Selfishly Prepaying in Financial Credit Networks

This article studies how to learn approximate Nash equilibrium (NE) from static historical datasets by empirical game-theoretic analysis (EGTA), which provides a simulation-based framework to model complex multiagent interactions. Generally, EGTA requires plentiful interactions with the environment or simulator to estimate a cogent and tractable game model approximating the underlying game. However, these exploratory interactions often suffer from low data utilization efficiency and may not be feasible in risk-sensitive applications. To address these problems, this article investigates a new EGTA paradigm for offline settings and introduces a novel algorithm called conservative offline policy space response oracle (COPSRO) to identify NE from fixed datasets without active data collection. COPSRO initiates by extracting a set of strategies from the offline dataset to construct an overcomplete strategy population, achieving an approximation to the policy space of the original game. Then, COPSRO integrates the conservative critic (CC) to tackle the challenge of overestimation inherent in offline learning scenarios. Additionally, it devises the offline NE solver to iteratively compute approximate NE. Consequently, COPSRO can ascertain equilibrium strategies without real-world interaction, markedly enhancing its utility in risk-averse settings. This article provides both theoretical analysis and empirical evaluation to demonstrate the effectiveness and superiority of COPSRO across various real-world tasks in the offline setting. Our method surpasses existing approaches in terms of convergence and exploitability, especially when the coverage ration of dataset is low (20% or 10%).

We examine the extent to which rescue strategies within a banking system can reduce systemic risk. We focus on donations from solvent banks to banks in distress, which can in principle reduce losses and prevent default cascades. We build an agent-based model to simulate the ensuing strategic game on a randomly generated financial network, where nodes represent banks and edges represent inter-bank liabilities. Each bank independently decides whether to rescue (and whom) to maximise their payoffs. We analyse the rescue strategies adopted by the banks at equilibrium, using empirical game-theoretic analysis. Our results show that donations can indeed reduce systemic risk when the equilibrium strategy profile is adopted. Individual donations can benefit multiple banks in the network. Our results also indicate that lower default costs and small-variance liabilities tend to decrease the incentives to donate. We furthermore examine the impact of the banks’ rationality on the effects of rescue, finding that banks behaving rationally use their funds for rescues more efficiently than banks that behave irrationally.

The exposure of banks to systemic risk in financial networks usually requires large bailouts of taxpayer money with long-lasting and damaging societal consequences. We examine whether the banking network can reduce systemic risk from within by selfishly cancelling the debts of banks in distress. This operation can in principle reduce losses and prevent default cascades. We define an abstract model to simulate the ensuing strategic game on randomly generated financial networks, where each systemically important bank independently decides how likely it is to cancel some debts of insolvent banks. We compute the equilibrium of the induced empirical game with the empirical game-theoretic analysis and analyse its efficiency by measuring the price of anarchy. Our results show that selfish debt cancellation can reduce systemic risk when adopting the equilibrium strategy profile. However, our results also indicate that the efficiency of the equilibrium can be low and relatively few banks cancel debts at equilibrium, and we explain the reason for this through analysis of the banks’ incentives and game dynamics.

Random matrix theory is applied to study the correlation between different financial correlation coefficient matrices in the financial field. Correlation coefficient matrix is a key factor for constructing a network. In this paper we relate the random matrix theory to the network construction to study the financial networks model in terms of the random matrix. We select the stock data of Shanghai stock market, and divide them into four stages. We discuss the statistical properties of eigenvalues in financial correlation coefficient matrix and random matrix based on the random matrix theory, and improve the existing denoising method to construct the correlation coefficient matrix and to make it more suitable for building financial networks. After that we can build the financial network model. Then we analyze and compare the original financial network, the denoising financial network and the noise financial network in terms of the random matrix theory and the key node of networks. It is found that the primary important information is still in the original network, and the noise information corresponds to the information which the nodes of small degree in the original network include. Finally we analyze the topological structure of the financial networks, such as the minimum spanning tree, the motif and community structure. We also find that the topological properties of the improved financial networks are more remarkable and the topological structure is more compact.

Macro-Prudential Regulation from the Perspective of the Financial Network

This study defines the cognitive structure of financial networks research by looking at the terms scholars have used over the last 31 years of research. Applying co-word analysis over a sample of 680 records extracted from Scopus database, the findings reveal that that the literature on financial networks is made up of six thematic clusters, these are: Dynamics of financial networks, Complex banking networks, Risk and contagion in financial networks, Topology and structure of financial networks, Main methodologies and techniques and Post-crisis responses. In addition to laying the groundwork for understanding financial networks through a retrospective view, this study shows how bibliometrics (especially co-words analysis) can contribute to reducing conceptual ambiguity in emerging research fields, such as financial networks research, and open doors for future research.

Financial risk contagion based on dynamic multi-layer network between banks and firms

Contagion in Financial Networks

Financial networks, bank efficiency and risk-taking

The Missing Links: A Global Study on Uncovering Financial Network Structures from Partial Data

We present a general methodology to automate the search for equilibrium strategies in games derived from computational experimentation. Our approach interleaves empirical game-theoretic analysis with reinforcement learning. We apply this methodology to the classic Continuous Double Auction game, conducting the most comprehensive CDA strategic study published to date. Empirical game analysis confirms prior findings about the relative performance of known strategies. Reinforcement learning derives new bidding strategies from the empirical equilibrium environment. Iterative application of this approach yields strategies stronger than any other published CDA bidding policy, culminating in a new Nash equilibrium supported exclusively by our learned strategies.

Effective anomaly detection is crucial for the success of many AI-based solutions in the financial domain, including e.g. fraud detection and risk modeling. Identifying anomaly from financial transaction networks is one of the challenging tasks that can be cast as a special instance of anomaly detection in networks. Existing methods typically attempt to detect only node-level anomalies, and assume prior knowledge to extract representative features for identifying anomalies. However, there exist collective fraudulent behaviors at the level of subgraphs rather than individual node. A ring structure for money laundering and a tree structure for pyramid schemes would be common examples. Also, in practice it is difficult to decide which features are more representative beforehand. In this paper, we introduce SADE (Subgraph Anomaly DEtection) framework that addresses these needs. SADE consists of two steps: 1) role-guided subgraph embedding, and 2) subgraph anomaly detection. Our approach for learning the subgraph embeddings allows to preserve both the local structure of subgraphs and the global structure of entire network by making use of global roles and local connections of nodes. The learnt representation allows effective use of the state of art anomaly detection approaches. Our extensive experiments on synthetic and real-world financial transaction networks demonstrate the effectiveness of SADE in learning subgraph embeddings without requiring any prior knowledge and detecting anomalous subgraphs.

991012657369603412 HKUST Electronic Theses Essays in financial networks and contagion by Jialiang Lin thesis 2018 x, 40 pages : illustrations ; 30 cm Chapter 1 studies the interplay of network structure and imperfect information in financial contagion. Financial crisis may…Read more ›

Spectral signatures of structural change in financial networks

Rethinking Financial Contagion