Saying social media is more accurate than polling--even polling that is considered out of touch--might be a stretch. But you could certainly say social media is differently accurate than polling, and that forecasters in future elections might do well to listen more closely for a signal in the noise of our feeds.

In the days leading up to the presidential election, Brigade, the voter engagement app backed by Napster founder Sean Parker, found itself in possession of data seemingly indicating a Donald Trump win when polls were still saying, "Nah." As of of Nov. 3, 78 percent of presidential vote pledges on the platform were for Trump. But even CEO Matt Mahan was surprised when Trump was elected president.

In the days since the election, the San Francisco-based startup has had a chance to comb through its data to find patterns in voting preferences of users. Crossover voting, specifically Democrats voting for Trump, appeared to play a role in swinging some states red, according to the startup.

Brigade has roughly 200,000 users verified as registered voters, plus additional users who are not verified. Users skew conservative. Among users registered as Republican, 94.5 percent pledged for Trump in the final months before the election according to Brigade, and 40 percent of registered Democrat users pledged for the candidate.

In North Carolina and Pennsylvania, users registered as Democrats were 25 percent and 15 percent more likely, respectively, to pledge Trump on the app. The candidate won both swing states. In Nevada, a swing state that went for Hillary Clinton, registered Democrat users were 30 percent less likely to pledge their votes for Trump.

Brigade's data also corresponded to Trump under- and over-performing in certain states as compared to polling by journalist Nate Silver's data website FiveThirtyEight.

In North Carolina, Trump beat FiveThirtyEight's forecasted voting rate by 4.5 percent; in Pennsylvania, by 4.9 percent. In Nevada by contrast, he underperformed the site's projections by 1.2 percent.

Trump also overperformed in Kentucky and Maine, states where Democrats were more likely to cross over, according to Brigade. In Kentucky he overperformed FiveThirtyEight's predictions by 11.8 percent and 85 percent more Democrats pledged for Trump. In Maine, he overperformed by 4.7 percent and 14 percent of Democrats pledged for Trump. Maine went blue, Kentucky, red.

In blue states New Mexico and California, where Democrat crossover was less likely than the national rate, Trump underperformed. In New Mexico he underperformed by 2.5 percent and 21 percent fewer users crossed over to vote for him, and in California he underperformed by 5.4 percent and 60 percent fewer Democrats crossed over than overall nationally among Brigade users.

Brigade spokesman Andrew Noyes notes, "One demographic tidbit has surfaced as we've begun probing deeper: In states with outcomes that didn't match the polling results (i.e. Trump did better than expected), we saw white women who were registered as Democrats pledge to vote for Trump on Brigade at a much greater rate (170%) than the country as a whole."

While Noyes says it's a stretch to read the data point as an indication that white women swung the election in Trump's favor, he suggests it relates to Clinton having more trouble in winning white women voters than her campaign apparently counted on.

Published on: Nov 14, 2016