Active Management: The best of both worlds?
Will applying systematic techniques to traditional discretionary active processes lead to better outcomes for investors?
• Managers are trying to combine quantitative investment with traditional active management
• The approach is sometimes referred to as ‘quantamental’
• Quantitative tools have always been used by fundamental managers but quantamental managers are doing something different
• The trend is driven by the availability of big data and artificial intelligence
A recent announcement by BNP Paribas Asset Management (BNPP AM) about a reshuffle within the business has brought the word ‘quantamental’ back into focus. BNPP AM used it in the context of combining quantitative expertise with fundamental research capabilities.
The expression is not as new as it may sound. In fact, according to Google Trends, it was already in use before the financial crisis and reached its peak in terms of Google searches in February 2006. The number of searches then declined steadily until it started rising again in recent months.
The concept is associated with the increasing collaboration between purveyors of quant investment and traditional active managers. BNPP AM has combined the two approaches within a newly-founded investment group called Multi-Asset, Quantitative and Solutions (MAQS). MAQS has 130-staff working on a diverse range of strategies all based on the idea that blending quantitative investment and fundamental active management is the way forward.
Denis Panel, who will be in charge of the group, says: “For us, quantamental is not a buzzword. It is not just about applying a quant filter to increase the quality of investment decisions. For us it represents a full organisation and a relevant framework that combines quant research teams with investment management teams working hand in hand. Quantamental means we are integrating directly new quant techniques in a fundamental portfolio.”
Judging by the recent spate of similar announcements by asset management firms, this could be a year in which such developments become significant. And the increasing collaboration between quants and fundamental stockpickers could be the response to the crisis of traditional active management. Stockpickers have suffered in recent years owing to a market rally that rewarded low-cost passive and index funds, as well accusations of ‘benchmark hugging’.
To understand what combining quant and fundamental is really about it is first necessary to clarify what these firms are trying to do. Depending on how quant and fundamental investing are defined, the merging the of the two approaches does not appear to be radically new.
Traditionally, quant managers gain an information advantage by processing large amounts of information about hundreds or thousands of stocks in a short time. They identify patterns and market anomalies – or ‘signals’ – that tell them where alpha returns are to be found. Traditional active managers tend to focus on fewer stocks and try to understand them inside out by getting to know the businesses and managers they invest in.
They also focus on the big picture, such as trying to identify the next commercial or economic development that would benefit certain businesses. This is how they search for an alpha-generating information advantage. There can also be a profound difference in terms of investment time horizon. Technology allows quants to exploit short-term price dislocations in a way that non-quants cannot. This makes many quant investors more focused on the short term.
Both kinds of investors, however, are trying to identify stocks that will outperform based on fundamental analysis. So in this sense combining quant and fundamental does not sound radical. Quant investors often see themselves as fundamental investors, and see no contradiction between the two approaches. Gideon Smith, CIO for Europe at Rosenberg Equities, AXA Investment Management’s quant business, points out: “We would call ourselves fundamental investors at heart, we just want to use technology and data to automate the process.”
Jai Jacob, managing director and portfolio manager at Lazard Asset Management, agrees: “A quant investor will tell you that they are fundamental investors. They are looking at income statements and balance sheets in extreme detail, much like a traditional fundamental analyst does.” The difference is, argues Jacob, that a quant can cover thousands of stocks in a consistent manner, something a traditional fundamental analyst simply will not be able to do.
A new concept?
There are different ideas about what might make ‘quantamental’ a new concept. For BNPP AM, it is about optmising exposure to the beta and alpha elements of a portfolio.
Panel explains: “Thanks to new quant techniques, we can build an allocation based on three separate elements: pure beta, through indices and benchmarks, smart beta or factors and uncorrelated alpha.”
At the same time, the idea has to do with risk management. Panel continues: “The quantamental approach puts risk at the heart of investment. Our Parvest Diversified Dynamic strategy is a good example of this. It is a combination of a pure fundamental approach with elements of discretionary management and dynamic hedging of the portfolio, but the asset allocation model is based on a 7.5% maximum volatility target. It has produced good results for our institutional clients.”
But another way to apply quant and fundamental approaches together could be to combine discretionary positions and systematic positions within a single portfolio. Although it seems hard to imagine how this can be achieved, it is what many investors do in practice. For example, investors such as PKA and AP2 have implemented a risk factor-based approach. These portfolios essentially combine elements of discretionary and systematic management.
Even the most sceptical about quantamental say this can be achieved. Andrew Dyson, CEO of quant house QMA, says: “If you think about how clients will design their portfolios going forward, it is likely that the core/satellite approach will continue to exist, but the core will be increasingly a computer-driven multi-factor-type portfolio, while the satellite might be very concentrated portfolios that lend themselves to discretionary management. This is a very logical design. The two approaches are complementary.”
Another leading quant house, AQR, has argued that discretionary and systematic approaches are complementary. In a recent paper, the firm said: “Neither systematic nor discretionary managers are inherently superior. Each has the ability to deliver good investment outcomes and […] there is little evidence that one approach is better than the other. The historical correlations between excess returns from systematic and discretionary managers are low, which suggests that many investors may benefit from incorporating both types into their allocations.”
Jacob says: “Quant investment can help in each of the different stages of building a portfolio: screening, scoring companies, deciding how much weight to assign to each stock and managing risk. Quant exists to adjust for human biases, which can be very dangerous.
He says there is little debate on one area: “Quants have trouble measuring regime changes and big disruptions that can span many years. Those signals not present in balance sheet and income statement data,” Jacob continues. “That’s why combining quant and fundamental is a very interesting field, because the players of both sides are polarised and have been at odds in the past. They are beginning to see there is some truth to what the other side is saying.”
Yet Dyson is sceptical, saying: “I think quantamental could be a flawed concept, depending on what it means. If all it means is using quant inputs when building a fundamental portfolio, then that’s just moving with the times, it’s not a revolution. If it means that the actual stock selection process, as opposed to the data points used has elements of both quant and fundamental, then there is a risk of getting caught in the middle. I think it’s not great from a performance perspective and that it doesn’t fit into how clients are thinking about their portfolio.”
“A quant investor will tell you that they are fundamental investors. They are looking at income statements and balance sheets in extreme detail”
The increased use of quantitative tools in risk management of traditional fundamental portfolios makes sense, he says, but that again is nothing new.
Jacob adds: “If you define things on a spectrum, any combination of quant and fundamental is really fundamental. If you are taking away the ability of the machine to make a final decision, then you are reintroducing the potential for human bias. This can be good or bad, depending on the ability of the manager.”
It remains to be seen whether marrying quant and traditional active approaches within the same firm will produce results. Dyson says: “It is questionable whether one firm will be worldclass both at building quantitative and highly-concentrated portfolios. You have to pick a specialism.”
Smith adds: “These are organisational, almost cultural challenges. It isn’t a technology question, it’s a business one, and it’s about how a firm organises its teams, how it brings data scientists and quant teams closer to traditional portfolio management teams. Experienced quant managers may have an initial advantage. We are seeing many firms rushing to build a quant capability and, even better, putting that quant capability close to that traditional asset management capability. That has always been the way we work and we think it’s paying dividends.”
It is clear, however, that these trends are a result of the fast technological progress of recent years. Smith says: “There has been a shift, and it’s related to the availability of new types of data and information, and the potential for new types of computer modelling around artificial intelligence techniques.”
Rapid progress in computing and data science empowers quant investment. Thanks to the proliferation of big data, managers can look for useful information about firms within alternative sets of data that complement widely available financial data. This data can consist, for instance, of language used on news and social media. Also, by using widely available machine-learning techniques, managers have greater ability to find useful market signals. There is a strong case, therefore, for using elements of quant investment in any portfolio. Asset managers that ignore this progress in technology could be left behind.
There are many practical examples of how big data can benefit quantitative investment. BNPP AM’s Panel says the firm is making more use of natural language processing techniques to scan news and social media, looking for live information on firms. Panel explains: “We believe that news and information found on social media are definitely elements to take into account to do what we call ‘nowcasting’, which is different from forecasting. We want to use live information to enrich the quality of the market timing models that we have in our portfolios.”
Panel adds that data scientists employed by the firm are building alternative sets of data based on information available on the web. “They are working on how to use this data to build sentiment indicators on companies,” he says.
Big data is growing in tandem with artificial intelligence and machine learning, because these are the tools that data scientists are often using to examine large databases. But artificial intelligence can also be used to improve on existing models based mainly on financial data. AXA IM’s Smith says his firm is using artificial neural networks, one of the many practical examples of AI, to identify stocks that are likely to surprise in terms of volatility. This is part of a defensive, low-volatility strategy.
Smith explains: “We have always had a tail-risk filter, but we had built the model in a linear way, and the truth is the risk we’re looking at is not linear in its nature. We are looking for fat-tail, black-swan-type events. We are now using a neural network model that learns how to make decisions. We feed it the same information we fed the traditional model, but the results are better, particularly in times of distress. The traditional model gets it right 73% of the times, while the neural network model gets it right 86% of the times.”
The technology allowing to do this is widely available, says Smith. To get computing power, anyone can use online services such as Amazon Web Services to hire extra processing power. Google’s TensorFlow gives access to neural network technology. “Teenagers can build a neural network in their bedroom for a school project”, he says. “Building the model is relatively easy. But the challenge for an investment manager is validating it, monitoring it, getting the data in, taking the signals out and using them in an investment process. That is much harder.”
Quant managers, experienced or otherwise, will always have to make an effort to explain how their models work. Smith says: “We are regularly challenged by those who claim that we are building black boxes. I don’t like that, because I think we work really hard to explain exactly how we value companies, how we forecast earnings and how we make investment decisions. But we do have to work harder to explain how neural networks model work.”
He continues: “It is making good decisions, because we train it and test its behaviour, but we have to go further. It is about our fiduciary duty to our clients. Therefore, a lot of work goes into understanding how exactly this model is doing what it does. We work on visualisation tools, we build proxy models, we are careful with the data so as not to introduce biases. We have to jump through all these hoops before we let it anywhere near clients’ portfolios.”
Perhaps the ‘non-quant’ element of quantamental consists in the ability of humans to make decisions that machines cannot make. As Smith points out, big data and AI have the ability to handle unstructured information, from news stories to ESG data.
But the kind of information that is used as input for computerised models is up to each manager. This is clearly one of the most important human-made decisions. Similarly, computers cannot teach an asset management firm how to make sure their quant teams and traditional portfolio managers get along.