Investigating the effectiveness of value at risk (VaR) models in portfolio risk management for diversified portfolios

Abstract

Value at Risk (VaR) has become a widely accepted risk management tool in the financial industry, providing a single metric that represents the maximum potential loss of a portfolio over a specific time horizon and confidence level (Kenton, 2024). However, the reliability and effectiveness of VaR models have been the subject to an ongoing debate, particularly when applied to diversified investment portfolios. The study aims to address the research gap by conducting a comprehensive evaluation of three prominent VaR models – historical simulation, variance-covariance, and Monte Carlo Simulations- and their suitability for diversified portfolios (Harper, 2024) The analysis focuses on highlighting the theoretical foundations, methodologies, strengths and limitations of each VaR model, as well as evaluating their effectiveness in predicting portfolio losses for diversified portfolios with varying asset class compositions. Using back-testing techniques, the accuracy and reliability of the VaR models are assessed. The analysis revealed that the Historical Simulation method provided the most reliable and adaptable estimates of VaR for the portfolio. The Variance-Covariance and Monte Carlo methods proved effective alternatives when normality assumptions were reasonably met. The findings provide valuable insights for portfolio managers and risk analysts on the most appropriate VaR models to use in managing the risk of diversified investment portfolios.

Keywords: Value at Risk (VaR), portfolio risk management, diversified portfolios, asset allocation, back-testing