Conditional graphical models for systemic risk estimation

• Published in: Expert systems with applications Vol. 43; pp. 165 - 174
• Main Authors: Cerchiello, Paola, Giudici, Paolo
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.01.2016
• Subjects: Banking; Banks; Conditional independence; Correlation; Estimates; Expert systems; Financial risk management; Gaussian; Network models; Networks; Risk; 99-00; Network models; Financial risk management; Conditional independence; 00-01
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2015.08.047

•The paper provides a stochastic framework for financial network models.•It is based on a conditional graphical Gaussian model.•Systemic risk is decomposed into country risk plus bank specific risk.•It is the first paper that considers more data sources in systemic risks estimation. Financial network models are a useful tool to model interconnectedness and systemic risks in banking and finance. Recently, graphical Gaussian models have been shown to improve the estimation of network models and, consequently, the interpretation of systemic risks. This paper provides a novel graphical Gaussian model to estimate systemic risks. The model is characterised by two main innovations, with respect to the recent literature: it estimates risks considering jointly market data and balance sheet data, in an integrated perspective; it decomposes the conditional dependencies between financial institutions into correlations between countries and correlations between institutions, within countries. The model has been applied to study systemic risks among the largest European banks, with the aim of identifying central institutions, more subject to contagion or, conversely, whose failure could result in further distress or breakdowns in the whole system. The results show that, in the transmission of systemic risk, there is a strong country effect, that reflects the weakness or the strength of the underlying economies. Besides the country effect, the most central banks are those larger in size.