Why Macro Investing Is Becoming More Systematic ft. George Patterson

[00:00:01] George

You always have to ask yourself, you know, are the assumptions behind the model still valid? Has there been a regime shift in the data? That’s the kind of thing we worry about every day when we’re using models. We still see, like, an overall positive macro environment for risk assets but, in many ways, the US economy has been surprisingly resilient despite all of the challenges that have impacted it.

[00:00:27] Intro

Welcome to Top Traders Unplugged. In markets success doesn’t come from predicting what happens next, it comes from being prepared for what you can’t predict.

In each episode we go deep with some of the world’s most thoughtful minds in investing, economics, and beyond to understand how they think, how they prepare, and how they decide, and the experiences that shaped how they see the world. No noise, no short-cuts, just real conversations to help you think better and invest with confidence.

[00:01:01] Alan

Welcome back to Top Traders Unplugged. My name is Alan Dunne and today I’m delighted to be joined by George Patterson. George is a Managing Director and CIO at PGIM Quantitative Solutions. He oversees all portfolio management and research for quant equity and multi-asset teams. He’s been in the markets for many years now, previously was at Axioma identifying buy side trends and market opportunities. He was CIO for quant investments at Bank of Montreal Global Asset Management and earlier in his career was co-founder of Menta Capital, a quant equity hedge fund. His background is in physics. He has a PhD in physics and spent time at NASA earlier in his career. So, George, great to have you with us. Great to have a proper rocket scientist to chat to. How are you?

[00:01:53] George

Very good, thank you for a great opportunity to be here.

[00:01:55] Alan

Good stuff. Well, I gave our listeners kind of the highlights from your career. But I mean as I mentioned, you started off in physics and then you worked at NASA. What got you interested in that? And then how did you make the leap into finance?

[00:02:11] George

You know, I was always interested in physics from a very young age. It was just an area that I thought brought together math and, you know, science and there was a lot of technology involved in, you know, how you do experiments and it just, for me, was like the combination of many different fields. I was always very interested, I enjoyed it.

However, I knew I was not going to have a traditional academic career. I knew I wanted to do something in industry, and I didn’t really know what it was. And it’s, you know, kind of really by luck I found some colleagues that understood my skill set and what I could offer.

Most investment firms were not interested in hiring a physics PhD. But there was a group, it was at Barclays Global Investors, in the mid-90s, and they looked at me and they said, yeah, we think we can figure out what to do with you. And that was really where I cut my teeth.

[00:03:12] Alan

Interesting. And you think, I mean I come from an economics background so kind of view markets from that kind of perspective. And then I guess, people from a quant background, I mean, do you take a different lens when you’re looking at price movement, price returns in financial markets?

[00:03:31] George

So, I’m a strong believer in having a multi-disciplinary team. So, you know, I’d say when I was early in my career I was very focused on the pure quantitative aspects and, you know, what did the model say? But over the years I’ve come to realize that you really have to kind of combine many different views. Quantitative is a very useful tool but you also have to pay attention to what’s going on in the market.

So, really my goal, and particularly with my team, is to have people with economics and finance backgrounds but also computer science, mathematics, you know, different fields because people bring different insights to the problem. So that’s one of the things we try and combine at PGIM Quantitative Solutions.

[00:04:13] Alan

Very good. Well, give us a sense on PGIM Quantitative Solutions in terms of the size and scale. I know there’s kind of two elements to it. There’s a quant equity and the multi asset side. But in terms of kind of assets under management or the types of clients you deal with.

[00:04:28] George

Sure. So, we’re the quantitative solutions group within PGIM. We’re actually one of the early pioneers in this space. We just celebrated our 50 year anniversary and, actually, many of our track records go back 25, 30, even 40 years. So, we really were a pioneer in this space in many ways.

The business actually originally started with multi-asset and focusing on, you know, systematic applications of multi-asset which obviously has evolved over the years. That’s roughly 50% of our US$110 billion. And then quantitative equity, which is much more along the lines of, you know, benchmark relative strategies, you know, kind of typically with you know, tracking error anywhere between 2% and 4%.

[00:05:16] Alan

Good stuff. And, I mean, as part of your role you’re responsible for portfolio management, research, the full gamut of investment activities?

[00:05:28] George

Correct. Yeah. So, again, from my perspective it combines many different skill sets. There’s the investment side, obviously research – research is one of my passions, trading, again, all of these areas have really evolved quite a bit over the past 30 years or so that I’ve been involved in this industry.

And obviously technology plays a very different role. You really need to have them embedded in the investment team if you’re going to have a quantitative firm. I think, actually, for any firm these days, technology really needs to be embedded into the process given the rate of change.

[00:06:03] Alan

Yeah, so you mentioned you started off at BGI in the mid-‘90s and, you know, so that’s the same time I started off myself, but sobering to think it’s 30 years ago now, or whatever. You know, a lot of change in the market since then, as you say, in terms of the technology, market participants, market microstructure. I mean, when you reflect back, what do you think the macro investment landscape, how has it changed over that period?

[00:06:31] George

Well, there’s been a lot of change. Right? I mean, when I think about what the main themes of change are, is one is just data. In the early days we were scraping to find data, you know, people didn’t really understand the use. They were not thinking systematically. And it was also a time when… that was kind of the ‘90s which was really the era of the big macro traders, you know, like Soros, and Druckenmiller, and Robertson, and really, you know, making big bets.

But over the years, I mean, if you think about it today, we have GDP now, we have web scraping prices, you know, you can track behavior, there’s geospatial. I mean just the amount of… I mean, really, the rise of technology has just enabled us to have a much tighter view on what consumers are doing, which is obviously critical for understanding where markets are going.

And I think the consequence of that too is that, you know, the big opportunities that the macro concentrated traders had in the ‘90s really just don’t exist today. You have to be much more nimble. I mean, macro investing has become much more systematic than it was in the ‘90s.

[00:07:48] Alan

Yeah, I mean, I don’t think you heard about too many systematic macro traders back then. I think it’s fair to say it’s a product of more recent decades. I mean, obviously with that change in microstructure and availability of data, etc. have you seen an evolution into strategies over time as well?

[00:08:13] George

Of course. I would say there’s definitely been an evolution of the strategies in the sense that we’re able to get much more data that we’re able to proxy for things. At the end of the day we’re still, I’d say we’re still trying to think about what are the drivers that are going to impact markets, things like inflation, things like growth? But we’re able to get much better proxies to measure those characteristics.

So, you know, it used to be that you’d rely on government data that came out once a month or once a quarter. Now, with language processing, you can be monitoring news feeds, you can be looking at papers from across the globe, looking at foreign languages and monitoring news stories about inflation or what are people really focusing on.

So, I’d say the overall thesis that drives the investments is similar, but the way that we’re able to get data and proxy and measure those has just really exploded.

[00:09:12] Alan

Yeah, well, maybe just taking it into the current day, I mean, obviously you’re running multi asset solutions and, I guess, in the multi asset space, the 60/40 has been the benchmark for a very long period of time and worked very well for many decades, particularly in the 2010s. And then as we’ve come into this decade, you know, a theme has been the end of 60/40 and, you know, obviously a shifting macro backdrop.

How do you think about that? Is it a case of, you know, people talk about 60/20/20, or 60/30/… whatever it is, you know, is that the way or what are your thoughts?

[00:09:55] George

So, people have been talking about the demise of 60/40 for years. I mean, this is not something new and there are definitely periods of time when it doesn’t work. Like, if you look at ’94, or if you look at like 2022 where, again, you had sudden moves in the bond market that really caused bonds to not work as a defensive instrument. I think 60/40 is probably, on one hand, it’s still probably going to be around for a long period of time. However, there are many more ways these days to be able to get access to diversification that you just couldn’t get 30 years ago.

So, depending on whether you’re looking at… if you’re an institution, you can become much more involved in overlays or solutions that are going to give you a little bit of convexity or downside protection or, as you know, return stacking. If you’re a retail investor, there are a lot of products that you can get either, I’d say, in the category of derivative income and derivative protection, that are really not new strategies but they’ve been democratized from strategies that were usually only available through private banks.

[00:11:03] Alan

Yeah, and I mean is that, you know, if a client comes to you with a blank piece of paper looking to build out a multi asset solution, is kind of 60/40 still the starting point with some kind of overlay on top of that or how do you approach the problem?

[00:11:21] George

Well, it really depends on the client. I mean most of the time clients are coming to us with a problem. I mean, the solutions part of PGIM Quantitative Solutions means that most of our mandates are customized. So, most of the time a client, typically an institution, endowment, foundation, comes to us with a challenge such as, you know, I have this in my portfolio, how do I get back to my strategic alignment? You know, I’ve got a blend of publics and privates, what’s the best way to solve this particular problem?

And a lot of times we’re really, you know, thinking across the board, not just 60/40, but how do we get there? Sometimes with derivatives, sometimes with options, you know, looking to solve those problems really on a much more of a customized basis. Rarely do we get somebody that just comes in and says, you know, what do you recommend?

[00:12:11] Alan

Yeah, okay, fair enough. I mean are you seeing… anecdotally it feels like institutions have heavily loaded up in private markets and maybe are overly exposed there. I mean, in terms of the kind of institutions coming into you, do you think that’s a theme that you’re seeing?

[00:12:29] George

Definitely. So, we see a lot of a very common request from an institution is, I have this portfolio, a large portfolio of privates, maybe concentrate in a particular area, and privates can offer really interesting long-term return opportunities. However, people forget that once you’ve gotten invested in them, you largely can’t trade them unless you’re willing to accept a material discount. So, a lot of the conversations we have are, I have this portfolio of privates, I want to get back to my strategic allocation but it’s got to be liquid. I’d love to have some alpha in the process and I want something that’s disciplined and systematic, help me get there. So, that’s really in the multi asset space. Those are a lot of our conversations.

Liquidity is a key component because, again, most investors already have relatively mature portfolios.

[00:13:25] Alan

Okay, interesting. So, obviously, I mean linked to the questions around 60/40 is the, I suppose, pick-up in inflation we’ve seen in recent years. Obviously, we had an inflation spike in 2021, 2022, inflation’s come down since then but is still above target, and it’s picked up again recently.

I mean, obviously that now seems to dominate the thought process a little bit more. Is that what you’re seeing as well? And I mean, what does that mean in terms of a portfolio construction perspective?

[00:13:56] George

So really, since the beginning of the military actions in the Middle East, the market has really… there’s been one principal component. It’s been, you know, oil’s up/equity down, or oil up/inflation up/equities down, or vice versa. And that has really been the dominating risk factor that’s driven the news. You just kind of have to look at one asset in the morning and you can figure out what everything else is doing.

Inflation, again, what we rely on is a number of like advanced language models to kind of see how people are talking about this in the marketplace as well as monitoring flows in the marketplace. And it’s clear that inflation has become a risk factor. It’s not quite at the level where, oh, this is going to cause a recession, or this is kind of cause for a major downturn. It is definitely going to weigh on the economy and shave a bit off of GDP.

But I think, typically the rule of thumb is around 4%. If inflation is under 4%, you can still do okay with equities. Obviously, commodities become more important. If it starts getting above 4%, that’s really when we start seeing kind of material impact, both in terms of fiscal policies but also in terms of markets. And it doesn’t seem like we’re going to get there just yet in the near term. But again, this is something that we monitor on a daily basis.

[00:15:22] Alan

And then, I guess, from a portfolio construction perspective, presumably then you have more clients looking to actively hedge inflation risk or are cognizant of that risk?

Is that fair to say?

[00:15:37] George

So, the way I would say that we think about it is that there’s kind of a mid-horizon view and then there’s a much shorter-term horizon view and the mid-horizon view is much more tied to fundamentals. So, there, you know, we’re looking at the multitude of data that you can get: forward interest rates, you know, GDP, corporate profitability, you know, employment, jobs.

There’s a whole host of numbers that we use to then try and figure out what regime are we in, in kind of like the mid-horizon. And then we combine that with a shorter-term tactical view that, again, is going to basically try and pivot between, you know, risks about inflation. Obviously, you know, having some risk about inflation will mean that, you know… We’re still pro risk in general, but the risk concerns around inflation have caused us to kind of shrink some of our active exposures just because of really the tactical overlay on top of that.

But, you know, if you look at the other economic data, we’re still generally in a benign pro risk environment. If you look at any of the macro indicators, you know, any of the forward rates, I mean people are expecting this to continue for a short period of time. But yeah, three to six months, end of the year, we expect things, they’re not going to go back to where they were, but they’re probably not going to get worse.

[00:17:03] Alan

Yeah, fair enough. But just to come back to the inflation bit, I mean, obviously the kind of traditional way people might hedge or have exposures to assets that would benefit from inflation, maybe real assets, property infrastructure, etc., and then, obviously, in the liquid space, is it via gold I guess, or commodity futures, or how would you kind of construct a basket or a portfolio? Or is it more on the equity side, more energy stocks, etc.? Or how do you think about kind of creating some kind of liquid inflation hedge?

[00:17:35] George

It’s really a combination. It’s really a combination. Most of the time we think about how to hedge against inflation using positions, as you said. So, again, probably the best tool is commodities to be able to hedge. And obviously, if you look at the return of energy, it’s been kind of huge returns since basically the over year-to-date just given the overall geopolitical risk. But again, commodities are still one of the best liquid tools, right?

You know, real estate, infrastructure, etc., those all give you some long-term hedging against inflation. Equities tend to do well up to a certain threshold. So again, we have been… And actually, you know, some of our commodity, we were already… We always include some commodities given the inflation hedging nature of them.

Commodities, if you ask me, I kind of call them the forgotten stepchild of multi-asset. Everybody thinks about stocks, and bonds, and real estate, but again, commodities oftentimes give you a good amount of inflation protection. Really what you’ve got to worry about is the inflation shocks and that’s when commodities really can perform in the short-term.

[00:18:50] Alan

Yeah, fair enough. You mentioned there about regimes and identifying regimes and I know that’s something you’ve done a lot of research and work on. And, I mean, it feels like the macro regime has undergone a fairly significant shift, if you look back to the 2010s. Obviously we had secular stagnation, we had low inflation, low growth.

Now, obviously since COVID we went into a higher nominal GDP environment for a while, but now there’s, at least, a little bit stronger growth and certainly higher and more volatile inflation. So, what are the ways you kind of categorize the economic and market regimes?

[00:19:35] George

So, for us again, we’re relying on a number of techniques to use to categorize the regimes. The one that we’re using most recently (just to throw in a little bit of mathematics) is kind of a Gaussian mixture model that basically views the world, views the economy, as some sort of unobservable state. And then we use all of the different data points to kind of classify and say what type of regime are we in?

And again, you know, that tends to be, I’d say, a medium-term view. We don’t use any like market-based data there because markets can move much faster than the overall economy.

So, again, that view tends to put us in a moderately pro risk environment. This is something that… What I would say is when you’re in the world of multi-asset, you really only have so many instruments that you can use. So again, it’s useful to have a model, but, I mean, one of the challenges that you have when using models is you always have to ask yourself, you know, are the assumptions behind the model still valid? Has there been a regime shift in the data? That’s the kind of thing we worry about every day when we’re using models.

We still see like an overall positive macro environment for risk assets. Definitely risk is on the horizon, there’s no question. But in many ways the US economy has been surprisingly resilient despite all of the challenges that have impacted it.

[00:21:02] Alan

I mean people talk about regimes and also people talk about, I suppose, there’s different terminology. Some people call it different quads, or people talk about the different stages, disinflationary boom, or inflationary boom, and then stagflation or recession, etc. I mean, do you kind of categorize markets in that way and then try and map that to asset performance? Or obviously, you’re talking about this Gaussian approach, which sounds much more, I guess, statistical in nature. Is that it?

[00:21:36] George

So, we have multiple approaches. So yes, we do have kind of more traditional, where are we in inflation? Is inflation high and rising, or high and falling? Again, we have those traditional. And then we also have more statistical methods.

[00:21:51] Alan

Yeah.

[00:21:52] George

Rarely do you find that one model is the best model. Oftentimes what you really want to do is you want to combine multiple models together. So, behind the scenes, rarely is there one model that says, you know, here’s the one scenario that we think is the highest likelihood. Oftentimes there’s really multiple models running in the background. And again, you want to come… I mean, just like different data sources.

We’re looking to, you know… We don’t have a portfolio that’s just a growth portfolio or just an inflation portfolio. We’re trying to balance all of those things together. You know, we think about relative valuation; we think about growth, but at the same time, it’s the same thing with models, you know, like, rarely, do you want to have one model.

And again, if you look at these… I love to look at this website, Kaggle, that has these kind of model building competitions. If you look at the winners of those projects that they have, they’re all multimodal. The winners are usually like a multimodal approach. So, rarely is there one approach that is, okay, this is the one answer. The world is just not that clean.

[00:23:03] Alan

No, no. I mean, absolutely. I mean, it’s a segue to the next question, which is, I mean, if you look at today and maybe look at it from a regime perspective, or a market perspective, and parallels with the past, I mean, have you observations? Because as you say, there’s always observations with lots of different periods, but not one exactly. Are your models pointing to any kind of parallels?

[00:23:26] George

I think people do get a little too attached to, okay, my particular analysis says that, we’re in a regime just like a particular year in the past. And, again, I think this is probably, I’d say, the bias of like human pattern matching. What I would say is, we spend a lot of time doing the research, building the models, and then we largely use the models in the real world.

And again, you’re always going to have models that have differing views. Rarely do we build things where everything’s going to point in the same direction. But really the effort and the thinking about all of that happen at the model construction stage. And once the model output is being produced and recommending a position, we’re always asking ourselves, are the assumptions correct? But if the assumptions are correct, we’re going to use the model.

Now, my job, and the hardest part of my job, is making sure my team sticks to their research and sticks to their process. And this is the hardest part of investing, the market’s working against you. You have a process, and you’re getting nervous, and it’s not working, you’re losing money, people are picking up the phone, what’s going on, that’s the time that people tend to panic. And the hardest part of this job is being patient and sticking with your guns. And if you have faith in your process, sticking with your process, that is the single hardest part of this job, particularly when everyone’s questioning you.

[00:24:59] Alan

Absolutely. Fair enough. At the same time, you did touch on kind of regime changes or structural shifts and all of that. For any model builder, I guess, that’s always top of your mind. I mean, are the models all kind of fundamental or do you respect the market trend or the market price action? Are there any kind of pure price inputs into the models kind of more technical in nature?

[00:25:28] George

So, there are some… I would say most of the model, probably 80%, is really derived on fundamentals. Looking at fundamental data and combining it in unique ways that give us our recommended positioning. So, this kind of goes back to exactly what I was saying before. If you have a purely statistical model, and it tells you to go long Europe and short the US, and it’s just a purely statistical model, there’s going to be a period of time when that doesn’t work.

And when it doesn’t work, you have to ask yourself, what do I do? And if you don’t understand it, if it’s a purely statistical black box model, that’s when you’re going to get nervous and you’re going to unplug it because you don’t understand it.

If you have a model that has fundamental data and you’ve checked the insights as you’re building the model, and as you build the model, you’re thinking through, like, does this make sense, does this type of… what do I expect my priors to be, and is the model coming out in the way that does that?

So, it’s not just like, hey, I run a regression, and I take the coefficients, and I throw them in the new model. As you’re building the model, you need to really think about what is it that you’re trying to tease out here and how does this align with your insights? And this is really true with the macro side because the degrees of freedom are very, very, very few. So, when you’re doing that, that is what allows you to stick with your model when times are tough. Right?

[00:26:52] Alan

Okay.

[00:26:53] George

We do have some small statistical components just because, again, you need to pay attention to those. They tend to be much shorter term.

[00:27:00] Alan

And have there been periods where, obviously, we’ve had big dislocations? Maybe COVID is the obvious example where we had big moves in markets seemingly divorced from economic fundamentals, because obviously there was a pandemic out there that didn’t immediately get reflected in economic data. Was that a challenging period or did that prompt you to have to intervene? Or how did you deal with that as a fundamental macro modeler?

[00:27:29] George

Yeah, so again, there were situations where… COVID was very challenging. So, on the macro side, I would say some of the biggest challenge was that certain data was not released on time. I mean, everything was shut down. So, certain time series were not being updated. And that is where you have to ask yourself, okay, you know, here’s my model. It was built with a certain set of assumptions. How would the model work today, you know, based on what we might estimate this data to be? And those are the periods of time that were, yes. You have to think a bit beyond just, okay, what does the model say based on some historical data that may not be accurate?

I mean, what I always tell my portfolio managers at the end of the day is we’re implementing a certain philosophy. The model is a tool that enables us to be much more efficient in doing that. But at the same time, the investment philosophy comes first. And it’s periods of time like COVID that you have to be a bit more hands-on and say, wait a minute, what is really happening in the market here? The model has not updated quickly and we might need to intervene.

That was definitely true on the quant equity side, where, you know, you saw, again, cruise ships, airlines, literally couldn’t dock, or landed on the tarmac and weren’t taking off. There you had to intervene in the model because the inputs were just not updated.

[00:28:53] Alan

Yeah. Okay. So, in terms of kind of integrating that into traditional multi-asset, so, effectively you’re taking client portfolios which generally have a lot of either privates, or equity risk, or something like a 60/40, and then running, effectively, a quant macro or a systematic macro as an overlay using derivatives. Is that the kind of most common sense?

[00:29:19] George

That’s part of it. So multi-asset is extremely varied. So, there are parts of it that are just long-only, as you mentioned. You know, kind of 60/40, 70/30 type benchmarks using active management on the different sleeves. There are other components that are, you know, long/short, you know, more of a global macro. And then there, there are overlays. There are also a number of option based strategies that are designed to either give protection or give convexity during periods of market stress. So, the multi-asset team really has this large mandate across many different types of asset classes. Typically much more on the pro-risk side I would say of the equation.

[00:30:07] Alan

Yeah, very good. And I mean, in terms of that kind of downside protection, is that kind of, I suppose, buying protection, or going long volatility, or just thinking of people who often point to, obviously, long vol did very well in COVID, so that was fantastic. But then you contrast that with 2022 when the equity market went down, but volatility was moderately high without ever spiking very high, and long vol strategies didn’t provide an enormous amount of convexity in that period. So, if somebody is looking for downside protection, how do you think about that from an options perspective?

[00:30:47] George

So, there are a couple ways we do it. So, just going out and buying puts is expensive. So, again, typically we tend to be more active, I’d say, in kind of using call… in replacing exposure with call, replacing some equity exposure with kind of effectively some call exposure, typically on the long term. And what you’ll do is you’ll get more protection on the downside.

And it’s really how you rebalance that allows you to get the downside protection and do it in a very cost-effective manner. That’s the challenge is always like how do you pay for the downside protection?

There are also other dynamic ways that you can effectively do it with either kind of more defensive equity strategies or sector rotation that kind of gives you a bit more downside protection as the market moves down. But it’s very expensive to just go out and buy a call or go out and buy a put.

[00:31:48] Alan

You talked about the evolution that we’ve seen in markets, in data, in technology, etc. Obviously, everybody’s talking about AI at the moment. And obviously, from a quant perspective, people have been using machine learning already for quite a long time. Give us a sense on that evolution of the use of machine learning and AI in your research process.

[00:32:14] George

Sure. So, this is an area where, again, the language has changed so much over the years. What exactly is machine learning? You can argue that people have been doing machine learning for decades. If you estimate a regression and you update the parameters with some sort of a rule, is that machine learning? Well, that’s basic machine learning.

So, we’ve been involved in various aspects of that for, really, for many years. For us, I would say most of the focus in the current environment is around language models. That’s just been a very rich source of data and a rich source of alpha that we found. And this is both for quant equity and for multi asset.

So again, in early days it was like counting positive words versus negative words, which is called the bag of words method. Then we moved on to like BERT and FinBERT, you know, which measured sentiment based on news. And now we’ve moved to using LLMs to basically ascertain, you know, what types of themes are playing out in the markets.

[00:33:22] Alan

Okay.

[00:33:23] George

And you know, again, with everything, the goal is for this to be as transparent as possible. You give up a little transparency when you start using some of these methods because they are more statistical. But we’re always looking to make sure that we can trace from what the recommended positioning is all the way down to the raw underlying data, just for transparency.

[00:33:44] Alan

Okay. I mean in terms of the use of LLMs, obviously people have been using it to read a lot of, say, Fed speakers, or earnings releases, or things like that. I mean, there are, I guess, obvious applications. Beyond that, not sharing the secret sauce, are there any other kind of obvious ways that you’re using it or you’re hearing people are using it?

[00:34:12] George

Well, it’s an area of focus for us because if you think about it, so much of human knowledge is encoded in text. I mean, you can get Compustat data on quarterly earnings, and you can get earnings estimates from various vendors and you can measure kind of web traffic. But again, there’s a huge… If you think about it, most of the world focuses around written text.

So, you know, whether it’s an analyst report, whether it’s a company website that is talking about product descriptions, whether it’s, you know, a 10k or a product release, just the amount of information in there is very rich. And the ability to extract things, both on the alpha side, but more importantly on the risk side, have just really advanced by leaps and bounds. And so that’s really, I would say, the area where we see the most relevant research in the current environment.

There are lots of applications obviously, you know, accessing research, summarizing, kind of consolidating data in a timely manner. Obviously, the software developers are big fans, accelerating a lot of their process.

[00:35:27] Alan

Yeah. Now, obviously you mentioned simple tasks like counting, you know, counting frequency or whatever, and then positive to negative. So, I mean, do you think LLMs are at the point now of being good at evaluating things, say, for example, the tone of a Fed speaker or something like that?

[00:35:46] George

For things? Yes, I would say for certain tasks, yes. Keep in mind, this is a statistical model, right? So, all of these models are only as good as the data that’d been trained on and, you know, you have to remember that all of these LLMs are basically trying to predict the next token. That’s all they’re doing. They’ve got these tokens that they’ve seen in the past and they can predict the next token.

So, for things like human speech, yes, they’re very good at saying, really what you’re saying is how does this compare with the past? You know, what is the pattern of speech and does that tend to be positive or negative? So, for things like that, where you’re trying to evaluate something, yes, LLMs are good. Actually, BERT is surprisingly good, or FinBERT’s surprisingly good in terms of extracting sentiment.

So, there are a number of areas like that where, you know, yes, LLMs have gotten to the point where you can really use them as effective tools in analyzing either speech or text documents and really probably has become the state of the art.

[00:36:47] Alan

Okay. I mean, you do hear a lot of people talking about the use of bots and how bots maybe will overreact to something and you’ll see this, you know, short term spike or sell-off in the market because the bots read something which proved to be incorrect. I mean, are they valid observations? Do you see that or not?

[00:37:09] George

You know, it’s hard to know because we just don’t know exactly who’s doing what. I would suspect that, yes, there are automated strategies that are probably, you know, in need of continual refinement and maybe driving prices in the short-term.

I mean we, again, this is no different than somebody who fat fingers a trade, you know, and accidentally moves a market kind of in an extreme manner. You know, if somebody’s programmed a bot to do something and drives the price in an extreme manner.

So, is it happening? Yes, it’s likely. It’s a little hard to show evidence that it is indeed the case. At PGIM Quantitative Solutions, we use a lot of these advanced tools but it’s still overseen. There’s still, what we call, a human in the loop. It’s still portfolio manager reviewing trades and ensuring the portfolio is in the same spirit as the overall investment philosophy. It’s not, kind of unconstrained robo trading.

[00:38:10] Alan

Yeah. And I mean a big debate at the moment is the economic impact of all of this. I mean, do you see… Presumably it’s making your researchers more efficient, more productive. Does that mean less hiring need over time, or is that an oversimplification, or how do you read that?

[00:38:31] George

So, I don’t view this as like a net destructor of jobs. I view it, like, with every technology change… I think if you go back and look at historical patterns, you know, around the introduction of the steam engine, and, you know, the automobile, and light bulbs, and things of this nature, or kind of machine automation, there have been lots of times that it’s basically said it’s going to put everyone out of work.

And what’s happened is yes, certain types of jobs may get eliminated or reduced, but other jobs are created. Don’t forget, building these models is not something that is done easily. It costs hundreds of thousands, if not millions of dollars of time to calibrate these models and I don’t know if anyone’s really making money in this space yet. There are a lot of people that have very high hopes of commercializing this.

So, I think it’s going to make people more productive. It’s probably going to eliminate some positions, but it’s going to create new positions. I mean, again, if you think back to, like, the introduction of the Internet, I don’t think anybody thought of the types of things that were going to come out of the Internet. And you know, in the early days it’s always a little rough. You know, you think, oh yeah, this is kind of hokey, it’s not good enough, if you think about like the first cell phones that were bulky, or you think about the first iPhone. And again, it really was kind of like, yeah, we can’t see this thing taking over the world. We still need digital cameras. Well, guess what? Nowadays we don’t because the technology continues to get better. I think that’s the trend we’re going to see.

[00:40:14] Alan

Yeah, yeah, interesting. Maybe moving into, you know, current markets a little bit more. Just curious to see or get a sense on what you’re seeing in markets. Obviously, we’ve seen quite a number of shifts on the rate sides in the last number of months. If you go back to February, we had all of the talk about the Citrini report and yields went down.

And then obviously, as we went into March, we had to strike in Iran, and we had a big reversal in rates. And we’ve had market going from pricing and rate cuts to possible rate hikes. I mean, how has that, you know, impacted performance positioning? What are your thoughts on that?

[00:40:59] George

So, again, this is the challenge. And I think this is just the way the modern world is. You have to be very nimble, right? I don’t think that there’s such a thing as a forecast, a year out, that isn’t going to get revised along the way. And if you look at like, again, if you’re looking at future interest rates, you know, moves as predicted by forward curves, you know, you can go back and look at that historically and rarely is it accurate. I mean, oftentimes they’re forecasting hikes which then never materialize or they’re forecasting cuts that never really materialize.

So, again, this is just where we are right now. We are very data dependent. So yes, inflation is elevated. It’s not at the level that, again, it’s not at the level that we see it causing like a major downturn or a recession. It is a risk factor. I think the other challenge too, that’s probably going to pay out over a longer term, is just the overall fiscal situation.

You know, where are we in terms of debt to GDP, what are the long-term policies? There are a number of issues that probably need to be resolved and it doesn’t seem like really either party is tackling them. But at some point they are going to become painful enough, we are going to hit the so-called tipping point. The hard question is where is that and when does it happen? It’s not easy… There’s no magic formula for that, which is, I think, why you just need to be very nimble in this world.

[00:42:35] Alan

And is that something you can model or not? Or I mean something like the term premium has gone from being significant to being negative; at a time now, it seems to be growing again. And that would kind of encapsulate those deficit concerns. Is that something you specifically look at?

[00:42:54] George

Yes, I mean, listen, with everything, you can attempt to model it. The question is, how good is the model? You know, people forget that in the world of quantitative investing our models are, you know… I like to say that, you know, we have a little bit of an edge. We don’t have a huge edge. We want to be right a little bit more than 50% of the time.

And, you know, I think it goes back to the old baseball analogy, the Billy Bean Moneyball analogy of like getting people on first. So, our goal, really, is to be right and kind of be right on average a little bit more often than not. And over time that adds up. And it’s also a much more smoother ride because you’re not taking these big bets along the way.

So, again, do we try and model these things? Yes. It’s very hard for us… Rarely do we come out with a big call that’s like, yes, this is grossly overvalued or this is grossly undervalued.

[00:43:58] Alan

I mean there’s a bit of a debate in markets amongst quant researchers, maybe about six months ago, I think it was, around simplicity versus complexity. And for a long time the mantra was always parsimonious models, fewer inputs, keep it simple, they’re more robust. etc. And then there’s a research piece that came out saying, no, that’s not correct. Actually, maybe there is a case for more complexity, which kicked off a whole debate. I can’t say I followed all of it, but I was aware of it. And as somebody more in the weeds on all of this stuff, what’s your perspective on that?

[00:44:38] George

Yeah, so generally what I say is we want just enough complexity to basically do what we need to do. So, like, I tend to go more on the simplistic side. However, you know, you have to do a good job of representing the world, and modeling the situation, and thinking about that. When you’re building the strategy, again, you’re not just running regressions. I mean, you know, there are a lot of people that say, hey, let’s just go run a backtest.

Well, that’s interesting. That’s a great way of solving the in-sample problem. The real problem we’re trying to solve is what happens next. And the absolute worst thing you can do is optimize in-sample performance. When you’re building the model, you need to be thinking about what happened in this sample, what is the environment, what did this capture, is that likely something that’s going to persist? If you take that approach, you’re much more likely to be successful in the future.

And again, you need to have enough complexity to represent the world, or attempt to. We’re never really representing the full world, but we want to have enough complexity to represent the richness of what’s out there. But you don’t want to overdo it because, again, that’s just like a loaded weapon and someone’s going to blow their toe off.

[00:46:06] Alan

And I mean, from a research process perspective, how do you guard against that? Presumably you have a whole bunch of researchers who are motivated to try and come up with something to present to the investment committee. And obviously they’re going to be influenced by their experience in markets, effectively solving in-samples. So, what are the guardrails?

[00:46:25] George

Yeah, so listen, this is again one of the challenges. And again, you typically have a lot of young members of the team that are looking to make a name and they’re basically saying, well, how do I turn this from an IR of 1 to an IR of 1.5, and how do I make this a little bit better? My question to them is always, what do you have to do to kill this research? Like, what assumptions do you have to make in order for this to go away? Like give me the scenario that this fails.

So, we always like to have somebody that’s thinking about, like, what’s the argument against this? And we have a relatively healthy debate internally. Again, it’s a fine line. On one hand having a systematic process is great, and at the same time you have to be careful not to over optimize for in-sample. I can’t tell you how many times I’ve seen this where somebody’s built a strategy, optimized it to fit a particular period of time.

The truth of the matter is, there’s so much noise in the data in the real world that you can find… This has always been my challenge with economics is you can always find some period to prove some thesis that you want, but it may not work all the time. And this is kind of both the curse and this is what makes markets interesting, right? They don’t stay the same and what people are doing today is very different than what they were doing 20 and 30 years ago.

[00:47:55] Alan

As you say, markets evolve. I mean, apart from having some complexity in there, are there other mechanisms to build in adaptiveness into models? Is that just kind of looking back on parameters that have worked and then updating for that, or how do you think about kind of adaptivity?

[00:48:17] George

Yeah, so we’re always looking, so we’re always monitoring what’s working… I have this very… It’s a very humbling exercise that I force my team to do, which is that, you know, we’ve done research and we basically say okay, when we did the research, how did the signal work? What were the parameters during that period of development? And then if you freeze it at that point, once we’ve gone live, what has been the out-of-sample performance? And it’s tough, right?

I mean, again, there is this bias to optimize in-sample and it’s a very humbling meeting because there are lots of signals that have been like positive, positive, positive, and then, you know, rarely does it go negative exactly. It has happened, but usually it’s, like, positive but not as positive as we thought it was.

But my view is, again, you can’t solve the in-sample problem. And you know, whenever we look at parameters and we say, hey, something decayed over three months in sample over this simulation, I’m always pushing people to say, well, if that’s the case, today’s parameter should probably be somewhat shorter. Right? You can’t be assuming that what was in sample is going to work just because you calibrated that in sample. You have to extrapolate forward.

So again, there’s lots of things. If you look at policy response, policy response is much faster now. You know, if markets are, again, it’s extreme with this administration in terms of like markets are panicking and people will start tweeting. You know, it used to be that with trends, policy response took time, and they did not move fast.

But really, if you think about it, GFC, I mean, liquidity. GFC kind of slowly got worse over time and then really accelerated and instigated policy response to kind of unlock liquidity in the market. If you think about Silicon Valley Bank, we had a bank run that occurred overnight. This really never happened before.

And again, similarly with COVID we had kind of a market move in just really a record short period of time. If you panicked at the bottom of COVID you really missed out on the rebound.

So, we’re always trying to think how we’ve done research, it’s worked in-sample, what parameters do we need to adjust, or do we need to kind of think about ways of shutting something off because the current environment is much faster. So, it’s something that you need to build in. We try not to wait until something has failed. Because of that reason, we try and design it in from the beginning.

[00:51:02] Alan

I mean, another change we’ve seen in the last while has been more retail participation in the stock market. I mean, I don’t know, does that impact kind of macro markets? Presumably it may impact your long/short equity, equity market neutral stuff. Again, is that something that you’re actively thinking about more and more?

[00:51:20] George

We do, yes, exactly. You know, I’d say probably a bit more on the equity side, just given that in the US you tend to have these meme stocks. And again, in the long run the models work, but again, you can have periods of time. And again, this is not a new insight. You can be right in the long run, but you just can’t afford to maintain the position in the short run.

So yeah, again, particularly in the US where you get, you know, I like to say that, you know, in the US you get these little pockets of hopes and dreams, where companies can be losing money and have no near term forecasted profitability, but people get excited about the story and invest. And again, this is one of the features.

This is actually one of the features, right? Because if you think about other markets, again, people say, oh, we’ve got these irrational investors and, you know, they’re chasing things in the short term. And I’m like, that’s a feature, that’s not a problem.

If you think about the world of bonds, the world of bond trading is largely an institutional market, right? And again, there you don’t have large, extreme moves driven by kind of speculation, maybe apart from bonds that are in default from Venezuela or Argentina or something like that. But you know, it’s much more of one institutional investor trading against another.

Whereas in the equity market you have retail investors and that’s really creates a lot of opportunities. The trick is obviously staying out, away from that steamroller.

[00:52:58] Alan

We’ve talked a little bit about kind of the shifts in markets of late, particularly on the rate side. And you mentioned the perspective that the overall macro environment is still kind of pro risk based on all of the parameters and models you look at on average.

But I mean, would you say is there anywhere investors are under appreciating risk in markets at the moment or overlooking opportunity, anything that’s standing out or any signals you’re getting from that side? Obviously, we’re in an environment where we’re seeing this kind of rolling wave of exuberance in the stock market. Do you think that’s a warning or is it just reflecting structural trends?

[00:53:45] George

There are definitely some pockets of excess. I mean the thing about the US in particular is that earnings have been solid. I mean, you know, if you look at what is interesting, again, I hear lots of commentary against kind of the mega cap names. But you know, if I think back to the tech bubble, the tech bubble was all about names that were not profitable. Right? I mean, we were talking about valuing companies based on clicks and eyeballs. The, the mega cap names that are doing well today all have cash flows. These are all profitable businesses.

So, listen, I think there’s a debate about, you know, is the current CapEx cycle going to be worth it? And I think that, to be honest, that’s something we’re just going to have to see. Like does this CapEx get rewarded? Does AI turn out to be a bit of a boom? I think it is going to help some areas. Where’s their opportunity? Again, I think people have probably ignored emerging markets for a period of time.

For many years emerging markets was always a story of, oh, it’s the next best thing, but the earnings never materialized. But I think we are seeing changes in emerging markets, and they’ve been out of favor for so long.

We do see, I would say for institutions, most of the incoming questions we get are about emerging markets because they also see, if you think about long term GDP growth, you know, long term profitability, and the fact that the asset class has been ignored for some time, we see that as a long-term strategic area of growth.

[00:55:20] Alan

Okay. I mean how do you see quant investing evolve from here? Obviously as you say, it’s evolved so much since the 1990s, obviously in terms of data, then the Internet, and then, obviously now, we’ve got machine learning and now the LLMs. Is quantum computing the next thing or what will it look like in 10 years time?

[00:55:49] George

Yeah, listen, yeah… quantum computing, I don’t know… I mean, again, people tell me that things like fusion are really just around the corner. But as we know that’s been the case for the past 50 years it’s been just around the corner, and quantum computing.

Look, I think some of these technologies are very interesting. What I would say is there’s a big difference between making something work in the lab and then scaling it up in the real world. Right? Making something work in the real world, it has to be pretty robust. You know, it can’t be like a scientific curiosity.

It has got to be something where, you know, it’s robust enough to work in the real world. They’ve figured it out. I mean, why have semiconductors been so successful? It’s because these semiconductor fab companies have figured out how to do like extreme clean room automation, you know, all of the many processes of making chips. And, again, it’s become a huge industry and there’s a lot of technology that has gone into like the commercialization of this.

And, you know, again, it took years to go from the early chips that we saw, you know, in the days of the 6502 processor to where we are today. And it’s going to take time for some of these technologies. Even if it works in the lab today, it’s going to take time for it to become commercially viable.

[00:57:22] Alan

Okay, good.

So just before we wrap up, we always like to ask our guests for some reflections on career and, you know, maybe advice for people looking to build a career in quant. I mean, any things you’ve read or done that have been influential on you.

[00:57:39] George

So, what I would say, so my general advice to younger folks interested in this particular field or interested, really… A lot of this applies to, I’d say, any technical field, A, learn how to communicate, learn how to write, and learn how to speak. I personally underappreciated these skills early in my career.

But what I’ve seen is that being able to communicate, whether it’s a podcast or writing, is something that is going to be a huge determinant of your success over time. You know, like being able to communicate simply, effectively, being able to read the room, understand the audience, you know, how do you interact with, and how do you interact.

The second thing I would say is build a network, Right? None of my jobs have come from like a blind application. You know, never once, I think I applied blindly to graduate school, but I think after that, going to NASA, going to Barclays Global Investors, they were always connections that I had with individuals that ultimately led to a role. So, the trick about a network is you need to build it when you don’t need it, because by the time you need it, it’s too late to build it.

And I think the last thing I’d say is just continual learning. Whatever we’re doing today, I wasn’t taught in grad school or undergrad. I mean I didn’t study economics. But you can learn it. Right? So, you have to just be committed to evolving. You know, whether it’s using new tools, whether it’s picking up ideas, you have to be committed to learning and pushing yourself. And look, it’s not an easy process, but I think that’s how you stay relevant in industries in general.

[00:59:36] Alan

Good stuff. Well, thanks very much for coming on George. It’s been great to get your insights on the whole world of quant and quant multi-asset and quant macro and all that goes with it.

So, it will be fascinating to see how that whole world evolves in the next while. People can obviously follow you, and PGIM, and PGIM Quant Solutions, to read and hear more about your work. I certainly encourage everybody to do so because it’s a fascinating field.

But from all of us here at Top Traders Unplugged, thanks for tuning in and we’ll be back soon with more content.

[01:00:11] George

Thank you very much.

[01:00:14] Ending

Thanks for listening to Top Traders Unplugged. If you feel you learned something of value from today’s episode, the best way to stay updated is to go on over to your favorite podcast platform and follow the show so that you’ll be sure to get all the new episodes as they’re released. We have some amazing guests lined up for you and to ensure our show continues to grow, please leave us an honest rating and review. It only takes a minute and it’s the best way to show us you love the podcast. We’ll see you next time on Top Traders Unplugged.

This podcast expresses the views of it’s hosts and the guests appearing on the podcast as of the date of its recording, and such views are subject to change without notice. Top Traders Unplugged do not have any duty or obligation to update the information contained herein. Furthermore, Top Traders Unplugged make no representation to its accuracy and it shall not be assumed that past investment performance is an indication of future results. Moreover, wherever there is a potential for profit, there is also the possibility of loss.

This content is made available for educational purposes only and should not be used for any other purpose.

The information contained in this podcast does not constitute and should not be construed as investment advice or an offering of advisory services, or an offer to sell or a solicitation to buy any securities or related financial instruments in any jurisdiction. Certain information contained herein concerning economic trends, performance and other data is based on or derived from information provided by independent third-party sources. Top Traders Unplugged may believe that the sources from which such information are obtained are reliable. However, Top Traders Unplugged cannot guarantee the accuracy of such information and has not independently verified the accuracy or completeness of such information or the assumptions on which such information is based.

This podcast, including the information contained herein, may not be reproduced, copied, republished or posted in whole or in part in any form without the prior written consent of Top Traders Unplugged.