SunGard: Macroeconomic Stress Testing - How BancorpSouth Built and Implemented a Strategy
Posted: 6 January 2014
Starting this fall, all U.S. banks with $10 to $50 billion in assets must conduct annual stress tests of their full balance sheets using a set of three macroeconomic scenarios prescribed by their regulators under the Dodd-Frank Act.
This deadline represented a challenge for the management team at BancorpSouth, a $13-billion-asset regional bank headquartered in Tupelo, Mississippi. To meet this challenge, the bank spent the first half of 2013 in close partnership with modeling consultants and solution experts from SunGard, building and implementing an enterprise-wide stress-testing program with the dual aim of making sure it could run the regulators’ scenarios and also improve management information regarding the effect of severe stresses on the institution.
The program was necessarily ambitious, embracing all the bank’s main business lines, including middle-market commercial and industrial (C&I), small business, commercial real estate (CRE), construction/development, and consumer loans. There was no gold standard for stress-testing programs and no well-worn route to follow - especially for relationship-oriented regional and smaller banks. Owing to the open-ended nature of the endeavor, strategic decisions confronted the bank’s project team from the start.
This article tells the story of how BancorpSouth found its way toward some answers, tackling several key issues such as:
- Whether to build a top-down or bottom-up modeling framework, given the associated data challenges.
- How to integrate the stress-testing framework by leveraging and building on the bank’s existing core credit and balance-sheet management systems and processes.
- Ensuring that qualitative factors, management judgment, and governance concerns would receive as much attention as the quantitative modeling.
The emphasis here will be on the bank’s credit risk modeling and its integration with the overall balance-sheet management framework.