Acceso usuarios
Real Estate Investment Trusts (REITs), traditionally known as an asset of low volatility, have been undergoing a period of unprecedentedly high volatility due to the current financial crisis. This has increased the need to search for appropriate methods to cope with extreme risks. This study aims to meet this need by comparing the performances of several commonly used methods in predicting the conditional Value at Risk (VaR) and Expected Shortfall (ES) for REITs. Our competing methods cover all three broad categories (i.e. nonparametric, parametric and semiparametric) classified by Manganelli and Engle (2004) and display a varying degree of complexity. Overall, our results show that the trio of EGARCH skewed t(EGARCH, Exponential Generalized Autoregressive Conditional Heteroscedacity), GARCH t, and GARCH EVT (EVT, Extreme Value Theory) provide the most reliable forecasts among all methods considered. Their good performance, with only a few exceptions, holds up for a variety of quantiles and is robust to the size of the moving window used to make the forecasts. We also find that GARCH normal and RiskMetrics of J.P. Morgan are the worst performers. Filtered Historical Simulation (FHS) models fall somewhere in between.
