Deep Learning in Banking

Chapter 5Financial Contagion and Network Models

Summary

After reading this chapter, and going through the case study, you will be able to:

• Understand the role that networks play in financial contagion.
• Know how to structure networks from various types of data and measure the properties of nodes in a network.
• Understand the theory and main components behind various types of graph neural networks.
• Develop a financial contagion model with graph neural networks.

Everything is connected, and that is especially true for the banking system (Jackson, 2008). The connections are sometimes clear, such as monetary transactions between companies or shared ownership, or they can be disguised when two or more companies are subject to the same market conditions. Financial institutions lend money to each other and organizations, and the world’s countries are connected in an intricate web of sovereign debt. If one of the entities experiences a shock of some sort, the impact of that shock will affect those connected to it and even the entire network. This is known as financial contagion, or spillover risk, which is the spread of economic or financial shocks from one market, country, or region to others (Gai and Kapadia, 2010; Elliott et al., 2014).

However, connections do not need to be direct or actual, such as transactions or loans, for risk to spread, as entities may be subject to the same risks through shared features. For example, two borrowers living in the same area or neighborhood ...