Tech Absolutism and Convenient Trade-Offs in Political Advertising

Last month, Twitter’s CEO Jack Dorsey announced the company would ban political advertising across the entire platform. He listed several reasons for the policy, including how political ads are detrimental to organic reach, can create risks in the voting process, and are purveyors of false or misleading information. The policy, which comes into force today, applies not only to candidate ads, but also issue ads – to a point. Dorsey argued that issue ads could allow ad buyers to circumvent Twitter’s objective: to ensure that reach is earned, not bought.

Although there was initially broad support for Twitter’s decision to ban political ads, concerns remain. The examples of legislative issues provided by Twitter, including taxation, gun control, social security, and trade, will never be fully exhaustive. In Facebook, the EU, and Election Integrity, I argued a similar point in the context of the European Parliamentary elections: that subjectively-defined issues will always vary depending on the country in question, given the vastly different legal and regulatory regimes involved:

This rollout extends beyond campaign ads by including issue ads: relevant, important, or highly-debated topics like get-out-the-vote campaigns, ballot initiatives, or referendums. To prevent foreign interference in the EU Parliamentary elections, Facebook requires that all political advertisers go through a country-specific authorization process wherein they submit documents that run technical checks to confirm identity and location.

One shortcoming of this framework is the seemingly arbitrary process according to which issues of national importance are chosen, and how to determine if the parameters are too restrictive or too broad. Facebook’s current list for the EU’s top issues – which it admits is subject to change – includes six topics: immigration, civil and social rights, political values, security and foreign policy, the economy, and environmental politics. In contrast, the equivalent list for the US contains more than twice as many issues of importance, going to show the opaque nature of how political issue ads are chosen and defined.

In Facebook’s case, the same core problem resurfaced: how to identify which ads are political. If the main issue is paid political reach, Twitter’s approach of restricting political action committees (PACs), Super PACs, and 501(c)(4)s from advertising on its platform seems like a sensible one. But if the objective is to solve broader societal challenges like misinformation and restoring civic discourse (apparently also a motivation for the decision), Twitter’s political ad ban is misguided at best. Corporations, nonprofit organizations, and other ‘apolitical’ actors will continue paying for reach and effecting long-term change over legislative outcomes.

That’s not to say there haven’t been efforts already to address the lack of transparency around political ads. In 2018, Twitter launched the Ads Transparency Center to provide its users with insights into ads, with details like spend, number of impressions, and targeting demographics. But issue ads fell under a separate policy. Vijaya Gadde, Twitter’s global lead for legal, policy, and trust and safety, argued this would allow Twitter to achieve a more “nuanced approach to transparency that is mindful of the inherent difference between political and issue-oriented advertising campaigns.”

It’s important to consider the extent to which Twitter’s political ad ban is a direct result from the company’s inability to deliver that nuanced approach. Before the ban, Twitter’s ad policy read as follows:

Political Content: Twitter permits political advertising, which includes political campaigning and issue advertising, but there may be additional country-level restrictions. In addition to Twitter Ads policies, all political content must comply with applicable laws regarding disclosure and content requirements, eligibility restrictions, and blackout dates for the countries where they advertise.

An advantage of the new policy for Twitter is that it will no longer have to police country-level restrictions. Instead, it can remove any content that “references a candidate, political party, elected or appointed government official, election, referendum, ballot measure, legislation, regulation, directive, or judicial outcome.” This definition is broad, vague, and ultimately bolsters Facebook’s argument: that it should be up to the government, and not social media executives, to set the parameters around content restrictions.

Issue Ad-dendum

One of the grey areas around Twitter’s political ad ban is the extent to which it would apply to issue ads, and what constitutes an issue ad. When prompted on this question, Gadde provided the following definition:

  1. Ads that refer to an election or a candidate, or
  2. Ads that advocate for or against legislative issues of national importance (such as: climate change, healthcare, immigration, national security, taxes)

There are naturally caveats to this. Ads supporting voter registration will be permitted, alongside what Twitter has called “cause-based advertising,” or issue ads by another name. These ads will be restricted and not prohibited, the argument goes, because they can be effective tools in the proliferation of civic discourse and outweigh many of the related challenges (microtargeting, misinformation, etc.). This was outlined on Twitter’s updated content policy page:

Twitter restricts the promotion of and requires advertiser certification for ads that educate, raise awareness, and/or call for people to take action in connection with civic engagement, economic growth, environmental stewardship, or social equity causes. We have made this decision based on the following two beliefs:

• Advertising should not be used to drive political, judicial, legislative, or regulatory outcomes; however, cause-based advertising can facilitate public conversation around important topics.

• Advertising that uses micro-targeting presents entirely new challenges to civic discourse that are not yet fully understood. 

This raises some thorny issues. First, if scientific concepts like climate change are defined as legislative issues, it could give weight to arguments denying its existence, creating a framework that moderates ads according to political disagreements in a specific country rather than factual contributions. This could lead to incumbency advantage by favoring representatives or existing legislation over their challengers and alternate proposals. Climate initiatives of a journalistic or multilateral nature could fall by the wayside in the U.S. while being acceptable in many other countries.

A second concern is where to draw the line around corporate ads. Twitter outlines the following restrictions around the advertisements of for-profit organizations: 1) they should not have the primary objective of “driving political, judicial, legislative, or regulatory outcomes” (whether this occurs incidentally is seemingly unimportant), and 2) advertisements must be tied to the publicly stated principles or values of the organization in question. For instance, Juul would be permitted to run ads in line with its mission to “improve the lives of the world’s one billion adult smokers by eliminating cigarettes,” so long as it doesn’t explicitly promote a piece of legislation that would, say, overturn a ban on e-cigarettes.

The third problem is enforcement – particularly in the U.S., where these issues are most salient. Twitter stated the following ways according to which it intends to ensure compliance with its new policy; from its Restricted Content Policies page:

  • The completion of its advertiser certification process, including identification and proof of being located in the U.S. (e.g. Employer Identification Number, mailing address, or government-issued ID) and additional requirements for handles to be consistent with a company or individual’s online presence.
  • Additional restrictions on targeting by geo, keyword, and interest: ZIP code level targeting will be prohibited, as will terms associated with political content, leanings, and affiliations (like liberal, conservative, and more). There is still a lack of clarity on how Twitter will police the context in which these words are used.

The restrictions on micro-targeting drew a sharp contrast with Facebook, where access to vast troves of data ensure that advertisers can be highly effective in targeting a specific segment of the population. According to Ellen Weintraub, the chair of the Federal Election Commission, the ideal policy would be a ban on micro-targeting, which would still allow political ads, “while deterring disinformation campaigns, restoring transparency and protecting the robust marketplace of ideas” (a similar system for restricting data strategies can be found in Germany).

But beyond the oft-repeated argument that political message reach should be earned and not bought, micro-targeting and political ads have simply not been a lucrative strategy for Twitter. Therein lies the actual problem with the company’s political ad ban: it represents a convenient trade-off with a status quo in which paid reach thrives, but does not resolve the core issues at the heart of Twitter’s decision. There are two reasons for this: 1) ads run by corporations and nonprofits can still prevent organic reach from being elevated across the platform; 2) ads from the public, private, and academic sectors, while not expressly political, could still reverberate across voting decisions and the collective consciousness.

Credit Score

Over the past few years, Twitter has been shifting its focus away from candidate campaigns and towards advocacy groups, federal agencies, and NGOs. This shift was the result of a lower “direct-response” rate of the political ads on Twitter’s platform compared to Facebook, where political ads typically lead to a higher conversion rate into campaign donations and email lists. Ned Segal, the CFO of Twitter, retorted that the ban was based on principle and not money, providing the example of ad spend for the 2018 midterm elections, which amounted to less than $3 million.

The connection is not evident. If anything, it illustrates how much less of a sacrifice it is for Twitter to ban political ads than platforms like Facebook. Ben Thompson, author and founder of tech newsletter Stratechery, argues that Twitter’s decision is a strategy credit, which he defines as “an uncomplicated decision that makes a company look good relative to [others] who face much more significant trade-offs.”

The full extent of the trade-offs Twitter faces is not yet clear. Its third quarter ad revenue was reported at $702 million (8% lower than analyst expectations). But with political ads representing a minuscule portion of the company’s ad revenue, and the PR generated by taking a seemingly principled stance, Dorsey’s decision seems like the epitome of a strategy credit. The timing was also convenient, with Twitter announcing the ban right as Facebook was about to report their third-quarter earnings.

Other critics argue that instead of banning political ads outright, platforms like Twitter or Google should instead evaluate the authenticity of claims being made; some lawmakers have, naturally, argued that an ideal solution would be to start fact-checking paid reach:

But norms around large companies fact-checking candidates or political groups tend to raise eyebrows. Axios framed Snapchat’s recent decision to dedicate a team to political fact-checking as putting it “at a middle ground between Twitter’s ban and Facebook’s highly-criticized policy of not fact-checking political ads at all.” That is inaccurate. Both companies’ approaches (a full ban and a hands-off approach) allow them to maintain a respective moral clarity by framing the issue differently: promoting organic reach vs. championing free expression. Anything in between is a lose-lose position, and lends credence to the argument that tech executives should not be the arbiters of content.

Framing the Problem

Alex Stamos, the former chief security officer at Facebook, says that while advertising is the most dangerous part of social media platforms, a full ban on political ads is not the solution to what ails them; from Columbia Journalism Review:

1. Tech platforms should absolutely not fact-check candidates’ organic (unpaid) speech

2. Tech platforms should have an open and transparent standard for facts in advertisements that gives them the least leeway possible to take down candidate speech

3. Tech platforms should enforce those rules as transparently as possible, preferably explaining why they made any given decision and laying out their reasoning (which should be precedent-setting)

4. It might be smart for #2 to be synchronized with the cable channels and other media who are making these decisions

Stamos goes on to recommend a legal floor on the advertising segment size for political ads – for instance, 10,000 people for a presidential election or 1,000 for a Congressional one. I agree with Stamos that the main risk of online ads (political or not) is not misinformation or the breakdown of civic discourse, but instead “the ability to target very small groups of people.” The legal floor on ad segment size would, on the one hand, diminish the effectiveness of false or highly targeted news coming from political ads. But it would also reduce the vast market for voter data across these platforms, which itself has led to several concerns related to privacy or election-meddling. 

Although legislative action on micro-targeting has been minimal, there has been some progress over the past few years. The Honest Ads Act, a bill sponsored by Senators Amy Klobuchar (D-MN) and Mark Warner (D-VA), would enforce disclaimers for digital ads with the objective of increasing transparency. It would require a public archive maintained by the Federal Election Commission (FEC) of election-related ads for candidates and legislative issues, and contain clear disclaimers revealing the individual or organization who paid for the ad. There are some shortcomings, however; from the Stanford Cyber Policy Center:

Currently, the most significant drawback of the Honest Ads Act is that the draft legislation places the critical responsibility of defining a political ad or an “issue of national legislative importance” entirely with the social media platforms themselves. Disclosure of issue advocacy represents a dramatic shift in the law, and it is too significant to trust private companies with defining which issues rise to the level of warranting advertising disclosure. As Facebook has moved in this direction, the firm has run into an array of line-drawing problems, including (1) addressing media organizations that boost news stories; (2) potentially designating charitable activity as political if, for example, its advertisements are related to health; or (3) managing product ads that touch on politics, such as a recent Nike ad featuring Colin Kaepernick, a Budweiser ad mentioning immigration, or an Amazon ad promoting a political book. Strong arguments could be made in favor of disclosure in all or just some of these cases, but such decisions should not depend on an individual company’s definition. […]

The second drawback of the proposed legislation concerns the disclosure of targeting information. While the Honest Ads Act is premised on a conception of targeting in which advertisers specify demographic categories and/or geographic regions, targeted online advertising has moved beyond categories of users to individual lists of users. The most sophisticated political consultants and parties now curate lists of individuals, along with email addresses to identify them, so as to send individualized messages to them. These lists are then turned over to Facebook and Google who promise to deliver the advertisement to a list of people (a “custom audience”) representing a large share of the targets. […]

Data regarding who was exposed to an ad is equally if not more important than targeting information. As targeting increasingly moves away from categories and towards individuals, advertisers or platforms cannot be expected to reveal the names of people who are targeted by the ad. “Exposure disclosure” should instead be required at a smaller level, such as zip code, census block, precinct level, or even at the county or district level. Talented enterprising analysts – and opposing campaigns – may still be able to identify some individuals from this geographical data, but the specific characteristics of these individuals would remain concealed. Although platforms tend to balk at such micro-level disclosure because it reveals the “secret sauce” of advertisers, the innate surgical precision of effective individual-level targeting remains a key problem with digital advertising, and exposure disclosure is the only way to truly understand the dynamics of modern campaigning. At a minimum, policy makers and the platforms should consider calibrating disclosure to the level of ad targeting, in order to ensure that the more micro-targeted an ad, the greater the disclosure obligation on the spender.

This last point is of particular interest, given the efforts at Twitter (and more recently, Google) to limit the use of micro-targeting on their platforms. But there are still differences in how they will do so; for instance, Google is prohibiting targeting users according to their political leanings with data from public voting records, but will allow targeting users by age, gender, and postal/ZIP code. Contextual targeting, the practice of serving ads to users relating to stories they’re reading or watching, will still be permitted. Twitter, on the other hand, would evade “exposure disclosure” entirely.

As tech companies attempt to navigate the absolutist spectrum of political advertising, it’s worth restating their intended aims:

  • Twitter operates according to the principle that “reach should be earned, not bought.” In other words, it wants to promote organic reach (in the political realm, at least) over any ads paid for by PACs, Super PACs, and 501(c)(4)s. Secondary harms, like the weakening of civic discourse or the spread of misinformation, are important but not the main driver of the policy.
  • Google wants to “improve voters’ confidence” in political ads they encounter on the platform by cracking down on micro-targeting. Its stated goal is to restore confidence in digital advertising and electoral processes globally. The policy would make ads less effective and more costly, and start rolling out in the U.K. within a week, ahead of the General Election.
  • Facebook maintains that ad buyers should be able to run any ads (social, political, electoral), provided they comply with applicable laws and the company’s authorization process. But this historically hands-off approach is being tested, as Facebook is reportedly consulting ad buyers on ways it could limit microtargeting to hamper fake news across the platform.

All companies acknowledge these policies will continue to evolve over time. But the debate over digital political ads has already drawn philosophical battle lines on the respective harms of each policy, whether it’s about the spread of misinformation, restoring democratic norms, free expression, or paid political reach. Facebook has long argued that the government, not tech executives, should be setting the parameters around content. But with each company hardening their stance on digital ad regulation, government involvement looks increasingly unlikely to effect change.