Image courtesy of Pixabay.
On November 5, 2024, American citizens will cast their votes in the presidential election. Casting a ballot is inevitably influenced by the weeks and months of candidates’ campaigning that precedes it: commercials, billboards, yard signs, and mail displaying Harris’s and Trump’s policies and beliefs. But how much do we really know about how campaigns are designed and the scientific strategies that inform them? Joshua Kalla, an associate professor of political science and statistics and data science, teaches a course called “Data Science for Political Campaigns,” an interdisciplinary class offered every fall that explores the structural and scientific underpinnings of electoral campaigns and how they drive voting decisions.
Kalla’s course approaches both past and present political campaigns with the rigor of data science analysis. As an introductory course, it is meant to be an introduction to data skills, computational social science, coding, and statistics, as well as an application of those topics to the world of elections and campaigns.
Before introducing this course, Kalla worked in data analytics for political campaigns and researched how campaign tactics can impact voting decisions. “One of my colleagues once told me that a lot of my research ruined cocktail parties,” Kalla said. People have inflated expectations about how influential the things that they’re working on—what they are funding, what doors they’re knocking on, or what TV ads they’re buying—really are, he explained. “But a lot of my work is about how most of what’s done in American campaigns tends to be minimally to not effective at all,” Kalla said. This revelation about the minimal impact played by election campaigns is just one of the many insights students are able to garner.
In this course, students study polling, create representative samples for surveys, analyze election results, and explore persuasion techniques. One topic Kalla examines is likely voter screening. In a campaign, the only important polling demographic is the people who actually vote in the election, otherwise known as likely voters. Data scientists can’t simply poll everyone who claims they will vote because, in reality, these predictions often fall short. Therefore, samples for polling are built based on voters’ history in order to minimize the chance that a non-voter gets a say in a poll that won’t impact them.
Beyond polling, accuracy in surveys is another topic that arises frequently in the class and is a fundamental aspect of understanding political data science. Politicians rely heavily on public surveys to gauge voter satisfaction. Because of this reliance, students consider how details of a survey, including how questions are phrased, can critically impact responses. Using non-leading questions, data scientists can gather more unbiased data, allowing politicians to accurately analyze the effectiveness of their campaigns.
Kalla’s 130-student course meets once a week and offers an interactive environment filled with student engagement and excitement. Each 110-minute class meets in a half-lecture and half-group lab format. “The labs are a way for them to practice those coding skills and the substantive skills of what we’re covering in class,” Kalla said. The final project prompts students to find real data in the world of politics and answer a question about it using the data analysis and coding skills that they’ve learned throughout the class.
This class allows students interested in politics to engage with it in a data-intensive way, serving as an entry point into the computational social sciences. “And on the flip side, there are a lot of students who come from a statistics and STEM background who are craving to learn more about applications of the kind of stuff that they’re learning, and this is a great opportunity for them to do that,” Kalla said. So, whether you come from a background in social science, statistics, or STEM, PLSC 347 acts as a great meeting point for coders, STEM enthusiasts, and passionate politicians alike. Because who knows? Maybe you, with the use of data science and Python 3, could help elect the next president of the United States.