Tech Perspectives

Network Analysis: How to Mitigate Unexpected Risk

Understanding the interconnectedness of different actors in the financial markets is essential in assessing aggregate market, credit and liquidity risks. Risk managers can use this knowledge to map out anomalies, concentrations and other outliers that can cause the disruptions that precipitate unanticipated risk events.

Friday, May 28, 2021

By Peter Went


Some of us hate it, others love it, but at least we can all agree that networking is a very important business skill. The ability to accumulate connections that provide relevant business information and enable access to the right type of insights are among the keys to a good network.

This business-networking approach also applies to financial networks. By depicting and analyzing their complex and maze-like makeup of linkages, financial institutions can gain an understanding of how networks operate - while simultaneously reducing the risk of unforeseen events.

Financial networks can be described as a collection of nodes (actors in the financial markets) and their bilateral links (transactions and exposures), which connect actors both directly and indirectly. Actors are the typical financial market participants, ranging from systemically important financial institutions and asset managers to central counterparties, exchanges, and the individuals accessing these markets. The bilateral links are determined by exposures from trading, business, or a combination of multiple (mutual) relationships.

Peter Went

Network analysis provides a framework for analyzing the magnitude, direction, change, and dynamics between these agents and their exposures. In network analysis, there is a difference between exposures and connections that are direct and indirect. A straightforward example is the interbank market, where buyers and sellers are transacting with each other on the assumption that the other side performs as agreed.

The ability to settle direct bilateral transactions, as agreed, is, in fact, critical to the smooth functioning of the financial markets. Settlement failures can easily yield negative ripple effects that can spread widely among multiple indirectly linked participants in the interbank and other markets. What's more, they can potentially precipitate widespread market disruptions - which can, in turn, lead to market panics.

Understanding the structure, components and actors in any financial network can help mitigate contagion - via, for, example, providing risk managers with the tools needed to comprehend connections and the various patterns of relationships.

Analyzing Networks

Modeling of the exposures in the system, and across its participants, is a critical exercise. This is the best way to identify the reach of an institution (as a function of its exposures) and to capture a fuller picture of the complexity of its transactions.

There are several approaches in network theory that can guide us in the mapping, aggregating and modeling of these financial networks. In the interbank market, for example, interactions can be modeled based on linkages that are created mainly through a series of short-term transactions. Banks that practice this type of modeling can reassess their exposures on a daily basis, as they trade with other institutions.

Directional changes in the exposures of a dynamic financial network reflect daily deal and transaction flow, as well as the repetitive nature of interactions across connections, links and agreements.

Financial network analysis cannot be easily conducted in isolation, at the individual institution level, as data sourcing is limited to the aggregate bilateral exposures where an institution has a direct relationship. Moreover, because of confidentiality rules, accessing exposure information across other (indirect) counterparties is not always possible.

Regulators, however, can help resolve this issue. Through their oversight and enforcement functions, regulators have access to confidential information about all exposures. Moreover, they can map out financial networks.

Besides regulators, the only entities that can aggregate direct and indirect exposures (providing a network view of the financial systems) are clearing houses, trading venues, payment systems and settlement systems.

Given these restrictions, how can risk managers at individual institutions model the interconnectedness across networks and estimate the exposure levels for each market participant? Moreover, how can they quantify the bilateral exposures between market participants, assess the aggregate level of multi-lateral exposure for market participants, identify potentially critical exposures, and assess potential mitigants (collateral, etc.) to reduce the level of exposures against individual market participants? Herein lies the critical challenge.

The simplistic answer is simulation of dynamic financial networks, leveraging the results from the vast academic and practical literature. Risk simulations do not only assess relationships (and their magnitude), but also identify critical exposures. They borrow liberally from physics and epidemiology methods (to describe the formation, structure, evolution, and dynamics of networks), and the results of such simulations provide insights to the centrality of actors in the financial networks.

The key actors are usually the large (systemically important) banks, financial institutions and financial institutions that are interconnected globally across multiple financial networks. However, other non-bank financial institutions (like asset managers and insurers) can also play important roles in different networks.

The Importance of Diversification

Generally, we do know that increasing the number of connections between actors reduces the risk that the failure of one specific financial actor in the network will bring down other parts of the financial system. In fact, the higher the number of the interlinked connections in a system, the more likely that the relationships mitigate the risk of contagion between actors.

Institutions can play a key role in the network by intermediating transactions and exposures, thereby sharing and effectively redistributing the aggregate risks across the system. By spreading out risks across more parties, central actors can make an interconnected financial system more resilient to random shocks.

Shocks, of course, can impact individual financial actors by driving rapid market liquidations that depress prices and distort both market (i.e., trading) and funding liquidity. However, with a diversified approach, these shocks are not difficult to prevent. Diversification, in short, is key.

There are multiple established use cases for network analysis and central counterparty exposure modeling. The network modeling of large value payment systems, for example, has provided valuable insights into the scope of complex bilateral exposures between banks and other financial actors participating in these networks. The scope of these payment network studies also encompasses global payment systems, such as SWIFT.

For regulators monitoring the potential for contagion in the system, understanding transaction flows (for compliance and money laundering purposes) is particularly important.

Parting Thoughts

In response to the pandemic, many of us have become overnight armchair epidemiologists, monitoring and assessing the unexpected transmission of various COVID-19 mutations across the globe. To gain a deeper understanding of the net worth of our financial networks, we should adopt a similar approach to the complex analysis of the connections between financial institutions and markets.

We know that systemic risks emerge from different directions across financial networks. The causes of these risks are often interconnected and highly correlated. That's why it's extremely important for market participants to gain a better understanding of the network of linked exposures. Indeed, this is a key component in comprehending the extent of the actual exposures across markets.

Peter Went (PhD, CFA) lectures at Columbia University on disruptive technologies, such as artificial intelligence and machine learning and their impact on financial services and financial risk management.


BylawsCode of ConductPrivacy NoticeTerms of Use © 2024 Global Association of Risk Professionals