My guest in this episode is Adam Butler, Chief Investment Officer at ReSolve Asset Management. Adam’s story is the near quintessential example of my belief that every investor’s approach is colored by their experience. From nearly blowing up his firm’s omnibus account at his first job, experiencing the tech wreck first hand, and going all in on the commodity and emerging market super cycle narrative, it took “three frying pans to the face” – his words, not mine – to finally rebuild his mental framework from the bottom up. The evolution of his thinking ultimately lead him to embrace what he believes is the ultimate gift: embracing uncertainty in strategy specifications as a means of exploiting the benefits of diversification.
Transcript
Corey Hoffstein 00:03
Hello and welcome, everyone. I’m Corey Hoffstein. And this is flirting with models, the podcast that pulls back the curtain to discover the human factor behind the quantitative strategy.
Narrator 00:15
Corey Hoffstein Is the co founder and chief investment officer of newfound research due to industry regulations. He will not discuss any of newfound researches funds on this podcast all opinions expressed by podcast participants are solely their own opinion and do not reflect the opinion of newfound research. This podcast is for informational purposes only and should not be relied upon as a basis for investment decisions. Clients of newfound research may maintain positions and securities discussed in this podcast for more information is it think newfound.com.
Corey Hoffstein 00:46
My guest in this episode is Adam Butler, Chief Investment Officer at resolve Asset Management. Adam story is a near quintessential example of my belief that every investors approach is colored by their experience from nearly blowing up his firm’s omnibus account at his first job experiencing the tech wreck firsthand. And going all in on the commodity and emerging market supercycle narrative, it took three frying pans to the face his words, not mine, to finally rebuild his mental framework from the bottom up. The evolution of his thinking ultimately led him to embrace what he believes is the ultimate gift, embracing uncertainty and strategy specification as a means of exploiting the benefits of diversification. I hope you find this conversation just as entertaining and enlightening, as I did. Adam, really excited to have you here with me today. And really looking forward to this chat.
Adam Butler 01:44
It’s gonna be a good bit of fun. The last time you and I chatted at length, I think was over beers in LA. And that went on for hours and hours. And we covered a lot of ground and it was really fun. So looking forward to
Corey Hoffstein 01:57
it. So Adam, if you’ll let me I’d really love to begin, set the tone for our listeners with a juxtaposition that I think is sort of fun. And it begins with a blog post you wrote titled The narrative is reality. And in that post you write, to quote for me in the mid Knotts, Dawn was the guardian of the trend, and I was a full on card carrying member of this cult. Now Don revere refers to Don Cox who was a very prominent strategist in the early 2000s, particularly around the commodity and emerging market supercycle narrative. And you go on to say in the blog post, quote, I tracked the relative cost curve for onshore versus deepwater drilling, as well as the lease process for different classes of exploration and production platforms. I watched the crack spread and the term structure of crude futures. I watched Saudi CDs as a leading indicator of oil price movements. I read Dennis gartman. Now I want to juxtapose that with a draft of an article you sent me just a few weeks ago, it was actually a citation in that draft, where you cite a paper titled cleaning large correlation matrices, tools from random matrix theory, which is 165, page monograph for practitioners all about techniques derived from random matrix theory that they can use to deal with a clean ill conditioned correlation matrices. And I bring up this juxtaposition only because I think it’s so well highlights the evolution of your thinking and your thinking and investments and why I’m so excited to have this conversation with you today. And so with that, I’d really just love to start at the beginning. Can you take us back to when you really first started getting interested in investing?
Adam Butler 03:43
Well, first, a few points of clarification. First of all, you bring that up, it’s like some sort of confessional. For me. It’s hard to admit that I was that guy. But that’s the truth. You know, and I think you’ve got to journey through, you got to make some mistakes before you can see the light. So I actually wanted to be a doctor. And so in school, I started in engineering, I lasted for about a week, I realized the engineers were not my people. And I went and actually I went through philosophy, physics, psychology, and I landed in psychology. I went through sort of two years, I had never heard of the stock market, like my parents were in the medical field and had no interest in markets. They didn’t grow up thinking about markets. I discovered markets in my third year over a dinner with a family friend, and started by investing using stock bulletin boards online. You know, the early internet. I guess I must have had a pretty enthusiastic and compelling pitch because in university I was living with four housemates and three of their parents gave me a couple 1000 bucks to invest using my bulletin board method in small mining stocks. And so I think I had 3000 bucks and I gave him back about 800 bucks. Three months later, but I didn’t learn any lessons from that. I went on and entered in some trading competitions that were hosted by one of those early discount brokers in Canada, and it was a Canada wide competition. And we had three months to turn a half a million bucks in fake money into as much money as you could by trading stocks and options. So I proceeded to trade extremely high beta tech stocks using out of the money call options, and turned half a million bucks into 1,000,002. Hughley out of luck, it has happened to be on the right side of the market with leverage at the time. And then I repeated the success in the next competition, then I turned that into a well, I parlayed that into a career on a large trading floor. So that’s sort of the that’s the beginning.
Corey Hoffstein 05:52
Just to clarify what error was this? It sounds like it was the late 90s.
Adam Butler 05:57
It was it was 9596.
Corey Hoffstein 06:01
So you win this first competition, you win the second competition, and you parlay that into a trading floor job.
Adam Butler 06:09
Correct. So I learned all the wrong lessons, I came on to the trading floor thinking I was God’s gift to trading. Well, there’s an interesting story in my interview process, which we can delve into if we have time. But I started on the trading desk. As a junior analyst. And our our My job was to come up with ideas for advisors, and like trade switch ideas, because really our we were a revenue generating desk. And we were to generate revenue by helping advisors to come up with ideas to switch out of one stock and into another to generate commissions. So I had a very small Omni Bus account to give me a small trading account, just to sort of stay in the flow of the market, right. So I was supposed to trade and try to generate some profit for the desk. But really, that only was account wasn’t really meant as the primary profit center for the desk, it was really just to stay in the flow helped me generate ideas, that sort of thing. And I had a very small risk budget appropriately in the beginning, but it was late 97. And from late 97, to about August of 98. You could throw a dart at the NASDAQ board and just keep hitting massive winners. And so that’s what I did. But I thought that wasn’t darts, I thought it was really skillful selection of high quality companies. So I was really successful, just purely out of luck. And so I turned that small couple 100 grand into 400 grand pretty quickly. And then they gave me a larger risk budget because they thought that my success meant that I was skillful. And I just kept throwing darts and winning. And they gave me a larger risk budget. Eventually, I was working with the full omnibus account, and I turned the 2 million bucks into about eight and a half million. And that was about right into the teeth of long term capital management, the Russian default, the Thai baht crisis in September, October 1998. And that’s things really started crashing down around me, all of a sudden, all the trades, the the way out of the money options, positions that I had on, started going against me, I had leveraged positions on in some individual stocks. They weren’t against me, but I just kept throwing darts. And if anything, it was a mad scramble, I started throwing more darts to try and make up for my losses. So in the span of just a few weeks, that 8 million odd dropped to under 2 million. Perhaps unsurprisingly, I didn’t last much longer on the trading desk. That was definitely a scarring experience for me. And I didn’t actually go back into finance until 2005. I
Corey Hoffstein 08:43
can only imagine that must have been a pretty traumatic experience for you. But if I’m doing my math correctly, 2005 would have been almost seven years later. So what did you do to fill the gap?
Adam Butler 08:53
Well, I was sort of shell shocked for a little while, it took me a little while to discover that I wasn’t gonna get back into finance to figure out what I wanted to do. Instead, I’d always really enjoyed computer programming. So I went back and did a program in computer programming and learned Java and VBA, HTML and some integrated development environments from IBM, which were pretty new at the time, and then ended up landing a position at IBM in their professional services division, they started to be often pretty well the worst job I’ve ever done, which was quality assurance for the programming team where I lasted for all of about six weeks before they moved me into an account management position with our consulting division, which was much much more interesting because I got to interface with all of the different members of the development team from the architect to the programming lead to design and the business analysts as well. And so that was really interesting because I got to experience firsthand what the disconnect is between the business side of an engagement and the technical side of engagement and, you know, really sales and business I don’t speak the same language as programming and development. And to be able to translate is actually a really neat skill. And it’s a skill that I use all the time to this day.
Corey Hoffstein 10:10
So it’s really interesting to me here, with the complete benefit of hindsight knowing where your career trajectory is taking you, you’re somewhat unintentionally building the foundations of a quantitative investing platform, you’ve got the psychology undergrad, which which will eventually be very useful for Behavioral Finance. You’ve built out the programming skill set, the communication side of bridging the technical and the non technical, you have your experience investing. So it’s interesting. While totally unintentional, you have laid the foundation for becoming a quantitative investor. But the tie back to the narrative. So this is the early days of the.com era, if I’ve done my math correctly here. I know there’s a lot of pressure at that time to go start up. You were obviously at big tech, did you ever feel that push to throw your hat in the startup arena?
Adam Butler 11:01
Yeah, I’ve been I’ve been communicating, some friends of mine had had moved into the startup Arena in the internet startups, and had put up a bunch of illusionary money in their pockets right in the form of company shares. So I eventually decided to take the leap, and I joined a small startup Internet startup on their account management team, and was there to see the share price run up from about seven bucks to the mid 30s. And the headcount at the company go from about 30 to 150, in a span of about 18 months. And then literally in three months, we went from 150 to about 40, the share price went from 35 to 10, or 12. And then the next three months, the headcount went to 20. And the share price went to five or six. And eventually I was walking the street, my wife at the time was working at Boston Consulting Group, which had a huge internet consulting practice in Toronto, I think it was their headquarters for Internet consulting. And so they had really built up headcount and had some really extravagant off sites, and it was boom times for them. So mid 2000, early 2001, they did a massive cut. And so my wife and I were directionless set around the same time. And so we decided that we take that opportunity to travel and so we went together to Bangkok, Thailand for a couple of years. Fortunately, we had some friends that were working at a large private school in Bangkok, and they submitted our resumes to their management and put in a good word for us. And we secured a position there and it was this. It was a boy school 6000 Boys From kindergarten to grade 12. So is literally like the energy of a nuclear weapon went off every every day. It was just incredible energy, and I just had a blast. I taught math and physics to English immersion students from grade seven to grade 10. While I was there, just a tremendous experience.
Corey Hoffstein 12:59
So with the complete benefit of hindsight here, we know you’re eventually going to get caught up in this global commodity and emerging market, supercycle narrative. And I find it really interesting that you are living in Thailand, when the beginning of that BRIC narrative really took off. And so I guess my question is, have you been keeping your pulse on the finger of the market this entire time? Were you made aware of the supercycle narrative because you were living in Thailand? Or was it really something else entirely that got you back into the markets.
Adam Butler 13:28
I was absolutely obsessive about tracking the markets. But it wasn’t because I was in Thailand, I think Thailand missed out on a lot of that boom cycle hadn’t really recovered from its currency devaluation. And it didn’t have any real natural resources to speak of. So, you know, it wasn’t one of the BRICs. It wasn’t one of the economies that really benefited from that supercycle. But I was obsessively reading the monthly reports from Don Cox at the time who you mentioned earlier in our chat. So he had come up with this thesis about this commodity and emerging market, supercycle and his thesis rested on this, saying that he had that he repeated all the time. And it went, the best investment opportunities come from areas where those who know it best love, at least because they’ve been disappointed most. And he was referring to the executives of the major mining company, there was no way in hell, they were going to start committing any meaningful capital to new projects in the resource area, because they just been burned so badly from scaling up their production into the teeth of the last resource downswing. So his thesis was that there is this enormous new source of demand which arised from or arose from the migration of rural Chinese and rural Brazilian and rural Indians into this new middle class where they would begin to embrace consumerism in the way that the West had embraced it 100 years earlier. And that this was going to prompt this massive supercycle in demand for iron ore, crude oil. But also, the migration from primarily rice and vegetable based diet to more of a meat based diet was going to mean that there was going to a huge demand for grain feeds, and the fertilizer components to be able to ramp up the production of grain feeds. And so you wanted to invest in phosphates and potassium and, and potassium and phosphate mines that produce those and those are fairly concentrated industries around the world, there’s only a handful of potassium mines, the largest phosphate mines are in Morocco, and in the southern US, and they’re all owned by the same three or four companies. I cottoned on to that early and as an advisor did very, very well for clients and the OH 60708 early oh eight period by having some knowledge of the Canadian oil sands benefited from this whole narrative of crude oil scarcity. And there was some rumor that the major Saudi oil fields were not nearly so well stocked, as the Saudi government had been suggesting. And so there was going to be this major crunch in crude production and their first crude price is going to skyrocket. And we saw some of that play out, right. And I was able to take advantage that as an expert in that narrative, but of course, we all know that that narrative ended cataclysmically in 2008.
Corey Hoffstein 16:36
So what makes that narrative you outline sound so convincing is that it’s not just the first degree investment opportunities. It’s the second and the third and the knock on effects that make it sound so much more convincing, because it is researched in so much greater depth than those first degree connections would suggest.
Adam Butler 16:57
Well, it’s a perfect example of the illusion of knowledge. The more information that you gather about something, the more you feel like you have agency over it. And I mean, that certainly was the case for me in so many endeavors. But definitely, that whole commodity supercycle I was in deep, and I knew everything there was to know I could quote Don Cox, I could quote from the annual reports for BHP Billiton and Rio Tinto and all the major oil sands companies, I knew all the major reserves of all the major oil fields, like natural gas, the substitution ratio for petroleum versus natural gas, all that kind of stuff. And I thought, because I had all this information, that I was better positioned to predict the future. That’s the critical mistake I think that investors make.
Corey Hoffstein 17:42
So I almost feel bad continuing the story, because we know how it’s going to end. We know that eventually this supercycle story is going to fizzle out, but not before you actually almost seem to be proven correct, right? Because post 2007 After financials sold off, the commodity and energy sector continued to do incredibly well into early 2008, which I can only imagine at that point serve to confirm your biases even more.
Adam Butler 18:11
Oh, absolutely. It was huge confirmation of my narrative. And Dawn was early to the whole housing collapse and what the impact was going to be on us financials. And so, you know, I was early to line puts on Fannie and Freddie and some of the European banks. So that helped. But but at the same time, you still had this massive run in the potassium, the potash companies, the fertilizers, the oil, sand companies, were still in a massive run right up into June of 2008. So I was under the impression that this commodity supercycle was going to persist in the face of this credit and housing collapse. And so that ended up being obviously a grave error, and would have been a catastrophic error for me had I not also had some of these other hedges on in the bank puts and just general short financials. My partner Mike always says that crisis is what prompts change, right? Nobody goes to God on prom night. So it was this perpetual prom night from 2006 to 2008. It was prom night from mid 97 and mid 98. It was prom night at IBM and at the startup in late 90s, early 2000s. So you know, those are three frying pans to the face, like you can’t look yourself in the mirror and say, think anything other than wow, I was under a major misunderstanding there. I really wasn’t sure I was going to come back to finance after the 2008 crisis because I’d sort of built my beliefs about myself and the value that I could provide to clients around the idea that I could beat the market through a deeper understanding of the machine. And when I discovered that it didn’t matter how deep my understanding was, that the machine can’t be understood from a reductionist framework. I really didn’t know where to turn. And that was a really, really challenging time for me. And I almost went back to law school and did a complete shift in career and was only through the Well, I mean, just having a great business partner, Mike filbert, just sort of stumbling on some really neat new ways to think about the world, that I was able to figure out how to navigate back to a productive career in finance.
Corey Hoffstein 20:24
Can you expand upon that for me? What were some of these neat new ways of thinking about the world? Well, yeah,
Adam Butler 20:29
I mean, like I said, right, crisis necessitates change. And when I discovered that my previous way of thinking about the world was fatally flawed, then I had no framework. I had no compass. I was an empty vessel, and only when you’re empty, are you ready to receive. So around that time, it was actually while I was on vacation, I was lying by a pool at the South Seas hotel in Miami, South Beach. And I stumbled on this presentation through the Long Now Foundation by a fellow named Philip Tetlock. And for those who don’t know, Dr. Tetlock story, his career, the way I think about any sort of had two phases, and I’m much more familiar with his first phase than his second. But the first phase was when he graduated from university, his specialty was applied psychology, I think I could be slightly off on any of these facts, but I’m directionally in the right direction, to graduates with a degree in Applied Psychology goes to work in the intelligence community in Washington. And he starts out as an analyst, and he’s documenting the forecasts, you know, what all of the major military generals and heads of the intelligence agencies have to say about what’s going on in the Russian pulborough. And they’re all making forecasts about what they think is going to happen. So he’s documenting this, taking notes. And he’s responsible for circulating the notes, and reading them out at these next intelligence briefings. And what he discovered after a few years of doing this was that these senior officials, the senior experts make forecasts about what was going to happen. And the next meeting, he’d read out what their forecasts were, and they didn’t jive at all, with what had actually played out. And so he began to wonder whether or not you know, these are the people at the very top of their field, the intelligence and military, and they can’t figure it out at all. So can anybody figure it out. And so I guess he went back into academia. And he pulled together a resource to do this long term experiment. And he recruited 280 experts from a wide variety of fields. Now, I want to say about 1516 years of experience, on average, average education level master’s degree, these are top journalists and economists and intelligence and military officials, etc. senior people in politics, and he asked them 100 questions each, and they were, they were really well conceived questions. They have very specific answers. And then the people were all we also asked to give their probability that they believe is going to happen. I don’t know. Do you think that inflation, the European will rise above 3%? In the next five years? Yes or no? Okay, great. What probability Do you ascribe to your answer? Oh, 65%. Okay, thank you. So, so 100 questions like that. And then after 15 years, he aggregated all of his all of these answers. So it’s 28,000 sample at 28,000 questions, right? So you draw some pretty meaningful conclusions. And the top four or five conclusions were number one that experts are no better at forecasting than you would expect from random guesses that they are universally overconfident. Though there are certain experts that are better calibrated than other experts in terms of their level of confidence, even though they’re typically they’re not a lot more accurate. None of them not one single person demonstrated an ability to forecast better than random guessing. So there were no outliers whatsoever, some more interesting discoveries. For example, experts were less well calibrated in their forecasts when making forecasts in their own field of experts. The experts that were cited most frequently in the news, who are who appeared more frequently on the radio or on TV, were markedly less well calibrated in their forecasts, then the experts that toil in obscurity, if you translate that to how you view the world, what you take from what experts are saying on TV, or what you read newspaper, obviously, this has a profound effect on your interpretation of events. And so one of the interesting things was that Dr. Tetlock ran some really simple algorithms alongside these experts in order to see whether these algorithms had better forecasting ability. So stuff like just linear regression right will the trend the current dominant trend persist, mean reversion? Will the phenomenon eventually revert to the long term mean? And in the short term The trend based methods worked really, really well. And over the long term, he discovered that most things revert to their long term mean. And these very simple rules based methods massively outperformed the experts and delivered forecasts that were meaningfully better than random guesses. That was one major piece of the puzzle that began to drive me towards systematic thinking. So it was about that time that I started digging into some of the papers on systematic investing. And one of the first ones that I stumbled on which I to this day think is one of the first ones that all new quants stumbled on. Was this paper, a quantitative? What is it by Faber?
Corey Hoffstein 25:44
Oh, Mepps paper, a quantitative approach to tactical asset allocation, quantitative
Adam Butler 25:47
approach to tactical asset allocation? Exactly right. Yep. That really set up a foundation for what turned out to be a completely almost a brand new career, even though I was in the same field. So we started out using, in practice, many of the techniques that Faber described as a quantitative approach to tactical asset allocation and his sector relative strength paper. And we didn’t have any real programming expertise. We didn’t have any real tools at the time, I got the most that I could get out of Excel. But it ended up saying Excel is unbelievably tedious. I know I’m offending like a huge cross section of quantitatively oriented thinkers out there, but I find Excel just to be excruciating to work with. So I outsourced some of the research to another group of quants who had been using more advanced tools and who were able to run some strategy tests for us. So it’s just really simple stuff like taking the MSCI a cross section and liquid MSCI equity indexes and, and running a relative sprang slash timing, quote, unquote, tool momentum type strategy on them. And it worked. Okay, it was over parameterised, we used one specific look back because it happened to work best in sample, but it was I mean, it looked really good. And so we put it to work, and we put to work long short, and actual client portfolios. And for the first eight or nine months, we looked like heroes. In fact, I remember a period in 2011, I think it was August of 2011. The markets had a really steep dip, but it was foreshadowed markets have already begun to sort of roll over. And so our models had taken us short into August trading. And so we were short when the markets dropped precipitously, which meant that we went up in client accounts, I think we made gains of nine or 10%. And then, of course, in September, the Fed intervened, I want to say it was a speech by Bernanke or something like that, anyway, something happened. And markets rallied by 12, or 13%. And our accounts dropped by the same amount. So we gave it all back and more. And this was very difficult for clients to understand or accept. So we abandoned the the long, short version of the strategy. And the you know, that’s probably one of the small milestones that led us to believe that we probably shouldn’t try to be overly specific and how we develop the models. And we also need to be conscious of something that you talk about a lot, right, which is the the best strategy is, is a good strategy that clients can stick with. And this long, short approach was not something that we felt the clients could stick with long term. So you know, we moved from that. It was around that time. But when I was sort of thinking, this is not as far as we can go with this concept. There are better ways I’ve been reading some of the work from guys like David Verratti, I learned that David was in Toronto. So we started having drinks and lunches. And eventually, I reached out to him and said, Hey, would you like to come and join the team? It was bringing David on and the conversations between David and Rodrigo and Mike, where we really began to understand the importance of building an investment strategy from first principles.
Corey Hoffstein 28:59
So I’ve heard you use this phrase first principles a number of times. Can you explain what you mean by first principles?
Adam Butler 29:06
It’s about first having an understanding or really a belief system, about how the markets that you are contemplating investing in are most likely to work. So what is the basic relationship between risk and return? So if you look at the asset class level, so big muscle groups like global government bonds, US versus international aefi versus US versus emerging equities, or maybe you can break it down into regions, the major categories of commodities, gold, some of the major currencies, the actual asset classes typically have a return that is linearly a function of risk. So higher risk assets tend to have a higher return. So the slope of the capital market line is positive, and the assets generally line up belong that line fairly closely. So they’re a pretty good fit. So as a general state into first principles, we believe that in the asset class space that returns are a function of risk and that that relationship is broadly linear. axiomatically means that all major asset classes have approximately the same Sharpe ratio. And so then, once you’ve got this basic belief, then you soon can sort of turn to the portfolio optimization machine and say, what type of optimization allows you to build a mean variance optimal portfolio using only the assumption that all of the asset classes have the same Sharpe ratio? And there are some obvious answers. If you believe that you can forecast volatility, and that the correlations between the assets are broadly similar than an inverse volatility weighted portfolio of those asset classes approximates the maximum Sharpe ratio portfolio. If you believe that you can measure covariances, some measure both volatility and correlation and generate reasonable estimates for those that are better than random. And you believe again that that asset classes have approximately equal Sharpe ratios, then equal risk contribution portfolio is the max Sharpe ex ante Max Sharpe portfolio conditional on the fact that you don’t have any active use on returns. Now, if you do have active use on returns, then you can use a mean variance framework. And so this is really the evolution to adaptive asset allocation. Adaptive asset allocation is a mean variance optimization based on some systematic active use. So in our early thinking, we used momentum, cross sectional asset class momentum to inform our active use. So we wanted to maximize exposure to assets with positive momentum, and minimize exposure to assets with zero or negative momentum. But we also want to minimize overall portfolio volatility, making use of information that we can glean from the correlation matrix from diversification. And so that really is adaptive asset allocation in nutshell, we’re trying to maximize the characteristics of the portfolio that we want to emphasize, like momentum, but you can expand it to enterprise value EBIT da or price to sales, or trend or carry when it maximize those characteristics, while minimizing overall portfolio volatility, which is all you do is part of the mean variance optimization
Corey Hoffstein 32:30
very often, when mean variance, optimization comes up. And really just optimization in general, there’s this debate between the simple and the complex. And there’s a whole lot of literature dedicated to try to find the balance and the out of sample success of very simple and naive methodologies like one over n. And acknowledging that optimization techniques are very often unintentionally, error maximizers that they will take those statistics about which we are most uncertain, and unintentionally overweight them. And I know this is an area, you’ve waxed philosophical about quite a bit, your firm makes heavy usage of optimization techniques. And so I was hoping you could just spend some time exploring this concept. How do you guys find the balance between the simple and the complex? And how do you address the fact that very often, the statistics of which you are trying to use as sources of information within your portfolio construction are often shrouded in this distribution of uncertainty?
Adam Butler 33:33
There’s really two different concepts embedded in that statement, right? One is the question of when and how is optimization likely to deliver better results than naive methods? And to how to best make use of ensemble methods? So probably we should unpack that those different concepts separately, right. And we could start with optimization. But we started to get there with this discussion of the optimization machine. But I mean, really the question of whether naive or optimal diversification through numerical optimization, which one of those are along the continuum between those? What’s most effective is a function of what we believe to be true about our investment universe. So for example, if you look at one of the most popular papers that weighs in on this question, a paper called optimal versus naive diversification by Dima Gallagher, Lappi and Hoople in 2009. They examine the performance of portfolios form using naive methods like equal weight, one of our n, relative to some very complex optimizations, using based on shrinkage and all kinds of different complicated applications, but they apply it to a really equity centric universe. So for example, one of the universes that they run this naive versus optimization based process on his 10 industry groups from the ken French library, which anyone can download, I would encourage you to download it and look into this yourself. So I just finished actually running some tests on that universe. And what did Miguel and his crew found was that optimization is, is not as useful as just one of our end methods and allocating to this to this industry group. And candidly, I started my investigation kind of skeptical of that claim, because they use some really strange parameterizations. They’re using five and 10 year monthly look backs. So I mean, the information decay on their volatility and correlation estimates based on that length of look back, obviously raised questions about whether there’s any information content at all. So I thought that we could use some, you know, use daily data which is provided for free, use shorter look back horizons, and come up with better results. So I ran it and the fact is, I couldn’t the results from the optimization methods did not work as well. We ran minimum variance Max versification, in verse fall in verse, very variance, ERC and a couple of heuristic methods like the hierarchical minimum variance, and we couldn’t make heads or tails of it, none of them many difference, the equal way completely dominated.
Corey Hoffstein 36:27
So this is a topic near and dear to my heart, we run a number of tactical sector strategies here at New Found and our research confirms much of the same, which is that the added informational benefit you get from statistics like volatility and correlation is often offset by the uncertainty with which you are measuring those statistics. And so a naive one over N approach proves to be incredibly robust out of sample. And if we take that a step further, and consider introducing trading signals, whether its value, momentum, or trend, that one over n is important, not only because we’re trying to balance sector risk, but also we’re trying to balance model risk. So by way of example, let’s consider a market cap weighted sector portfolio that we were going to run a trend following strategy on in that sort of portfolio, your trend following signal on the utility sector would have very little influence on the overall portfolio success or failure. Whereas the accuracy of your Trend call on something like the technology sector, or the financial sector, would have an overwhelming impact compared to the rest of the sectors.
Adam Butler 37:41
So let’s unpack these results. Right. So you’ve got this 10 assets, 10 industry groups, all US portfolios of US stocks, we tried a bunch of optimizations, none of them helped most of them delivered worse results than one over the hell’s going on. The culprit, it turns out is that there is no information in the correlation. So if you decompose the sources of risk through principal component analysis of 10 Industry portfolios, you find that about 90% of the risk is in the first principal component. And that, in fact, the first principal component is the only statistically significant or economically meaningful source of risk. So in fact, if you regress the returns for each of the industries on the s&p 500, and then you extract the residuals, those residuals contain no new information about the structure of those assets. Which is kind of interesting, because what it says from an economic interpretation is that industries, at least according to how they’re classified by Fama, French, are basically just random portfolios of stocks, they don’t introduce any more information about the structure of the market than you would get from just constructing any random 10 diversified portfolios of stocks, right. So that was your instinct. And actually, there’s a way you can derive the number of independent risks in the portfolio as the square of the diversification ratio. So you find the maximum diversified portfolio, and you take the square of the diversification ratio of that portfolio. If you do that, for the 10. Industry universe, you discover that there’s less than one and a half independent bets across those 10 assets. So then we thought, okay, well, that’s interesting. I wonder if we can get better results from another of the universe’s the 25 portfolios sorted on size and book to market also from and in fact, we do get slightly better results, but even across 25 portfolios sorted on both sides and book to market. There’s less than two independent sources of return. What is the purpose of of optimization it is to maximize the opportunity for diversification? Well, the fact that there’s one and a half or two bets across these 10 or 20 Five different assets means that there’s no opportunity for diversification. Now, if you run it on, for example, the asset universe that we use for a global risk parity portfolio, now you’ve got five independent bets across 12 assets. If you run it on our 48, futures universe, you’ve got 13, independent bets. So the broad lesson is that mean variance optimization is extremely useful, where there are several different independent sources of risk. If risk is all dominated by one risk factor, then there is no opportunity for diversification. And therefore, optimization is truly just operating on random noise. And so it’s, it’s obviously not going to be useful.
Corey Hoffstein 40:42
So I’m going to spin you up a bit here because you have almost a fanatical obsession with asset allocation. Whereas most quants, and most of the literature tends to be published on the security selection side, there’s the whole factor Xu 500, plus different and unique security selection factors. And you’ve taken the opposite view and written at length about that that focus is wrong, that the real opportunity for investors to generate alpha is actually at the asset allocation level. And it’s somewhat ties back to this discussion we’re having around the independent number of bets, and that empirically, the opportunity to benefit from diversification is much larger at the asset allocation level than it is at the security selection level. But I know that you’ve also proposed before and interesting limits to arbitrage argument. And I’d love for you to go down that path a bit and explain why you believe that for investors who are willing to have a more flexible allocation policy, there’s actually a much more significant Alpha opportunity.
Adam Butler 41:47
Thanks for that introduction. And, I mean, this is definitely one of the foundational principles that we espouse at resolve. But it’s just the idea that 99.99% of all computational cognitive energy in markets is devoted to security selection. And this is true, because that is the way the industry is structured. So if you look at the organizational structure of any large pension plan or endowment, or of the portfolio’s of most individual private wealth clients, what you see is that the portfolio has a strategic asset allocation that is set by an investment committee probably in the case of pensions verified by actuaries. In the case of private clients, it’s verified by the compliance departments at the brokerage houses, so that it is suitable to what the client has expressed their objectives are and their risk tolerances are. And once that strategic asset allocation is set, it doesn’t vary very much. If it does vary. It varies incrementally through time based on changes in age or changes in financial situation of the client, or based on a change in leadership of the institution. But it certainly doesn’t change very much based on tactical decisions of the Investment Committee. By the same token, if you go one level deeper into each of the different sleeves, so you’ve got typically an equity sleeve or a growth assets leave, you’ve got a fixed income sleeve, and you’ve got an alternative sleeve, sometimes that alternative sleeve is arbitrarily divided between stuff like infrastructure and private equity and venture cap and so called hedge funds, whatever that means. But within the equity space, now, there’s an enormous latitude for tracking you’re the purpose of the people that are hired within the equity sleeve at the institutions, or that are hired as mutual fund managers, or SMA managers for private wealth clients, is to take active risk in pursuit of active returns. So there’s an enormous tolerance for tracking, you’re at the individual security selection level relative to the tracking year that is tolerated at the asset allocation level. And what that means is that the vast majority of capital that is seeking to arbitrage the opportunities in markets is seeking to arbitrage within the security space. And there is extreme constraints on those investors that are able to arbitrage the mispricings at the asset allocation level. And so the opportunity at the asset allocation level, to generate alpha by harvesting Miss pricings, or taking advantage of Miss pricings is much larger, and in our opinion, also much more sustainable, because it’s just very, very few players out there that are actively looking to arbitrage those opportunities.
Corey Hoffstein 44:30
So we’ve been circling the drain a bit on this adaptive asset allocation philosophy, but I know that for you, the actual implementation is a critical component of the philosophy itself. You have a chapter titled in your book adaptive asset allocation. All we know is that we know nothing. And I know this appreciation for uncertainty, and embracing randomness is a big piece of the implementation puzzle. So I want to make sure we don’t close out Got any discussion on this topic without taking the time to actually talk about how we take some of these lessons and move from the philosophical to the practical? Yeah,
Adam Butler 45:13
I mean, I think that this is a critical piece of the puzzle that many systematic investors neglect. It’s this first step, which is asking the question, is there anything magical about the way that I’ve specified the problem? And just going back to the 200, day moving average with the 10 month moving average for MEB? Right? Is there anything magic about the 10 month moving average, many papers on trend are written using a 12 month look back horizon. And that’s a part of the scientific process, right, the ivory tower requires that you build on the papers that have been written before yours. And so if the first paper that comes out on momentum, like Jagadish, and Tippmann, comes out using 12 month momentum with a skip month, well, I’m gonna say 95% of all papers written on momentum since 1992, were specified in exactly the same way. Because the idea is to build on hold that specification constant, that definition of momentum constant, and explore things that touch on that specification, right. But you know, Jagadish, and Tippmann, they randomly decided that they’re going to use 12 Minus one as their specification. So use lots, but then Carhart use 12, minus one, and that took off. But the point being that there’s nothing magical about 12 Minus one, no remarks prove that Stambaugh proved that a specified other models that work just as well, the guys that PIM then for yet, and the guys are robeco use idiosyncratic momentum. Like there’s lots of different ways that are just as useful tool, but it’s nothing magical. But total minus one, there’s nothing magical about 12 month look back for trend following or a 200 day moving average. So we sort of start by thinking about why is that 12 month, look back horizon meaningful? Is it? And if it isn’t, why are we married to it? Why are we married to 10 months or 12 months, especially if you examine the literature, and you’ll see that a one month look back horizon for trend over the past 200 years has been just as effective as a 12 month look back for trend three months has been just as effective. Sure, they’ve got Sharpe ratios that differ by 10 or 15 basis points, but the standard error or the Sharpe ratio is on the order of 30 or 35 basis points. So in other words, they are statistically indistinguishable from one another. So why are we going to zero in on a 12 month look back? Well, we know that there are five or 10, or 15, or 20 year periods, when a three month or a one month or an eight month, look back is going to completely dominate a 12 month look back,
Corey Hoffstein 47:36
I always laugh a little because there is this expectation or this thought that there is some magic parameter out there, as far as it relates to some of these technical trading strategies, like trend following. And yet, if you were to turn the conversation to something like value investing, there isn’t the same expectation that people understand and appreciate that there is no single valuation metric that can capture the full picture of a company’s valuation. And it really is true in the very same way for something like trend following that there’s no magic in the 10 month moving average or the 200 day moving average, it’s just a model. And it may capture some aspects of trend. But there are going to be times when it does not work the same way that certain value measures wouldn’t would not work.
Adam Butler 48:23
Yeah, we actually started when we tried to introduce this idea of ensemble, we often have used the value concepts to try and anchor them because people are familiar with value. And they’re not many people are not so familiar with trend or momentum type specification. I think most people that read maps paper think that they have if it’s sort of the first paper first few that they’ve read, in that field believe that they’ve stumbled on this holy grail, right? There is something truly magical about this 10 month moving average, right markets just are predestined to behave in a way that makes them respond to a cross of the 10 month moving average for 10 months is about 200 trading days. So the 200 day moving average. So everyone I think starts in this for most people starting this really deterministic place where it’s the parameters that matter, right? It’s this 10 month moving average is magic. It’s not that this general concept may have some merit, it’s that this particular signal is the holy grail. And once you sort of back up and realize that there’s no deterministic reason why market should trend at a 12 month horizon, but then not trend at a six month horizon. Then you also begin to realize that there’s this gift that these guys aren’t taking advantage of. and a gift is the fact that the six month strategy and the 12 month strategy and you know, let’s get one level granular, more granular the 26 day strategy and 192 day strategy have the same expected Sharpe ratio, but they are not perfectly correlated with one another In fact, our head quant Andrew Butler has just derived the like analytically what the expected correlation is between trends of different horizons and for holding periods of different horizons to. And so in reality, you can get about two and a half independent bets just by using different look backs or a variety of different look backs, combined with a variety of different holding periods in order to form your portfolios. So you get this massive diversification free lunch. And even better. Now, I don’t have to make any decisions based on what’s worked best in the past. In order to specify my model, I can use them all I can say, I’m going to take some random number of look backs to measure my trend or my momentum of somewhere between whatever two and seven. This sample picks for now I’m going to pick four random look backs. Okay, we use a space filling algorithm, but we still it’s randomized, random look backs between, basically the range that momentum and trend has been identified 20 days to about 300 days, randomly select four numbers. Maybe it’s 29 and 61 and 270 days, right. And so now I’ve got four different momentum look backs. Well, I’m going to have to put those together somehow I’m just Am I just going to average their raw returns? Am I going to average their Sharpe ratios? Or their Omega ratios? Am I going to examine the time about trend? So that’s the quantity, how am I going to transform it? Am I just going to use the raw numbers of plugging into the optimizer? Or am I going to say that assets that are above the median at each lookback horizon? Get a one and below the median, get a negative one and then just add up those ones and negative ones across the four look backs and feed that score into the optimizer? Right? God? Oh, no, they all seem like they’re perfectly useful. It’s just examining the market using different assumptions about the granularity of information, right? Rank might be perfectly legitimate, is rank better than raw better than binary? Well, they all seem to work just as well in simulation, and they’re all theoretically just as reasonable. So we use all of them. And then what’s the best optimization to use? do we maximize the Sharpe ratio? Do we find the mean variance optimal portfolio at our target volatility, do we just make the assumption that all the assets that we like have equal expected return and therefore the minimum variance portfolio of those assets is the mean variance optimal portfolio, there’s actually five or six different ways to optimize it and get the same theoretical mean variance optimal portfolio turns out in simulation, they all seem to work just as well. So when you put all the combinations of different look backs, different numbers of look backs, different transforms different optimization methods together, you get a large number of different sub strategies, they all aren’t perfectly correlated to one another. And so you, again, you get this major diversification benefit, you get more diverse portfolios, you get portfolios that work in more diverse environments, you’re less reliant on a particular type of or length of trend, or, you know, you just less reliant on this any sort of specification, you get this more robust underdetermined or non deterministic portfolio that works better in practice, than 80% of the sub strategies. So it’s this amazing phenomenon. If you look back at the back test, across all of the different combinations of our strategy, you got some sub strategies that back tested a point nine one sharp, we’ve got other sub strategies the back test at 1.67 Sharpe, the standard error is point three, seven. So I cannot say with any statistical meaning that the 1.67 Sharpe strategy is materially better than our point nine one Sharpe strategy, I have no confidence that the point nine one Sharpe strategy won’t outperform the 1.67 sharp strategy in live trading. So I should just use them all. And even better, if I’ve got 100 sub strategies. And the 50th percentile is 1.2 sharp. When you put them all together, I might get 1.45 or 1.5 sharp, which is above the 80th percentile of all of the individual sub strategies without me having to choose any parameter specifications at all.
Corey Hoffstein 54:19
So this conversation is reminded me of a quote from Aaron Brown, who last time I checked was the head of risk at AQR and the quote goes something to the effect of randomness is something we create, to learn about something deterministic. And what’s really interesting here to me what it sounds like is you are using randomness in a way and really trying to exploit diversification that with the expectation that for every statistic you’re trying to estimate there is this shroud and this distribution of uncertainty around it, that by leveraging distribution and taking all these random samples, you can try to add average all that noise out and come up with a much more stable estimate, whether it’s have some sort of timing, signal momentum value, even things like volatility and correlation.
Adam Butler 55:12
Yeah, it really is. It’s just an advanced signal extraction technology. It’s using ensemble methods to extract a higher level of signal from the noise.
Corey Hoffstein 55:22
So the mental model that I use for diversification is really like a 3d graph. When most people talk about diversification very often, they’re talking about asset class diversification. Or if you’re a quant, you might talk about factor diversification. Or if you are a macro investor, you might talk about some of the macro economic influences and diversifying across those exposures. But that’s really just one axis of diversification. When I think about it, that there’s these other axes along which you can diversify, there is the how, which would be your process. And those might be the signals you use to manage your investments. And then there’s also the when, which is when are you investing? What are the opportunities at that point of investment? How frequently are you rebalancing? How long are you holding for? And if you believe that there is diversification opportunity, among those different approaches, when you invest at different times what asset classes you use, that by expanding your diversification from really one access to three, you can in theory, dramatically improve your Sharpe ratio?
Adam Butler 56:31
Well, this is the thing, right? What it does is it increases the ex ante Sharpe ratio of the portfolio, because you’re not relying on any particular specification, you’ve just got a lot more faith in the live trading of a portfolio, if you’ve made very few decisions about how that portfolio should be formed, right? It’s funny because a lot of people sort of preach kiss, right, keep it simple, stupid. And using one moving average, like a 10 month moving average, or something or one look back period for trend. It sounds like it’s simpler. But what it is, is it’s more fragile, because it’s forced you to make a really critical decision. And that decision is going to have profound impacts over the performance of that portfolio, especially in the short term. Like if you’ve got a strategy formed on a 12 month moving average, or 12 month look back horizon and a strategy formed on a three month look back horizon are going to provide the same terminal wealth, they have the same expected ex ante Sharpe ratio. But over the next 10 years, a 12 month look back horizon could be just God awful, while a three month look back horizon could be magical. And those who chose 12 month because it has worked over the very long term, have failed to realize the long term may not apply in the short term. And very, very few investors have an infinite time horizon to allow that time infinite average to manifest.
Corey Hoffstein 57:58
So a fun question, I think, to ask of people who work in the asset allocation space is if you are building a portfolio of individual securities, let’s say I asked you to build a portfolio of large cap US equities. How would you think about approaching that problem?
Adam Butler 58:14
I do believe that there are inefficiencies in the security space, my belief in any particular type of inefficiencies are very loosely held. So I believe that there is a value premium, I have no preference or confidence that that value premium is better expressed using sorts on book value, or free cash flow, or earnings or sales or enterprise value, sorry, EBIT da, I think all of them work just as well. I do believe in momentum. But I don’t have any strongly held beliefs about how to specify it, I would probably use both six month and 12 month momentum, I would use residual momentum as well. Those are really the factors that I probably hold most dear. I’m highly enamored with some of the research that has come from Losang with his Q factor model. And so I would probably seek to emphasize firms with high ROI E and low investment. So I think the way that I would approach it is I would take my list of factors, it’s probably four, maybe five, and then I would pick a diverse set of specifications for each of those factors. And I would perform sorts. So I would take the top 5% of stocks sorted on each of the value factors, top 5% of stocks sorted on each of the minimum specifications, top 5% on are we top 5% on investment, the different ways to measure investment. There’s different ways to measure are we so I probably got maybe 15 or 20, sort of groups of 10 or 20 stocks that I like and then I would form a robust minimum variance portfolio of those stocks that have characteristics that I like. And that just expresses the belief that I believe that all of those stocks that I selected will outperform the broader market. But I have an equal expected return for all of those stocks and therefore, the mean variance optimal expression of the fact that I think all of those stocks have the same expected return results in a minimum variance portfolio as a first stab at it. I probably approach problem that way.
Corey Hoffstein 1:00:31
All right, Adam, last question for you. If you were an investment strategy, it can be any investment strategy could be passive. It could be value investing, global, macro, merger, arbitrage, whatever you want. If you were an investment strategy, what would you be and why?
1:00:47
I’m pretty sure I would be a deep value contrarian strategy. I’m kind of the ultimate curmudgeon, I’m just generally very cynical and skeptical and tend to run against the grain. And so I think, if I was to anthropomorphize myself as a as an investment strategy, then I’d have to go with contrarian value
1:01:09
out of it’s been a lot of fun chatting with you. I really appreciate you taking the time. Thank you.
1:01:14
Thanks, Corey it has been a lot of fun.
1:01:16
Thank you for listening to my conversation with Adam Butler. I hope you enjoyed. You can find more of Adam on his blog, www dot Gestalt yahoo.com and on Twitter under the handle Gestalt you you can find show notes for this episode and more at www dot flirting with models.com/podcast. Finally, if you enjoyed this podcast, I’d urge you to share it with others, whether by email or social media, and leave us a review on iTunes.