Can Social Data Predict A Presidential Election? An Examination of the Iowa Caucus

Vote buttons

Phil Burch Phil Burch, Former Contributor

On Monday, Iowa got the US Presidential Election cycle into its next phase: Presidential Primary Season! As I outlined in my previous blog post, I wanted to explore primary results over the course of 2016 to see if social volume on Twitter and Facebook could outperform traditional polling data.

Since the election cycle will give this author 50(!) independent test cases, I am going to see if we can predict who will win the actual election in November based off of demographic insights, social volume, and share of voice.

For Iowa specifically, you can review the results from Monday’s caucus here, and here. Of the coverage, there were a few items I found particularly interesting:

  • Overall turnout was 15.70% of eligible voters participated (186,874 Republicans vs. 171,109 Democrats)
  • There was an increase of about 65,000 Republican voters that participated in the Caucus vis-a-vis 2012
  • Despite 45% of GOP participants stating that this was their first Caucus, Donald Trump ultimately did not win over a majority of these first time participants (30% to Trump, 23% to Cruz, and 22% to Rubio). Many commentators (including this author) are very surprised that Trump did not dominate this demographic of new participants
  • Marco Rubio had a very strong showing by winning 7 delegates. NPR cites that because Trump did not participate in the final debate, many GOP rivals went after Ted Cruz instead which helps explains Rubio’s surge in the Caucus
  • Bernie Sanders dominated the youth vote over Hillary Clinton (84% to 14%, respectively), but over 60% of caucus-goers were over age 45. Hillary dominated this demographic

 

What Did We Find on Facebook and Twitter?

For the social media analysis, I looked at Twitter data, Facebook data, and polling data from January 5 to January 31. Because Facebook now allows us to deeply understand demographic information via Sysomos Scout, I also found the most relevant insights through Facebook content.

As the results of the Democratic race showed, Bernie Sanders dominated Hillary Clinton in the 18-24 year old demographic (61% vs. 39% of mindshare on Facebook) and 25-34 year old demographic (57% vs 43%). However, when looking at older demographics, Clinton crushed Sanders – which does align with what the Facebook volumes suggest.

Share Of Voice Of Democratic Candidates Being Discussed on Facebook By 18 - 24 Year OldsShare Of Voice Of Democratic Candidates Being Discussed on Facebook By 25-34 Year OldsShare Of Voice Of Democratic Candidates Being Discussed on Facebook By 35 - 44 Year OldsShare Of Voice Of Democratic Candidates Being Discussed on Facebook By 45 - 54 Year OldsShare Of Voice Of Democratic Candidates Being Discussed on Facebook By 55 - 64 Year OldsShare Of Voice Of Democratic Candidates Being Discussed on Facebook By 65+ Year Olds

What Does This Mean?

As one would expect, voter participation does skew heavily to older voters. And as stated above, the huge edge Clinton had in social volume amongst older Facebook users suggests that Clinton was much more effective converting mindshare into actual votes.

If Sanders can continue to drive the youth vote, he may be able to defeat Clinton throughout the primaries. He will have to have a much more compelling message to “middle of the road Democrats” in these older age brackets, though.

On the Republican side, Facebook social volume does not necessarily reflect what took place in Iowa nearly as well. To some degree, Donald Trump’s notoriety and his ubiquity in the media are certainly having an impact on social volumes overall in Iowa (44% of ALL political conversations on Facebook in Iowa referenced Mr. Trump). While social mindshare was dramatically in his favor, he was only able to finish in second place in the caucus. This likely means that a significant amount of his social volume on Facebook is negative.

On Twitter, however, Trump’s Twitter volume from users in Iowa was a bit closer to how he finished in the caucus. He did have a majority of Twitter mentions, but this was roughly equivalent to the total breakout of Cruz’s volume and Rubio’s combined (12,417 for Trump and 12,874 for Cruz and Rubio).

Traditional Polling vs Twitter Volume for Republican CandidatesTraditional Polling vs Twitter Volume for Democratic Candidates

I will continue to monitor these results as we move closer to Super Tuesday and the party conventions. Based on our analysis in Iowa, there does appear to be some linkage between mindshare and turnout for the older demographic of voters.

As the Democratic race illustrates, older Facebook commenters appear to be more likely to convert to actual votes. Obviously, this means that these older demographics in social volume could prove to be a strong indicator as we continue to inch closer November.