How does home-based teleworking reshape commuting and activity participation? An empirical study in Shanghai using machine learning

Despite the growing adoption of teleworking, empirical studies remain limited in developing nations, including China. However, the rapid rise of teleworking, accelerated by the COVID-19 pandemic, highlights the need to understand its impacts. This study examines how teleworking influences commuting behavior and time reallocation, identifying key factors and their effects. Using survey data from Shanghai, we first analyze commuting differences across teleworking patterns, revealing significant effects on travel modes, departure times, weekly commuting days, and trip-chaining complexity. We then develop XGBoost models to investigate how individuals reallocate saved commuting time across four activity types: in-home, out-of-home mandatory, leisure, and maintenance activities. SHAP and Partial Dependence Plot analyses identify four core factors—age, daily Internet use, commercial housing density, and company density—as critical determinants of activity choices. The findings suggest that while teleworking may reduce commuting trips, it primarily restructures travel patterns rather than simply decreasing trip frequency. These findings will help transport planner develop adaptive measures to accommodate evolving mobility behaviors in an increasingly telework-oriented society.