Master Thesis
What about traditional theories? The effect of
endorsements and product placements on
content enjoyment in a social media
environment.
Bram Verleur
10433090
Statement of Originality
This document is written by Student Bram Verleur 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.
Acknowledgements
This thesis is the last step for me in completing my Master degree in Business Administration at the University of Amsterdam in the specialization direction of Marketing. I would like to extend my gratitude to the intelligent, motivated and inspiring teachers of the University of Amsterdam Business School for providing such interesting lectures on very interesting topics. Most of all I would like to thank Dr. Jonne Guyt, my thesis supervisor. Not only does he have a great deal of knowledge and expertise in a wide array of topics, he is also a very pleasurable person to work with. He has been a great support for me in writing my thesis.
Additionally, I would like to thank my family and friends, for supporting me all along the way. Specifically, I would like to thank my brother, Wouter Verleur, for helping me in writing the more technical side of the scraper code I needed. Wouter, I couldn´t have done it without your support.
I hope you as a reader will enjoy my thesis, and that it will add to your knowledge on social media and product placement.
Yours sincerely, Bram Verleur 27th of January 2017
Abstract
Product placement has been around for a long time, and much research has been done on this topic. However, the majority of these studies took place in traditional media environments, where fewer studies focused on content on websites, and none on social media. Little research has been done on the effect of product placement on content enjoyment, and the area of social media has not been touched upon at all in regard to this topic. This study aims to provide insights on how product placements on the social medium Facebook affects the content enjoyment of Facebook posts in the form of likes and subsequently the following of 20 celebrities. The concepts of transportation theory, the role of matchup and media integration are discussed. The study shows that product placement has significant negative effects on the short-term content enjoyment of posts in the form of likes, and perhaps more importantly, it has indirect and direct negative effects on the following of influencers. The average post that contains product placement leads to a 77,6% decline in content enjoyment for that post compared to if it did not contain product placement. Every case of product placement leads to 0,021% less followers of a page, effectively eroding the fan base that is being used as an asset. These effects are moderated by different factors. The moderating roles of perceived fit and content quality are discussed in this paper. It appears that on social media the effects of perceived fit and the role of matchup are less strong than in traditional media. Content quality however, is essential on social media. It has strong direct influences on the content enjoyment, as well as indirect influences via other constructs on the following.
Table of contents
Introduction 5
Literature review 7
Product placements and endorsements 7
Content enjoyment 8
Movies, TV & Social Media 10
Effects of content enjoyment on following 12
Perceived fit 12
Content quality 13
Obviousness and personal or third party product placements 14
Conceptual model 15
Method 16
Data collection and coding 16
Choice of social medium 17
Chosen celebrities 18
Participants 18
Sample procedure questionnaire 18
Results 19
Variables 19
Validity and reliability 20
Short-term effects 21
Main effects 21
Moderated effects 22
Other analyses 23
Long-term effects 26
Controls & Method of analysis 26
Main effects including moderators 28
Dynamics and robustness 29
Discussion/Conclusion 30
Managerial implications 36
Limitations 37
Directions for future research & Final remarks 38
Introduction
Product placement is a practise that has been around for a while. It is a commonly used marketing tool that aims to incorporate brands purposefully into editorial content (Schneider & Cornwell, 2005). In fact, the earliest reported product placement occurred in 1896, with the deliberate integration of the Lever brothers’ (half of the to be formed multinational Unilever) ‘Sunlight Soap’ into several films made by the Lumière brothers (de Gregorio & Sung, 2010). Over the years, product placement has evolved. During the great depression, movie makers were eager to drive total production expenditure down by showcasing products in films (Brett, 1995). In specific cases, such as the 1931 film ‘It pays to advertise’, this led to negative publicity about the surrender of media content to commercial interests (Brett, 1995). After a long period of little to no product placement, a rebirth was catalysed by various factors (e.g. the diminishing role of movie studios, emergence of independent producers and location-based movie production) during the 1960s and 70s (Balasubramanian, Karrh & Patwardhan, 2006). Growth spurts during the 1980s and 90s have ultimately led to a multibillion dollar industry, with global spending in branded entertainment totalling at an estimated $10.6 billion in 2014 (PQ Media, 2015). The global spending on placements in digital media integrations have even risen a massive 35.8% in 2014, compared to a 13.6% rise of the total product placement category (PQ Media, 2015). With a total spending that vast, and a growth rate that high, it can be considered an important part of the marketing mix.
The rapidly increasing expenditure on digital media product placements and endorsements seems to signal that marketers think that it works, however little to no academic research shows that product placement actually has an effect on consumers’ attitudinal and behavioural factors important in the social media environment. In this research, the event of product placement will be studied in a social media environment, specifically on Facebook, in its effect on content enjoyment of posts in the form of likes (of the post) and the number of
followers (of the page on Facebook). Content enjoyment of product placement posts can be seen as attitude towards the ad and is thus, as classical conditioning states, important for the attitude towards the brand. The total number of followers is made up by likes and unlikes. Likes and unlikes are the decisions of Facebook account holders to follow a page on Facebook to keep up to date with the information and posts provided on that page.
The literature up until now has focused on the effect of product placement on absorption (Wiles & Danielova, 2009) and its effect within transportation theory (Green et al, 2004), brand attitudes, -recognition and -memory, and behaviours in response to the product placement on the product that is advertised (De Gregorio & Sung, 2010; Peters & Leshner, 2013; Dens et al., 2012; Reijmersdal, Neijens & Smit, 2009) and less on the effect of the product placement on perceptions of the media carrier (Rosengren & Dahlén, 2013; Goldstein et al., 2014). Little research has been done on the effect of product placement on content enjoyment, and the area of social media has not been touched upon at all in regard to this topic. The relationship between the endorser and the product placement will be examined in its effect on the relationship between product placement and content enjoyment as perceived fit. The quality of the content itself will also be added as a moderating factor in the relationship between product placement and content enjoyment. The effect of content enjoyment on ad avoidance, in the form of mutations in followers or likers, will also be studied. An effect is expected of falling content enjoyment on ad avoidance. Within product placement variations are also present, it is expected that the obviousness of the product placement, thus the obviousness of the fact that it is a product placement, will play a role. Product placements for the endorser themselves are also expected to be viewed as different than when they are endorsing a product for a third party. The general research question following from the previous section is:
To what extent does product placement affect content enjoyment of Facebook posts and subsequently the following of the endorser?
The answers to this research question will provide knowledge in a topical area which has little to no research to date. The relevant theories on product placement and endorsements will be tested in a social media environment. The answers may open up a whole new area of research and might help explain if and why product placement on social media works as it does on traditional media. As the expenditure on digital product placements is rising so rapidly, this answer is also of uttermost importance to marketers distributing their budget across channels and might help in allocating the budget to certain campaigns and media-outlets over other communication messages.
Literature review
Below, the present study is placed within the context of the current literature on product placement and endorsement. Firstly, the main concepts of product placements, endorsement and content enjoyment are defined, after which we turn to the differences between traditional and social media. Subsequently, the moderating and direct roles of perceived fit and content quality are discussed. Also, the concepts of obviousness and the party for which the endorsement is taking place are evaluated in their relation with product placement. Finally, the research questions and hypotheses are presented in a relational model.
Product placement and endorsements
Brands are often used as tools by writers, directors, set designers and other creative professionals to communicate specific meaning to audiences, to help set the timeframe, or to express characters’ personality traits (Balasubramanian, Karrh & Patwardhan, 2006). However generally, the intentions are not just that. Product placement or endorsements are often used as a promotional tool, a paid attempt to influence audiences and to change the
diverse. Ferraro and Avery (2000) define it as the inclusion of brands in movies or television scripts. Schneider, Systems and Cornwell (2005) view product placement as a purposeful incorporation of a brand into editorial content. The definition of Schneider et al (2005) still holds in a social media environment, as the timeline of a social media page can be viewed as editorial content in which product placement can be incorporated. The way of paying for the product placement does not matter in this study, whether it’s a barter-type kind of deal, an incorporation payed for with a currency, or another motivation. The individuals who are viewing the content are not aware of the exact motivations to place the product on the page either way, even if they are aware of the post being a product placement.
Whilst differing on a number of points, endorsements and product placements are seen as one and the same construct in this study. In a social media environment product placements are more personal, as the editorial content which they are integrated in is linked to an entity with its own values. Product placements in social media can thus be seen as said entity endorsing a product or brand.
Content enjoyment
Enjoyability is a concept that is easy to define for an individual person, however, the specifics of enjoyability are constructed differently per person. If one would ask ten different people what enjoyability is, it is likely that the answers would not be in line with each other. Raney (2003) says that enjoyment broadly refers to a pleasurable affective response to a stimulus. However, not only affective factors are important, cognitive factors also influence a persons’ evaluation of media (Raney, 2002). In this study, content enjoyment is viewed as the amount of likes a post on Facebook gets. When a person on Facebook clicks on the like button, they are expressing their appreciation or support for a post and thus enjoyed the content in some way or another. In this study the content enjoyment will be assessed on post-level. So,
timeline. A Facebook post consist of multiple elements: The name of the person or page posting, a small image which is their profile picture or avatar, a timestamp, the privacy settings, an arrow with a drop down menu with some options for action if the post offends you or you want to get notifications about the post, and the content of the post: Text, or a combination of text and a video, an image, an image-carousel, a link with or without a preview image. In their natural environment the amount of likes, comments and shares and a like, comment and share button are also shown underneath the post, but these are not part of the post itself. When a person clicks on the comments button, other comments also are shown automatically. An example of a Facebook post in its natural environment is given in Figure 1.
Transportation theory specifies that the experience of being taken into a narrative world is a key aspect of experiencing media. This transportation or absorption is critical to media (content) enjoyment - the benefits that might come from media exposure - and the conditions under which enjoyment is more or less likely to occur (Green, Brock & Kaufman, 2004). To enjoy content, it is thus critical to be absorbed by the content. To believe what you are seeing, reading or hearing. Absorption leaves little motivation and ability for other tasks, such as processing product placement (Wiles & Danielova, 2009). If a lack of cognitive resources is indeed driving the lack of attention to product placement, the reverse of this statement may be correct as well. Hence, it can be assumed that processing product placement leaves less cognitive capacity for absorption and the content is thus enjoyed less. However, if the reverse holds for the social media is not certain.
Movies, TV & Social Media
While, as discussed in the previous chapter, the current literature leaves us with a relatively strong assumption that product placement might hamper absorption; in the social media environment this might be completely different. Product placement in a movie or in a magazine can be seen as incorporated into, or part of the content itself, however on social media it might be perceived as incorporated in a different way. On a social media platform every post is a different story that is told, a different piece of content. The general topic of the page might be the same, but most of the time the individual posts on a page do not share the same storyline. In a more coherent medium such as a movie, product placement can take place within the storyline and is processed simultaneously with the original content. Because viewers’ primary goal is to process program content (Brechman, Bellman, Schewda & Varan, 2015) and as said before, absorption leaves little motivation and ability for processing product placement (Wiles & Danielova, 2009) consumers allocate less cognitive resources towards these processing activities. This is also in line with the limited capacity model of mediated
processing of Lang (2000). However, if we assume that social media product placement does not need to be processed together with the content that consumers are motivated to process, so within the same storyline, this might lead to a completely different effect. If this is true, product placement does not need to fight with the original content for the attention of the consumer, as the product placement is the content that is to be processed. It is interesting to see whether or not the processing of social media posts has the same effect as the processing of product placement integrated within the media on content enjoyment. And thus, if it can be assumed that the concerned theories about traditional media hold for social media. It is expected that the theories also hold for social media, this leads to the following hypothesis:
H1: Product placement makes content enjoyment fall in a social media environment
Effects of content enjoyment on following
The goal of product placements is obviously to induce a positive attitude towards the brand or product and ultimately influence the purchasing behaviour of the consumer or its network towards the placed product or brand. But product placements might not always be beneficial for the sender of the message, as consumers’ response to the placement might be affected by factors within the placement itself. Goldstein, Suri, McAfee, Ekstrand-Abueg & Diaz (2014) found that annoying ads cause drop-out. They found that people needed to be compensated more than the advertisements’ actual revenue to stay on the page. People also notice and complain more about annoying ads and are more likely to abandon the sites on which they are surfing. One might say that if a person finds the placement annoying, they are not enjoying it, and thus the content enjoyment falls. El-Adly (2010) found, when doing a research about TV-ads, that the more negative attitudes about TV-ads are, the more likely people are to avoid the ads. So, if people dislike the content and thus enjoy the content less, they are more likely to avoid the ads. Transferring this to the social media environment: If content enjoyment of the
they might remove the provider of the annoying ads out of their timeline, which can be seen as the ultimate form of ad avoidance. It is interesting to find out if this page-leaving behaviour also transfers to the extreme of unliking someone consumers were a fan of at first, which naturally is the opposite of what social media influencers are trying to achieve. The hypothesis that follows is:
H2: Content enjoyment positively influences the existing growth trend in amount of
followers.
If we see the celebrity endorser as a brand, which they essentially are, we can view the action of unfollowing or unliking the celebrity as a brand divorce. The Merriam-Webster dictionary defines a divorce as “the ending of a marriage by a legal process” or “a complete separation between two things”. As shown by the definition, divorce is quite a drastic decision. Whilst a person first liked a page, they are now withdrawing their appreciation because of annoyance. From Srinivasan, Rutz and Pauwels (2016) it became clear that whilst a like only has a small positive effect on sales, a dislike has a very large negative effect on sales. The effect of dislikes is thus disproportionally high, and dislikes are something that should therefore be avoided.
Perceived fit
As Kamins (1990) already described in his article about the matchup hypothesis, endorsers are more effective when there is a fit between the endorser and the endorsed product. When an endorser is e.g. attractive, this endorser is more effective in endorsing a product that is used to enhance the attractiveness of the consumer. Whilst early research focussed mostly on attractiveness, later research also included factors such as expertise and appropriateness (Till & Busles, 2000). The role of the matchup hypothesis was recently evaluated again by Wright (2016), who confirmed that the role of matchup was still present, and mediated by the
appropriateness of the endorser for the endorsed product. Wright (2016) also found that the social adaptation theory is of relevance in the matchup hypothesis. When the endorser was appropriate and thus fitting for the endorsed product, consumers used the endorser to help in forming brand evaluations, when there was a mismatch, consumers relied solely on product attribute evaluations. As the role of matchup and thus the fit between the endorser and the endorsed product has been confirmed time after time, it is expected that the perceived fit between the endorser and the posted content will be a moderating factor in the relationship between the product placement and content enjoyment and it will have a direct effect on the content enjoyment as well. The endorser plays a big role in the evaluation of the brand, so it is expected that they play a big role in the evaluation of the content itself as well. The hypotheses are thus:
H3a: When the perceived fit is higher, the content enjoyment will also be higher.
H3b: When the perceived fit is higher the effect product placement has on the
content enjoyment is less negative.
Content quality
Rosengren and Dahlén (2013) found that the advertising content matters for the evaluation of the media vehicle. In their study, they found that high-end advertising increases the evaluation of a magazine. The value added by the actual advertising content causes this effect. This means that this content is of such quality, that the content itself increases the content enjoyment. It is expected that this also holds in a social media environment, a moderating effect of content quality is thus expected. As the content quality of advertising is so important for the content enjoyment, it is logical that this also holds for content quality in general. This means a direct effect of content quality on content enjoyment is also expected. Concluding, it can be stated on the basis of the previous paragraph, that the quality of the content is likely to
matter for the attitude towards the post, which is measured by content enjoyment in this study, the hypotheses that follows are thus:
H4a: When the quality of the content is higher, the content enjoyment is also higher.
H4b: When the quality of the content is higher the effect product placement has on the
content enjoyment is less negative.
Obviousness and personal or third party product placements
Product placement as such is not a unilateral construct. Some product placements are more obvious than others and some are endorsing the self instead of third parties. It is expected that these two factor have an influence on the effect product placement has on content enjoyment. Obviousness in product placement has been examined by D’Astous and Chartier (2000) as the noticeable manifestation of the placement. They found that dependent on the integration of the placement, the obviousness had a greater impact on liking when integration was high, and made product placement more unacceptable when the integration was low. So when the integration within the content was high, noticeable manifestation actually increased liking. When it was not integrated well within the content, making it stand out more obviously, it made the product placement more unacceptable, which decreased liking. Additionally, Dens, de Pelsmacker, Wouter and Purniwirawan (2012) found that brand prominence in movies can have a negative effect on the attitude towards the placed product, this is in line with previous research by Cowley and Barron (2008). In a prominent product placement, it is obvious that the product placement is there as a persuasion cue, which might cause the audience to start worrying about the reasons for the placement (Dens et al., 2012). Concluding, we can expect the following relation:
H5: When a product placement is obvious (non-obvious) it has a negative (positive)
Endorsements do not only take place for third parties, something endorsers have produced themselves can also be endorsed. Examples could be Beyoncé (a singer) launching and album, Kylie Jenner (a public figure/reality soap star) launching a new lip shade, or Katherine Heigl (an actress) promoting her blog. As said before, the role of matchup plays a big role in effectiveness of the endorser for the endorsed product (Kamins, 1990; Wright, 2016). The matchup between an endorser and a product they themselves launched is obviously very high, as it is their own project. The following relation is thus expected:
H6: When a product placement is directed at the endorser themselves (third party) it
has a positive (negative) effect on the content enjoyment.
Conceptual model
The model that can be formed on basis of the researched relations is shown in figure 2.
Figure 2. The effect of product placement on content enjoyment and following
+ + +
- + +
Method
To answer the research question a multiple approach was taken. Data was scraped from Facebook by communicating with the application programming interface (API) of Facebook, social media posts were coded, and an online survey was distributed. To collect the Facebook data two data scrapers were written in the program R. A data scraper collects data from the appointed online environment. This data is real-world data and the participants are not aware that anonymous data is collected about them, making it the ultimate natural environment data.
Product placement Content enjoyment Following
Perceived fit Content Quality Obviousness
Data collection and coding
The first scraper scanned the social media pages of 20 selected celebrities and collected the number of followers three times per day in the analysis period 29-06-2016 to 05-09-2016. In total, the data consists of 171 time points for each celebrity, resulting in 3420 observations. The second scraper collected all the Facebook posts posted by the celebrities in the analysis period, including amount of likes, comments and shares, from their public Facebook pages. The scraper to collect the posts was run sometime after the last following scraper, to allow lagged post likes for newer posts to exist as they do for older posts. In total 1882 posts were analysed, an average of 94 per celebrity. The data was subsequently classified as being non-product placement or non-product placement, classified as being obvious non-product placement or not and classified as being product placement about the endorser itself or third parties. The Facebook posts were coded (by the author) on a five question, 1 to 7 Likert scale for the constructs “Perceived Fit” and “Content Quality”.1 Because perceived fit is often researched
in the context of a specific aspect of fit in combination with source credibility, such as attractiveness cues for an attractive endorser, a new scale had to be constructed to be able to measure more general perceived fit. This scale was constructed on the base of the relatedness scale of Sengupta, Goodstein and Boringer (1997), and the perceived expertise, trustworthiness and attractiveness scale of Ohanian (1990). The common aspect of these scales was that they describe specific aspects of the content in combination with aspects of the endorser. This common aspect was taken and a new more general scale was constructed to measure perceived fit of the endorser with the posted content. The number of aspects that relate to content quality that a Facebook post can reflect, such as fonts, imagery, lay-out and multimedia functioning (Akadwabu & Palvia, 2002) is very limited. Because of this a new scale, focussing on the aspects that can vary between posts, was created. The scale was based
1 Due to the high amount of data and difficulties in obtaining consistent and reliable responses on these scales
on user-perceived web quality scale of Aladwani and Palvia (2002). The aspects taken into account are the quality of the imagery and the quality of the post itself, thus covering the imagery and the text part of the post. Both scales can be found in appendix III.
Additionally, a benchmark survey was conducted via an online questionnaire to be able to control for possible bias in the coding of the posts. A randomly selected stratified sample was taken from the Facebook post database. An online questionnaire is a good way to inexpensively reach a lot of people in a short amount of time, with little to no costs (Wright, 2005). Getting a good sample could be a problem with an online questionnaire, because participants without internet cannot be reached (Wright, 2005). This poses no threat to this study as the study takes place in a social media environment. The data that was collected via the questionnaire was collected between 15-11-2016 and 23-11-2016 to try to limit the amount of possible time-related confounding variables that might influence the reliability of this part of the study negatively.
Choice of social medium
The data was collected from the social medium Facebook. Facebook is the largest social medium, with 1.71 billion active users globally (Statista, 2016). Facebook has not restricted the richness of data access via the API recently and had not announced any plans to do so in the period of data collection. In January 2016 Facebook was the most popular in all but 7 of the worlds social media using countries (Similarweb, 2016). It was thus the most consistent access point of social data whilst also having the largest and most omnipresent active user base, which led to the decision of this social medium being the medium of choice.
Chosen celebrities
The 20 chosen celebrities were selected on six aspects: i) they had to post regularly, ii) had to endorse regularly on their timeline, iii) had to post in English most of the time and needed to
iv) had to have a considerable amount of likes on their page (above 1mln.), v) had to have a real life besides their commercial activities and vi) had to still be in the public eye. Of the celebrities 11 were male and 9 were female. They self-categorised as Actor/Director (n = 6), Musician/Band (n = 5), Public Figure (n = 4), Athlete (n = 4) and Entertainer (n = 1). A list of the chosen celebrities can be found in the appendix I.
Participants
The scraper collected data from pages of celebrities. The input of this page, the likes, comments, shares and number of followers is created by the visitors of their page. The data input of Facebook is thus generated by the likers, commenters, sharers of the posts and the likers of the page. These participants are anonymous, so there is no full information publicly available concerning their demographics. What is known about these persons is that they liked one or more posts on the selected pages in the period of 29-06-2016 – 05-09-2016. During this period they thus must have had an internet connection, a Facebook account, and have seen one or more posts published by the selected pages.
Sample procedure benchmark questionnaire
A convenience sample was used to select the participants for the questionnaire. A convenience sample is a method of sampling in which participants are collected on the basis of their relative reachability. With convenience sampling the external validity could fall because only easy to reach participants will be selected. However, this method of sampling enables the investigator to reach a lot of participants in a limited time (Bryman, 2008). To reach the participants mainly Facebook was used, but face-to-face invites were also used. By means of a call on Facebook participants can be reached easily, quickly and inexpensively. The demographics of the sample, questionnaire and coding guide can be found in appendix II and III.
Results
Variables
For this study two different datasets were used. One dataset contains the data on post level for short-term effects, the other contains the data per scrapemoment for long-term effects. The posts in the long-term dataset were assigned to the specific moments the datascraper pulled data from the social media environment. Due to the interval used (eight hours between scrape moments), it could happen that a celebrity had multiple posts between scrape moments. As a result, there are several observations with multiple posts per celebrity per scrapemoment, as well as one post or none. The used variables for both the short-term and long-term analysis are described in table 1.
Table 1. Measures and level of measures of used independent and dependent variables
Variable Measure Level
Product Placement 0 = No, 1 = Yes Dichotomous
Obviousness* 0 = No, 1 = Yes Dichotomous
Third Party* 0 = Self, 1 = Third Party Dichotomous
Perceived Fit 1 to 7 continuous Likert scale Interval/Ratio Content Quality 1 to 7 continuous Likert scale Interval/Ratio
Likes Numeric Ratio
Comments Numeric Ratio
Shares Numeric Ratio
Followers Numeric Ratio
Interaction Perceived Fit 1 to 7 continuous Likert Scale * Product placement
Interval/Ratio Interaction Content
Quality
1 to 7 continuous Likert Scale * Product placement
Interval/Ratio
Note: * Variables ‘Obvious’ and ‘Third Party’ can only take place if variable ‘Product Placement’ is 1. If ‘Product Placement’ is 0, ‘Obvious’ and ‘Third Party’ are also 0.
Besides the described variables, a dummy was added for every individual celebrity for the short-term analysis. This way the individual differences between the celebrities are controlled for. For the long-term analysis, the individual differences were controlled for by mean centering the followers. Dummies were also added for scrapemoments that followed an error in the datascraper that lasted for more than 3 scrapemoments (1 day). This led to 3 dummies being added.
Validity and reliability
The data collected is real world data, collected without the participants knowing. This amounts to an absence of bias caused by the data collection itself. Most threats to internal validity caused by data collection are thus not present. A history effect might be present because of the data collection period of 69 days. It is however expected that the number of celebrities used (N = 20) and the number of followers (N = 595 mln.) they have among them control for most of these possible effects. The threat to external validity mainly comes from the fact that the celebrities used in the research are popular in western culture, but might not all be as well known in non-western societies. This might threaten the generalizability of this study.
A comparison of the coding from the benchmark survey and the manual coding shows a high overlap (a Cohen’s Kappa of .76). The (minor) difference between the benchmark and the author might be caused by the fact that the author was trained better in distinguishing product placement and had more background information and knowledge concerning the celebrities. The construct Perceived Fit had a Cronbach’s alpha of α = 0.954 for the participant coders. The construct Content Quality construct had a Cronbach’s of α = 0.991 for the participant coders.
Short-term effects Controls
Dummies for each individual celebrity were added to all short-term analyses to be able to control for the differences between them.
Table 2. The effect of product placement on the amount of likes on a post Amount of likes on post
Main effect With moderators
Variables B SE p B SE p Constant 101244 14912 < 0,001 60034 15539 < 0,001 Product placement -114114 27198 < 0,001 -44656 28182 0,113 Obvious -27675 26034 0,288 -73097 25936 0,005 Third Party 82434 18397 < 0,001 57358 18620 0,002 Perceived Fit 14985 5666 0,008 11712 8545 0,171 Content Quality 26856 5450 < 0,001 68206 7723 < 0,001 PPL*PF - - - 10294 11189 0,358 PPL*CQ - - - -78695 10510 < 0,001 (N = 1814)
Note: Variables ‘Product placement’ and ‘Obvious’: Yes (1) No (0), ‘Third Party’: Third Party (1) Self (0), ‘Perceived Fit’ and ‘Content Quality’: scale from 1 to 7 where 1 is lowest and 7 is highest. Significance level: p < 0,05 (two-tailed). To control for the individual celebrity effects, 20 dummy variables were added to these analyses, one for every celebrity.
Main: R = 0,798, R² = 0,637, With moderators: R = 0,810, R² = 0,656
Main effects
A linear regression was performed to investigate if product placement makes content enjoyment fall in a social media environment (H1). The model was statistically significant
(F (24,1790) = 130,88; p < 0,001) and explained 63,7% of the variance in content enjoyment. The output of this analysis can be seen in table 2. From the analysis, it became apparent that when product placement is present, the likes on a post are significantly lower than when it is not present. H1 thus holds. Within product placement there are also differences, some product
(H6). Obviosity does not significantly influence the direct effect product placement has, hence
H5 does not hold. When a product placement is of a third party the effect product placement
has, is significantly less negative in comparison with a product placement about the self. The observed relationship for H6 is thus the opposite of the expected relationship based on the
literature review.
Within the linear regression to determine the effects of product placement on content enjoyment, the perceived fit of the post with the celebrity and the quality of the posted content itself were also analysed. It was expected that when the perceived fit was higher, the content enjoyment was higher as well (H3a). When looking at the output, it can be seen that the
perceived fit has a significant positive direct effect on the content enjoyment and the hypothesis thus holds. When the content quality was higher, it was also expected that the content enjoyment was higher (H4a). This hypothesis holds, as can be seen in the output. The
content quality has a significant positive direct on content enjoyment.
Moderated effects
A second linear regression analysis was performed for the effects perceived fit (H3b) and
content quality (H4b) have on the effect product placement has on content enjoyment. The
model was statistically significant (F (26,1788) = 131,05; p < 0,001) and explained 65,6% of the variance in content enjoyment. The output of this analysis can be seen in table 2. Perceived fit did not have a significant effect on the effect the presence of product placement has on content enjoyment. H3b does not hold. Content quality however does change the effect
product placement has on content enjoyment significantly. The negative effect product placement has on content enjoyment is significantly enlarged. However, the relationship is not as expected in H4b. Higher content quality leads to a larger negative effect of product
moderators are added, obviosity however does have a significant negative effect on the effect product placement has on the content enjoyment.
The moderated analysis seems to show that product placement does not have a significant effect anymore on content enjoyment. However, product placement still has an effect via the moderator PPL*CQ. When there is no product placement, this dichotomous variable has a value of 0, meaning that the sum of PPL*CQ is also zero. However, when there is product placement, this variable has a value of 1, causing PPL*CQ to have a value higher than zero. When the content quality is compared with PPL*CQ the effect of product placement becomes visible. The product placement thus still has a significant effect via the moderator PPL*CQ and the content quality.
Other analyses
Whilst the comments and shares are not directly important for the expressed content enjoyment, they do correlate with the content enjoyment in the form of likes on posts significantly, as can be seen in table 3. What makes comments and shares of interest is their effect on the long-term. It is thus also interesting to know what factors influence the comments and shares on the short-term, as this flows through as a long-term effect.
Table 3. Pearson correlation for likes, comments and shares
Measures Likes Comments Shares
Likes -
Comments .823* -
Shares .517* .780* -
Note. * = p <.001
Main effects comments
To investigate the main effects the model has on comments, a linear regression was performed. The output of this analysis can be found in appendix IV. The significant effect of
product placement on the amount of comments and the effect that the obviousness of the placement has on this effect have been graphically represented in plot 1. Besides this effect, only content quality had a direct significant effect on the amount of comments, the better the content quality was, the higher the amount of comments.
Plot 1. Graphical presentation of the significant interaction between product placement and a
placement about a third party or the self on the amount of comments on a post
Moderated effects comments
A second linear regression was performed to investigate the effects perceived fit and content quality have on the effect product placement has on comments. The output of this analysis can be found in appendix IV. The significant interaction effect of content quality with product placement on the amount of comments has been graphically represented in plot 2. Other variables did not have a significant effect on the amount of comments on a post.
0 100 200 300 400 500 600 700 800
Third Party Self
Am o u n t o f co m m en ts Product placement No product placement
Plot 2. Graphical presentation of the significant interaction between product placement and
content quality on its effect on the amount of comments on a post
Main effects shares
To investigate the main effects the model has on shares, a linear regression was performed. The output of this analysis can be found in appendix V. The significant effect of product placement on the amount of shares and the effect of obviousness has been graphically represented in plot 3. Besides this effect, only content quality had a direct significant effect on the amount of shares, the better the content quality was, the higher the shares.
Plot 3. Graphical presentation of the significant interaction between product placement and a
placement about a third party or self on the amount of shares of a post
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 1 2 3 4 5 6 7 Am o u n t o f co m m en ts
Level of content quality
Product placement No product placement 0 200 400 600 800 1000 1200 1400 1600 1800 2000
Third Party Self
Am o u n t o f sha re s Product placement No product placement
Moderated effects share
A second linear regression was performed to investigate the effects perceived fit and content quality have on the effect product placement has on shares. The output of this analysis can be found in appendix V. The significant interaction effect of content quality with product placement on the amount of comments has been graphically represented in plot 4. Other variables did not have a significant effect on the amount of comments on a post. Perceived fit does have a positive marginally significant direct effect on the amount of shares when moderators are added, but without the significance of its moderating counterpart this is difficult to interpret.
Plot 4. Graphical presentation of the significant interaction between product placement and
content quality on its effect on the amount of shares of a post
Long-term effects Controls
Dummies were added for three scrapemoments that were expected to confound the results, namely scrapemoments that had a distance of more than 1 day from the previous scrapemoment. When the distance between scrapemoments was too large, the difference between them became a confounding factor. The dependent variable ‘followers on page’ was
-5000 0 5000 10000 15000 20000 25000 1 2 3 4 5 6 7 Am o u n t o f sha re s
Level of content quality
Product placement No product placement
mean centred to control for differences between the absolute number of followers of the celebrities. The first difference was taken from scrapemoment to scrapemoment, this way the differences between the scrapemoments were taken instead of the cumulative effects. Using the first difference cancels out the continuous trend of an increasing number of followers, allowing for focus on the actual effects of the variables of interest.
Method of analysis
A mediation effects is expected between dependent variable 1 ‘content enjoyment’ and dependent variable 2 ‘number of followers on page’, however this cannot be fully tested because the amount of observations for the dependent variables are different (DV1 = 1814 observations, DV2 = 3420 observations). This difference is caused by the level at which the data is analysed. For the short-term DV1 this is on post level, whilst on the long-term DV2 the posts are aggregated per scrapemoment. Whilst this data structure prohibits a more encompassing analysis, it is expected that this more encompassing analysis method does not lead to very different results.
Table 4. The effect of product placement on the number of followers of a page Number of followers on page
Main effect With moderators
Variables B SE p B SE p Constant -1481,2 416,7 < 0,001 -1538,8 417,1 < 0,001 Product placement -5328,6 2179,7 0,015 -4755,4 2229,5 0,040 Obvious 4605,9 2212,2 0,037 3833,2 2252,7 0,089 Third Party 1205,9 1310,0 0,357 921,6 1135 0,490 Perceived Fit 805,2 740,0 0,277 1083,6 1090,6 0,320 Content Quality -143,3 738,3 0,846 648,5 1037,4 0,532 Content enjoyment 0,032 0,004 < 0,001 0,032 0,004 <0,001 Comments -1,202 0,389 0,002 -1,219 0,389 0,002 Shares 0,111 0,088 0,207 0,101 0,088 0,252 PPL*PF - - - -378,4 1497,7 0,801 PPL*CQ - - - -1561,8 1447,2 0,532 (N = 3420)
Note: Dependent Variable is first differenced and mean centred. Variables are aggregated per scrapemoment. Variables ‘Product placement’, ‘Obvious’ and ‘Third Party’ are the sum of
observations per scrapemoment, ‘Perceived Fit’, ‘Content Quality’: mean centred from 1 to 7 scale where 1 is lowest and 7 is highest. ‘Content Enjoyment’ (Likes) , ‘Shares’ and ‘Comments’ are the mean amount of units per scrapemoment. Significance level: p < 0,05 (two-tailed). Dummy variables for 3 scrapemoments that had a distance of 1 day or more from the previous scrapemoment have been added.
Main: R = 0,338, R² = 0,114, With moderators: R = 0,339, R² = 0,115
Main effects including moderators
A linear regression analysis was performed to investigate if content enjoyment has an effect on the number of followers of a Facebook page (H2). The model was statistically significant
(F (13,3386) = 33,80; p < 0,001) and explained 11,5% of the variance in content enjoyment. The output of this analysis can be seen in table 4. The content enjoyment was expected to have a positive influence on the existing growth trend in number of followers (H2). This
means that when content enjoyment increases, the number of followers also increases. From the results it can be seen that this is in fact true, H2 holds. However, this is not the only result
found. The amount of comments a post gets has a significantly negative effect on the number of followers. The amount of shares on a post did not matter significantly. The number of product placements had a significant negative effect on the number of followers. When there were more product placements per scrapemoment the difference in number of followers between that scrapemoment and the previous one became more negative. Whether it was an obvious placement or whether it was a placement about the self or a third party did not matter significantly. Also the perceived fit between the content and the celebrity and the quality of the content itself did not matter significantly for the difference between two scrapemoments. The indirect effects of perceived fit and content quality also do not exert any significant influence via an effect on product placement.
Dynamics and Robustness
To check if and when the effects of product placement fade out on the long-term, lags can be added. In Table 5, the full long-term model was tested with instead of a direct product placement variable, a lagged one. Not all independent variables are reported in the table, the main interest of these analyses is the lagged effect. For an overview of the variables used see the notes in the table.
Table 5. The effect of time-lagged product placements on the number of followers of a page
Number of followers on page
Main effect With moderators
Variables B SE p B SE p PPL lag 1 -1986,3 2119,6 0,349 -2080,3 2120,1 0,327 PPL lag 2 -1096,2 2126,6 0,606 -1014,5 2123,1 0,633 PPL lag 3 -1720,9 2136,0 0,421 -1713,5 2135,6 0,422 PPL lag 4 -1456,4 2147,9 0,498 -1472,1 2147,1 0,493 PPL lag 5 -2178,6 2147,0 0,310 -2127,3 2146,6 0,322
Lag 1 (N = 3400), Lag 2 (N = 3380), Lag 3 (N = 3360), Lag 4 (N = 3340), Lag 5 (N = 3320) Note: The lags of variables ‘Obvious’, ‘Third Party’ and the variables ‘Perceived Fit’, ‘Content Quality’, ‘Content Enjoyment’ (likes), ‘Shares’ and ‘Comments’ were also added to the analysis Significance level: p < 0,05 (two-tailed).
As can be seen in the table above, none of the lagged product placement variables are able to exert any significant influence on the followers of a celebrity on the next scrapemoment. This in effect means that the significant direct negative effects of the product placement on the followers wear off after eight hours.
Discussion/Conclusion
With the results from the analysis the main research question: ‘To what extent does product
placement affect content enjoyment of Facebook posts and subsequently the following of the endorser?’ can be answered. In short: It can be concluded that product placement has a
significant negative effect on both the short-term, in the form of a decreased content enjoyment, and on the long-term, both direct and indirect, in the form of Facebook-page unliking. The relational model however, turned out to be slightly different from what was expected. The relational model with all significant relationships and their direction is shown in figure 3. The direction and significance level of the short-term effects are summarized in table 6, the long-term effects are summarized in table 7.
Figure 3. Conceptual model of significant relations between variables. -+ + + + + -- + + Short-term effects
Table 6. Summary of effects and direction for the short-term analyses
Output Independent variable Significance level Direction Content enjoyment (likes) Product placement < 0,001 - Product placement Obvious 0,288 n.s. Product placement Third party < 0,001 + Perceived fit 0,008 + Content quality < 0,001 + PPL*PF 0,358 n.s. PPL*CQ < 0,001 +
Comments Product placement 0,001 -
Product placement Obvious 0,881 n.s. Product placement Third party 0,015 + Perceived fit 0,098 n.s. Content quality < 0,001 + PPL*PF 0,715 n.s. PPL*CQ < 0,001 +
Shares Product placement 0,003 -
Product placement Obvious 0,484 n.s. Product placement Third party 0,002 + Perceived fit 0,104 n.s. Content quality < 0,001 + PPL*PF 0,715 n.s. PPL*CQ 0,002 +
Product placement Content enjoyment Following
Perceived fit Content Quality Product placement
Self or third party +
Shares Comments
As can be seen in table 6, product placement has a significant negative effect on content enjoyment. The first hypothesis is thus proven to be true. This means that in a social media environment, product placement affects content enjoyment negatively, just as it does in various non-social media environments. However, this effect is not a simple a to b effect. When a product placement is an endorsement for a third party instead of for the celebrity him/herself the product placement was evaluated as more positive. This is the opposite of what was expected. It could be that third-party placements were not perceived as significantly different posts by the likers compared to placements about the self and some not analysed characteristics of the posts itself caused the likes to be higher when the post is about a third party. A scenario in which paying sponsors demand the best timeslots for a celebrity would not be an unlikely one. Neither unlikely is a scenario in which sponsors have optimized their posts to achieve the most amounts of likes, comments and shares. Oddly, the obviosity of the product placement does not matter for the effect product placement has on the content enjoyment. An obvious placement is not enjoyed significantly more or less than a non-obvious placement in a social media environment. This might indicate that non-obviousness as seen in traditional media is wholly different than on social media. Product placement on social media does not take place within a continuous storyline that is the norm in media such as movies, it takes place in individual posts. This means that on social media it might not be integrated in such a way for obviousness to matter. However, when moderators where added, the obviousness did matter for the effect product placement has on the content enjoyment. This could signal a relation between the content quality and the obviousness of a post. However, the examination of this relation is beyond the scope of this study.
As expected, the perceived fit between the endorser and the posted content is important. This is in line with the matchup theory. However, for the amount of comments and shares, perceived fit is not significant anymore. The ELM of Petty and Cacioppo (1986) can help to
explain the situation. Commenting and sharing require more effort than clicking on a like button to like something does. So, it might be that the underlying processes of commenting and sharing might be a more central route process than liking is, causing perceived fit to be less important than other aspects of a post such as content quality. Content quality is important across all output categories, proving that the strong assumptions about the effect of content quality derived from the literature also hold in a social media environment. If the content is of poor quality, all output factors are significantly lower than they are when the content is of high quality. Content quality might be more important than perceived fit because of the amount of space taken by the content. Within a post, the content is taking up way more space than the name and profile picture of the endorser is, and might thus be evaluated more intensively.
The effect of perceived fit on the effect product placement has on the content enjoyment is not significant, nor is this effect present for comments and shares. This means that when a product placement is present, the perceived fit between that post and the celebrity does not matter. The fact that this is contrary to what literature would predict, might indicate that there is a difference between social media and traditional media. This needs to be researched further, as it can provide interesting insights in the processes that make product placement work the way it does in the social media environment. An interaction however is present between product placement and content quality on the content enjoyment, comments and shares. This significant positive relation confirms the importance of content quality. When the content quality was higher, the negative effects of product placement on content enjoyment were way less negative than they were when content quality was lower. However, the coefficient of the positive effect of content quality was still larger when product placement was not present compared to when it was, again stressing the negative effects of product placement on the content enjoyment, comments and shares.
Long-term effects
Table 7. Summary of effects and direction for the long-term analysis
Output Independent variable Significance level Direction
Followers on page Product placement 0,040 -
Product placement Obvious 0,089 n.s. Product placement Third party 0,490 n.s. Perceived fit 0,320 n.s. Content quality 0,532 n.s. Content enjoyment < 0,001 + Comments 0,002 - Shares 0,252 n.s. PPL*PF 0,801 n.s. PPL*CQ 0,532 n.s.
As expected, product placements have a significant direct negative effect on the number of followers of a page. The more product placements a celebrity posts, the lower the number of followers is when compared to the previous scrapemoment. This does not always mean that product placements have an absolute negative effect on the total number of followers, but it does mean that they have a negative effect on the growth trend. Product placements can cause the growth trend to stagnate, make the growth disappear or even turn it into a decline. It did not matter significantly for the number of followers if the celebrities endorsed themselves or third parties in their product placements, nor did it matter if the product placement was obvious or not. This might indicate that followers find product placements annoying, no matter what. Content quality and perceived fit do not have a significant effect on the number of followers. This might be caused by the fact that the constructs content quality and perceived fit are very post focused and in effect do not move beyond the posts itself. Therefore they do not matter directly for the number of followers of a page. Product placement posts have direct negative consequences, however these consequences do not carry on for long. Whilst followers lost are probably not acquired again soon, the effect of the product placement on the number of followers ceases to have a significant negative effect
after one scrapemoment.
The output variables of the short-term analysis, namely content enjoyment, comments and shares also have effects on the following themselves. Whilst the amount shares do not have a significant effect on the number of followers of a page, the content enjoyment and comments do have a significant effect. The content enjoyment has a positive significant effect on the number of followers, every like causes the number of followers to increase a small bit. Because the number of followers is obviously measured in round numbers, this effect cannot be observed per individual like, but as a group they definitely do have an effect. The amount of comments on the contrary, has a negative effect on the number of followers. This is odd. When looking at the viral nature of social media and the traction comments cause, a similar effect to likes would be expected. However, comments can also be used to express dislike or disgust, contrary to the unilateral construct that is a like. When a post gets a lot of comments, this might thus mean that a lot of people are expressing their negative feelings. On the other hand, a recent research by Burke and Develin (2016) found that whilst positive posts receive more likes, negative posts receive more comments. This might explain the negative effects of the amount of comments on following. If a celebrity constantly places negative posts, the followers might be inclined to remove the negativity from their lives by unfollowing, especially because they do not have a real personal bond with the celebrity. Another explanation might be the recent advent of people tagging their friends in posts. Whilst it has not been researched yet, there definitely is some negative feedback about this constant tagging of friends. Especially for larger Facebook pages, it can be lots of irrelevant names and no actual meaningful content. This might add annoyance and lead to unfollowing in itself.
The direct long-term effects are complemented by their term counterparts. The short-term direct negative effects of product placement on content enjoyment and comments cause
Product placement thus has a multitude of negative effects on important output factors in the social media environment. Some of these negative effects can be made weaker by strengthening the content quality, however the effect always stays visible in the output factors.
Managerial implications
From this research, it becomes clear that product placement has a negative effect on both the amount of traction a post generates in the form of likes, comments and shares, as on the number of followers. The average post that contains product placement, compared to when it does not contain product placement, causes the likes to go down by 114.114 (77,6%), the comments by 1.600 (87,7%) and the amount of shares to drop below zero, with 4.491 (125,3%) less shares if the average post contains a product placement. On the long-term, every case of product placement causes the followers to decline with 4755 (0,016%). However, as likes and comments also have significant effects on the long-term the product placement also proceeds as an indirect effect on the number of followers. Via likes, product placement causes a decline of 3652 (0,012%) followers, whilst the decline in amount of comments lead to an increase of 1950 (0,007%) followers. Does this mean that marketers should stop using product placement on social media altogether? No. Marketers however should be cautious when using product placement, and make sure the quality of the content is very high. Otherwise they might cause their promoters to lose their following and effect, and perhaps even cause subsequent negative effects to the company itself. Influencers should be more wary than marketers should be, it is their traction and following that is on the line, and not the companies following. Influencers should focus on not just being used as promoters, and post their own content as well to counteract some of the negative effects. Content quality is still an important factor in the social media environment, just as it is in traditional media. Whilst perceived fit is not a statistically significant influence factor for the amount of comments and shares, it is for the content enjoyment expressed as likes. Because the role of
matchup has been proved time after time, fit should not be thrown overboard if the aim of a post is more on comments and shares than likes. Likes is a unilateral construct, only going up when it is positive, whilst comments and shares allow for other messages to occur. When the fit is low, a company might be ridiculed in shares or comments, causing negative effects on factors important for the company, whilst not having a negative effect on the amount of shares and comments. All in all, marketers should be careful when applying product placement and influencers should be wary of product placements on their pages. Whilst the negative effect does not carry on for long, followers lost are probably not reacquired anytime soon. If marketers or influencers choose to apply product placement, they should vary product placements with non-product placements posts and they should focus on keeping the quality of the content as high as possible. The content quality negates some of the negative effects of product placements, both on the short and the long-term, however it can never negate all of the negative effects. Effectively, the negative effect of product placement can’t be negated by improving characteristics of the product placement post itself.
Limitations
Whilst the sample size, internal and external validity of this study were generally considered reasonable, there are some limitations to the data that should be improved when doing future research. The data for the long-term was aggregated per scrapemoment. This caused an average effect for the posts per scrapemoment, instead of individual long-term effects. In future research the scrape frequency should be increased to be able to assign a maximum of 1 post per scrapemoment. Not only does this make the effects more specific and clear, it also makes the data analysis easier to conduct and interpret and allows for a more encompassing analysis. To improve the generalizability, more celebrities from more countries could be added. Only well-known western celebrities were added. Whilst these western celebrities are probably known around the world, this might not be distributed evenly around the globe.
Also, when adding more celebrities, differences between different categories of celebrities (e.g. movie stars, musicians, public figure etc.) or the type of content (photo, status, video, etc.) can be made. The coding might be improved by adding multiple well-trained coders that are well-read into the subject. This might make the coder bias even smaller.
Directions for future research
As product placement in social media environments on content enjoyment and following was more or less an open field in the research literature, lots of concepts can be researched further. The fundamental difference in storyline integration of product placements between social media posts and continuous media is an interesting topic to be researched. A lot of research topics can be found in the area of underlying motivations and reasons for liking, commenting and sharing. These motivations might explain some of the more odd results found in this study. Another interesting aspect of the social medium Facebook, is the recent integration of reactions additionally to the unilateral construct ‘like’. The quick reaction that is a like has been split into the standard like with an addition of 6 emotional responses, ranging from very positive to very negative. Peripheral negative effects can also be measured now. In general, it is interesting to take the sentiment into account when doing a research on social media. Sentiments can explore underlying reasons with more depth than counts can. As the role of perceived fit between posts and celebrities seems to be not all that important on social media, contrary to what is predicted by literature, this is an interesting topic to do further research on. In addition, the distinction to which people are aware of the presence of product placement in posts is a concept that could use further study.
Final remarks
I am aware that more insights could have been taken from the collected datasets, however I chose not to do so, as these insights were not very relevant for the main research question. For this research no funding was received from any external entities.
