Return of the Trend: 'It's all about Correlation'
Niels: I had a question the other day from someone here based in Switzerland, Roman, in fact, and he asked about people's perception about Trend Following that perhaps it's performed poorly in the last couple of years because there's been too much money chasing Trend Following after the great year of 2008. When you hear something like that, what comes to mind?
Katy: Well, I actually just wrote an article which is coming out for Eurex on this exact topic, and I call it Return of the Trend: It's All About Correlation. I'll just give you a view on this. If you look at a Trend Following system, any portfolio system in general, we depend on correlation. We depend on the diversification across markets, and regardless of looking at the capacity in the industry, not even thinking about that, if you take a graph of correlations pre in the last twenty years, it almost looks like a step function. So up until 2008 the correlations are pretty low across all futures markets, and they just shot through the roof in 2008 and they stayed there until earlier this year, or until late 2013. If you think about portfolio construction, the returns of Trend Following is driven by divergence and we've had some divergence over this period of time: quantitative easing, all sorts of events like the nuclear meltdown in Japan, those are diversion events, but in this sharp ratio is also the diversification and the risk. When you construct a portfolio, it's not only the volatility, which has been low, but also the correlation across assets that allows you to have proper diversification. Correlation being high means diversification is low, which means that even though there maybe some trends, there's a lot of risk because it's sort of like a one trade world. If you look at that, that coincides with a period that has been difficult for Trend Following strategies. So they do have profits in some areas, but there was just not enough diversification across their portfolio I think to support their performance as consistent with history.
Niels: Sure, sure. The next area I want to talk about is what usually is the trading program, but today, in our conversation it will be about the trading strategy or the model, however you phrase it, that you've used in your study that really represents the performance over this long period. Tell me about how you and Alex constructed this and feel free to go into as much detail as you want.
Katy: OK, yeah, this is a very important and good question, Niels. One of the chapters of our book that I'm most excited about is actually chapter 3, and this is one called Systematic Trend Following Basics. I found that when I looked at most other books on trend following, and those sort of descriptions, it's very hard to find a specific formula that you could use as "this is the formula" for Trend Following. So what we did, instead, is we tried to create a framework with one formula.
...we focus on creating one formula for Position Sizing, which is a function of several key variables.—Katy Kaminski
This one formula obviously, as I said was science and art, can be ejacted and made much more complex by any particular manager, but we actually used this formula to build a framework for style analysis later in the book. So we started off by asking what are the four key question of trend following? We need to determine when to enter? How large of a position to take on? How to get out of a position and how much risk to allocate to any particular position? We then go on to define the five key building blocks of a trend following system, and these include data processing (which is pretty straight forward), signal generation, position sizing, market allocation, and execution. In this process, the most important part that we focus on is position sizing. So we focus on creating one formula for position sizing which is a function of several key variables. So this function is that the nominal position amount is a sizing function which tells you the size, and we leave that very open in the book because that is a lot of art right there. So the size of the position is a sizing function, times the risk loading (this allows you to gage leverage up and down), times the capital allocated to that particular market, divided by the total adjusted dollar risk, times the dollar risk of a position. So, basically let me do it in laymen terms - I'll do it as if I was telling it to my husband. So what you do is you say, OK the sizing function tells you how strong your strength, your conviction is. So if your models, like your moving average models say that this is a very strong trend, then your sizing function weights based on those rules. The risk loading is just a simple loading of risk across markets, which allows you to lever the position up and down. So if you want to have more or a larger exposure, then you can expand every position at the same level. The capital allocated; that's simple, it's just a dollar amount that you put into the position. You're going to multiply that by the dollar risk of that one contract and then, because not all futures contracts are created equal, some are more volatile than others, we need to divide it by the total dollar risk of the position. This allows you to adjust, if you have something like lean hogs compared to oil, you're going to adjust the one that's more volatile down because you need to put less capital in it to expose yourself to the same amount of risk as the other contract.
Niels: Sure. Now you mentioned signal generation. I guess there are two main forms of signal generation, I guess in trend following, one that uses moving averages, the one that uses price breakout channels, what did you use in your study and why?
Katy: So we actually left that open as part of the sizing function, and then later in the book we will apply that to the sizing function. So if we do an example, we'll say that this is a moving average system, which means that the moving averages were used to create the sizing function. And we also explained that it's obviously not going to be perhaps one moving average, but it could be a basket of moving averages, where you take this average over a range. This is where art comes in here. So the sizing function is the science but the application of that depends on how sophisticated you get in terms of creating that function.
Niels: I think you also mentioned something about... I think a lot of people are focused on where did you buy something, and if you talk to people in general, and they talk about their investments, they would always say, "oh, I've just bought IBM, or I've just bought Facebook stocks," but you rarely hear them say, "I've just sold Facebook, or I've just sold IBM," but you actually put a lot of emphasis, I think, on the exit of a trade relative to the entry point. Talk to me a little bit about that and why this is important for investors to be aware of.
Katy: If you think about... a lot of people think that predictability is coming into play when you are talking about using it in a trend following strategy. If you're following a trend, you're not necessarily starting it. So I think the point of a trend following strategy is there's both the entry and there's the exit and we explicitly explain that these decisions are different and actually the exit is a very important part of the whole process. What we do to do this analysis, and this is kind of an exciting part of the book too. Is that in chapter 5, after we explain this concept of the divergence and divergent risk taking and the importance of cutting your losses. We look at trading systems that have agnostic entry rules. What I mean is that, imagine that, instead of actually using any sort of indicator to get in, you randomly flip a coin to which market to get into. It turns out when you look at that system there's still performance that comes from getting out of the position. So that tells you something about the driving of the getting out. The exit decision and how important it is. You see that that varies over time. Then we also look at the entry and try and determine, OK, is there some predictability? So we look at a concept called trend leakage. The idea is how often is a trend following systems position positively correlated with a future trend position? We see that trend leakage actually does exist and it's also time bearing, so for some periods of time some of these trends seem to leak out into market prices, but there's also plenty of times where trend leakage is actually rather weak. What that tells us is that, in some scenarios, you may actually have some trends leaking out if you use a systematic approach, but there's other times where it's actually... and a lot of times where it's really that sort of getting out of the position when the trends are disappearing as well, that drive a lot of performance.
Trend Following is long Divergence, so if inflation causes Divergence, great for the strategy. It's really about understanding that if inflation will drive divergence and prices, then Trend Following should work.—Katy Kaminski
Niels: Right and I think that's the key to understand, those few words that you said at the end that exits to a large extent can be the driver of the performance. Environment: we talked about it already about when these things don't work and when they work, but I want to talk about environment in a different context because investors, and in particular institutional investors they often talk about inflationary versus deflationary environments and how is this all going to shape up. I noticed in the few pages that I did see from your book that you actually have a chart of annual inflation rate in the US and the UK. I think it goes back to the 1700s actually, and then you show the performance of a trend following system in these different periods. Talk to me a little bit about these kinds of environments and how that is framed in your and Alex's mind when it comes to trend following.
Katy: Well, I guess this goes back again to the same philosophy of trend following. Trend following is long divergence, so if inflation causes divergence, great for the strategy. Maybe not great for everybody, but it's really about understanding that if inflation will drive divergence and prices, then it should work. I don't have the polar view on low inflation; high inflation is good. I just think that this is a type of strategy that is meant to adapt to spectacular moves. So if we see that over time, as a result of the inflation, then inflationary environments of different types will be interesting for us. The same is true for interest rates and I get that question a lot.
Niels: So if I understand you correctly, Katy, what you're saying is it doesn't really matter whether we have deflation or inflation, and obviously for the last 40, 50 years we've pretty much had inflationary environments of some degree, but recently, it very much looked like deflation could be back on the table. But what you're saying is it doesn't really matter as long as it creates divergence, it's OK for trend followers.
Katy: When you look at the statistics from 1720 to about 2013, which is a roughly 300 year period. We see that the average return during a low inflation to deflationary environment is about 10.4%. The 5% to 10% inflation is about 10.1% and for high inflation, it's actually about 15%. So what I would say is that extreme moves... extreme environments tend to have divergence and thus opportunities. So from history it doesn't look like there's less or more momentum in a sort of low inflation vs. moderate, but it seems to be that high inflation, really high, and possibly extremely low, or extreme periods of deflation may actually cause things to be interesting for someone who is divergent.
…we say Unhidden Risks are risks that come up in price, so price risk. Hidden risks are risks that do not come up in price, so they inflate Sharpe Ratios.—Katy Kaminski
Niels: Katy, the next thing I want to talk about is risk management and based on everything you've done and the way you look at these things, I wanted to find out how you define risk and what is the important risk to look at when you look at these strategies?
Katy: Yes, and this is a very important question because in my opinion I think risk management is the greatest asset of the CTA industry. What I mean by that is that risk management and being sophisticated about that is the value added that a manager has over someone hiring two guys and saying OK, build this trend following system. Read Katy's book, read this book, implement this. Having a sophisticated understanding and very good risk system and an allocation... understanding of how to allocate risk over time is one of the greatest attributes of professional management. Let me give you an example of this, which I think we go over with in chapter 5 of our book. So the example works the following way: imagine that you, one year from now, had perfect knowledge of the price of one particular market. So you know that oil's going to trade at X in one year. Now that seems like really valuable information, but if you just take that position and hold it, over time, the risk of that position could be huge, because markets could go up or down, or up or down, over time and you may actually get completely wiped out based on just the price movements. So what we do is we sort of add risk management to that trade and sort of start managing risk. What you see is that as you manage risk, over time (so maybe half of it is using the forecast and half is not using the forecast) the sharp ratio approach improves dramatically over time. Drawdowns also improve. So risk management is the way that... the major value added of any trend following system and it's the art that a manager adds to any sort of simple product or adding two guys on a team and telling them to, "hey code this." The systems are, in theory, not complicated, but in practice, having a very sophisticated approach to risk and understanding how to dynamically adjust it properly is where you can really add value.
Niels: Sure. Risk can come out in many different ways, meaning people can look at standard deviation, value at risk, margin to equity, risk to stops, whatever it might be. Is there something that, from your point of view would give you more comfort knowing or looking at when you look at a strategy or a manager?
Katy: Understanding how they change their risk. How did they allocate risk? Having a good idea of what causes them to allocate more or less risk to a particular position is very important. I think the real danger is when you have dynamic leveraging, or when you have positions growing at risk and accelerating. That's what we're trying to avoid. So understanding if risk allocation is actually conditional or not. We actually touch on this in the book as well when we talk about drawdowns and we also talk about leveraging over time. We discuss how you can look at leverage inter-day or day to day and sort of determine if leverage is a function of past P&L or not, can tell you something about how the manager takes his or her positions.
Niels: Sure, you also mentioned something called hidden and unhidden risk, what do you mean by that?
Katy: Oh yes, that's one of my favorite topics. I like this topic from going back to... I wrote a paper for the CME about this, and it was about 2011, and I was thinking of this a lot. Trend following strategies, and futures strategies in general contain much less of these hidden risks. So what are unhidden and hidden risks? In our book, we say unhidden risks are risks that come up in price, so price risk. Hidden risk are risks that come up not in price, so they inflate sharp ratios. So liquidity risk is a hidden risk, because it doesn't show up until it shows up, and it's very hard to measure prior to it arriving. Credit risk is also hidden. It's very hard to estimate credit quality, so it will come up in shocks and sort of out of the blue and the numbers will not. No matter how hard you try and estimate default probabilities and things like that, there are always very few observations, which means that it's very, very hard to actually calibrate properly, and risk adjust for credit risk. Another important one is leverage. Leverage can be very transparent, but there also are ways to imbed risk in leverage, and we go through that in detail. We say leverage should be very transparent, but there are actually methodologies like using dynamic leveraging where you can increase risk by creating some cyclicality in your leverage application that won't show up on a slower frequency. So what I mean is that if you're losing you double your bets. Most trend following systems don't do that, but some do. If you do the right analysis, you can see that it makes sense because it creates some cyclicality in the application of leverage over time, which is fine as long as it doesn't catch you in the wrong direction, in the wrong moment.
I think that Momentum, just like Value, just like other sort of risk premia, are time-bearing, so hopefully we're back in that time again.—Katy Kaminski
Niels: I think it's actually quite topical. You and I, we're talking at the end of October of 2014 and from what I hear, certain strategies out there in October had a very rough time. I'm thinking here about option strategy which, again managed futures, which is a word that is being thrown around a lot, covers many different strategies not just trend following, and part of that universe is option strategies. Talking about divergent and convergent strategies, it's a good example, so it's just interesting about that particular area because clearly there are risks there in those kinds of strategies that investors may not be aware of, and we'll see how the month ends, but it looks like it's going to be one of those months where some of these hidden risks have come out.
Katy: Well Niels, if I give a little comment on that, just to kind of bring this back to Derivatives 101 or something like that. If you look at an option strategy, it applies dynamic leveraging. So let's say that you want to invest in a call option. The delta of the call option increases as a function of P&L. So that means that, in some sense, you build your positions and the leverage increases with an option strategy and so what that shows is that a month like this where maybe your leverage might have hurt you, an options strategy would suffer, but a trend following strategy that doesn't do that, won't.
Niels: Sure, absolutely. Great example, thank you.
Katy: The manager that does that in their strategy will probably have a harder time than a trend follower that doesn't use any dynamic leveraging.
Niels: Yeah, true, and speaking of drawdowns, which is typically the next thing I talk to people about, in your opinion, from a 30,000 foot view, have you come up with any measure that could explain or give a framework for investors in terms of what kind of drawdowns should they expect from a trend follower without the red lights going completely berserk?
Katy: What I would say is that, if you look at trend followers, they have a lot more drawdowns over time than long equity, but they're way shorter. So if you look at a history of a drawdown picture, and we do this in the beginning of our book of an equity strategy, long equity versus trend following, you'll see that there's a lot of small drawdowns and they're very often. This is because most of the time I would say that a lot of times... if you go back to the same analogy of venture capital versus private equity, right, 2/3 of the time or maybe over 50% of the time there may not be any opportunities. But when there're opportunities you gain way more than you've lost. So as a result, you're going to have to expect that a strategy like trend following that has lots of price risk and no other hidden risk, because hidden risks - what do they do - they create huge drawdowns and they happen rarely. This type of strategy is very systematic, only has exposure to price risk, and thus has lots of small drawdowns over time which are compensated by larger returns.
Niels: You say that, and I accept it, of course, coming from that world, but on the other hand, looking at the last few years, what we did see was that many managers including those who have been around for 20, 30, 40 years, saw in most cases much bigger drawdowns than they have seen before and longer drawdowns. I know that it's very dangerous to start calling for the death of trend following and this time it's different is some really dangerous words to use, but how do you, having done all this research and studying, how do you frame the last few years for CTAs and trend followers, and the performance, and the expanded drawdowns and prolonged drawdowns, how do you frame that in the overall picture? Is it just because we feel it's a little bit different because we're looking at the last 20 years, that we remember, or was it a bit different this time?
Katy: Well I think it's perfect you ask that because an academic from Edinburgh, there's a new paper out that he's by, I think it's Hutchinson and O'Brien, and this particular paper is called Is this Time Different for CTAs, and I think it's a very good paper to go and look at because here's what they say. They look at many different past crisis periods for trend following and CTAs, and they show that the performance of trend following tends to be somewhat depressed post the crisis, and this happens to be the case for every single crisis that they looked at. It got me thinking a little bit, listening to them talk about it in that, in some sense we spend a lot of time here talking about debt overhang and the difficult period post-crisis recovery. It's sort of a recessionary period, and it seems that after the profits are made from the divergence in a crisis, there is a period where markets have to re-stabilize, people have to... the market ecology has to readjust. I would say that it just happens to be that this particular period the crisis was so bad and we see that when we look at what's happening in financial politics that it's still sort of sorting itself out. I'm sure you might actually find some similar results if you looked at the great depression, which also took a lot of time to recover. So it's quite possible that these sort of delays and reestablishing what is the new paradigm in financial markets. How long will we be sitting with still trying to deal with EMIR and still trying to figure all these things out? It's not surprising to me that that's the case, but honestly with correlations coming down and with a lot of these issues starting to get sorted out in financial markets and in the financial industry as a whole, we are seeing again that the strategy seems to be bouncing back, so that's why, for example I wrote Return of the Trend. I think that momentum, just like value, just like other sort of risk premia are time bearing, so hopefully we're back in that time again.
Niels: Now, in terms of drawdowns, you need to help me here because drawdowns, for a manager of course, clearly create some emotional roller coaster. We know that, and we learn to deal with them over time. I think one of the biggest challenges for us really are to get investors comfortable and help them through the emotional roller coaster, because what often happens is they tend to redeem or reduce their investment at the worst possible time. They almost become trend followers on trend followers, meaning that they buy high and sell low, which is not a good strategy, so how do we educate them a bit, and how do we explain that a drawdown is not quite the same as a drawdown in an equity market, where there's kind of an open, and the risk - it could go to zero like what we have seen with some stocks. How do we explain that, do you think in a language that would make them comfortable about being in a drawdown and maybe even see it as a buying opportunity?
Katy: So we actually... this is a really important question, and it's actually one of the chapters of our book as well, it's called Dynamic Allocation to Trend Following. What's interesting about this is that if you take the concept that risk premia are time bearing, and there's some cyclicality, you should actually buy low and sell high, and investors tend to do the opposite. They buy high and sell low. This is because, if they understand... if we can explain better the divergent/convergent type of concept. If you have a strategy like trend following that over time, they actually tend to be mean reverting. The strategy over a long run is somewhat mean reverting. So momentum, actually, goes in waves, so if that's the case, then you need to sort of try and make sure that you actually buy in a mean reversion sense. So you shouldn't trend follow trend following; you should actually do the opposite. As an investor, you should buy low and sell high. So when I talk to certain investors who maybe are very familiar with this, what I see now is that in the last couple of months, some of them have started to say, well equity markets are at alltime highs historically in relative terms. That makes me concerned, so I'm going to readjust and start thinking about adding more CTAs and those that did that really profited. So those who were able to lock in some of their profits on the equity market and start thinking about alternatives, really, really did well. The intuition was pretty clear. Equity markets are at really major spectacular wins, what's the chances it's going to go that much higher? Those who I talked to who did that, they really sort of profited from that.