117 Strategic Risk Management ft. Cam Harvey & Rob Carver

[00:00:07] Cam

Empirically, we’ve had these academic papers that show that volatility targeting does just as you said, it’s really, really effective. It improves Sharpe ratios. However, that research was applied to equity markets and people read these academic papers. We know, for example, that trend following firms use volatility targeting routinely in terms of their signals. So this is really important. It works for equity, but it doesn’t work for everything. You need to step back and figure out why.

[00:00:52] Intro

Imagine spending an hour with the world’s greatest traders. Imagine learning from their experiences, their successes and their failures. Imagine no more. Welcome to Top Traders Unplugged, the place where you can learn from the best hedge fund managers in the world so you can take your manager due diligence or investment career to the next level.

Before we begin today’s conversation, remember to keep two things in mind. All the discussion we will have about investment performance is about the past, and past performance does not guarantee or even infer anything about future performance.
Also, understand that there’s a significant risk of financial loss with all investment strategies and you need to request and understand the specific risks from the investment manager about their product before you make investment decisions.

Here’s your host, veteran hedge fund manager, Niels Kaastrup-Larsen.

[00:01:49] Niels

Cam, it’s great to have you back on the podcast. I can’t believe it’s been more than a year since we last spoke. How are things where you are today?

[00:01:56] Cam

Everything is going well. I’ve got a bunch of new research going and my teaching begins in January.

[00:02:06] Niels

Exciting. Now, Rob and I, we are definitely very excited about our conversation today since we’re going to talk about a really important topic, namely, how to design and manage the risk in a portfolio, which of course, relates to your latest book that you published with Sandy Rattray and Otto van Hemert, two of your colleagues at Man Group. But before we dive into all of this, perhaps I can start with you, Rob, because I’m interested and curious to know what your memories of meeting Cam were for the first time and also maybe whether you knew his work before you actually met him in person.

[00:02:42] Rob

Yeah. This goes back to when I first joined AHL in 2006. My first job was to put together a systematic global macro strategy. So I started looking at all the research. I guess back then, 99.9% of academic research was equity strategies, long-short equity strategies, and there wasn’t really anything on what I was interested in, which was cross-country stuff. I think I actually put ‘global tactical asset allocation’ into Google Scholar and Cam had actually written a paper on that exact title. I started reading his other stuff and he had done lots of things on cross-country equity and bond returns, which was exactly what I was looking for.

So, I was raving about what a great guy he was, and how interesting and how, unlike a lot of academics, his work was really relevant to what we were doing. I guess that raving had some effect because a few years later, we actually hired Cam as an academic advisor and I was able to meet him when he came to London, which for me, was like meeting one of the Beatles. So, Cam probably has a slightly different impression of meeting me.

[00:03:52] Cam

A way different impression.

[00:03:55] Niels

Yeah. I was going to ask you that actually, Cam. Do you remember your first meeting with Rob?

[00:04:00] Cam

Absolutely. It was really interesting because I’d been working with a very, very small quant group at GLG, and then Man bought GLG. All of a sudden, I basically walk into a place with just so many quants. So I was working with like three or four people, and then all of a sudden, I’m looking at a range of PhDs that’s larger than my colleagues at Duke University. What was really interesting was the diversity of backgrounds. So this was just great for me. You had people trained in mathematics and statistics and physics and chemistry, all of this stuff, quantitative, and very few people with a financial economics background.

So it was great for me, and I remember my first meeting like it was yesterday. I was just so excited to be able to work with a quant group of this caliber. It still amazes me the sort of human capital that they’re able to gather.

[00:05:27] Niels

Yeah, absolutely. It’s fascinating. Actually, it’s quite interesting, in general, that all of these sciences have made their way to finance. That’s not how we thought about it 20 years ago I guess, but yeah.

[00:05:39] Cam

It’s kind of interesting, because there’s always been a good crossover of mathematics and physics in finance. The key 1952 Markowitz portfolio optimization paper, that was just like an OR paper; no economic content really.

[00:06:04] Rob

Operations Research for the people who aren’t familiar with the acronym.

[00:06:07] Cam

Yeah, exactly. You think of all the work in derivatives pricing, Merton and Black and Scholes, that’s essentially applied physics. It’s like the heat equation. So there’s always been this really strong ability of people in financial economics to pick up on stuff that is outside of economics and bring it in. Indeed, today, the advances that we’re seeing in machine learning are just another example of that. So it’s really important. You can’t just be siloed in this business where all you do is economics or finance. You need to be well-versed on the latest advances in other scientific fields.

[00:07:07] Niels

Absolutely. Anything you want, Rob, to add before I dive in?

[00:07:11] Rob

No. I think we’ve mentioned Markowitz so I think it’s a good thing to start diving into this book, which I guess you can trace all the way back to Markowitz in terms of its intellectual and theoretical underpinning. So yeah, over to you, Niels.

[00:07:27] Niels

I’m curious, Cam, there’s a lot of books, I guess, that have been written on the topic in terms of portfolio construction and risk management. So I’m interested in what motivated you to write this book with your colleagues and also, where did you want it to be different maybe from some of the books that are already out there?

[00:07:48] Cam

There’s no shortage of books on portfolio formation, no shortage of books on risk management. So why did we do this book? Actually, I do like to take this back to Markowitz, and indeed, my own research 21 years ago. Markowitz is a great idea. As I said, it’s foundational for what we do in finance, but it’s very limited in terms of its abilities. It’s limited in two major aspects.

To be clear, this paper in 1952 recognizes these limitations. So Harry Markowitz, who I’m very proud to be a co-author with Harry on one paper last year, which was a great honor for me to write a paper with him, somebody I’ve looked up to and interacted with for many years. But he points out two key limitations. The first one is that the inputs have to be known.

The inputs to a portfolio optimization include the expected return; so a portfolio optimization setup to actually form your portfolio for a good position for the future. You don’t care about what has happened in the past, it’s about the future. So you need, when you’re doing this optimization, to know what the expected returns are, to know what the volatilities are, to know what the correlations are, and then you get the best possible portfolio in terms of highest target return for a given level of volatility.

So, one thing that is often not realized in the practice is that there’s no uncertainty here. You have to know exactly what the expected return is. Indeed, if you change that expected return, even to a small degree, it could lead to a completely different weighting of assets. So this is definitely a limitation of this technique in that there’s no uncertainty. My research in Quantitative Finance (that’s an academic journal) basically tried to deal with that uncertainty by implementing a Bayesian sort of approach for portfolio construction.

Think of the Bayesian construction as not having a point estimate for the expected return, but you’ve got a distribution. So actually, you’ve got an expected return but there’s a 10% probability it’d be less than x and a 10% probability it’d be greater than y. You put that into the actual problem and allow for that uncertainty, and that’s what we did.

It’s the second limitation that’s more important to me, and I think, was foundational for the research projects that I did with Man Group on this topic. The second limitation is explicitly in the paper, I think it’s on page 91, where Markowitz basically is showing that the portfolio that maximizes the expected return for a level of volatility is optimal, and he gives conditions whereby it’s optimal.

Essentially, the conditions are that the investor really only cares about the expected return and the variance or the volatility. That’s it. He actually writes down that this would not be optimal if the investor has got a preference for a higher moment like skewness. So, basically, he writes on the equation that says that there is no preference for skewness. Think of skewness (positive skew) is where you’ve got very large positive returns, which people like, so right tail positive surprise. A negative skew is where you’ve got these very large drawdowns that are much different than what you would expect for, let’s say, a normal distribution.

It turns out that, number one, the data are actually such that they don’t adhere to a simple normal distribution, and more importantly, number two, investors have a clear preference for positive skew and they very much dislike a negative skew. So this assumption in Markowitz is really limiting. Indeed, you open textbooks today, and you see the same diagram that you see in the 1952 Journal of Finance paper, where you’ve got, on one axis, the y-axis, the expected return, and on the x-axis, the volatility, you’ve got a curve, the optimal portfolio, all this stuff. Again, this is a world where there’s no preference for skew.

What I did in my paper in 2000, in the Journal of Finance, is that I extended the Markowitz model in a very simple way, and that was that you’ve got preferences for the expected return, the volatility, which you don’t like, and the skew which you do like. So basically, you inject this third dimension into this. Think of it this way. You might have a number of portfolios that have the same expected return and the same volatility. Which one do you choose? Well, you choose the one that’s got the highest possible skew. The one you dislike the most is the one with the negative skew.

So basically, you’ve got this three dimensional optimization, you take this into account, and when you do this, all of a sudden, you’re managing not just the expected returns and the volatility, but also the drawdowns. So think of this book as sketching a strategy. What we actually detail at the beginning is that risk management is often done as an afterthought. The risk team often is not even located in the same area or maybe on a different floor than the portfolio team. The risk team often kicks in when there’s a problem.

What we advocate in the book is a holistic approach where at the very beginning in setting your portfolio, you take the skew (or as we call it, the positive or negative convexity) into account in forming your portfolio. That, I think, is the contribution, that it’s one of the few books that takes this view that in setting your portfolio, you need to take asset convexity and portfolio convexity into account. If you do that, it’s got very good properties in terms of performance in markets that are extremely uncertain and volatile.

[00:15:59] Rob

Obviously, to us people in this trend following world, this is all very familiar stuff. We’re always trying to produce strategies with positive skew. I think one of the problems I’m always having is that you often have a tradeoff between Sharpe ratio and positive skew. A lot of the debates we have on the podcast actually are between people like Jerry, who lean more towards going for the variations of strategies that have very positive skew. I’m like, “Well, I’m prepared to give a little bit that positive skew if I can get enough Sharpe ratio.”

So yeah, it’s a very interesting debate. Of course, a lot of investors aren’t even aware of this stuff even though you’re right, it’s kind of debating their preferences, isn’t it?

[00:16:46] Cam

Yeah. So, it’s just really interesting, Rob. This issue of Sharpe ratio is basically, that the Sharpe ratio is the Markowitz, 1952. That’s all it is. It’s the expected return divided by the volatility, or expected excess return divided by the volatility. That’s all it is. So it’s kind of obvious, when you look at these different strategies. You can see this by just looking at hedge fund strategies, that there’s a wide range of Sharpe ratios by strategy. So, why is it that everybody just doesn’t pile into the highest Sharpe ratio? Well, the reason is simple. The highest Sharpe ratio strategies often have the most negative skews.

[00:16:46] Rob

They also require the most leverage often as well, right?

[00:17:35] Cam

Which feeds into this. Which feeds into the negative skew. So, basically, what you need to do is effectively to adjust that Sharpe ratio for the level of skew. That’s the right way to do it. That is not easy to do, but that’s the right way to do it – just to select.

I see this all the time, investors who sometimes say, “We like this strategy because it got a higher Sharpe ratio.” But the Sharpe ratio is not telling you everything about the strategy. You need to look at some downside measures also, and that strategy that’s got the highest Sharpe ratio might be strongly dominated by something that’s got a lower Sharpe ratio. So this is really important and it’s kind of like a simple way to think about it.

You’ve got a strategy where you’ve been writing some put options and the market isn’t that volatile, so you get the Sharpe ratio that’s really high. But then, some volatility kicks in, the market goes down, you get basically hammered and you get the massive drawdown. That’s just the negative skew. So any strategy like that, when you’re writing options, you’re just inducing that negative convexity. That, again, is part of what we do in our paper where we detail different strategies and how they contribute to the convexity.

Rob, actually, I do recollect one situation, where you were involved, and it was actually really insightful for me. I’m sitting in a room at AHL and we’re considering adding a new strategy to the product. We looked at the dynamics of it and I believe you were the one that said, “That looks really good, but it actually is a negative skew strategy and that will reduce the overall skew of our product and that’s not what we want.” So that was exactly- in practice, I saw that happening and that was really impactful for me to say, “Hey, let’s do a book on this. This is really important.”

[00:19:59] Rob

That’s Good to know. I’m going to dig into the book in a bit more detail, Cam. So it’s actually quite interesting because I ordered your book and we were due to have the podcast this week. We were actually recording this probably a few days ahead of when it will be released, I guess, as normal. So today’s actually the Wednesday after Black Friday, which for me personally (and for quite a lot of people), was actually a real Black Friday in the sense that we had a pretty substantial drawdown in our portfolios.

So, I’m reading your book on Friday afternoon with one eye on the book and one eye on my live P&L, which wasn’t pretty I can tell you. What was the first chapter I see in your book? It’s about managing portfolios in a crisis. So, that was an amazing coincidence. I’ve got a few related questions. The thing about crises, they’re all different. The events on Friday felt to me like something that you almost couldn’t hedge against. It was something that really came out of the blue.

In the book, you describe different ways that people have traditionally hedged portfolios against crises and some ways they may not have considered. Maybe you could talk a little bit about that. How do we even think about these crises? These things don’t happen very often. It’s not like a simple linear world where we can just measure a statistic like Sharpe ratio or a correlation or a linear statistic. We’re talking about episodic events. So I’m just curious how you think we should sort of manage these hedges in our portfolio in terms of how they would guard against different crises and what the advantages and disadvantages are. I would have really loved to have a magic bullet that would have saved me on Friday, although I suspect one doesn’t exist.

[00:21:55] Cam

It definitely exists, and it’s very clear in our book. If you had put options, long put options, then you’re completely protected. You’re right, in the book, we go through various different ways that you can hedge and the most effective way is just basically to buy the put options. When the market tanks, you get a great payoff. The problem, of course, is that with any hedging strategy, you need to take into account the cost of the hedge. So those put options did great on that Friday, on Black Friday, however, what about the other days? What about over the last number of years? So one thing that we do in our paper and in our book is to look at various different episodes, which we labeled crises.

There are actually definitions. One definition is just like a market drawdown. That’s maybe the most popular. We also look at recession as a crisis also, and those are different datings. Sometimes they coincide and sometimes they don’t, but recessions are important also.

If you look at a strategy consistently buying put options, that strategy is a disaster in terms of the drag in returns. You’re going to be haircutting your portfolio performance by like 10% a year in these non-crisis periods. So are you willing to pay 10% of your portfolio to have that 100% protection in the inevitable drawdown? I don’t think so.

So it’s really important that yes, you can completely hedge, but that’s so costly that nobody would actually do it. What we need to do is to explore different strategies that are not as reliable as let’s say a put option, but they’re vastly cheaper than the put option. So you put that portfolio together, and maybe it’s done in a way that mitigates the drawdown. So it might be not a 40% drawdown but a 30% drawdown, which is a very significant difference, a 10% difference.

Often, managers are measured versus their peers, and that would be significant outperformance. So you basically make that choice and the cost, of course, depends upon how close you want to get to complete protection.

[00:24:55] Niels

Did you want to dive a bit deeper, Rob? Do you have a follow-up?

[00:24:59] Rob

I guess, yeah. What are the hedging strategies that you discuss? What are the main pluses and minuses that we need to think about in terms of what we should be using? Obviously, with put options, they’re the gold standard, but also the most expensive. So I’m guessing that, as we skew in Sharpe ratio, there’s a trade-off between the different ways we can do this hedging. So the one extreme, we’ve got no hedging at all, it doesn’t cost anything, provides no protection. On the other extreme, we’ve got buying put options, 100% protection, very expensive. So, what does that continuum look like in terms of the options that are in the middle of between those two extremes?

[00:25:39] Cam

Again, we discussed so many different passive strategies and inactive strategies. Coming back to the discussion of trend following, it’s obvious that trend following theoretically and empirically has a positive convexity sort of behavior. Think of trend following as when the market is going up, you’re buying. That effectively is like dynamically replicating a long call option. When the market is going down, you’re selling. So that is like dynamically replicating a long put option.

You put those two together and you get a long straddle. We all know that the long straddle has got positive convexity. So you’re going to be making money when the market is going up and you’re going to be making money when the market goes down.

In the middle is where you’re not doing as well. In the middle is the low volatility state where you’re actually paying the premium for the straddle. So it’s kind of interesting that if you look at the history of trend following, it’s remarkable that you do get this protection. So you do see, empirically, the positive convexity. You do better in the negative states. I described a straddle as a symmetric straddle, but it turns out that you do much better in the negative states when the market is going down than in the very positive states when the market is going up sharply. But nevertheless, the negative states, that’s where you really want that protection.

It is true, in low volatile periods, that the strategies have very low or potentially small negative returns. But this sort of strategy has historically been very effective in providing a hedge to a crisis. So the key thing, of course, is the cost. As I said, in low volatility states, you get a very low return, but it is a return. It’s small, but it’s positive.

That’s much different than the put option where you’ve got a strong negative return in the non-crisis states. So you get something in the crisis state and the non-crisis state, on average, even though there could be years where you get zero or small negative returns.

Again, it’s kind of interesting. Going back to our Sharpe ratio discussion, I’ve heard this where people say that there were a number of years ( two or three years in a row,) where the trend following returns were really low and people were saying, “Why should I invest in this when I can get double digit-returns in just a long equity portfolio?” They make statements like, “This trend strategy is underperforming the market?” You’re right that if you have this view that the way that you assess performance is just looking at a Sharpe ratio, then, yeah, the Sharpe ratio, during this period, was much lower. But you need to take these other dimensions into account.

So yes, maybe you’re below the market or maybe it’s a single-digit return or maybe it’s even a 2% return. But when the market turns and goes negative, these strategies tend to do really well, so they provide that hedge. So it is completely inappropriate to compare Sharpe ratios of, let’s say, trend following to a long equity portfolio, because the trend following, it’s got different properties – this hedging property that’s really important. So you need to take that into account. I think that’s a really good example of an active strategy that doesn’t have a lot of cost, that provides a level of crisis protection.

There is a paper that’s not included in the book (because we wrote it after the book) with similar analysis, with the main risks that we face – one of the main risks that we face today, and that is inflation. It’s also the case that during inflation surges, the trend following strategies tend to do well whereas the equity markets do poorly in inflation surges.

[00:30:43] Niels

This is super interesting, Cam. I want to stay with the topic of trend following a little bit more, because one of the things we discuss a lot on the podcast is really the role of diversification and the role it plays in terms of sitting inside a trend following strategy. Of course, we often discuss the diversification in terms of markets that we trade as a key component, but also in terms of the different timeframes that typical trend following models apply.

But I know that at AHL, for example, they’ve also really embraced the concept of diversification across models and how they can play a key role. So I’m interested if you can talk a little bit about that, because this is where we are a little bit divided, actually, in our thought process. Some of us like Rob and me, we’re probably more in the camp of saying, yeah, you should actually blend different types of trend following models, while some of our other co-hosts, they are more in the camp of diversifying across just timeframes and markets. So I’d love to hear your thoughts about this and what dimension you think diversification across models brings to the table.

[00:31:59] Cam

I don’t just have thoughts on this, I’ve got active research on this. So hopefully, you can provide a link to a new paper that is called Momentum Turning Points. We do a very simple exercise in this work and the models are highly simplified; think of looking at momentum for equity returns by a fast momentum would be just looking at the past month’s return and taking position based upon that. So, if it’s positive, you’re long, if it’s negative, you’re short. A slow momentum would be, let’s say, a 12-month momentum and 12-month lookback.

You can look at the performance of the 1-month, you can look at the performance of the 12-month, and we look at actually all of them in between, including the 6-month. But then you compare that to a diversified portfolio of the 1-month and the 12-month, and it’s actually dramatically different. So in our paper, we look at this portfolio of momentum strategies, and it almost always outperforms either the fast or the slow. We do this also across different countries. It works in all the countries except for one, and that’s Japan, but nothing really works in Japan, no surprise, and it is not really that far off anyways.

So this idea of having this blend of strategies is really important. Effectively, what it does is think of all these strategies as having signal and noise. When you have a composite of different signals, you are diversifying some of that noise away. So it’s very much just like any other portfolio of individual assets. You get this diversification effect where the noise is canceled out and you’re able to get a much sharper composite signal. The camp that I’m in, basically, I don’t even consider it a camp. It’s basic stuff. It goes back to Markowitz, 1952.

Let me just say one thing that’s kind of related to this, when we talk about diversification, diversification is something that people in finance really believe they understand. It’s one of the most basic concepts. I’ve kind of described it that when you have a portfolio of different assets, the noise cancels out, you reduce the volatility and that’s good for you. But I want to circle back to the motivation for my book. I basically make the case with my co-authors that people really don’t understand diversification, because it’s not just about reducing the volatility, it’s about getting rid of some of the tail risks also.

So there’s a different level of diversification when it comes to the skew. You need to manage that too. So when I think about a diversified portfolio, yes, I want to reduce the noise, I want to reduce the variance, but I also want to increase the skew, and that’s really important in terms of the portfolio management. So it all links back to the motivation for the book.

[00:36:23] Niels

Just as a quick follow-up before Rob joins us with another set of thoughts. I heard Sandy talk at some point about that, since he joined AHL and to now, there’s been this massive expansion of, of different models. I think he was quoting a number of 300, 400 different types of models. I don’t know whether he was talking about Man as a whole or just AHL as a group. But my question is if we want to stay true to trend following though, isn’t there a limit to how much diversification we can get in terms of different model types? Because in some ways, you would think that there is a limit to how many ways you can do trend following.

[00:37:07] Cam

Yeah, this is really important and this, basically, is what I mentioned earlier about Rob’s question at one of the meetings where we’re considering a new signal. Oh, well, that’s got negative skew and that actually takes away from our portfolio of positive skew. What we’re offering to investors is something with positive skew. So that is the foundation of trend following, that when markets go down, there should be a payoff. That’s what people expect, at least on average, not every single time, but at least on average.

So there are plenty of great strategies out there that have got this negative skew, they probably shouldn’t go into a trend following product because it’s going to drift the style. You don’t want style drift. It can be a disaster. So people buying into a trend following strategy thinking that it will do well when markets go down, then the portfolio constructors add all these signals that are non-trend following signals, and then all of a sudden, when we have a drawdown, there’s no hedge, the strategy actually goes down and could go down by even more. So you definitely don’t want to do that because investors want basically what is being sold to them.

It’s perfectly fine to have another product that might have some trend following component but that has got other stuff in it that definitely will perform differently. So maybe the return on averages is higher over the longer horizon but it doesn’t provide the same level as protection. I think that you need to be really careful about this style drift because it could lead to investors being disappointed.

[00:39:16] Rob

One thing I must say, and this may sound a bit cheesy, the paper you wrote with Harry Markowitz and also Rob Arnott is ‘the bomb’ paper. I say to my students, “If you only read one academic paper this semester, read this form because it’s so relevant to what we do.” It’s called Backtesting in the Era of Machine Learning, I think, if my memory is correct.

[00:39:46] Cam

Yes, that’s correct.

[00:39:48] Rob

Anyway, back to the book, for those of you reading along at home on Chapter Three, Page 71 in the print edition which is entitled, Risk Management and Volatility Targeting. Again, this really caught my attention because volatility targeting is another very controversial and popular subject on the podcast, although, in fact, really, when we’re discussing it on the podcast, it’s really a question of degree and frequency. So we’re mainly divided into camps of people who vol target when they open positions and then keep the positions maintained, or people (and I’m in this camp) who actually dynamically adjust their positions every day according to the vol target. But we’re basically doing the same thing really, it’s just a question of frequency.

Volatility targeting is one of these very simple things that we can do which does produce, actually (and if I’m being honest here), probably the most dramatic improvement in risk-adjusted returns when going from, say, a long-only portfolio to vol-adjusted long-only portfolio. It’s quite impressive actually, because when you layer on conditioning signals like momentum, and carry, and so on and so forth, it obviously improves further. But the improvement from volatility targeting alone is really quite striking for something so simple. So maybe you could explore that a little bit for us.

[00:41:06] Cam

Sure. There’s a number of interesting things in that chapter. The first thing is really important, and that is that, empirically, we’ve had these academic papers that show that volatility targeting does just as you said. It’s really, really effective. It improves Sharpe ratios. However, that research was applied to equity markets and people read these academic papers. By the way, I’ve been teaching this in my asset allocation course for 15 years. Indeed, I never thought that, “Oh, well, is it publishable?” We know that, for example, trend following firms use volatility targeting routinely in terms of their signals.

So this is really important. It works for equity, but it doesn’t work for everything. You need to step back and figure out why. It works for equity. It works for equity-like strategies like credit. But when you go to fixed income in general, like government bonds and stuff like that, the strategy just doesn’t work. So let’s step back and figure out what’s going on in terms of theoretical mechanics and then, statistically, what’s happening.

[00:42:51] Rob

Just very briefly, just to double check that what you’re talking about here is what you described as the leverage effect in the book. Because that was a term I hadn’t come across before.

[00:43:00] Cam

So that’s the theoretical basis for this. It actually goes back to an idea of Fischer Black, the late Fischer Black. Basically, think about equity as being like a call option; you need to pay off your debt, and whatever is left over, you get or the equity holder actually gets. That’s the foundation of the Black-Scholes 1973 paper. But think about what happens when the equity market goes down or the price of a stock goes down. When that happens, the leverage of the firm goes up. So think of a debt equity ratio as the value of the equity goes down (the denominator goes down), and the debt equity ratio goes up, and effectively, you become riskier.

So the idea here is that’s bad news. If you’re riskier, then expected returns need to change, and that basically is the leverage effect. The idea here is that we expect to see a negative relation between the returns and the volatility. So when volatility is spiking, this is basically increasing leverage and markets are tanking, the prices are going down to get a higher expected return. It’s kind of confusing that to get a higher expected return, the price needs to go down, just like for a bond to get a higher yield to maturity, the price needs to go down. So this is really fundamental to the idea.

As to what’s actually going on, it’s kind of interesting because if volatility is increasing, then you decrease your dollar positioning. So, you’ve got equity investment, volatility is increasing, you want to keep a constant level of volatility, therefore, you need to reduce the dollar weight in equity. So as the volatility is increasing, you’re selling. This is interesting because it links back to the trend following. When the volatility is increasing, it’s almost always the result of the market going down. As the market is going down, the volatility increases and you’re selling, you’re reducing your equity position.

That sounds a lot like what I described as the dynamically replicated long-put option. The volatility targeting is serving a very similar function to what a trend following actually does. In the other way, it’s kind of the same thing. As volatility is going down, that’s usually the markets going up, you need to increase your investment in the equity, and that’s dynamically replicating the long-call option. So, the volatility targeting naturally induces this positive skew or positive convexity. Again, this is important to realize that the theoretical effect should only apply to risky assets like equity and credit and this just doesn’t have much traction for other asset classes.

That’s the basic idea in this paper that we focus on volatility targeting because it is a positive convexity strategy. So think about our book as providing some methods for investment. Volatility targeting is a method. So, this is something that you operate your portfolio with a rule like that. We also provide analysis of the inputs for the portfolio. So you’ve got the inputs and then what do you do with the inputs? The book deals with both.

[00:47:34] Rob

That’s very helpful. Because yeah, it’s just something that I’ve always done without really thinking about it. So that was very helpful. Thank you.

[00:47:41] Niels

I wouldn’t mind if we could stay on the topic of volatility a little bit further, because one of the things that I find interesting is, and I don’t know the exact sequence of events, but if we think about volatility going back in time, it starts out as being this measure of riskiness. That’s kind of how we started to think about it. But then suddenly, at some point, I’m not entirely sure when that happens, it becomes an input in our models, and suddenly, it becomes an input in everybody’s models.

I’m curious to know how you think of this change and whether that structurally changes something, because at the same time as this one component becomes a key player in our arsenal of inputs, a big part of the world has gone quant at the same time. So you’ve got more assets following models, you got this one input playing an important role for all of these models. How do you think about that in terms of riskiness?

[00:48:54] Cam

Let me rephrase your question a little bit, and that is, your chapter shows all the advantages of volatility targeting, people read it, and those that aren’t doing it are convinced now so everybody starts to do the volatility targeting or a very substantial number of large investors do this. Suppose the market goes down, but not by a lot, let’s say 5%. Because everybody is doing this volatility targeting, the minus 5% is linked to an increase in volatility. So we need to reduce risk. Everybody, at the same time, is basically selling equities.

When this happens, because a large number of people are selling, this could be effectively self-fulfilling. Because people are selling, then the market goes down even further. So the argument, and it’s only an argument, is that maybe that 5% correction was appropriate given the new information that came into the market. But you’re driven below the 5% to maybe negative 9% because of this mechanical situation where people are dumping their equity to achieve their volatility target. This makes the market less efficient. The correct price should be minus 5%, but we’ve seen minus 9%.

This argument also, just to be clear, applies to trend following identically. A large number of people are trend following, the market goes down by 5%, the trend following signal says to sell, and potentially, the market is driven to minus 9% and not reflecting the efficiency of the market.

So this is basically a story that’s told, and I think the story doesn’t really have much merit. Let me tell you why. There are many very smart investors out there. They have got their own models and if they detect that some information came into the market, the market corrected 5%, If they see the market going below that 5%, they’re buyers. So these are value investors. The market is cheap, and if you see this deviation, they’re just going to buy.

So, we’re not going to see this large deviation that’s induced by volatility targeting investors or the trend following investors. There are plenty of other investors that if they see or they detect that the market prices have deviated from what they consider the correct range, then they will be buyers and will take us back. So I’m pretty confident of that. Indeed, if you think of the volatility targeting, while theoretically justified, it is a strategy that is a mechanical strategy. It’s a statistical strategy. The same thing with trend following. The market return is negative, therefore, you’re a seller.

There’s no other information that goes into that other than that algorithmic type of rule. Whereas there’s plenty of other investors out there that if they see some deviation, they will come in and either buy or sell depending on what side you’re actually on. So I think that the size of these players is so large that it is just not credible to make the case that either volatility targeting or trend following leads to a market that is less efficient. It doesn’t make any sense to me when I consider it. Do I believe that the market is efficient all the time? Obviously not. It’s never efficient. University of Chicago PhD, my chairman, is Eugene Fama.

[00:53:44] Niels

They’re going to take your PhD away from you, Cam.

[00:53:46] Cam

No, no, they’re not. That’s the key thing, that people misunderstand what the efficient market hypothesis actually is. Markets are efficient in basically very stylized, very special conditions that are completely unrealistic. It’s just a theoretical construct. What really matters is just how efficient the markets actually are. So, it’s a relative measure. No market is efficient, and certainly, my dissertation chair does not believe that all markets are efficient at any point in time.

So there will be some deviations and I guess the efficiency of the market just depends upon how fast players come in, when there’s significant deviations, and drive prices back to a reasonable range. We don’t know what the true price is, but we might have an idea of a confidence range, a 90% confidence range.

[00:54:52] Niels

Yeah, it’s quite interesting. The whole thing about volatility in general, I think a lot of investors focus a lot of their time on the point of volatility, but they end up equating that with risk saying it’s very risky, it’s very volatile. But I think that the risk we really should be worrying about is obviously the drawdown or the risk of drawdown, so to speak. I know you touch on that in the book as well, so I don’t want to take all the spoils away. I know Rob has more questions, but I think it’s an important distinction between those two things.

[00:55:27] Cam

Think of drawdown as very much related to skewness. It’s different because when we think about a maximum drawdown, it’s just a single episode, whereas a skewness measure uses the entire data to actually come up with a measure, but it’s related. If you’ve got something, a strategy with a large maximum drawdown historically, that strategy likely has negative convexity or negative skew.

Drawdown is a popular measure in the practice of finance. It’s interesting, in the academic textbooks, you almost never see a mention as a statistic. So we thought that it would be good to do a chapter on drawdown because it’s just so consistent with the theme of the book that you need to manage not just the expected return and the volatility, but you need to also look at this third measure on the third dimension of skewness. Just to be clear here, we think of high volatility as bad. But volatility essentially doesn’t distinguish between big positive returns and big negative returns.

For example, if there’s a surprise large positive return, then that actually increases the volatility. But that’s good volatility. So given this ambiguity that some volatility is good, some volatility is bad, it’s really important to go to this third measure that tries to asign the volatility as good or bad.

[00:57:27] Rob

That’s the thing that people always forget. You’ve got to look at skew, because obviously, standard deviation is symmetry, and that doesn’t exist. You need some way of measuring how much we departed away from that. Just shifting gears again slightly, another thing in your book, which again, is one of those things that we just do, which is rebalancing of course. Rebalancing is quite similar to vol targeting. There’s a theoretical idea of Shannon’s Demon, which means you can basically make money by buying and selling as the asset follows this stylized random walk.

But of course, the interesting thing is when you go into the empirical side of this, is that when you’re rebalancing, what you’re actually doing of course is fighting momentum. Because you’ve got something that’s going up in price, and happily, every month, you rebalance your portfolio and you’re selling this thing consistently. So you’re kind of fighting against the momentum.

I guess there’s two ways of dealing with this problem. The way I would do it is obviously just to apply a momentum filter to my strategy anyway. So, although I would be rebalancing for some notional expected risk, that expected risk will be changed according to the momentum of the different assets I’ve got. So if equities have been going up a lot, then I’d be rebalancing, but the size of my equity risk could be increasing over time because of momentum. So, those two effects would possibly offset or I’d even end up being a net buyer of equities as they went up, possibly.

Now, in your book, you discuss a different approach, which is about delaying rebalancing, which to me, seems a little bit of a backdoor way of getting momentum into a long-only portfolio. Is that a fair characterization or do you think it’s a genuinely useful and interesting thing to do?

[00:59:18] Cam

Well, I believe it’s interesting because it’s in the book.

[00:59:25] Rob

It’s definitely interesting, but to me, it seems like a slightly secretive way of getting momentum into a portfolio. So we’re not doing momentum, but we are delaying our rebalancing.

[00:59:34] Cam

We need to think about the big picture here. I mentioned diversification as something that investors really believe they understand. There’s another concept that they believe they understand and that’s rebalancing. So those two concepts, probably if I did a survey of investors, those are going to be right at the very top in terms of what people are sure they understand. Our chapter makes the case that people actually don’t understand rebalancing. So this is another method contribution. The volatility targeting is one method contribution and the rebalancing is another method contribution. So you’re correct that this is related to momentum.

Let’s actually go through an example of what happens with rebalancing. Suppose you’ve got a 60/40 portfolio and the equity market is tanking. So this is the situation where we’re not doing volatility targeting, we’ve just got a cap-weighted 60/40 portfolio, and then when the equity market is tanking, you need to rebalance. So your portfolio has become 50/50 and you need to buy equity to get it back up to the 60/40. Think about that dynamic and the dynamic is that the equity market is falling. What do you do? You’re buying. That is exactly the opposite of what you would do in momentum. As you said, it’s fighting a momentum.

This mechanical rebalancing strategy is analogous to… because when the market is going down, you’re actually buying dynamically replicating a short-put option. When the market is going up, you’re selling, and that’s analogous to dynamically replicating a short-call option. So you put those together, and you get a short straddle which is negative convexity. What we show, both theoretically and empirically, is that the rebalancing induces this negative convexity and mechanical rebalancing actually increases the risk of your portfolio.

Indeed, we’ve got plenty of examples in the chapter where we show that the drawdowns, the worst drawdowns, are actually made worse by this mechanical rebalancing. The chapter basically says there’s a number of things that you might consider in terms of mitigating this increased risk.

The first thing is that I think readers will, after going through the chapter, understand that a mechanical rebalancing strategy is not what you want to do. They probably had no idea that this actually increases drawdowns; so it increases the risk on your portfolio. So what do you do? We can go through some heuristics, like you do only a fraction of the mechanical rebalancing every month or every quarter or you have some banding.

Our empirical results suggest that that is minimal effect, it’s likely not even worth doing. Then we consider basically counteracting this extra risk that’s induced by the rebalancing by allocating to a positive convexity strategy like trend following. So basically, the rebalancing is inducing anti-momentum. You cancel that out by allocating to a momentum strategy. That works fine, but there are two issues here.

The number one issue is that it’s kind of weird to have a strategy where you’re doing this mechanical buying and selling but then you’ve got to allocate to something else to effectively undo this buying and selling. That’s number one.

Number two is that, if you’re a very large investor like a pension or a government pension (which could be a trillion-dollar portfolio), there’s just not enough capacity in these markets to actually counteract the rebalancing with an allocation to trend strategies. So, the other idea in the chapter is to use the information in the trend signal without actually investing explicitly in the trend product, but use that information strategically to time your rebalancing. A simple example here is the equity market is selling off and your mechanical rebalancing says we need to buy.

Well, you look at a simple trend signal and the trend signal says the trend following would actually suggest that this drawdown is going to continue. Given that signal, then you delay your rebalancing. So you wait. You use that information strategically. When the trend signal reverses and says we’re at a turning point, that is the point where you actually do your rebalancing. So you’re using the information in the signal to change from mechanical rebalancing to strategic rebalancing.

While it’s not as effective as just a pure allocation to trend, it is very effective overall and you get rid of a lot of this negative convexity that rebalancing induces. So, in terms of all of the chapters in the book, I think that this chapter has got the most to offer. It’s got tangible value in terms of portfolio management for both large and small portfolios.

[01:06:45] Rob

There’s an analogy here, isn’t there? Because a lot of big funds, they won’t use FX carriers as an investment strategy, but on the other hand, they will tactically do their hedging, their currency hedging, with that in mind. So they wouldn’t necessarily hedge if it was going to cost them a lot of money in carry. So yeah, maybe I was being a bit unfair when I said it was sneaking in through the backdoor.

There is another kind of issue here, isn’t there? Because it’s to do with timing. I guess most of these funds rebalance monthly, which is pretty much the worst possible frequency in terms of fighting momentum, isn’t it? Because if they rebalance daily or rebalance annually or every couple of years, then you’d probably be outside of the trendiness period and into the mean reversion period. There wouldn’t be so much of a drag on returns.

[01:07:35] Niels

Rob, I’ve got only a couple more questions because I know we need to be mindful of Cam’s time, but did you want to talk about anything in machine learning that’s more of your streak than mine?

[01:07:47] Rob

Machine learning, goodness me. The only reason I thought of it was, I call it your last substantive chapter in the book, Cam, because there’s a chapter at the end which is about the- was the true out-of-sample evidence of the COVID crisis, the COVID-19 crisis, which is really good to see in that book, obviously, because we rarely get to see the true out-of-sample with academic research. Whereas, of course, as practitioners, we’re living out of sample every day of our lives.

But the chapter before that which is man versus machine, I thought was very interesting because you have a take on this eternal debate about whether it’s better to have a systematic strategy or a discretionary strategy. Can our simple computer rules really replicate, match or even beat these competitors? So that was just the question, really. But in the interest of time, Niels, maybe we’ll cut what I’ve just said out and go into your question instead.

[01:08:54] Niels

No, no, no. We certainly won’t leave your wonderful comments out, Rob. But I am going to try and pack three questions into one final one. In one way, I want to remind the listeners of our conversation about a year and a half ago, where we touched upon inflation. I want to tie that into a couple of things we’ve already talked about today. Some of the things that I observe, and I’m sure you do as well, is on one side, we talk about diversification being a great tool, so to speak.

On the other hand, when we look at how markets are today, they’re probably less diversified today than they’ve ever been. They’re dominated by some very large players. They are dominated by some very large companies, so to speak. So there is a little bit of an issue there, I think, the emergence of inflation and then on top of that, I’d love you to put that into perspective of the 60/40 portfolio as that still seems to be quite a dominant allocation out there. I don’t know if that’s even possible to put these three topics into one answer. But if anyone can do it, Cam, it’s going to be you.

[01:10:21] Cam

Well, that is probably going to take to like another podcast to go through all of that.

[01:10:27] Niels

Right. I know.

[01:10:28] Cam

But let me focus on the inflation a little bit because I know it’s really important. I’ve been very active on LinkedIn, basically criticizing the political and Fed spin of this transitory inflation. It doesn’t make any sense. My paper that looks at inflation, that hopefully you can link to, basically shows that the 60/40 portfolio gets hammered when there’s unexpected inflation. So inflation, we look at data over the past 95 years from three different countries, but focusing on the US. We’ve had, when we published the paper, eight searches. We’re now in number nine, and that’s serious.

When we look at these surge periods, equities do poorly and then it’s kind of obvious that fixed income does poorly. If inflation surges, that usually leads to higher rates and higher rates are obviously bad for bonds. So the 60/40 portfolio is poorly positioned in terms of an inflation surge. What can you do about that? Well, the first thing that you can do is to reallocate within your equities to sectors that are more resilient to inflation.

For example, consumer durables is the worst sector in terms of performance during inflation surges. The average return during a surge is, like negative 41%. It’s really bad. So if you want to reposition your portfolio, that is an obvious sector to do. But we also show that various different factor strategies like quality strategy, so tilt your portfolio more towards quality, that has got some ability to hedge the unexpected inflation also. We look at various different asset classes. I’ve already mentioned equities that does poorly, but there is a diversity of performance within equities. Fixed income, unconditionally does poorly.

Commodities are resilient to inflation. Indeed, some commodities are the foundation of inflation. We look at various different passive strategies, but also many active strategies too. The factor strategy that I described, quality is an active strategy. Trend following, something we’ve talked about a lot already today, that tends to have very good inflation hedging properties. It’s kind of a very strange time that the Fed has got this target of 2%. They’ve made all the statements that are not very credible, to me.

My LinkedIn really focuses on the credibility of the CPI numbers that are even coming out. So the latest print is year-over-year 6.2%. But if you look at the components, it’s really interesting, because the largest component is shelter. Think of that as housing or rental equivalents. The rate of inflation for shelter is 2.7%. It is actually contributing to a lower CPI. So 2.7% is way less than, let’s say, the 6.2%. But just a casual examination of the data suggests a massive disconnect. The Case-Shiller 20-City housing price index, year-over-year is up 19.9%. If you look at rental metrics, the year-over-year change in rents is up over 10%.

So how does that square with the 2.7%? Really, the way I look at it is that those increased costs will just make their way into the shelter component and we will have elevated inflation for quite a while. What does this mean? Inflation, this isn’t really news, but it’s only recently that the Fed has abandoned the word transitory, which I never believed. Indeed, it just doesn’t make any sense that it’s 100% transitory. Certainly, some of the increase is transitory, but it’s not 100% transitory. This is serious stuff, and indeed, it is an ideal time to reposition your portfolio to basically provide some sort of hedge for higher-than-expected inflation.

I think it’s naive to think that inflation will be 2% over the next five years. The bond market seems to disagree with my view, but of course, the bond market is being impacted by $120 billion of monthly purchases by the Fed. So I believe that we are basically in the 3% range for the next five years and that should be a trigger for investors to basically reexamine the positioning of their portfolio. Are they set up to withstand a surge in inflation that’s not just for a few months, but potentially, for a few years?

[01:16:38] Niels

Absolutely. That is absolutely perfect. Cam, I’m going to end it on this note. Thanks ever so much for spending some of your time with us today. We really appreciate this, as I’m sure all our listeners do. By the way, make sure you follow, subscribe, and support Cam’s work, and of course, make sure you get a copy of his latest book Strategic Risk Management.

As you can tell from today’s conversation, designing your portfolio and managing your risk is perhaps more important than ever before. From Rob and me, thanks ever so much for listening, and we look forward to being back with you on the next episode. In the meantime, take care of yourself and take care of each other.

[01:17:20] Ending

Thanks for listening to Top Traders Unplugged. If you feel you learned something of value from today’s episode, the best way to stay updated is to go on over to iTunes and subscribe to the show so that you’ll be sure to get all the new episodes as they’re released.

We have some amazing guests lined up for you and to ensure our show continues to grow, please leave us an honest rating and review on iTunes. It only takes a minute and it’s the best way to show us you love the podcast. We’ll see you next time on Top Traders Unplugged.