不同相依結構下兩種波動模型之動態避險

Abstract

過去一年多以來全世界都經歷了全球的金融危機，所有相關的金融市場，都面對了前所未見的市場波動。在這樣的情況下，避險對於風險管理的重要性和需求也與日俱增。對於持有現貨資產的投資人，想要規避風險，直覺的方式就是利用相對應的期貨契約，而避險也被認為是期貨的主要功能之一。隨著期貨市場的活絡，如何決定最佳的避險比率也成為主要研究的課題，從過去的文獻中，提出許多的動態避險模型，它們大致上能做出比原始模型更好的避險的效果，但改善的幅度卻是隨著使用的動態避險模型和評估期間有所不同。在這篇論文中，在不考慮交易成本的前提下，我們針對兩個方面，報酬基礎和變幅基礎的波動模型分別在兩種不同的相依結構DCC和Copula下的樣本內和樣本外避險結果進行比較。根據實證的結果觀察，變幅基礎的模型特別在波動大的市場下有較佳的預測樣本外短期波動的能力，另一方面報酬基礎的模型則是在樣本內和樣本外都保持一定的避險表現，雖然不一定是最佳的避險模型，但是有很好的一致性，對於避險而言這是非常重要的性質。 而Copula對於DCC模型而言，在樣本內外的避險結果，並沒有展現出明顯的優勢， 推測可能的原因是計算條件共變異數時，採用了數值積分而產生的小幅計算誤差，即可能抵銷了因為使用Copula相依結構因而放寬了限制帶來的好處。或是這也表示著有更符合現實，符合實際分配的函數或架構值得進一步研究。

The world has experienced the global financial crisis and extremely volatile impact for the past year. Under this circumstance, the demands for risk management increase. For investor who possesses spot assets, it is natural to think of the corresponding futures contract for hedging purpose. Way to determine the optimal ratios for hedging purpose has become an important task. Dynamic hedging models generally result in better risk reduction compared to conventional method. However, the performances differ. This thesis attempts to make the comparison of the hedging performances cross two dimensions, return-based against range-based and DCC against Copula among both in-sample and out-of-sample results. According to the empirical results, while the range-based models in general demonstrate relatively better out-of-sample forecasting power under volatile environment, the return-based models can be considered to be more consistent cross markets and cross period of time. As for the copula-based model, the improvement over the DCC-based model is somehow insignificant. The numerical integral for generating covariance maybe blamed for this. The limitation on the computation accuracy could bring inaccuracy on hedging ratio and thus neutralize the possible benefit brought by using more realistic distributions.