Analysing mesoscale structures in economic and financial networks

Modelling, characterising and detecting the structure of complex networks is of primary importance to understand the dynamics of the systems considered. This is especially true for economic and financial networks, whose structural organisation deeply affects their resilience to shock propagation. Many real world networks are characterised by the presence of mesoscale structures: while a lot of attention has been focused on the community structure, many real-world networks are characterised by core-periphery, bow-tie and bipartite structures, especially so economic and financial networks. In this thesis we present new methods to model and detect these mesoscale structures. We apply these methods to characterise the structure of real-world economic and financial networks. Using maximum entropy networks encoding different levels of information, we model the structure of the international trade network and of national interbank exposures networks. We find that constraining local information is enough to reconstruct the mesoscale structure of these networks: hence, we introduce a new method to detect statistically significant bimodular structures, based on the connectivity within and between network modules. We also apply our method to multiplex networks. In particular, to unravel different types of corporate networks, we construct a new multilayer dataset of company interactions: we find that the disaggregate network describes a small corporate world, but that these different company interactions are characterised by vastly different topological properties.