Town Hall Analytics: Data Science Reveals Viewers’ Emotions by Categorizing Language Ahead of Debate
Updated: Oct 22
4Sight Analysis Shows Trump Uses More Words that Evoke Fear and Anger; Biden Uses Words that Evoke More Hope and Joy
In the past, campaign strategists have analyzed consumer sentiment from Presidential debates to determine how the public feels about a candidate’s performance and message. However, measuring sentiment alone requires oversimplification of data and does not tell the entire story.
Last week’s dueling town halls with President Trump and former Vice President Biden offered an opportunity to gauge some of the specific emotions behind the candidates’ words and how audiences feel about the presidential candidates’ messages. Our company, 4Sight, mines user-generated content (UGC) for brands in the private sector, but we leveraged our methodology to extract objective learnings about the candidates in a non-partisan way. To do so, the data scientists at 4Sight went beyond measuring sentiment from these events to measuring 8 different emotions. The same analysis, when applied to brands and products, offers insight into how consumers feel about brands. The results have important takeaways for brands as well as for the candidates themselves.
The 4Sight team analyzed both presidential candidates’ transcripts from last Thursday’s town halls, as well as thousands of YouTube comments from both events to gain an insight into the emotions evoked by each. Words were categorized into eight emotions: Anger, Disgust, Fear, Trust, Joy, Surprise, Anticipation and Sadness. Viewers were not categorized by their support for either candidate within this particular analysis.
Specifically, 4Sight used a two-step process:
Extract: 4Sight pulled thousands of viewer comments from YouTube in addition to candidate transcripts of their town halls.
Harness: 4Sight leveraged the National Research Council’s emotional lexicon, an academic, crowdsourced emotional lexicon that partners word associations across 8 different emotional vectors: Fear, Disgust, Sadness, Anger, Trust, Joy, Anticipation, and Surprise. Through crowdsourcing, words are paired with different emotions e.g. Protect is a Trust word. We also used a proprietary methodology for extracting “word wheels”, which highlight the top ten words for each candidate by emotion. The size of the word's position on the wheel indicates the percentage it was used.
For example, let's look at anticipation to understand the “word wheels”. Based on this simple graphic, in this example, we can see that Biden used words that elicit anticipation far more - such as vote, plan, and opportunity. Commenters on Biden’s town hall video also used more anticipation words than commenters on Trump’s video, by echoing words such as vote and plan.
Town Hall – Candidate Findings:
In his Town Hall, President Trump’s language focused on words that correspond to words that are more tied to Anger (e.g., denounce, terrible), Disgust (e.g., illegal, hoax), and Fear (e.g., criminal). For example, 12% of Trump’s words evoke emotions of Fear vs only 10% for Biden. Biden’s language, on the other hand, reflected verbiage categorized with Trust (e.g., vote and provide) and Joy (e.g., hope and progress). In Biden’s case, 28% of his words evoke emotions of Trust vs only 22% for Trump.
President Trump's linguistic choices focused on words that correspond more with anger and disgust, and commenters on his town hall tended to reflect the same language.
Former Vice President Biden's linguistic choices focused on words that correspond more with joy and trust, and though viewers of his town hall did not often reflect the same language in their comments, there were still many comments echoing the hopeful tone he tried to convey.
Note: Each percentage shown above corresponds to the percentage of words in relation to the sum of all words used that evoke emotion. For example, of all emotionally-coded language that Trump spoke, 10% were words that the NRC classifies as “Anger” words.
Town Hall – Viewer Findings:
The words that viewers of Trump’s Town Hall event used closely matched the same type of emotions in the candidate’s comments. Again, those viewer comments were not segmented based on whether they were a Biden or Trump supporter or undecided in this analysis. The most common emotions we found based on language usage were Anger, Fear and Disgust - words such as terrible, hoax, illegal and criminal. Trump spoke more words associated with trust, but viewers did not repeat those words as often as they did language associated with Anger, Fear and Disgust.
Biden used more words associated with Anticipation (a proxy for hope) and Joy than Trump, words like peaceful, create, save, and progress. However, viewers did not repeat his language usage in their responses as often compared to Trump. Bear in mind that the comments from viewers are not categorized by their support for either candidate in this analysis, so it’s possible that Trump supporters responded on YouTube more often than Biden supporters for his town hall event - something that could be analyzed further.
Words to Watch on This Week’s Debate:
The presidential candidates are scheduled to spar onstage again on Oct. 22. Here’s what we’ll be watching for:
Team Trump: If Trump’s team concludes that he needs more balance on his messaging toward Hope and Joy, then we should expect him to evoke more of those emotions with more of a Ronald Reagan “Morning in America” vibe, envisioning a joyful, bright world with his second term as president - showing optimism about the economy, the pandemic response, foreign policy, etc.
Team Biden: If Biden’s team is looking at the same 4Sight data, they might consider increasing his usage of words like hope, vote, truth, justice, peace and respect to increase emotions toward trusting him as a candidate and president. Or they could go the other direction and seek to leverage language that evokes more Anger and Fear, an emotional strategy shift that would align more with Trump’s previous messaging.
Why it matters: Emotions Associated with Specific Words Build (or destroy) Brands
Everyone has a point of view on the upcoming election. The point of leveraging the 4Sight capability in this case was not to take one side or the other, but rather to objectively analyze the language of candidates and viewers. It’s relevant by itself, but also for brands. Brands can use the same data analysis on consumer reviews and comments in social media to gain a competitive advantage in their category.
Word choices are also important to build brand equity; thus products and politicians alike should be intentional about the words they use or avoid. A proactive understanding of which words can positively affect the desired outcome (through a product’s star ratings or how followers engage on social media) serves as a powerful tool in building a brand’s reputation.