信用溢酬債券投資模型：以科技及原物料類股為例

Abstract

本文主要說明如何從債券指數中投資被市場低估的債券，建構類似於原來指數，但報酬率較佳、風險較低的投資組合，以結構式模型(Structural-Form Model)的Merton model(1974)得出公司違約機率，再配合違約時的損失程度、發行人的系統性風險程度與經濟環境的相關性等相關參數評估信用溢酬。 當得出每一檔債券的信用溢酬後，接著衡量在每單位的預期損失下，有多少的超額信用風險溢酬，同時將指數中的成份債券依不同的存續期間做分類，每類中的債券再依超額信用風險溢酬做排序，並投資前20%的債券做為我們的投資組合。 實證利用Bloomberg的科技類股、原物料類股債券指數做benchmark，針對我們的投資組合作回溯測試，無論是科技類股或原物料類股，其報酬均優於指數、風險也較小，每年報酬比指數更好且更加穩定，同時能有效降低最大回檔時的損失，也能夠提高Sharpe ratio，尾端風險也顯著下降，特別是在風險較高、年化報酬較低的原物料類股有更佳的效果。

This article shows how to invest the undervalued bonds from different indices. We want to construct the portfolios which is similar to indices, but we can gain higher return and lower risk portfolios. By adopting Merton model to obtain the default probability and other parameters like loss given default(LGD) and systematic risk, we can evaluate the credit spread. When obtaining every credit spreads of bonds, we should measure how much excess credit spread we can get in one unit of loss given default. Meantime, we must use different durations to classify bonds, and then sort bonds by excess credit spreads in every duration. Finally, we invest the top 20% bonds of excess credit spreads to construct portfolios. We use Bloomberg technology and material bond indices as benchmarks. The return and risk of our model portfolios are significantly superior to benchmarks in our back-testing. We find that year-to-year returns of our model portfolios are better and more stable. Also, our model portfolios can reduce the max drawdowns effectively, improve Sharpe ratio, and lower tail risks. Especially, our model has better effect on the material bond index because there are higher risk and lower return in this index.