Value Investing is a popular investment approach dating back to the seminal work of Benjamin Graham and David Dodd in the 1930s[1]. Historically, it has been predominantly applied to the management of equity portfolios. Today, value-tilted equity funds of different kinds are easily available to both professional and retail investors[2].

Fixed income has been (very) late to the party. The serious research on the viability of Value Investing in the fixed income space started only about 10-15 years ago (I discussed the reasons for it in this article)[3]. Since then, the body of research has grown substantially but remains minuscule compared to what is available in the equity space. Similarly, value-tilted systematic fixed income products are now available to the public (mostly packaged under the umbrella of multi-factor portfolios) but their size and market penetration are still well below their discretionary fixed income peers, opening up the possibility for strong secular growth in the space[4].

In this series of articles, I will present the rationale, mechanics, and investment results of applying systematic value investing in the fixed income space. Let’s dive straight into it.

What is Systematic Value Investing?

Value Investing in the fixed income space encompasses the idea that undervalued bonds tend to outperform overvalued bonds.

This is not a revolutionary idea. Open up the pitch book of any of the largest discretionary fixed income managers and you will see a detailed description of their large credit research and portfolio management teams whose primary task is to identify mispriced securities using fundamental analysis.[5]

Systematic value strategies are no different in spirit but apply the idea of capturing mispriced bonds differently. They represent a data-driven investment approach that relies on statistically significant and repeatable dependencies. Organized as algorithms, systematic value strategies evaluate large sets of bond and non-bond specific data (e.g., equity and accounting information) and use it to derive fundamental insights upon which portfolios are constructed (more on this in the next article).

There are many advantages that systematic value strategies offer vs their discretionary counterparts, including their breadth of coverage and limited susceptibility to behavioral biases (I have written in detail about this here). Systematic value strategies are not infallible though – they rely on the assumption that the analyzed data is correct, representative, and complete. Any violation of these assumptions poses the risk of selecting a “value trap” type of security that fails to deliver the expected outperformance. A potential way to mitigate this risk is combining systematic and traditional discretionary approaches in a cohesive quantamental investment process (as I discussed at length here).

Why do Systematic Value Strategies work?

“Value Investing can be used on a systematic basis and is often seen as part of a suite of investment styles that aim to capture thematic risk premia” (Ben Dor et. al., 2021, page 413). In simple words, systematic value strategies offer access to the value risk premium whose addition to traditional fixed income portfolios may be beneficial due to its low correlation to global macro risk factors[6].

There is no consensus as to why the value risk premium exists and works in practice. The existing literature offers two explanations – the value risk premium 1) is a legitimate compensation for a specific undiversifiable type of risk or 2) results from cognitive errors that investors persistently make[7]. Let’s dig a little deeper and get a better feel of what each explanation entails.

Value fixed income strategies entail going long a basket of “cheap” bonds and shorting a basket of “expensive” bonds (or overweighting “cheap”/underweighting “expensive” bonds in long-only portfolios). If the expensive bonds are (in aggregate) risker than the cheaper bonds (i.e. riskier in an underversifiable, systematic way), then the value risk premium is nothing more than rational compensation for investors bearing that risk; it is a compensation for the possibility that when investors’ overall wealth falls (like in a period of market crisis), the value strategy may also lose money if cheaper bonds end up underperforming expensive ones[8].

On the contrary, if the value risk premium results from cognitive errors, then it relies on temporary irrational behavior leading to security mispricings which, in turn, reverse when market efficiency is restored. Well-documented phenomena[9] like investor overreaction and myopia can be the driving force for such mispricings and are also well observed in credit markets[10].

Which of these explanations is true has been argued by researchers for ages. Personally, I side with Asness (2015) who proposes that these explanations are not mutually exclusive and their relative impact can vary through time.

Do Systematic Value Strategies deliver persistent results?

Up until recently, it was common to discuss the “death of value investing” which – in all fairness – came on the back of years of disappointing performance, particularly by equity value strategies. At its core, this was a discussion about whether or not the value risk premium still works. Critiques of value investing have attributed its weak performance to the fact that A) the strategy may be overcrowded and the risk premium is being arbitraged away, and/or B) the strategy has suffered due to unfavorable macroeconomic conditions[11]. Although this discussion has been very equity-centric, clarifying these questions from a fixed income point of view is important, so let’s dive into the existing literature and see what it has to say about it.

Unsurprisingly, there are not many studies investigating the time variability of Value in the fixed income space although a paper by Ilmanen et. al (2021) provides some useful insights. The authors of the paper use almost a century’s worth of data (Jul 1926 – Nov 2020) to conduct a broad study on the time variation of different factor risk premia (Value, Momentum, Defensive, Carry) across different asset classes (Equities, Fixed Income, Commodities, etc). Looking at their results for Fixed Income/Value, one can get a sense of how the value risk premium in fixed income has performed through time and whether or not there is evidence of it withering away in recent years.

Using the entire sample, the authors find that Fixed Income Value has averaged a 1.3% return p.a. (t-stats. 2.84) with a standard deviation of 4.6%, and a Sharpe Ratio of 0.29. However, these aggregate numbers may be hiding the temporal dynamics in the value risk factor’s performance. For example, arbitrage activities may have eliminated Value’s positive expected return in recent years, hence the 1.3% long-term average return may only reflect past periods when the market structure was completely different. Therefore, it is important to look beyond these aggregate statistics, but before doing that, let’s take a moment and consider what may happen when a strategy gets overcrowded.

A paper by Asness (2015) provides, in my view, a useful framework. From an investor’s point of view, the attractiveness of a strategy is a function of two things – 1) its long-term expected return, and 2) its volatility. In theory, when a strategy like Value gets crowded, its long-term expected return may go down because investors pile into the strategy and arbitrage away the spread between the overvalued and undervalued securities. At the same time, an increase in the popularity of a strategy may attract a lot of “tourist money” which quickly abandons the strategy when returns turn negative. Such “boom and bust cycles” exacerbate the natural ups and downs of a strategy and thus effectively magnify its volatility. Therefore, if we wanted to know if a strategy continues to “work”, we need to understand primarily how its return distribution (in terms of mean and standard deviation) has changed through time.

In essence, this is precisely what the study by Ilmanen et. al. (2021) attempts to find out. First, the authors investigate if there is a shift (in a statistical sense) in Fixed Income Value’s (FI Value hereafter) mean return through time. For illustration, here is what a shift/structural break in the mean could look like:

Using the Chow test and a Bonferroni correction, Ilmanen et. al. (2021)  test the hypothesis of structural breaks in the FI Value data. What you see in the table below are the p-values from this test – the lower the value, the higher the probability for a shift in the mean. As can be observed, the test results for FI Value reject the hypothesis of a structural break in the data, indicating a lack of evidence for a shift in FI Value’s mean return over time.

Similar evidence (or, actually, lack thereof) is produced by a second test that Ilmanen et al. conduct. The underlying theory is that after the “discovery” of a style factor (i.e. after the seminal paper describing that factor is published), the average returns may be statistically different from the average returns prior to the discovery. Taking advantage of the long time series at hand, the authors are in a position to measure and test this hypothesis. In the case of FI Value, the authors find (again) no evidence for a shift in the mean of the return distribution [12].

Next, using the same technique as for the mean,  Ilmanen et al. (2021) uncover the existence of structural breaks in the volatility of FI Value. Unfortunately, the study doesn’t investigate the reasons for this phenomenon (something I consider a weakness), but thankfully the data from the paper is publicly available, which allows for some further analysis.

What you see below is a chart of the realized 5-year rolling volatility based on monthly data for the period June 1931 and February 2022 used in the Ilmanen et. al. (2021) study.

Looking at this chart, there are two observations that can be made.

First, the chart illustrates quite clearly the time-variation of FI Value’s volatility and thus confirms what Ilmanen et al. (2021) discover. Looking at the chart, it is easy to spot the extended periods of heightened volatility but also the periods where volatility has been low. Even if we agree with Ilmanen et. al. (2021) that there has been no degradation in the FI Value expected return, the Sharpe Ratio of the strategy can be vastly different from one period to another due to the time variation in its volatility. This is important for portfolio construction reasons (especially in the context of multi-factor portfolios), but also for business-related reasons as an upward volatility shift may leave investors with a strategy that yields a materially lower Sharpe Ratio than what was originally pitched to them.

Second, looking at the chart, it is difficult to see evidence of a monotonic increase in FI Value’s volatility. As a reminder, one of the side effects of overcrowding may be an increase in the strategy’s volatility. A linear estimate of the long-term trend indicates a negative slope (t-stats = 10.96), which speaks against the notion of rising volatility.  Even if we discard older data (for which Ilmanen et. al confess is of lower quality) and only focus on the last 20 years when the data is more reliable, it is still difficult to see a pattern of increasing volatility.

Conclusion and investment implications

It is now broadly accepted that systematic value strategies in the fixed income space “work” even though the reasons for that (cognitive biases or genuine risk compensation) are still debated. As systematic value investing in the fixed income space grows in popularity, investors may legitimately be concerned about the long-term sustainability of the strategy. With that being said, the little empirical research at hand finds no evidence that the value risk premium in FI has been arbitraged away – hardly a surprise given the relatively small footprint that systematic value strategies currently have on the market. However, the empirical evidence also suggests that the FI Value strategy may go through extended periods of higher or lower volatility. Not only is this relevant for multi-factor portfolio construction problems but also for the investors in the strategy who should keep their eyes wide open, knowing that the risk-adjusted return of the strategy may fluctuate substantially between decades.

Coming up next

In the next part of this series, I will discuss some of the seminal papers that uncovered the value risk premium in the fixed income space. I will take a closer look at how the value signals are generated, how the strategies are designed, and what type of investment results can be achieved both in the corporate and sovereign space. Stay tuned!

Do you like this new and shorter format? As this website is still in its early stages, I responded to subscribers‘ feedback and wrote this article to be shorter and more easily digestible that the previous ones. Let me know in the comments below if you prefer this format.

Notes

[1] I am referring here to the Benjamin Graham and David Dodd’s book “Security analysis” first published in 1934

[2] A casual search on Morningstar returns more than 40 results for active Value equity funds in the U.S. (SOURCE: Morningstar, as of May 2022). Not included in this figure are the active and passive ETFs available to both retail and institutional investors. In Europe, the situation is similar – as a retail investor based in Germany, I have access to 20 Value Equity ETFs (as of May 2022) with a total AuM of almost $17 bn.

[3] To the best of my knowledge, one of the earliest papers to investigate Value investing in fixed income markets is a paper by Correia, Richardson, and Tune from 2012 titled “Value investing in credit markets”. Shortly thereafter, in 2013, Asness, Moskowitz, and Pedersen further solidify this by providing evidence for the existence of the value risk premium in a broad range of asset classes, including fixed income.

[4] According to International Finance Review using data from evestment, the total AuM invested in quantitative credit strategies is about $52bn as of 2021. Although this number doubled since 2016, it still remains minuscule compared to the $66tn investible fixed income universe. SOURCE: International Finance Review, evestment. As of January 2022.

[5] This statement is based on a research I conducted while working for the Emerging Markets Strategy team at PIMCO.

[6] See Exhibit 4 in Style Investing in Fixed Income by Brooks et al. (2018) as an example of the low loadings of Value to global risk factors observed in both government and corporate bonds.

[7] See Asness (2015) and Israel et. al. (2021, page 2).

[8] SOURCE: Asness (2015)

[9] Perhaps the best record of bihavioral and cognitive biases can be found in Daniel Kahneman’s book “Thinking Fast and Slow” (Kahnemann, 2011). Studies have shown that professional investors are not immune to such biases, see e.g., Kudrytsev et. al. (2013), Zahera and Bansal (2018), Jain et. al. (2015), and Shukla et. al. (2020)

[10] I make this statement based on my ~8 years of experience working for PIMCO.

[11] See Israel et. al (2021, page 1)

[12] See Ilmanen et. al (2021), page 50

References

Asness, C., Moskowitz, T. and Pedersen, L., 2013. Value and Momentum Everywhere. The Journal of Finance, 68(3), pp.929-985.

Asness, C., 2015. How Can a Strategy Still Work If Everyone Knows About It?. [online] Aqr.com.

Ben Dor, A., Desclée, A., Dynkin, L., Hyman, J. and Polbennikov, S., 2021. Systematic investing in credit. Hoboken, New Jersey: John Wiley & Sons, Inc

Brooks, J., Palhares, D. and Richardson, S., 2018. Style Investing in Fixed Income. The Journal of Portfolio Management, 44(4), pp.127-139.

Correia, M., Richardson, S. and Tuna, A., 2011. Value Investing in Credit Markets. Review of Accounting Studies, 17(3), pp.572-609.

Graham, B. and Dodd, D., 1934. Security analysis. 1st ed. New York City: McGraw Hill.

Ilmanen, A., Israel, R., Moskowitz, T., Thapar, A. and Wang, F., 2021. Factor Premia and Factor Timing: A Century of Evidence. Journal of Investment Management, 19(4), page 15-57.

Israel, R., Laursen, K. and Richardson, S., 2021. Is (Systematic) Value Investing Dead?. The Journal of Portfolio Management, 47(2).

Kahneman, D., 2013. Thinking, Fast and Slow. New York, United States of America: Farrar, Straus and Giroux.

Kudryavtsev, A., Cohen, G. and Hon-Snir, S., 2013. ‘Rational’or’Intuitive’: Are behavioral biases correlated across stock market investors?. Contemporary economics, 7(2), pp.31-53.

Shukla, A., Rushdi, D., Jamal, N., Katiyar, D. and Chandra, R., 2020. Impact of behavioral biases on investment decisions ‘a systematic review’. International Journal of Management, 11(4).

Zahera, S.A. and Bansal, R., 2018. Do investors exhibit behavioral biases in investment decision making? A systematic review. Qualitative Research in Financial Markets.

Disclaimer

The information in this article is for personal, non-commercial use only. All content of this article is for general informational purposes only. No information included in this article is investment advice of any kind. Plamen Todorov (“The Provider”) is a publisher and not an investment advisor. The Provider does not make any investment recommendations, nor does the Provider intend to influence users of this article in their personal investment decisions. None of the ideas in this article are meant to be construed as professional and / or financial advice. Readers of this article are solely responsible for their investment decisions. In exchange for using this Website and reading this article, you agree not to hold the Provider or any third party service provider which is connected to this website liable for any possible claim for damages arising from any decision you make based on information or material made available to you through this website.
© 2022 Systematic Bonds. All rights reserved. Plamen Todorov holds the international copyright to all content on this website. Any usage of text or images from this website is permitted only with the prior written consent of Plamen Todorov.