Dr. Joseph Simonian Sheds Light on the Future of Quantitative Finance
RR: What trends in the quant world have caught your attention recently?
Overall, there is an attempt to keep up with the latest development with AI and ML. Because many quant firms (especially legacy managers) both small and large have adopted these technologies and techniques. They’re trying to keep up / catch up because their investment process needs serious revamping. And of course applying data science to finance.
Factor strategies, there have been various firms that have performed differently. Many firms run similar factor strategies. There have been papers defending some factor approaches, but i think this is misguided. Running factor strategies require self-reflection on processes. Why do some work, why do some not work? It reinforces successful strategies.
ESG investing as it pertains to quants and how ESG considerations can be incorporated?
Alternative Data, especially in the ML and data science world. Many firms are trying to build signals to compliment existing strategies with Alt Data.
RR: If you had to choose one or two, what is, or are, the biggest unsolved problems in quantitative finance? And what progress, if any, have we made?
This is more of a methodological problem. People are asking, do certain factors still work? But, one must first understand that finance is not like natural science. First, there is a much faster evolution. Second, certain factors may have behaved a certain way for only a short while. So the problem investors are trying to solve is whether factors still work.
The problem should be how to position in finance so that no matter how the world evolves we profit and how to set up and build models in finance that adapts as input data is adapted. Science is more or less static. But, in finance and econ, variables evolve much more rapidly. One has to build models that will adapt to the data as it changes. People still rely on the same static models.
ML and data science has the potential to evolve as data changes. Another unsolved problem in practice for me is how to incorporate quantitative process data that is very important but not always amenable to quantitative information. Such as fundamental views, geopolitical risk, etc.
Unless you want to adhere to a wholly quantitative process, there are limitations, it's confined. You need good discipline to know what to let in. How to mix quantitative and qualitative inputs?
This is the challenge a lot of investors have.
Going back to AI and ML; NLP may have a formal consideration of qualitative information and could be one way of incorporating qualitative information Those are main practical problems, in theory, host of problems.
RR: In your opinion, what role will machine learning play in the future, and how might that affect the markets?
ML will play a bigger role in various aspects, we already see a large role in terms of automated processes. ML is also playing already a role in terms of reviewing documents, ie. contracts or other compliance documents.
On the risk management side, there is a lot that ML techniques can add in terms of source of risk and managing risk. Identifying the primary drivers of portfolio / strategy risk is crucial.
ML is very good at picking out patterns and drivers, several firms are using ML to derive alpha and signals in a more robust way. In quant shops that survive and thrive, more ML in alpha or risk management processes is what is occurring.
More traditional firms will find themselves at a disadvantage. Already differences in performance in quantitative firms vs firms that have not evolved to data science and ML yet are becoming apparent. And the reason is because a lot of traditional statistical models are not good at predicting the future. Statistical models are much better at explaining the past, econometrics and forecasting was a second function
ML is very good at predicting, but not at explaining the past. That’s why ML is a huge player in finance.
RR: What are your thoughts on the passive investing boom?
Passive investing as a whole plays a role in terms of you may want to have a passive position in the market for a good reason to get a basic exposure to an asset class. In general, active management provides a lot of functions that passive strategies can’t provide, such as risk management. Passive strategies have no drivers. With active strategies, the team can manage dislocation of markets.
Statistics says active underperforms passive, but that only says that they are different. It is difficult to fashion a process that outperforms the market in the long-term, but they exist!
Passive has a role, but a good active manager provides good value
Written by Gihyen Eom
Edited by Alexander Fleiss