Application of Dynamic Copula-GARCH models in Portfolio Allocation

[[abstract]]本文旨在創造一個用於配適金融資產的最佳模型。為了這個目的，本研究根據 Bayesian Information Criterion (BIC) 來選擇最適的動態Copula-GARCH模型以及從Hansen's skew t分配(STD)、skewed generalized t分配(SGT) 和generalized hyperbolic分配(GH)中選擇最佳殘差分配。由本研究所使用的日資料─MSCI歐洲指數、道瓊工業指數、S&P500黃金指數、MSCI世界能源指數和MSCI新興市場指數─顯示，動態條件相關性(DCC)copula搭配Student's t分配最能捕捉到資產的特性。透過相同權重投資法(the equal-weighted portfolio)、最小波動度投資法(the minimum-variance portfolio)、相同風險貢獻投資法(the equally-weighted risk contributions portfolio)、最大夏普指數投資法(the maximum Sharpe ratio portfolio)以及最大CRRA效用投資法(the maximum CRRA utility portfolio)計算出權重後，便運用於樣本外期間之真實報酬來比較績效。本研究的結果顯示出最大CRRA效用投資法不僅可以考慮到投資者對於風險的偏好外，在全樣本時期(1995年1月2日到2012年12月31日)或金融風暴期間(2008年6月到2009年3月)都有不錯的表現。[[abstract]]In this paper, we evaluate the use of optimization frameworks to allocate financial assets. Comparing time-varying copula–GARCH models with the Hansen's skew t distribution (STD), the skewed generalized t distribution (SGT) and the generalized hyperbolic (GH) distribution by Bayesian Information Criterion (BIC), daily data of MSCI Europe Index (EUR), Dow Jones Industry (DJ), S&P500 Gold Index (Gold), MSCI World Energy Index (ENG), MSCI Emerging Markets Index (EM) support using the Dynamic Conditional Correlation (DCC) copula with Student's t distribution as best model to capture the characteristic of assets during January 2, 1995 through December 31, 2012. The equal-weighted portfolio (1/n), the minimum-variance portfolio (MV), the equally-weighted risk contributions portfolio (ERC), the maximum Sharpe ratio portfolio (MSR) and the maximum CRRA utility portfolio (CRRA) are used to estimate the optimal weights applying in out-sample. Our results provide that the best strategies is the CRRA portfolio which consider about risk appetite of investor as well as has maximum return in both entire period and financial crisis (from June 2008 to March 2009)