EVALUATION OF MULTI-ASSET VALUE AT RISK: EVIDENCE FROM TAIWAN

Under the internal model approach (IMA) stipulated by Basel II, financial institutions are allowed to develop and employ proprietary internal models to evaluate various risk. However, the flexibility to develop a proprietary model leads to the question of which computing method delivers the most accurate and reliable estimates of value at risk (VaR). This research employs the new backtesting method proposed by Pérignon and Smith (2008) to determine the best method for computing integrated value at risk. It tests three major VaR computation methods – historical simulation, Monte Carlo simulation, and variance-covariance methods. The portfolio on which VaR is computed includes equities, government bonds, foreign exchange, and index options, all of which are commonly traded by financial institutions. The empirical analysis indicates that historical simulation is the best VaR computation method, which is consistent with the result of Pérignon and Smith (2008).