Letter #91: Ray Iwanowski (2019)
Letter #91: Ray Iwanowski (2019)
Founder of SECOR and Co-CIO of GSAM Quant Investment Strategies | Conversation with Dean Curnutt on Alpha Exchange
Today’s letter is the transcript of a conversation between Ray Iwanowski and Dean Curnutt for the Alpha Exchange podcast. In it, Ray discusses how he went from studying math and Russian literature to finance, being struck with inspiration after coming across the Black-Scholes equation and option pricing theory, his journey from Penn to Chicago to GSAM, learning from Eugene Fama, working with Cliff Asness, factor investing, machine learning, the 2007 quant crisis, crowding risks, factor timing, and the correlation of momentum and value strategies. He also dives into big data, machine learning, and the volatility risk premium, how it has evolved over time, and the reflexive properties of volatility.
Ray is a Co-Founder of SECOR Asset Management, a global asset management firm that partners with clients across three business areas: quantitative investing, portfolio solutions and pension advisory. Previously, he served as Chief Investment Officer of SECOR’s quantitative businesses. Prior to founding SECOR, Ray was Co-Chief Investment Officer of the Quantitative Investment Strategies (QIS) group at Goldman Sachs Asset Management (GSAM). While at Goldman Sachs, Ray was named Managing Director in 1999 and Partner in 2004. Before GSAM, Ray worked at Salomon Brothers as head of the Fixed Income Derivatives Client Research Group and member of the Bond Portfolio Analysis Research Group. At Salomon, Ray authored a number of publications on fixed income asset allocation and fixed-income derivatives. He also worked in First Boston’s Fixed-Income Portfolio Strategies Group, specializing in asset-liability issues for banks and thrifts, as well as valuation of mortgage-backed securities and their derivatives.
Transcript
Dean Curnutt: Hello, this is Dean Curnutt, and welcome to The Alpha Exchange, where we explore topics and financial markets associated with managing risk generating return, and the deployment of capital in the alternative investment industry. There's not much natural intersection between the study of mathematics and Russian literature. But for the ever curious mind of Ray Iwanowski, the Wharton School provided exposure to both. Ultimately, Ray's interest in math and physics would lead him to finance, where he came upon the Black Scholes equation in option pricing theory. After a stint in fixed income research focused on modeling mortgages, Ray set upon the PhD program at the University of Chicago in the early 1990s, a vibrant time for advancement in the empirical study of asset pricing. Utilizing the toolkit he developed, Ray landed at Goldman Sachs Asset Management, where he ultimately co-ran the firm's Global Alpha business. Today, Ray is cofounder and CIO of SECOR Asset Management, a firm that provides customized portfolio solutions to institutional clients around the world. My conversation with Ray considers the current state of factor investing in light of the increasingly competitive search for alpha. In the process, we look back on the 2007 quant crisis, exploring the questions of factor timing, crowding risks, and the correlation of momentum and value strategies. We also look forward as Ray shares his views on harnessing data and utilizing artificial intelligence and machine learning. Lastly, we delve into the volatility risk premium, how it has evolved over time, and the reflexive properties of volatility. I thoroughly enjoyed this discussion and the perspective Ray offered through his experience as a quant investor. Now, my conversation with Ray Iwanowski, on this episode of the Alpha Exchange.
Dean Curnutt: Ray, it's great to be in your office today, and welcome to The Alpha Exchange.
Ray Iwanowski: I really appreciate the opportunity to participate.
Dean Curnutt: Absolutely. Lots of interesting things for us to talk about. I always like to get started by learning more about your career path, how you came to really embrace the discipline of finance. So sort of take us back to the beginning. What is it that struck that chord for you with respect to finance?
Ray Iwanowski: Like a lot of things in life, a lot of people's career paths are sort of driven by perhaps being at a certain place at a certain time. And I could definitely say that about my career as a quant investor. I was at University of Pennsylvania as an undergraduate in the mid 1980s. And I had a wide range of interests, ranging from mathematics to Russian literature. And one thing I did not have interest in, when I entered Penn, was business or finance. So I was not in the Wharton School. And I was studying, I really enjoyed mathematics. And those other things like Russian literature, were really interesting to me, but I couldn't see a career path there. So as time started evolving, I was learning that over at the Wharton School, there was classes in finance where people were starting to apply mathematics or physics or other things to solving financial problems, trading issues, investing, things like that. So that piqued my interest. And I went over and took a couple classes, and one example of the types of things that we learned there that was transformational to me was, we learned about, for example, the Black Scholes equation. And as a math student, if you know the evolution of Black Scholes or the derivation of Black Scholes, Black and Scholes showed that option pricing formula could come out of a solution for the heat equation in physics. And I found that fascinating. And at the time, there was a view that maybe this could give me an edge in investing in trading. And my motivation to get into it was not to try to make a lot of money and things like that, it was really seemed like this was like a competition where you can get an edge and win through the use of applied mathematics and other types of statistics and other types of quantitative techniques. So that was the impetus for me to get into finance and investing. Now, it turns out, for all of us who know, just knowing the Black Scholes equation, even back in the 80s, wasn't really an edge. That that edge went away pretty quickly. But it did sort of motivate me to get into this business and try to find other edges and other interesting problems to solve and invest in.
Dean Curnutt: I've been teaching a class at my own alma mater, St. John's, and do some work on the Black Scholes equation. We don't derive it, but I present the assumptions. Some of these assumptions are pretty easy to skate past, you know they're not true. We'd all like to pay no taxes. We'd all like to bear no transactions costs. What I try to get the students to really focus on is this this assumption of the normal distribution, that stock returns are normally distributed, and what do we see time and time again? So I think that's a topic that we'll come back to over the course of our conversation, but it's one in which, as you say, they solved something. Given their set of assumptions, they prove that this is the answer. But how seriously do you take the assumptions around the normal distribution? I think that's a big question.
Ray Iwanowski: That's a great point. And that's, in my opinion, part of the the fun of investing and trading and markets and using quant techniques. I think, like Black Scholes, take another example, things I learned in graduate school, like the concept of market efficiency. Those concepts are based on assumptions, as you said, that are almost definitely false. But when when people critique these things, they'll say, this doesn't work because that assumption is false. But like, as one of my professors at Chicago used to always say, It's a model, it's not a reality. So models will, by definition, have flaws. And what's really, the good users of models think about is, Is this a framework to think about it, is Black Scholes or market efficiencies, good frameworks to think about it? And then if you think there's deviations, that might be your edge, but you want to think deeply about what those deviations are, and what's the impact. So you wouldn't say, if you were pricing an option, you wouldn't say that because markets generally aren't necessarily normally distributed, all the other things that come out of Black Scholes, inputs like volatility, or market efficiency, where you say, It's really, really hard to beat the market for a sustained period of time, are probably good frameworks to have. And then, to the extent that you deviate, you should look back and say, Does that make sense relative to this starting point, which is a reasonably well thought out model.
Dean Curnutt: So obviously, you took to finance during your undergraduate period. What was post-coming out of Penn? Take us through the early parts of your professional career.
Ray Iwanowski: So I was fortunate enough to get a job in fixed income research at First Boston. Now it's part of Credit Suisse. And the more quantitative areas was in fixed income, and I worked in fixed income derivatives and also the the mortgage market was taking off and mortgage markets, it was becoming better known that quantitative techniques are valuable there, and things like understanding the prepayment option, or using statistics to sort of model how prepayments work. And so I spent a couple years in fixed income research at First Boston. Really enjoyed it a lot, but it motivated me to say there's more to learn. And it was a good time to go to graduate school. So I then joined the PhD program at the University of Chicago. And that was, for me, a transformational event. There I learned a lot of the underpinnings that to this day drive how I think about modeling and the markets. There was a lot of exciting--this is in the early 1990s now--there's a lot of exciting things going on at University Chicago, the professors, and in the PhD program, around factor investing, market efficiency, innovations in how fixed income is priced. So, to me, that was just an incredibly intellectually awakening period. I made a lot of, as we could talk about, a lot of friends and network that also helped in terms of the development of my career. I ended up not finishing my PhD. I passed all the exams and the coursework, but I got itchy to get back into the industry, so I left without finishing the thesis, and got my MBA, and then then worked at Salomon Brothers.
Dean Curnutt: So I was at Chicago in the MBA program in the mid 90s, and also found it to be so enriching, just the framework for thinking about finance. And of course, Gene Fama was the academic and intellectual leader of the faculty. What was that experience, just in the Ph. D. program? Where were your areas of focus? Was it asset pricing, derivatives, sort of the theory of financial decision making? Where was your time really spent?
Ray Iwanowski: So first of all, the beautiful thing about it was just the breadth of things you could actually learn. That was my motivation for going to graduate school, because I was learning fun things and working on fun things in fixed income research even at First Boston. And so it wasn't a lack of fun problems to learn. But I felt that I'd benefit by getting more breadth. So not only, we'll come to the specific things in a second, but learning econometrics or even learning microeconomics from some of the best. So you mentioned Gene Fama. He was one of my mentors, and a lot of people that were in the program. But I took a couple courses from Gary Becker, who won a Nobel Prize. And then some of the econometrics and statistics professors are among the best in the world. So it was just an incredibly fertile environment for thinking. And so a lot of times, even this stage of my career, much later, you pull things off the shelf that you had learned there. And then the other thing is a set, a network of friends and colleagues and fellow PhD students. It happened to be a bunch of us. Academics want PhD students to kind of groom more academics. And I think that era at University Chicago finance department, there might have been a little bit of disappointment from our professors that a lot of us went into the industry. So one of my friends and ultimately, my boss, when I worked at Goldman Sachs, was Cliff Asness. And probably a lot of people listening to this podcast know him as the founder of AQR. A number of the other founding partners of AQR were colleagues of mine. Some of the--my cohead at Goldman Sachs, Mark Carhart's work was in the PhD Program at the time. So like a lot of good academic experiences, you develop a network of friends and colleagues that you admire and you go on to either work with them or correspond with them in future life. And just getting to your question about other things that you learned. I think there's actually a piece that Cliff Asness and John Liu wrote in the institutional investor that I highly recommend, if anyone gets a copy of it, after Gene Fama did win the Nobel Prize, and they summarized their view of Professor Fama's contributions to their career, as well as where we might deviate. So a lot of people who remember, understand, Chicago school of finance and economics, and particularly Professor Fama, everything's grounded in the concept of efficient market hypothesis. And even then, even in the 1990s, even with people like Gene Fama, nobody took the extreme assumption that markets are perfectly efficient. But you know, it was a very good starting point, again, to just say, It's really, really hard to beat the market. And if you think you're beating the market, are you really just taking certain risks that you don't really want to sign up for? And so their piece does a nice job of saying, These are the things that we learned and our foundation that came from academic finance over the years, including Professor Fama. And here's the mild deviations we take from that in trying to kind of provide additional premium, risk-return trade-offs. And maybe one might argue that markets are certainly not, not even perfectly efficient, there's persistent inefficiencies that can be harvested and delivered.
Dean Curnutt: Right. So post the experience at Chicago, you're at Goldman Sachs. We had a couple of pretty interesting risk events in the 90s. Nothing like the 2008 event, but maybe an appetizer through things like the Asia Crisis in 97. And, of course, LTCM. So as someone who was a specialist in fixed income research and derivatives, what can you remember about the LTCM fiasco?
Ray Iwanowski: Yeah, I think backing it up, but even a little, when I came out of Chicago, I went to work at Salomon Brothers in fixed income research. And it was similar kind of work that I had done at First Boston. But now I had the added toolkit, if you will, from the things I learned at Chicago. And then Solomon was a place where quantitative fixed income research was born. Again, my mentor at Salomon was a guy named Marty Liebowitz, who's kind of considered the father of a lot of bond analytics. So we were doing interesting things. And I think one sort of lesson that was sort of bubbling up at that time, that became clear later, when LTCM hit, by then I was at Goldman Sachs Asset Management. But you could think of it, derivatives, for those of us who, those were the early days of derivatives markets, and those of us who are quantitative and think deeply about it, we think derivatives are great tools for a number of things: to get exposures, to manage your risk, to hedge your risk. So there's all kinds of textbook usages of derivatives and high value. And to this day, I stand by that. There was also kind of a frenzy of structuring derivatives to get certain exposures that, obviously, Wall Street was benefiting from a lot. Tranching things, tranching risks, those concepts are all good, but I think things got way overblown, people underappreciated what was the embedded leverage. And so there was a period that when you look back, people will say, the structurers and the derivative traders were making a lot of money for what they would take out of these instruments, and then customers were buying, or investors were buying, these things in ways that they underappreciated the risks or they under appreciated what the broker-dealer community was actually taking out. So in the benefit of hindsight, and some of the things you mentioned, as well as like, the Orange County experience in 1994. 1994 was the year where yield curve moved violently and a lot of people had difficulties. You could kind of take a look back and say, what started as a very good concept by people underappreciating, people getting overly exuberant, caused a lot of pain and losses that people underappreciated what the risks were. And then, LTCM. By then I was cohead of the quant business at Goldman Sachs Asset Management. I had gone to work for Cliff Asness. And then when he left and started AQR, I was named cohead with Mark Carhart. First of all, we navigated through that experience. We didn't have the difficulties that LTCM did. But there, that was a good case study and a lot of lessons, if you think about it. One is appreciating, again, having leveraged positions, what are risks that aren't captured by standard measures like volatility when you take a lot of leverage. And then, having really sizable positions, especially OTC derivatives positions, where counterparties have some control over what happens to you. So we were fortunate enough to not have a really bad period during the LTCM period, and in fact, we were in a position where we were to actually take some positions on the unwinds in some of the volatility markets. But if you're managing money quantitatively, the people that LTCM were among the most highly respected quants in the world. And, sort of mistakes of assumptions that drove their process were learning experiences for all of those who would do it subsequently. And so you spent a lot more time thinking about tail risk, counterparty risk, margining risks, and things like that, from that experience.
Dean Curnutt: What's interesting is, you step back and you think about LTCM. And there's so many lessons there, right? I think It's the all star team that you'd never seen any collaboration amongst the rock stars from the [Salomon bond arm] team, but also folks like David Mullins, you had Robert Merton, and Myron Scholes. What more could you really ask for in terms of putting together this never seen before team? And what's to me so fascinating looking back on it, is the sheer size of the portfolio. And as steeped as they were in risk management and modeling, to really miss the, maybe the misunderstanding of the risk that your trades themselves become part of the market risk dynamic, and back to Black Scholes assumptions, one of them is markets are infinitely liquid. They're not. When you're short $80M of five year Vega, it's hard to buy back, like you said, especially when It's when it's OTC. And the other part was, there wasn't the concept of risk-on, risk-off back then. That word was not in the lexicon, but that was an entirely risk-on portfolio. It was short vol in every format you can imagine. And that, to me, is just, looking back on it, is pretty interesting.
Ray Iwanowski: I think you just touched on a couple very important lessons. Like first of all, as you said, those guys you mentioned, were, for those of us who are in quantitative finance, were heroes of ours. I mentioned the Black Scholes equation, right? All the other people you mentioned as well. So you you can't look at it and be, in the seat I was in, and say, Oh, those guys are not that smart. Because that's just not true. I think, some lessons, it was one of the early times to let to have me think about the limitation of models. So I think quant investing gets, in those days, certainly, but even now, it's vastly misunderstood in the sense that people get overly enamored with it when it's doing well, and saying, This is this magic box, we solve markets. That's just not true. Nobody solves markets. But at the margin, using a systematic process to make the investment decisions to manage a risk is better than not, in my humble opinion. But there's ways that the models can lead you in a place that, you used the term earlier, assumptions, that turned out to be wrong, or just an outcome that you hadn't seen before. So it teaches you, and we could get to the 10 years later, 11 years later, the quant crisis, 10 years later, you have to push on your assumptions and think about whether things that are driving the model haven't been seen before, but could could materialize. And I think for those guys, counterparty risk. And then you mentioned something that was definitely a learning experience. I'll get quanty for a second, to restate what you said. I think many quant models, at the time and maybe even now, assume pricing is exogenous, meaning you observe market prices and then you act on those observed set of market prices where a lot of times, especially if you're sizable, or fast forward 10 years, especially if you're doing the same things as others, is pricing becomes endogenous. As you try to trade, you're moving prices in a direction against you. And then that's sort of your classic liquidity spiral where things can go horribly wrong. And then there's things we learned about fixed income markets at the time. So, those of us who were relatively unaffected by that period would look at their experience and say, Okay, now my new process, call it, the 1999 2000 version, has built in these safeguards relative to what we learned from them. The problem is you can't, it's hard to envision everything. So again, I think you intend to talk about the quant crisis, those of us who were affected by the quant crisis, when we reflect back on our assumptions, yeah, we fixed it so we didn't get hit by the LTCM problem, but there's other things that materialized of the similar kind of circumstance, but in a different place that I think we underappreciated.
Dean Curnutt: Right. Let's talk about that. I think, for me, and I'm not a quant, I found the August 2007, it's a fascinating development, because even though markets were volatile, and you had things happening, sort of the precursor to the big financial crisis, it was happening in 2007, you started to see the liquidity concerns and subprime and BNP had their, it was August 2007, where they had to gate on some basic fixed income that they just couldn't unwind. You started to see the cracks. But that quant crisis really happened beneath the surface. It wasn't like the VIX went to 60 during August 07, it was probably low, very low 20s. So to me, the learning experience was what, boy, a lot of these strategies wound up being in very much the same stuff at the same time. How do you think that that occurred? Was that just the success of these strategies for a number of years that they just got too big? Broadly speaking, what are the lessons that we can derive from that?
Ray Iwanowski: Yeah, and I think the lessons we derive from that that I'll describe are ones that thoughtful quants, that are managing money now, ourselves included, should build those lessons into your process every day. So some of the things I'm going to describe, speaking for our process, and I know other people are that way, tend to have protections on what I'm about to say that, this experience deeply affected how we think about models and markets. I think one advantage of having an experienced quant manager versus somebody who's relatively new, even if their technical skills are really strong, is having lived through these periods, kind of motivates a clarity of thought on some things that you hadn't thought about before. But you got to it, you started talking about size. First of all, let me say, make a blanket statement that investment managers, quant and otherwise tend to vastly overstate the capacity and the strategies that they're in. So, speaking for ourselves, we ran a pretty sizable business at Goldman Sachs, but it's unfair to say we didn't think about capacity, or we didn't think about potential tail risks of being too big. We actually, as we started getting bigger and bigger, agonized over it. But I think, in retrospect, there's a lot of ways that being too big can hurt you. And it might not hurt you on an ongoing basis, but then you get into a situation where the world is changing in some form, or something's happening, and there's a lot of ill effects of being too big that I won't spend a ton of time on but, just in general. And I think what happened, specifically at that point in time is, speaking for myself, ourselves, but also probably broadly for the industry, we vastly underestimate how much money was actually being run in these strategies. So we analyzed it, but you see your competitors, at that time, we underappreciated how much quant money was being run by prop desks of the banks, at that time. Before Dodd Frank, there were tons of money being run by prop desks. And there's a lot of these sort of multi strat hedge funds, big hedge funds, where because of the success of quant, there were people saying, We want some of that too, and so we're going to develop some quant capabilities. So the first thing is just the sheer dollars. Second thing is we underestimated, you mentioned, being in the same positions, the nature of our factors, the nature of our signals. We knew a lot of the concepts that were in our models are ones that people still talk about today, things like value and momentum. We knew that any MBA who studied at University Chicago would learn about value and momentum and can build those models in a spreadsheet. We felt that our versions of it, although cousins and reasonably correlated, there was enough differentiation that, it's not like we weren't going to be correlated, but we would think, by this differentiation, we would be in slightly or more than slightly different positions. I think what happened was, we overstated how different our models, our factors, actually were. And then as you get bigger and bigger, again, you got to start going for positions that are liquid enough. And that tends to be a problem and put you in the similar positions. And then the third thing is, and I think you want to talk about crowding again at some point, but there's a lot of subtleties about, What, I'll call it complexities, of what do crowded positions mean? So if you are in a position and people are crowding into it, you might get a good, you might get some good results by the rising tide lifting all boats so to speak. And then on an ongoing basis, there's nothing necessarily problematic about being in a crowded position. And then you also have to philosophize on like, What do you mean by crowded? If somebody's buying, somebody selling. So I think the biggest problem with, what the real risk of being in crowded positions are, is, if being in these crowded positions are susceptible to large dislocations, if a bunch of people who are also in these positions, feel the need to liquidate at the same time. So everybody's heading for the exits, and even if you're not heading for the exits, your pricing is being affected. So if you have a leveraged position, or you have margin, you might be forced to sell if the position values are deteriorating in the short term. And so I think one of the things that happened in the quant crisis is, when we looked across the industry, we were generally longer term players, our factors were a little bit longer, we had lock ups on our capital, our clients were generally happy with us, they weren't going to pull the plug right away as things gone south. But some of these other players I mentioned, maybe a prop desk, maybe a quant sleeve of a multi strat, might say, as dispositions deteriorated, or even as things are going on in other spaces, like credit, which was less liquid, there appeared to be a goal that some of these big entities that were running quant money to liquidate pretty quickly. And that started as sort of a pricing spiral, that then once the prices started deteriorating, and we were in similar positions, we either from a risk management standpoint where we were losing enough money that we needed to adjust our positions, or even on a margin call basis, needed to trade. So once you're in that situation where you need to trade into the storm, really bad things can happen to you.
Dean Curnutt: August 2007. Is it motivated by one, maybe this long, very benign credit cycle where people's books got bigger and bigger, and so maybe quant is just some version of that? People's credit books were huge, there was a lot of there were a lot of positions, right? There was a lot of risk being taken, not just in quant, but we all know the broker-dealer value at risks, their daily risks were quite high. So was it a function of just the size as a result of just this leveraging behavior that occurred over a couple of years? Is it related at all to the breakdown of credit risk through housing? And then, maybe a little bit unrelated, but is it a function of the starting point of the valuation of the factors? And this is, we'll talk more about this, but was it not just the size, but the pricing of the factors got maybe overpriced? I'm curious what you think on that.
Ray Iwanowski: So there's sort of two elements to this right. One was idiosyncratic to the quant space, what I just described before where there was a period in August of 07, that a bunch of people felt the need to push to get out of the markets. And that did influence, to get out of the factors, it did influence pricing. So there was a lot of storm happening around quant, as you mentioned, it wasn't even, like other parts of the markets were flat, you were just seeing the stocks that were held by quants moving a lot. So there's an idiosyncratic component. There was a broader macro component that I think also had a lot to do with this. So subprime crisis was happening a little bit earlier, and I think that motivated some of this need for some of the players to derisk and instead hit credit. But also like, for example, in retrospect, we adapted our processes to be a little bit more dynamic, to react to some things that might be early warning signs of bad things to happen in your space. So when I when I tell the story, I use the analogy, there's a canary in the mine and the canary dies, then you should get out of the mine. So subprime quite crisis was happening, or starting, and some of the reactions of ours and others, I'm sure, were like, Oh, thank God we don't have subprime exposure. That would be meaningfully affected. But we should have said, Is this the canary in the mine that says maybe we shouldn't carry certain risk factors like currency carry related positions, that history has demonstrated that in periods of disruption, like the subprime, you probably don't want to be in carry factors or trades. And that, the client specific thing happened in the summer of 2007. And then a lot of it either rebounded or stabilized. And then you saw the same thing happen to other spaces in 2008, like in credit. I mean, when you think about it, the quant crisis of 07 meaningfully impacted me and others who were managing in many ways, we lost, we had outsized losses relative to expectation, our clients were disappointed, we were all disappointed that our processes performed that way. And there was a lot of focus on it in the press. It sort of paled in comparison to what happened in other spaces over a prolonged period in 2008. So the same lessons to be learned about having positions that are affected by other people heading for the exits, a lot of credit hedge funds, or funds, or any kind of fixed income, felt that at degrees of magnitude higher than what we did, and their positions were significantly less liquid than what we had. So it's a lesson that non quants should take about what happened in quant. And for me, subsequent to the clock crisis, as I was developing our process here at SECOR, we think about, Okay, we're not so necessarily susceptible to if 2007 were to happen again, but how about these lessons in other kinds of liquidity crunches? Is there anything analogous to that in our process that can get us into trouble.
Dean Curnutt: So maybe step back a little bit, just on factors in general, and maybe it would be helpful to hear from your perspective, when you think about almost the taxonomy of factors, we were talking a little bit earlier, and some folks have different views on the number of factors out there. But there's quite a few, if you really ask someone to generate an exhaustive list for the the non-quant, we're associated more with things like value and momentum, and we talked a little bit earlier, just about the subset. From your standpoint, what are the core, if you were to really try to, again, to the non-quant, What does the core list of factors look like?
Ray Iwanowski: Let me start by saying, I think the increased interest in factor investing, and thinking about factors in terms of risk is a great thing, unequivocally, for the investment industry. So the fact that pension funds are thinking about it, I think there's a lot of good things that have come out of that. And I think it's also good that some players in the asset management industry are offering sort of pure factor exposures and things that people might want in their portfolio or might help them manage the risk. And I think that's a great innovation that some firms have done really well, kind of building out products there. And through all kinds of noise, that's unlikely to go away. And that's an innovation that's valuable for the investment industry. As I was talking about with derivatives, sometimes these sort of enthusiasm or even frenzies get way overblown. And so there's this influx of trying to find hundreds of factors, and maybe factors that might have some some issues or some problems. So let me mention some of those. There's an interesting debate going on, or conversation going on, in academia, where, and by the way, most of what I'm going to describe is generally been researched or debated in US large cap stocks. So a lot of factor investing, what we do, and what other people do is, you might find these factors that are valuable in fixed income or currencies or commodities. But a lot of the sort of the academic literature focuses on US stocks, I'll focus on that for this conversation. And so like, people have documented that there's 300-400 factors that have, kind of concepts, factored-based concepts that have been published in credible academic journals. And there's sort of a cottage industry of PhD students or academics will say, Here's a new factor that prices US stocks that's different than the old Fama French value or momentum, and then the industry has latched on to it. So some of the people who have kind of said here, value and momentum are good factors. Now, there's people who will give you a laundry list of 300 factors and say, We'll get you exposures to any of these that you want. So some of the academic call it backlash, if you will, is some some good work that people, for example, there's a an academic called Campbell Harvey who had pointed out that a lot of these types of factors that had been found, might have been sort of data mined. And in academic finance, data mine used to be a derogatory term. Now, people who are employing machine learning techniques and stuff, say, We want to data mine. But what it meant in academic finance was finding spurious relationships in the data by kind of overfitting or over-mining it. So Harvey and his coauthors show that there's a lot of factors that have been rejected. So when you're trying a lot of things, you're likely to stumble on some things. So they postulated that the bar needs to be raised to accept factors. And I think that's a good thing. And then other academics have shown that a lot of these other 300 factors are subsumed by a smaller subset. So sort of the academic literature and some of the alt beta, factor based, premium based products, kind of focus on about five concepts. And those, value, momentum, you mentioned there's quality, there's volatility-based, and I think that's not a bad sort of subset to start. Now, what people like ourselves, who, we sort of manage active quant, in some kind of hedge fund form, would say, Okay, maybe you can get those factors by, fairly liquidly, fairly cheaply, from certain kinds of firms. Can we find things, can we find signals, I like to call them attractive risk return trade offs, because there's a lot of semantic debate about, Is it really a factor? Is it really a risk premium? Or is it an anomaly? Is it an alpha, not a free lunch, but a cheap lunch? I think it's probably healthy to stay away from that debate in the sense that I'm not gonna sit here and say, Is this a factor that's priced? Or is it just people being irrational? But we're trying to find, whether it's factor--not everything we do is factor-based, by the way, we do some strategies where we look at trading around an economic data release, or some of the things we do in volatility aren't technically factor-based. But for the sake of this conversation, we'll just focus on factors. And do you find things that are incremental to those five or so concepts that maybe are capturing phenomenon that's not those things? And some of the most successful quant hedge funds in the world, I think, clearly demonstrate that they're finding regularities in the markets that are different than just value and momentum and quality.
Dean Curnutt: Right. And a lot of the early work post-finding some of the sources of return that couldn't be explained by the basic CAPM model that started to go international. And you noticed things, that it was consistent, some factors that were persisting internationally. What about cross asset?
Ray Iwanowski: So one of the things that we look at is, you could call it macro, but we apply factor technology on thinking about commodities markets versus equities markets versus bond markets. Now, one interesting thing, like one thing I like to, I keep getting at is, it's sort of the experienced practitioners, there's a lot of, again, to use the phrase, the devil in the details, in putting together these things. So for example, what I just described, if you're just gonna look at a value factor in commodities, how do you compare that to a value factor in currencies, how do you compare that to a value factor in equities? So there's a normalization process that different people may do differently. But it's still a value concept. But how you express value, there's some art to that. So I think that's valuable, and certainly something we do. So when you say, like, are you bringing value by using a value factor? Well, if we look cross asset class, it's not something that just the pure equity value factors will necessarily capture. In fact, a lot of times, the same concept in different asset classes are pretty lowly correlated with each other, unless maybe there's a total value meltdown in the world or something like that, like there was in the tech bubble. One thing I also forgot to mention, when you're saying like, how you think about this is, there's also some work being done in academia, and certainly, all practitioners should be worried about this, in terms of deterioration of factors. So once people know about something, if it does indeed look to be a good risk-return trade-off or a premium, there's a likelihood that many people will ultimately pile into it. And if you kind of reflect on well, why do you think this premium existed in the first place, a lot of us like to find some intuition about it. So I'll give you one example. There's a lot of academic and practitioner work that said, Low volatility stocks outperform high volatility stocks, which is counter-theoretical, right? Because you're supposed to be compensated for risk. And then some good people have written papers on, for example, a former colleague of mine, Antti Ilmanen, wrote a paper about demonstrating that people could be leverage-averse and lottery-seeking. So if you want to, if you like the payoffs of lotteries that work well once in a while, but you you want to get higher returns, but you can't leverage, then buying high beta or high volatility stocks is the way to go. So there's enough people that really want those lottery tickets that they drive the price up of high volatility stocks. So if you say, I just want to have a good risk-return trade-off, I could buy low vol stocks, and if I want more vol, I can leverage it up, because I'm not leverage-averse. Totally makes sense, right? So you have the empirics, you have some conceptual backing of it. Now what happens if multiples of billions of dollars then go out on the low vol side? And I'm not suggesting that that's what's happening right now, but you can see some world where enough money taking the other side of that trade, at the very least, brings down the magnitude of the premium, and could make the premium ultimately go negative. So there's some academic work, there's two guys named McLean and Pontiff who wrote, they studied once a factor gets--it used to be gets published, but now, a lot of these academic papers are on SSRN. And once It's been released, what's the performance? And what they find is, there's generally deterioration, but it doesn't go away. And then sometimes there's deterioration, then it comes back, because I think people then get frustrated and get out.
Dean Curnutt: The low vol, or min vol, that phenomenon I've found very interesting. And at least empirically, the demonstration of, Take the low vol portfolio and leverage it up a little bit, and you're gonna do better on a risk-return basis, I find just somewhat compelling. And yet, the ETF-ization of everything. So now they have something, I think the ticker IS USMV, that's your low vol ETF, and the thing got crowded, they got a lot of assets, and I don't want to say it blew up, but it was much more volatile than it should have been. And it felt like a crowded unwind. And what we saw going into it, a lot of the stocks in that were consumer staples stocks. And so the valuation discrepancy, let's say between consumer staples and the S&P got really distorted. In other words, people were paying too much for a low vol or min vol.
Ray Iwanowski: So your--this anecdote that you just said, and I can't vouch--I don't look at this particular ETF, but it touches on a lot of important things when you're thinking about factor investing, and in this factor in particular. So when I keep saying the devils in the details, or sometimes I'll say like, when you talk about factor construction, no two snowflakes are exactly alike. But this one is even more stark, because what a lot of good quants do, including ourselves, is when we construct a factor, let's say a low vol type factor, you could just say I want to buy the lowest volatility stocks and sell the highest ones, or you might say I also want to control for any other kind of factor effect, or any other kind of risk effect. So people who do that, which is what we do, there's a technical term called orthogonalize, you say, if low vol happens to be bad value, you don't want those stocks. You want low vol that's neutral value. Or low vol that's neutral momentum. So you clean out the effects of other factors. Because otherwise, if you just buy the low, you might have a whole bunch of other things in there. And one that you touched on, is sector risk. So sometimes utilities will be on one side and tech stocks will be on the other. So again, thoughtful quants sometimes say, I want to neutralize this, so I'm not taking any outright sector bets, because that's what... And then, I think the other thing that you touched on, you touched on two things. One is, sometimes, and low vol is one, certain economic environments or market environments may prove to be good or bad for the factor. So the first thing we would do is we'd say, after orthogonalizing it appropriately so that you don't have these sector bets, do you find that your low vol environments tend to work well and low vol stocks tend to work well in high vol or low vol, good markets or bad, certain macroeconomic conditions. And some people will then say, I maybe want to tie my exposure to these factors based on those. I'm not saying that it's easy to do, or it's necessarily fruitful, but that's another thing to think about, that sometimes people document that risk-on risk-off environments affect these these factors differently. And then you mentioned crowding. Again, I think there's a lot of complexity to the impact of crowding. So I won't, even for things I know more, I wouldn't vouch for saying that the min vol was actually crowded. People tend to say, Oh, it moves a lot, it must be crowded. It doesn't necessarily mean that. And as I mentioned before, the impact of crowding is only real and meaningful if there's people unwinding that actually adversely affect the prices. So like, even things I'm more familiar with, I generally, I'm very cautious on concluding that crowding is the reason why something doesn't work over a particular period, even if you know it's probably crowded.
Dean Curnutt: Yeah, it's easy to always ex-post explain it.
Ray Iwanowski: That's how people--when quant does poorly these days, rather than saying, Wow, those factors didn't work, it's always, Oh, quants must have unwound their crowded positions. And I think way way more often than not, those are either, the impact is either overblown or not insistent.
Dean Curnutt: Let's shift to vol and vol risk premium. And you mentioned that it is almost a quasi factor. So before we talk about the vol risk premium, I'm curious, when you look at that, is any portion of the factor universe have characteristics that are like the vol risk premium? And again, when I think about vol, I think credit spreads, and I do think about the economic cycle, and so forth. Is any portion of factors like that?
Ray Iwanowski: Yeah. So first of all, for the listener, the way I would characterize the vol risk premium is you get a premium for selling volatility on average over time, in a lot of asset classes. And so we'll come to how I look at that in a second, but yes, I think that's what you just touched on is one of the important things of risk management. So you have a bunch of factors, and some of them, their characteristics might be a lot like the vol risk premium, which I'll describe. And that in and of itself might not be so bad. So you think you're diversified, right? Now, if the state of the world where the bad things happen to you and those factors is different than the vol risk premium, then you still have some diversification. If it's the same time and the same circumstance, then you may, you're diversifying, you're leveraging up to get to a certain risk. And instead of really having diversification, you might be doubling, tripling, quadrupling down on the same phenomenon. And that's obviously bad, and could could lead you to tail risk. So the direct answer to your question is Yes, we think a number of other factors have those characteristics, and also may be even worse, moving at the same time that the vol risk premium does. So what we need to do is size those positions appropriately. So if that circumstance happens, the pain is mitigated, and also construct the strategies to maybe have either hedges or timing models or security selections, that could also mitigate the impact, both in the vol risk premium stuff, as well as the other factors. And that touches on how we think about it. If you think about the general concept that I just described, where you get paid to sell vol, that in and of itself, like, to just a general practitioner or investor, should not say, Oh, I want to do it then, because what comes attached with that generally, is some kind of tail risk that when markets really go bad, you get really hurt. And so another way to say it is, it's selling insurance against the markets, effectively. And if that happens at a circumstance that's a bad time for other things that are going on for you, for your beta portfolio, for your job, then it's really bad to carry. So a rational person might say, at any level of vol risk premium, 1%, 2%, 5%, 10%, I don't want to take that risk, given the tail risk that's embedded in there. And so those of us who actually do, if you're thoughtful, you're not naive and say, Oh, this is a free lunch. You say, I'm either going to size this, or manage the portfolio, hedge it in a particular way, that the downsides are acceptable to me, and the premium, even after doing all the hedges, is still juicy enough. And that's the way we think of it. We have a number of volatility strategies in a number of asset classes. And the basic tenant of them all is, on average, you want to be capturing the vol risk premium. You want to be selling. However, to manage the tail risk, we might put on certain other types of hedges, we might go long things against the shorts that we don't think are going to have a high premium, but maybe their premium's not negative, or slightly negative, such that the mitigates the risks effectively, and we still have some of the premium. We also do try to find timing models that don't always work, but say, going back to my canary in the mines analogy, Uh oh, there's a canary in in the mines that died. It's time to reduce or get out of our vol risk premium. And then like I said, there's there's also ways like, Dean, I know you're involved in options, where you can transform where your tail risk comes from. So if you're going to sell vol, and we won't get that technical here, but you might have some degrees of freedom on both the strikes, and the maturities, and other things, where you say, the tail risk isn't gonna go away, but I might be able to transform it to a different state of the world where it's not when the market crashes. And that's some things we think about, too. We say, Where do we want to take our our tail risk?
Dean Curnutt: The vol risk premium is pretty persistent over time. We know that folks have famously blown up selling vol, LTCM being one of them. And the global financial crisis was one big short vol unwind, mostly in credit. And it was, to me, fascinating that folks were selling vol late late 06, early 07, at some of the skinniest levels you've ever seen. That didn't turn out well. And yet, some of the best profits in vol were right after the crisis. If you if you started initiating vol sales at 45 or 50 in 2009, boy, that was pretty profitable. And then if you fast forward to 2017, which, this is where I really want to get your thoughts, this was one of the best vol selling years in the history of markets, amidst an 11 VIX. And [realize] was six and a half in the S&P. How do you manage the almost paradox or the dichotomy between a trade that's working so well, it's carrying so well, but you're doing it at a thinner and thinner levels, how do you approach that?
Ray Iwanowski: Yeah, you touched on a challenge of complexity of even something that sounds simple, like vol selling. So if I'm telling you, I'm earning the premium selling vol for taking the tail risk, and then all of a sudden the premium gets small, you may say that would be a good time to reduce or get out. And then if the premium gets huge, you might say that's a good time, if you have the dry powder, to get in. Unfortunately, the vol markets, or I should say fortunately, because maybe there's benefits to knowing something I'm about the say is, they don't quite work that way, in the sense that, as you referred to, there's persistent periods that even if your premium gets very skinny, you still get a nice premium. And the tail risk just doesn't hurt you. So you might say, Well, there's also a perception that the tail risk is less. In other words, if you're starting at 10 vol, it doesn't go necessarily to 30 right away. So it's not necessarily, even if it's intuitive, as you said, that selling vol at low levels is a bad trade. Conversely, when you mentioned the global financial crisis, but if you remember, vol spikes up, and then there's points in time where it would spike up more. So you correctly said, there was a point in time where it was so high, and then the world stabilized, and you got a really juicy premium. But if you would have sold it half way up, and a lot of things, it's not only just in vol, like that happened in mortgages and things, you go, Ahh, this is such a juicy spread for mortgages, and then another shoe dropped. So you have to be careful, because if you sell it halfway up, and then it goes crazy more, you can really suffer pretty bad losses. So what we do is, we think more about these longer term dynamics. So again, without giving away how we think of it, there's a point in time, maybe after that first wave, and some other things happen, where it becomes more obvious that it's now a great time to do. And we do tend to get more cautious when the premium gets low, but sometimes that has to do with like our perception of the tail risk, like we always we always focus on our appetite for the tail risk, rather than saying, Well, the premium's only 1% when it's on average, five. So those are important things to think about. But things don't work so cleanly. Like you can't just say, Oh, when it's 10, I'm going to buy, and when it's 30, I'm going to sell. And if, on average, it's 15, talking about the VIX levels, I'm going to sell there, too. You either need to just ride it out, and earn the vol risk premium, irrespective of the levels, which isn't necessarily a bad strategy, and then just manage your your appetite for the tail risk when it happens. Or if you're going to do everything we just talked about, you could call it a factor timing model, if you will, if you're gonna employ a factor timing model, you have to be appropriately cautious about how exactly, what exactly you think you're seeing.
Dean Curnutt: So vol is, from my standpoint, one of the really reflexive assets. It's this thing that, for example, into 2017, it was so profitable to be short vol. The Sharpe ratios on a lot of these systematic strategies were very high. We watched this XIV get bigger and bigger AUM, and we learned ex-post that the man on the street was long this stock that they probably thought was a stock, and it was just printing money. And so the question is, especially with vol, it seems like if you're short it, you want to know who else is short. And you want to know what the reaction function is of folks that could get marked down very quickly. The XIV is this incredibly short trigger, which we could see, you saw the AUM, you could do the math and say, Well, if the VIX goes from 12 to 18, like, that's it. And yet, people wound up kind of getting caught up in it anyway. How do you measure the crowdedness, I suppose, in vol, as well.
Ray Iwanowski: You gave a good example of what seems obvious that, using the word again, endogeneity of pricing for certain players that was sort of exacerbating their demise, if you will. So just like I mentioned it in the quant crisis, where a bunch of people had to head for the exits. And it made, the situation worse, that XIV situation where there was some mechanical things that needed to be done, as the VIX move that one, push the markets even further, and then in some cases, caused certain instruments to need to be shut down or whatever. So there's a lot of focus on that. And if you're an investor, you can do what you suggested, which is to say, I'm very cognizant of all the players in the market, and I'm very worried that these ETFs are getting some traction, and I'm going to get out, that's kind of hard to do. And again, this is a concept of what we would call factor timing. And you might not be that effective at it. And some of those times when you might say, Oh, I need to get out of here now, because there's more people, you're effectively throwing the baby out with the bathwater, and you get, another way to say it, you're getting these false positives. And so you leave premium on the table for a long time, because every single time it looks like something's happening, you get out and you also incur trading costs to do that. So there's risks to doing what you suggested. The other thing you can do, which, I think is is sensible, is to size your positions, and hedge in a particular way, and set yourself up so that you don't have to head for the exits when everybody else does. And in that case, you either take your short term pain at a modest level where it doesn't make your fund perform so poorly that your clients are all going to redeem, that you don't get margin calls. And you just ride it out. So you think about that instance that you mentioned, in February of 2018, I guess. And there was a level, the VIX was going up, and then there was this level, I forget what it was, like 50 or something, where there was some people trading it at the end of the day or overnight, and then even the next day or two, it came back, maybe not to where it started, but it came back a lot. And then eventually it came back to where it started. So if you didn't need to get out, and you didn't have a process to get out, it was interesting to watch the VIX hit 50. And maybe that would have been like, if you were, if you happened to be there, you could sell it. But for those of us who had short positions at that time, that wasn't really that meaningful. Like, you even hear it in other spaces, like, flash crash happens, right? In equity. And, of course, there's a lot of concern in terms of market microstructure, a lot of concerns about the entities that were trading in that. But if you didn't have the need to actually execute at those prices, it was an interesting thing to watch. It's something that scares you about equity market trading. But, it was like a non event. So a lot of what I'm saying here is, you want to think about the time horizon of which you will hold these positions, and the circumstances and well-constructed processes, my opinion, don't have a mechanism where they have to trade into a period that's disrupted. So if the guys who are crowded are unwinding, you sit there, you watch it, you see how it affects your positions negatively. But It's not enough to make you panic, and then you just stay in it. Or even if you have the wherewithal, you go out and do the other side of the trade.
Dean Curnutt: As we finish up. And this has been an awesome conversation, I'm curious, what's on the research docket for you? What are some of the things you're investigating? What are you excited about in terms of your firm's research efforts?
Ray Iwanowski: A couple of things that we're spending time on, were some of the things that we talked about earlier. So, how do you measure alpha deterioration? How can you be convinced you have something, and then it starts going away? Especially because some of the things that we do, we've taken from the academic literature, or we've done modifications of some academic work. The data mining issues that I talked about earlier, like, you have several hundred factors, but how many of those are really real? So we're doing some work that it's a trade off between, if you raise too high of a bar, you may be leaving alpha on the table. On the other hand, if you accept, like a lot of factors blindly, you might be incurring certain types of risks. And also there's trading costs that can add up. So those are some things we're doing. One area, one genre, and probably having a quant conversation where you'd be remiss if you didn't bring up these buzzwords, there's a lot of buzzwords in quant land, and investment land, around things that are buzzwords in general these days so AI, machine learning, natural language processing, big data. There's a lot of interest, and a lot of people out there, putting together quant processes where they're claiming that these are great alpha sources or whatever, black boxes that are beating the markets or whatever. The way we think about it, the way we look at it is, It's always good to consider expanding the set of techniques that you use to find value. So we have some very simple techniques that we use, and then there's some things that maybe we learned that, simple types of methodologies, regression-based, and otherwise that we learned, 25-30 years ago when we were at Chicago. And then there's this view that there's these new techniques in machine learning. Now, the first thing I'll say is, none of these are, very few of these are really new. Some of the techniques in machine learning have been around for 50 years, most of them have been around at least 30 years. But it's worth, in my opinion, it's worth a relook, in the sense that we now do have more data. So a lot of times, 30 years ago, 20 years ago, when I looked at them, and we rejected these techniques, a lot of the problems were, there's problems around overfitting, there's problems around confidence due to lack of data, that now, with so much data, faster computing power, and in fairness, the literature, the machine learning literature, has addressed some of these things in terms of overfitting issues, that it makes sense to think about expanding our toolbox, using natural language processing to go to Edgar filings or whatever, and find information. But I think one of the things that thoughtful quants like, again, there's there's a frenzy, if you will, of people having very strong claims about the value of machine learning. I think It's valuable to say, What do we think these techniques are teaching us versus what we already know? Are there things like, like where we find it valuable building in nonlinear relationships. Everybody knows that most things don't work perfectly linearly, and what the old school kind of conversation wouldn't be, Okay, but ignoring nonlinearities doesn't cost you that much, and thinking of it linearly, gets you a lot. So then the next wave is, if we find some interesting techniques that can actually confidently model nonlinearities, or cross effects, or things like that, you might want to add it to a process. And that's what we do. Now, you also want to say, what's the downside of putting this thing in? It's less transparent, it's less intuitive, can it get me into trouble? A lot of what we've seen, and again, I don't want to denigrate machine learning techniques, as we use them, and it's part of our research agenda, we feel that a lot of times, people are just discovering a new version of momentum. And that's not necessarily the worst thing, like if you say, simple momentum gets you, the types that the factor products will get you 80-90% of the way there, but I've gotten 100% of the way there with a really interesting variation that came out of the machine learning literature, I think that's great. I think the problem is, you dupe yourself into thinking you found all these innovative things, and in fact, they're only momentum in disguise. And there's been some literature that suggests, empirical work, that momentum does tend to crash. So when the momentum crash comes, will you, again, as I mentioned about the quant crisis, are you doubling, tripling, quadrupling on the same on the same phenomenon? And that's what you have to be careful of. So I think we're, we spent the time to learn the sort of the breadth of techniques in machine learning, and some of these other text processing, not--we wouldn't call ourselves experts, but we know it well enough. And we've looked at alternative data. And we're cautiously introducing some of these things at the margin. But if you said to me, How much of the driver of your process are with these techniques, it would be fairly minimal. And we also use some of the techniques in risk management. I don't want to take too much more time on it, but we find some of the techniques that come out of machine learning literature valuable, and trying to find these latent correlations that might not be in the easily accessible data.
Dean Curnutt: Alright, this has been an awesome conversation. Thank you very much.
Ray Iwanowski: Thank you for the time. I appreciate you having me, Dean.
Wrap-up
If you’ve got any thoughts, questions, or feedback, please drop me a line - I would love to chat! You can find me on twitter at @kevg1412 or my email at kevin@12mv2.com.
If you're a fan of business or technology in general, please check out some of my other projects!
Speedwell Research — Comprehensive research on great public companies including Copart, Constellation Software, Floor & Decor, Meta, RH, interesting new frameworks like the Consumer’s Hierarchy of Preferences (Part 1, Part 2, Part 3), and much more.
Cloud Valley — Easy to read, in-depth biographies that explore the defining moments, investments, and life decisions of investing, business, and tech legends like Dan Loeb, Bob Iger, Steve Jurvetson, and Cyan Banister.
DJY Research — Comprehensive research on publicly-traded Asian companies like Alibaba, Tencent, Nintendo, Sea Limited (FREE SAMPLE), Coupang (FREE SAMPLE), and more.
Compilations — “A national treasure — for every country.”
Memos — A selection of some of my favorite investor memos.
Bookshelves — Your favorite investors’/operators’ favorite books.