Analysing mesoscale structures in economic and financial networks

Modelling, characterising and detecting the structure of complexnetworks is of primary importance to understand thedynamics of the systems considered. This is especially truefor economic and financial networks, whose structural organisationdeeply affects their resilience to shock propagation.Many real world networks are characterised by the presenceof mesoscale structures: while a lot of attention has been focusedon the community structure, many real-world networksare characterised by core-periphery, bow-tie and bipartite structures,especially so economic and financial networks. In thisthesis we present new methods to model and detect thesemesoscale structures. We apply these methods to characterisethe structure of real-world economic and financial networks.Using maximum entropy networks encoding different levelsof information, we model the structure of the internationaltrade network and of national interbank exposures networks.We find that constraining local information is enough to reconstructthe mesoscale structure of these networks: hence,we introduce a new method to detect statistically significantbimodular structures, based on the connectivity within andbetween network modules. We also apply our method to multiplexnetworks. In particular, to unravel different types ofcorporate networks, we construct a new multilayer datasetof company interactions: we find that the disaggregate networkdescribes a small corporate world, but that these differentcompany interactions are characterised by vastly differenttopological properties.