Political Ads 2020:
A Tale of Two Platforms

This is a companion analysis to the Political Advertising Transparency Project’s Spring Monitoring Report. This analysis, established by BBB National Programs’ Digital Advertising Accountability Program (DAAP), breaks down publicly available digital political advertising data from Facebook and Google as part of an ongoing effort to ensure campaigns meet industry-standard political advertising best practices.


SECTION 1 | Executive Summary

 

 

Digital political advertising has become a critical – and controversial – part of the U.S. election cycle. Once subject to the space and time constraints of print, television, and radio, political advertising has exploded into the “wild west” of the digital age, with the total digital ad spend this campaign cycle already doubling what was spent in 2016.1 Across websites and mobile apps, from the walled gardens of social media networks to the mixed content of the open web, candidates, campaigns, and political action committees navigate vast and intersecting networks to get their messages out to activists, funders, and ultimately, the voters.

 

In the Digital Advertising Accountability Program (DAAP) Political Advertising Transparency Project’s compliance reports, we monitor websites and mobile apps for political ads and their associated political advertisers across the United States and report on the advertising platforms used, geographic targeting, type of campaign, and the prevalence of ad transparency in line with best practices for online express political advocacy, called the Political Advertising Principles.2 You can access our Spring Monitoring Report here.  

 

To complement that work, this separate reporting analyzes publicly accessible political advertisement databases provided by major platforms to examine the ten most prominent digital advertisers identified in our Spring Monitoring Report. Specifically, we analyzed the political ads reporting data provided by Facebook and Google (see Section 2), who together capture 78% of digital political ad spend and who both have their own, slightly divergent approaches to political ads transparency.3  

 

 

Key Findings 

 

 

  • For the set of entities we examined within Facebook’s data, most political advertisers are targeting women, except for the Trump Make America Great Again Committee, and adults over the age of 55, except for Bernie 2020.  

 

  • The biggest difference between the datasets publicly available from Google and Facebook is ad spend transparency and impression4 data.   

 

  • Google provides precise weekly spending data for each campaign, rounded to the nearest $100, compared to Facebook’s broad spending ranges that make a precise estimate somewhat harder to come by. In contrast, Facebook provides a narrow range of impression data for its ads, whereas Google provides a wide range of impressions that make it difficult to gauge the impact of each Google Ad.  

 

  • Further, Facebook provides precise geographic percentage breakdowns for impression data, whereas Google only provides data as to whether an ad was targeted to a location, but not the percentage of impressions attributed to that location.   


 

Platform Differences 

 

 

While advertising overall constitutes approximately 98% of Facebook’s revenue and 70% of Google’s revenue, political ads account for only approximately 1% of each company’s total advertising revenue. Though both companies are digital ad giants, they have different business models, products, and means of reaching users. Facebook is a global social media network with a community of 2.6 billion users, which includes its eponymous platform, Instagram, and WhatsApp. The company has a number of advertising tools that combine data generated through use of its products with data obtained from other sources, including a pixel for websites and a software development kit (SDK) for apps that helps Facebook associate data about its users with their other experiences online.

 

Further, in June 2020, Facebook stated that it would implement a feature on its platforms that allows people in the United States to opt out of seeing social issue, electoral, or political ads from candidates or political action committees in their Facebook or Instagram feeds.5 Whether this change will have an impact in overall political ad spend for Facebook remains to be seen. 

 

Google is known for its eponymous search engine, the video sharing platform YouTube, and a powerful display network that delivers ads across two million websites worldwide. Like Facebook, Google has powerful ad tools to reach users with ads tailored to their interests, including pixels for the web and mobile SDKs. Notably, Google also claims that it "[...] never allow[s] granular microtargeting of political ads on [its] platforms [...] " and offers "[...] basic political targeting capabilities to verified advertisers, such as serving ads based on public voter records and general political affiliations (left-leaning, right-leaning, and independent)."6 

 

In sum, Facebook is focused on delivering customized advertising experiences to users of its social media networks, whereas Google is focused on targeting ads to users as they hop from its search engine to the myriad of disparate websites, videos, and apps in their online experience. 

 

We started our analysis with the top 10 advertisers found in our own datasets. Then we used data from Google and Facebook to dig deeper into the advertising habits and tactics of those top ten advertisers to expand our understanding of their reach and impact.  

 

 

Political Ad Libraries Differences and Limitations 

 

 

Facebook’s political Ad Library data is accessible via an application programming interface (API). Once retrieved, the data is specific. For each campaign, the data is presented in the form of individual advertisements. Spending and impression data on the advertisement are given as ranges (e.g., a given advertisement might have received between 5,000 - 10,000 impressions and had $100 - $199 spent on it). Each region (generally U.S. states for political ads) and demographic associated with the ad have percentages attached (e.g., 10% of the ad’s impressions were in New York, 40% of the audience were male, and 20% were between the ages of 25 - 34). As the impression and spending ranges are relatively narrow, it is possible to combine them with the location and demographic data to draw conclusions about a campaign’s targeting strategy and advertising budget priorities.  

 

The Google political ads Transparency Report data is comparatively easier to retrieve. An automatically generated set of CSV files, compatible with spreadsheet software, is located on Google’s website and may be downloaded without restriction. The data itself, however, is somewhat less precise. The data is presented in the form of individual advertisements, like Facebook, but are inconsistent (geographic data such as state names, ZIP codes, congressional districts, and metropolitan areas are freely mixed) and presented without any indication as to what proportion of impressions can be attributed to a given region. The demographic data provided shows the ad’s target demographics (e.g., male, female, unknown, etc.) but again, without any indication of proportional impressions. In addition, impression and spending data is provided in much larger ranges than the corresponding Facebook data (e.g., 10,000 - 100,000 impressions, $100 -$1,000 dollars, etc.). Consequently, it is comparatively less meaningful to draw conclusions about a given campaign, based on its aggregated impression, spending, region, and demographic data.  

 

In one way, however, Google, does provide clearer information than Facebook. In addition to the data on individual advertisements, Google also provides precise weekly spending data for each campaign, rounded to the nearest $100 (e.g., on the week starting January 19, 2020, the Amy McGrath campaign spent $21,100). Comparatively, the equivalent Facebook data can only be estimated from the provided advertisement and spending data ranges. 

 

 

About This Analysis

 

 

This companion analysis is produced as part of DAAP's political advertising transparency project. In 2018, DAAP, a division of BBB National Programs, was tasked with enforcing the Digital Advertising Alliance (DAA)'s best practices for online express political advocacy7 called the Political Advertising Principles. In pursuit of this mission, DAAP monitors express advocacy political ads across websites, social media, and mobile apps to assess political advertiser8 compliance with those Principles.  

 

DAAP notes that in an effort to deepen our understanding of the digital political advertising market and explore the state of transparency in the industry, we examined data provided by major political ad platforms. While this data was not specifically applicable to compliance with the Political Advertising Principles, and thus outside the scope of our enforcement efforts, it appeared to be of public interest. Therefore, because we had access to these publicly available transparency tools, we decided to publish this analysis of additional platform data alongside our usual compliance assessment to illuminate some of the practices of political advertisers in this space. Note that none of this information is meant to express an opinion about any of the candidates, campaigns, or other political advertisers, nor are we opining on the quality, correctness, or adequacy of any platform tools. 

 

 

SECTION 2 | Political Ads Data from Google and Facebook

 

What Facebook's Data Tells Us 

 

 

Facebook’s political ad library9 includes every active or inactive ad on social issues, elections, or politics that have run on the platform since May 2018. This ad library provides a downloadable Ad Library API, which allows certain users10 to access data about ads from countries where the ad library is live, including Brazil, Canada, European Union countries, India, Israel, Ukraine, the United Kingdom, and the United States. 


Data available from the Facebook Ad Library API include: 

 

  • Ad creatives
  • Ad performance data, including total amount spent (range)
  • Total impressions an ad received (range)
  • Demographics: age, gender, and location of users reached (percentage)
  • Dates the ad ran 

 

 

The Context 

 

 

It is important to note that this period of analysis includes Super Tuesday, which led to great variability among the political ad spend of various advertisers. Specifically, for ad spend, there is a big change in the data both where billionaire Mike Bloomberg entered and exited the Democratic presidential race.  

 

The data below are based on estimated impressions.11

 

 

 

Figure 1: Gender Impressions 

Except for Bernie 2020 and Trump Make America Great Again Committee, most political advertisements on Facebook from our monitoring set skewed heavily toward gaining most of their impressions from women.  

 

 

 

Figure 2: Age Impressions 

When looking at advertisements based on the age range of users, standouts on Facebook included Bernie 2020 advertisements, which skewed toward users in the 18 - 34 age range, and Jeff Merkley for Oregon advertisements, which were quite the opposite and skewed toward users age 55+. Interestingly, the Biden for President and Trump Make America Great Again  Committee advertisers each had a relatively even spread of impressions across all age groups, though the Biden for President data skewed slightly younger. 

 

 

 

Figure 3: Estimated Weekly Ad Spend 

We were able to estimate weekly ad spend on Facebook for each of our identified political advertisers12 and found that, unsurprisingly, Mike Bloomberg 2020 represents a significant outlier. From January 1 to May 24, he spent approximately twice as much as his combined fellow democratic primary candidates from out dataset, a figure which includes Democratic nominee Joe Biden’s continuous spend. 


 

 

Figure 4: Estimated Weekly Ad Spend (without Mike Bloomberg 2020)

Because of the Bloomberg outlier, we also created the following dataset to show the relative ad spend of the political advertisers excluding Bloomberg. Here we can see more clearly that Biden 2020 had a significant increase in ad spend in the weeks leading up to the Super Tuesday contest, as well as the drop-off of other candidates’ spending following Super Tuesday. 

 

 

 

Figure 5: Ad Spend United States Heat Map 

To see what ad spending looked like for each state, we overlapped the weekly ad spend data with the number of impressions received in each state.13 We note that Texas, California, and Florida appear to be major foci of this time period’s ad spend.  

 

 

 

Figure 6: Top Facebook Advertiser Based on Ad Spend 

When we looked at who was spending that money, we found that Mike Bloomberg 2020 and Trump Make America Great Again Committee were two top spenders in many of the states across the country. Biden for President appeared to be a top spender in some of the northeastern states, and Bernie 2020 was a top spender in Iowa—the first state to vote in the Democratic primary contest.  

 

 

 

Figure 7: Top Facebook Advertiser Based on Ad Spend (without Mike Bloomberg 2020)

When you remove the Mike Bloomberg 2020 outlier, the top political ad spend per state across the United States belongs to the Trump Make America Great Again Committee, with Biden for President and Bernie 2020 being top spenders in northeast, west, and parts of the Midwest.  

 

 

 

Figure 8: Top Facebook Advertiser Based on Estimated Impressions 

Here, we have the top Facebook political advertisers from our monitoring set based on estimated impression data. 

 

 

 

Figure 9: Top Facebook Advertiser Based on Estimated Impressions (without Mike Bloomberg 2020) 

Here, we have the top Facebook political advertisers from our monitoring set based on estimated impression data with Mike Bloomberg 2020 removed.  


What Google’s Data Tells Us 

 

 

Google provides a Political Advertising Transparency Report14 for political advertising that takes place on Google search, YouTube, and Google’s partner properties on the open web. This database includes ads that feature a current officeholder or candidate for an elected federal or state office, federal or state political party, or state ballot measure, initiative, or proposition that qualifies for the ballot in a state. 

 

Data provided in the Google Transparency Report includes: 

 

  • Precise weekly ad spend for each political advertiser 

  • Number of unique Google Ads per political advertiser 

  • Data on locations where each Google Ad was targeted 

 

 

 

 


Figure 10: Weekly Ad Spend 

As with Facebook data, Mike Bloomberg 2020 represents a significant outlier in weekly ad spend compared to other political advertisers, including democratic primary competitors, from January 1 to March 1.  

 

 

 


Figure 11: Weekly Ad Spend without Mike Bloomberg 2020 

Because of the Bloomberg outlier, here again we created the following dataset to show the relative ad spend of the political advertisers excluding Bloomberg. We note that both Biden for President and Bernie 2020 had a significant increase in ad spend in the weeks leading up to the Super Tuesday contest, which was to be expected, and the Trump Make America Great Again Committee increased its political ad spend significantly in the week of May 24. 

 

 

 

 

Like Facebook, Google provides location data about political ad targeting. While some ads are targeted to the United States generally, other unique ads are targeted to a single state, and some even focused on specific zip codes or metro areas. Using that data, we have identified the political advertisers with the most unique ads per state. We note that these descriptions do not include impression data, which is highly difficult to estimate based on what Google provides. Further, we note that ads targeted to a certain state or region does not preclude them from being viewed elsewhere.15 

 

 

As the Google dataset does not provide sufficiently precise impression or spending data that can be attributed to specific states, we have instead looked to the number of unique advertisements each campaign targeted to specific states. This approach is not perfect: a few advertisements in the data set are targeted at more than one state, the provided geographic data is inconsistent (as previously described), and campaigns use individual ads differently (e.g., one campaign may place one unique advertisement for a few hours targeting one state, while another may place one unique advertisement for several days targeting multiple states). Nevertheless, we feel that the provided data is adequate to gain a sense of the approaches taken by different campaigns in their geographic targeting in Google ads, and the emphasis each campaign has placed on targeting specific states. 

 

 


Figure 12: Top Unique Ad Advertisers 

 


Figure 13: Top Unique Ad Advertisers (without Mike Bloomberg 2020) 

Once we removed the Mike Bloomberg 2020 outlier from the data, we found a significant number of states where no Google ads were targeted, and that Biden for President focused its targeting efforts on the Midwest and on the coastal states. 

 

 

 

Figure 14: Top Unique Ad Advertisers, Pre-Super Tuesday  

 

 

 

Figure 15: Top Unique Ad Advertisers, Post-Super Tuesday  


 

 

 

Contact Us

*Required fields






For questions about this report or the principles that DAAP enforces, please send us a note through the Contact Us form.

  1. Butler/Till Media Services, Inc., Chasing Media Impact During a Political Ad Storm: A 2020 Election Outlook Media Guide 3 (Nov 4, 2019), https://www.butlertill.com/wp-content/uploads/ButlerTill_ChasingMedia_Impact_2020.pdf.
  2. Digital Advertising Alliance, Application of the Self-Regulatory Principles of Transparency & Accountability to Political Advertising § I.A (2018), https://aboutpoliticalads.org/sites/politic/files/DAA_files/DAA_Self-Regulatory_Principles_for_Political_Advertising_May2018.pdf (“Express Advocacy means paid-for communications that unmistakably urge the election or defeat of one or more clearly identified candidate(s) for a federal or statewide election.”); See also id. § I.C (“Political Advertisement means an advertisement that contains Express Advocacy.”).
  3. eMarketer Editors, Facebook Dominates 2019/2020 Political Ad Spending, eMarketer (Feb. 24, 2020), https://www.emarketer.com/content/facebook-dominates-2019-2020-political-ad-spending.
  4. “An impression is counted as the number of times an instance of an ad is on screen for the first time. … In a few cases when it can't be determined whether ads are on screen, such as on feature mobile phones, impressions are counted when ads are delivered to devices.” Impressions, Facebook, https://www.facebook.com/business/help/675615482516035 (last visited July 9, 2020). See also “How often your ad is shown. An impression is counted each time your ad is shown on a search result page or other site on the Google Network.” Impressions: Definition, Google, https://support.google.com/google-ads/answer/6320?hl=en#:~:text=An%20impression%20is%20counted%20each,your%20ad%20may%20be%20shown.(last visited July 9, 2020).
  5. Mike Isaac, “Now You Can Opt Out of Seeing Political Ads on Facebook,” New York Times, (June 16, 2020), https://www.nytimes.com/2020/06/16/technology/opt-out-political-ads-facebook.html
  6. Scott Spencer, An Update on our Political Ads Policy, Google (Nov. 20, 2019), https://www.blog.google/technology/ads/update-our-political-ads-policy/.
  7. Digital Advertising Alliance, supra, § I.A (“Express Advocacy means paid-for communications that unmistakably urge the election or defeat of one or more clearly identified candidate(s) for a federal or statewide election.”).
  8. Id. § I.B (“Political Advertiser means a person or entity that pays for the display of a Political Advertisement.”).
  9. We note that Facebook and Google classify political ads differently than DAAP – DAAP’s compliance monitoring focuses only on express advocacy political ads (e.g. ads that advocate for the election or defeat of a particular candidate). Facebook ads library includes data on “every active and inactive ad about social issues, elections or politics”[1] and Google’s political ads library encompasses ads that “feature a current officeholder or candidate for an elected federal or state office, federal or state political party, or state ballot measure, initiative, or proposition that qualifies for the ballot in a state.” Ad Library Frequently Asked Questions, Facebook, https://www.facebook.com/ads/library/?active_status=all&ad_type=all&country=US&impression_search_field=has_impressions_lifetime (last visited July 9, 2020); Transparency Report, Google, https://transparencyreport.google.com/political-ads/region/US (last visited July 9, 2020).
  10. The Facebook Ad Library API allows access to the raw data underpinning its Ad Library Report, including individual ads. To gain access to the API, Facebook requires that an individual be authenticated to the same extent as would be required to place political ads on Facebook. This includes confirmation of identity and location (by sending a picture of photo ID), creation of a Facebook Developer account (which ties into an ordinary Facebook account), and ultimately confirmation of a phone number and/or credit card. See Ad Library API, Facebook, https://www.facebook.com/ads/library/api (last visited July 9, 2020).
  11. The Facebook Ad Library API does not provide the exact number of impressions for a given political ad, but rather a range of impressions, such as “10,000-14,999” impressions. See Graph API Reference v7.0: Archived Ad, Facebook, https://developers.facebook.com/docs/graph-api/reference/archived-ad/ (last visited July 9, 2020). For purposes of this analysis, we estimated the number of impressions for each ad by using the rounded midpoint for each range (e.g, an estimated 12,500 impressions for the range “10,000-14,999). Subsequent analysis, including Monte Carlo simulations, suggests that for most major candidates running many ads with a few impressions each (a typical pattern), this method produces a margin of error of less than 5%, which we deemed acceptable for purposes of this analysis.
  12. Weekly ad spend was estimated in a similar fashion to impression data, as individual advertisement data is only provided in a range (e.g., $0-99). We first took the rounded midpoint of the range (e.g., $50) for each ad, and allotted the money in equal portions to each day the ad was run. Most ads only ran for a few days at most. We then grouped total daily ad spend for each candidate into 7-day weekly “baskets”, starting on December 29, 2019, and continuing to late May 2020. Subsequent analysis and comparison to patterns in other available data (e.g., Google’s weekly ad spend data, which is exact) suggests that this method is adequate for the purpose of gaining a broad understanding of spending patterns between candidates over the given time period.
  13. Ad spend for each state was estimated by taking the estimated spend for each ad (by the same method used for daily spend; see n. 12), and allotting the money proportionately to states according to the number of impressions received in each state. For instance, for an ad with an estimated spend of $100, which had 5% of its impressions in Oklahoma, we would allot $5 of spend for that ad to Oklahoma. We then totaled spend for each state, both for each candidate, and overall. Given the Facebook ad targeting methodology, which uses an ”auction” to spend available funds on impressions in targeted populations (see, e.g., About Ad Auctions, Facebook, https://www.facebook.com/business/help/430291176997542 (last visited July 9, 2020)), and that most ads were likely targeted to populations in specific states, we felt that this method of estimating spend per state was acceptable for purposes of gaining a broad overview of spending across states and between candidates.
  14. Transparency Report, Google, https://transparencyreport.google.com/political-ads/region/US (last visited July 9, 2020).
  15. DAAP methodology for location.
    1. If the name of a state is somewhere in the list of locations, find the first match, then set the ad to that state. (Lots of states have: either the name of the state up front, OR a congressional district including the name of the state (e.g., CA1,California,USA), OR a metro area including the name of the state (e.g., Minneapolis,Minnesota,USA)).
    2. For all remaining unmatched ads, if a valid zip code corresponding to a state is in the list of locations, find the first match, then set the ad to that state.
    3. For all remaining unmatched ads, if a match to the following list is found, find the first match, then set the ad to that state:

      "63001": "Missouri",

      "Progreso": "Texas",

      "Austin": "Texas",

      "Dallas": "Texas",

      "El Paso": "Texas",

      "Houston": "Texas",

      "San Diego": "California",

      "Palm Beach": "Florida",

      "Denver": "Colorado"

    4. For the following list, drop the ad:

      "Hail Province",

      "Bourg-la-Reine",

      "Colombes",

      "Saint-Cloud",

      "Vanves",

      "United States",

      "Not targeted"

    5. Drop all unmatched ads