That’s the question being asked around the world as shocked commentators try to make sense of Donald Trump’s win. While, no doubt, many will talk about a divided America, the angry Midwest, and the trend against globalisation, there is one fact that is now abundantly clear. The 2016 US Election is the most polled event in history, and yet those polls failed to predict the outcome. Instead, accurate social media analysis delivered the correct prediction, and can no longer be ignored as the key to understanding the voice of the people.
The polls’ poor performance in the US elections is not an isolated event. They consistently and confidently called a Remain win in the UK’s Brexit referendum, while DataEQ’s social media analysis warned of a majority looking to leave the EU. Whilst the world that relied on polls was surprised, DataEQ was not.
Polling is flawed in a number of ways. To start with, few outsiders realise that many polls contain no more than a thousand participants. While statistically this should be enough to gauge a much larger population, it takes only a small number of people to introduce a significant bias. Consider also how this bias is introduced. Many polls are telephone polls, yet many Americans no longer possess landlines. The lists of participants are often dated and tend to be predominately urban. Behavioural psychologists have documented that, when asked by a stranger for an opinion, people often lie or dilute their position, uncomfortable with their own views or prejudices. Significantly, polling cannot measure anger, enthusiasm or the “volume” of commitment expressed in the answers. These emotions determine the likelihood of people acting on their views on Election Day. And most of all, polls take time to conduct, compile and re-weight, leading them to make predictions on old news. There is good reason why candidates campaign feverishly up to Election Day - the events and momentum of the last few days going into the voting booth really do matter.
Social media seems to have come of age as being increasingly representative of the electorate at large and no longer the domain of the angry and extreme. As such it has three great advantages in being used to augment and build better polling methodologies into the future.
One, the number of participants is much, much greater with millions of people voicing their views. DataEQ, for example, tracked over 37.6 million conversations involving approximately 4 million individual authors during October and the first week of November.
Two, these views are unsolicited and spontaneous and represent an undiluted vote for or against a candidate.
Three, they can be captured and measured almost immediately, giving an idea of how people are feeling right now. The key is measuring them accurately, and this is the skill DataEQ has mastered.
By using a massively scaled crowdsourcing approach, with humans verifying expressed sentiment, DataEQ are able to look accurately at sarcasm, innuendo, implied meaning and varied language to produce a result that is now proven to be reliable, representative and trustworthy.
In the weeks leading up to both the Brexit referendum and the US Election, that crowdsourcing approach made one thing very clear - the polls were not measuring a large swathe of the electorate who felt disenfranchised, left out and angry. Sentiment about jobs, immigration, globalisation, and the perception that the US was run by an out of touch and exclusive elite, was overwhelmingly expressed on social media.
For example, in many key US battleground states, DataEQ was measuring 9 authors pro-Trump or anti-Hillary to one against or for her. Michigan, where almost all polls predicted a Democratic win, was trending particularly strongly in the other direction. The strength of this data was impossible to ignore, and DataEQ called a Trump win against almost all major political commentators, just as they had with Brexit. And, as with Brexit, accurately measured social media beat traditional polling and called the correct outcome.
The world should not have been surprised, and the key players could have known far earlier how people really felt had social media been included and measured accurately.