Social media data provide a veritable playground of questions for those of us who pay attention to politics and policy.
- Are there identifiable patterns of who wins and loses in how campaigns use social media?
- Can social media level the playing field for candidates who have less money than their opponents?
- How important is social media, particularly Twitter, in defining an issue in the press and public opinion?
These are some of the questions that consultants, data analysts and candidates spend time and money trying to answer.
In February 2013, Indiana University completed a study (More Tweets, More Votes: Social Media as a Quantitative Indicator of Political Behavior) which found a statistically significant relationship between the number of mentions of a candidate on Twitter and the likelihood of them winning their election. We decided to see if the finding held true in the latest Texas election cycle.
Looking at the Republican primary results for the lieutenant governor race between Dan Patrick, David Dewhurst, Todd Staples and Jerry Patterson, we searched the number of mentions from Twitter users in Texas over the date range of October 1, 2013 to March 1, 2014 – just before the primary voting day. We found that the percent of Twitter mentions showed a high correlation to the percentage of the vote a candidate actually received.
*There is a correlation coefficient of 0.986, suggesting a highly dependent relationship. A correlation coefficient (r) shows how closely two sets of data are related and range from -1 to +1, where -1 would be complete opposites and +1 would be exactly the same. A score of zero shows that there is no relation between the two sets.
**Search set included the candidate’s full name and Twitter handle.
Next we wanted to see if this correlation held true for less visible runoff races. We checked to see the relationship between votes and Twitter mentions in less notable runoff races finding that in eight of the 17 instances it did. Many of the races affirmed the findings of Indiana University’s study including the following:
While these findings don’t prove that having more Twitter mentions causes a candidate to win an election, it does indicate a correlation between number of Twitter mentions and vote totals. The findings also underline the opportunity to analyze the unstructured data that is created when people and organizations use social media in politics and policy.
And, ultimately, the findings validate the idea that a compelling and strategic social media campaign is no longer a luxury; it is a necessity for any well-run campaign.
With early voting taking place today, we are interested in seeing if the runoff election between Dewhurst and Patrick proves to hold true; stay tuned for updates.