透過您的圖書館登入
IP:18.221.187.121
  • 學位論文

應用高頻率資料提升波動模型預測能力之研究

Improving Predictive Ability of Volatility Models with High-frequency Data

指導教授 : 邱建良
共同指導教授 : 劉洪鈞(Hung-Chun Liu)

摘要


本研究以美國個股(微軟、亞馬遜)、股價指數(S&P 500指數、那斯達克綜合指數)與指數型股票基金(道瓊工業平均指數基金)自2001年1月至2010年5月之日資料為實證標的,探討加入日變幅(PK)、已實現波動率(RV)、已實現變幅(RRV)與已實現雙冪次變異(RBP)等波動估計式對於GARCH模型樣本外預測能力的提升效果。分別以PK及RV作為市場真實波動的代理變數,並採用各種損失函數評估各波動模型的預測績效。 本文以五分鐘頻率之日內資料來估計RV、RRV與RBP,進一步探討高頻率資料的效果。實證發現,各種波動估計式對GARCH模型波動預測存在不同程度的提升效果,其中又以RV、RRV及RBP等高頻日內資料為基礎之波動估計式表現較佳。而以RV及PK作為真實波動性之代理變數下的預測績效比較結果趨於一致,證明預測結果的穩健性,也另外說明除了RV適合作為真實波動性之代理變數外,PK也是個不錯選擇。 另外,波動估計式對於提升波動模型預測績效的程度會因標的不同而 有所差異,本文研究發現在個股方面的提升效果最佳,此結果隱含在波動性較大的標的下更能顯現波動估計式的效果,因此投資者在投資高波動性之金融商品時,可應用波動估計式提升模型的波動性預測績效。

並列摘要


This paper augments the GARCH models with the PK range, realized volatility (RV), realized range volatility (RRV) and realized bipower variation (RBP). We investigate the impact of these volatility estimators by examining their out-of-sample forecast-improved. The data for our empirical study consists of individual stocks (Amazon and Microsoft), stock indices (S&P 500 and Nasdaq) and exchange traded fund (Dow Jones Industrial Average ETF) price quotes covering the period from 16 January 2001 to 28 May 2010. The forecast performance evaluation is relied on several loss functions and utilizing PK and RV as a proxy for true volatility. RV, RRV and RBP are intraday-based Volatilities which are obtained from intraday prices at 5-min frequency. So we can study the effect of high frequency data. Empirical results indicate that all volatility estimators can improve predictive ability. Especially, the inclusion of intraday-based volatility measure in GARCH models notably improves forecasts. In most cases, the forecasting performances of models are almost consistent which are robust to alternative proxy measures, indicating that the PK is a useful alternative to the RV since daily high-low price data are readily available for most financial assets. Additionally, the degree of incremental predictive content of these volatility estimators varies from the data used. The volatility estimator provides the most incremental predictive content on individual stock. It implies that the volatility estimator can be more advantageous with higher volatility commodity. Thus, market practitioners can exploit the information content implied by these volatility estimators to improve forecast accuracy of models when they invest in financial instruments with higher volatility.

參考文獻


1. 王祝三、莊益源、張鐘霖(2003),「波動率模型預測能力的比較--以臺指選擇權為例」,台灣金融財務季刊第四輯第二期,頁41-63。
4. 林秀蓉(2008),「股價波動性預測」,淡江大學財務金融學系碩士學位論文。
5. 李沃牆、張克群(2006),「比較不同波動率模型下台灣股票選擇權之評價績效」,真理財經學報,第14期,頁71-96。
7. 周雨田、巫春洲、劉炳麟(2004),「動態波動模型預測能力之比較與實證」,財金論文叢刊,第一期,頁1-23。
8. 賈景宇(2001),「台灣創新型認購權證在不同波動性模型下之比較」,中原大學企業管理學系碩士學位論文。

被引用紀錄


周益賢(2012)。運用日內資料提升選擇權價格預測準確性之研究〔碩士論文,淡江大學〕。華藝線上圖書館。https://doi.org/10.6846/TKU.2012.00914

延伸閱讀