Memes + model: which topics drove the election

I was late to the party on memes, preferring multivariate models to help me understand what matters most in marketing and political effectiveness. Memes did capture my interest in 2016, when our analysis of the elections pointed to their power to capture agendas among stakeholders in the dynamic political ecosystem. Sure, these memes are simplistic and lack nuance, but they do point to the topics dominating elections and jive with the most comprehensive analysis of the last two US presidential elections, now published in the Journal of Business Research and forthcoming in the Handbook of Political Marketing.

Based on over 200 million posts, our structural topic models identified the key topics of the candidates’ owned media, paid media such as TV ads, and earned media such as offline word of mouth and online conversations. LASSO variable selection and Vector Autoregressive Models show how much each variable drove the most accurate polls.

What did we find? In both elections, fear appeals (e.g. Trump on terror, Clinton and Biden on Trump's Russian ties) were a lot less effective than how people felt about the economy. The same holds for 2024, and in decades of earlier political research. However, how people feel about the economy is shaped not so much by statistics or proposed policies, but by how these are discussed among friends and family (offline word-of-mouth), in online conversations (social media and disinformation and to a much lesser extent, in mainstream media. And a lot of these sources implied the United States of America is on fire:

Let’s organize our 2024 topics by marketing’s 4Ps: product, price, place and promotion. As Representative Dean Phillips—the only elected Democrat who mounted a primary bid to unseat President Biden this year—put it ?when asked by a?Washington Post?reporter what the party must do to reinvent itself, “We have good product and terrible packaging and distribution.” Note the omission of PRICE, which was a key factor in this election. Here's my breakdown:

a) Product: how can candidates who were in power, demonstrate innovation?

b) Price: who is to blame for high inflation, and slow response to Hurricane Helene damages?

c) Place: did offline word-of-mouth and Joe Rogan's podcast tip the election?

d) Promotion: timelines, media coverage and the candidate's messaging

PRODUCT: the last three presidential elections were all about CHANGE, which turned the good old incumbent advantage into an incumbent deficit. Both Harris and Trump were in power before, respectively as vice president and as president. A key question that continued to come up in memes: why didn't they already fix the things they are now promising to fix when elected?

In such election climate, which candidate is seen as most innovative? While the policies proposed by Harris/Waltz were endorsed by Nobel Prize laureates, we are talking minor instead of major product innovation. Senator Chris Murphy of Connecticut, in a long?thread ?on X outlined that his party “skips past the way people are feeling … and straight to uninspiring solutions … that do little to actually upset the status quo of who has power and who doesn’t.” Moreover, the Democrats’ product was highly segmented this year, with specific proposals targeting specific demographics. Instead, Trump mostly ran on one big platform for all:

?

PRICE: a key complaint in 2024 was that the price is too high. Many voters were upset about inflation, with the prices of several products being highlighted in memes:

But is the government to blame for inflation? Some memes pointed to overspending:

While others pointed out that the president does not control inflation or gas prices:

The key turning point for us to call the election were the conversations around the flood damage by Hurricane Helene in North Carolina. Many memes implied that US tax dollars go to foreigners instead of US citizens.

And AI-generated images such as the below blamed Kamala Harris for fund depletion:

While others pointed out it was Republicans who voted against more FEMA resources, and against measures that should ease climate change:

PLACE: while the Harris campaign held concerts, went door-to-door and broadcasted many ads, Trump came on podcasts, such as Joe Rogan’s, and veered deep into ‘blue states’. According to this unscientific poll by my colleague AJ Rollsy, the Joe Rogan appearance actually swayed votes:

Leading up to the election, Trump voters trusted friends and family over traditional media, a Northeastern University survey found: friends and family (29%) and the news media (26%) are the top sources Americans use for voting information.?Of course such self-reports should be taken with a grain of salt. However, our econometric model also demonstrated a large election impact of offline word-of-mouth (WOM) in 2016 and 2020, with pro-Trump WOM dominating in 2016 and 2024.

Even more interesting is the survey's breakdown by demographics and voting behavior. Americans ages 18 to 24 and those with less formal education lean more on personal networks; and are more likely to vote for Trump, while those 65 and older and those with higher education and income favor the news media, and are more likely to vote for Harris. The late candidacy of Harris likely made it harder for positive WOM about her to spread and have a real impact. Instead, her campaign relied on paid promotion:

PROMOTION: both parties had plenty of money for ads, but their timelines, media coverage and messaging strongly differed. On the timeline, marketing research has long demonstrated the long-term effects of marketing communication , reaching over double of the short-term effect within months or even years. Trump had plenty of awareness going into election season, while Harris was less known. Meanwhile, Democrat Kamala Harris — who only became the party’s nominee three months ago — failed to break through the media bubble.

“It was sort of like ‘too little too late,’ to establish a national candidacy — you used to spend an enormous amount of time and money, and they certainly spent money, but they didn’t have a lot of time,” Northeastern professor WIhbey commented.

Just as in 2016 and 2020, Trump substantially benefitted from more media coverage across the political spectrum:

Kassova and Addy rightly title their October 25th Fortune piece "How the Democrat-leaning news media is unwittingly aiding Trump" . Our research showed the same, with the New York Times playing an important role getting Trump elected the first time around - even their lukewarm endorsement for Hillary Clinton benefitted Trump instead. In 2024, we saw a substantial drop in media mentions of Harris after the initial strong media coverage getting to know Harris at the start of her campaign, including her mother's story about coconut trees , and Charli XCS's Brat comment, which also showed up in memes:

Finally, the messaging of the candidates strongly differed, with the Trump campaign going for the same message across demographics: Make America Great Again, and only Trump can do so:

?As nicely captured in this word cloud of Trump's ads on Meta platforms, which also featured strong calls-for-action: request a ballot, make a plan, vote at the polls.

Instead, Harris' messaging was more segmented by audience - a tactic that helped her get elected in California. Fear about Trump's election and Project 2025, abortion rights, middle class economic struggles and health care were such targeted themes picked up in this word cloud of the Harris ad campaigns on the Meta platform:

However, it did not work in a national election where memes spread across targeted segments, diffusing the specific messages. In interviews for national audiences, Harris often seemed to avoid taking a stance.

In marketing, the Ehrenberg-Bass Institute has long pointed out that light buyers are key to brand growth, and that they need continuous distinctive messaging with straightforward talking points. The biggest mistake of marketers is to believe the consumer cares as much about the brand and category as they do. Likewise, much has been made of this election of low-information voters, who care not about elaborate policy proposals but about decide based on general thoughts and feelings, as visually captured in memes:

?