Systemic risk measures and contagion models

The main theme of the thesis is systemic risk measurement. This extremely young field of research has gained a lot of attention in recent times from academics and practioners alike because of global financial crises. The main contributions of the thesis can be grouped in four broad items. Firstly, we propose a novel categorization of the risk measures advanced in recent years based on the modeling assumptions they rely upon. We identified four categories: measures based on portfolio theory, econometric indicators, network analysis and measures based on multivariate default distribution. The second set of contributions regards the CIMDO framework, a methodology heavily used in systemic risk studies. We proved a new theoretical independence result that significantly extended previous ones. We also performed a comprehensive stability study where every input of the methodology was considered and whose conclusions should serve as guidance for future CIMDO users. The third contribution is a new contagion model that is both tractable and flexible enough to be used with heterogeneous portfolios. We provided several theoretical results with respect to both marginal and joint default distribution. We also detailed a recursive algorithm to calculate the portfolio loss distribution in an efficient way. An application to the problem of pricing and hedging CDO products is hence shown. Lastly, we introduced a new systemic risk measure in the context of contagion models called contagion loss ratio (CLR). It is based on attributing losses to either idiosyncratic or infection-driven events and represents the percentage of the total portfolio losses due to contagion. We showed how to calculate it in a variety of models and presented an application to the problem of banking stability.