Do low-risk bonds, like low-risk stocks, tend to outperform their high-risk counterparts? In their September 2013 paper entitled “Low-Risk Anomalies in Global Fixed Income: Evidence from Major Broad Markets”, Raul Leote de Carvalho, Patrick Dugnolle, Xiao Lu and Pierre Moulin investigate whether low-risk beats high-risk for different measures of risk and different bond segments. They consider only measures of risk that account for the fact that the risk of a bond inherently decreases as it approaches maturity, emphasizing duration-times-yield (yield elasticity). They focus on corporate investment grade bonds denominated in dollars, euros, pounds or yen, but also consider government and high-yield corporate bonds worldwide. Each month, they rank a selected category of bonds by risk into fifths (quintile portfolios). For calculation of monthly quintile returns, they weight individual bond returns by market capitalization. They reinvest coupons the end of the month. They focus on quintile portfolio Sharpe ratios to test the risk-performance relationship. Using monthly risk data and returns for 85,442 individual bonds during January 1997 through December 2012 (192 months), they find that:
- Over the sample period, for the base case of the corporate investment grade bond segment with risk measured by duration-times-yield:
- The low-risk bond portfolio (bottom quintile) is much less volatile and has a much higher gross Sharpe ratio than the high-risk bond portfolio (top quintile).
- The low-risk (high-risk) portfolio has a very low (extremely high) beta relative to the bond segment capitalization-weighted index.
- The low-risk (high-risk) portfolio has the largest positive (most negative) gross alpha relative to the bond segment index. The outperformance of the low-risk portfolio is not due to industry, market capitalization or bond rating tilts.
- The impact of estimated transaction costs due to portfolio turnover is low (high) for the low-risk (high-risk) portfolio.
- However, during the bond bull market encompassed by the sample period, the positive alpha of the low-risk portfolio is not large enough to offset the performance drag of the low beta in terms of return. The low-risk portfolio underperforms the bond segment index on this basis.
- Assuming that the bond bull market is over, the positive alpha and low beta of low-risk bond portfolios will be attractive compared to the overall bond market. A bond portfolio with duration-times-yield modestly below that of the market should beat (match) the market index when the index performs poorly (well) based on return, and beat the market based on Sharpe ratio.
- Duration-times-yield is a particularly efficient metric for sorting bonds to exploit the low-risk anomaly, but results are generally consistent for other measures of risk.
- Results are very consistent for other bond market segments and aggregates of segments.
In summary, evidence indicates that a bond portfolio with risk (duration-times-yield) modestly below that of the bond market may be attractive in the coming post-bull market environment.
More exuberantly, the authors speculate that “this is probably the best time in many years to actually profit from the low-risk anomaly from an absolute point of view. It is not unrealistic to expect that the next few years will bring a normalization of the long-term Sharpe ratio of fixed income asset classes back to…+0.3…from the levels often in excess of +0.7 seen in our back-tests. A lower remuneration of high beta fixed income, perhaps even with negative returns, makes investing in low-risk fixed income with positive alpha and low beta even more appealing.”
Cautions regarding findings include:
- As noted in the paper, a positive alpha and high Sharpe ratio for (unleveraged) low-risk bonds does not translate to market-beating returns over the (bull market) sample period.
- While the paper assesses trading (portfolio rebalancing) frictions, reported returns and Sharpe ratios are gross. Net returns and Sharpe ratios would be lower.
- Most investors cannot construct their own diversified low-risk bond portfolios, so would have to locate a low-risk bond fund with attendant management and administrative fees.
- Given the number of alternative risk measures and market segments considered in the study, results for the best-performing combinations may impound material data snooping bias (luck), thereby overstating performance expectations.
