Research

News-Related Social Media Use, Political Knowledge, and Participation in the 2016 Election

2018 - 2019

For my senior Honors thesis research at UNC Chapel Hill, I looked into the role that social media played in the 2016 election. I worked under Dr. Doug Lauen and Dr. Rebecca Kreitzer, both of the UNC Public Policy Department. I used Stata and R for analysis and figure-making. You can read the abstract below:

Social media’s role in the 2016 election was one of many aspects of that election which make it unique. Record numbers of Americans were active social media users in 2015-2016, and many used the platforms not only for entertainment, but to learn about the news. Social media were also used by candidates in novel ways and to unprecedented extents, and the quality of information distributed through social media came into question. Previous studies have shown that news-related social media use leads to increases in political knowledge and likelihood of voting, but the unique circumstances surrounding the 2016 election make generalization of those findings to that election perilous. Therefore, this thesis used national survey data from the 2016 Cooperative Congressional Election Survey to investigate the relationships between news-related social media use, political knowledge, and participation specifically in the 2016 election. Exploratory factor analyses, logit regressions, and marginal effects were used. Analysis found that consuming social media news content did predict higher likelihood of voting in 2016, and it found that that effect was partially mediated by an increase in political knowledge. Additionally, evidence suggested that whether someone shared, commented on, or forwarded news-related content – in addition to consuming it – was found to moderate each of the aforementioned relationships, though that finding was not statistically significant. The relationships between social media use, knowledge, and voting also varied by state and by party. Further research should consider personality in analysis and use experimental, rather than observational, methods to the extent possible.

Predicting the Fates of Bills from the 112th - 115th Congresses

Spring 2019

In this term project for UNC INLS 625 Predictive Analytics, I combined data from the ProPublica Congress API, govtrack.us, and DW_NOMINATE to predict what happened to bills from the 112th to 115th Congresses. I used Python, R, and KNIME to scrape, organize, and clean the data, then apply k-Means cluster analysis, RandomForest decision tree modeling, Naïve Bayesian modeling, and logit regression.

The link embedded in the title above will take you to my report, published using RMarkdown and GitHub. Below, you can watch the video(s) I made for the class, in which I describe each step of the project: data collection, preprocessing, and exploration, analysis, results, and conclusions. The first (and longer) video I made is at left; a briefer version is on the right.