Social Media Virtual Special Issue Journal of Advertising

Journal of Advertising

Social Media Advertising and eWOM from 2011 to present

Guest Editor: Louisa Ha, Bowling Green State University

Social media are the fastest growing advertising media worldwide.  Just Facebook’s 2016 global advertising revenue is already 27 billion U.S. dollars, a 58% growth from 2015 (Statista, 2017). Despite social media’s heavy reliance of advertising as revenue source, the study of advertising on social media is still scarce compared to other media.  Partly it is due to the difficulty of measuring the ads because they are customized and programmed based on individual attributes.  Another problem is that social media ads are frequently in native format, similar to the social media content, and vary in forms from sponsored posts (Facebook) to promoted tweets (Twitter) to sponsored filters and stories (Snapchat).  

Journal of Advertising (JA) has published articles addressing social media both as an advertising medium and as a non-paid electronic word-of-mouth (eWOM) medium.  This virtual theme collection selected 10 articles that are useful reference and for future research in social media advertising and e-WOM published 2011-2017. 

First we examine articles that study social media advertising as advertiser controlled advertising messages.  Wojdynski & Evans’ (2016) eye-tracking experiment found very few participants can distinguish native ads from editorial content.  They also found a position effect that a middle-positioned disclosure attracts greater visual attention and likelihood of fixation compared to top- and bottom-positioned disclosures which have been believed to have stronger attention. Although recognizing advertising disclosure did affect negatively on the credibility of the native ad as a news story, but in general, most people did not know the content was sponsored.  The persuasive effect is also low for sponsored content. Their study results indicate the need to develop sponsor disclosure standards based on empirical evidence to avoid lowering the credibility of the media.   

The viral or easy to spread and share aspect of social media has attracted several JA scholars to examine social media advertising from a viral advertising perspective. Cho, Huh and Faber (2014) used a field experiment to examine the influence of sender trust (interpersonal relationship trust) and advertiser trust (mediated and calculated trust) on four stages of e-mail advertising effects of a tax preparer service in the United States.  The four stages of e-mail advertising effects are 1) attention to e-mail, 2) pre-exposure perception of the ad (informativeness, entertainment, irritation and perceived risk, 3) voluntary exposure to e-mail, 4) attitude toward e-mail content and attitude toward the brand. They found an overwhelming dominating effect of sender trust across all measures of advertising effects, and advertiser trust only affects pre-exposure perception.   Perceived entertainment and risk are important factor in determining ad exposure.  So the viral effect of advertising is highly dependent on the relationship of the sender to the recipient especially at the attention and exposure stage.

Vanwesenbeeck, Walrave and Ponnet's (2016) experiment specifically examined how parental mediation style may affect young adolescent’s response to social network game advertising.  Their study in Belgium confirmed that autonomy parental mediation style is more effective than controlling parental mediation style in developing persuasion knowledge of young people ages 10-14.  Inconsistent parental mediation style is the worst in outcome with reactance from young people. Only when young people have critical attitude toward advertising will they be less persuaded by the social network game advertising. Persuasion knowledge does not discourage request for advertised products embedded in the games.

Most articles published in JA on social media focused on the user-generated content characteristic of social media and saw them as a mix of interpersonal and media communication.  They viewed social media content as eWOM generated by consumers.  Jin and Phua (2014) conducted two between-subject lab experiments to demonstrate the effects of different attributes of celebrities’ tweets on brands as influential eWOM on their followers.  Both experiments used semi-real celebrities and the researchers found that celebrities’ tweets influence their purchase intention and product involvement.  The number of followers increases the credibility of the celebrities on product purchase intention but negatively influences the subjects’ intention to spread the message.  When the message is negative, the subjects are also more likely to spread the word.  Social identification is a strong mediator of such effect.  So it’s important that the user identifies with the celebrity and the celebrity is perceived as positive role model to the users in using celebrities to endorse a product as testimonials. It adds to the celebrity endorsement research on how celebrity effects works on Twitter.

Minton et al.’s (2012) three-country survey of consumers shows how cultures influence social media use and motivations for sustainable behaviors such as recycling, organic food purchase and donation to green charities. They found that collectivist culture such as South Korea has the highest use of social media than Germany and the U.S.  Responsibility motives for sustainable behaviors were similar across the three markets but involvement motives are country specific.  The study also shows the difference between Facebook and Twitter in complexity of consumer interaction and the different motives for sustainable behavior:  Twitter users have higher responsibility motives while Facebook users have higher involvement motives.

Both Campbell et al. (2011) and Lawrence, Fournier and Brunel’s (2013) studied a specific form of social media advertising: consumer-generated advertising (CGA), which consumers create ad-like brand-focused messages with the intention of informing, persuading and reminding others. While Campbell et al. (2011) used only a data-mining tool Leximancer to quantitatively examine consumers’ responses to the four specific CGAs and came up with the typology along the two dimensions of Conceptual-Emotional and Collaborative – Oppositionary. Lawrence, Fournier and Brunel’s (2013) study combined qualitative thematic analysis, two experiments and a survey to confirm the main benefit of CGA.  The CGA comments can be positive or negative as per the typology of Campbell et al. (2011): 1) The inquiry, 2) The Laudation, 3) The Debate and 4) The Flame.  Interestingly, both studies found comments around those CGA ads are mostly not about the brand, but the consumer as the ad creator and viewpoints shown in the ads. Lawrence, Fournier and Brunel’s (2013) confirmed the main benefits of CGA are the perceived trustworthiness of the creator as independent from the advertiser and the consumers’ engagement with the creator and the ad, not because of consumers’ identification with the creator.     

Fernado, Suganthi and Sivakumaran (2014) used greenwashed ads with questionable environment claims as a specific case to examine if blogs by online news organizations and environmental protection NGOs set the agenda for specialized greenwashing blog users in the U.S.  The frequency of greenwashing blog discussion is the outcome measure of agenda-setting effect. The study first used thematic analysis to determine coding themes for the subsequent content analysis.  Testing the blog postings with cross-lagged correlation, they were able to find that U.S. expert consumer forum Greenwashingindex.com mostly set the agenda for Greenpeace, the NGO, and the Guardian, the British newspaper; while US news organization New York Times and Greenwashingindex.com mutually set the agenda for each other.  Green advertisers can learn the process of how their ads can be perceived as greenwashed and how skepticism is diffused to different stakeholders and to take proactive measures to prevent it.

Liu, Burns and Hou (2017) utilized a big data approach to examine the 1.7 million tweets as eWOM of top four brands owned by S & P 500 companies in five industries that people use daily.  It integrates the Latent Dirichlet Allocation (LDA) topic modeling data mining technique and Sentiment Analysis to create a framework that are more than descriptive but diagnostic in nature.  It identifies issues and concerns of specific brands as well as merits of specific brands that help in generating insights and directions for brand management.   The study emphasizes the importance of data preparation in transforming unstructured texts to meaningful categories using machine learning approaches and demonstrates a rigorous data cleaning method to remove spam tweets, URLs, non-essential grammatical elements and stop words. Their study show important difference in consumer concerns for different industries.  Their results show the telecommunications industry has the most negative eWOM on Twitter on their customer service. They validated the results with other sources of information about the telecommunications industry and industry experts.  Although negative comments are important, the authors also emphasize positive comments as the best practices of the brands that managers should capture as well.

Because trust is an important concept in social media, Roy et al. (2017) developed an algorithm called Trust in Social Media Score (TSM) which combines survey data and computational data to measure trust in social media. It can be used to identify who are the most trusting (likely to trust others in social media) and the most trustworthy (most followed/trusted) in social network sites.  The study tested the TSM’s computation efficiency and precision in three types of social network sites:  1) social networks sites that do not require pre-existing network such as Twitter, 2) social network sites that offer users to comment but does not expect users directly interacting with one another such as Epinion, and 3) multiplayer interactive online games such as EverQuest II.  In addition to measure trust based on computation data of those sites, TSM also weight-adjusted by the user’s trust-decision involvement as a contingency variable because the authors argue users give more trust to sites that they have high involvement than sites that they have lower involvement.  The level of the site involvement was obtained from survey data.  Although their survey data set was small and from students only, it illustrated the importance to add the user’s attitude input in big data research because not all sites are of equal importance to people.  The main contribution of this study is to propose the negative reinforcement property of trust — the importance of selective links as the proxy of trustworthiness and authoritativeness of the poster and offers an effective way to compute those scores.  Those who are followed by those that follow (link) many people but have very few followers are less authoritative than those followed by people who have selective links.  It can be used to identify who are most vulnerable in viral advertising and those who are likely to spread viral messages or even rumors.   It can also identify most trusted actors in the network to endorse advertising message virally.

These JA articles lay important groundwork for researchers to study social media advertising and eWOM.  More is needed to study how consumers perceive and respond to the different forms of social media advertising, and the interaction between traditional media, digital advertising and eWOM in social media.

Statista (2017). Facebook's advertising revenue worldwide from 2009 to 2016 (in million U.S. dollars)

Access articles chosen by guest editor Louisa Ha for FREE until August 31, 2017.