Program: Data Science Master's
Location: Not Specified (remote)
Student: Peter Torpey
This case study uses the North Carolina State Board of Elections data to predict presidential election turnout among new voters. The primary objective is to introduce the reader to the valuable information held within state provided voter files and explore methods that can be utilized to identify these voters in future elections. Of the over 890,000 new registrants in the state of North Carolina during 2020, a sample of 50,000 voters was selected at random and used for this case study. Out of the four models examined, Boosting produced the lowest misclassification error rate of 18.15% and its predictors were further analyzed. This information can be utilized by political campaigns, data companies, and other entities who specialize in political microtargeting to predict voter turnout.