Quant Risk and Regulations - a potpourri of multiple disciplines

Mid-90s witnessed the influx of Physics and Maths PhDs to Wall Street in large number. The likes of Goldman, Merrill Lynch and Morgan Stanley would make beelines to the pristine campuses of Columbia, Harvard and Wharton to hire their Quant PhDs as associates to do the complex Maths behind Derivative models.

Many of the academically oriented quants (Maths and Physics PhDs) decided to take a shot at this new-found career option with the I-banks to try out what they learnt in the classrooms taught by the likes of Nassim Taleb and Jim Gathedral over the years.

Quants would come to office every morning (at times during weekends), do some stochastic calculus for computing volatility of the stocks or for building derivative models, and take out a fat package back home at the end of the pay cycle! They would top it up with exotic holidays with regular intervals across the globe!

It was indeed a paradise on earth for most of them.

No wonder, they were known as the masters of the universe around those days!

However, quant world is not bereft of idiosyncrasies and conundrums though!

While donning the hat of model risk practitioners (a stimulating area in the high-voltage world of regulation today), we often tend to forget the origin/history of the models in question.

For instance, take a look at the concept of random walk, which is ubiquitous in the natural sciences.

The first theoretical description of a random walk in the natural sciences was performed in 1905 by Einstein in his famous paper dealing with Avogadro number. Subsequently, the mathematics of the random walk was made more rigorous by Wiener (the by-now famous Wiener process used in Black-Scholes’ pricing equation).

The random walk concept has spread across almost all the areas of natural sciences over the years!

Interestingly, the first formalization of a random walk was not in a publication by Einstein, but in a doctoral thesis by Bachelier, a French mathematician, whose advisor was Pointcare, one of the greatest mathematicians of his time!!

The thesis was about pricing of options in speculative markets! At the risk of going on a tangent, we can’t perhaps emphasise more on the source/origin of the fundamental nature of the models being used in varied contexts!

Let us be mindful of these nuances always!

During my student days in Europe, I was a member of an inter-disciplinary group called CINEF (Centre for Inter-disciplinary Research on Economics and Financial Engineering). Economics, Physics Computer Science and Psychology students would sit and discuss the business problems together and take a very holistic approach to come out with a practical solution. I remember having done it for leading investment banks on a regular basis and coming up with very creative solutions. All of us, the enthusiastic lot of students used to gain enormously in the process of interacting with students from other disciplines.

When I came back to India to pursue a career in Analytics 2 decades ago, that trend was still new and confined to the corridors of only select few global banks! Recruitment used to be mostly stereo-typed and old-fashioned tilted towards the disciplines of Computer Science and Finance in silos!

Fortunately, the trend seems to have taken an interesting turn over the last decade.

Today, I am happy to see so many youngsters contributing to the field of Risk management, having completed degrees from multiple fields, inter alia, Economics, Physics, Statistics, mathematics, Data Science and the dual-degrees in Engineering and Data Science from prestigious campuses in India.

I believe this trend will continue due to the complex nature of requirement, particularly in mathematically demanding disciplines like risk management and regulatory modelling which needs quantitative rigour with solid precision and accuracy. A quant desk where a programmer discusses the nitty-gritty of Comprehensive Capital Analysis and Review (CCAR) regulations and the Pre Provision Net Revenue (PPNR) models, or multiple approaches to tackle Counter party Credit Risk (CCR) with the statisticians, economists and the mathematicians works out well in terms of meeting regulatory compliance with speed and accuracy.

Today, in a world driven by AI/ML with increased scrutiny on data, demand from regulators like PRA, APRA and Federal Reserve has shot up substantially. Even regulators in geographies of Middle East like CBUAE have been very demanding in terms of the kind of models built including the type and the usage of data by the banks in the region. This is particularly true for regulations like IFRS9.

Over the years, I have also noticed an interesting tendency among the programmers to limit themselves by rushing to arrive at the output quickly without diving into the nuances of a particular model, esp. the associated domain aspect.

A simple example is running a SAS or a Python code for building a regression model without paying any heed to statistics/mathematics behind the model.

On the other hand, mathematically trained quants often tend to ignore working on the nitty-gritty of data.

However, I always say that without love for data, an analytics practitioner can’t be a complete professional!

In fact, successful modelers always have a good grip over the data being used. If we recall the old adage, “Garbage-in Garbage-Out”, and the model is as good as the data, we will go a long way in creating an army of successful AI/analytics professionals, who would always be in great demand in this technology-driven world! This is particularly relevant today in the context of crucial data regulations like GDPR and CCPA confronting the organisations across the globe.

Without an iota of doubt, we have come a long way!

Today, banks have been dealing with model risk regulations in US like SR11-7 which emphasises on the right usage of the models by the banks. During the process, the vital area of model risk, a discipline that involves building, rebuilding and validating the models thoroughly has evolved significantly encompassing the data being used along with the models in question.

In an era of AI/ML and Generative AI, without being enamoured by it, we need to be cautious of the fundamentals behind the models being used.

In conclusion, appropriateness of data being used, and the fundamentals of the models are always the keys to the applied world, esp. with respect to the banking regulations!

---

About the Author

Diganta Saikia consults banks globally in the areas of risk management and regulatory compliance, with deep expertise in model risk, data governance, and quantitative analytics. The views expressed in this article are personal and do not represent those of any institution.

For feedback or inquiries, Diganta can be reached at s_diganta@yahoo.com.