Explaining Youth Driver Licensing Determinants Using XGBoost and SHAP

Abstract

This study explores the factors influencing driver's license acquisition among young individuals and examines its broader implications for mobility, safety, and sustainability. Leveraging nationally representative survey data on Millennials and Generation Z, we apply eXtreme Gradient Boosting (XGBoost) and SHapley Additive Explanations (SHAP) to identify key socioeconomic determinants of teenage driver's license attainment. Our findings reveal consistent predictors across both generations, including the percentage of licensed family members, household income per capita, educational attainment, and public transit ridership. We identify meaningful dose-response relationships, such as the increasing influence of licensed household members beyond a 0.75 threshold and the higher likelihood of licensing among individuals with some college or an associate degree. Additionally, household income exhibits a positive association with licensing within a specific range but declines at higher income levels. Beyond predictive accuracy, this study offers valuable insights into overcoming empirical challenges in transportation research through nonparametric machine learning models. Our findings provide a nuanced understanding of youth mobility behaviors, informing planning and policy strategies to support equitable access to driver education, multimodal transportation options, and sustainable mobility solutions.