Evaluating Transparency in Platform Political Advertising Policies
Political advertising is a key component of election-related content on social media platforms. However, while traditional forms of media such as television and radio advertising requirements are comparatively well-regulated by the Federal Election Commission (FEC), online advertising guidelines are less clear. The most recent substantial FEC guidance for online political advertising disclaimers was issued in 2006 and the commission, which currently lacks a quorum, is unlikely to change those practices anytime soon. As a result, the policies created by the largest first-party advertising platforms substitute for democratically debated rules.
Scholars at the the Center for Information, Technology, and Public Life (CITAP) have conducted extensive research into social media platform’s ad policies, raising the point that platforms such as Facebook, Twitter, Google, and Snapchat have already “gone far beyond what is required of them by federal law” in terms of their public ad library archives. That said, as CITAP and other scholars argue, there are still sizable barriers to studying and understanding the policies that govern paid political speech on social media platforms.
This blog post builds on the foundational work of other scholars to focus specifically on three key components of digital political advertising: the review process for political ads, platforms’ transparency for information about those ads, and unique loopholes for different types of speakers running political ads. Regarding the last component, recent investigations by EIP have found loopholes in Google’s advertising policy that are cause for concern. For example, news organizations, regardless of whether their ad content is election related, are exempt from being considered election-related ads and thus their content is not incorporated into Google’s ads transparency report, which operates as its ads library. We end with a series of recommendations to help address some gaps we highlight in our analysis.
Key Takeaways
We examined the policies of 17 different first-party platforms. Of these platforms, 7 allow political advertising: Facebook, Google, Snapchat, Reddit, Hulu, Roku, and Pandora. The remaining 9 platforms do not allow political advertising: Twitter, Pinterest, TikTok, Gab, Discord, WhatsApp, Telegram, Spotify, NextDoor, and LinkedIn.
Platforms need to provide more transparency and accessibility for outside review of their political advertising. As it stands, it is difficult to assess the impact or reach of political ads on any given platform, and no platform provides what we would consider sufficient transparency to enable a reliable understanding of the political ad ecosystem and its impact. It is critical for both researchers and the public to have access to detailed information to more clearly understand this increasingly popular medium for political advertising.
There are two interesting loopholes related to speaker types in online political ads: first, Facebook / Instagram and Google do not include paid posts by content creators in their ad libraries. Second, Google exempts media organizations from their definition of election advertisements, resulting in their ads not being displayed in Google’s ad library.
What is a Political Ad?
When analyzing the platforms’ ad policies, it is important to remember there are many ways to define what is political, and therefore what is political advertising. The FEC has different subsets for political advertising based on the type of entity running the ad, political content, and communication channel. Each subset has different reporting and regulation requirements at both the federal and state level.
The relevant definition of political advertising for platforms is what the FEC defines as public communication, which refers to ads on mediums including, “communication placed for a fee on another person’s website.” Public communication includes “electioneering communications,” which refers to an advertisement that clearly refers to a candidate for federal office. Platform definitions of political advertising generally fall in line with the FEC’s definition. Certain platforms such as Facebook, Reddit and Snapchat expand beyond the FEC definition to include social issue ads under their political ad policy. Google, however, does not include social issue ads. Instead, Google defines political ads as ads that feature federal and state candidates, parties, and ads about specific ballot measures.
Platforms make political ads available in their Ad Libraries, so this designation matters. As Fowler et al. highlight in their chapter on political advertising in Social Media and Democracy: The State of the Field and Prospects for Reform, the FEC has very poor standards for itemizing how money is spent on political advertising and therefore we don’t have a good way of understanding all the ads campaigns are running. Additionally, the scholars argue that there is a clear reporting gap for outside entities (i.e. not registered by the FEC) running issue ads online. While Ad Libraries are not a substitution for federally reported advertisement spending, they provide the public a more detailed look at how campaigns are spending money on paid online political speech. Therefore, what is made available in these ad libraries is critical.
Comparing Ad Scope, Review Process, and Non-Permissible Content
There are many components to a political ad: targeting capabilities, permissible content, reactions to state laws, and verification and authorization of advertisers can be found in CITAP’s helpful comparison table. Adding to this analysis, our chart below examines the process for reviewing political ads and what type of content is not allowed. Our last column explicitly looks at whether platforms will ban early claims of victory in their ads.
Table 1: The evaluation of the platforms’ policies is informed by Facebook’s Ad Policies, About Social Issues, and Fact Checking Program Policies pages, Google’s Political Content, About the ad approval process, Advertising Policy on Misrepresentation, and political ads policy pages, Snapchat’s Advertising Policies, Reddit’s Advertising Policy, Hulu’s Ad policy, Roku Ad Policy Pandora Ad Policy.
Table 1 shows that while most of the scope for political ads is well defined, those for streaming services are not: Hulu, Roku, and Pandora do not define what a political ad is so it is unclear if ads falling under this category would meet FEC standards. Hulu’s ad policy states that the advertiser, not Hulu, “is responsible for ensuring that Ad(s) comply with all applicable laws, regulations, and industry guidelines.” Pandora’s policies are also vague, stating that political ads may have additional required targeting and/or content restrictions, but do not explain what those might be or what type of political content that will apply to.
Another important takeaway from this chart is that while all platforms do not allow misleading or false content in its political ads, Facebook and Google explicitly do not allow ads that falsely claim the results of the election before they are called. As of September 23, 2020 Facebook announced that it will, “reject political ads that claim victory before the results of the 2020 election have been declared.” A few months prior, in November 2019, Google added a similar claim about election results to the examples of content that it does not allow, including demonstrably false information about “election results.” After reviewing the other platforms’ policies, we found that the platforms that accept political ads might take down this type of content due to their existing policies on misleading and false content. However, we argue that platforms should clarify if this type of content will not be allowed in political ads.
Comparing Platform Transparency Policies
In response to growing public criticism and demand for transparency, major social media platforms have launched advertisement transparency libraries. These libraries generally show a history of political and issue advertisements run by all groups, including the advertisement as it would be presented on the platform. While most social media platforms have launched such libraries in recent years, the level of detail provided and data accessibility varies widely by platform, including who can access the data, how you can access the data, and what type of information is in the library.
Table 2: The evaluation of the transparency policies by platform was informed by their advertising policies and Ad Library portals linked here: Facebook Ad Library, Google Transparency Report, Twitter Ads Transparency Center, Snapchat Ads Library, Reddit Political Ads Transparency Community, Hulu, Roku Ad Policy, and Pandora Ad Policy.
Why it’s important to have an export of ad creatives in a machine-readable format: Political advertisement libraries are most useful if researchers can export advertisements and study them programmatically at scale. Facebook’s Ad Library, for instance, has received sharp criticism from Mozilla researchers, as Facebook does not offer a simple means to export all advertisements. Other platforms received “partial” designations as ad creatives are linked in exports but not immediately accessible, and must be manually downloaded or scraped. Only Twitter and Roku offer exports with a full archive of the advertisement run.
Archive of taken-down advertisements: Advertisements that are taken down by platforms are particularly valuable to researchers and the public, as they provide insight into potentially untruthful claims made by politicians or other advertisers. While platforms like Facebook and Snapchat do make available ads even if they are taken down, others such as Google only display a warning that the advertisement violated the platform’s policies, making it impossible to view the original advertisement.
Political Ads Policy Based on Speaker
One unresolved dimension of political ads that has come up in our ongoing monitoring is how the type of speaker intersects with the policy. Mike Bloomberg’s paid-influencer meme campaign and a surge in ads by the news outlet “Epoch Times,” show that entities not typically considered political actors can push election-related content.
There is not much guidance from the FEC on how to deal with entities not registered to the Commission. Known political entities, such as Federal candidates, political parties, Super PACs, and registered political committees that spend or raise a certain amount have legal requirements to report ad expenditure, and ad disclosure. In the chart below, we evaluate whether platform ad policies address these different types of entities.
Content Creators and Branded Content
Speech from content creators mostly applies to user-generated content platforms such as Facebook and Instagram, YouTube, and Snapchat. On Facebook and Instagram, creators can post “branded content” or content that “features or is influenced by a business partner for an exchange of value.” The content becomes a Facebook or Instagram ad once the business partner makes an ad featuring the content creator’s material.
Figure 1: A screenshot taken September 23, 2020 of a sponsored post run by Mike Bloomberg. The Instagram account, grapejuiceboys has 3.1 million followers, therefore this ad has the potential to reach a large audience. However, this ad and others like it, do not show up in Facebook’s Ad Library.
Branded content as a separate content category was created after Mike Bloomberg’s advertisement campaign with prominent accounts on Instagram. Bloomberg’s practice was highly criticized but has since fallen out of the news cycle. Yet there are still important loopholes within this advertising model that have not been addressed. Branded content on its own does not end up in the Ad Library, because it is technically not an ad (meaning Facebook does not make money on sponsored content). Instead, its inclusion is contingent on the business partner creating an advertisement with the content. Branded content must have the “sponsored by” disclaimer, but sponsored content does not end up in the Ad Library.
Branded content for presidential candidates is visible through CrowdTangle, a tool primarily used by researchers to study public accounts on Facebook and Instagram. While this transparency covers two very important campaigns from Joe Biden and President Donald Trump, we cannot trace sponsored political content from other sources who may be influencing political discourse through branded content. Not being able to track this content in the Ad Library means we miss other important details such as ad spend and targeting. Researchers might miss political content being pushed through sponsored content.
Similarly on YouTube, content creators can run paid promotions on their channels, which include paid product placements, sponsorships, and endorsements. Paid promotions must follow YouTube’s ad policies, but nothing exempts political content from being promoted through this medium. Paid promotions also seem to not be included in Google’s transparency report. As on Facebook and Instagram, this leaves a loophole for paid political speech to go unflagged.
Google’s News Organization Exemption
Media properties offer another loophole. Our team has identified multiple cases of misleading advertisements on Google created by media outlets. On September 10, we wrote about multiple misleading headlines from the Washington Times that cast doubt on mail-in voting. More recently, on September 21, we wrote about an ad run by a non-profit using a headline from a debunked article about voter fraud in Florida. These advertisements do not show up in Google’s transparency report. This is because political ads from news organizations are not included in the company’s definition of an election ad.
Until recently, the Epoch Times, another news outlet flagged for sharing conspiratorial content, was excluded from Google’s transparency report. In August 2019, Facebook banned ads from Epoch Times after “trying to evade its advertising transparency rules,” as the New York Times reports. (It should be noted that this “evasion” was through the sponsored content model on Facebook.) The media outlet shifted its advertising to YouTube, and is running promoted tweets on Twitter.
Now that we are able to view the Epoch Times’ ads in aggregate, it is easier to see what type of political ads the news organization is running. As of September 24, 2020, we found that 88 out of 848 ads or about 10% of ads violated Google’s political ad policy. While we are unable to see what the content was that violated YouTube’s ad policy, researchers and YouTube users now know more about this media outlet and the veracity of information it is sharing through its ads.
Figure 2: A screenshot taken September 24, 2020 from Google’s Transparency Report on ads shared by the Epoch Times. Epoch Times’ inclusion in Google’s Ad Library is an example of why allowing public access to ads run by media properties is helpful in understanding the type of content pushed by the media outlet.
Policy Recommendations
As shown in Table 2, platforms vary widely on the type and accessibility of the data published (if any). However, allowing researchers and users to access this data is critical to holding platforms accountable and ensuring advertisers aren’t violating these policies. Providing more information to users about how they are targeted helps create a more informed public.
We advocate that platforms develop advertisement transparency libraries that:
Allow exporting all advertisements in a standard machine-readable format. Such an export should either include full ad creatives in the export, or link to machine-readable forms of the creative (e.g., a raw image or video rather than a separate webpage)
Allow access to all advertisements that ran on the platform, even if they violate the platform’s advertising policies
Expand the definition of political advertising to include ads about the election and tabulation process
Include detailed targeting information, so that researchers can understand the potential impact of any given ad
The challenge of defining what constitutes a political advertisement has led to inconsistency across platforms and some platforms reacting two slowly to rapidly developing narratives. The fundamental challenge here suggests that a more appropriate, long-term solution would be mandated transparency for all online advertisements without regard for topic. Such a decision would be highly controversial among the advertisers who comprise the customers of these platforms, making a legislative solution by democratic bodies necessary to enforce compliance.
Contributors to this post: Jack Cable, Carly Miller, Shelby Perkins, Stanford Internet Observatory