An empirical evaluation of value at risk by scenario simulation

"Scenario simulation was proposed by Jamshidian and Zhu (1997) as a method to separate computationally intensive portfolio revaluations from the simulation step in VaR by Monte Carlo. For multicurrency interest rate derivatives portfolios examined in this paper, the relative performance of scenario simulation is erratic when compared with standard Monte Carlo results. Although by design the discrete distributions used in scenario simulation converge to their continuous distributions, convergence appears to be slow, with irregular oscillations that depend on portfolio characteristics and the correlation structure of the risk factors. Periodic validation of scenario-simulated VaR results by cross-checking with other methods is advisable"--Office of the Comptroller of the Currency web site.