Trading Insights: Why macro matters for systematic investors

[Music]

Ralph Sueppel: Macro is everywhere, and I think we all know that. Even if you are focused on single-name equity, you know that the evolution of monetary policy, the rise of financial leverage in the economy in which you are trading, have a profound impact on directional and relative returns. So if you take these indicators, you already have a far more informed process in managing your directional equity exposure.

Eloise Goulder: Hi, welcome to ‘Market Matters,’ our markets series here on J.P. Morgan’s Making Sense. I'm Eloise Goulder, and today I'm so pleased to be joined by Ralph Sueppel, who is managing director at Macrosynergy, a London-based macroeconomic research and technology company, to talk about the investing process and how the J.P. Morgan Macrosynergy quantamental system, otherwise known as JPMaQS, can be used to harness macro-quantamental data in the investing process. So Ralph, thank you so much for joining us here today.

Ralph Sueppel: The pleasure is all mine. Thanks a lot for having me.

Eloise Goulder: Could you start by walking us through your background and your role at MacroSynergy today?

Ralph Sueppel: My background is quantitative finance and economics. I started out as a macroeconomist for J.P. Morgan and the European Central Bank. I worked as a macrostrategist for Merrill Lynch and UBS. And then I managed money for two large hedge funds before finally joining Macrosynergy. So what do I do at Macrosynergy? My role is to manage the development and research of macro-quantamental factors and macro-systematic strategies. The systematic and efficient use of macroeconomic information is underdeveloped, and we have strong reason and evidence to believe that better and more systematic use of that information makes the financial system more efficient, price more meaningful, and crisis less likely.

Eloise Goulder: That's fascinating. And can you explain what JPMaQS is as a service?

Ralph Sueppel: So JPMaQS is a service that makes it easy to use macroeconomic point-in-time information for building systematic strategies. At its heart, a data system that tracks for each point in time, and a couple of decades back in history, what the market knew about all types of important economic concept, anything from short-term inflation rates to now past the GDP growth rates to external balances to terms of trade. All of these things are being tracked point-in-time. That in itself makes JPMaQS unique, because in order to do so, you need to recover the full time series of underlying economic indicators for each day in the past, or at least approximated very closely. This is what we call vintages. Based on these vintages, we can calculate for any given day the concurrent economic indicator that would have been available to the market. And that allows to do scientific analysis and to do meaningful back tests. So that is the heart of JPMaQS. Now around this data set, we have built a support infrastructure which consists of experts that work with all the clients It includes tutorials, use cases. It includes seminars.

Eloise Goulder: And the fact that this data is point in time is critical, I understand, because without it any back test is rendered meaningless.

Ralph Sueppel: At the very minimum, it's risky. So if you don't have that, you spend a lot of money in approximating information states. And realistically, you cut a lot of corners. So what's the problem with that? If you don't use exact information states, you usually commit two types of errors. We call them type 1 and type 2. Type 1 is bad. Type 2 is very bad. So type 1 is bad insofar as it denotes an error where you don't see the value of an indicator because the revised series that are available today without exact knowledge on the release date are not what has driven the market in the past. Now, that's bad because you don't see trading opportunities. The type 2 error is a lot worse because it's the error where you see value in revised data sets, but there actually wasn't any. And that means you are operating strategies that just don't come anywhere near to the risk-adjusted return generation that you expect. Now, even if by good luck or by good fortune, you don't incur any of these two errors, you never know. And that means when you operate a quantamental strategy live, you're never sure whether your back tests are actually good estimates. And that means you tend to lose faith in the moments when your strategy is underperforming. You don't have the confidence to go with a strategy through the difficult and troubled times if you don't have a high-quality back test.

Eloise Goulder: Absolutely. Well, whenever a strategy underperforms, we like to ask, is this a natural underperformance that one should wear, one should live with, perhaps one should even double down on, or is this strategy underperformance a function of drift versus the back test for reasons like revisions? And so often revisions, can be a problem. So this idea that by using point-in-time data, you can be confident that specifically when it comes to revisions, that is not the source of drift versus back test, that can really give you confidence to stick with a strategy.  

Ralph Sueppel: It's very expensive to produce that type of data set. Bear in mind, these are billions and billions of data points. You need to literally go back into vintage data warehouses. And in this best case, they are available electronically, but they are still in messy format. There may be incorrect vintages. They may be mislabeled. They may be missing vintages. And that is before you go to the original sources of the data, going to individual statistics offices, research institutions, and I’m asking them, where are your historic data? Sometimes they are available electronically. Sometimes they come in CSV files. Sometimes they can't be found. So you have to go through these tedious, time-consuming and money-consuming processes. You have considerable fixed costs. And so we needed an agreement between an ex-buy-side institution, like Macrosynergy, and a central large sell-side institution, like J.P. Morgan, to bring the first global quantamental system to life. So JPMaQS is property of J.P. Morgan. The documentation of JPMaQS can be found on J.P. Morgan Markets, and there you can learn everything about the system and its content. The role of Macrosynergy is twofold. We advise and oversee the development of that system, of the data and the delivery to clients. And in addition to that, Macrosynergy is today the leading company for the know-how on how to build systematic strategies with actual macroeconomic information.

Eloise Goulder: Well, it's so exciting in terms of where this can take the investment community and obviously such an enormous and complex feat to create this full vintage warehouse of all of these different economic indicators over such a long time horizon. So let's move on to the know-how then and let's move on to the alpha signals. One would assume, given this enormous investment that there would be considerable alpha cross-asset. Can you speak to where you found the most power so far, i.e. which asset classes have perhaps been most responsive to these macro inputs and which macro variables have proved to be the most important?

Ralph Sueppel: My answer to your first question where that alpha can be found, is simple. It's everywhere, quite literally, and we have documented that. The whole field is in its infancy. There are a lot of pioneer gains, I think, for the next 10, if not 20 years, to be earned. We have what we call a macro-quantamental academy hosted by Macrosynergy. Now the most important part of that academy is a section that we call use cases, which provides empirical evidence for the relevance of macro factors across various asset classes, where we document with the Python code, with all necessary details, the value of macro-quantamental factors in fixed income, foreign exchange, equity, in commodity futures, and in credit. I think we're now close to 45 use cases with simple proofs of concept, where simple plausibility-based, theory-based hypotheses translate into significant economic value. Macro is everywhere, and I think we all know that. That said, of course the most popular applications are in the field of fixed income and foreign exchange. To give some examples directional trends in fixed income markets, are very dependent on the trends of inflation, the trends in economic activity, the trends in money and credit growth. Relative performances of fixed income markets are, to a significant degree, shaped by external balances, international investment positions, terms of trade. The behavior of yield curves is strongly influenced and predictable through point-in-time information on the state of the business cycle. And if you like trading more on a daily or weekly frequency, then we have shown a strong relation between what we call information state changes, second derivatives of quantamental indicators, or economic surprises, and subsequent daily and returns in fixed income markets. So that's why I say ubiquitous. You can take almost any sensible economic hypothesis and translate it into a quantamental strategy with the necessary care and with good principles and research. I would say you have a success ratio of over 50%.

Eloise Goulder: It's fascinating and very exciting. Perhaps not a surprise that the most obvious examples of alpha you found to date have been in the fixed income and the foreign exchange markets. Can we turn to equities? Because from my perspective, there's so many macro drivers of equities, but they're a little bit more complex in that certain macro drivers will impact certain equities at certain points of time. And equally, they don't always go in the same direction. So for example a high growth and perhaps higher inflation environment could support a certain type of cyclical equity. On the other hand, if that cyclical equity is also, let's say, an exporter and the currency has strengthened amid that high growth, high inflation environment, that could have a negative impact on their earnings. So it's a multifaceted problem statement. How have you found your system, your toolkit to be fruitful in equity alpha opportunities?

Ralph Sueppel: I actually have to say, it's proved to be not as complicated as we originally thought with one or two big exceptions. The directional trends in equity markets are partly predictable based on the economic situation of the large economies in particular. I say partly, not everything can be predicted, COVID-19 could not be predicted, but very prominent factors that condition forecasts for month or quarter ahead equity developments include, for example, the expansion of money and credit. They include short-term dynamics of sentiment indicators. Similarly, negative predictors include high operating rates, low unemployment rates, labor market tightness, high wage growth, and a state of the business cycle where aggregate demand as indicated by consumer data or national account aggregate demand data exceeds clearly the long-term trend rate of growth. So if you take these four types of indicators, you already have a far more informed process in managing your directional equity exposure. So the next stage is prediction of cross-country differences. We have just recently published our third note on that matter. And if all you do is to arbitrage different countries based on economic strategies, you already create significant long-term value. Now, where it gets a little more complicated is the field of cross-sector equity allocation. And I initially thought this is a very difficult field because there's almost no theoretical guidance except for some common-sense argument of what economic factor predicts the out-performance of what sector. You won't find many academic articles that make a connection between economic developments and sectoral equity performance. What we found, however, is that when we go through the whole list of macro-quantamental indicators, there are about 50 categories of indicators that matter at least for the out-performance of one of the large GICS sectors. And then we thought, well, since we don't really know how important these are, and since there are so many of them, we need help. And that help has been provided by machine learning. And I was originally quite skeptical with so little theory, can the machine learning process succeed? And boy, was I wrong. It was one of the most humbling experiences in my long career to see how much more effective a sequential machine learning process can be. Of course, the process still needs to be guided. You use something that has plausibility. But you can delegate a lot of decisions to the statistical process, such as deciding the hyperparameters and parameters, and ultimately delivering the signal. And through that process, we have seen an enormous and powerful impact of a large number of macro factors on sectoral equity performance. We live and learn. And macro has proven also in this case to be ubiquitous.

Eloise Goulder: Well, this age-old debate as to whether domain-specific knowledge is critical or whether the machines with enough history and enough reliable point-in-time data can actually provide value on the relationship between the input variables and the stocks is coming to life again with your example. And at least the impact of the machine, even if it's not just the machine alone, is significant.

Ralph Sueppel: It is an important debate in the field of macro investment. You see, we have the disadvantage we don't have as much data on macroeconomic business cycles in the age of liquid financial markets so we need to make do with limited amount of data. And that has, I think, for the longest time discouraged the more aggressive use of machine learning techniques for macro strategies. However, this can be greatly mitigated by two circumstances. First we can look at the behavior of similar asset markets across anything between 12 and 40 countries. In other words, we can use panel analysis, panel applications of random forests, And use the diverse experiences of different countries to derive more stable hypotheses. Second, you need to be parsimonious. You need to keep your machine learning process on a short leash. And what that means is you assign it a specific purpose. In most of our published use cases, we have given machine learning, specifically sequential machine learning, one dominant purpose. And that was to select and weigh from among a broad range of plausible trading factors. And at that specific task, with reasonable restrictions, such as restrictions to regularization, on this specific task, machine learning has excelled. It has delivered two things. First, it has discovered things that maybe we as even experienced human researchers would not have discovered. And second, it has given back tests greater credibility. It's not just about optimizing, per se, but it's about making decisions only point in time. Machine learning adds another layer to that, where not only the indicators used for the strategies are point in time, but also the factors built on them are built point in time. And the trading signal that's ultimately derived is also built point in time.

Eloise Goulder: Yes because you can't use your domain knowledge today to know that a certain macro variable was important in the context of a certain historical period. You want to rely on the machines at that point in time, which didn't have that knowledge to determine which were the significant variables.

Ralph Sueppel: Yes it is more credible.

Eloise Goulder: Yes. Out of interest, have you looked into the relationship between these macro factors and single stock equities? And have you thought about the problem that single stock equities change over time? They might divest certain businesses, grow other businesses, and therefore their sensitivity to certain macro factors will change.

Ralph Sueppel: Yes, we did. Unlike a fixed income contract, a stock changes characteristics and it can be addressed through something like time-weighted least squares, whereby the relevant history that you assign to past relations is decaying. I think after we saw the relevance of macro for cross-country and cross-sector, we have now more confidence the relative performance of individual stocks is predictable.

Eloise Goulder: We've spoken so far about the systematic usage of JPMaQS in the investment decision. But of course, discretionary investors care deeply about the macro economy. And in fact, it comes up so often in financial narratives of discretionary investors. So Ralph, is there a use case for JPMaQS within the discretionary investing community as well?

Ralph Sueppel: There is. And I increasingly suspect it may actually be the biggest use case. And so we have often portfolio managers or analysts whose assignment is to build something like scorecards, systematic visualizations and evaluations of global macroeconomic developments. And responding to that trend, the Quantamental Academy now also contains a set of scorecards, examples how to build these scorecards, with Python code based on cross-country macro-quantum indicators. It's a big field.

Eloise Goulder: So I have to ask you our age old question, what's more important, the data or the model? And from what I can hear, both are absolutely critical to everything you believe about alpha generation in financial markets. Can you weigh the relative importance of one versus the other?

Ralph Sueppel: That goes very deep. I would say any good strategy development project is an exercise in logic and humility. And maybe in our field, besides from saying we need both, the humility is more in shortage. So logic is what guides you to a good model, a good framework of thinking. And of course, it's indispensable. But what we lack is humility. And I think the way it usually goes wrong is one of two. So either we say, oh, this is so simple, it can't possibly work. That's probably the most common mistake that prevents people from making money in the macro space. It is that simple. What we lack is not intelligence. What we lack is consistency and focus. And that can be learned only by looking at the data. You can guess without data, but you can never run with conviction trading strategy without the data. The second way it goes wrong is that we think we know it all, and that my hypothesis is far better than your hypothesis. Again, the data helps us to become humble. There's much more knowledge in the herd than it is in my little house. And we all have to come to that point. And the easiest way is with good data, solid analysis, where we can build our know-how from an open mind and the acknowledgement of empirical results.

Eloise Goulder: It really rings true to me, both the examples you give, the fact that some relationships just seem too obvious. We can almost look past them. We think the markets are already efficient in that regard. But that's simply not true if you can apply something very consistently. But also we assume we know it all. And I think this comes back to recency bias. This obsession with the here and now, the macro variable du jour. So right now, perhaps that is the impact of tariffs on growth and inflation. And yet there are many other macro variables across the entire world that are somewhat driving assets right now. Or if they're not driving them right now, they could well drive them in the near future.

Ralph Sueppel: I keep referring people to the academic concept of rational inattention because that theory says most people do not pay attention to most relevant developments, because they are perfectly rational. Procuring information and implementing information is expensive. The concept is information overload. It's painful feeling overwhelmed with all the things that one ought to consider. And so most important information is being ignored, to save costs and emotional pain. Over and above that, we have certain structural impediments in our industry that prevent the efficient use of information. For example, in order to make good use of macro data, you need to understand and like macroeconomics a little bit. I have great good fortune that I work with the most advanced financial institutions in the world. Very few people are far advanced on the road of using macro efficiently. And yet everyone in the academic world seems to believe that we all do it. So there's a big disconnect.

Eloise Goulder: Well, I think that's a really wonderful place to close this discussion, because it's so heartening and it's so exciting if you are in the pursuit of alpha generation to know that opportunities and dislocations do still exist. And using this macroeconomic information with the full complexity and sophistication that you have put together in the JPMaQS system can be so incredibly fruitful in the investment process. So thank you so much, Ralph, for taking the time to speak with us about this today.

Ralph Sueppel: The pleasure is all mine.

Eloise Goulder: Thank you also to our listeners for tuning in to this biweekly podcast from our group. If you'd like to learn more about the JPMaQS system or Ralph's work on systematic strategies using this data set, then please do see the dedicated JPMaQS page on J.P. Morgan Markets or indeed reach out to your data and analytics sales representative. And with that, we'll close. Thank you.

[Music]

Voiceover: Thanks for listening to ‘Market Matters.’ If you’ve enjoyed this conversation, we hope you’ll review, rate, and subscribe to J.P. Morgan’s Making Sense to stay on top of the latest industry news and trends, available on Apple Podcasts, Spotify, and YouTube.

The views expressed in this podcast may not necessarily reflect the views of J.P. Morgan Chase & Co and its affiliates (together “J.P. Morgan”), they are not the product of J.P. Morgan’s Research Department and do not constitute a recommendation, advice, or an offer or a solicitation to buy or sell any security or financial instrument.  This podcast is intended for institutional and professional investors only and is not intended for retail investor use, it is provided for information purposes only. Referenced products and services in this podcast may not be suitable for you and may not be available in all jurisdictions.  J.P. Morgan may make markets and trade as principal in securities and other asset classes and financial products that may have been discussed.  For additional disclaimers and regulatory disclosures, please visit: www.jpmorgan.com/disclosures/salesandtradingdisclaimer. For the avoidance of doubt, opinions expressed by any external speakers are the personal views of those speakers and do not represent the views of J.P. Morgan.

© 2025 JPMorgan Chase & Company. All rights reserved.

[End of episode]