Development of a Methodology for Validating Diversification Strategies of Financial Institutions (focus on Value At Risk & Expected Shortfall models in Market Risk Capital Requirements and Portfolio Margining)

As a result of the increasing number of links between financial institutions and because of the resulting contagion effect, a loss suffered by one company can significantly impact other institutions. This has been particularly well observed during the last financial crises. As a consequence, financial institutions are required to mitigate the risk associated with their activities by different regulations. Portfolio Margins and Market Risk Capital Requirements are examples of risk mitigations. However, in order to be efficient, one should ensure that the risk that those mitigations aim to cover is properly quantified. Hence, regulators ask to external institutions, called validators, to validate models of financial institutions. In this thesis, I propose a methodology to help validating diversification strategies of financial institutions in Value At Risk or Expected Shortfall models. More precisely, I focus on those models for Market Risk Capital Requirements and Portfolio Margining. Risk Dynamics (part of Mckinsey and Company), the company where I made my internship, is a validator in risk management. The methodology I developed aims in that context to provide a practical approach for validating the models of its clients. And indeed, by using my methodology on its clients models, Risk Dynamics is now better equipped to state whether the strategy is accepted or rejected as a good risk mitigation. The thesis is structured in several parts. First a theoretical part covers the background material necessary to understand validation and diversification (mainly based on the concept of dependence). Then, the key concept of diversification strategy is explained together with its importance in risk management models and the reasons why it should be validated. Thereafter, Portfolio Margining and Market Risk are studied with a focus on Value At Risk and associated Expected Shortfall models. On that basis, I then present the methodology I developed and then evidence its usefulness through an empirical real case application with a fictive portfolio.