There are many complexities to negotiate in applying factor-investing strategies to credit markets
- Applying a factor-investing approach to credit is not straightforward
- The data challenges in credit are particularly complex
- The existing credit factors are supported by extensive academic research
- Factor approaches to credit are likely to become more widespread
The success of quantitative equity funds and smart beta products in the equity markets has established the credibility of factor investing. But can the credit markets be approached in the same way? The simple answer is yes, but both the academic evidence and the practical challenges reveal that the transition is not straightforward.
Value, momentum, quality and size factors are well established in the equity markets and firms such as Robeco started researching their applicability to the credit markets a couple of decades ago. While there is still debate over which group of factors is most effective in the credit markets these four equity factors also seem to be popular in credit, although what they represent can differ in detail. NN Investment Partners (NNIP), for example, uses value, momentum and size together with low risk and carry in its credit-factor strategies.
Robeco, in contrast, uses value, momentum size, quality and low risk. Most factor approaches, says Harald Henke, team manager, fixed-income research, at Quoniam, seem to incorporate some measure of value, momentum/sentiment, and quality/low risk.
In the case of equity markets, researchers have also found many other factors that appear to work on back tests on large databases of historical data, but fail when applied to out-of-sample data. In the case of credit markets, there are additional hurdles to be overcome before even starting to undertake any analysis of historical data.
While the driving forces that can create factor anomalies can be similar across all asset classes, the data challenges in the credit markets are more complex. Companies issue multiple bonds; the characteristics of a bond change with the passage of time; and transaction costs are higher than seen in equity markets and can vary significantly with time. NNIP, for example, therefore only started its work on factor investing by first creating a proprietary database four years ago, says Willem van Dommelen, head of factor investing at NNIP.
In one sense, the analysis of factors should be easier in credit markets than equity markets as the spread on a corporate bond means that the risk premium is directly observable. For equities, in contrast, it has to be estimated by making model assumptions. This should, in theory, lead to less noisy factor premia estimates for bonds, says Henke. But this is offset by the much lower volatility in investment-grade corporate bonds compared with equities, particularly when related to transaction costs in the respective asset classes. That means outperformance is more difficult to achieve in a fixed-income universe.
Moreover, says Patrick Houweling, a portfolio manager at Robeco, in bear markets, credit spreads are wide and there are examples of over-reactions and mis-pricing with large differences between high-risk and low-risk names. In bull markets, spreads are compressed, making it more difficult to generate alpha. Bond-factor research also has to contend with regular blow-ups in specific sectors that can hit all names indiscriminately – in 2015 it was energy, in 2008, banking and in 2001, telecoms.
Avoiding data mining is a key consideration in any research. One reality check is to try and explain why factors exist. There could be behavourial biases on the part of investors, who are often willing to chase higher-risk products in the hope of higher returns, exemplified most clearly in the popularity of lottery tickets where the expected return is negative.
There are also market structures that restrict investors in their choice of investments through currency and regulatory barriers as well as risk limits. Factors need to be robust and insensitive to exact definitions and most importantly, they need to be strong enough to overcome transactions costs, given that any factor strategy will inevitably increase turnover beyond a purely passive buy-and-hold approach.
The existing credit factors are supported by extensive academic research. While these factors are continuously refined, van Dommelen, for one, does not see it as likely that a significant number of new factors can be practically applied in the credit markets. There is, however, potential in the further integration of ESG in their investment process (where so far it is largely used for excluding issuers with a high controversy score).
“Our US high-yield strategy incorporates our factor knowledge in every part of its investment process”
With the current enthusiasm for ESG approaches, ensuring that performance does not suffer by taking an ESG approach is a worthwhile objective in its own right.
While there may be some general consensus on which factors are the most relevant to credit markets, there is plenty of scope for disagreements on the most appropriate implementation of them. The value factor in equities, for example, is measured by relating variables like the book value of equity or corporate earnings to the stock price.
In fixed income, value measures are usually more risk-related quantities using metrics such as the likelihood of default in comparison to bond spreads. “In order to come up with a well-functioning value factor one has to dig deeper and go beyond simple specifications of value,” says Henke.
NNIP, for example, assesses value through a combination of long-term mean-reversion and a cross-sectional regression of spreads, controlling for several characteristics such as duration, size, volatility, industry and rating. Other firms will have their own approaches to determining value and other factors so that, while the underlying philosophy may be the same, the implementation and even the results could be radically different.
The great advantage of using factors in credit is similar to the advantages of quantitative approaches in equities. It is possible to analyse the universe of opportunities on a consistent basis. The depth of analysis on any specific security and name will clearly be much less than would be available using traditional fundamental credit analysis. NNIP uses factor analysis both in conjunction with traditional credit analysis and also on a stand-alone dedicated basis for its US investment-grade portfolios. The latter, says Sebastiaan Reinders, lead portfolio manager, US high yield, has 200 issuers, whereas its high-yield portfolio, which also uses traditional credit analysis, is far more concentrated with about 100 names.
Once factors have been defined and calculated for a universe of securities, how should they be used in constructing a portfolio? Managers can decide on whether portfolios should be constructed based on selecting the highest-scoring securities for each factor separately and then combining sub-portfolios, or through combining them to produce multi-factor portfolios.
Quoniam finds that it achieves the best performance using a multi-factor fair-value strategy rather than a mix of single factors. For investors, comparing factor-driven approaches against benchmark indices can create problems as cap-weighted indices introduce many distortions to the investment process, most notably that weaker credits end up having higher weightings.
The size of dedicated portfolios using credit factors may still be small but the use of credit factors may expand for other reasons. NNIP sees factor analysis as complementary to traditional credit analysis that enables more efficient screening of a database.
Factor analysis of a portfolio also provides a useful risk-analysis framework as well as a source of idea generation: “Our US high-yield strategy incorporates our factor knowledge in every part of its investment process,” says Reinders. “Analysts and portfolio managers use daily refreshed scores as one of their tools for idea generation. They use portfolio-factor exposures in their risk-management tools and involve the factor scoring in the sizing of their credit exposures.”
It seems likely that credit-factor analysis will become more widely available with analytics developed initially in-house by fund managers becoming more widely available. The market share of factor-based credit strategies is also likely to increase. But given the complexities of the credit markets, it is likely to be a challenge for smart-beta credit ETFs to ever match the success of smart-beta equity ETFs.
Credit: Reading the cycle
- Currently reading
Factor Investing: Factors to consider