How Brigade Voters Forecasted a Trump Victory When Polls Didn’t
The 2016 presidential election will be studied by pollsters, political campaigns and historians for generations to come — and for good reason. Nearly all pre-election polls called for a resounding Hillary Clinton victory on November 8, but in a stunning turn of events, billionaire businessman Donald Trump became the 45th President of the United States.
So what happened and why did the polls get it wrong?
A factor could be Democrats in key states who cast their votes for Trump — a population that many overlooked or undercounted, but one that was making their choice for Commander-in-Chief known on Brigade for a number of weeks leading up to Election Day.
Here’s a recap of what we know, what we don’t and what’s next.
In the final weeks of the general election, 94.5% of registered Republicans on Brigade pledged for Trump; only 2.2% pledged for Clinton — ratios roughly in line with the national average. However, during the same time period, we saw nearly 40% of registered Democrats on Brigade pledge to vote for Trump, while only 55% pledged to vote for Clinton. Note: We’re able to verify their identities by matching them to state voter records.
Because Brigade is an open, peer-to-peer platform, we’ve seen significant inclinations toward different ideologies over time, ranging from a wave of Bernie Sanders backers who swept through during the primaries to a surge in Trump supporters during the general election. We believe these shifts are indicative of voter enthusiasm, and in the homestretch of this cycle, our user base leaned right and vote-pledging didn’t line up with national polls.
Fast forward to Election Night. States that had been considered winnable by Clinton fell to Trump and by early Wednesday morning, the U.S. had an unexpected president-elect. We quickly realized we might be onto something that the polls, pundits and campaigns themselves missed and our engineers started looking for some signals in the noise.
To account for the conservative skew from users who joined in the weeks leading up to the election, we normalized our user activity data and looked for unexpected trends. Here’s what we found:
Registered Democrats on Brigade were significantly more likely to pledge their vote for Trump in states where he beat pollsters’ forecasts.
For example:
- In North Carolina, they were 25% more likely to pledge for Trump and 15% more likely to do the same in Pennsylvania. Trump beat FiveThirtyEight’s widely cited forecasts in those states by 4.5% and 4.9%, respectively.
- In Kentucky, Trump beat FiveThirtyEight’s forecast by 11.8% and we had 85% more Democrats cross over to his camp.
- In Maine, Trump beat FiveThirtyEight’s forecast by 4.7% and we had 14% more Democrats pledge for Trump.
Conversely, in states where Clinton outperformed pollsters’ predictions, Democrats were less likely to cross over to pledge their vote for Trump.
For example:
- Registered Democrats on Brigade were 30% less likely to pledge for Trump in Nevada and Trump underperformed FiveThirtyEight’s forecast in that state by 1.2%.
- In New Mexico, Trump underperformed FiveThirtyEight’s forecast by 2.5% and we had 21% fewer Democrats change sides.
- In California, Trump underperformed FiveThirtyEight’s forecast by 5.4% and we had 60% fewer Democrats pledge for Trump.
What else do we know about Democrats on Brigade who crossed over to pledge for Trump?
Based on an analysis of their prior voting behavior, we know crossover Trump pledgers were significantly less likely than Democrats who pledged to vote for Clinton to have voted in a recent party primary or general election. This points to Trump’s ability to activate Democrats who usually don’t vote, and compel them to cross over to vote for him.
Democrats who pledged for Trump largely seem to align ideologically with his other supporters on Brigade. They choose to engage with topics like defense, immigration and gun rights on Brigade vs. civil rights, health care and education, which are the top three favorite categories among Democrats on Brigade overall.
We also know that in states with outcomes that didn’t match the polling results (i.e. where Trump did better than expected), we saw white women registered as Democrats pledge to vote for Trump on Brigade at a much greater rate (170%) than the country as a whole. While it’s a stretch to infer that white women swung the election in Trump’s favor, it’s an indication that Clinton had more trouble winning them over than she counted on.
Our team will continue to learn more about these crossover voters as well as nearly 200,000 other users whose voter records are integrated with Brigade. In the weeks and months to come, we’ll be studying our social GOTV (“get out the vote”) efforts — referenced here — to determine which voters on the platform followed through on their Election Day commitments and which candidates, messages and channels were most effective at driving turnout.
More on that (and other developments) soon!
— Thank you Aleks Mistratov, Stephanie Bian, Dave Kincade, Jim He, Hao Su, Shai Haim & Mark Cooper for your contributions to to this project.
