Detecting anomalous payments networks: A dimensionality reduction approach

The speaker, Carlos León, introduces the presentation and his role as a principal researcher at Banco de la República. He discusses the department's focus on financial infrastructure and information exploitation, mentioning previous research on financial entities' interactions and risks. The upcoming presentation will cover the topology of cross-holdings in the Colombian financial system, with a draft article in progress.

Carlos León elaborates on the concept of cross-holdings, explaining how they indicate interdependent relationships between financial entities due to mutual asset ownership. He emphasizes the potential for financial instability due to reciprocal and circular dependencies, known as feedback effects, and plans to study their occurrence in the Colombian financial network.

The session discusses cross-holdings with a focus on specific financial instruments: bonds, deposit certificates, and shares. The study finds that reciprocal and transitive relationships in the Colombian financial system are scarce, implying limited cyclical interdependencies. Leon highlights how this study is a preliminary step toward assessing the financial system's stability considering cross-holdings.

The presentation continues with a literature review on cross-holdings, noting their overlooked role in systemic risk. Most studies are theoretical or simulations, with few analyzing real data. The current research leverages data from 2016 to 2019 on various financial assets, highlighting their significance relative to Colombia’s interbank market.

Leon discusses the research methodologies, including network analysis metrics for assessing reciprocity and transitivity, which gauge the extent of mutual relationships and closed triangles within financial networks. He introduces measures like network density and survival rate as indicators of network stability and potential systemic risk.

The speaker outlines their dataset, including bonds, certificates, and equities issued and held by financial entities, detailing their structure and relevance. He explains how this data forms multiple layers within the network and emphasizes the use of network topology to unravel interdependencies, serving as a crucial factor in understanding financial risk.

The analysis sees the correlations among various financial assets networks, revealing that deposit certificates contribute most to the network, followed by bonds and equities. Topological properties indicate that the network is generally sparse, with low density, reciprocity, and transitivity, implying a limited systemic risk from cross-holdings in Colombia.

The speaker highlights the findings on reciprocity and transitivity across the network layers, noting the near-absence of such relationships in equities due to regulatory restrictions. Anomalies such as entities holding their own issued certificates are explained, adding nuances to the systemic risk landscape.

Further analysis covers network stability via survival rates, showing a high persistence of cross-holdings over time. The study’s dynamic aspect reveals that both connections and gaps within the network remain stable weekly, suggesting a robust but limited web of financial interdependence.

The presentation compares different network layers using correlation indices, identifying that deposit certificates heavily influence the overall network’s properties. Despite low inter-layer similarity, the results suggest that systemic risks from cross-holdings are currently minimal but warrant continued monitoring.

Leon discusses the use of a null model to compare observed networks with random graphs, assessing the significance of observed metrics like density and reciprocity. The analysis shows these metrics have similar magnitudes to random models, suggesting the inherent low risk posed by cross-holdings.

The session moves into the interpretation of findings, emphasizing that the Colombian financial system’s network shows low circular interdependencies, hinting at minimal systemic risk due to cross-holdings alone. However, further investigation into the monetary value and real-sector integration is necessary for comprehensive risk assessment.

Questions and remarks from participants address methodological approaches and comparison with other financial systems. Discussions include the stability of such networks during crises, underscoring the importance of understanding external influences and regulatory impacts on financial interlinkages.

Final discussions acknowledge the study's relevance in providing insights into potential network stability indicators for financial systems. Carlos León reiterates the exploratory nature of the study, emphasizing ongoing efforts to incorporate economic impacts and broader exposure types into systemic risk evaluations.