The fragility of financial systems was starkly demonstrated in early 2023 through a cascade of major bank failures in the United States, including the second, third, and fourth largest collapses in the US history. The highly interdependent financial networks and the associated high systemic risk have been deemed the cause of the crashes.
While established algorithms of systemic risk calculation have provided foundational principles for understanding network interactions and crisis propagation, they exhibit limited flexibility due to their simplified treatment of financial instruments. Factors such as recovery rate, time structure of debt, credit quality, and interest rates have been heavily neglected, despite being fundamental components of debt valuation in financial mathematics.
Our research aims to integrate those essential debt valuation factors into systemic risk analysis and assess their impact on the US financial system’s stability. We focus particularly on identifying critical points where cascading failures might occur due to system interdependencies. We achieve this by incorporating a robust class of reduced-form models into the comprehensive NEVA framework, demonstrating how our methods can detect cascading failures that standard approaches might miss.
The influence of the recovery and credit quality parameters on the risk assessment of the United States financial system is presented in Figure 1. The impact of the credit quality is more significant. Effects occurring around critical thresholds of 6% and 8% shock values exhibit heightened severity, evidenced by a larger number of defaults and more substantial shifts. Moreover, following the critical threshold of 6%, the amount of indirect defaults remains consistently around 3, significantly contributing to the ongoing amplification of the crisis. In fact, from the very beginning of crisis around the 6% point, the network effect at least doubles the crisis magnitude most of the time up to the 8% threshold of sharp increase in direct defaults.
In the context of topology of financial system under consideration, network effects primarily induce shifts in crises and amplify pre-existing ones. A noticeable shift occurs around the point of 6% shock magnitude, important to consider when assessing system stability. However, this shift does not substantially alter system behavior. Based on these findings, there is insufficient evidence to posit that systemic effects serve as a primary source of severe crises; rather, they amplify or increase probability of occurrence for crises caused by direct shocks, without fundamentally altering the system’s nature. Consequently, there is a basis to infer that network effects do not inherently pose a qualitative danger to the financial system of United States. It seems reasonable to state that, for instance, the incorporation of safety margins into critical point calculations conducted by standard non-network techniques is able to provide a sufficiently accurate assessment of secure area boundaries, without the actual need to resort to systemic risk methods. Nevertheless, financial systems are complex systems, with multiple layers of intertwining relationships. In reality, the dynamics of the interest rates and their relationships with the behavior of the banking system was crucial aspect of the 2008 subrime mortgage crisis, the following sovereign euro crisis, as well as recent 2023 American banking crisis. This feature is reflected in the proposed model, as it is the incorporation of those relationships into the network dynamics that allows to obtain the cascading failure stemming from interactions between banks.
This effect is illustrated in Figure 2. The incorporation of interest rate dependency leads to notable transpositions of the critical points in comparison to the previous approach. Specifically, the mild critical point transitions from 6% to a range spanning 2%-5%, while the severe crisis point shifts from 8% to a range of 3%-7%, attributable to extensive cascading failures induced by systemic feedback mechanisms. This difference underscores a fundamental discrepancy between models with and without interest rate dependencies on market conditions. The crunch of the financial system varies significantly, as the inclusion of interest rates precipitates a collapse driven by the common exposure and indirectly by the interplay between interest rates and the banking environment. Consequently, neglecting interest rates and their interconnections with the banking system in modeling can obscure risk recognition, resulting in a substantial underestimation of actual risk and eventual system collapse due to decisions predicated on inadequate analysis.
Overall, the incorporation of debt valuation factors into the network dynamics significantly influences the results of the risk assessment of the American financial system. The substantial impact of debt valuation processes, particularly with recovery and credit quality parameters, was demonstrated, showcasing the adaptability of the proposed approach across diverse modeling environments, in contrast to the approaches not taking into account those features. The relationship between the interest rates and the banking environment integrated into the model dynamics allowed to uncover the potential for cascading failures and a collapse of financial system stemming from interdependencies among its participants, a phenomenon not captured by other models. These results underscore the complexity inherent in the study of systemic risk, where the exclusion of certain factors can lead to significant deviations in risk assessments, posing a danger of misjudging risks.
By Kamil Fortuna (Wrocław University of Science and Technology)