Value at Risk: a Trade-off between Accuracy and Computational Time

Since the 90’s, the Basle and the Basle II committee has required banks to calculate their VaR periodically to maintain a sufficient capital in order to face eventual losses estimated by VaR. Frequently, the risk managers have to choose between the accuracy and the computational time of the VaR. And these choices depend meanly on the nature of their portfolios and the assumptions related to the VaR models. This study introduces a practical and tangible approach of the VaR computation by exploring and detailing the Vba program of each VaR model. As we supposed, the results illustrate that the most accurate methods are the most computationally intensive one. The matter was to evaluate various methods in order to appreciate their respective efficiencies in actual processes. Furthermore, three solutions have been considered to improve the trade-off between accuracy and time of computation: The first solution focuses on advanced methods such as the Copula approach and the Fourier’s transform. The second solution is the solution given by the Graphical Unit Processor that permits to improve the Monte Carlo simulation. The last solution is the most recent solution that focuses on improving the quality and the computational time of the random numbers, which is determinant in the Monte Carlo simulation. This new technology, which is not yet available, appears to be promising in the near future