Originally posted by Jordi Zamora ’15 on the Barcelona GSE Data Scientists blog.
The study focused on understanding the underlying structure of a network of messages between financial institutions in different countries. It looked at how the network was affected by various recent economic events and evaluated the robustness of the system over time.
The data set underpinning the study contains standard MT103 SWIFT messages from 1 January 2003 and 31 July 2013, a period characterised by extreme economic turmoil. Each message represents a single customer credit transfer from bank to bank. The data is aggregated at the country level.
Samantha showed us different statistical analyses of the data set. The analysis of the data in terms of a complex weighted network was particularly interesting. In the network, each node represented a country and the edges connecting two different nodes were weighted according to the amount of messages those country exchanged in a given time period. The resulting network follows approximately a Core-Peripheral structure, that is, some nodes are fully connected with each other (the so-called core) while some others are mostly connected only to a node of the core: these are the peripheral nodes. Interestingly, events such as the introduction of new regulations or the beginning of the financial crisis was clearly reflected in the links and even more striking this network structure was resilient during the period studied. This work showcases a novel approach to understanding the structure of the complex financial system and the findings may provide a way to help improve the global service.
The discussion also identified some opportunities for further research. For example, we discussed why the degree distribution does not behaves as other related financial networks, and why the number of links decreases while the number of messages has a clearly increasing trend. These questions, and others that emerge, may provide ideas for further research and modelling work in this area.