I recently served as a guest lecturer for an undergraduate finance course at my alma mater. At the start of the class, I thought back to when I was a student and how interesting it was when a real-world practitioner visited the theory-based classroom. By the end of the class, I was reminded that the real world can benefit from understanding the theory that underlies practice. When I got back to my desk later that day, I read an article on zero duration¹ ETFs that sparked an idea to test the theoretical underpinnings of the strategy.
Zero duration ETFs are fairly new products that represent a form of factor-based investing. These ETFs have long positions in predominantly U.S. dollar-denominated bonds and duration-equivalent short positions in U.S. Treasuries (or Treasury futures). The goal of this strategy is to eliminate exposure to risk-free interest rates (the Term factor) while maintaining exposure to corporate bond spreads (the Credit factor).
Our intent was to test how well theoretical factor research translated into the very real performance of these ETFs. To create our definitions for the factors, we used a methodology similar to that of Fama and French.² We regressed each ETF’s weekly excess return over Treasury bills on the Term and Credit factors, plus the equity market factor (Equity).
Here’s what we found:
First, these products greatly dampened sensitivity to changes in long-term interest rates but did not eliminate the exposure altogether. The figure above indicates that if long-term bonds underperformed T-bills by 100 basis points (i.e., long-term rates rose more than short-term rates), these products underperformed T-bills by around 11 basis points, all else being equal. Why was there still exposure to the Term factor? When theory meets reality, executing a hedge is difficult. Minor, hard-to-avoid implementation issues related to duration matching likely caused some key rate exposures to remain. Also, hedging incurs transaction costs. Finally—and it’s not a minor or uncommon issue—there might be differences in how we defined and measured the factor. It seems that the ETFs eliminated most but not all interest rate exposure, an example of the real-life frictions that can hinder theory.
Second, the results show that the zero duration ETFs were all sensitive to credit risk and that the regression coefficients for the high-yield ETFs were generally greater than those for the investment-grade³ ETFs. Neither of these findings was surprising because the ETFs’ goal is to emphasize exposure to the Credit factor, and the Credit factor should have a more profound effect on high-yield bonds because high-yield bonds carry lower credit quality than investment-grade bonds (i.e., they are more sensitive to credit risk). It was perhaps surprising that more of the regression coefficients were not closer to 1. Theory might suggest that these ETFs should have nearly a one-to-one exposure to the Credit factor.
A closer look at the Credit factor can also help investors understand one of the risks associated with such a strategy. Over the long term, investors may expect a factor to earn a return premium, but it is reasonable that they should also expect periods of near-term underperformance, which is illustrated by the figure below.
Finally, the results show that each of the high-yield ETFs had significant exposure to equity market risk and the inclusion of the Equity factor meaningfully improved the explanatory power of our regression models. What might be the reason? High-yield bonds have been shown to share characteristics of both bonds and stocks. Why is this important? Well, as the graph below demonstrates, the Credit and Equity factors appear to be fairly well correlated. If stocks were to underperform, high-yield bonds would likely suffer as well and therefore would not provide the diversification benefits usually provided by investment-grade bonds. While many practitioners might understand this theoretically, dealing with it during the portfolio construction process can be challenging.