Stability analysis of financial contagion due to overlapping portfolios

Fabio Caccioli, Munik Shrestha, Cristopher Moore, J. Doyne FarmerarXiv 2012 · arXiv

Abstract

This paper presents a mathematical framework to quantify how financial "fire sales" propagate through a network of overlapping portfolios. By mapping market dynamics onto a generalized branching process, the authors identify a "contagion window" where the system is uniquely vulnerable. The study reveals a diversification paradox: while diversification may protect individual institutions, it can destabilize the global market, creating a "robust yet fragile" environment where failures are rare but catastrophic.

Introduction

Traditional models of financial contagion primarily address counterparty risk—the direct "domino effect" of one bank failing to meet its obligations to another. However, the 2008 financial crisis highlighted the importance of indirect contagion through overlapping portfolios. When a stressed bank liquidates assets, the resulting "fire sale" depresses prices, harming the balance sheets of other institutions holding the same assets. This "liquidity spiral" can lead to systemic collapse even in the absence of direct lending ties. This paper aims to quantify the conditions—such as leverage, diversification, and market crowding—under which local shocks escalate into global crises.

Methodology

The authors represent the financial system as a bipartite network consisting of two distinct node types: Banks and Assets. A link between a bank and an asset indicates that the bank holds that asset in its portfolio.

Bipartite network representation of banks and assets.

To analyze the stability of this system, the authors employ several analytical tools:

  1. Generalized Branching Processes: The spread of bankruptcy is modeled similarly to a biological outbreak using Galton-Watson processes. A stability matrix is constructed where entries represent the expected number of bank failures of a certain type triggered by the failure of another.
  2. Eigenvalue Analysis: The system's stability is dictated by the largest eigenvalue ($\xi_1$) of the stability matrix. A value of $\xi_1 > 1$ indicates the potential for a global cascade.
  3. Synthetic Ensembles: Due to the lack of empirical datasets, the researchers use bipartite Erdős-Rényi graphs for Monte Carlo simulations. This allows for systematic variation of parameters like the number of banks ($N$), assets ($M$), and average diversification ($\mu_b$).

Key Insights and Results

The study identifies a "Diversification Paradox." At low levels of diversification, the network is too fragmented for shocks to spread. At very high levels, banks become robust because their exposure to any single asset is minimal. However, in the intermediate "Danger Zone," banks are interconnected enough to transmit shocks but not robust enough to withstand them.

This leads to a "robust yet fragile" state: the probability of a cascade is low, but if one occurs, it is catastrophic, potentially wiping out the entire network. Furthermore, the authors identify a "leverage floor" (approximately 12x in their model) below which the system remains unconditionally stable. Market "crowding"—the ratio of banks to assets—also plays a critical role, typically shifting the stability window and increasing instability at lower diversification levels.

The analytical branching process provides a strong match for numerical results, though it slightly underestimates risk by ignoring network loops. Interestingly, the contagion window remains invariant whether the initial shock is a toxic asset price drop or a specific bank failure, as the network's internal dynamics dominate once a cascade begins.

Conclusion and Future Work

While the model provides a robust framework for understanding systemic risk, it relies on several simplifying assumptions. The "tree-like" assumption ignores the amplifying effect of circular dependencies (loops) in real-world networks. Additionally, the model assumes static portfolios, whereas real-world "stop-loss" orders might accelerate contagion. Future research could incorporate heterogeneous bank sizes ("Too Big to Fail") and calibrate the model using real-world regulatory data for macroprudential stress testing.

Summary updated

Abstract

Common asset holdings are widely believed to have been the primary vector of contagion in the recent financial crisis. We develop a network approach to the amplification of financial contagion due to the combination of overlapping portfolios and leverage, and we show how it can be understood in terms of a generalized branching process. By studying a stylized model we estimate the circumstances under which systemic instabilities are likely to occur as a function of parameters such as leverage, market crowding, diversification, and market impact. Although diversification may be good for individual institutions, it can create dangerous systemic effects, and as a result financial contagion gets worse with too much diversification. Under our model there is a critical threshold for leverage; below it financial networks are always stable, and above it the unstable region grows as leverage increases. The financial system exhibits "robust yet fragile" behavior, with regions of the parameter space where contagion is rare but catastrophic whenever it occurs. Our model and methods of analysis can be calibrated to real data and provide simple yet powerful tools for macroprudential stress testing.
BibTeX
@misc{caccioli2012stability,
  author = {Fabio Caccioli and Munik Shrestha and Cristopher Moore and J. Doyne Farmer},
  title = {{Stability analysis of financial contagion due to overlapping portfolios}},
  year = {2012},
  eprint = {1210.5987},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1210.5987},
}
Figure