When we started to monitor the Presidential Election in July, we quickly saw a significant divergence between the traditional polls and what the Eyesover system was reporting. As early as August, the Eyesover system was consistently reporting that Trump had an advantage in support over Clinton.
The two driving factors for this result were:
1) Trump consistently garnered 2-3 times as many mentions online as Clinton. This factor is important since there are numerous studies, in addition to our own research, that conclude the number of mentions a political candidate receives online is positively correlated to the candidate’s actual support on election day.
2) The general sentiment level towards the two candidates was very similar (albeit negative). It is rather obvious, and our research agrees, that sentiment towards a candidate is positively correlated with election day support.
While Trump's support held firm through to Election Day, Eyesover picked up a significant surge in Clinton’s support and sentiment during the last week making the popular vote a dead heat.
This surge coincided with the deterioration of the Libertarian support, indicating that as voters were making their final decisions, many who had been looking at the third parties as an option, decided in the end to vote for Clinton.
In terms of the Electoral College, Eyesover found Trump had an unexpected lead in Michigan, a small lead in Florida with the swing states of North Carolina, Virginia, Pennsylvania, and Ohio all within 4% or less. When we allotted a State’s Electoral College votes to a candidate with a lead of 4% or more, we projected 195 for Trump, 188 for Clinton, and 155 votes in states either too close to call, or we did not have enough content to make allow the system to make a projection. In other words, it was clear there was a path to victory for Trump.