Dan Jelicic - Mastering Factor Timing

00:00:06 [Dan Jelicic]
Stocks are over long-term persistent. they behave, there's positive autocorrelation, in other words, they trend. They trend, and we know that stock momentum is a very interesting factor that's often used. But over short-term horizons, so less than one month, stocks are anti-persistent. In other words, they revert. And price reversals is basically the models that will be used there.
00:00:30 [Vincent Weber]
Hello, everyone, and welcome to another episode of Resonanz Spotlight, the podcast where we explore investment strategies, learn the story behind them, and meet the experts who create them. I'm your host, Vincent, and today we are discussing a topic that has been generating a lot of interest among investment professionals, factor timing. Factor timing is about adjusting an investment portfolio to focus on certain factors like value, momentum or quality based on what's happening in the market right now. The goal is to improve returns by choosing the factors that are likely to perform best in the near future. But why is factor timing so important at the moment? The reason is a unique market condition we are seeing. Traditional investment methods are being challenged because some factors that used to be reliable are now less stable.
At the same time, the popularity of smart beta ETFs means that factor timing, once used only by a few quantitative managers, is now a common strategy for both big institutions and individual investors. This has led to an important debate. Can factor timing consistently deliver better returns? Or is it too risky in today's unpredictable market? Some people think that trying to time factors add too much complexity and risk, while others believe that with the right data and models, it can significantly improve portfolio performance. In today's episode, we'll explore these critical questions with our guest, Dan Jelicic, a seasoned fund manager and a pioneer in the field of factor timing.
By the end of this episode, we hope to provide a clear answer on whether factor timing is a valuable tool to your investment strategy, or if it's better to stick with a more traditional approach. Dan Jelicic is the Chief Investment Officer and Portfolio Manager of the Trium Avala Strategies. With over 20 years of experience in quant finance, Dan has made significant contributions to the field, with his strategies earning multiple awards. Dan's pioneer work began in 2002 at Sabre, where he launched one of Europe's first quant equity market neutral funds. Before Sabre, Dan designed and managed a European equity market neutral fund at ABN Amro and worked as a quant researcher at J.P. Morgan. So without further delay, let's welcome Dan Jelicic to the show. Dan, good to have you here today. Thank you, Vincent.
00:02:58 [Dan Jelicic]
Yes, delighted to be here and that you are addressing this very important topic that was of great significance to me.
00:03:05 [Vincent Weber]
Yeah, so Dan, let's start this episode by learning a bit more about you in particular. So what initially drew you to the world of finance and investment management?
00:03:16 [Dan Jelicic]
Yes, in early days, indeed, I was training to be an actuary and worked mainly in last liability modeling but there was a field that was of particular interest to me at that time which was asset liability modeling where i came across a first quant model actually which was modeling assets based on effectively inflation and one of the starting point i then realized that actually the asset side of things is a lot more interesting and I then took a little risk in terms of moving on to investments and got a master's in quantitative trading and finance. That was in the mid-90s. And after that, I got a job in J.P. Morgan Fund Management as a quant researcher.
I was lucky enough to dare address a number of multiple asset classes, but mainly my emphasis was in equities and currencies. And from there on, then I moved to ABN Amro and so on. And then, obviously moved on to see things that perhaps at that time were quite unusual, and maybe not seen by many others.
00:04:32 [Vincent Weber]
So were you already doing a quantitative job back then? I don't know if the term was already worded use. Would you have described yourself as a discretionary manager?
00:04:45 [Dan Jelicic]
I was always thinking in that direction in terms of, say, how to use numerical methods to find solutions, which computers obviously are of great help there. But when I moved on to asset management in J.P. Morgan, J.P. Morgan was indeed using quantitative tools, and I was part of the quant group, research group, which was developing tools for them.
00:05:07 [Vincent Weber]
You've been pioneering the de facto timing or what you used to call it, maybe the style arbitrage approach. So how did you get there? Was there a particular moment, the tipping points, or was it a gradual evolution?
00:05:25 [Dan Jelicic]
Yes. From JP Morgan, I moved to Airbnb Amaro Asset Management. And there I had two roles. One was strategic advisory, which was based on my actual skills, helping them producing models for long-term pension funds, advising directly the CIO. But the other one was Alternatives Group, where I was managing, sorry, when I was actually doing product design. So one of my first roles was, okay, ABN AMRO had very good internal analysts, and essentially the main thing I did was to produce a value overlay, which was In a sense, I realized that because of a lot of persistence in value factor, I could switch that off and on with a lot of confidence in being successful in doing so. And that was the product I designed. So the next step was who's going to manage this product?
And the CIO offered me to manage it. And that's where I said yes. And I started managing this product. This started in June 2020. And it turned out that. It was after the major inflection points from growth to value. It made me realize, A, that factors can be timed, and B, that with a little bit of more quantitative screening, you can actually input an awful lot of other factors and build my own products, which is basically how I went to Saba. Saba wanted somebody to run factors. They already had a quite successful startup product, and they wanted somebody to run factors for them. And I started Style Arbitrage in 2002.
So it's very interesting because, firstly, because of the topic of the organization and also because of some recent academic papers that have been written calling them, on factor momentum. So, yeah, so that's how we, that's how we started back then.
00:07:22 [Vincent Weber]
Okay, great. Before we dive deeper into
00:07:27 [Dan Jelicic]
really about factor timing maybe just taking a step back can you just describe for us the idea behind your style factor performance persistence and yes of course okay let's go back to the basics here okay what are the factors factors essentially are groups of stocks built around certain attributes that these stocks have in common. For example, book to price. So what we do is we construct effectively long-short portfolios where we go long cheap stocks and short expensive stocks. So many, many, many multiple factors can be built around many, many other different attributes like this. And we can then look at the behavior of these factors and behavior of stocks and what is the difference and what are the commonalities. So we look at this.
Usually, I draw a quadrant with short-term, long-term horizon and stocks and factors, but I can describe that quite easily here. Basically, stocks are over long-term persistent. They behave, there's positive autocorrelation, in other words, they trend. They trend, and we know that stock momentum is a very interesting factor that's often used. Over short-term horizon, so less than one month, stocks are anti-persistent. In other words, they revert. Price reversals is basically the models to be used there. And, in particular, a very lucrative area of market neutral investing is startup, which indeed is just focusing on short-term time horizons, sometimes intra-daytime horizons. Okay, yes. Overall, below one month, stocks definitely revert. And over three months and onwards, especially six months onwards, they definitely trend. Now, what is the situation in factors?
Now, factors, as I said, they're just groups of stocks. So you'd expect, OK, maybe they behave the same way, but they actually don't. Over the long term, yes, they do. OK, factor trend. But over short term or less than one month, factors also trend. In other words, there is this positive persistence, positive autocorrelation, even over the period of less than one month. OK, now that was very fascinating to me back then and still fascinating to me now.
now why is this okay so and there is enough explanation even in the literature back then what's what's going on why do we have trends well some people have impacted information and the others created start the trend the others get information later jump on to this again and then so on and so on until people buy just because they've seen something's been going up and you basically get these herd effects So this is very much true for most assets, in fact, if you look at Bitcoin or something like that, which trades purely on sentiment. But in terms of stock factors, definitely is the case, but not for individual stocks.
00:10:28 [Vincent Weber]
So what's your approach to time factors?
00:10:32 [Dan Jelicic]
Okay, so because most factors, but not all, most factors have positive autocorrelation. The techniques to exploit this basically, make money out of positive autocorrelation. So what are these techniques? Most obvious one is econometric models, like a REMA model, for example, but also trend following models, which are commonly used in currency. Remember, I mentioned that I've seen how currency traders exploit trends in currencies. And furthermore, I met a lot of interesting people who had a very good idea about that. And that's basically what we're doing. We're looking at Back in 2002, we started effectively looking at trend funneling models such as MACD, moving average convergence divergence models. And I'm pleased to say that is still being used.
We have, of course, expanded over the period of years, adding a number of proprietary models that best capture the time series as well as cross-sectional properties of factors. this original approach is still here. And I think that's still working very well. So I think what's also important is very careful implementation. one has to be very careful and not overdo it. you can find that you're always, by necessarily overweighting certain factor more and more and more, you're going towards more concentration. So you don't want to lose the benefit of diversification at the same time.
So you always, it's a trade-off between adding alpha bet making factors better and losing a little bit of diversification so that's why this complexity people maybe get over concern about the complexity of doing this because you really need to know what you're doing and what are the real issues with this okay of course the complexity i mentioned to some extent a little bit earlier in the sense that It's a cross-sectional issue. It's not just a single-factor issue. You have to be careful in blending this with respect to not losing the effects of diversification. That's one point. Some obvious points are also there in terms of cost of implementation.
I did, over the years, come across a number of papers, and I'm sure academics have been noticing and have been producing papers saying that, yes, there is. positive autocorrelation but it's hard to exploit it because of the costs of doing that because of course you need to trade a lot of stocks right to do this okay and i've seen some large assumptions for costs in in academic papers so for people like us seasoned investors who've been around the costs are much much much lower but of course it's true if you have a 10 billion plus portfolio right you will be struggling to to perhaps optimally do this in terms of what you would be optimally fine at one billion portfolio okay because it does it does take some trading and of course if you're going
to move the market the other people are going to front run you okay it's as simple as that it's also these effects are probably getting stronger down the size curve so i say it's probably a little bit stronger at mid cups rather than large caps. So you can't just say, oh, well, I'm only going to do it for the top 100 stocks because, of course, you can do that. But unfortunately, the effects are not as strong as for the lower 1,000 stocks. So there are these costs that are actually a little bit difficult. But even if you're 10 billion, you can still probably do it somewhat, but at a smaller scale. Other issues that can be a problem. is overtiming. I mentioned it earlier, but I'll say it again because it's very important.
People can see the autocorrelation, but if they go too much into this area, there is a risk of too much concentration and actually not helping the overall Sharpe ratio. And a final point, I should say, is the inflection point. That has always been the case when basically momentum breaks down. Of course, if you are now overemphasizing value, value, value, and value breaks down, you will be losing more money than if you just had a static exposure to value. So how do you protect against that? That, I think, is a big, big problem potentially for investors. And I'm not just talking about factor timing. So far, we've been talking about factor timing based on persistence, which I'd say would be one of the main.
for factor timing, but I've seen a lot of other factor timing models. You can, for example, use valuation spread. You can say quality looks very cheap at the moment, so we should overemphasize quality. You can also look at volatility of factors. You can say I'm going to emphasize the low vol compared to high vol. So, so overall, either one of these cases, you can find in the end of the day, you have this potential inflection point which is a risk. So overall, yeah, I would say these are the main pitfalls.
00:15:46 [Vincent Weber]
Right. Yeah. So you mentioned a lot of quant techniques, when you talk about econometrics and your actuarial background and covariance metrics, but I haven't heard the words machine learning or AI. So is there a particular reason for this?
00:16:08 [Dan Jelicic]
Not really. I mean, the question is a bread and butter question for me about exploiting effectively the herd effects in factors. And now we have a proliferation of factor strategies and even stronger herd effects. So I'm kind of getting quite excited because it's my pet topic and I know how to exploit it. But is machine learning helpful? Potentially, but I think we're not saying no. We're agnostic to machine learning, of course. And we indeed have some models that are looking at stock level machine learning stuff that exploit some nonlinear properties in stocks. But this is stuff that this is, I heard effects that are really best captured by the models that have been tried and tested over many years and enhanced over many years as well.
So we feel we have a strong understanding of what's going on there. Besides, in terms of AI, which is obviously making a huge impact. in other parts of the world, I'd say for quant managers, these machine learning techniques have been around for many, many years because we simply deal with data. I think the trick here is that people have managed to create data outside the financial world where there was no data before. Natural language processing for text and various other data. collection and so on for various other things. I mean, medicine, amazing, and so on. But for us, we've always been dealing with data. So we already have been dealing with numbers. So we already had that first step that people are finding out in different areas.
00:17:43 [Vincent Weber]
And what future trends do you foresee in quant equity investing, particularly concerning factor timing?
00:17:53 [Dan Jelicic]
It's very interesting. Going forward, okay. Just to use French summary of this. I think what we've seen over the years is essentially we're coming back to the same thing. People do not change behavior that easily. We can see that in the case of Bitcoin. What is Bitcoin? It's entirely sentiment driven. But the trends in Bitcoin are enormous. Just because people follow these trends, it's a herd effect. It's asymmetric information that I mentioned earlier. And what we are now seeing in factors is very much, the follow-up story on trends and people observing factor and effectively jumping on factors when they see it working and causing more hard effects. Even the big rotation people are talking about. It's very much now coming up.
When it's coming up, are they going to cut interest rates by how many cuts and so on? The big rotation, as they're talking about from large caps to breadth of the market to the diversification. And finally, opportunities for good opportunities for stock pickers to add value. That's actually an example of the factor rotation. It's an example of what we've seen since 2021 is massive outperformance in large caps for three years now. And what we expect to see once this turns, and it's starting to turn already, like it's kind of, we are now close to, but it's quite mild inflection point. We're going to likely to see a year of continuation of the performance of mid caps and small caps. So we need to see some evidence in terms of, of course.
it's kind of all of this is linked to the economic regimes okay so yes when interest rates are low it's good for small cups when these are high large cups but people jump on this before and cause this trend so it's very hard to actually use economic models to forecast this anyhow ultimately the understanding behind this But what we are now seeing is potentially the star rotation. Also, what I would say is the emphasis and the popularity now, factor timing, is possibly going to create even stronger trends. Now, what goes with stronger trends, of course, we might even see the inflection points that are even stronger, which is bad news, like when we have big reversals of these trends.
I should say that for us that's probably good news because i need to say a few things about inflection points inflection points are can be quite bad but in terms of what we do we effectively use the price reversal models earlier on i was talking about opportunities for stocks that do revert over a short period of time well if the factors revert as well over that quick turnaround then the effects for stocks are even stronger So because we exploit that already, at that time when we need it most, the price reversal model works at its best. So examples of that are, say, November 2020 or March 2009 or August 2007. I've been around for many years and I can remember all these things.
And in all of these inflection points, our price reversal models have done very well to protect. the fund from it and i think that's one way of doing it of course you need to run them simultaneously and again for very large managers it's very hard to do that because obviously it's a lot of trading as well to capture that anomaly as well but overall exciting times because we expect if anything stronger stronger factor momentum then thank you so much for sharing your insight and experiences with us today it's been a pleasure having you on the show thank you
00:21:49 [Vincent Weber]
really appreciate the opportunity to talk about my pet topic thanks and to our listener thank you for tuning in to another episode of resonance spotlight stay tuned for more expert insight in our upcoming episodes until next time i'm vincent and this is resonance spotlight