Machine learning firm dynamics: failure, success, and growth

The thesis explores the application of cutting-edge, tree-based machine-learning techniques to analyze and predict the dynamics of firm failure, success, and growth. It is structured around three key areas: the detection of non-viable firms, the analysis of factors driving the future success of startups, and the examination of firm growth dynamics while addressing methodological limitations. Given the endemic missingness affecting firm-level data—often highly informative in prediction—particular attention is dedicated to the handling and usage of missing values within predictive frameworks. By integrating predictive analytics with empirical insights, the thesis seeks to enhance understanding and transparency in firm dynamics, offering insights for both academic research and practical applications.