Republican Senate Takeover Was Predicted Through Social Data


By now you’ve probably heard that Republicans won more than enough seats to take over the Senate, just as we predicted last week. Our swing state predictions were correct for four of the six states: Kentucky, New Hampshire, Iowa, and Georgia.

Our prediction for Kansas was that the Independent candidate, Greg Orman, would win. Instead, Republican Pat Roberts swept all but three counties in the state, though Orman was a close second in that race.

We were also fairly certain (65%) that Democrat Kay Hagan would have a win in North Carolina, when instead, Republican Thom Tillis won the Senate seat. Hagan was approximately 2% points behind Tillis, which still shows the reflective power of social data when it comes to figuring out how voters may lean on Election Day.

Georgia Prediction

Last week, we predicted a win for Republican candidate David Perdue in Georgia, based on total shares, velocity of sharing, and sentiment of the content. It turns out, Perdue won, and our data was almost dead on.


AddThis Prediction Actual Results

Android & Twitter Ruled Election Coverage

Aside from the county-by-county voting results, we also looked at how people shared and clicked on election-related content. Here are some interesting nuggets we found:

  • People were 32% more likely to engage with election content on mobile than on desktop
  • People were 26% more likely to use and Android device versus iOS device to engage with content about the election
  • Politically conservative people are 25% more likely to use Android than iOS
  • The news of the Senate win was 10% more likely to be spread on Twitter in the U.S., but the rest of the world was 20% more likely to use Facebook