流動性對不動產投資信託之影響

未考量流動性風險的風險值可能低估風險，且估計99％信賴水準的風險值，更應將高階動差納入風險值模型，以捕捉報酬率和流動性風險呈現非常態分配的現象，本文以相對價差衡量流動性風險，並將流動性效果區分為有無加入相對價差波動率，估計兩種流動性調整風險值，另利用Cornish-Fisher展開式納入報酬率或相對價差的非常態分配效果，估計修正風險值和三種修正流動性調整風險值。共採用七種風險值模型，衡量十二檔在香港和台灣上市的不動產投資信託之風險，進行回顧測試，發現流動性效果和非常態效果會受到不動產投資信託市場區位、流動性特徵和持有天期的影響。在99％信賴水準時，一日風險值完全低估REIT市場的真實風險，而流動性調整風險值可降低失敗率和超額損失，尤其是流動性相對較差的香港REIT和台灣REIT，風險值納入流動性效果可通過失敗率檢定，納入報酬率或相對價差的非常態效果後，可準確預測兩個不動產投資信託市場的真實風險。整體而言，在99％信賴水準下，非常態分配效果高於流動性效果，而相對價差波動率會影響流動性調整風險值的準確性，當持有天期愈短時，流動性效果和非常態效果提升風險值績效的能力愈大。

關鍵字

流動性風險； Cornish-Fisher ；風險值；流動性調整風險值；不動產投資信託

並列摘要

The goal of this thesis is to examine the forecasting performances of liquidity-adjusted VaR and Cornish-Fisher modified VaR/LVaR for Honkong and Taiwan REIT markets. The empirical evidence shows that the liquidity effect measured by proportional quoted spread will improve the predicted efficiency of VaR and LVaR with lower failure frequency and excess loss. With high confidence level, VaR seems to generally underestimate risk in all REIT. There is also slight underestimation for LVaR in specific REIT. Moreover, Cornish-Fisher modified VaR reduces the underestimated probability of VaR in 99% confidence level. Because of the fat-tail and skewness of return and spread, using the Cornish-Fisher approach to account for non-normality in both price and liquidity risk produces much accurate results. We also compare the average failure frequencies between the 1-day VaR and 10-day VaR, and find that the non-normality effects for 1-day VaR are higher than those for longer horizons in 99% confidence level. Overall, liquidity effect and non-normality effect constitute a large part of total risk in high confidence level and at shorter horizons.