Essays in the Econometric Analysis of Systemic Risk Measures

This thesis aims at the study of systemic risk measurement, which became crucial after the 2007 ? 2009 financial crisis. The objective of the thesis is twofold: (i) we address the issue of assessing the accuracy of systemic risk measures, (ii) we investigate the role of the long-range dependence in systemic risk forecasting, under both methodological and empirical perspectives. From the methodological point of view, we propose two appropriate loss functions, the Tail Tick Loss function and the Tail Mean Square Error, specifically designed to evaluate the CoVaR and MES accuracy, respectively. Moreover, we introduce a comprehensive model called Asymmetric-Component-GARCH (ACGARCH), which is able to capture both the leverage effect and long-range dependence. An empirical analysis of different bivariate volatility models to the daily returns of 91 US financial institutions in the period 2000 ? 2012 confirms the need of employing appropriate loss functions to evaluate systemic risk accuracy and to discriminate among different competing models. Moreover, empirical results encourage the usage of the ACGARCH model in the systemic risk framework.