The Quant edge: Why systematic investing matters

Quin Smith:

Welcome to the Curious Podcast. I'm Quin Smith, Head of Distribution for Australia and New Zealand at First Sentier Group. Joining me today from RQI Investors is Dr. Joanna Nash, Head of Portfolio Management, and Dr. David Walsh, Head of Investments at RQI. Welcome, Jo and David.

David Walsh:

Thanks, Quin.

Joanna Nash:

Thanks for having us.

Quin Smith:

In today's episode, we'll dive into quantitative investing, the role of AI, and what role quant equities can play alongside fundamental investing.

Speaker 4:

This podcast is intended for institutional and professional audiences globally and financial advisers in Australia. Any advice within this material has been prepared without taking account of the objectives, financial situation or needs of any particular person. Before acting on any advice, seek the advice of a registered financial adviser and consider the appropriateness of the advice having regard to your objectives, financial situation or needs.  References to specific securities should not be construed as a recommendation to buy or sell such securities. 

Quin Smith:

Probably the best place to start, can you talk us through what is quant?

David Walsh:

Yeah, it's one of those issues that, questions that comes up quite often because quant can be perceived as very black box and very data and systematically driven without real investment insight. Just the data and the systems tell you what to do. Properly done, quant investing is not a black box at all. It's the systematic application of good investment ideas. So we take good investment ideas, things that work, to try and differentiate between good stocks and bad stocks. And we use data and systems and technology to implement those in the most efficient way that we can. That's kind of the best way of thinking of it.

Quin Smith:

Jo, anything to add there?

Joanna Nash:

Yeah, I think what David said is correct. We always want to have some investment insight behind it. And I think what people perceive it is is unknown, but let's take an example of one of those investment insights.

We look at a corporate culture, right? So the idea is that culture's driven from the top, companies with good corporate culture are going to outperform because the employees are going to be more efficient. They're going to be wanting to be more productive. So that's the sort of the investment insight we're trying to capture. Now you may go, "Oh, how does a quant capture that?" It's a fairly qualitative aspect.

So this is where some of the newer AI tools help us. We can use sort of natural language processing to see how people speak, to try and draw culture out of that and see whether they're sort of the way that they speak when they're in their prepared speeches versus when they speak off the cuff are the same or not. And that can then help us see how well embedded that culture is. And then we can sort of rank companies on how good their culture is versus others. And that's one of the insights or one of the ways we look at companies. And as you can see, it's very much fundamentally driven, it's very much investment insight driven, which then helps us to be able to analyse these companies.

Quin Smith:

Let's dive into that a little bit there with what you're talking through around the difference. So maybe, Dave, if I crossed you around, why is it important? Why is quant important? How is it different to other types of investing?

David Walsh:

Well, it's more about exposures to ideas than it is about individual stocks. So fundamental investing is more about finding an individual stock that the investor believes will outperform. And then holding a stock, waiting for a catalyst, or perhaps looking for the long-term delivery of returns, whatever the reasons for that fundamental investing is.

In terms of quant, we don't think of individual stocks quite the same way. Our portfolios are made up of them. We think more about exposures, tilting the portfolio towards a good idea rather than a good stock, controlling for risk we don't want to bet on. So quant investing kind of different from fundamental investing in that we tilt towards ideas, but control the risk that we don't want to take at the same time. So I think that maximising return and minimising risk at the same time concept lies at the heart of what quantum investing's all about.

Quin Smith:

That heart of quantitative investing, it's actually proven to be very effective in recent periods. More a cheeky question, can quant continue to deliver?

David Walsh:

Yes. And the answer to that is yes, but of course, but quantum investing is about systematic delivery of good ideas, like I said. If those good ideas are not paying off for whatever reason, the market is being, let's put it inverted commas, irrational, betting on things that we don't think were systematic long-term drivers of returns, then quant could underperform. But those sort of things have been shown many times in the past to be thematic, things that pass by. And returning to good economic ideas, companies that are cheaper than their peer groups should outperform, companies with better quality should outperform, companies who are on upgrade cycles should outperform, companies with better culture should outperform, those things are tilts towards ideas, given risk, that we can take and it can add value.

Now, if the market goes to a situation where starts rewarding expensive junk, then we're not going to outperform. But those sort of things are episodic and aren't consistent through time. So the answer is yes, quant should be able to outperform consistently through time, but there are periods when it won't be, of course.

Joanna Nash:

Look, and I think you'll see, if you look at quant's recent performance, it's been really, really strong, right? And that's where there's been a lot of attention coming into quant. And I think it's very much to David's point, he spoke about the different aspects of a company we look at and try and get different exposures to that.

What we've seen recently is that all those different aspects have been performing very, very strongly. So that sort of made, I guess quant highlight this additional return that we haven't, especially when you compare it to fundamental managers, has been very, very stark, that difference. So when people sort of go, "Okay, we're going to enter this quant winter again that we've seen previously," I don't think that's going to be the case because yes, we may not get as strong of performance because maybe every single aspect won't be outperforming at exactly the same time. But as David said, you'll get that consistent sort of performance from every one of those different exposures, which will then give us that sort of strong core performance that we'll get to the portfolio. So it may not be as strong as it has been, but I think we'll still get that strong performance coming through.

David Walsh:

Yeah, that's pretty important to highlight that because our alpha signals pick return or control risk are not all lined up on the same bets. So if we're looking at companies which are cheap compared to their peer groups, they may be companies that have sold off recently, in which case the momentum is not as good. So we've got momentum signals associated with what the market thinks is going to be good. We're buying companies that are trending upwards and selling companies that are trending downwards. But they may be expensive companies, which case we want to bet against them. So it's a combination of what we think these insights is, diversified across our composite alpha model that gives us the consistent returns over time. I'm not betting on one theme, we're betting on multiple themes and it's diversification of those themes, like value momentum, that seem to work well through the cycle.

Quin Smith:

It wouldn't be a conversation without bringing AI into the realm. It would be great to understand from a quantitative systematic investing perspective, how is AI used and how does it benefit your process?

David Walsh:

AI is a term that everybody's throwing around, of course, it's applying every part of our lives. We're seeing it all the time. Machine learning has been around a lot longer than the concept of AI. So the fact that machines, computers can learn to act in certain ways, to behave in certain ways or to differentiate between group A and group B is not a new concept. Taken further down the track, you end up with tools like AI, which generate things which are new or to interpret things in ways which are much more complicated. But at its heart, it's still machine learning.

So we use machine learning/AI tools in a bunch of places in our models, in our processes. Primarily, it's associated with improving access to research and alpha ideas, but also in terms of portfolio construction, in terms of risk, in terms of implementation, primarily alpha. So we can get at the alpha ideas in a much more systematic and complex way using machine learning AI tools than we could in the past.

For example, the ability to process unstructured text is a good example. So unstructured text like corporate filings or news items or corporate transcripts from earnings calls, those sort of things can be processed by URI reading them and going, "I think it's good news. I think it's bad news," or rudimentary machine learning tools that allow us to say, "Well, there's more good news in this than there is bad news." The machine learning tool does that. Or very sophisticated models based in AI, in inverted commas, using what are called large language models to interpret sentiment in very complicated ways. And that part is I think tool that allows us to extend that one idea we know is right, using machine learning much further than we ever could in the past.

Joanna Nash:

I think your point exactly there, what it's allowed us to do is opened up different data sources to us, and that's really important. So when you think about quants, people go, "Oh, it's got to be based on numbers, so it's going to be balance sheets or accounts or maybe some analyst numbers." They don't know about all the other stuff about the company, so how could they possibly analyse the company? The ability to read text, the ability to look at unstructured data, credit card transactions, looking at visuals and be able to interpret those has then opened up all those different data sources to quants to then help us better analyse the company. So in actual fact, these tools have made us sort of become more like fundamental investors in the sense that we can analyse a whole lot more information, which then allows us to be able to better determine which of the stocks we want to overweight and which ones we want to underweight.

David Walsh:

There's probably one other point I'd make about our use of machine learning, and that is that these tools, our traditional tools tend to be what's called linear in a sense that you change one unit and you get, say, two units of outperformance. In that linearity is embedded in a lot of the models that we use, but machine learning allows us to capture much more non-linear characteristics. So if you move one unit, you get two units upgrade. But if you move two units, you get six units upgrade. That non-linearity can be captured by machine learning, it's harder to do in more traditional models.

To give you an example of that, outliers in analyst forecasts can contain a lot more information than people who had consensus. So the impact of an upgrade or a downgrade when you're well away from consensus can mean an awful lot more than if you're at consensus. Machine learning tool is much more able to pick that up and extract that in a non-linear way than a simplistic model could do.

Joanna Nash:

Yeah. And markets aren't linear.

David Walsh:

No.

Joanna Nash:

They're a lot of complicated, so we need those more complicated tools.

Quin Smith:

I'm going to try and get you to crystal ball a little bit here. We're talking about how you're using AI or large language models today. What does that look like in three to five years from now?

David Walsh:

It's very hard crystal balling, these sort of things are difficult because the pace of change since ChatGPT broke onto the scene a couple of years ago has been so dramatic, with more recent changes around all kinds of things associated with Claude and with other things which have changed the way people are working in a lot of ways.

The impact of that in a productivity sense for companies is a little difficult to measure. So the extent to which it reflects on the sort of universe we're investing and how our models change to capture that productivity is quite hard to see. But in explicitly, the sort of things we would see changing would be the robustness of our models. We would expect that the next three to five years, we'll see an increased robustness in models because we can apply many, many more insights. That's one positive you get from that, because we've got many more insights captured in more ways, the data and tools allow us to do that.

The negative from that is there's a lot more ability, a lower barrier to entry, if you want to think of it that way, for other people to come in, smaller groups to come in, use these tools and data to exploit what might be considered to be in sample or data mining, so to exploit ideas they can see without really applying the insight. So volatility will probably increase because there's more AI in the market. So that's changes we'll see, probably more increased volatility, more chasing of older ideas, the change in the way we think of application of AI in our processes will continue to evolve.

And one last thing I think we might see, we've not really seen much change in a way that trading processes have evolved during AI. We've talked about it a lot, but trading's still algorithmically based with systematic measures and so on to go with that. I can see the use of agentic AI expanding much more into trading algorithms and the ability for agentic AIs to trade on our behalf.

I can explain what agentic AIs are if you'd like. Agentic AI, so a normal AI model would be you ask it a question, it gives you an answer. With an agentic AI, you ask it a question, you'd say, "I give you agency to act on my behalf." So it acts as an agent for you and does things automatically that you've prescribed without you having to touch it. Reads data, analyses it, does the action, that's what agentic AI is trying to do. In the trading sense, if you build an algo, an algorithm, it kind of does that, is a much more intelligent version of those algos, versions of those algos. I can see that as being a big change in trading over the next three to five years.

Joanna Nash:

What I will add though is I think we'll still need quant PMs, hopefully.

David Walsh:

Oh, yeah. Yeah.

Joanna Nash:

So as David's alluded to, quant is based on those investment ideas and the computers can't bring up those investment ideas, so you still need the people behind them. May make us more efficient, we may be able to trade better, as Dave was saying, but you still need the people behind to get the ideas to get that investment insight into the portfolio. So thankfully, hopefully the job's still there.

David Walsh:

No, Jo's made a really, really good point. One of the things associated with this movement towards AI and machine learning and tools like this is the risk for career development for people in the industry, that people who code for a living may be squeezed out, that quant PMs might have positions at risk, whatever. I don't see that at all happening. I see what it will do is it'll sharpen up the need for better people, that people will matter even more in the future than they do now because the day-to-day work is taken away and the insight and the ability to think through these problems is even more important now than it was in the past.

So in some sense, the day-to-day, I won't put drudgery in there, but that's kind of the concept, is that things which are done automatically, which can be automated by AI and by machine learning tools will be done that way. And then the people will generate the insights, will generate the ideas that could be used for investments or whatever the process improvements, whether it's being quantitative development or portfolio management or research, those tools, those skills will still be in demand, probably more so than they are now, relatively speaking. So your job's safe, Jo, I think.

Joanna Nash:

Oh, good, good.

Quin Smith:

Great to hear.

Joanna Nash:

Yes.

Quin Smith:

Moving tack a little, where are you seeing quantitative equities strategies being used in portfolios now?

Joanna Nash:

Yep. Well, we've been seeing in a number of ways where clients are coming up to us and asking us about the different ways that they can incorporate it. I think one of the positive aspects with quant is it provides that risk adjusted return. So clients can come in and say, "Okay, I'm looking for this level of risk," and you can provide a portfolio that provides that level of risk, which then allows them within their overall portfolio construction to sort of then go, "Okay, I can have something with a higher tracking error over here because I know I've got this sort of core allocation, and quants often at that core allocation of providing this controlled level of risk within my portfolio."

The other way we're seeing is that there was a big shift towards passive investment previously. People are wanting now to take on some alpha, but again, they're concerned on the risk. They don't want to take on too much risk. So what we're seeing is a lot of inquiries and a lot of people moving to sort of what we call enhanced quant or enhanced active. So this is lower tracking error, but it's providing that consistent return but approximately level of risk. And in actual fact, quant works really, really well in that process because things like what we call the long only constraint, not being able to short stocks doesn't impact on the portfolios as much there.

Quin Smith:

Tying it all together here, you both spoke to portfolio construction and risk modelling. Why is that more important or as important than, say, for instance, picking a theme?

David Walsh:

Yeah. So I think we've mentioned a couple of times about how important the insights are. And we talk of insights and generally narrow ourselves down a little bit to talk about alpha insights, cross-sectional stock picking, this kind of stock will outperform that stock or this theme will outperform, this better value stock should outperform their peers, things like that.

But the way in which you blend your ideas, the way you construct your alpha models, the way you build portfolios, trading off risk and return against that alpha, the way in which you understand risk, those sort of things are critically important no matter what your alpha sources are. So you should be trying to build those into your process and innovate as much as you can in a quant process on how you build alphas, how you build portfolios, how you understand risk. Even aside from the alpha insights important, which of course are important because you've got to implement them, right? That's the idea, but that's critically important.

If you chase the theme of too much AI, both in terms of an investment insight, I should chase AI because it's going to outperform, or I should use AI more, that kind of abstracts from the actual problem that you try and take an alpha and build into a portfolio which outperforms. So the alpha idea may incorporate things like using better AI to generate the alpha models or chasing earnings associated with companies which are generating AI, but it doesn't get you away from the problem. And what quant investors are particularly good at is managing that combination of ideas like constructing portfolios in a risk controlled way. I think that is true. Whatever theme you talk about, whether it's an oil theme or AI or blockchain or whatever you're picking, those sort of themes can come and go, but the underlying processes are still going to be the same.

Quin Smith:

Jo, I'm interested in some research insights. The team is obviously conducting research all the time. Is there anything you can share with us that is recent?

Joanna Nash:

So the corporate culture one, which we spoke out recently, that was sort of added into the model sort of late last year, and that's been one of the more recent works. I think some of the other aspects where we've been doing some work is often using machine learning tools to improve a signal we've already got. So that's also part of a research. It's not always necessarily finding the next best idea, love to do that, but it's also improving the way we implement it, which is what Dave was just speaking about.

So where previously, we may have looked at four different components, helping to explain, say, valuations. Now with machine learning tools, we can look at 50 or 100 different components that help us explain valuation tools, but also have that non-linear aspect to them as well. So what you then can do and what we've found with some of that research is that we get a similar performance from the signal, but we lose the drawdown, which is obviously very, very important.

So that's not as exciting as the next big alpha insight, but often it's that portfolio construction side, it's that alpha signal construction side that actually makes the big difference because you can have 1,000 amazing insights, but if you don't get exposure to them in the portfolio, there's no point having them.

Quin Smith:

Joanna, David, thank you for joining us today.

Joanna Nash:

Thanks for having us.

David Walsh:

No problem. Thank you.

Quin Smith:

And thank you for joining us. Please remember to follow the Curious Podcast. And for more information on RQI investors, head to our website, firstsentierinvestors.com.au.

We communicate and conduct business through different legal entities in different locations, please refer to the notes section of the podcast platform you use for more information. 

For Singapore only, the podcast shall only be used in accordance with the applicable laws in Singapore. In Singapore, the podcast is issued by First Sentier Investors (Singapore) whose company registration number is 196900420D. This advertisement or material has not been reviewed by the Monetary Authority of Singapore.   

First Sentier Group. Registration number 53507290B, First Sentier Investors. Registration number 53236800B. FSSA Investment Managers. Registration number 53314080C. Stewart Investors. Registration number 53310114W. RQI Investors. Registration number 53472532E and Igneo Infrastructure Partners. Registration number 53447928J are the business names of First Sentier Investors (Singapore). 

For Hong Kong, this podcast is issued by First Sentier Investors (Hong Kong) Limited and has not been reviewed by the Securities & Futures Commission in Hong Kong. First Sentier Group, First Sentier Investors, FSSA Investment Managers, Stewart Investors, RQI Investors and Igneo Infrastructure Partners are the business names of First Sentier Investors (Hong Kong) Limited.