We do not guarantee the accuracy of this transcript.
Patrick Houweling (Patrick): I'm ready.
Erika van der Merwe (Erika): Patrick, do you see yourself as a quant investor or a credit investor?
Patrick: Do I have to choose? Because we're both. No, we’re both.
Erika: And exactly to your point, right? You need to have both skills.
This podcast is for professional investors only.
Erika: Quant investing in credit is still a relatively new approach, certainly compared with the track record that quants have in equities. Well, what's it all about and what are the unique challenges of this investment style? And, can it cope with shocks such as Covid-19 and major shifts in global policy regimes?
Male voice: Welcome to a new episode of the Robeco podcast.
Erika: Patrick Houweling, co-head of Quant Fixed Income at Robeco, has been a quant investor for many years and he's my guest. Welcome, Patrick.
Patrick: Hi, Erika. Nice to be here.
Erika: Patrick, what is quant investing? We've had a few conversations in this podcast series with quant equity investors, and there were a lot of references to factors, models and data. What does your domain look like?
Patrick: Well, it's very similar to quant equity, but it's also very different. So it has a lot of overlaps, but there are also notable differences. Maybe first we start with the similarities, because there are quite a few. It's all about, what we say, rules-based and evidence-based investing. So we construct portfolios, we select bonds to go into the portfolio based on predefined rules, and these rules are based on historical evidence. So we do a lot of research on historical data to find which rules have worked and which didn't. And these are called the factors. So this is all very similar and should sound familiar to you. But there's also differences in the sense that bonds are not stocks. And we need to sometimes adopt factor definitions to the specific asset class.
Erika: We'll elaborate on that in a bit, but I just wanted to make the point that a quant approach to credits is relatively new, as I said in the introduction. But you yourself are no newcomer to the discipline. You personally have been at it for more than 18 years at Robeco, which makes you a pioneer in the field. You're also a co-author of the chapter on factor investing in fixed income in Frank Fabozzi’s Handbook of Fixed Income Securities. So what's been the history of this capability at Robeco and specifically also your role in it?
Patrick: Yes, it's indeed relatively new to the outside world, but at Robeco we have been doing research on this for over 20 years. So basically in the early days of quant investing at Robeco, we very quickly, after we found that factors work in equity markets, my predecessors, even before me, they were already asking the question: Can we do this on other asset classes as well? And in particular on corporate bonds, because what they have in common with equities is that we're talking about the same underlying firms, the same underlying assets. So that's basically when the research started, at the end of the 1990s. And well, the first results were very encouraging and also led to a first academic publication by Robeco authors already in 2001. So this may be the very first academic publication on factor investing in corporate bonds. And since then we have done a lot of research, of course. And even though initially, these models were mostly used as idea generators for the fundamental portfolio managers that we have at Robeco, the models kept on improving and our confidence in the models also kept increasing. So at some point we started investing on this in a sub-portfolio, so to say an in-house client. And from 2012 onwards, we have also started investing money from external clients based on our multi-factor strategies. So that was maybe the most visible to the outside world. But before 2012, we had already been working on this for over a decade.
Erika: So what's the ultimate point, then, of a quant’s approach to credits? Is the idea to eliminate human bias? Before you answer, to illustrate the point, here’s an extract from the podcast The CIO Agenda, featuring Rob Lam. He's Co-Head of Credit at Man Numeric.
Rob Lam: 99% of high yield assets are managed using traditional discretionary processes. And when I think about that, inherently when the market composition is so one-sided, inefficiencies and distortions are naturally created. And so we believe that there's a really large opportunity to go against the grain here. But I think you're right, it is hard and it's hard to exploit these inefficiencies in the credit market because of some of those complexities in some of the systematic processes’ steps, right? And includes everything from being able to capture those key drivers of credit spreads to modeling the cost of execution and the probability of fill, which are actually quite unique to credit.
Erika: So he's emphasizing there the inefficiencies, which, to my mind, also relates to human biases that create some of those inefficiencies. But he's also emphasizing that it’s really complex in the credit markets. How do you see it?
Patrick: Yeah, there's a huge body of literature that confirms these behavioral biases. So things like overconfidence or herding behavior, overreaction, etcetera. And these are really specific to people, not specific to the asset class. So if we believe that people overreact, then we should be able to see this back in equity markets, corporate bond markets, government bond markets, commodity markets, etc. So in all the different asset classes in financial markets. So in that sense, it's not different. But there are indeed some operational challenges that are unique to fixed income markets. So whereas most firms have only one stock, a typical firm has multiple bonds and they can have different maturities, different currencies, different coupons, different seniorities in the capital structure. So that alone makes it more difficult to deal with corporate bonds than with stocks, because you have basically information at the company level and you have heterogeneity, differences at the security level. And next to that, as alluded to by the fragment that you just played to us, there’s the illiquidity challenge as well. So most bonds are not traded every day, some not even every month. There's huge differences between the liquidity of individual bonds. Transaction costs are also more sizable relative to the alpha that you can reap. So these are all things that you need to take into account in the research phase already, because you want the research to be as realistic as possible, but also and definitely in the execution, because you cannot just make a buy list of securities and assume that everything will be traded, everything can be bought in the market. So you need to take liquidity into account. So overall, there’s quite some cogs to the machine. So you have the factor model, you have the portfolio construction, you have liquidity and transaction cost modelling. But you also have a risk model, for example, that you need or you want to say something about performance attribution to try and understand the performance after the fact. So there's a lot of things that you need to work on, and this definitely provides barriers to entry. So a newcomer would have to develop a whole big machine, which takes years, which really takes years to develop.
Erika: So it's very clear then in your response there that it's about far more than just the academic research and the clever modelling and the mathematics and the programming, right. And time in the game and history and institutional knowledge really helps you and gives you that advantage.
Patrick: Yes, and therefore, it's really nice I think that at Robeco, we have both a strong quant capability and a strong fixed income capability. And it's really at the boundaries of these two capabilities that we have the quant fixed income teams. So we tap into the legacy and the heritage of Robeco in quant investing, but also its strengths in fixed income investing. And you really need both. And sometimes we see newcomers to the market that lack one of these pieces of experience. So they may have a quant capability in equity and are new to fixed income. So then you make all the rookie mistakes in fixed income, or they're an experienced fixed income investor and they think, well, maybe we can do some quant and then you make the rookie mistakes in quant. So it really requires both, and both at a high level, to do quant fixed income well.
Erika: So does it follow then, Patrick, that quant models behave differently from fundamental credit portfolios? How strong is the case for diversification from your work?
Patrick: That's interesting, because what we see is that on the one hand, quant investing is all about systematizing what fundamental investors could also do. So we are not reinventing the wheel. We're not working with a black box. So a fundamental credit investor looks at things like leverage or free cash flow, or at the credit spread relative to the rating, or all these measures. We also use them, but we use them in a systematic way, so we can apply them efficiently to the entire investment universe. So that, you would say, is very similar to fundamental investing. But the differences are also clear in the sense that fundamental investors are typically concentrating their efforts on a, let’s say, more large-cap, a more liquid part of the universe, and they typically follow one investment style. So they are, for example, value investors. And that is then very visible in their portfolio. In the case of quant investing, we have the full breadth of the universe. We can basically invest in everything that has data, so we have much more breadth, we have more investment opportunities. But also because it's a systematic portfolio construction process, we can efficiently incorporate multiple styles or multiple factors in the portfolio. So it is not only about value, but also about low risk and quality and momentum, et cetera. And that is very difficult to do. It's not impossible, but it's very difficult to do in a more manual workflow, to take into account different styles or different characteristics. And in a systematic quant investment process that's quite natural to do, because everything is rules based and systematic and data driven. So it is more efficient to do it like that. And these two things, that we have the full breadth of the universe and that we are multi-factor, whereas fundamental managers are mostly single factor, that makes that we have a different return stream. So it diversifies versus typical fundamental managers. And this is what we have also investigated and shown empirically. So if you look at a peer group of fundamentally managed global corporate funds and you compare their returns to a multi-factor fund, our funds, for example, then you see that the correlation is very low and from time to time even negative. So it's a good diversifier for an asset owner that has outsourced their corporate bond allocation to multiple external managers. And it would be really advisable for them to not only have fundamental managers in their pool, but also to have a quant manager next to that, because the overall return stream will become more stable because of the diversification benefits.
Erika: Patrick, picking up some of the points that you made about factors. Which factors specifically are important for quant credits, and is it totally comparable with equities? You alluded to that.
Patrick: Yeah. So it's really the usual suspects: low-risk, quality, value, momentum and size. So these are the five factors that we found to work in credit markets. And they're also, as I said, the usual suspects, they're also found to work in equity markets. And we also see that they are similar. They're not identical, but they're similar. So a value equity portfolio is positively correlated to a value credit portfolio and tends to outperform and underperform in similar parts of the cycle. That’s not to say that they are exactly the same, but over a very long investment horizon, you see that the factors across asset classes are positively correlated.
Erika: Okay well, let's take that further, because in our conversations with quant equity investors, there's plenty of references to the anguish of underperformance, particularly with the value factor. What's been your experience in this?
Patrick: Yeah, that's why I stressed the long-term investment horizon just now, because there's no guarantee of course, that factors work each and every year. And indeed, the quant equity investors had a bit of a rough stretch with their value factor, because that basically underperformed for quite a prolonged period of time. And my colleagues at Robeco and also other authors externally have done research on this. So why did value not work for such a long period of time? And the answer is big tech. So big tech firms, Apple, Microsoft, etc., they are very expensive from a value perspective, they are trading at high multiples because the market is so optimistic about their growth. And so the quant portfolios did not invest in the big tech stocks. But guess what? The expensive stocks became even more expensive, so this contributed to the underperformance of the value factor, especially because the tech companies are so dominant in the index. I believe they are now 15-20% of equity indices. So it's a large part of the market where the value factor was wrong and therefore value didn't work for a long time. I'm oversimplifying a bit because I'm not an equity investor, but this is what I understand from it. And this problem didn't occur on the credit side, because credit investors are not necessarily interested in growth of the company, they're just interested in stability, getting their coupons and getting their notional back at the end of the day.
Erika: How do you keep your mind in tip-top shape for clear thinking and reasoning? Because I must say with my interactions with you, you are super clear and crisp and I would imagine, suffer no fools. So how do you stay at that level?
Patrick: That’s a very difficult question.
Erika: I believe you go for lunchtime walks and that you are a regular lunchtime walker, so I imagine that's an important part of that discipline.
Patrick: That's what you're alluding to. That’s a kind of personal hobby of mine, to take regular walks. And yeah, there was an expression once that I read that it's say, physical wandering during walking is also mental wandering. So if you walk outside and you kind of forget your work, then you can also relax your brain in a way. So I do this quite structurally, as you can expect from a quant, I do this every day, multiple times and then you get the step counter. And for me that works, it helps to relax. And I think unconsciously, you also think about your work, even if you're not really concentrating on it. Your brain in the background continues to process it while you're doing other things. And while you're thinking about other things.
Erika: When it comes to your work, what keeps you up at night? What's the one thing that worries you?
Patrick: Performance. In the end, it's all about performance. So that's, I think the big difference between being a researcher in academia and being a researcher and portfolio manager in the industry, because in the end, we're investing clients’ money. They trust us and we, in a way, make them promises. We do our backtests, we tell them how we operate. And this comes with expectations. And in the end, we want to meet those expectations and, if possible, exceed them.
Erika: So Patrick, building on that point, the reality is that you’re working with models, and that's a simplification of the real world. You’re also working with data that's always imperfect no matter what you do with it. And I believe this is a real problem for quant investors. So how do you manage those imperfections, whether in a practical sense or in a philosophical sense? So I found a quote of yours where you say: in credit investing, one should hope for the best, but prepare for the worst. So it sounds like a mental attitude of always kind of being alert and on the lookout and seeing where you can tweak and improve.
Patrick: Yes, so we're on the one hand quite ambitious and we want to make very good models and deliver very good performance. But on the other hand, we also are very humble, and that has to do with the fact that the model is a simplification of reality. So we don't want to be too religious and trust our models blindly, because we know that you cannot capture everything in a quantitative model. We also know that data that goes into the model could potentially be wrong or outdated or no longer be representative. So there can be various reasons why you shouldn't blindly follow the model and you want to be bit street smart about it, and not too naive. And this is where our place in the Robeco organization comes at hand because we are closely linked to the fundamental fixed income teams. So before we make an investment based on our quantitative investment process, we have one of our fundamental analysts take a last look, and the question is really simple: did the model overlook something? And we ask basically, the analysts, are there extra risks beyond the scope of the model? Is the data that feeds into the model still representative? Are there reasons to overrule the model? And this is something that we are extremely cautious about, because we have done the back-tests, we've seen that models work, so it should really be an exception that we overrule the model, there should be a really good reason for this. So that's why we have put down some guidelines, what are the circumstances in which we can overrule the model?
Erika: And how often does that happen?
Patrick: It happens maybe 5% of the cases, less often in investment grade, so the higher-rated firms, and more often in high yield, the lower credit quality firms. So on average, about 5%. And some examples maybe make this come more alive. So in the beginning of last year, we saw in the model that Wirecard was scoring very well. So it had good quality, good momentum, valuation was attractive, etc. So it was one of the potential buys in our portfolios. And then we asked the analyst: Did we overlook something? And then the analyst said: Yes, you did, because there is talks in the newspapers of fraud, accounting problems, poor governance, et cetera. So we did not invest in Wirecard. Well, looking back now, that was a good decision because Wirecard went bankrupt. But, if you are a pure quant investor, you would have bought it and you would have fallen into this trap. And what makes this difficult for a quant is that it's about fraud and about the quality of the company, of the governance structure of the company. So this is very difficult to put in a quantitative model or to backtest empirically, you don't have historical data on it, but it can have a material impact on the firm and therefore on your investments.
Erika: Then what about major shifts or, you know, regime changes? I think over the past year or more, we've spoken often about Covid-19 and what that might mean to models, so you needn't go into that. But for instance, what about the role of inflation? We are now at a point of great uncertainty over what the future is for inflation, and there may be a regime shift here. How would a quant model deal with that?
Patrick: So in the end, we believe that it's all about human behavior and financial markets pricing in certain risks or events or news about companies or the macro economy as a whole. And in that sense, there's nothing new. But on the other hand, we sometimes worry whether the historical data that we have contains enough information about future possible events. And one such worry could indeed be inflation, because we have been in a very-low-inflation regime for the past few decades. And now we are seeing headline inflation figures above 5% again in the US, for example, which is the highest number in 25 years. So we did do some research on this, and what we did basically is look at our historical data that we have and ask the question: Did the factors behave differently in times of rising inflation or decreasing inflation? And then we must acknowledge that we didn't really have periods of really high inflation, of 10%, which we saw in the 1970s. But we have a lot of observations on declining and increasing inflation in our data sets. And the reassuring conclusion was that, well, in the end our multi-factor strategy was quite robust to these inflationary regimes. So we did see some differences, so some factors worked better in terms of rising inflation and other factors in times of decreasing inflation. Other factors have more all-weather performance, so they were quite robust. But on the strategy level, so combining multiple factors in the portfolio, we saw that the strategy was quite robust, so we were quite happy to see that result.
Erika: And I would imagine that this could be an example of other such events, right, so that the principle could be applied to other similar concepts.
Patrick: Yes, exactly.
Erika: So, Patrick, the shift to sustainable investing is relatively new for fundamental investors, even though Robeco has been at it for many years. How does sustainability feature in factor credits?
Patrick: Well, sustainability is well integrated in our multi-factor credit strategies, and we incorporate different sustainability dimensions in our investment process. So think of ESG scores, carbon emissions, but also other environmental footprints like water use or waste disposal. These are all integrated in the investment process. Also, of course, the Robeco exclusion policy. So we don't invest in companies that have shown controversial behavior or produce certain products like tobacco and controversial weapons. The challenge here is that it's not always possible to derive conclusions from historical data, because who cared about carbon emissions in the 1990s? But still, we are kind of expected to provide an answer to this question. So how did a carbon constraint on your portfolio affect historical risk and return? Will it make your backtest better or worse? And that is typically asked from us as quant investors. So we can go some way, in the sense that we can do some simulations with historical data to the extent that we have it, or simulated data, and then we can make some general conclusions about such and such a constraint that impacts the risk or the return of the strategy. So in case of carbon, for example, we saw that historically it's not impacting the risk and return a lot, because the distribution of carbon emissions is very asymmetrical. There's a small group of heavy polluters. So if you exclude those, then your portfolio average goes down significantly already. So that helps to make an assessment of how feasible it is to make such a change to your investment process, to add a sustainability constraint, in this case a carbon constraint. So this was the quant side. On the other hand, we have more the fundamental beliefs of Robeco as a firm, and we do believe as a firm that climate risk, for example, is very important, that companies face transition risks and that you have winners and losers or leaders and laggards. And from that perspective, we strongly believe that we should invest in companies that are at the forefront of this revolution, rather than companies that lag behind. So this is maybe a bit unnatural for a quant investor because you're asking to implement fundamental beliefs in your investment portfolio, which is otherwise very systematic. But it is something we do because this is maybe a structural break versus the past where we can use only little historical data to draw inferences for the future, and clients are also demanding this from us. So they're not just asking anymore for a portfolio, but they're typically asking for a sustainable portfolio, and sustainability is maybe not fully standardized. So if you have 10 clients, they have 10 different wishes. But this is something we can accommodate. So we have all the data, et cetera, we have the expertise, so we can build these customized portfolios for clients that meet different sustainability requirements.
Erika: Right. Sounds to me like a major innovation that you've designed and implemented over the past year or two. Looking ahead, so in closing, Patrick, looking ahead over, say the coming five years, what do you expect to be some of the biggest innovations in your team's work?
Patrick: Yeah, so it will definitely be on sustainability. No doubt about that. Everybody is talking about it. Everybody's thinking about it. There will be more data, better data, maybe also worse data, but that also makes it interesting from an intellectual point of view. So there will be a lot of research on sustainability. But the big trend in quant investing is alternative data or big data, as it's sometimes called. So as quants, we used to rely a lot on numerical data, structured data. So think of financial markets, information like bond and equity prices, or think of accounting data like the profitability or the earnings of firms. So all very structured and numerical data and the big change in the quant space has been the emergence of alternative data. So new data sets like credit card information, the text in corporate filings, satellite images. So there's a whole range of applications possible. And there again, the challenge is the depth and breadth of the data. So can you back-test these things? If you talk about satellite images, well maybe you have only five years of data so difficult to back-test. Maybe it only applies to a small number of companies in your universe, so maybe only companies that have a physical location where people will come with their car to visit the store. So then you can see from the satellite: is the parking lot busy or not? But this is not useful for a power plant. So new data come with new challenges. But some of the newer data, or the alternative data does have a long history, and we're spending most of our efforts there. And one example would be textual data. So companies have filed annual reports for a long time. There is history available and we can process these texts and see what companies write about themselves, and it's quite informative. And do they use a lot of positive words or negative words? Are they concise or lengthy? So that's a whole new area of research that we're looking forward to, to learn new things and to again improve our models.
Erika: Sounds as though there's plenty of work for you to do in the years ahead. Patrick, thank you so much for this conversation. Lovely talking to you.
Patrick: Thanks, Erika.
Male voice: Thanks for joining this Robeco podcast. Please tune in next time as well. Important information: this publication is intended for professional investors. This podcast was brought to you by Robeco and in the US by Robeco Institutional Asset Management US Inc, a Delaware corporation, as well as an investment advisor registered with the US Securities and Exchange Commission. The podcast is intended only for institutional investors. Robeco Institutional Asset Management US is a wholly owned subsidiary of ORIX Corporation Europe N.V., a Dutch investment management firm located in Rotterdam, the Netherlands. Robeco Institutional Asset Management B.V. has a license as manager of UCITS and AIFS for the Netherlands Authority for the Financial Markets in Amsterdam.
当資料は情報提供を目的として、Robeco Institutional Asset Management B.V.が作成した英文資料、もしくはその英文資料をロベコ・ジャパン株式会社が翻訳したものです。資料中の個別の金融商品の売買の勧誘や推奨等を目的とするものではありません。記載された情報は十分信頼できるものであると考えておりますが、その正確性、完全性を保証するものではありません。意見や見通しはあくまで作成日における弊社の判断に基づくものであり、今後予告なしに変更されることがあります。運用状況、市場動向、意見等は、過去の一時点あるいは過去の一定期間についてのものであり、過去の実績は将来の運用成果を保証または示唆するものではありません。また、記載された投資方針・戦略等は全ての投資家の皆様に適合するとは限りません。当資料は法律、税務、会計面での助言の提供を意図するものではありません。
商号等： ロベコ・ジャパン株式会社 金融商品取引業者 関東財務局長（金商）第２７８０号
加入協会： 一般社団法人 日本投資顧問業協会