| Summary: | 碩士 === 元智大學 === 財務金融研究所 === 92 === This paper employs four VaR methodologies, the equally weighted moving average (SMA), the exponentially weighted moving average (EWMA), historical simulation and bootstrap methods to measure the downside market risk of a total of twelve real estate investment trust (REITs) portfolios. The one-day horizon and two-week holding period VaRs are estimated separately for both 95% and 99% confidence levels. Then, we apply the backtesting of Basel rules to verify these models. We find that the bootstrap method produces the most precise VaRs at the 99% confidence level. In addition, the four models don’t forecast VaRs accurately for each portfolio at the 95% confidence level. For the portfolios formed by different property types or different leverage ratios, we suggest that one-day VaRs should be estimated by the historical simulation or the bootstrap method at the 99% confidence level. Finally, the parametric approaches produce more reliable two-week holding period VaRs at a 95% confidence level.
|
|---|
