Financial Network Stability and Structure: Econometric and Network Analysis

Since the Global Financial Crisis, the literature of financial networks analysis has been trying to investigate the changes in the financial networks structure, that led to the instability of the financial system. The Global Financial Crisis followed by the Great Recession costed taxpayers an unprecedented $14 trillion (Alessandri and Haldane, 2009), austerity and downturns in GDP. The dynamics of the financial networks transferred the collapse of a US housing market bubble into a large meltdown of the financial systems globally. The study of systemic risk and macro-prudential policy has come to the forefront to model and manage the negative externalities of monetary, fiscal and financial sector activities that can lead to system wide instabilities and failure. The dimensions of crisis propagation have been modelled as those that can spread cross-sectionally in domino like failures with global scope, or build up over time, as in asset bubbles. The cross sectional propagation of shocks that occur due to non-payment of debt or other financial obligations with the failure of a financial intermediary or a sovereign leading to the failure of other economic entities, is called financial contagion. Cross sectional analysis of financial contagion can be done using statistical methods or by network analysis. The latter gives a structural model of the interconnections in terms of financial obligations. This dissertation uses both approaches to model financial contagion. The applications include the study of systemic risk in Eurozone Sovereign crisis, the US CDS market and the global banking network. This is organized in three self-contained chapters Our contribution to the literature begins with the study of the dynamics of the market of the Credit Default Swap (CDS) contracts for selected Eurozone sovereigns and the UK. The EWMA correlation analysis and the Granger-causality test demonstrate that there was contagion effect since correlations and cross-county interdependencies increased after August 2007. Furthermore, the IRF analysis shows that among PIIGS, the CDS spreads of Spain and Ireland have the biggest impact on the European CDS spreads, whereas the UK is found not be a source of sovereign contagion to the Eurozone. Next we perform the empirical reconstruction of the US CDS network based on the real-world data obtained from the FDIC Call Reports, and study the propagation of contagion, assuming different network structures. The financial network shows a highly tiered core-periphery structure. We find that network topology matters for the stability of the financial system. The “too interconnected to fail” phenomenon is discussed and shown to be the result of highly tiered network with central core of so called super-spreaders. In this type of network the contagion is found to be short, without multiple waves, but with very high losses brought by the core of the network. Finally we study a global banking network (GBN) model based on the Markose (2012) eigen-pair approach and propose a systemic risk indices (SRI) which provide early warning signals for systemic instability and also the rank order of the systemic importance and vulnerability of the banking systems. The empirical model is based on BIS Consolidated Banking Statistics for the exposures of 19 national banking systems to the same number of debtor countries and the data obtained from Bankscope for the equity capital of these 19 national banking systems. The SRI is based on the ratio of the netted cross-border exposures of the national banking systems to their respective equity capital. The eigen-pair method stipulates that if the maximum eigenvalue of the network exceeds the capital threshold, there is cause for concern of a contagion. This is compared with the loss multiplier SRI proposed by Castrén and Rancan (2012). The latter is found to have no early warning capabilities and peaks well after the onset of the crisis in 2009 while the eigen-pair SRI gives ample warning by late 2006 that the cross border liabilities was unsustainable in respect of the equity capital of the national banking systems. We contribute to the literature by highlighting the efficacy of the network approach to systemic stability analysis of GBNs. In particular we develop an eigen-pair approach for GBNs and prove its usefulness in an early warning context.