日內資料對黃金指數型基金波動性的增量預測價值

Incremental Value of Intraday Data in Improving Volatility Forecasts of GLD Returns

劉洪鈞(Hung-Chun Liu) ； 姜淑美(Shu-Mei Chiang) ； 白東岳(Tung-Yueh Pai)

《績效與策略研究》 8卷2期 (2011/10) Pp. 11-29

https://doi.org/10.6736/JPSR.201110_8(2).0002

摘要

金融史上看似不可能發生，但實際卻又發生的黑天鵝事件，凸顯資產波動預測的重要性及市場風險管理的必要性。近年美元持續走弱，加上美國幾波量化寬鬆政策，遂使得黃金指數基金成為亂世的最佳避險金融商品。同時，資訊科技發達使得金融商品日內交易資料的取得漸趨便利。因此，如何善用這些資訊提升現存波動模型的預測效益，值得進一步探討。本文以GLD為研究標的，使用GARCH（1,1）為波動模型架構，在其條件變異數方程式中加入PK、RV、RBP、RRV等波動估計式，探討前述各種交易資訊能否提升GARCH模型對GLD樣本外波動預測的準確性。實證結果指出RRV為最具效率的日真實波動代理變數。其次，各種波動估計式，對GARCH模型的波動預測存在不同程度的提升效果。多數情況下，GARCH-RBP表現最佳、依序為GARCH-RV、GARCH-RRV、GARCH-PK及GARCH模型，隱含日內交易資訊能有效增進GARCH模型的樣本外預測績效，而日交易資訊，其提升效果較不明顯。最後，選擇權的交易者可應用GARCH-RBP模型來增進其投資部位的操作績效。本研究結果可使市場參與者瞭解各波動估計式所隱含之資訊內容的潛在價值，有助於提升資產的波動預測效率，更可應用風險值相關分析尋求最適的風險管理模型，以提供投資人進行風險計提的模型基礎。

關鍵字

GLD ； 變幅 ； 已實現變幅 ； 已實現波動 ； 已實現雙冪次變異

並列摘要

Various financial crises over the past few decades, which are typically regarded as the black swan events, have highlighted the significance of volatility forecasting and the necessity of market risk management. Recently, given the relative weakness of the U.S. dollar along with the quantitative easing monetary policy by the US government, the SPDR Gold Shares (Ticker: GLD) has become increasingly prevalent among global investors as a financial hedging instrument during troubled times. In the meanwhile, the broad availability of financial market data at intraday frequency has inspired research into their potential value as a source of information for volatility forecasting. Thus, it deserves to further examine the promotion in expected benefit of existing volatility model by making good use of this information. This study aims to propose the introduction of various daily- and intraday-based volatility estimators (PK, RV, RBP and RRV) into the conditional variance of GARCH(1,1) framework to explore the benefit and incremental value of the intraday trading information that is embodied in those estimators for improving out-of-sample volatility forecasts of GLD returns at daily horizon over the period from 2005 to 2010. Empirical results indicate that the realized range (RRV) is the most efficient proxy for daily volatility. Secondly, the inclusion of each volatility estimator considered shows an improvement in the GARCH model with certain degree. Specifically, the GARCH-RBP model generates the most accurate daily volatility forecasts, followed by the GARCH-RV, GARCH-RRV, GARCH-PK and GARCH (the benchmark) in most cases. Such evidence indicates that the GARCH volatility forecasts can be improved much more with the additional information contained in intraday-based volatility estimators than in daily-based one. Finally, the mean mixed error criterion (MME) suggests that the GARCH-RBP model facilitates option traders for improving the performance of their trading position.These results can enable the market participants understand the potential values of information content in various volatility estimators. It can not only help to heave the volatility forecasting efficiency of assets, but also apply related analysis of Value-at-Risk to seek optimum risk management model to provide the investors with the base of risk- models.

並列關鍵字

GLD ； Range ； Realized range ； Realized volatility ； Realized bipower variation

參考文獻

邱懷義(2004)。SPDRs 在新增 NYSE 市場後的定價效率與錯價因素之比較(碩士論文)。淡江大學財務金融研究所。

周詩婷(2008)。DIA 之不對稱 GARCH 市場風險值之研究(碩士論文)。臺灣大學財務金融學研究所。

連欣儀(2008)。SPY 之不對稱 GARCH 市場風險值之研究(碩士論文)。臺灣大學財務金融學研究所。

陳冠臻(2004)。新增 NYSE 市場對 SPY、QQQ 及 DIA 流動性之影響(碩士論文)。淡江大學財務金融研究所。

Alexander, C.,Barbosa, A.(2008).Hedging index exchange traded funds.Journal of Banking and Finance.32,326-337.