Systemic Risk in the Financial System: Insights from Network Science

Ross A. Hammond


The recent financial crisis was characterized by a rapid widespread destabilization of the financial system, with little apparent warning. The crisis has led to fresh attempts to understand the structure of the financial system and what made it vulnerable to instability. One important characteristic of the system is the set of links between actors that form networks of connectivity. The structure of these networks may have enabled disruptions initially affecting only a few financial actors to rapidly spill over into a system-wide crisis.

Network structure may have played a role in the financial crisis in at least three ways:

1. Financial networks may have lacked robustness, hampering the ability of the system as a whole to continue functioning even when a few central actors stopped functioning;

2. The pattern of network links may have made the system especially susceptible to contagion (both through formal transactional links between institutions, and through informal social networks that connect both customers and individuals within institutions) ;

3. A lack of diversity in financial networks may have impaired their resilience, or ability to adapt to a new financial environment by reshaping themselves to recover functionality.

Over the past few decades, a large multidisciplinary scientific literature has developed around the mathematical study of networks—spanning fields as diverse as ecology, physics, sociology, and epidemiology. Although relatively new, this field has begun to amass a body of important findings and insights into how networks behave. In this paper, I give an overview of several key findings from network science, addressing in particular the three attributes of networks outlined above (robustness, contagion, and resilience). Where relevant, I will point out general lessons from this literature that may be of interest to regulators and reformers of the financial system—suggesting the types of structures or patterns that might be encouraged or avoided, and the kinds of data that might be useful to understand and monitor network stability.