Measuring Swing Voters Using Supervised Machine Learning

• In Event: Understanding Vote Choice

Abstract

Theory has long suggested that swing voting is a response to cross-pressures arising from a complicated mix of individual attributes and contextual factors. Unfortunately, existing regression-based approaches are unable to explore the complex combinations of demographic, policy, and political factors that produce swing voters in American elections. This gap between theory and practice motivates our use of an ensemble of flexible machine learning methods to analyze swing voters in the 2012 and 2016 US presidential elections. As past work has demonstrated, machine learning methods excel in their ability to uncover complex interactions and nonlinearities present in the data generating process. I apply these methods to derive a more valid and granular measure of swing voting propensity than one would find from employing a more conventional probit regression model. I then use Shapley values and other feature elimination methods to tease out the estimated relationships between the predictor variables and swing voter propensity, demonstrating how machine learning methods can provide substantive insight into voting behavior.

Author

• Christopher Hare, University of California, Davis