Twitter Topic Modeling and the 2016 Presidential Campaigns Kelsey S. O Neill and Thomas W. Miller Northwestern University School of Professional Studies July 5, 2016
Introduction Many organizations today face a glut of data. In addition to records of business transactions, there are customer call reports and support requests, as well as data from the web and social media. Much of the data is text, and making sense of text presents a special challenge. Data scientists use topic modeling to explore large quantities of text. They use computers to analyze thousands of documents, identifying common themes. Northwestern University School of Professional Studies offers training in data science across a number of graduate programs. Research from one Northwestern program, the Master of Science in Predictive Analytics (MSPA), has focused on text analytics of the 2016 Presidential campaign. One master's thesis project from the Predictive Analytics program is analyzing Twitter data from the three remaining candidates going into the Republican and Democratic conventions. Thousands of Twitter documents (tweets) have been observed from the three remaining candidates from March through June 2016. Preliminary results suggest that discussion on the social medium Twitter may be organized around about fifteen topics. Topic Overview and Descriptions The words of the candidates on Twitter reflect an American- centric campaign with topics dominated by domestic policy. Among the remaining three candidates, the conversation is heavily influenced by Trump- initiated messaging and rhetoric. Here is a grouped list of topics. Topics within groups are sorted by prevalence. Topics Relating to the 2016 Presidential Campaign Hillary Clinton: Mostly Trump messages about Hillary Clinton, often preceded by the adjective "crooked." Trump Rhetoric: A variety of Trump- initiated topics and responses by others about those topics. These include criticisms of Elizabeth Warren, plans for immigrant bans based on religion or ethnicity, and the Trump University lawsuit or Judge Curiel. Republican Candidates: Trump mentions of Ted Cruz, John Kasich, and the Republican establishment. Trump Slogan: These tweets are dominated by Donald Trump blasts of "MAKE AMERICA GREAT AGAIN," his campaign slogan. Bernie Sanders: Mentions of Bernie Sanders, many initiated by Trump. Get Out the Vote: Targeted messages focused on primary elections across the states.
Topics Relating to Economic Policy Living Wage: Messages calling attention to American's working long hours at low wages and having little savings. InequalityTop One Percent: One of two identified topics dealing with income and wealth inequality. These focus on the top one percent of the population. InequalityPoverty: Another of the two topics dealing with income and wealth inequality. These focus on the middle class and those living in poverty. Jobs and Trade: Messages concerning the loss of jobs and tying the loss of jobs to trade agreements. Regulation and Control: Calls for governmental involvement of various kinds, including such diverse themes as gun control and financial regulation. Wall Street: Calls for more regulation of Wall Street. Topics Relating to Social and Environmental Well- Being Health Care: Calls for guaranteed health care for all, paid sick leave and vacation time. Public Education: Calls to expand governmental support of public education, including the idea of tuition- free college education at public institutions. Climate Change: Calls to act boldly in response to the worldwide threat of climate change. Analysis of Candidates Donald Trump Trump focuses on topics that are closely related to his campaign and competition (Republican Candidates, Hillary Clinton, Trump Slogan, Bernie Sanders). Additionally, he discusses jobs moving overseas, American trade and competition in the global marketplace (Jobs and Trade). Topics Lead by Trump Bernie Sanders Hillary Clinton Republican Candidates Trump Slogan Jobs and Trade
Hillary Clinton Hillary Clinton focuses on herself (Hillary Clinton) and Donald Trump, primarily drawing comparisons between the two as candidates in the Trump Rhetoric and Trump Slogan topics. Aside from Trump, she focuses on Regulation and Control. She also shares several topics with Bernie Sanders, most notably Public Education, Health Care, and Climate Change. Topics Lead by Clinton Regulation and Control Trump Rhetoric Bernie Sanders Bernie Sanders is more issues- focused than Trump or Hillary. He does not focus on himself or Hillary, although he discusses Trump Rhetoric and Trump Slogan a fair amount. He focuses heavily on Health Care, Living Wage, Income InequalityTop One Percent, Income InequalityPoverty, Education, and Climate Change. Sanders is also most prominent in the Get Out the Vote Topics Lead by Sanders Get Out the Vote Climate Change Wall Street Public Education Income InequalityTop One Percent Income InequalityPoverty Health Care Living Wage Comparisons and Contrasts Most of the 15 topics do have a strong association with one or more of the candidates. We can see that Trump is less issues- based with his Twitter campaign, and more people- focused. He specifically mentions Republican Candidates Ted Cruz and John Kasich, and that topic (Republican Candidates) peaks during the months of March and April when they were still considered to be his competition. Cruz and Kasich dropped out of the race on May 3 rd and 4 th, respectively. Similarly, during the months of May and June, he focuses on presumptive Democratic nominee Hillary Clinton is shared with Trump discusses Bernie Sanders and
supporters showing up and causing trouble at his events, and threatens that he will send his protesters to distinct topic among Trump Slogan). Trump Rhetoric proves to be driving the conversation as the other candidates find reason to discuss and counter it (in the months of May and especially June). As pillars of his campaign, Trump talks about domestic job loss, trade agreements and America as a competitor abroad. He mentions Obama frequently and discusses his foreign policies and his failure to call Hillary Clinton and Bernie Sanders (as Democrats) share several topics including Public Education, Health Care, Living Wage, and Climate Change. Bernie Sanders is the most issues- focused candidate (as he is prominent in 7 out of 9 issues- based topics), although his topics are interrelated and all focus around income inequality and equal opportunity. Hillary has a larger focus on issues than Trump, but she spends more time discussing herself and Trump than Bernie does. Hillary and Trump seem to turn towards each other specifically in the months of May and June as they both see themselves as the presumptive nominees for their parties.
Topics Shared by Two Candidates Climate Change Hillary and Bernie Public Education Hillary and Bernie Health Care Hillary and Bernie Living Wage Hillary and Bernie Hillary Clinton Hillary and Trump Topics Shared by All Three Candidates Trump Rhetoric Trump Slogan Get Out the Vote Jobs and Trade Regulation and Control Topics More Prevalent at Certain Times in 2016 Republican Candidates March, April - Trump InequalityTop One Percent June - Bernie Trump Rhetoric June Hillary, Bernie Hillary Clinton May, June Trump, Hillary Living Wage May Hillary, Bernie Conclusion What distinguishes political analysis coming out of the Northwestern program is that it is data- driven. Identified topics are not the opinions of political analysts or experts. Rather, topics reviewed in the Northwestern research emerge from the words of the candidates themselves, words posted to social media. It is text data and computer algorithms that discover general topics and themes. This is political data science, empirical research rather than political commentary and expert opinion.