考量波動度特性、外溢效果與資訊傳遞之國際投資組合分析

Abstract

Markowitz(1952)平均數-變異數分析架構中，最重要的參數即為變異數共變異數矩陣。 Black and Litterman(1992)應用月資料在資本資產定價模型之均衡假設下，納入國際證券、債 券與外匯市場進行投資組合最適化分析，Andersen and Bollerslev (1997)研究指出，日內高頻 資料較日資料、週資料與月資料擁有更豐富的資訊內涵，因此本計畫主要採用高頻資料，進 行波動度的估計，建立投資組合變異數-共變異數矩陣之較佳估計式。在國際投資組合之分析 議題上，Sarno and Valente(2005)、Ane and Labidi (2006)及Guidolin and Hyde(2006)均指出考 量跨國市場關聯性之重要性，因此本研究將採用線性向量自我迴歸模型(VAR)與非線性的馬 可夫轉換自我迴歸模型(MS-VAR)分析資產報酬與波動度之外溢效果，最後進行國際投資組合 之分析。本計畫預計分為三期來進行，計畫第一年，應用高頻資料建立各市場波動度指數、 剖析波動度之特性並分析與標的指數報酬率之間的統計因果性，計畫第二年：分析標的資產 報酬率與波動度指數在跨市場與跨資產之間的外溢效果與資訊傳遞，計畫第三年：以計畫第 一年建立的各資產波動度指數模型，估計投資組合之變異數-共變異數矩陣，並納入計畫第二 年分析之外溢效果進行國際投資組合之分析。

In the framework of Markovwitz』 mean-variance analysis, variance-covariance matrix is a critical analytic problem. Black and Litterman (1992) incorporate the equilibrium returns for equities, bonds and currencies to drive the portfolio optimization process. Andersen and Bollerslev (1997) indicate that the estimation of variance-covariance with intraday data have much more information content than with daily, weekly and monthly data. High frequency data will be employed to estimate the volatility in different assets in this research. The constant estimator of variance-covariance matrix could also be derived. Sarno and Valente(2005), Ane and Labidi (2006), Guidolin and Hyde(2006) specify the importance of market linkages among countries. Therefore, the linear vector autoregressive model (VAR) and the Markov switching autoregressive model (MS-VAR) would be applied to capture the spillover effect of asset returns and its』 volatility. The international portfolio analysis would then be carried out. The project will be scheduled to three sub-periods. In the first year, we will construct volatility index in different market and the causality relationship between volatility index and its』 underlying index return will be analyzed. In the second year, we will examine the spillover effect and information flow among market. In the last year, we will estimate the portfolio variance-covariance matrix by using volatility index constructed in the first year and incorporate the spillover effect examined in the second year to analyze the international portfolio.