Blog Posts — May 7, 2020

Corporate Credit Portfolio Construction: Targeting low-beta names during the COVID-19 Market Crisis

by Dieter Vandenbussche and David Antonio

Authors: Dieter Vandenbussche and David Antonio

USD Corporate Credit – style factors drive risk profile

In a previous post (here) we analyzed the risk of US High Yield and Investment Grade portfolios using the Axioma Factor-based Fixed Income Risk Model (FFIM), which is calibrated using cross-sectional techniques and incorporates style factors. We saw that in the both cases the style factors, and in particular the beta style factor, drove the risk profile at the peak of the crisis. For example, results for USD Investment Grade shown below:

In this post we investigate how a portfolio constructed to optimize exposure to the Beta style factor performed through the market crisis.

What do the FFIM factors represent?

The FFIM is a spread-based cross-sectional model that includes categorical factors (market intercept, local currency market, country and sector) and non-categorical factors (quality and style). The factors explain the returns of issuer spread curves i.e. capture the market credit risk. The beta style factor exposures are calculated by regressing a 1-year history of issuer returns against 1-year of the global market returns (average global issuer spread movements). An issuer whose spread co-movements with the global market are more volatile will have a high beta score. Raw beta scores are standardized to ensure the estimation universe has a capitalization-weighted mean of zero and standard deviation of one. The beta factor return then reflects the issuer spread movements of high/low beta names in excess of the market, and can be thought of as a long-short basket of curves. The FFIM Beta factor is analogous to the market sensitivity factor in Axioma’s suite of equity models.

Let’s look at an example: the charts below show the movement of the Beta factor (green) alongside the market intercept (blue). The average market spread rises considerably during the COVID-19 crisis. In contrast the Beta factor exhibits a sharp tightening of spreads, followed by a recovery. The factor movements demonstrate that low-Beta names experienced a lower spread increase relative to the market, and that the size of the movement in the Beta factor was unprecedented in recent history.

From Issuer Spreads to Portfolio Construction

Having identified spread behavior of low-Beta names, can we translate this to a bond portfolio that beats its benchmark? In order to test this we minimize the exposure to this Beta factor of USD High Yield and Investment Grade portfolios using the Axioma Portfolio Optimizer, and evaluate performance relative to a liquid benchmark. The following constraints are used:

  1. DTS neutrality – no active exposure to the credit market;
  2. Target annualized tracking error of 2%;
  3. Maximum issuer holding of 1% i.e. at least 100 names in the portfolio;
  4. Daily rebalancing with a 10% two-way turnover limit.

We do recognize the challenges associated with fixed-income trading, especially during times of reduced liquidity. The portfolios constructed here are not necessarily tradable, but serve to illustrate how to relate the spread factor movements to bond portfolio returns.

Optimized Portfolio Return Attribution:

For both IG and HY the optimized portfolios active returns track the behavior of the Beta factor: strong positive active returns while Beta returns are negative, and negative active returns when the factor reverses, leading ultimately to a total active return of approximately zero over the period – as ever indicating that timing is key! In both cases the Beta factor contribution explains a large majority of the total factor contribution, and bond-specific factors do not appear to contribute much. This highlights the fact that the optimized portfolios do indeed tilt effectively toward the Beta factor. The charts below show the return attribution for each portfolio:

Optimized Portfolio Risk Attribution:

Finally, when we measure the risk of the optimized portfolios we see the Style factors dominating the attribution, with a small proportion also attributable to issuer specific factors. Both the IG and HY portfolios have an active risk of 2%, showing that the TE constraint was binding. The variance attribution chart below shows that for both the IG and HY portfolios the style factor accounts for over 88% of the variance, and of this the Beta factor accounts for 90% (IG) and 94% (HY).  The optimized portfolios are indeed tilting effectively toward the Beta factor, and the constraints are effective in eliminating any unintended exposures. Note in particular the absence of any contribution from the Market and Currency Intercept factors, indicating the active portfolios are DTS neutral.


The Axioma FFIM provides a means to accurately measure credit risk using a small number of cross-sectional spread factors. These factors include a series of style factors, and through use of the Axioma Portfolio Optimizer we can construct portfolios with high exposure to these factors. The Beta, or market sensitivity, factor has exhibited extreme moves in recent months and an optimized portfolio that targets low-Beta names is shown to be able to translate that factor movement into a portfolio of bonds that, if timed correctly, outperformed its benchmark.