U.S presidential election 2016 : what can we tell about public opinion from 140 characters?

Finding the extent of measurable benefits of social media on political outcomes is not easy. My paper aims to understand the relationship between Twitter data and public opinion for the U.S 2016 Presidential Candidates, Donald Trump and Hilary Clinton. This seeks to evaluate how user influence online (Twitter) translates to real-world or offline behavior (public opinion polls). I will consider the tweet engagement metrics of ‘retweet count’ and ‘favorite count’ of a candidate as a proxy for collective attention to him or her, identify the dynamics of the tweet engagement and how they are correlated to public perceptions. The key question is: Does the political use of Twitter have a positive correlation to public opinion? The methodology used is linear regression between two variables: tweet engagement metrics (independent variable) and Gallup’s Daily Ratings (dependent variable) from 20th July - 7th November 2016. Data sources used include (a) Public Twitter data from Twitter’s API (Application Programming Interface), (b) data mining and analysis using Python and Tweepy. (c) Regression analysis using Excel’s statPlus. In this study, I attempt to find valid and reliable data from tweet engagement metrics that can accurately reflect public sentiment. Based on my findings, there is a positive relationship between the two variables but the degree of correlation is weak.Bachelor of Art