Prof Jennifer Stromer-Galley
Professor at Syracuse University
Dr Patricia Rossini
Derby Fellow at the Department of Communication and Media, University of Liverpool
Dr Brian McKernan
Research assistant professor in the School of Information Studies at Syracuse University
Dr Jeff Hemsley
Associate Professor at the iSchool at Syracuse University
Section 5: Social media
- Media and social media platforms finally begin to embrace their roles as democratic gatekeepers
- Did social media make us more or less politically unequal in 2020?
- Platform transparency in the fight against disinformation
- Why Trump’s determination to sow doubt about data undermines democracy
- A banner year for advertising and a look at differences across platforms
- How Joe Biden conveyed empathy
- The debates and the election conversation on Twitter
- Did the economy, COVID-19, or Black Lives Matter to the Senate candidates in 2020?
- Leadership through showmanship: Trump’s ability to coin nicknames for opponents on Twitter
- Election countdown: Instagram’s role in visualizing the 2020 campaign
- Candidates did lackluster youth targeting on Instagram
- College students, political engagement and Snapchat in the 2020 general election
- Detecting emotions in Facebook political ads with computer vision
In 2016, much concern was raised about advertising on social media, especially Facebook. A Bloomberg story at the end of the 2016 election noted that the Trump campaign used Facebook advertising specifically to target pockets of Democratic voters to demobilize them from voting for Hillary Clinton. Without any way to systematically monitor ads on Facebook, researchers were left with anecdotes to determine who was targeted and what was said.
In 2018, Facebook created an ad library for journalists and researchers. Our project, Illuminating 2020, tracked the spending, targets, and content of ad buys on Facebook and Instagram and created a dashboard to visualize the data.
Using computational techniques to classify the content in ads, with at least 75% accuracy for each category, we find noteworthy differences in how the campaigns communicated on the platforms. Our research suggests that advertising largely reflected the demographic trends of the Republican and Democratic Parties and the rhetorical trends of the candidates. Unfortunately, because of the limitations of data reporting from Facebook, it is challenging to ascertain fine-grained micro-targeting practices.
Overall Spending and Trends
Facebook advertising was a major focus of total overall ad spending by Donald Trump and Joe Biden. As much as $1.5 billion was spent overall on advertising, from television to radio to digital media, between April 6 and October 25. A major share of that spending was on Facebook. Our analysis suggests that, of the ads associated with their official campaign pages, the Biden campaign spent $76.9 million between June 1 and November 1, and the Trump campaign spent $87.2 million. This amount is significantly more than the campaigns spent in 2016: Trump spent an estimated $44 million and Clinton spent $28 million between June and November, according to a Facebook report.
Although the Trump campaign spent more on Facebook overall, between October 5 and November 1, 2020 Biden outspent Trump 4:3. Biden had his largest new ad buys on Facebook over the second presidential debate, which took place on October 22nd, spending over $10 million as compared with Trump’s $7.5 million (See Figure 1).
When looking across the five months of the campaign, from June 1 until November 1, 2020, the overall targeting of demographic groups matched the demographics of the political parties. Biden overweighted women in his targeted spending, which reflects the gender skew in the Democratic party (See Figure 2). The two campaigns had distinctly different age targeting strategies. The Trump campaign prioritized ad spending to people 45 and older, while the Biden campaign focused heavily on the 25-44 age brackets, mirroring the differences in age demographics of the two parties. Millennials are more likely to lean Democrat, while the Silent Generation is more likely to lean Republican. Neither campaign focused resources heavily on the youngest voters – those in the 18-24 age bracket – even though this group most heavily uses social media, especially Instagram. Both campaigns heavily targeted the 65+ category, even though less than 50% of Americans in that age bracket are on any social media platforms. To us, that speaks to the power of social media advertising for micro-targeting the nations’ most motivated voters (See Figure 3). Lacking in Facebook data is information about education level, race and ethnicity, religion, and location data beyond the level of the state to ascertain micro-targeting strategies.
At a high level, Trump’s advertisements on Facebook looked much like his campaign rhetoric. He attacked his opponents – Biden, Harris, and the news media – and did so often with an uncivil tone. The Biden advertisements tended to emphasize his personality and ability to lead and were overwhelmingly civil.
Nearly all ads call targets to action in the text – ranging from fundraising to polls to petitions. Beyond that, there are distinct differences in messaging strategy. Biden spent more on advocacy ads, while Trump spent more on attack ads. When Trump attacked, he was more likely to attack Biden’s policies, while when Biden attacked, he was more likely to attack Trump’s persona – his character, personality, and ability to lead.
We also analyzed whether an ad exhibited uncivil language. We operationalize incivility as hateful, disparaging, and derogatory remarks targeted at another individual or group. Nearly one-third of Trump’s ads contain evidence of incivility, while a fraction of Biden’s ads were uncivil in tone. (Figure 4)
Our analysis suggests that party demographics drove targeting strategy. We also found that the rhetoric in the advertisements largely matched the rhetoric by the candidates on the campaign trail. Trump’s ads show evidence of the same personal attacks and self-aggrandizement as his campaign speeches. The challenge for researchers is the lack of fine-grained targeting data to ascertain whether the campaigns undertook the same demobilization approaches in 2020 as reported in 2016. While we gain a high-level view, the extent of microtargeting is unfortunately left to guesswork.