• United States presidential race in 2016 was full of twists both in the Democrat and Republican preliminaries and Clinton-Trump battle after they have been named the nominees of the DNC and GOP.

  • Among hundreds of surveys and polls, very few had actually guessed the outcome of the elections. Up until the last days of the campaigns majority of the nationwide polls were showing Clinton ahead of Trump. The point differential was fluctuating significantly, though, from .2% to 7%, which made the race alive up to the election day.

• The race was so interesting and exciting that I wanted to do my own analysis.

Speech analysis of Clinton and Trump

• • I downloaded the debate transcripts for all three battles between Hillary Clinton and Donald Trump from Politico.

  • The first analysis I did was purely numerical as in analyzing number of sentences and words spoken by each candidate during these debates.

  • The table below shows that Clinton uses on average 3.5 words more in her sentences than her opponent. 

  • Trump prefers short sentences but that does not necessarily mean he was overshadowed by Clinton in the debates. On the contrary, in total, he spoke significantly more words than Clinton in each of their first two battles, while, Clinton appears to have outspoken Trump in the last debate. 

  • The last debate is the only one that analysts had not confidently declare her the winner. 

Word clouds

• • NOTE: For a statistically sound analysis, I ignored the top 100 words in the Brown Corpus (excluding a few such as information, contact, business, service, and find, which may be relevant to the topic of my research. I also excluded three other top 100 words he/him/his from the "ignore list", as we might relate those to Trump - Clinton's she/her were already not among the top 100 Brown Corpus words).

  • The word frequencies show that Trump had tremendous consistency over the course of three debates to steer the direction of the discussion to talk to people's nationalistic concerns, repeating particular keywords such as people, country, community, border, Iran, Assad, Mosul, ISIS etc. He successfully defamed Obama administration and everything they did during their tenure at the White House, with strong keywords such as bad, disaster, tremendous.  

  • Clinton, on the other hand, attempted to talk to women and American families in general. Seeing Trump's success in polls, she decided to address nationalistic sides of American public in the last debate, which appear to be too little too late.

  • Another note is that Trump rarely spoke of Hillary Clinton's name (except for the second debate) and consistently addressed her as "she" when he has to; while, Clinton addressed her opponent much more frequently and by using his first name.

Social media reactions to Clinton-Trump debates

• • After the first debate, I thought it would be interesting to investigate American public's live reactions to these debates. So, I decided to stream-collect some data from Twitter during the subsequent two debates in October.

  • I live-streamed just under 1.5 million tweets for my analysis. After eliminating retweets and obviously spam tweets, I ended up having a clean, ready-to-analyze, data set of 350 thousand tweets per debate.

  • For each debate day, these 350 thousand tweets produced about 5 million words to be analyzed.

Word clouds

• • NOTE: For a statistically sound analysis, I ignored the top 100 words in the Brown Corpus (excluding a few such as information, contact, business, service, and find, which may be relevant to the topic of my research. I also excluded three other top 100 words he/him/his from the "ignore list", as we might relate those to Trump - Clinton's she/her were already not among the top 100 Brown Corpus words)

  • I first generate a random map of words, where the font size emphasizes the frequency of the word in the data set. 

  • These word clouds are clear indications that Trump was talked about much more frequently than Clinton.

October 9, 2016

October 19, 2016

• • Political experts and voters argued that Clinton won the debate on October 9th, or it was a tie for Trump at best.

  • My opinion was just the opposite. I believed Trump had just achieved what he wanted in all three of his debates with Clinton. 

  • From the candidate panels in GOP preliminaries and his interviews, it was obvious that Trump aimed to make splash with everything he said. After all, he is a businessman and to him there is no bad publicity. He wanted to be talked about, and he did so in both debates.

  • Trump was the big winner of the final debate on October 19th, which is also evident in the corresponding word cloud.

  • His impact on the social media users is more pronounced in the picture. Not only do his names (donald, trump, realdonaldtrump; even he, his, and him) but also the topics he wanted to be emphasize (wikileaks, abortion, moderator etc) are highly mentioned. 

  Debate performances

• • We can also analyze candidates' performances throughout the debates via their twitter mentions.

  • Each debate lasted about 90 minutes and I recorded tweets from 10 minutes prior and ~20 minutes after the debates, for a total of 120 exact minutes in each debate.

  • Histograms show that Trump is - apart from several minutes at the beginnings of the debates - clearly the more talked-about candidate.

  • It is evident that Clinton prepared well for the first debate. However, she was not able to respond as well to Trump's attacks in time. 

• • Trump is a natural manipulator, and very good at improvising and twisting topics. He never let Clinton to take the attention away from him.

• Even at his last say at the end of the October 19 debate, when Trump implied that he would not respect the outcome in case Clinton wins, she could not use this to create a social media blast.