In this case study we aim to provide a new perspective into how language can be an important predictor of success or failure. To do so, we turned our attention to the U.S. Senate elections from November 2014 and analyzed Tweets sent out by the candidates. We then separated the winners from the losers to find out how their communication differed.
For this case study we have analyzed all the tweets sent out from January 1st to November 2nd (2014) from the accounts of various candidates who participated in the U.S. Senate elections.
According to the results of the elections, 5 of them were among the winners and 7 of them lost the elections. Statistics are presented in the following two tables:
|Senator||Account||Number of tweets|
|Cory Gardner (R)||@repcorygardner||378|
|Gary Peters (D)||@RepGaryPeters||198|
|James Lankford (R)||@RepLankford||320|
|Shelley Moore Capito (R)||@RepShelley||406|
|Steve Daines (R)||@SteveDaines||390|
|Senator||Account||Number of tweets|
|Bill Cassidy (R)||@BillCassidy||550|
|Bill Owens (D)||@BillOwensNY||147|
|Bruce Braley (D)||@BruceBraley||1450|
|Jim Matheson (D)||@RepJimMatheson||42|
|Jon Runyan (R)||@RepJonRunyan||187|
|Mike McIntyre (D)||@RepMikeMcIntyre||69|
|Mike Michaud (D)||@Michaud2014||511|
We were most interested to find out what characteristics did the messages of the winners have when compared to those on the losing side. We knew that if we could identify characteristics that were common to the group of winners and common to the group of losers, we could then easily compare the two sides. This would be extremely important because these characteristics could tell us a lot about the preferences of the voters and during the next election may be used as a guide for positioning the candidates.
Based on the analysis of 2956 tweets analyzed, our system calculated the occurrence of words for each of the evaluated categories and, after the election results were announced, winners and losers were evaluated relative to the established norms.
According to the analysis results for the categories representing the six basic emotions, we noticed that voters generally preferred the candidates who didn’t hold back when it came to expressing their feelings.
From these results, we can already draw two conclusions:
- The expression of emotions attracts voters. They prefer to see in front of them real people and not emotionless beings.
- A specific set of strongly expressed emotions – Fear, Sadness and Surprise – associated with criticism of the existing status and promises of change, which is quite natural for election participants, can play an important role in winning the hearts of the voters.
ATTITUDE (SEMANTIC DIFERENTIAL)
When measuring the attitudes expressed by each candidate we used the three main semantic differential scales: Positive-Negative, Active-Passive and Strong-Weak. These are the three main scales on which we unconsciously evaluate all people and phenomena of the world.
The very high values obtained in all of the semantic differential categories suggest that the winners of the elections demonstrated the best ability to navigate in the world, and this attracted the voters. Voters need orientation and preference for those who they offer these guidelines.
After analyzing the candidates’ tweets, one of the first things we noticed was the huge difference between some of the candidates when it comes to values that describe their communication style (representational systems).
The winners were constantly using all of the communication channels (visual, auditory, kinesthetic & rational) when communicating, while the losing candidates seemed unaware of this detail. This leads us to believe that the election winners were better informed about how to use all channels of information transmission and may have been working closely with professionals in order to be more influential.
When it comes to measuring the level of sincerity, the candidates who found themselves on the winning side used more clear language in their tweets and avoided hiding behind words that have little or no real meaning.
It is no coincidence that the winning candidates, who expressed their emotions more clearly, were also the ones to be perceived as more sincere, while the losing candidates, who seemed more emotionless, were detected by our system as trying to hide their truthful opinions.
The evaluation of the timeline categories reflects what we had already expected: that voters are interested in the future changes and generally prefer those candidates who talk to them more about what’s to come.
It’s also important to note that the candidates who lost the elections failed to paint a clear picture about the future that they intend to bring and focused most of their attention on the present.
The motivation direction categories reflect how the reader is being motivated: either by moving away from problems (Away From) or by moving towards achieving goals (Towards).
In terms of the direction of motivation expressed throughout the tweets, our analysis results show one major difference: the winners focused a lot more on creating a clear picture of the problems that exist and criticized more the current way of dealing with these problems, which they intend to help solve. This also explains why the winners obtained very high values for negative emotions (Anger, Fear, Sadness) and why it had a positive impact for them.
The perceptual positions results give us insights into the points of view that the senate candidates adopted. This information is very important when the goal is to influence the perception of the audience, or, in our case, when looking to understand how the winners successfully managed to influence the voters’ perception of them.
In our opinion, the candidates who were elected did three things differently from those who lost the elections:
- High values for first position: this means that they talked more about themselves and weren’t afraid to express their opinions and thoughts. They took responsibility for their views and showed themselves more as who they are. As a result, they were perceived as being better leaders.
- Very high values for third position: the winners also talked more about others, they used a lot of criticism and opposed other views and opinions more than those on the losing side.
- Very high values for fourth position: they connected better with the audience and talked more about what they can achieve together. By using more frequently the point of view of “we” or “us”, they created better rapport with the audience and shared better their vision with the voters.
After going through the text analysis results for every category, we can easily notice the differences in communication between the candidates who won the elections for the senate and those who lost them.
We would like to point out that, based on this comparison and also on our extended research in this field, our system can prove to be an important, as well as accurate, tool for achieving greater influence and getting better results. It can also be used to analyze texts in order to predict results or outcomes, and that is something which we will present in greater detail in a separate case study.