本計畫嘗試將Yeh, Wang and Kuan (2007) 所提出的單變量以分量為基礎設算之波動性，透過無套利條件的運用，擴充至多維度共變異矩陣的估計。因為以分量為基礎設算之波動性其本身所具備的一些優點，諸如與模型設定無關、對市場微結構因子干擾的頑強性、以及不受價格不連續的跳躍因子影響這些特性，幾乎可以推論以分量為基礎設算之共變異矩陣不僅會較文獻中諸如: realized volatility, realized bi-power variation, 以及range-based estimator 等來得準確，而且會表現得較好。文中也列舉出我們日後將在研究中針對各個估計式進行績效比較的模擬實驗，和利用外匯匯率進行之實證分析。我們也討論若干財務實證中能應用此估計式得出新意的一些潛在議題。文末總結本計劃之研究流程與預期之貢獻。 ; This proposal extends the univariate quantile-based realized volatility by Yeh, Wang and Kuan (2007) into multivariate cases by allowing for quantile-based estimator for covariance under no arbitrage condition. With the nice disclosed properties of model-free, being robust to market microstructure noises and jump components in prices, the quantile-based estimator for realized variance-covariance bounds to be more accurate and shall outperform the existing estimators utilizing high frequency, namely realized volatility, realized bi-power variation and range-based estimators. We sketch the tentative horse racing of the aforementioned estimators to entertain in Monte Carlo experiments and empirical studies with foreign exchange rate data in examining the performance of the arbitrage-free quantile-based realized variance-covariance. We discuss some potential applications in empirical finance subsequently. The tempted research agenda and expected contributions from completion of this project are summarized in the end of this proposal. ; 研究期間 9708 ~ 9807
