AFTAs 2020: Best Risk Management Initiative—Morgan Stanley

Many stress-testing tools were put through the wringer this year due to the coronavirus pandemic. Despite that, Morgan Stanley delivered a stress loss approximation capability, a key addition to its risk department toolkit. The bank built the Stress Testing Engine (STE) to help its risk management teams understand portfolio stress loss impacts under a given market scenario with a quick turnaround time. STE performs stress loss approximations using a wide range of methodologies across all market risk factors.

A spokesperson for the firm explains that the engine played a key role in prioritizing risk management use-cases that need quick turnaround times and flexibility. The engine uses a mix of risk measures combined with a high-performance computing framework and estimation techniques. 

The design of the new infrastructure is intended to address various business and technology challenges related to data volumes, computational resources for scenario-based pricing models, as well as time-to-market for new scenarios and new measures. 

Morgan Stanley made a few technical design decisions, one of them being transitioning to a massively parallel processing database platform. This allowed it to provide more granular risk measure information, capacity for huge amounts of data needed for analysis, and more accurate shock calibration. The new platform also features a subscription API and utilities to extract data more efficiently. The results are made available at highly granular levels to assist in analysis and at aggregated levels for senior management and regulatory reports.  

It took Morgan Stanley about 18 months to build and roll out the new infrastructure. It used Agile tools and techniques to support incremental delivery, including adopting DevOps guidelines to automate delivery and improve testing timelines by eliminating manual steps.

Looking ahead, the firm will invest in enhancing the workflow to design and estimate a scenario for a given portfolio using in-memory computations, which will speed-up decision making. It will also enhance estimation methodologies to improve complex risk capture.