Measuring and Modelling the Spatial Patterns of Firms: Integrating Spatial Statistics and Machine Learning for Firm-Level Analysis

This thesis is developed as a collection of three independent chapters/papers and studies how productivity differences emerge and persist across firms and places, with a focus on industrial clustering and agglomeration. The first paper devel- ops an empirical framework based on bipartite network representations of firms to characterize local productive structures of innovative startups in Lombardy. The second paper develops a deep clustering pipeline to perform bootstrap analysis of high-tech firms in Lombardy. The third paper links micro-level firm information to meso- and macro-level patterns of specialisation, the analysis identifies regularities in diversification and analyses their impact on the labour productivity.