風險平價與其他傳統投資組合之績效分析

近年來業界開發出許多資產配置方法以擊敗市場大盤指數為目標，風險平價法就是其中一例，在橋水基金績效卓越的操作下贏得世人的目光，學界、業界爭相研究探索其中科學邏輯。1952 年馬克維茲的平均數變異數理論開創了資產管理的一片天，但因參數的估計太過敏感且不精確，於是乎實際績效與理論上有顯著差異。不同的是，風險平價透過控制風險，將每一個資產對投資組合的風險貢獻設為相同，進而計算出權重，不需要平均報酬的估計。本研究想知道風險平價法與傳統平均數變異數理論、最小變異數理論、均等權重法投資績效的差異，使用R語言實證美國主要30 種產業3369 家公司從1926 年以來的月資料，利用樣本外的測試計算累積報酬、標準差、夏普比率、卡瑪比率、最大回檔等。最後驗證風險平價法是否能如橋水基金全天候策略般均勻分散風險，在經濟不同現象中仍能賺取超額報酬，得出符合理論上標準差位於最小變異數與均等權重法之間的結論，且夏普比率最高、累積報酬第二高，低均等權重3 個百分點。

關鍵字

風險平價；最小變異數；平均數變異數；均等權重；資產配置

並列摘要

Risk Parity portfolio has been risen above since the late 90s curtsey of the All-Weather asset allocation fund by the biggest hedge fund in the world, Bridgewater Associates. Motivated by the economic result, the academia started to put together a fundamental theorem of risk parity portfolio. The idea of risk parity is to equally weigh the asset class by its risk contribution to the portfolio. It sets the weight so each asset contributes the same expected fluctuation of the portfolio. Traditional asset allocation theory is deeprooted in the mean-variance optimization (MVO) framework by Harry Markowitz. However, the MVO approach requires the ex-ante expected return and covariance matrixes which are hard to estimate with accuracy. Empirical estimates based on historical data show that the MVO approach is not favourable. Often results in over or under estimation of returns and variances. The object of the paper is using R language to back test the result of four different asset allocation strategy, Risk Parity, MVO, Minimum Variance, and 1/N. We formed the portfolio by selected 30 major industries in US consisted of 3,369 firms spanned from 1926 to 2020. We conducted out-of-sample test using the monthly return. Results show that indeed the RP approach has the highest Sharpe ratio and the reasonable low standard deviation which is consistent to the previous studies.