How Big Data Got “Trumped”
“Torture the data, and it will confess to anything.” – Ronald Coase
The day after the 2016 U.S. presidential elections, many democratic and non-republican voters woke up with a feeling of dread and confusion. Questions of how this could have happened ran through their heads. Undoubtedly, they didn’t know many Trump supporters. Also undoubtedly, data swore that Clinton would win the election without fail. Many voters fell into believing both big data and the data they gathered from their day-to-day life. However, studies are increasingly showing that big data cannot always be trusted.
Donald Trump did not have to map data points to know that he was going to win. He, instead, used intuition. He tapped into the emotions of a large population of the United States and made promises of change that was particular to those groups. Where he won, and where Clinton supporters failed, was in two facets of the election. First, Clinton supporters and their constituents oversold the precision of big data. Their failure was to understand the very nature of statistics and their fallibility. To Clinton supporters, this data confirmed victory, but ignored the potential flaws.
A historically divided election distracted analysts and the general public from the real heart of the matter – that it did not update traditional statistical methods to reflect the difference in this election. Due to its uniqueness from any other election to this date, the data really should have been interpreted in more than just the traditional way. Trump’s election into office does not indicate that big data is dead, merely that humans often cannot predict the many outcomes of the data itself.
The United States election of 2016 was a unique event that probably could not have been predicted with our current data-driven mindset. For data scientists to succeed in the future, it is imperative that they come to a better understanding of human nature and sentiment.