What inspired you to enter the field of quant investing?
“The close connection initially developed through my studies and PhD journey. Being able to combine econometric theory with real-world investments is fascinating to me. At university, the approach is academic by nature and the focus is predominantly on theoretical aspects as opposed to their practical application. Quant investing is essentially an interconnection between the two or a mixture of both worlds. It provides a platform for us to really undertake interesting research which can make a tangible difference for investors.”
Why did you decide to join Robeco?
“When I joined the industry back in 2014 following my PhD, I was aware that Robeco was one of the flagships in terms of quant investing. Naturally, I read a lot of Robeco publications and found their research to be cutting-edge and ahead of the curve. Therefore, I was genuinely pleased when I was approached with the opportunity to join the Robeco Quant Fixed Income team. Moreover, I had previously met several of my present colleagues at different conferences and those interactions left a lasting impression on me. Not only did they come across as smart, but they were also nice, regular people. So I had already gathered that it would be a good professional fit for me.”
“This view has been reinforced in my short time here so far. In particular, the flat hierarchical structure has impressed me, as decisions are made from a bottom-up perspective. So, as an individual and a team as a whole, you have an influential stake in the entire investment process. In my brief experience, I have found that the corporate culture at Robeco encourages people to take ownership of responsibilities, as they are not constrained by top-down decision-making.”
Does this help to create an environment that is conducive to innovation?
“Absolutely. For example, the main projects in our 2022 research agenda were chosen by majority vote. This approach allows team members to come up with their own ideas and, if they can convince the team of their value-add, they stand a good chance of being taken onboard and included in the agenda. This encourages innovation, and the majority vote approach also ensures collective buy-in from the team, which is helpful when carrying out the project.”
How do you see the future of quant investing?
“The field of quant investing is constantly evolving. While the broad, fundamental framework of factor investing generally ties everything together, some of the individual elements frequently change or new ones emerge. For example, investors are increasingly interested in the possibilities offered by novel data sources or techniques and they are also more involved with sustainable investing. But if you are ahead of the curve, perform cutting-edge research, have state-of-the-art infrastructure and employ smart people, then the questions that are posed by constant change are fun to solve. In fact, keeping abreast of the evolution is what gives certain quant investors an edge. If you fail to take into account the future needs of investors, it is likely that you will be left behind.”
“But because of the constant evolutionary process, quant investors have to adapt. They have to reassess their previous findings, come up with new ideas and maybe draw different conclusions. As Robeco quant investors, this makes our job challenging, yet interesting. And I believe the company has the ingredients in place to successfully navigate the changing terrain, as it has a culture that promotes innovation, the right infrastructure, institutional knowledge in quantitative and sustainable investing and a great mix of people who challenge the status quo.”
Stay informed on Quant investing
Receive our Robeco newsletter and be the first one to get the latest insights, or build the greenest portfolio.
Can novel techniques such as machine learning fit into an investment process?
“It is important to note that machine learning (ML) techniques were not originally developed for financial data. These methods typically work really well when you have lots of independent data observations and a high signal-to-noise ratio. But the problem with trying to explain future returns, for example, is that you don’t necessarily have a vast set of independent observations. Securities generally move in the same direction due to their correlation, while the time series return data for individual securities is also typically highly correlated. Moreover, datasets usually have a lot of noise and few signals.”
“That said, ML techniques are useful for specific applications such as liquidity or risk measurement as well as return or volatility forecasting. However, you also need to blend these techniques with traditional statistical approaches and incorporate economic insights into the relevant models to reduce the dimensionality of the problems you are solving. This is easier than solving an unconstrained problem. Therefore, it is important to refrain from blindly applying ML techniques to these applications, but to also incorporate your financial expertise.”
Are traditional factors old-fashioned or outdated?
“Absolutely not. Factors defined by traditional data and methods are still the foundation of financial markets. They are the return and risk drivers for corporate bonds, government bonds, equities and many more financial instruments. But new techniques, such as ML, help us to better understand how factors work and interact with each other. They can extract information from large datasets that would otherwise remain hidden when using traditional techniques.”
“This additional information can provide more clarity on how to exploit factors more effectively to achieve better risk-adjusted performance. Moreover, it can reveal some of the shortcomings of factor investing. Although factors deliver a premium over the long term, they are susceptible to periods of underperformance. Thus, these methods can uncover some of the less understood risks associated with factors.”
Why are new data sources integrated slowly into quant models?
“Alternative datasets are still relatively new and need to be studied carefully. They usually do not have a long history, so it is really hard to test whether they are useful or not. To do so, you need at least a full investment cycle to evaluate whether new data signals work well as standalone factors or complement your model. Furthermore, it is important to understand whether these signals perform better or worse in varying economic conditions and why that is the case. Also, many datasets offer limited breadth, as they are only relevant for small parts of the investment universe – like satellite images. This makes them less applicable for quant models, which require breadth and long-term empirical evidence. Thus, being cautious is better than being wrong.”