A comprehensive risk management review for financial instruments using different value at risk approaches: which methodology improves market risk valuation?

Trabajo Fin de Grado en Economía y Finanzas . Curso Académico 2019-2020In recent years, there has been an incrementing need within the financial industry to make use of more sophisticated models to quantify the associated risk in any investment or financial activity, with the goal of achieving an adequate risk management and control in decision-making processes. Accordingly, throughout this academic research project, we present a review of the different methodologies surrounding the Value at Risk framework, one of the most common tools in financial risk analysis and quantification. We perform a deep analysis from standard approaches for measuring VaR to the more complex techniques. We will also review some backtesting procedures used to evaluate VaR models. Therefore, the main focus of this research will be to implement a theoretical and practical analysis of several value at risk methodologies and discuss their respective behaviour under real life scenarios including low but also high volatility periods, such as the one fostered by the recent Covid-19 pandemic. To carry out the investigation, historic data from the daily log returns of the Dow Jones Index will exploited through the open-source software R. Results in this paper suggest that the GARCH (1,1) model parametric approach to VaR is the best method for forecasting VaR, especially under the Student-T distribution assumption of returns. The Historical Simulation non-parametric approach, as well as the Moving Average Volatility model, also had promising results under relatively stable circumstances, but showed their weaknesses when the situation changed and volatility in financial markets dramatically increased as a consequence of the current health crisis. For its part, empirical literature highlights the lack of accuracy of the traditional Riskmetrics methodology, fact that was also observed here, where we obtained very discouraging results under such approach. Lastly, it seems that the Extreme Value Theory significantly underestimated risk, resulting in a surprisingly bad performanc