You will find a“diversification effects”column in the reports which shows the risk reduction effect if an asset negatively correlated with the portfolio is constructed. You can also infer this information from the positive or negative sign of the incremental VaR.
Assuming that all variables are independent and identically distributed (so that time conforms to the laws of arithmetic) and because variance is a square value, the total VaR is calculated as daily VaR multiplied by the square root of the number of days for which the assets are held.
Yield values come from three sources: benchmark government bond yields, spline-fitted zero-coupon yields, and interest rate swap (IRS) yields.
Performance index VaR is the benchmark VaR per unit of investment target. It is useful when comparing the risk of a given target with a benchmark instrument (e.g., securities vs. weighted stock market indices). It can be interpreted as the increase in the target’s risk exposure per additional unit of VaR.
Geometric Brownian Motion (GBM) is used for equity and foreign exchange factors, whereas the Vasicek Model is used for interest rate factors owing to their tendency of mean reversion.
To update a portfolio, simply reupload it to the system. Historical values will remain in the system. Warning: Do not delete the portfolio. Doing so will also delete all historical reports from the system.
VaR is the quantification of unrealized gain/loss, so RAROC (risk-adjusted return on capital) values also pertain to unrealized gain/loss only. Realized gains and losses are no longer categorized as risks, so the system does not include them in RAROC calculation.
The best method really depends on the asset type. Variance and covariance methods are best suited for linear assets, while historical simulation and the Monte Carlo method are recommended for portfolios with a significant proportion of non-linear assets. In addition, the configuration of parameters also influences accuracy. Backtesting can be used to determine the best model for your portfolios.
Due to the increasing likelihood of extreme market events (also called “fat-tail distribution”), and because the historical simulation method is based on actual market performance, risk values may sometimes appear larger than if normal distribution is assumed (as is the case with other statistical methods).
The system currently provides transaction data from the stock markets of seven East Asian countries. For stocks traded in other countries, you will need to upload the daily transaction data yourself and set the corresponding exchange rate in order for the system to perform calculations.