting the vignettes, we are able to mitigate or even eliminate these factors.
2. Decision Point: Which Type of EVM is suitable?
In their research, Aguinis & Bradley (2014) distinguish between two types of EVM.
Firstly, Paper people studies which “consist of presenting participants with vignettes typically in written form (and hence their name) and then asking participants to make explicit decisions, judgments, and choices or express behavioral preferences” (Aguinis & Bradley, 2014, p.354).
Secondly, there are policy capturing and conjoint analysis studies, in which the variables are carefully manipulated and who show the implicit processes and outcomes. It is an experimental method “for handling situations in which a decision maker has to deal with options that simultaneously vary across two or more attributes” (Green, Krieger, and Wind 2001, p. S57) In these types of studies, the researchers are especially interested in the “ordering of a dependent variable” (Green & Aro, 1971, p. 355).
In conclusion, the paper people studies in online form will be used since this research aims to measure respondents’ perception of the marketing strategies on Facebook rather than asking them to rank the vignette to any order.
3. Decision Point: Choosing the Type of Research Design
In this decision point, it is important to choose the type of experimental design. In general, the vignette studies can be divided into three categories, particularly, within-person, between-person, and mixed research design (Atzmüller and Steiner, 2010). Despite the recommendation of Aguinis & Bradley (2014) to use within-person design, in which the respondents view the equal number of vignettes and the comparison is made within the person’s response (Atzmüller and Steiner, 2010), we decided to use between-person design, in which each participant is allocated only one vignette (Atzmüller and Steiner, 2010).
When all three vignettes would be presented to each respondent, the answers could be biased. If they knew we are studying the comparison between company-generated content, electronic word of mouth, and user-generated advertisement, they might prefer user-generated content not because they are persuaded it is more suitable for them but rather because they might expect worse attitude toward advertisement and company-generated content in general. Another reason why we think that the between-person is more appropriate in this case, is the shorter length of the survey. Shorter survey is also more likely to be finished, than the longer survey and have also lower rate of careless responses (Meade & Craig, 2011).
4. Decision Point: The Right Level of Immersion
One of the main disadvantage of the EVM is that the results are often specific to those in vignettes but are not applicable in broader reality (Hughes & Huby, 2002).
Moreover, self-reported surveys are often criticized for being vague and unreliable. In order to mitigate these limitations, Alexander and Becker (1978) suggest to create realistic scenarios. Therefore, we used the real text from official Nike e-shop. Facebook posts were created including this text and the picture from the official Nike Facebook webpage. In the version of the UGC, the picture of the same shoes in the similar setting was taken, in order to
be able to make compare the results of the vignettes. The screenshots of the Facebook posts are attached in Appendix 1.
Nike brand was chosen in the survey for several reasons. Firstly, it is a well-known brand among wide range of population, including both generational cohorts. Secondly, healthy lifestyle seems to be popular nowadays, and therefore, we assumed that since their attitude toward the product, the running/training shoes, is positive, they will be more likely to finish the survey.
5. Decision Point: Specifying the Number and Levels of Manipulated Factors
The relevant factors were derived using the “actual derived cases” approach as recommended in Aguinis and Bradley (2014) and are summarized in the conceptual model.
6. Decision Point: Choosing the Number of Vignettes
As already explained in the third decision point, in this research, we are using the between-person research design, in which only one vignette is shown to every respondent.
Implementing an EVM
7. Decision Point: Specifying the Sample and Number of Participants
The population of this research are Facebook users. In 2016, around 30% of the whole population used a social network at least once a month (eMarketer, 2016), which makes it an important tool for marketing purposes for researchers as well as companies. Currently, the biggest social network is Facebook which was the first one surpassing the amount of 1 billion active monthly users in 2012. As of the end of of 2016 there were 1.86 billion monthly active users registered (Statista, n. d.). Regarding demographic distribution, in 2012 the largest part (29.7%) was comprised from people age 25 to 34 (Zephoria, 2017), members of Generation Y.
frame, the most suitable method is non-probability convenience sampling. Since the focus in this study is on Generation Y and also the previous Generation X, this might be the most appropriate method to achieve sufficient number of respondents in both generations. The researcher will try to reach and ask as many people as possible to fill in the questionnaire, with a goal of 200 Facebook users. For this reason, along with administration by e-mails and Facebook, snowballing and self-selection method will be used. Regarding the response rate, the researcher will aim to reach Facebook users by posting and therefore, this rate is very difficult to estimate.
8. Decision Point: Choosing the Setting and Timing for Administration
The survey was administered via anonymous link, which was posted on Facebook in numerous Facebook groups and personal profiles. The distribution started on 5th April, 2017 and was finished after one month, 5th May, 2017. After two weeks, 19th April and 23th April, the survey link was administered again in the same groups as a reminder. No monetary award was offered, but non-monetary one in the form of a short summary of result from the thesis was promised, if the respondent decided to leave his or her email at the end of the survey.
9. Decision Point: Choosing the Best Method for Analyzing the Data
In the data analysis, Aguinis and Bradley (2014) recommends the ANOVA technique to be used when doing between-subject comparison. In addition, regression analysis might be performed. Because the regression analysis provides more complex results, we decided to conduct the regressions rather than analysis of variance (ANOVA).
Reporting Results of an EVM Study
10. Decision Point: Level of Transparency of the Results and Methodology
In order to reveal as many details about the methodology as possible we provide the description of these ten decision points we followed as well as the pictures of vignettes, three Facebook posts (Appendix 1).
Figure 2. Steps and decision points how to design, implement, and analyze an experimental vignette study
3.2 Measurement of Variables
Translation, back-translation procedure
The questionnaire will be translated to Slovak language in order to make the survey more comprehensible and for the Facebook users in Slovakia, especially for people from Generation X. The procedure of back-translation will be used for this purpose.
Marketing on SNS
The study focuses on three different marketing channels on Facebook, specifically company-generated content, company-generated content shared by another user (me), and customer-generated content. These are the independent dummy variables with value 1 if the particular Facebook post was shown to the person in the questionnaire and value 0 otherwise. Each respondent can view only one of the three posts in the survey. The following questions related to the screenshot of the post.
Willingness to Pay
The dependent variable of the study is respondents’ willingness to pay for the Nike shoes depicted in the survey based on the shown screenshot of one of the Facebook posts. The question of “How much are you willing to pay for these shoes?” is required to answered and only numerical values are accepted. There is no restriction on the value, except of the requirement to insert the amount in Euro (€) currency. If the person is not willing to pay anything, he or she can insert 0, which is however not explicitly written in the questionnaire in order to avoid anchoring bias.
In a situation when people are asked to assess the value of an item, an anchoring bias might occur. Wilson (2012) states that “with anchoring bias, a participant's assessment of the
value of an item is influenced by seemingly irrelevant pieces of information5” (p. 65). Mentioning the zero value as an option might serve as one of these “pieces of information” affecting the respondent’s willingness to pay. Moreover, according to this theory, the first proposed value has a significant effect on the final decision and therefore, we refrained from providing more information how to answer the question, even though, it might have helped to improve its clarity.
Mediator: Trust
The first measurement of user responses implemented in this study is trust. The five-point Likert scale consisting of four sets of items was adopted from Chi (2011) validated by Soh, Reid, and King's (2009) ADTRUST scale. These sets measure how participants perceive affect, usefulness, reliability of marketing on social media platform called Facebook. Moreover, it also measured willingness to rely on the marketing on Facebook. Cronbach’s alpha achieved .94 (Chi, 2011) which is considered as high internal consistency of the items.
Mediator: Credibility
In various literature measuring trust in the advertisement in digital marketing, credibility is the main item. Therefore, the credibility is determined as a separate construct. All three items were measured using the adopted scale from Boateng and Okoe (2015), who slightly adapted the credibility scale as written in the validation paper of Chowdhury et al. (2006). The adopted version’s Cronbach’s alpha equaled .7 which is acceptable (Goforth, 2015).
Mediator: Attitude Toward Advertising
The second construct in this study is attitude toward social media advertising. To measure this construct, the four-item scale was adopted from originally validated in Taylor and Todd (1995). All items were measured with a five-point Likert scale (1 = strongly disagree, 5
= strongly agree). Cronbach’s alpha reached .84 demonstrating relatively high consistency of the items.
Moderator: Age
Moderating variable age is measured by self-reporting in the demography section at the end of the survey. By answering the question “What is your age?” people should report their current age in a numerical value. Since this question may be sensitive to some respondents, in order to decrease the dropping out rate, inserting “00” into the field is allowed. The anchoring bias is not present because there is no uncertainty in assessing his or her age. However, there might be the risk of careless response and tendency to speed up the filling of the questionnaire. To avoid of losing these responses, the categorical question with age ranges followed. In that case, the respondent is moved to the next question where the age is divided into 5 categories based on the generations’ definitions. The answer also included the option “Sorry, I don’t want to
answer this question”. Unfortunately, people who tick the answer cannot be included in our
analysis but were allowed to finish the whole questionnaire.
Asking to insert the age into the range of numbers can decrease people’s reluctance to answer it but gives us less accurate inputs for data analysis. Therefore, numerical as well as categorical questions are included.
Control variables
Control variables included Facebook usage, gender, nationality, ethnicity, work status, and education and all except the Facebook usage were asked at the end of the questionnaire.
Because some respondents may have more than one nationality, the question was in the format of the open questions where they could write all their nationalities, not only one of them. In the pilot testing, also the word ethnicity as a substitute for “nationality” was tested, but the
results showed it was not clear to many respondents, what is meant by “ethnicity”. Therefore, the original word “nationality” remained in the question.
In addition to the control variables above, after some feedback received from the respondents, one question about “knowing me” personally was added. The main purpose of this, is to control for potential bias it may cause, especially, to the results from versions about CGC shared and UGC.
That may occur because the text in these posts represents the personal opinion on the Nike shoes, not the official statement of the company. O’Reilly et al. (2016) tested the criteria on which people evaluate the trustworthiness of the product review (e-wom). They found that the willingness to rely on the e-wom is based on several determinants. These include source expertise, source trustworthiness, persona similarity, and usage similarity, which are being reviewed at that order (O’Reilly et al., 2016).
Because the focus of this study is on marketing on social media, not especially on testing all of these characteristics, and also in order not to discourage the respondents with a long survey, only the question “Do you know me (Michaela Halasova) personally?” was added.6
3.3. Statistical Procedure
In order to conduct the statistical data analysis, we used Statistical Software Package for Social Sciences (SPSS) (version 24) and Analysis for Moment Structures (AMOS).
The questionnaires were divided into complete and incomplete and scrutinised further in detail if they should be included in the final analysis. From the completed answers, we eliminated the careless responses detected by wrong answers to the question with bogus item
inserted in the second half of the questionnaire. No contra-indicative items were included in the questionnaire and therefore, no recoding of the variables was needed. Incomplete answers were deleted from the collected data and analysed separately in order to identify the reasons why the respondents dropped the survey.
In the preliminary analysis, we coded and captured the collected data, tested the reliability of the Likert scales with Cronbach alpha as the indicator of scale’s reliability. Moreover, to further validate the scales we performed factorial analysis. Descriptive statistics including means, frequencies, and standard deviations were also calculated (Table 3).
Pearson’s correlation coefficient analysis was conducted to measure the strength of the relationship among constructs (Table 4). The path way analysis with maximum likelihood estimation in AMOS was conducted on the structural model to test the hypotheses.
4. Results
4.1. Descriptive Statistics
A total number of 523 responses was recorded in one-month-long data collection period, out of which 268 passed the screening for completeness and accuracy. This equal 50.57% of all collected questionnaires. 54 responses were discarded because of inaccurate answer in the question with bogus item. 201 responses were unfinished and deleted from further analysis. However, to determine if there are any differences between samples of completes and not completed surveys, basic statistical analysis was conducted.
From this analysis, we found out that majority of them (75%) stopped the survey in the question when the vignette started, the one were the screenshot of Nike shoes was presented to respondent.
Regarding the respondents who finished the survey, majority of them were females (66.42%), were pursuing/had a bachelor (32.09%) or master (42.91%) degree and belonged to the age category from 19 to 27 years (63.43%). Moreover, 69.40% chose to fill in the questionnaire in English, compared to 30.60% who preferred to have the survey translated to Slovak language.
Knoll (2016) encourages use of non-student samples because they do not represent the whole population of social media users and therefore, we tried to focus on as many groups as possible. Finally, we managed to collect almost equal number of students (42.16%) and full-time employees (35.82%). The full description of demography is summarized in Table 3.
Table 3. Descriptive Statistics - Demography n % Age 0-18 8 2.99% 19-27 170 63.43% 28-36 53 19.78% 37-52 37 13.81% 53-71 3 1.12% 72+ 1 0.37% Language English 186 69.40% Slovak 82 30.60% Nationality Slovak 105 39.18% Dutch 30 11.19% Czech 19 7.09% Other 106 39.55%
Ethnicity American Indian or Alaska Native 1 0.37%
Asian 10 3.73%
Black or African American 1 0.37%
White/ Caucasian 224 83.58% Hispanic or Latino 7 2.61% Non-Hispanic 2 0.75% Other 19 7.09% Gender Male 90 33.58% Female 178 66.42%
Work status Employed full time 96 35.82%
Employed part time 17 6.34%
Unemployed looking for work 7 2.61%
Unemployed not looking for work 4 1.49%
Retired 2 0.75%
Student 113 42.16%
Disabled 1 0.37%
Entrepreneur or self-employed 28 10.45%
Education Less than high school 2 0.75%
High school graduate 20 7.46%
Some college 25 9.33%
Bachelor degree 86 32.09%
Master degree (not MBA) 115 42.91%
Professional degree (e.g. MBA) 17 6.34%
Doctorate 3 1.12%
4.2. Correlation Analysis
In table 4, means, standard deviations, and correlations of relevant variables are depicted. From the correlation matrix, we can conclude several observations. Firstly, willingness to pay is significantly and negatively correlated with age (r = -.184, p < .01) but the effect is small. Moreover, it is similarly correlated to education with the similar small negative effect (r = -.155, p < .05). On the other hand, positive correlation was found between willingness to pay and usefulness but the effect was also small in size (r = .158, p < .01).
Secondly, high correlation was observed among variable where Likert scales were used for their measurement. Trust is composed from reliability, usefulness, affect, and willingness to rely on marketing on SNS and therefore, significant correlations with large effect were expected. The exceptions were willingness to pay which showed correlations of medium size with reliability (r = .372, p < .01), usefulness (r = .435, p < .01), and affect (r = .343, p < .01). In order to further justify the reliability of the scales, we performed the exploratory factor analysis. First of all, we analysed credibility and attitude toward advertisement. All of the questionnaire items related to credibility showed high factor loadings in component 2, while the other construct, attitude toward advertisement had high factor loadings in component 1. In addition, even though credibility is often a separate construct, especially in advertisement and marketing in general, it might share some variance with trust. Since trust consists of four other constructs as in the original ADTRUST scale of Soh, Reid, and King's (2009), for easier interpretation we performed four exploratory factor analyses with credibility. The results from the factor analysis are summarized in Tables 5. All items showed low cross-loading and therefore, no items were dropped from the analysis.
4.3. Regression Analysis
In order to test the hypotheses, firstly, we estimated the full structural model using AMOS including all control variables and mediating variables. Secondly, the moderating effect, was tested. The results from both analyses are summarized in the table 6 and 7. The effects of covariates are reported in table 8.
Direct effects of marketing on SNS on willingness to pay
First of all, the direct effects of the marketing strategies on SNS on willingness to pay were tested. User-generated advertisement served as a control variable and therefore, it’s effect is not explicitly stated in table 6.
The indicators of goodness of fit did not show good goodness of fit (CMIN/DF = 124.58, CFI = .604, RMSEA = .680). However, when we further tried to edit the model by deleting some paths, which were insignificant, there were no major differences which would make the model better. As a result, being aware of these limitations, we proceeded with the analysis of the results from the original model.
Hypothesis 1 suggests that there is a negative relationship between company-generated content and willingness to pay, such that the relationship is weaker when the content is generated by the user. We find that the company-generated content on Nike Facebook page decreases the willingness to pay by 0.257 (p<0.01) when compared to user-generated content. Therefore, hypothesis 1 is supported.
Another hypothesis formulating the direct effect of marketing strategies on SNSs on willingness to pay is Hypothesis 7. It suggests that there is a negative relationship between company-generated content shared and willingness to pay, such that the relationship is weaker when the content is generated by the user. From regression analysis in AMOS we found a
significant negative effect (ß= -0.188, p<0.05). When we compare it to the results from the previous regression analysis in Hypothesis 1, the effect is less negative.
The last direct effect is formulated in Hypothesis 4. It suggests that there is a negative relationship between company-generated content and willingness to pay, such that the relationship is weaker when it is company-generated content shared by another user. Subtracting the estimates from previous two regressions, we can conclude the company-generated content decreases the willingness to pay by 0.069 more than company-company-generated content shared. Since both estimates are significant, their difference is also significant and thus, we can conclude that Hypothesis 4 is supported.
In summary, by sorting the researched marketing methods, the largest negative effect on willingness to pay has the company-generated content, company-generated content shared by another user (e-wom) stands in the middle, and user-generated advertisement as subset of user-generated content has the most favourable effect out of the studied marketing methods.
Table 6. Direct Effects of Marketing on SNS
Estimate S.E. C.R. Company-Generated Content vs. UGA ---> Willingness to pay -0.257** 0.091 -2.814 Company-Generated Content
Shared vs. UGA ---> Willingness to pay -0.188* 0.092 -2.053 Note: * p<0.05, ** p < 0.01, *** p<0.001; N=268
Indirect effects of marketing on SNS on willingness to pay
Consequently, as suggested by the hypotheses 2, 5, and 8, the indirect mediating effects of marketing on SNS on willingness to pay were examined. The moderating effects are formulated in hypotheses 3, 6, and 9. The results from mediations as well as moderation are summarised in table 7.
Mediating Effects
Hypothesis 2a suggests that there is a positive relationship between company-generated content shared by another user (e-wom) and trust, such that there is a higher level of trust when the content is generated by the user compared to when it is generated by the company, or when it is generated by the user. The effect was negative (ß = -0.022) and insignificant (p > 0.05). This contradicts our assumption of positive relationship. However, because the results were insignificants there is lack of evidence to accept or reject the hypothesis. In other words, the Hypothesis 2a is not supported.
Similar results were found when testing Hypothesis 5a, which suggests that there is a positive relationship between company-generated content shared by another user (e-wom) and credibility, such that there is a higher level of trust when the content is generated by the user compared to when it is generated by the company, or when it is generated by the user. The effect was also negative (ß = -0.125) and not significant (p > 0.05). Therefore, Hypothesis 5a is not supported, too.
The last tested mediator, attitude toward advertisement on SNS, is included in Hypothesis 8a. Hypothesis 8a suggests that there is a positive relationship between company-generated content shared by another user (e-wom) and attitude toward social media advertising, such that there is a higher level of attitude toward social media advertising when the content is generated by the user compared to when it is generated by the company, or when it is generated by the user. We found that company-generated content shared by another user (e-wom) decreases the attitude toward advertisement on SNS by 0.332 (p<0.01) than the user-generated advertisement. Thus, Hypothesis 8a is supported.
Furthermore, there is a positive significant effect of attitude toward advertisement on SNS on willingness to pay, which is in congruence with our Hypothesis 8b. The beta estimate in this regression reached 0.314 with p-value smaller than 0.001 what means high level of
significance. Thus, Hypothesis 8b is also supported. Hypothesis 8 states that the positive relationship between company-generated content shared by another user and willingness to pay is mediated by attitude toward social media advertising. Because Hypotheses 8a and 8b are supported we can conclude the whole set of Hypotheses 8a, 8b, and 8 is supported.
Total indirect effect with trust and credibility as mediators cannot be computed because Hypotheses 2a and 5a are not accepted. Therefore, we cannot accept neither reject the Hypotheses 2 and 5 which suggest that the positive relationship between company-generated content shared by another user (e-wom) and willingness to pay is mediated by trust and credibility, respectively.
However, the effect of mediators on our outcome variable, willingness to pay, as demonstrated in Hypotheses 2b, 5b. The positive relationship between trust and willingness to pay was confirmed (ß= 0.391, p<0.001). The positive effect of credibility on willingness to pay was also significant but the beta estimate was smallest out of the mediating variables (ß= 0.264, p<0.001). Therefore, we conclude that Hypotheses 2b and 5b are supported.
In conclusion, in this study, the relationship of company-generated content shared by another user and willingness to pay when compared to user-generated advertisement is significantly mediated only by attitude toward advertisement on SNS.
Moderating Effect
After testing for mediation, we investigated the moderated mediation with age as a moderator. Hypothesis 3 suggests that the positive relationship between company-generated content shared by another user (e-wom) and trust is moderated by generational cohort, such that this relationship is weaker for higher values of age. We found the effect was very small, positive (ß= 0.015), and insignificant. Therefore, Hypothesis 3 is not supported.
Hypothesis 6 describes the relationship of company-generated content shared by another user and credibility when moderated by age. Particularly in Hypothesis 6 we suggest that the positive relationship between company-generated content shared by another user (e-wom) and credibility is moderated by generational cohort, such that this relationship is weaker for higher values of age. As in the case of trust, the effect is small, positive, and insignificant (ß= 0.015).
Last but not least, Hypothesis 9 which propose that the positive relationship between company-generated content shared by another user (e-wom) and attitude toward social media advertising is moderated by generational cohort, such that this relationship is weaker for higher values of age. The results are almost the same as in the previous two hypotheses, and therefore, neither Hypothesis 9 is supported.
In conclusion, unfortunately, none of the interactions was significant and all of them were very small, which implies that there is no significant effect of age on trust, credibility, or attitude toward advertisement. Due to the lack of evidence from the data analysis, none of the hypotheses with moderated mediation (H3, H6, H9) are not supported.
The summary of all hypotheses including their formulation and the status of being supported or not by our analysis is provided in Appendix 2 at the end of the study.
Effects of Covariates
Last but not least, we tested the relationship of control variables on endogenous variables, particularly the dependent variable willingness to pay, and mediating variables trust, credibility, and attitude toward advertisement on SNS.
As presented in table 8, if the respondent knew the author of this study in person, he or she tended to respond more positively than respondents who did not report any relationship to the author. The effects were similar on all mediators, which were measured with scales. Specifically, people with relationship to the author reported 0.435 (p<0.01) higher value on attitude toward advertisement on SNS, 0.431 (p<0.001) on trust, and 0.397 (p<0.05) than those with no relationship to the author. However, even though the chosen values on the Likert scales were higher, the effect on willingness to pay was negative but not significant. As a result, the type of relationship with the person who created (CGA) or shared and recommended (e-wom)
Estimate S.E. C.R. Interaction term: Age*CGC ---> Attitude toward advertisement on SNS 0.051 0.073 0.697 Interaction term: Age*CGC shared ---> Attitude toward advertisement on SNS -0.01 0.068 -0.146 H5a Company-Generated Content ---> Attitude toward advertisement on SNS -0.127 0.143 -0.891 H5a Company-Generated Content Shared ---> Attitude toward advertisement on SNS -0.332** 0.143 -2.329 Age ---> Attitude toward advertisement on SNS 0.011 0.007 1.725
H2 Interaction term: Age*CGC ---> Trust -0.007 0.057 -0.118
Interaction term: Age*CGC shared ---> Credibility 0.08 0.07 1.151
Interaction term: Age*CGC ---> Credibility 0.003 0.074 0.037
H2 Interaction term: Age*CGC shared ---> Trust 0.015 0.054 0.28
H1a Company-Generated Content ---> Trust -0.022 0.113 -0.195
H3a Company-Generated Content ---> Credibility -0.23 0.146 -1.574
H1a Company-Generated Content Shared ---> Trust -0.064 0.113 -0.567
H3a Company-Generated Content Shared ---> Credibility -0.125 0.146 -0.855
Age ---> Trust -0.005 0.005 -1.054 Age ---> Credibility 0.001 0.007 0.202 H5b Attitude toward advertisement on SNS ---> Willingness to pay 0.314*** 0.039 8.032 H1b Trust ---> Willingness to pay 0.391*** 0.05 7.884 H3b Credibility ---> Willingness to pay 0.264*** 0.038 6.949 Interaction term: Age*CGC ---> Willingness to pay 0.035 0.046 0.757 Interaction term: Age*CGC shared ---> Willingness to pay -0.024 0.043 -0.551 Age ---> Willingness to pay -0.007 0.004 -1.691 Note: * p<0.05, ** p < 0.01, *** p<0.001; N=268 Table 7. Indirect Effects
Moreover, nationality of the consumers showed a significant effect only in the case of Czech nationality. The respondents from Czech Republic reported significantly lower values on credibility (ß= -0.495, p<0.05), trust (ß= -0.448, p<0.05), and attitude toward advertisement (ß= -0.842, p<0.001). However, they were willing to pay more as the effect on the willingness to pay was positive and significant (ß= 0.386, p<0.05).
Estimate S.E. C.R. Know me ---> Attitude toward advertisement on SNS 0.435** 0.159 2.729 Know me ---> Trust 0.431*** 0.126 3.435 Know me ---> Credibility 0,397* 0.163 2.431 Gender (male = 0, female =1) ---> Credibility -0.056 0.126 -0.439 Gender (male = 0, female =1) ---> Trust -0.027 0.097 -0.273 Gender (male = 0, female =1) ---> Attitude toward advertisement on SNS -0.087 0.123 -0.707 Slovak ---> Attitude toward advertisement on SNS -0.226 0.136 -1.665 Slovak ---> Trust -0.167 0.107 -1.562 Slovak ---> Credibility -0.158 0.139 -1.136 Tenure ---> Attitude toward advertisement on SNS 0.025 0.023 1.127 Tenure ---> Trust -0.005 0.018 -0.255 Tenure ---> Credibility 0.002 0.023 0.077 Education ---> Attitude toward advertisement on SNS -0.036 0.059 -0.616 Education ---> Trust -0.044 0.047 -0.948 Education ---> Credibility -0.013 0.061 -0.221 Ethnicity ---> Attitude toward advertisement on SNS -0.069 0.063 -1.097 Ethnicity ---> Trust 0.005 0.05 0.098 Ethnicity ---> Credibility 0.037 0.065 0.567 American ---> Attitude toward advertisement on SNS 0.078 0.345 0.225 American ---> Trust 0.298 0.273 1.094 American ---> Credibility 0.135 0.354 0.382 Dutch ---> Attitude toward advertisement on SNS 0.199 0.199 1.001 Dutch ---> Trust 0.265 0.157 1.689 Dutch ---> Credibility 0,3 0.204 1.475 Czech ---> Credibility -0.495* 0.261 -1.895 Czech ---> Trust -0.448* 0.201 -2.223 Czech ---> Attitude toward advertisement on SNS -0.842*** 0.255 -3.305 Know me ---> Willingness to pay -0.049 0.108 -0.452 Gender (male = 0, female =1) ---> Willingness to pay 0.037 0.078 0.476 Slovak ---> Willingness to pay 0.137 0.087 1.569 Tenure ---> Willingness to pay -0.014 0.014 -0.95 Education ---> Willingness to pay 0.009 0.038 0.234 Ethnicity ---> Willingness to pay 0.032 0.04 0.8 American ---> Willingness to pay 0.068 0.22 0.308 Dutch ---> Willingness to pay -0.042 0.127 -0.327 Czech ---> Willingness to pay 0.386* 0.168 2.291 Note: * p<0.05, ** p < 0.01, *** p<0.001; N=268 Table 8. Control variables
5. Discussion and Conclusion
In this thesis, we studied the differences in perceptions toward three forms of advertisement on social networking sites (SNS) across generational cohorts. Various relevant insights for researchers as well as managers and other professionals emerged from the data analysis. Key findings, implications for science and professionals, as well as limitations and further research suggestions, are discussed below.
5.1. Discussion of Results – Key Findings
As our main research question, we studied the differences in perceptions toward three forms of advertisement on social networking sites (SNS) across generational cohorts. Overall, no significant effect of age on the relationship between marketing on SNS and trust, credibility, and attitude toward advertisement on SNS was found. Therefore, no causal assertions could be concluded from this moderated mediation. One of the reasons why we failed to confirm or reject the Hypotheses 3, 6, and 9 suggesting that there are differences in the way how SNS users perceive advertisement on SNS, which we showed by measuring the levels of trust, credibility, and attitude toward advertisement can be low variance in the variable age.
Except of the impact of age on advertisements’ perception on SNS, we studied the direct effects of marketing on SNS on users’ willingness to pay. Hypotheses 1, 3, and 5 formulating these relationships were supported and therefore, we can draw some conclusion from them. As we expected based on the previous research (e.g. Pehlivan et al., 2011; Lawrence et al., 2013), our findings confirmed that the user-generated advertisement is the most favourable one when compared to other two marketing methods, which we studied. Company-generated content had the least favourable, decreasing the willingness to pay by largest amount. However, when the company-generated content was shared by another user including a
