Who Runs China? A Story Told by Machine Learning

Abstract

China’s politician selection process has been obscure and rarely discussed. This paper uses machine learning approaches to bring economic performance, political patronage and career path into one picture in understanding the provincial level politicians’ selection decision of China. Through mapping out the dynamic politician networks from 1995 to 2015, we find network communities are strongly determined by hometown ties. The predictions using machine learning algorithms including shrinkage and tree-based methods affirm the importance of economic performance, expand the conventional understanding of political patronage with the concept and measurements of social network, and point out the critical impact of career path, with largely reduced testing errors.