Worldwide, many people were shocked by the outcome of the U.S. presidential election of 2016. It’s been just over one full year since Trump’s electoral win and, although the race between Trump and Clinton was close, many of the electoral polls forecasted Clinton as the likely winner of the 2016 election (Perez, 2016). While the outcome of said election left many data scientists confused, many social media analytics firms are claiming that their measurements would have been a much better predictor (Perez, 2016).
Why were the electoral predictions so far off? Why did traditional prediction models fail to foresee Trump as the next president of the United States? The key, according to many social media analysts, was the large silent voter base that surrounded the Republican nominee. For better of for worse, Trump effectively leveraged social media throughout his campaign to establish press presence and, although Clinton outspent him in almost every traditional avenue, he seemed to have the internet under his right wing (Perez, 2016).
One of the analytics firms that claims to have seen it coming is Simply Measured, a social media metrics firm that specializes in monitoring (listening to) consumers to determine trends and collect meaningful data (Simply Measured, n.d.). Simply Measured collected a significant amount of data prior to and during the election in an attempt to make electoral predictions based on voter sentiment.
In the month leading up to the election, positive sentiments for Trump were, on average, much higher than they were for Clinton. Perhaps most important was the fact that Trump’s positive sentiment on the day of the election (November 8, 2017) was substantially higher than that of Hillary Clinton’s. On this day, Simply Measured software uncovered about 175,000 instances of positive sentiment toward Trump, but only about 120,000 instances of positive sentiment for Clinton.
Additionally, Simply Measured collected the same data trends for negative sentiments. Of course, Trump produced quite a large volume of bad press throughout the race. Some of Trump’s most controversial events can be seen in the chart below, indicated by the peaks in negative sentiment; however, it is important to note that there was a negligible disparity between negative feelings for Trump and Clinton during the week of the election. During the last few days prior to the election, each candidate had about 15,000 instances of negative sentiment based on Simply Measured’s data collection algorithms.
If we use both the positive and negative sentiment data in combination, we can produce a ratio of positive:negative sentiment for each candidate. On election day, Trump’s positive:negative sentiment ratio is 175,000:15,000, or about 12:1. On the same day, Clinton’s ratio is about 120,000:15,000, or 8:1. With this information, we can deduce that, based on all online press, Trump seemed to be generating more positive impact with voters than Clinton.
After reviewing these findings, we can return to the theory of silent voters, which helps to explain how traditional metrics may have painted an inaccurate portrait of the pre-election sentiment. Sarah Perez of Tech Crunch theorizes that the reason that the estimates were incorrectly skewed in Clinton’s favour may have been due to Trump supporters’ greater propensity to keep their sentiments private due to it being seen as taboo to support Trump (Perez, 2016). As a result, and clearly illustrated with the help of social media analytics like Simply Measured, it was not until a few days before the election that Trump’s supporters became more vocal and active, painting a clear and accurate picture of the likely outcome of the 2016 presidential election.
Lessons for Others
Trump’s surprising win of the 2016 election is a perfect example of how traditional sources of data collection and processing can be insufficient or inaccurate on their own. Social media metrics can be used in combination with other data collection methods in an attempt to unveil hidden information and provide a clearer outcome, especially as consumers move more of their daily activities online.
Industry: Social Media Metrics
Name of Organization Contact: N/A
Authored by: Lucas R. Coady
If you have concerns as to the accuracy of anything posted on this site, please send your concerns to Peter Carr, Program Director, Social Media for Business Performance.
Perez, S. (Nov 10, 2016). Analysis of social media did a better job at predicting Trump’s win than the polls, TechCrunch. Retrieved from https://techcrunch.com/2016/11/10/social-media-did-a-better-job-at-predicting-trumps-win-than-the-polls/
Simply Measured. (n.d.). Full-funnel social analytics: Insight to help you create, connect and convert. Retrieved from https://simplymeasured.com/#sm.0000gtfpm6kzbdrmzfk1nqld48y6v