台灣ETF投資組合配置之研究－改良後階層風險對等模型之應用

Abstract

本研究以台灣發行之各地股票ETF與各類債券ETF作為資產池，改良近期被JP Morgan譽為「必須加入資產配置工具箱」之階層平價方法，創建「改良階層平價方法」，並與歷史上著名資產配置模型，如等權重、最小變異數、風險平價、最多元化投資組合進行績效、風險、權重與喜好資產的比較分析。發現這些以風險為基礎之投資組合，皆傾向選擇「低波動之資產」，也就是此處的債券ETF種類，造成其於包含股票與債券之資產池中無法有很好的風險分散效果。 在這些以風險為基礎的資產配置方法比較部分，本研究改良之階層平價方法無論在股市相對多頭或空頭，皆能打敗原階層平價方法，為學術領域提供一項有效之資產配置模型。而等權重與風險平價投資組合因分別以權重、風險等值概念進行權重配置，而使其權重相較於其他方法，集中在少數資產與單一債券種類之問題較輕微，但其缺乏考量其他資產特性也讓其表現無論在股市相對空頭或多頭期間皆表現不佳。最小變異數與最多元化投資組合，則是所有方法之中較易選擇波動性高之資產，因此其於股市相對多頭期間有較高之投資組合報酬，但投資組合也因此遭受較高波動性。而階層風險平價系列方法則是其中較保守之方法，比其他方法更傾向選擇低波動之資產，但其於股市相對空頭期間獲得比其他方法更高之報酬，證明其資產配置能力。 雖然並沒有一種投資組合方法永遠勝過其他方法，但透過本文分析各方法之差異，我們更能夠對不同資產池之資產特性與景氣循環，視情況選擇適合之資產配置方法，達到更好的投資組合表現。

This study improves Hierarchical Risk Parity method, which JP Morgan said that should be added to asset allocation toolbox, by changing the algorithm of weight calculation. Comparing it with other historical models, including equally-weighted, minimum variance, risk parity, most diversified, hierarchical risk parity, this study finds that these risk-based portfolios tend to choose low-volatility assets, which are bond ETFs. This would make them unable to have a good risk dispersion, even if both stock ETFs and bond ETFs are in the universe. In the comparison of asset allocation methods, the improved Hierarchical Risk Parity defeat the original one whether the stock market is relative bull or bear. Because equally-weighted and risk parity portfolios are respectively weighted by weight equivalence and risk equivalence, their weights are equally distributed among all the assets. However, the lack of taking other characteristics into consideration also makes poor performance in a bull market. Minimum variance and most diversified portfolios tend to choose the assets with high volatility instead, so they have higher returns and therefore subject to higher volatility. Though the original and the improved hierarchical risk parity portfolios are more conservative, their brilliant asset allocation abilities are proved by the higher returns in a bear market. Although there is no such a portfolio that always outperforms others, by analyzing the differences between the methods, we could choose an appropriate asset allocation model for different asset characteristics or in different economic cycle to achieve better performance.