Discussion

Virtually all tobacco companies have a presence online with most maintaining brand websites, and many even offer their products for sale directly to consumers. These tend to portray tobacco use in a positive light and increasingly engage users with interactive features.19 Many also possess a youth orientation as shown in an analysis of e-cigarette websites with youth targeting described via images or claims of modernity (73%), increased social status (44%), enhanced social activity (32%), romance (31%) and use by celebrities (22%).20

In recent years, social media advertising has become a progressively more prevalent tool used by tobacco marketers.21 22 An analysis of 365 e-cigarette videos on YouTube found that 85% were sponsored by e-cigarette brands and these were viewed by 1.2 million youth under <18.23 In a study of nearly a half a million Texas middle and high school students, 52.5% reported exposure to tobacco-related social media.24 Exposure to tobacco promotion on social media among youth aged 11–18 was correlated with a more favourable attitude towards tobacco.25 A study of undergraduate college students between the ages of 18 and 25 years found that greater prevalence of e-cigarette use in an individual’s social network was associated with a more positive ‘smoking experience’.26 One study found that youth exposure to tobacco depictions on social media better predicts smoking tendency than exposure on television and movies.27 While this observation needs to be verified by others, it is remarkable given the abundant research demonstrating the potent influence of television and movies on smoking behaviour.

Active engagement with online tobacco material (eg, watching video, subscribing to newsletter, Facebook likes, retweets) was found in 2.9 million youth aged 12–17.28 Higher levels of online engagement were associated with greater susceptibility to use a tobacco product among tobacco never users as well as current and past tobacco users. A longitudinal study of 1742 never e-cigarette users reported that exposure to e-cigarette advertisements on social media had an effect on subsequent use of e-cigarettes.29 Social media has also been exploited in marketing alcoholic beverages with a transition in emphasis from exposure to engagement noted over recent years. An examination of the marketing of 20 alcohol brands on Facebook over a 2-year period found that the number of ‘fans’ increased by 52%.30 Another study of alcohol brand activity on Facebook found that companies engaged consumers in the co-creation of content that contributed to the brand’s marketing objectives.31

Evidence shows that the tobacco industry exploits social media as a means of circumventing advertising restrictions. In countries such as Australia where tobacco advertising to consumers is prohibited, Facebook has been shown to be a leading vehicle for tobacco marketers to circumvent the intended ban.32 Tobacco marketers sometimes surreptitiously pose as users on social media platforms. Employees of British American Tobacco created fan pages for their products, joined pages as fans and posted pictures of their products on Facebook.33 As barriers to regulatory solutions of youth online/social media tobacco exposure are substantial, focus upon enriching antismoking messages via these channels may offer the most effective countermeasure.34

The key observation in our study is that tobacco marketing is highly prevalent on Facebook and many such activities appear to be at variance with Facebook policy. Brand-sponsored pages by tobacco marketers are widespread on Facebook and these often include offers to sell tobacco products. Many also allow access to prohibited content by underage users. The brand-sponsored tobacco pages are dominated by e-cigarettes, hookah and cigars with the notable absence of traditional cigarettes. Images of tobacco products were nearly universal while a purchase link (eg, ‘shop now’) and/or sale promotion (eg, discount coupon) were quite frequent. While traditional cigarette brands did not maintain brand-sponsored pages, the Facebook pages of online tobacco vendors routinely sold cigarettes including leading brands such as Marlboro, Camel, Newport and American Spirit.

Because tobacco and alcohol marketers are prohibited by Facebook policy from paid promotion (eg, advertising, boosting) brands increasingly pursue an ‘organic’ and ‘conversational’ content strategy to engage users.30 Ongoing postings of fresh material on their brand-sponsored pages such as memes, sponsored events, product innovations and contests attract ‘likes’ with the goal of creating communities with shared interest in their wares who will snowball their message to friends. Such material is targeted to demographics most likely to use their products (eg, teens and young adults, adult smokers). The large number of likes on tobacco brand-sponsored Facebook pages observed in our study illustrates the successful implementation of this strategy.

Facebook’s policy requires age gates excluding those <18 for pages offering tobacco products for sale. Our study showed that age gates were absent from a large proportion of pages for e-cigarette and hookah vendors as well as online vendors of numerous tobacco products. While the legal age for tobacco sales in Europe and most US states is 18, in California (where this study was conducted) and four other US states (New Jersey, Oregon, Hawaii, Maine) the age is 21. With Facebook’s geolocation ability, it would be reasonable for their policy to conform to location regulations.

To police its content, Facebook relies on multiple methods, including automated algorithms, human screeners and user complaints. With some 2 billion users and enormous volumes of postings each day, Facebook has a daunting task in enforcing its policies.35 Our results show that Facebook policing of its tobacco policies is notably porous with a large volume of promotional materials escaping scrutiny. By searching Facebook’s ‘Help Community’ feature for comments relating to tobacco, it is clear from objections raised by would-be tobacco marketers that, at least in some instances, tobacco-related material is being excluded.11 The results of our study indicate that current screening methods are largely ineffective. Numerous comments on the Community Help feature seek to alert Facebook about the presence of tobacco material. These notifications are referred to an online feature to report a possible violation of Facebook policies. It seems clear that Facebook’s current method of inhibiting tobacco promotion relies excessively on people reporting tobacco promotion. It is of interest that the Help Community includes numerous protests from e-cigarette advocates arguing that promotion of vapour products should be permitted. Petition drives have been organised in an effort to convince Facebook to reverse its e-cigarette prohibition.36

This study has several limitations. Identification of tobacco brands via the Nielsen database, which tracks sales in convenience store channels, favours major national brands over smaller regional and online only ones. Limitations of our interface with Facebook imply that we may not have revealed some aspects of tobacco promotion on the platform. Our study is confined to public posts and may not detect ‘dark posts’ which are unpublished posts which do not appear on the brand’s page, but may be selectively configured to customise their message according to individual Facebook user’s characteristics. As the four profiles we created for this study frequented a large number of tobacco-related Facebook pages, it is probable that they were profiled to an unknown extent and thus the material we collected may have included targeted material. In assessing the popularity of tobacco brand-sponsored pages, we could not determine the potential effect of brand influencers who are often ‘social stars’ with a large number of followers. The age gating data may not represent actual user age as many underage users create profiles older than their actual age.37 This means that the proportion of underage youth being exposed is likely higher than our data would indicate. The four Facebook profiles created for our study covered a number of ages (13, 17, 18, 21) but all were male and located within California. Because Facebook tobacco pages may have content targeted by user characteristics such as gender, location (national, urban vs rural, ethnic predominance) and previous social media use patterns demographically different users may have encountered a different spectrum of tobacco promotion.

An alternative approach to our methodology of tracing tobacco promotion from the perspective of the individual user would be measuring interaction among large groups of users of varying interest. This would better capture relationships which flow from the algorithmic and participatory nature of social media.38 An automated, large-scale analysis of a substantial data set of user interactions among tobacco social media revealed that tobacco-related content grows exponentially while there is linear growth for the tobacco control group.21 In addition, the antitobacco community is more tightly connected, while the much larger protobacco community is sparsely connected.

Much attention has focused recently on Facebook’s efforts to manage inappropriate content such as hate speech and fake news. Following a number of highly publicised offensive postings, Facebook announced in early 2017 plans to expand its content management team by 3000 people, to a total of 7500 people.39 As identification of hate speech often necessitates human judgement, this makes it difficult to use automated algorithms.40 As of 2017, Facebook removes some 66 000 hate speech posts each week.41

By comparison to identification of rhetorically ambiguous concepts such as hate speech, which necessitate human judgement, enforcing Facebook’s tobacco policies would seem a much more manageable task. It seems reasonable that Facebook could invest engineering and/or human content managerial resources in developing an automated screening algorithm to detect tobacco product brand-sponsored pages. These pages contain readily identifiable features, both text and graphic, revealing their origin and intent. Distinguishing affinity groups (eg, those who happen to share a passion for cigars) from marketers is more challenging. To make such distinctions, machine learning algorithms working in partnership with human reviewers would be needed to improve vetting. Automated screening is especially important as it is not difficult for vendors to create new Facebook identities and use them to relaunch their banned content.

While Facebook’s tobacco policies are comprehensive in scope, they are at times ambiguous and occasionally contradictory. Nevertheless, it is laudable that they have policies intended to interdict tobacco promotion throughout its platform. Voluntary efforts by social media platforms are especially valuable given the regulatory challenges of constraining online tobacco promotion.34 Facebook’s commitment to continuous improvement in sustaining their guidelines (‘We are always working to get better at evaluating this content and enforcing our standards’) reflects their core commitment to community.13 Our hope is that realising the extent to which tobacco promotion evades their efforts to restrict it will encourage Facebook to undertake an effort to establish more comprehensive enforcement of their well-intentioned policies.