2021 | On Quants and Data Scientists in Finance

Over the years, I have often wondered what the difference is between quants and data scientists. These are both popular roles in Financial Institutions of various kinds.

I've worked in both roles at large global banks, as a quant analyst on a FICC desk on the buy side, and as Head of Data Science doing global securities research on the sell side. I've also spoken with a number of smart people in Finance about the difference, and usually the answer is that there isn't too much to distinguish the two roles. 

Similarities

The similarities are obvious: 

1. Both roles require a deep understanding of advanced mathematics, statistics, finance, etc.
2. Both require an ability to code, and code well, potentially in Python, R, C/C++/C#, maybe Julia
3. Both require that you are able to model things to reflect real world realities with good fidelity
4. Both require that you are able to articulate and explain things to non-quantitative audiences well
5. Both have some of the hardest interviews to get in

But what about the differences? 

As a data scientist, I've given talks at the Quant Forum (gathering of all quants) at a bank on data science methods, and it was the greatest fun I had giving a presentation. I felt right at home there. And received several collaboration requests from that session. So the overlap cannot be understated. But these two roles are different, and the differences only became clear to me as I read this book: 

 
Dave Epstein discusses the central thesis of his book in this video

The Concept of Range, and the Different Types of Domains

In this book, the author explains that there are some "kind" domains that allow for deep specialization, where experience helps you make better decisions and more effectively deliver value. The more time you spend in the field, the better you get at it. 

There are also "unkind" or "wild" domains where the opposite holds true. In some of these, having more experience might give you more confidence, but unless you are humble, leveraging old learnings in new problem contexts can lead to more errors i.e. experience doesn't really help in problem solving. 

He illustrates this idea with several examples - from sports using golf and Tiger Woods as an example of a "kind" domain, and with tennis and Roger Federer as an illustration of the "wild" one. He further underscores his point with Chess - grandmasters can reproduce a board with several pieces exactly with only 3 seconds to see it IF the pieces are placed such that the scenario can occur in a real chess game. However, if random pieces are randomly placed, they cannot replicate the board unless given significantly more time to study it, and in the latter case total chess novices can do a better job.

Another example he cites is Music. Did you know that Yo-Yo Ma, the world-renowned cellist tried several other instruments before settling on the cello? This works because music is a "wild" field - you can pick up broad patterns that might work generally, but then again, might not. Then when you specialize in a narrow "kind"-er domain (that of playing the cello in this case), you try those patterns to go deep and excel, keeping what works and discarding the rest.

So how is all this relevant? 

As a data scientist, even if you pick just one field, say finance, you develop and use lateral skills a lot. You may work with consumer spend data and basket analysis for retail, with Fed Beige Book data for global macro, with satellite data for real estate or oil/energy markets, with news data for other verticals, etc. You have a wide and deep knowledge, but you pick up sector knowledge as you go, working closely with specialists.

As a quant, you (usually) work on one desk, and specialize in one or two markets. You learn all you can possibly learn about those markets, and go as deep as you possibly can to identify, isolate (where possible), examine, model, and forecast particular variables and the relationships between them. You are a specialist in your sector.

A data scientist may not need stochastics, but should have the intellectual capacity to learn it. Most quants should know enough stochastics to be able to do their jobs well.

In Conclusion...

Both the quant as well as the data scientist roles in finance offer stimulating, intellectually challenging work. I guess in the end, your ability to make a strong positive contribution comes down to your preparation, expectations, firm culture, and the competence and quality of the team you work with.

How does this post match with your views? Please share in the comments below.