This is an edited transcript of our podcast episode with Nick Baltas, published on 19 April 2024. Nick Baltas is a managing director and head of R&D, cross-asset delta-one and commodity systematic trading strategies at Goldman Sachs. In this podcast, we discuss the difference between alpha, beta, smart beta, factors, the difference between a good backtest versus true risk premia, and why momentum makes money. While we have tried to make the transcript as accurate as possible, if you do notice any errors, let me know by email.
Summary
Bilal Hafeez: Welcome to Macro Hive conversations with Bilal Hafeez. Macro Hive uses natural and artificial intelligence to educate investors and provide actionable insights for all markets, from rates to FX to equities. For our latest insights, visit macrohive.com.
Before I start my conversation with this episode’s guests, I have three requests. First, please make sure to subscribe to this podcast show on Apple, Spotify or wherever you listen to podcasts. Leave some nice feedback and let your friends know about the show. For our latest insights and unlocked content, you can sign up for that at macrohive.com.
Finally, and my third request is that if you are a professional or institutional investor, do get in touch with me. We have a very high-octane research and analytics offering that includes access to our world class research team, trade ideas, AI models, and much, much more. You can email me at [email protected] or you can message me on Bloomberg for more details.
Now onto this episode’s guest, Nick Baltas. Nick is managing director and head of R&D, cross-asset, delta-one and commodity systematic trading strategies at Goldman Sachs. Prior to joining Goldmans in 2017, Nick was an executive director in the quant research unit of UBS, and before that he was a lecturer in finance at Imperial College Business School, a visiting lecturer at Queen Mary University of London, as well as a risk manager in a London based hedge fund.
Nick Baltas: The Origin Story
Now onto our conversation. Greetings and welcome Nick. It’s fantastic to have you on the podcast show.
Nick Baltas: Bilal, I’m so glad to be here and thank you so much for the invitation.
Bilal Hafeez: Great. Now, before we get into the meat of our conversation, I always like to ask my guests something about their origin story. So, can you tell me something about what you studied at university? Was it inevitable you’d end up in finance? And what are some of your career milestones up until your current role at Goldmans?
Nick Baltas: Yeah, for sure. Was it inevitable? I don’t know. Maybe it was a sequence of events or kind of a coincidence that I ended up being in finance. But I’m a computer engineer by training that turned into a financial economist or a quant – however you want to call it these days. And I guess my focus is to apply practical knowledge of math, stats optimizations, into the design of systematic strategies. And I try to partner with institutions, students, and investors across the globe to put together some solutions to solve specific investment problems.
Now, how did we go from a computer engineer to a quant, and having discussions with large pension funds and sovereign wealth funds? It was partly a path that I tried to follow, but I think it was also a coincidence that happened throughout that path. Effectively being an engineer and doing a PhD that somehow was related to the financial economic side but how we can look into engineering applied into finance.
And then came a time that I decided to move into the industry to apply some of my academic research into investment research. That was the time that I used to be at UBS. And then the more time went by writing about systematic strategies and quantitative investments, there came the structuring side of the world as to how you can convert those ideas into investable products.
And that’s how I moved into GS about seven years back, where basically my current role is to oversee some of our product suite on the systematic side, but also spending a lot of time with our clients, traveling globally to sit down with them and discuss.
So, milestones. Certainly, a PhD is a milestone.
Bilal Hafeez: It’s hard work, you know, it’s quite solitary work. It’s painful, but you did it!
Nick Baltas: Yeah, you have those evenings and the desperation moments of “what have I done with my life?”, you know, “why am I not progressing, my friends have real jobs and I’m still a student”. So that’s certainly a milestone, once you conclude that. But beyond that, I would say, genuinely, what keeps me motivated and excited about my day to day is just working with great people.
Is it a milestone? I don’t know. But to the extent that I can make some people’s lives more interesting and more engaging, and at the same time, I can learn from them and become a better person personally and professionally – I think that’s the most important thing. I’ll say that’s the milestone that I would just flag. And one I’m proud of.
Differentiating Between Terms
Bilal Hafeez: Fantastic. Yeah, I totally agree with that as well. Now you talk about systematic strategies and there’s this whole sort of debate around what’s alpha? What’s beta? Then you have smart beta, then you have factors. How do you think about all of these things? What’s these terms? Well, number one, what do you think about these terms? Are they the correct terms to use? And how do you make the distinction between them all?
Nick Baltas: So, I mean, a very nice question. Very good question. Let’s start from alpha. I think when people talk about alpha, there’s two notions here. There’s, the more subtle notion of alpha, which is basically investment skill. So oh, this manager has alpha, or like, this portfolio requires some sort of alpha to be improved in some respect. So, alpha in some notion is nothing more than skill investment. Skill in picking the best investment; being able to turn the portfolio around before the market regime shifts. And somehow, we associate skill with alpha.
But there’s also a quantification of alpha – which is nothing more than what we call a linear regression, that you have a return series that you try to project on a number of other return series and understand how much of the return you’re getting is, “exposure” to those other returns or factors.
I’m going to come back to that point, and how much of that is unexplained. So, if you cannot explain your return series by things that you would expect, that would explain your series, that’s basically your value add. So now there’s the link between the more subtle notion of alpha’s skill versus the quantified version of alpha being the residual of what I can explain. So what I cannot explain, to the extent this is positive, it’s the additional division I’m bringing beyond the stuff that I can explain.
Now, I think people mix up those two terms, and it’s not necessarily bad, but in some respect, somebody might talk about the alpha out of a regression, and the other one might talk about alpha in a more nuanced way. So that’s the first point I would flag.
Now beta, to your point, is the exposure to the stuff that I can explain. So, if I can explain that my return is coming from exposure to emerging market currencies, then my exposure to that is the beta. By the way, alpha and beta are the first two letters of the Greek alphabet that I happened to get to know from the age of five or six.
Bilal Hafeez: You’re Greek, aren’t you, Nick?
Nick Baltas: Exactly. Exactly. So people talk about alpha and beta and gamma and all those Greek letters, even in the option space. So alpha and beta, at the end of the day, were just two letters that have been put in place into a regression. And just purely by the fact that they were letters, and they associate a mathematical context of what I can explain and cannot explain, they were transferred into the investment world as the exposure to the explained and the unexplained – so beta and alpha in some respect.
Now, smart beta – I’m not necessarily sure whether there’s any dumb beta in place – but smart beta was this notion of: I’m performing – not because I’m getting exposed to the market, but because I’m getting exposed to other factors or other type of exposures, like emergency market risk or momentum investing, or it can be a variety of ways to explain an investment, but somehow that’s smarter than just buying the market.
So, the term came from the fact that buying the equity market is a good investment in itself. You’re getting compensated by taking on growth risk. But there are smarter ways you can be exposed to the market. Maybe you allocate more into smaller cap names because they have more ability to outperform, or maybe you allocate into a few more positive momentum single stock names.
Bilal Hafeez: And those are typically called factor models. There’s a jargon of the industry here.
Nick Baltas: It’s jargon, right. Because eventually you basically say, well, if I don’t hold the market, I hold something else. But you know, this something else can still explain my performance without that being completely unexplained – so it’s not like the alpha thing – it’s like it’s another beta to something else, which is not the market. And this is what we call factor.
Now, factor itself is our attempt to explain beyond the market. Now, explaining risk is different to explaining performance. And the fact that my risk might come from emerging market exposure or small cap exposure, or interest rate exposure is statement number one.
Statement number two is that if those factors and those exposures that I have inherently risky can be rewarded in the longer term with some positive return, then the risk factor also becomes a risk premium – so people talk about risk premium and factors almost interchangeably, but looking into the detail, it doesn’t necessarily mean that a factor that can explain performance at the same time should be rewarded.
One example here is, for example, counter risk. Counter risk in itself is not rewarded, but it’s certainly a way for us to explain where my risk lies. I have emerging market risk, or European deals or U.S. deals in my exposures. Summarising this whole context of what is return: return is a combination of exposures to investment themes.
In explaining risk, these become risk factors. To the extent that some of those risk factors capture some return reward, they also become risk premium. My exposure to those factors and/or risk premia are the beta components in a regression framework. Whatever is left out of it is the alpha.
And the way that academic literature has played out over the years is to try to explain the unexplained by looking into more and more and more factors. I’m pretty sure, the audience has seen the factors, or literature, maybe read about them, there’s like 500 factors and the attempt is to make that alpha zero. Because if that alpha becomes zero, then the model is correct in some respect. Otherwise, it’s incorrect.
How to Handle Data Mining
Bilal Hafeez: On that factor zoo, I look at a lot of academic literature, and every day there’s some new paper saying there’s this new factor that explains this or that. And this begs the question of data mining, versus true risk premia. Is there a theory behind these sorts of things? You’ve spent your life looking at risk premier, systematic returns, and obviously you want to make sure that it’s not just the backtest; that it will perform out of sample.
So how do you think about the data mining side? Because the returns may look good historically, and then you could say, okay, that’s rewarding me for some risk, but maybe there’s no rationale for it and you just happen to capture a period that had strong returns, but there’s no reason, theoretically or otherwise, for it to continue to behave well afterwards.
Nick Baltas: You’re very right. I think data mining is the largest scene in financial investments, specifically on the quantitative side. There’s this thing that people talk about, Instagram versus reality. I tend to talk about statistical fluke versus reality. And that is nothing more than this popular saying that says ‘if you torture the data enough, it will confess to anything’.
So the way that empirical research is typically conducted is in two ways: Either you start from a theory to test its existence, or lack thereof, using data. Or you start with data, you somehow come up with a pattern. And then the human brain has this tendency of trying to associate and explain empirical observations.
Now, the second part of my statement is the more dangerous one, whereby you run some analysis and you’re like ‘wow, that totally makes sense now’. I mean, because obviously this happened then, and then this happened. So, it clearly makes sense. But does it really?
So I think the way that we should be looking into empirical research should always start with some attempt to have a hypothesis as to why a turn should be there in the first place. Is it some market segmentation? Is there a reason whereby some underlying risk that exists in financial markets, and there is a segment of the market that is willing to take on that risk and another segment that is not willing to take on that risk, and therefore there is a natural transfer of a premium?
The fact that, for example, you have to pay for insurance is genuinely your rational decision not to be exposed in that particular risk. But hey, you’re actually paying for it. And whoever is willing to give that insurance over to you is the one benefiting from that premium. But they would be the ones to be paying out should the event occur. So, I guess another way of looking into the data mining concerns, the way I’m looking at it, in the way that also we discussed with the team and the way we build systematic strategies, is primarily a matter of culture beyond anything else.
I think it goes beyond the type of data we use, or the mathematical tools we use, or the optimization engines that we use. It genuinely goes beyond all of that, and it’s a matter of culture. So, people are smart, and people are well trained, and have done their PhDs, and they’ve done their university studies and so on and so forth. Once you train them and you have a conversation openly about, okay, now we’re actually data mining, it starts becoming part of the day to day.
I don’t think there’s an easy way to avoid data mining because guess what? There’s only one history. And of course, we can simulate the history as many times as we like, but there’s only one history we can actually test. So, kind of concluding all that – data mining is a sin. And it’s recognised as one of the key issues when empirical research is to be conducted.
I think it’s a combination of understanding financial markets and researching financial markets, forming hypothesis, trying to the extent possible to test those hypotheses still with the same data. But also, to be cognizant of the fact that any little gimmick we would do into any particular investment that historically has had a positive or otherwise effect, is still to a certain extent exposed to some data mining concerns. So, we know we have to call it out and making sure that at least we’re aware of it.
I’ll give you one example. We wrote some report about six, seven years back on mean reversion strategies. This was published into an academic journal. And recently we started looking into designing that strategy. Now, there’s a claim to be made here that back in 2017, we hadn’t seen the subsequent six years. So somehow there’s a claim that says, look, it’s been performing, so whatever we actually studied back then, it actually makes sense and so on and so forth.
But I’m always honest to say, look, if it had lost 30% since 2017, I doubt we would have launched it as a strategy. So, it’s not straightforward, but it’s certainly something.
Bilal Hafeez: Yeah, but it’s important to be clear about the culture you have and be transparent and honest and show your methodology, show your philosophy.
Explaining Momentum
Now let’s talk about one factor in that context, the momentum sort of strategies. So, momentum is a popular type of strategy where you follow the trend in some form, or past prices give you some information about future prices, which violates all versions of proficient markets hypothesis.
So, it seems to go against theory, what we would think is theory. Yet over different periods of time, over the very long run, it seems to deliver returns. And so, you have a strategy here that empirically seems to deliver returns, yet it violates one of the cornerstones of financial market theory. So how do you square that circle?
Nick Baltas: First of all, you’re absolutely right. Looking into past performance and predicting future performance goes genuinely against any notion of the official market hypothesis that basically says, whatever observed today has nothing to do with the past because it should be priced in. I guess for the bigger part of the listeners, they should be up to date with this notion of hypothesis.
The first attempt, or the biggest attempt of academic literature to explain momentum has been clearly through the behavioural side, and the recognition that maybe human beings are not rational in the way that they operate, and their decision making is fully loaded with cognitive biases. And there have been tons of research around underreaction to news or the slow diffusion of information into financial markets, or overconfidence to private information or subsequent overreaction.
So, to the extent that, and this is the claim here, that human beings and investors under react to the information flow and only gradually act upon it, that in itself inflicts into the price path this serial correlation. So, the empirical observation of the trend is nothing more than the exposition of the gradual digestion of information.
There’s also other effects, like the disposition effect, that says I’m selling my winners very quickly to crystallize my gains, but when a loss is happening, I don’t want to crystallize my loss, I’m going to stick to it, maybe double up on it. That in itself inflicts serial correlation.
So, the primary attempt in the literature to explain momentum, which to your point, and I definitely agree with it – it’s the longest standing empirical pattern we have, not only in single stocks, but across all the asset classes – is through the behavioural channel.
Now, not only that, interestingly, because of the way that momentum operates – which is a positive function of the demand of the assets. So, the more the price goes up, the more you want to buy. And the more you want to buy, the higher the price goes. And the higher the price goes, the more you want to buy. So, it’s like this self-fulfilling prophecy, this positive feedback loop. There are times specifically outside of equities that this self-fulfilling prophecy delivers what people call convexity and downside protection.
But before even going there, I guess the point that I would want to make to answer your question, is that it’s the behavioural side that have made people come to peace with the existence of momentum. And interestingly, it’s there and has been there. And this year has been one of the strongest single stock momentum quarters on the record.
Trend following in the more cross asset sense, has had a very, very strong year in 2022, helping substantially series capital allocation portfolios. So yeah, it is there. It has been recognized as one of the driving forces of systematic returns. But the puzzle remains. If somebody is not convinced by the behavioural explanation of momentum.
How Successful Are Momentum Strategies Really?
Bilal Hafeez: Now with momentum, there’s some asset classes like in the fixed income FX where momentum hasn’t worked as well; its returns have been poorer – is that a case of in the long run it will come back, the returns? Or is it somehow FX markets are more efficient, and that information diffusion is faster now?
Nick Baltas: It’s certainly true that, I guess before going into that point, just a subtle distinction – people talk about momentum in two different ways; There’s the so-called time series momentum, which is more like trend following. In other words, appreciating markets continue going up, depreciating markets go down.
And there’s also the cross-sectional nature, which is more, I think, akin to single stock investing that says past winners outperform past losers. That does not necessarily mean that rising prices will continue going up, it is more the relative outperformance that continues.
Now, I guess to your point, looking into currency markets, this is more about the trend following side of it. And whether from a trend following standpoint, we observe historically good returns. It is true that primarily bonds, maybe equities on the downside, maybe commodity at inflationary regimes, have been the ones to have delivered historically stronger returns.
The effects, yes, it has been weaker, in all fairness, but there have been times that temporarily have added a lot of good performance for a cross asset trend follower. I think the prime example here is maybe around 2022, between the summer and November, with a very strong rally in the dollar. That was a strong contributor.
The way I would look at it, is that in isolation, no it’s not a standalone pattern that delivers statistically significant returns. To the extent it doesn’t bleed outside of the times that it actually helps, it’s still certainly a positive contributor from a diversification standpoint to a cross asset construction. But it’s fair to say that it has not, in isolation, delivered returns similar to things like carry, for instance.
The Best Ways to Implement Momentum Strategies
Bilal Hafeez: More generally with some version of momentum or trend following strategies. You do have these characteristics where you do have this bleed, as you call it; day to day, you might lose money, and every now and then you get these big, big gains as you capture some kind of big moves. Are there some clever ways to get rid of that bleed and modify the strategy somehow? Has there been advances on that side or is that just overkill and you’re just data mining at that point?
Nick Baltas: I think there have been some attempts, and again, data mining is always the concern here. There have been some attempts. I would look at them quite differently when we talk about cross sectional momentum like in single stocks, versus more like cross asset trend following.
When you look into, let’s start from equities, buying your winners, selling your losers. Year to date, for example, a very strong performance, primarily coming from the artificial intelligence craze – the magnificent seven, however you want to call it. The winners keep on winning, and therefore the overall momentum in single stocks is outperforming very, very strongly.
Now, that type of momentum in the single stock world is occasionally suffering from crashes. And those crashes would happen if the winners lose, or more likely than not, the losers pick up. The losers play the catch-up game. So those that are lagging behind and you’re short, start outperforming.
Now to improve the performance or maybe reduce the probability of a crash, there are two ways somebody can look at it; One way is to reduce the exposure when the probability of a crash increases. It’s hard to claim that we can foresee a crash in single stock momentum, but can be associated, for example, with funding liquidity. And when that becomes scarce, you might end up having a situation that some hedge fund would want to reduce the exposure, might be a short squeeze. So that could be one way to look into dynamically scaling the exposure.
The more empirical or easier way was to volatility scale your exposure. So, the more volatile single stock momentum becomes, you lower your exposure. And that’s more like becoming part of the party, but not actually being 100% part of the party in some respect.
But also, some recent research suggests that better stock selection in single stock momentum can also allow you to build a more, I guess, immune portfolio to crashes. I’ll give you an example. There’s this behavioural bias whereby investors always compare, even mentally, where the stock price is today to the highest point over the last one year. And the further away you are, the more the temptation exists that there is room to get to if the price starts rallying.
So having single stocks that are far away from that and having losers that are far away from that level, makes the portfolio more risky because a rebound towards that level might make the portfolio perform worse. So there’s a filter that somebody can deploy in single stock momentum to move away from the stocks that are more likely to create a crash.
Now, if you look into the trend following camp, which is a completely different story, and I guess it’s even more related to your question, and interestingly, a pension fund asked me this question, ten days ago, they said ‘I like trend following because it kind of helps me with the downside, diversifies equity downturns, but it kind of does a job every five years or so. So, what do we do in between?’
I guess the point is, if it doesn’t bleed too much, it’s not bad to have a defensive component there in the first place. Every option based or contractual hedge is expensive. But beyond that, why would trend following not do well?
Well, two reasons. Either you have a flash correction or a flask reversion in the market, or the prices are kind of whip sewing. They don’t really move that much. So, the way to diversify the exposure in trend following without cannibalizing its ability to perform would be to add some reversion component in the design that somehow tries to capture a reversion in the prices, but also get some carry engine at the times that things are not moving.
So, if things do not move, you just need some carry to keep on improving your overall performance. So that’s how we look at it. In the single stock world, it’s more about risk management. In the latter, it’s more about complementing the strategy with other return streams.
Defining Mean Reversion
Bilal Hafeez: That’s great. Now you mentioned mean reversion. So, tell me more about mean reversion and how does it perform as a strategy? Is it popular or not? Which asset classes does it work well in?
Nick Baltas: So mean reversion, I think investment practitioners have been struggling to design a good mean reversion strategy. And that’s for a reason, because momentum, more likely than not, does the job. So how do you define mean reversion? Do we define mean reversion on a unaverred basis. It’s hard. So, the way that we think at least of mean reversion is more like on a cross sectional basis.
Ultimately, if you have large dispersion of price movements, there is this temptation that the strong – not necessarily outperformance – but those that are spiking up tend to revert back, and those that are tanking tend to recover. We’ve been using, for instance, the realized skewness, which is a measure of asymmetry.
But that’s not too dissimilar to how, for example, value works in single stocks. Value is nothing more than a reversion strategy. At the end of the day, you buy the cheap and you expect them to go up and you sell the expensive and you expect them to go down. But why did the stock become cheap?
Well, there’s two reasons for it. Either the book price went high, or the price dropped. But the research has shown that more than 90% of that activity comes from the price. So, the price fell rather than the book to price or like any fundamental value of valuation, went up. And the opposite I guess happens for the expensive stock.
So how do we look into reversion? To directly answer your question is purely as a cross sectional strategy, purely as a convergence strategy. And I think the challenge is to identify or measure the reversion part. Is it a longer-term return? Is it some form of comparison with fundamental value? Maybe in single stocks but not outside of it? What is fundamental value in commodities? It’s hard.
So, we look into more price-based data. I mentioned realized skewness as one of those that we have researched upon, published upon, and also utilized as a reversion signal.
So, I guess the last point I would say here is that it’s not a hedge for trend following strategies, it’s a diversifier. Trend following in itself captures principal component moves like I’m buying equities when equities are going up, but within the equities universe, I can still look into a cross sectional strategy. That is kind of adding some additional return or factor, if you were to kind of use the terminology we used earlier on. So that’s how we look into reversing strategies.
Effective Value Strategies
Bilal Hafeez: And when people say inequities have been underperforming for decades or over the last 10-15 years, and lots of quant funds have suffered as a result – and they keep saying it’s going to come back. Within your framework of looking at mean reversion, how do you approach that question?
Nick Baltas: I guess two reasons or two possibilities. Either we’re measuring value incorrectly. And there has been a lot of attempt over the last few years to revamp the way that we think about value – is it now in the period of intangibles, a cash flow based measure that becomes more important?
So, there’s one element that says value is not underperforming, period, but it’s actually underperforming in that definition, and I should just go back to the drawing board and redefine the way I’m measuring cheapness or expensiveness. But there’s also this other side of it, which is I’ve actually managed to capture a valuation right, but the regime is not accommodative of value investing in the present moment.
And I think consensus is somewhere between the two. The nonprofitable tech and the growth dynamics, and the post Covid rallies that we have seen in the high-tech world, alongside the fact that intangibles have become more relevant and therefore more cash flow based measures should be used. But even in that alternative definition, it’s not technically value skyrocketing. So, I think it’s a combination between those two, at least in my reading.
How to Approach Carry
Bilal Hafeez: Yeah. And then of course carry is the other big one that people like to follow. Momentum and carry, I would say, have got the core that everybody follows in some way or another. With carry, how do you like to implement that? And which asset classes does it work well in?
Nick Baltas: Similarly, I would look at it from the perspective of how do we design it? So, I look into carry also in a cross sectional manner. The reason for it is that I like to look into different sources of return in ways – not only in isolation, but also in combination, would give us a portfolio that is well behaved and neither cannibalize an existing return stream nor doubling on it.
The last thing I would want is to say I’m buying bonds because they’re appreciating in price in my trend following system, but I’m also buying bonds because they have no upward yield curves. So, do I actually do momentum? Do I do carry? Do I do a combination of the two? Or do I actually maximize my concentration into bonds? And then I have a significant exposure in the turning point.
So the way I kind of like to look into all those different premier, is maybe from a more kind of risk factor model perspective without explicitly doing so. And we spoke about factors earlier on in the podcast. The more factors we bring, the more decorrelation we would want to achieve in the overall construction. If you add the same factor again and again and again and again, the model becomes mis specified in the first place.
You wouldn’t run CAPM with S&P and FTSE and MSCI world, right? You just need one of them and then the excess return of any other market to S&P or to MSCI. So, my attempt to look into carry is to say, forget about directional moves. Try to lock in the differential in the yield, in currencies, in commodities, in interest rates, always go long half, bottom half, always try to find a way to rank markets. And obviously measuring carry is not straightforward, but subject to having a measure, look into a cross sectional definition of it.
Same thing like reversion, and then allow trend following to capture those conditional directional moves.
Bilal Hafeez: And in terms of measuring carry, in effect it’s fairly easy. You can just rank by short term yields and then for bond markets, would you look at the curve, like a kind of roll down yield?
Nick Baltas: Correct. But I guess the point here is the following – When does carry work, when things don’t move? Now you cannot force markets not to move, but you can do the best you can to reduce your exposure to directional market movements from the design standpoint. Now, the cross-sectionality allows you to achieve that in some respect, but also the signal itself can to some extent be informed of the possibility of a market move.
So let’s suppose that inflation is significantly higher than short term rates. What’s the danger here? Well, the fact is that you buy a bond and you’re getting hit by a yield increase. So somehow trying to intersect inflationary and economics variables in the carry definition is one way to protect yourself from buying what seemingly looks like a good carry opportunity. But it’s actually much more exposed to a potential spot market move.
Predicting Drawdowns
Bilal Hafeez: In that case, are you going down the path of trying to predict drawdowns? Or is it different would you say?
Nick Baltas: It’s more like avoiding the possibility of getting into a market that is more likely to experience a spot move.
Bilal Hafeez: Understood.
Nick Baltas: It’s almost like, eventually it looks like a real yield definition. It brings some element of valuation. If you look into commodities, carry is nothing more than the slope of the curve. But how about seasonal commodities? How about the fact that you observe very high price for natural gas over the winter? That shows in a term structure perspective, I guess a contango dynamic. You would want to go short that market.
But we should acknowledge the fact that seasonality does not suggest that things will stay as they are. It actually prices in a spot move, which is that price is going to go up. It’s the front that is going to move during wintertime. So if you just look blindingly into that curve. You say I’m going to go short in that case. As I enter winter, that’s the worst trade to do because it’s the spot that more likely than not, assuming winter realizes in the way that it’s expected to realize, that the price is going to go up.
So there has to be some seasonality acknowledgement and therefore accounting for it when we designed and measure carry
So, it’s those nuances in their respective markets that to the extent there is a downside exposure inherent either in the measurement or in the design, we should do the best we can to reduce that exposure.
Bilal Hafeez: And in terms of trying to predict drawdowns, I mean, many people have tried to do these risk aversion models, all these sorts of different things.
Nick Baltas: I mean, it’s hard. The reason why it’s cross sectional, it’s precisely to reduce the directionality. If you do, FX carry, you said it yourself, straightforward higher yielding currency versus lower yielding currency, that should perform. What is the underlying risk? Some sort of growth risk. But there is a dollar movement inherent in any portfolio. By building this RV, to some extent, you’re reducing the dollar exposure. So principal component number one.
Number two could be exposure to global growth shocks. So maybe there is some control to equity market moves in addition to the dollar neutrality. So, all those gimmicks in the design are here, at least in my view, to achieve one philosophical goal; reduce the exposure to spot moves, because there is no way we can predict them. And therefore, if they happen on average, they’d rather just cancel out in the design. And if they’re directional, by the way, let trend following deal with that. But not carry.
Identifying Macro Regimes
Bilal Hafeez: In terms of regimes, then let’s talk about macro regimes. And this has been a very popular topic, where people have said, okay, we’re in an inflation regime. And when we say that, does that mean that, that’s really a discretionary view and it’s a discretionary manager that should decide that? Or is there a role for systematic strategies, where you say something about, based on this regime, these systematic strategies will go well?
The other macro topic people are talking about is recession, hard landing recession. So how do you think about macro regimes and systematic strategies?
Nick Baltas: The identification of macro regimes is not empirically easy. We can talk about growth and inflation, but maybe after the fact. So there has been an attempt to obviously build now casting indicators in the industry which basically says, let me follow prices, production, new orders, credit spreads, financial and real economy activity, for us to have a gauge on to what growth suggests to be momentarily, and that maybe can allow us to classify time periods into economic regimes.
Because to the extent I can have a regime classification, maybe I can have a systematic strategy that can perform that particular regime. Maybe I do more carry if I’m in a medium growth, not inflationary regime. But if its highly inflationary, maybe I should allocate more into the commodity landscape.
Inflation itself is hard not only to measure, but also to decide upon its own regime. People have looked into levels of inflation. I think that’s half of the road. Others have looked into changes of inflation measures, and that’s another half of the road. Because going from zero to one, it’s not necessarily bad news for equities, but going from two to three, its actually bad news for equities.
So I think even defining inflation regimes would require a bit more work towards what’s the problematic level of inflation, but also what is the change in that inflation? And is it like the realized inflation or like the implied by the breakeven? Or do I use, a forecast based inflation, like a breakeven inflation, like a model based inflation. Is it like a CPI or CP or core inflation. So there’s so much choice.
Bilal Hafeez: I guess my question kind of presupposed that we know what regime we’re in, but you’re saying that in itself is the problem. We don’t know what regime – how do you measure the regime? There’s choices to be made, methodologies to be decided on. That’s step one, almost.
How do Macro Regimes Affect Systematic Strategies?
Nick Baltas: That’s step one. And step number two is then associating the regime with a systematic strategy. So, it’s almost like now doubling on the challenge. But I would not say it’s impossible. And we’ve done substantial amount of work for some of our institutional clients that want to have regime targeted portfolios on inflation growth. But even historically, the data is not rich enough in the sense that for like four decades we’ve had the Goldilocks.
So, you’re even trying to prove a point, but statistically speaking, you don’t have data, you don’t have the regimes that could allow you to do so. Maybe there is some sort of mechanical connection. For example, the fact that commodities move together with inflation, specifically, if there is any growth catalyst to it, it’s not a surprise, because at the end of the day, part of inflation is commodity prices, so, there’s this mechanical link between the two.
But certainly, there is this underlying risk of association of systematic themes with regimes, let alone the definition of the regimes themselves. I guess the last thing I would say is that we are using some growth indicators in some of our strategies. It’s been screaming recession for like a year now. So, guess what?
Bilal Hafeez: I’ve built lots of models over time predicting recessions, yield curve based models, this model, that model, and they all haven’t worked.
Nick Baltas: Exactly. I’m not against it. I think there is value looking into economic regimes, specifically because sometimes the market is turning faster than the prices do. And therefore, there is a statement to be made with regards to I’m positioning my asset allocation, I’m positioning my portfolio as a function of those implied movements, should the regime shift is correct.
But I think a combination of price-based data with economic regimes is maybe a better way forward, rather than just purely economic regimes.
Selecting the Right Systematic Strategy
Bilal Hafeez: Okay, now, when I look at the investment community broadly, I’m kind of exaggerating here, but you kind of seem to have the quant funds or systematic funds. So, either they could be the ultra geeky quant funds, high frequency, or then in the traditional asset management space, what you tend to find is you have a big asset management company. They have one fund, discretionary fund, headed by some big manager, and then they have another fund, which is their systematic funds.
So they brand them separately. That’s the systematic fund over there. And then that’s the discretionary fund. And you kind of have the two running separate from each other.
When it comes to all the strategies we’ve talked about here, is this a case of, it’s those systematic funds that you’re talking to, or would use this type of research? Or are there discretionary managers that will take on systematic strategies? So that’s with the fund managers. And then the higher level up is the pension funds, sovereign wealth funds, who are kind of the source of the money. How do they want to allocate to these sorts of strategies?
Nick Baltas: So, the strategies that we look at, and we build, are trying to systematize empirical patterns. So the success that we have really depends on how we can translate an empirical observation into a systematic program. Now, that in itself allows some of the asset managers to utilize the product as an access product.
So, if they want to do effects carry or trend following or maybe sell volatility and delta hedge it, that in itself, it’s nothing more than a source of return, that, to the extent it’s well designed – it allows them to do stuff that maybe a machine cannot do, which is, for instance, have a better understanding of the economic regime and not just looking into data. Or maybe it allows them to do dynamic allocation between those teams in a more efficient manner.
So, it’s not necessarily the case that we go one or the other. And we have seen our product and our strategies being utilized by asset managers and asset owners purely as vehicles, to express a particular investment objective in a transparent and liquid format.
Beyond that, we have seen our product being utilized by discretionary managers when they want to for example, outsource a hedging program. So maybe they run a discretionary equity model. They do long-short, based on their own views, very concentrated portfolios. But there’s an underlying growth exposure shock that they would want to hedge against.
Well, of course, they can buy put options on a kind of quasi regular basis, a six-month put option, and every six months buying a new put option. And that in itself is as manual as well as a path dependent process. Because you buy a put option today, for the next six months, you’re actually locked in on today’s strike price.
Or they can externalize that and say, well, how about I design a program, that systematically buys put options, delta hedges them, and therefore I have a convex profile that may become lower cost depending on the way that we design a systematic hedging program that they then just purely attach to their core mandate – that being, how can I best select the longs and the shorts from a single stock world in a highly concentrated manner? Ten names long, ten names short.
So that’s their main job. And there’s a solution to a potential need which is not their core skill set, neither the way that they would want to spend their time on. So, the utilization of our product is very broad and really changes on a client by client basis.
Strategic Versus Tactical Asset Allocation
Bilal Hafeez: How about something like strategic asset allocation then? I guess the example you gave could be used within that. But strategic allocation is that they set the high level – this is our allocation for the long run. At that level, that seems not systematic. It doesn’t seem like it will lend itself to systematic strategies.
Because you say, okay, we got 60/40 allocation, 60 to equities, 40 to bonds. Then the next level down is the TAA, the tactical asset management. And then you could incorporate some of these strategies that you’re talking about. But at the strategic asset allocation level, can you introduce systematic strategies or not?
Nick Baltas: If you were to ask me maybe three or four years ago, I’d tell you maybe not. Now, I think with the experience we’ve had post-Covid and the inflationary shocks that the economy went through, we have seen the reopening or actually the opening, I’m not sure if it was open in the past, but the opening of the dialogue around what the SAA is here to achieve.
What is the SAA in the first place? Why is there the anchoring of the 60/40? Is there anything more we could potentially do? Because the 60/40 portfolio at its core historically have achieved two plus one things. I’m very simplistic here, but basically buying equities and bonds historically have delivered, number one, compensation for taking on growth risk. Number two, compensation for holding your money for a longer term – that’s the term premium.
So the two more traditional sources of reward with obviously the underlying risks being very well acknowledged are here. But the plus one is the fact that the two were operating in this negative correlation dynamic, and therefore one was diversifying, the other making the combo the sweet spot of financial investments for big asset owners.
Now, not only have we seen those shifts in the correlation between the two playing out, but also we have seen them both failing very aggressively so in 2022. And that I think in itself increased awareness around inflation. And that in itself led to more maybe commodity oriented investments that we would need. That in itself lends itself to a more systematic way of designing a commodity solution.
Yes, we can maybe get commodity exposure via a benchmark index, but how about green transition and the cap that is operating in that manner? Should we redesign our commodity portfolios for SAA purposes? And that’s something we’re actually doing because you get inflation, you get diversification of equities and bonds, but also you can achieve – temporarily that’s not a risk factor, but it’s a theme – you can temporarily, benefit from the price appreciation.
And that in itself moves even one step further, that says, how about trend following for instance? I was speaking with a private bank recently and they told me the fact that we like trend following is that it kind of moves quite opportunistically in an unbiased manner without us having to act upon it. The minute we decide to change our SAA, it will take us a number of weeks or months to go to the IEC and take the decision, so on and so forth, and the market has moved already. I’d rather have a component that in itself is a bit more opportunistic there.
So, to answer now your question, it was not the place that we had seen systematic strategies being utilized historically in the SAA space. It was primarily for the overlays to the SAA. It was part of the old allocation that sits next to the SAA. It’s not TAA either, but it’s more like the alternatives. Now, some components of the systematic toolkit, I guess humbly start making a few steps more towards the SAA.
The TAA is a bit more nuanced. I think the utilization of systematic themes in the TAA is primarily in the volatility selling space. It’s very hard to time most of the systematic themes. Maybe somebody would have a chance of crystallizing some gain by entering a volatility selling program when the levels of volatility are elevated. It’s no surprise that year to date, interest rate volatility is substantially higher than all the rest. And that has been a very popular trade. But that’s again the tactical entry into what otherwise is a strategic allocation harvesting of alterios premium.
Can We Use AI in Systematic Strategies?
Bilal Hafeez: Yeah, understood. I just want to round off this part of our conversation with innovation in the systematic space. We’re hearing a lot about AI. Everyone’s talking about AI everywhere. And you have a computer engineering background, I’ve just learned as well. So I mean, do you think we can use AI in systematic strategies or not?
Nick Baltas: It’s a big question and a big debate, because AI in itself, it can help us get unstructured data in order, and get information faster. But that in itself stays at the data collection level. So instead of me using past prices, I can use sentiment harvested or extracted from earnings calls or media news or tweets or whatever. And this is something that I believe the industry has used over the last ten years or so. That stays still at the data level.
One step further would be to say, how do I have a collection of investments and I have a black box deciding upon it, or maybe it’s a crystal box these days. And I think there’s a lot of work that is happening in the space for explainability and transparency, and it can be used there. Or maybe it’s a better way of forecasting liquidity and forecasting risk. And that in itself is more about risk management and execution and tradability.
I think the way that we can plug in AI engines and algorithms really differs throughout that chain of getting the data or allocating, or then managing the liquidity and the tradability and the risk of the portfolio. I’m not yet convinced that we can completely take a step back and let the engine operate throughout that process of designing a strategy.
I heard recently in a podcast, and I couldn’t relate more with it – if somebody were to tell us that a computer will decide upon a medicine, would you actually take it. I guess the answer is no. Somehow you’d always need a human eye taking a look at it. Maybe that’s not the right thing, but I’m just putting it out there, because it actually struck me quite much. So, can we actually trust a machine without any controls upon it to run a financial investment?
It’s not clear to me, but certainly I can see the values it can bring in a variety of those stages. I think segmenting the investment process into building blocks, data collection, asset allocation and tradability and risk management of the portfolio. I can see applications. And we have actually already started scratching the surface. That’s my view on this. By all means. It’s extremely helpful in terms of building platforms, prototyping. So, the whole coding part, over day to day, we have already started seeing significant help already.
Bilal Hafeez: Yeah, that’s great.
Nick Baltas: It’s a big debate. It’s a big question.
Investment Advice
Bilal Hafeez: I’m sure in the next six months or twelve months, we’ll kind of find other ways of using it. Now, I did want to round off with a couple of personal questions. One is, what’s the best investment advice you’ve received from anyone else?
Nick Baltas: I think the best one that I would give goes as follows, precisely because literature has identified so many behavioural biases on anything we do, let alone investment. And I think there’s always this temptation of human beings to act upon the recent information, the recency effect, and the fact that memory becomes very short when you’re under tension.
The best investment advice I’ve ever received is the following – just anytime you do an investment, take a pen, take a paper and just write down what you expect this thing to do. What are the scenarios you’re expected to do badly? What are the scenarios you’re expected to do well? And then the first time something bad happens, go back to that script. If you wrote that down, that’s fine. And that’s a recipe to go beyond the emotional state that you can enter during a bad period, investment wise.
So that’s maybe the best one. And it was actually an asset owner that gave me that advice years back when we were discussing about some investments.
Advice for Graduates
Bilal Hafeez: No, that’s great. I like that a lot. Now, the other question, we do have some younger people listening to this podcast. What advice would you give to younger people who are going to graduate and leave college this year?
Nick Baltas: The first one, and I’ve only realized that after the fact that I actually paid the price for it. There’s so much knowledge that we learned during graduate studies or like postgraduate studies that at that point in time we don’t feel will be relevant. But then the amount of experience we build and when it becomes relevant, it will become relevant, it’s just mind blowing.
I think the example that I would bring myself is linear algebra. I remember being like, why the hell do we need all those matrices, blah, blah, blah. Guess what? It’s my day to day now when we do like optimizations. So, there’s a lot of knowledge. Maybe it’s the way that this knowledge is transferred, or at least was transferred 20 years back when I was a student. I think that’s the first one.
The second one – and that’s, again, more like a personal experience. You know, some people come to me and they ask me about career advice. Now you’ve done your PhD, you’ve done, the masters and so on and so forth, you know, should you do the same thing? You know, is it good advice?
I want to be very clear on that. And I’m always clear when people ask me, do not look at the goal and just sacrifice anything in between. Like you said it yourself, like, doing a PhD is an investment in itself. It’s not the means to achieve a goal. You don’t spend, let alone waste three, four, five years of your life to go through those emotional states of like, did I do the right thing? What is the opportunity cost? And so on and so forth. So I would say just try to enjoy every step of the way.
Certainly there are goals to meet and means to achieve those, but I would not sacrifice everything in between just for the goal. That would be my advice. I hope it’s not super high level, but I think it’s a very, very important one.
Book Recommendations
Bilal Hafeez: Yeah, that’s fantastic advice. Now, I do like reading books as well. So I mean, what are some of the books that have influenced you? Whether it’s in finance or outside of finance, just any domain.
Nick Baltas: I’d say I kind of thought that through, because I was expecting that to come around. Right. I guess I can answer both sides of your questions. On the investment side, the books that influence me. The ones that influence me, certainly ‘Expected Returns’ by Antti Illmanen and ‘Efficiently Inefficient’ by Lasse Pedersen.
Then people would know, and you know, and I’m pretty sure most of the listeners, that obviously I have a passion for systematic trend following. So ‘Trend Following with Managed Futures’ by Katie Kaminski was the one that I read back in the day when I was finishing my PhD.
But also, I believe the book that I’ve read, and it was complicated enough, but at the same time I think, re-dialed my brain from an engineer into a finance practitioner, was on the stochastic calculus front. The ‘Stochastic Calculus for Finance’ by Shreve, maybe the one by Baxter and Rennie (‘Financial Calculus’).
So these were books that got me closer to the stochastics of financial processes. Beyond that, one of the novels I remember vividly is ‘The Alchemist’, but that’s a classic one.
I guess from a management standpoint, there’s a book that I recently read which I enjoyed. It’s called ‘Extreme Ownership’. It’s by two marines, Jocko Willink and Leif Babin. They have military experience, and somehow through the military process there are lessons for business and life. So that was basically anything you would do, and managing a team and managing a business and managing your own time and life and so on. There’s an element of ownership, and the way that you express that ownership reflects upon you, but also upon your people.
That was a good one. And I think the last one that I also mentioned in the podcast a couple of months ago is one by Matthew Walker, which is called ‘Why We Sleep’. And that’s more like on the science side.
That had a big impact on me, because it basically says how human brain redials every evening and how important sleep is, and us being creatures that evolved over the years in principle, should have not had the need for a sleep. If that was just a rest exercise, human species should have found a more efficient way through evolution to find relief to being tired, as opposed to losing a third of the day. So it’s not a lost investment. There’s way more that happens during that time.
Bilal Hafeez: That’s fantastic. That’s excellent. We’ve had a fantastic conversation. I’ve learned a lot, as I was expecting. Now finally, what’s the best way for people to connect with you, follow your work, if people wanted to understand more about what you’re doing?
Nick Baltas: By all means, I’m trying to stay connected with social media, specifically LinkedIn. I’m not posting a lot, but I’m actually posting things that I find interesting; Maybe some of the research that we do, maybe are some of the podcasts that I find interesting. So that’s certainly one place, but by all means contacting me there it’s straightforward.
My email is also there. I’m always getting people contacting me, and I’m trying my best to always respond to every single one of them. So, by all means.
Bilal Hafeez: I’ll include a link in the show notes. It’s fantastic speaking to you, Nick, and you’re doing really interesting work. And good luck in all the work that you’re doing.
Nick Baltas: Thanks, Bilal. And that was again, it’s been a pleasure. And thanks so much for having me here. I’ve been following the show for some time, as I told you in our private conversation earlier on. So, I’m glad to be part of your show. Thanks a lot for having me.
Bilal Hafeez: Thanks a lot, Nick. Thanks for listening to the episode. Please subscribe to the podcast show on Apple, Spotify or wherever you listen to podcasts. Leave a five-star rating, a nice comment, and let other people know about the show. We’d be very, very grateful. Finally, sign up for our free newsletter at macrohive.com. We’ll be back soon, so tune in then.
(The commentary contained in the above article does not constitute an offer or a solicitation, or a recommendation to implement or liquidate an investment or to carry out any other transaction. It should not be used as a basis for any investment decision or other decision. Any investment decision should be based on appropriate professional advice specific to your needs.)