Lynn Strongin Dodds looks at the pros and cons of quantamental analysis.
Unearthing promising prospects is never easy but the past three years have been particularly challenging with a war in Ukraine, skyrocketing inflation and central banks tightening. The tried and tested research tools do not always deliver the goods and increasingly many are combining the best of both man and machine.
Instead of conducting research through separate prisms, the so called quantamental process incorporates fundamental research. The theory is that using computers to make the trading decisions will remove the human-error component, including emotional and/or cognitive biases However, not everything can be modelled as markets are dynamic and history does not always serve as a guide to the future. Equally as important, the algos are designed by humans, which means they are as successful as the developers behind those strategies.
On one side, there is the traditional fundamental bottom-up picking individual stocks based on criteria such as earnings, cash flow, capital expenditure, and market share. Analysts hope to exploit any mispricing to generate alpha or the excess return of an investment in relation to the return of the market or a benchmark.
On the other side, it also involves quantitative analysis which employs computer models and algorithms, as well as vast quantities of data, to determine trends and patterns. The objective is to predict future security price movements. The machines will also digest information more quickly than people and be able to weigh recent information more heavily than past data.
“Having fundamental thinking and intuition behind a quantitative approach is a vital pillar of our investment philosophy – the linkage between data and stock return performance expectations must have strong investment intuition, not simply be an observed statistical pattern in data without understanding,” says Olivia Engel, chief investment officer, active quantitative equity at State Street Global Advisors
The approach is not only the preserve of the equity or fixed income domains. For the past two years, State Street Global Markets have been experimenting with an approach called ‘Mind and Machine’ to make selections in FX currency pairs, according to Michael Metcalfe, head of macro strategy at the Boston based fund manager. “Using data on trend, valuation, equity returns, interest rates, market turbulence and investor positioning, we deploy three distinct machine learning techniques, random forest, boosting machine, and neural network to the challenge of picking the direction in FX rates over the coming month,” he adds.
He notes that each machine has a slightly different approach to the task, “just like a team of portfolio managers might have; one takes a top-down approach to understanding how FX rates are interacting with the variables, another focuses on error-correction, while the third takes a granular bottom-up view of the individual relationships.”
Raising the profile
Over the past 20 years, many buy side firms had deployed some version of quantamental analysis to combat the low, benign interest rate environment that plagued markets. However, BlackRock put it firmly on the investment map in 2017 when it announced it was making alterations to its research framework. The behemoth fund manager took roughly $30 billion in assets or around 11% of its active equity funds that were designated for a fundamental stock-picking method and incorporated a quantitative approach based on modelling and algorithms investing.
At the time CEO Larry Fink told reporters at the New York Times that “the democratisation of information has made it much harder for active management. We have to change the ecosystem – that means relying more on big data, artificial intelligence, factors and models within quant and traditional investment strategies.”
Nabeel Abdoula, deputy CIO of Fulcrum Asset Management also points to two key drivers for its growing popularity – a decisive empirical turn within modern academic macroeconomics and increased availability of cheap computing power to crunch the numbers.
“The average economics or finance paper twenty years ago was investigating highly stylised theoretical models detached from real-world data, and often required simplifying assumptions to be able to run on 1990s computers,” he adds. “Today, more than half of papers published in the top academic journals are empirical, and econ PhDs are also expert data scientists and coders, assembling datasets from disparate sources and testing their theories using high-performance computational methods.”
He notes that “the events of 2022 are further evidence of the difficulty in making short-term market predictions. Not only is it impossible to know for sure what macro events lie ahead, but also how markets will react to the things that happen.”
Engel also believes that “today, with advances in non-financial data available to inform investors, new ways of generating metrics that capture a fundamental investing idea in a systematic way are possible and can sometimes be uniquely applied to smaller groups of companies in similar industries.”
She adds, “For example, data about a biotech or pharmaceutical company’s clinical trial programme can systematically inform a view on which companies have research that has a higher likelihood of success. Additionally, data about brands and patents and can help to value intangible assets that are an increasingly important driver of stock market returns.”
The sustainability effect
Environmental, social and governance investing has also sharpened the focus on incorporating both disciplines, according to Raul Leote de Carvalho, deputy head of the Quant Research Group at BNP Paribas Asset Management. “We use both qualitative and quantitative inputs, and this is particularly useful for sustainability criteria which is a core component of our investment process,” he adds. “The scores are heavily quant based. We integrate ESG criteria into all our investments. However, portfolio construction is not just a black box, and you can’t just use what you learn in academia to solve a problem.”
De Carvalho notes that if BNPP AM can’t find the right tools it will build the solution itself. For example, the French fund manager has a proprietary model and dataset of corporate carbon emissions that was developed using machine learning. The reported data covers only about 20% of the company’s investment universe and the model can already predict Scope 1 and 2 emissions for the other 80%.
As with any method, there are challenges. The quality of data can vary and collecting the right information can be difficult. “In Europe, there are many exchanges and liquidity is highly fragmented,” says Max Hilton, Managing Director, Quantitative Solutions, MJ Hudson. “We tend to receive external vendor and stock exchange data but to get a better understanding and depth, you need to have an aggregated view and that can be challenging from a data perspective.”
There are also those who are happy to just stick to one or the other approach. In one camp there are the are the quant or hedge funds who prefer their high-powered models while the other comprises the fundamental based stock pickers.
“In a nutshell we believe that you can achieve better returns over a full market cycle by being a long term, valuation-driven investor using fundamental research to form independent insights and aim to build portfolios with diversifiers,” says Mark Preskett, Senior Portfolio Manager, Morningstar Investment Management Europe.
He adds that its “capital markets teams build assumptions for 200 stock groupings and 150 bond groupings – and for those groupings that are priced well below their intrinsic value, we carry out in-depth fundamental research. Those asset groupings that pass muster, we invest with conviction.
To ensure we have sufficient diversification within the portfolio, we rely on internally built tools and systems to help answer specific questions. For example, how do European financials typically behave in periods of high inflation? How do the high margins achieved by US IT stocks compare to history?”
