Optimal Monitoring and Mitigation of Systemic Risk in Financial Networks

Zhang Li, Xiaojun Lin, Borja Peleato-Inarrea, Ilya Pollak

Abstract

This paper addresses the challenge of optimally allocating a cash injection into a distressed financial network to minimize systemic risk. In a network where institutions are interconnected through debt, the failure of one can lead to a cascade of defaults. We model this as a mathematical optimization problem, showing that under a proportional payment regime, minimizing total unpaid liabilities is equivalent to a linear program (LP). We extend this to stochastic settings where external asset values are uncertain and to "all-or-nothing" payment regimes, which we formulate as a mixed-integer linear program (MILP). Furthermore, we develop a distributed algorithm that allows for global optimization without centralizing sensitive financial data.

The Cash Injection Allocation Problem

In the wake of financial crises, regulators often face the difficult task of deciding how to distribute limited emergency funds. The goal is typically to prevent a "domino effect" of contagion. This paper formalizes the "Cash Injection Allocation Problem": given a fixed budget $C$, how should funds be distributed across nodes (banks) to minimize a chosen risk metric, such as the total weighted sum of unpaid liabilities or the total number of defaults?

To model this, we define a financial network where each node $i$ has external assets and liabilities to other nodes. The variables used in our formulations are summarized in the table below.

Table of notation for the financial network model.

Optimization Frameworks

Deterministic Linear Programming

If we assume a "proportional payment" scheme—where a defaulting institution pays its creditors a pro-rata share of its remaining assets—the problem of minimizing total unpaid liabilities ($\sum w_i (\bar{p}_i - p_i)$) can be solved as a standard Linear Program (LP). This is a significant finding because LPs are computationally efficient and can be solved for very large networks.

Stochastic Programming for Stress Testing

In reality, the value of external assets ($e$) is often uncertain. To account for this, we formulate a two-stage stochastic linear program. In the first stage, the regulator decides the cash allocation $c$. In the second stage, the external assets are realized, and the clearing payments $p$ are determined. We solve this using Benders decomposition and projected stochastic gradient descent, allowing for robust planning under thousands of possible scenarios.

Mixed-Integer Programming for Hard Defaults

In some scenarios, a defaulting node might pay nothing at all (an "all-or-nothing" regime). This makes the problem NP-hard. However, we show it can be mapped to a Mixed-Integer Linear Program (MILP). Despite the theoretical complexity, modern solvers can find near-optimal solutions for networks of over 1,000 nodes in under two seconds.

Distributed Duality-Based Algorithm

A major hurdle in systemic risk management is the privacy of bank data. Banks are often reluctant to share their full balance sheets with a central regulator. We propose a distributed algorithm based on the Proximal Optimization Algorithm. By passing "threat indices"—multipliers representing the systemic impact of a node's distress—between neighbors, the network can converge to the global optimal cash allocation without any node revealing its private internal state to a central authority.

Key Insights and Results

  • The Threat Index: The optimal allocation strategy prioritizes nodes with the highest "threat index," which measures how much a marginal increase in a node's assets reduces total system-wide debt.
  • Computational Efficiency: Simulations of a US-scale core-periphery network (1,065 nodes) show that the MILP approach is highly practical, reaching solutions in seconds. The distributed algorithm, while slower due to communication latency, provides a viable path for privacy-preserving interventions.
  • Heuristic Performance: For minimizing the number of defaults, a reweighted $\ell_1$ minimization heuristic proved significantly more effective than simple greedy approaches, particularly in complex core-periphery topologies.

Conclusion

This research provides a rigorous quantitative foundation for bailout strategies. By moving from ad-hoc decisions to optimized allocations, regulators can significantly reduce systemic risk with the same amount of capital. Future work will look into dynamic models where debts mature over time and the inclusion of behavioral factors like moral hazard.

Summary updated

Abstract

This paper studies the problem of optimally allocating a cash injection into a financial system in distress. Given a one-period borrower-lender network in which all debts are due at the same time and have the same seniority, we address the problem of allocating a fixed amount of cash among the nodes to minimize the weighted sum of unpaid liabilities. Assuming all the loan amounts and asset values are fixed and that there are no bankruptcy costs, we show that this problem is equivalent to a linear program. We develop a duality-based distributed algorithm to solve it which is useful for applications where it is desirable to avoid centralized data gathering and computation. We also consider the problem of minimizing the expectation of the weighted sum of unpaid liabilities under the assumption that the net external asset holdings of all institutions are stochastic. We show that this problem is a two-stage stochastic linear program. To solve it, we develop two algorithms based on: Benders decomposition algorithm and projected stochastic gradient descent. We show that if the defaulting nodes never pay anything, the deterministic optimal cash injection allocation problem is an NP-hard mixed-integer linear program. However, modern optimization software enables the computation of very accurate solutions to this problem on a personal computer in a few seconds for network sizes comparable with the size of the US banking system. In addition, we address the problem of allocating the cash injection amount so as to minimize the number of nodes in default. For this problem, we develop two heuristic algorithms: a reweighted l1 minimization algorithm and a greedy algorithm. We illustrate these two algorithms using three synthetic network structures for which the optimal solution can be calculated exactly. We also compare these two algorithms on three types random networks which are more complex.
BibTeX
@misc{li2014optimal,
  author = {Zhang Li and Xiaojun Lin and Borja Peleato-Inarrea and Ilya Pollak},
  title = {{Optimal Monitoring and Mitigation of Systemic Risk in Financial Networks}},
  year = {2014},
  eprint = {1410.2570},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1410.2570},
}
Figure