How did Donald Trump outperform his polling numbers so handily with practically no ground operation on Election Day? User activity patterns on Brigade, Sean Parker's civic-engagement app, suggest one possible explanation.
During the primary season, Brigade ran a "peer-to-peer vote recruitment test" to measure how well supporters of each candidate were able to communicate their enthusiasm to others. The startup called that quality "virality" and quantified it with a "viral coefficient." It turned out that support for Trump and Bernie Sanders was approximately three times more 'viral' than support for Hillary Clinton.
As of the week before the general election, that was still very much the case. Trump's coefficient heading into November was 0.37, according to Brigade, and Clinton's was 0.12. That means every 100 users who pledged they would vote for Trump were able to persuade 37 additional users to also pledge for Trump.
When Brigade CEO Matt Mahan first cited the profusion of pro-Trump activity on his app to Inc., he was bringing it up as an indicator of how Trump users engage with social media and how the app grows usership. He wasn't at the time describing the data as a projection of election results.
"It's an open product," he said last week of the platform. "Once it takes off within a given community, it'll grow like crazy in a particular area."
But on Tuesday night, as Trump's win surprised Mahan and wetted the eyes of a largely liberal crowd of San Franciscans at an open-bar election night watch party in the startup's SoMa district warehouse office, reporters started getting in touch with Mahan, apparently to ask what Brigade may have tapped into that polls had overlooked.
Brigade had more than 170,000 users whose voter registration was verified as of last week, plus hundreds of thousands of accounts for non-verified users. Brigade reported as of November 3 that 78 percent of presidential vote pledges on the platform were for Trump. Polls told a different story, putting Hillary Clinton in the lead. Nate Silver's data website FiveThirtyEight gave Clinton a roughly two-thirds chance of winning on November 3, and on Election Day a roughly 71 percent chance of winning the presidency.
After taking a call in a conference room late Tuesday night as partygoers started leaving, Mahan acknowledged Brigade users' pledges may have been more indicative of the election outcome than the startup had imagined.
"But I'm also trying to temper expectations," he said, explaining that if Brigade had made a projection on the basis of its users' pledges, the company would have predicted a landslide victory for Trump. In fact, the election was close and Clinton appears to have won the popular vote.
"Our platform has looked surprisingly skewed to the right," he said. (A September post on liberal blog Daily Kos suggested Brigade was "exploited by the right.") "People have commented on it and asked why it's skewed."
Now it appears Brigade's platform isn't necessarily "skewed" or "exploited." The platform may be reflective of trends in sentiment and behavior. Mahan described the environment on Brigade as indicative of an enthusiasm gap, something he pointed to last week.
Twitter also consistently saw Trump take a greater share of voice among users than Clinton. Roughly two-thirds of conversation on Twitter about the second presidential debate was centered on Trump, according to statistics website Statistica.
In an email, Andrew Noyes, Brigade's director of communications, said that enthusiasm for Trump continued right up through the night of the presidential election, with Trump supporters "disproportionately more active than Clinton supporters on Brigade." In addition to pledging more votes, the users were more likely to append explanations to their pledges and to recruit others to support their chosen candidate.
"Our data team has also uncovered a disproportionately high number of registered Democrats on Brigade who pledged for Trump and a small number of registered Republicans who pledged for Clinton," added Noyes.