Saddle-point approach: backtesting VaR models in the presence of extreme losses

• Autor(a) principal: Gouveia, Ricardo João da Silva
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/17714
• topic: Backtesting; Value-at-risk; Time series; Risk management; Gestão do risco; Modelos de risco; Séries temporais

The Basel Committee for Banking Supervision requires every financial institution to carry out efficient Risk Management practices, so that these are able to face adverse days in the market and, thus, avoid another potential meltdown of the financial system, such as the 'Black Monday' in 1987 or the 'Subprime' crisis in 2007. To do so, traditional backtesting techniques assess the quality of commercial banks’ risk forecasts based on the number of the exceedances. However, these backtests are not sensitive to the size of the exceedances, which could lead to inaccurate risk models to be accepted. This way, this dissertation presents the Saddle-point backtest, a size-based procedure developed by Wong (2008) that evaluates risk models through the Tail-Risk-of-VaR. This approach is believed to constitute a reliable size counterpart to the Basel II Agreements, hence deserving an important role in backtesting. However, the Saddle-point backtest shows some drawbacks regarding its application to non-parametric risk models, which is explored throughout this dissertation’s empirical analysis.