A year after we worked with big data analytics company ICG Solutions to see how the contestants were faring in the primary debates, the process started again with the Sept. 26 debate between the two candidates for president of the United States, Democrat Hillary Clinton and Republican Donald Trump.
As in the past, we measured the sentiment as expressed by users on social media while the debates were in progress, and during the hours immediately after the debates we studied a social media sentiment analysis provided QSocialNow, a consumer trend analyst company based in Argentina that uses a somewhat different approach.
In addition, we confirmed the accuracy of the social media measurements by comparing them with a set of polls taken by Politico and Morning Consult, two Washington, D.C., news organizations that have begun publishing fast-turnaround polls.
What’s important on the highest possible level is that all three measures agreed that Clinton came out on top of Trump in Sept. 26 debate.
One significant difference between the three measurements is that while both the QSocialNow social media sentiment analysis and the Morning Consult/Politico polls report the outcome of the debates as a moment in time, the ICG big data analysis tool, LUX2016 tracks the event over time during the debates as well as reporting the overall outcome. LUX was able to show the changes in sentiment as debate played out.
“What we noticed is that during the first 30 minutes of the debate we had a hard time seeing any real sentiment gap between the two candidates,” said Louis Lyons, chief operating officer of ICG Solutions.
Lyons said that the measurements using LUX2016 revealed that the volume of social media activity from backers of both sides in the debates was approximately the same—which was not the case in the primary debates a year ago, when Trump was far ahead in social media engagement.
Lyons said that after the first 30 minutes of the debate, Clinton began to pull ahead of Trump in positive voter sentiment. “We felt that it was a pretty even match in the first 30 minutes, but then it went in her direction across parties, genders and ages,” Lyons said.
He noted that the personal attacks by one candidate against the other had little if any effect on the overall sentiment, although the more Trump attacked Clinton, the more his positive sentiment among voters declined.
Lyons also said that the results revealed a positive voter response when the debates went back to covering issues and moved away from personal attacks.
The results from QSocialNow were less specific, but that’s because the data was sampled differently. Pam Baker, author of “Data Divination: Big Data Strategies,” who was my co-presenter for a presentation on big data analysis at a recent Excellence in Journalism conference along with Lyons, explained what was going on.
Social Media Analysis Shows Clinton Beat Trump in Presidential Debate
“The one analysis (QSocialNow) appears to be of sentiment alone (meaning it doesn’t appear to be pegged to defining specifics such as demographics, political party affiliation, likely voters, etc.), so it is more of an indication of Twitter user feelings during the event than of a read on voter segment sentiment. Think of it like a big data-fueled mood ring,” Baker explained in an email.
“The Lux2016 analysis is more sophisticated and more meaningful as it offers a breakdown by political affiliation (such as independents, GOP moderates, conservatives, etc.) and demographics (such as women) as well as trend tracking and gap spread measurements. There are more specific insights here, even though it is still a sentiment analysis,” Baker explained.
Baker also noted there are limitations to the insights that social media sentiment analysis can provide. “These early glimpses often correlate with the findings of more scientific and broader polls done later—but not always. Still, measuring public mood during a debate can and often is useful, not the least of which for politicians trying to hone their election pitches, ads or debate performances.”
In this case, the social media sentiment analysis did track closely with the polls, which reported 49 percent of respondents saying Clinton was the winner while 26 percent giving the debate to Trump.
So, the measurements all indicate that Democrat Hillary Clinton was the winner of the presidential debates on Monday. But what does that mean? Mostly it means that Clinton had a good night.
The measurements may also tell both candidates (if they’re paying attention at all) that the negative attacks on each other didn’t work in changing voter sentiment, but that discussions of policy did work. Perhaps this could mean that we’ll see more discussions about issues, although it’s doubtful we’ll see a significant reduction in personal attacks.
But on the broader question of what the sentiment analysis of the first presidential debate means for the results of the general election in November, the answer is, it may not mean much. There are two more presidential and one vice presidential debates yet to come and anything can happen at those events. And of many other political developments and “October surprises” can have a profound effect on the election results.
If you’re asking yourself whether all of this big data analysis on the mood and sentiment of the people viewing televised debates means anything, it’s good to reflect on something that Baker told me while we discussed the results.
“The point is you can’t equate social media analysis with scientific polling and research. They are not interchangeable. Social media analysis is limited to self-reporting,” she said. “There is data, and there is knowledge,” Baker added, “They’re not necessarily the same thing.”