Revista Egitania Sciencia - Volume 18 | ARTICLE


Author: Vítor Manuel de Sousa Gabriel (

Publication: Revista Egitania Sciencia - Volume 18

This study compares the performance of several alternatives for modeling market risk, namely historical simulation, exponential weighted moving average, the Gaussian approach, Student-t distribution, the Cornish-Fisher approach and extreme value theory, considering three levels of confidence and the volatility estimates produced by conditional heteroscedasticity models, in a highly volatile and turbulent environment, such as that of the dotcom bubble burst and the Global Financial Crisis. To this end, twelve stock market indices were selected, corresponding to markets in various stages of development and in different locations, applying tests of unconditional coverage, independence and conditional coverage, in keeping with the Christoffersen (2003) approach. The RiskMetrics and historical simulation models produced less accurate predictions, revealing little flexibility in the face of volatility and to accommodate extreme returns. The t-Student assumption and extreme value theory were shown to be appropriate and valid alternatives for measuring risk regardless of the given confidence level, in the context of international risk management.

Keywords: Risk management, Value-at-Risk, Backtesting, Dotcom Bubble, Global Financial Crisis

               INSTITUTO POLITÉCNICO DA GUARDA | Av. Dr. Francisco Sá Carneiro, 50 | 6300 - 559 Guarda | Tel.+351271220100 | Fax +351271222690