Voter reaction by demographics

“Were it not for Republican gains (and Democratic losses) among blacks, Hispanics and Asians – Trump likely would have lost’ al Gharbi (2018, 511-512)

Different voter groups showed interesting shifts in the 4 months leading up to the 2016 US presidential election, as detailed in our paper ( Based on your vote, we highlight in this blog the breakdowns by income and ethnicity. Over the 4 months before the elections, our metric is the gap between Clinton and Trump in the daily USC Dornsife / LA Times polls, whose probabilistic design helped reduce response bias.

Screen Shot 2020-08-02 at 4.04.21 PM

As to income, the daily poll gap between Clinton and Trump shows the changing appeal of each candidate to voters identified as ‘low income’ (below $35K annually), middle income (between $35 and $75K annually) and high income (above $75K annually). While low and middle income voters consistently preferred respectively Clinton and Trump, it is the high income voters who switched, first from Trump to Clinton in August, then decidedly to Trump in mid September at the same time as Clinton’s ‘Basket of Deplorables’ statement leaked. Finally, all income groups turned more to Trump in the last weeks before the election, consistent with the movement in overall polls.

Screen Shot 2020-08-02 at 8.24.55 PM

As to ethnicity, black and white voters consistently preferred respectively Clinton and Trump throughout the election campaign. As noted in the opening quote, it is the Hispanic and ‘other’ (mostly Asian) voters that turned away from Clinton in the last 2 weeks before the election, around the time of Comey’s letter to Congress on the FBI investigation into Clinton’s emails.

Why did we observe such important swings for key demographics? We combined data on the candidates’ owned media (statements and social media posts), TV ads and donations with almost 100 million Twitter, Facebook and Instagram social media reactions, events (such as the FBI email investigation), their news coverage and fake news sharing. Our longitudinal analysis showed the importance of each variable in driving the polls.

The main poll gap driver was Fake News on Clinton’s emails, posted on her Facebook page. This was much more impactful on high earners (-20 points) than on middle (-0.24) and low income (0) groups, consistent with the high earners’ drop in Clinton polls over the campaign period and especially with Comey’s Letter to Congress of October 28. As to ethnicity, Hispanics (-46 points) were affected, not black or white voters (0).

Screen Shot 2020-08-03 at 5.27.49 PM

Candidates’ campaign actions showed vastly different effects by demographic. Clinton’s TV ads on the Economy and Gun Control worked especially well for low earners, but not at all for mid earners, who are instead affected by Trump’s ads against Terror. As to ethnicity, Clinton’s Economy ads worked well for black voters, but not for Hispanic voters. Trump’s Terror ads helped him attract mid earners, but pushed away black voters.

Screen Shot 2020-08-03 at 5.28.31 PM

Social media comments negative of Clinton, hurt her chances across demographics, with large impact for high earners and Hispanics, and a moderate impact for mid earners and other ethnicity.  In contrast, news coverage of Trump’s sex scandals only hurts his chances for ‘other races’, and even benefitted him for mid earners and for Hispanics. Finally, positive Trump sentiment on social media had rather small effects, with low income and ‘other ethnicity’ voters more affected than others. Interestingly, Clinton’s ‘Basket of Deplorables’ statement increased positive sentiment for Trump on Twitter and lowered the Clinton-Trump poll gap more than any other event. The harm was especially pronounced for high earners and for black voters.

Overall, the fake news impact was most pronounced for Hispanics and high earners – demographics who moved towards Trump in the last weeks before the election. The polls for Asian voters were more driven by social media sentiment on both candidates, which also turned in favor of Trump in the weeks leading up to the election. In contrast, other demographics were more affected by factors stable throughout the campaign. Low-earners were most affected by Clinton ads about the economy, while mid-earners were most affected by Trump TV ads on Terror.

What are the lessons for the 2020 election? First, neither of the candidates ‘owns’ any demographics’ vote. While they start at different positions based on their history with the demographic, the candidate’s actions and their supporters’ (including fake news sharing) will influence their relative standing from now till election day. Second, it is not so much the statement or leak that matters, but how it influences (fake) news coverage. Finally, platform and topic matters: different demographics prefer different platforms (e.g. TV versus social media) and were more influenced by respectively the Economy, Gun Control and Terror, just as Health and the Economy will be key issues this November.

How does this demographic breakdown look for the 2020 election? Currently collecting data to analyze just that, we are looking forward to your reaction.


3 thoughts on “Voter reaction by demographics

  1. Pingback: How topics and channels matter in U.S. presidential elections | Smarter Marketing gets Better Results

  2. Pingback: Is Trump a good marketer? | Smarter Marketing gets Better Results

  3. Pingback: 2020 Election impact of fake news | Smarter Marketing gets Better Results

Leave your comment here

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s