Random digraphs with given expected degree sequences: A model for economic networks

Building upon the growing interest for complex network theory, the main ambition of this paper is to contribute to a more effective application of network theory to economic phenomena, and particularly to financial networks. We depart from recent contributions on credit networks in two respects. In the first place, we constrain diversification with the out- and in-degree distribution of nodes, by adopting a suitable extension of the expected degree model for random weighted digraphs. In the second place, we focus on real networks by using this extension as null model for statistical analysis. With the help of statistical inference, we provide a rigorous answer to the following question: do real networks obey our null model? Further, we show that this answer is tightly connected to the existence of clusters or modules, as defined by Newman and Girvan (2004), within networks.

► We generalize the expected degree model in order to apply it to economic networks. ► We derive new statistical tests for economic networks, based on this theoretical model. ► Applying our tests to credit network data, we show these networks do not follow the model. ► These networks are poorly diversified and contain densely connected communities of nodes. ► Our work suggests ways to improve community detection algorithms for economic networks.