Since my last piece on Factor-Based Investing in the Corporate Bond Market, both the research into this topic and tools available have come a long way. Just as with our equity counterparts, rich new datasets such as environmental, social, and governance (ESG) and liquidity have found their place in both the portfolio construction and management workflows of the fixed-income portfolio manager. New quantitative tools have made the factor research process far more seamless than it was even just a few short years ago.
Despite the availability of tools, true factor-based corporate bond portfolios haven’t yet become as ubiquitous as traditional fixed-income strategies, which has left fertile ground for the development of new strategies in this space. In this piece, I explore the entire workflow, from bottom-up factor construction to backtesting a composite strategy.
Since our target asset class in this case is corporate bonds, we have a wide variety of factors available for research on both the security as well as the ultimate issuer level.
On the security level, I focused on two of the most commonly used factor styles—risk and momentum. Within the low-risk category, I focused mainly on highly rated short-dated debt, and for momentum I’ve used changes in spread, Duration Times Spread (DTS), and total return over various horizons. While I’ve used simple analytics and basic transformations, a full array of statistical and arithmetic functions is available for the more enterprising quant credit analyst.
The ability to analyze issuer-level characteristics opens a rich swath of factors focusing on liquidity, debt capital structure, and creditworthiness of corporate issuers. Here it’s important that your credit parent database is timestamped, as frequent M&A activity between issuers means the characteristics that might drive total return need to be linked to the correct period at the point in time under analysis. In my example, I’ve used the constituents of the Bloomberg Barclays Global Aggregate and their ultimate issuers.
For my analysis I’ve focused on two main issuer-level analytics types, ESG and issuer level debt position and use of cash.
All these factors combined will allow me to eventually test a strategy of whether low risk, high momentum, high ESG, liquid issues where the issuer uses cash to service debt promptly can outperform the universe benchmark.
As a starting point of my quantitative analysis, I created an equally-weighted composite factor for each group (low risk, high momentum, high ESG, liquidity, and equity). Fractile analysis of the composite factors can quickly reveal their efficacy as investible signals. Here you see that both momentum and low-risk show promise, as does equity to a lesser extent.
Alternatively, your analysis could start at the individual factor level, but I’ve found I can save time by starting from the top down.
Digging deeper into the momentum factor group, I supplemented standard total return momentum factors with momentum in spread to treasury, as well as momentum in DTS. Each of these factors was calculated for a lagged period up to one month prior to the period under analysis. It’s the momentum in narrowing spread and DTS that stand out as potential factors for further investigation.
Digging deeper into the equity factor group, the factors I tested focused on issuers that have a low overall debt burden (absolute, relative to EBITDA, and relative to cash flow), as well as those where cash is used to service outstanding debt rather than issuing new equity of paying dividends. With a simple fractile analysis, two stand out as potential signals—debt/EBITDA and debt/FFO.
That’s not to say that the other factors might not be useful; their signal strength might be more evident over a different horizon or as part of a multi-factor rank.
Since I had promising factors in several groups, I wanted a way to view the significance of all factors using two measures: information coefficient and Sharpe ratio. Running a scatter plot with these axes quickly shows that our significant factors are spread across the low-risk, momentum, and equity factor groups.
As with previous research in the equity market, we see that ESG doesn’t perform well as an investible factor and is better suited as a universe limitation criterion. Similarly, liquidity shows no clear use as a signal at the return horizon I tested.
Working towards our strategy, we can combine the above factors into a single composite. Historically these composite factors would have been combined using a simple weighting scheme, but techniques have improved greatly in the last few years. In creating my composite strategy, I used the trailing month information coefficient of each of the components as the weighting, yielding the below profile.
Running the total return for my strategy factor, we see the fractile profile matches what we would expect for a significant investible factor (fractile 1 outperforms fractile 2, fractile 2 outperforms fractile 3, etc.).
When comparing securities in the top fractile to the benchmark, we see significant outperformance over the evaluation period (more than 20%). The strategy was particularly effective at amplifying the market upswing since 2018.
While this strategy doesn’t apply real-world considerations such as trading costs, turnover limits, and other mandate limits, it does provide a useful jumping-off point for deeper analysis. To further sharpen the efficacy of the factor we could analyze how it performs during various macroeconomic environments (e.g., low inflation, high QE), to help predict how it might perform in these markets in the future.
In my next piece, I will explore how a portfolio can be built around this strategy using an optimization tool and then how the relative performance of this portfolio can be explained using a fixed-income attribution methodology.