For over two decades, Robeco researchers have been at the frontier of quantitative investing. They were among the first to document factor premiums in emerging markets; to report a worldwide low-risk anomaly; to fully integrate ESG into quant portfolios; and to exploit factors in corporate bond markets. In recent years, we also invested considerable resources in innovative technologies, leading to concrete advances in the integration of alternative data, artificial intelligence (AI) and cloud computing into the investment process. We talked with four researchers that have been involved in the latest developments in these areas.
Weili Zhou: “Not in the foreseeable future. True, the industry has seen massive amounts of money flowing into factor-based strategies. Also, the performance of some factors could be under pressure for some time. People often present these two arguments to advocate for alternative signals and strategies, with the concern that “traditional” factors may be overcrowded. ”
“But, as active managers, we see this more as an opportunity than a threat. We think crowding issues have more to do with the way factors are implemented in generic strategies. These often mean following the simple textbook definition of factors, investing in transparent public indices, or implementing trades on a few days in the year.”
“Top-notch strategies require continuous enhancement of signals, optimization processes, and execution. And innovation is a crucial way to strengthen our offering and to bring scalability to our business. Here, we don’t mean that machines can fully replace humans in research. But, under human supervision, machines can help detect or even explain new patterns that are often non-linear.”
“Machines can also make research production much more scalable. Thanks to AI and automation, what could be done in a day, may be done in an hour all this explains why Robeco is investing so heavily in these areas. This is really a broad-based effort, involving many more people than just the researchers within the company.”
Thom Marchesini: “We currently use several alternative data sources in our investment process. For example, we use a news sentiment signal based on aggregated information of media text analytics. This signal recommends our stock-selection model to buy stocks that have been positively mentioned in news articles, and sell stocks that have had a less positive coverage in the media.”
Before including new signals, we always take a critical look first
“But before including these signals, we always take a critical look first. In the news sentiment example, we noticed that the naïve aggregation of news sentiment leads to undesired biases. Companies that are positively mentioned in the news tend to be larger and more expensive stocks. After correcting for these exposures, we concluded that the news sentiment signal is a diversifying addition to our momentum strategies, in both developed and emerging markets.”
Thom Marchesini: “There are numerous alternative data providers and it is virtually impossible to evaluate them all. So, it is up to us to separate the wheat from the chaff. In that process, we stick to our investment philosophy: evidence, rationale, and prudence. We will not include new data sources into our investment decision process based only on beliefs. The source has to be backed by strong empirical evidence and sound economic rationale. If we find that it does not meet our standards, we will refrain from using it.”
Bastiaan van Gaalen: “Robeco has developed and enhanced its proprietary optimization algorithm to achieve the highest net performance. The algorithm runs based on hundreds of parameters, each of which has a different impact on the outcome. Normally, when we develop a new strategy or customize a solution, parameters are re-evaluated and adjusted.”
“In the past, we did this in a heuristic way, based on in-house experience. But thanks to the large increase in customization requests and our own ambitions in terms of innovation, we have started to introduce some clever machine learning algorithms to facilitate the search for optimal settings. In this large algorithm design space, it is not desirable to do a grid search due to the curse of dimensionality.”
“Therefore, we took inspiration from the latest developments in parameter tuning for machine learning algorithms, in other words Bayesian and evolutionary guided searches, that naturally converge to a solution based on previous results. We eventually settled for one that was the best in avoiding local optima.”
AI is competing with our most experienced researchers and is now part of our standard backtests
“We have tested this technique in a controlled environment, for example, for a customization with strict ESG constraints, for which you need to loosen certain constraints to be able to come to feasible solutions. AI is actually competing with our most experienced researchers and is now part of our standard backtests.”
Wouter Tilgenkamp: “Many academic papers use monthly rebalancing of portfolios to test the existence of factor premiums. This is still common practice in the asset management world But equity markets are open every week, five days a week. So, if we kept trading once a month only, we would cut ourselves short in finding the best opportunities.”
“This is why we’ve enhanced the investment process by scouting markets throughout every trading day for opportunities in stocks we deem attractive to buy or sell. More recently, we completed the implementation of so-called wish lists algorithmically. These are lists of stocks which we can trade when an opportunity pops up at any point of the day.”
“To do this in an efficient manner for hundreds of quant equity accounts, we needed scalability. Project ‘Obelix’ (Opportunistic Block Liquidity Xapture) was launched to continuously look for favorable block trades, with little market impact. This opportunity-searching algorithm now runs on the Cloud and we have seen a rapid increase in the amount of trades being executed, resulting in substantial savings for our clients.”
BY CLICKING ON “I AGREE”, I DECLARE I AM A WHOLESALE CLIENT AS DEFINED IN THE CORPORATIONS ACT 2001.
What is a Wholesale Client?
A person or entity is a “wholesale client” if they satisfy the requirements of section 761G of the Corporations Act.
This commonly includes a person or entity: