A global investment bank had fared relatively well during and immediately after the 2008 economic crisis, thanks to strong and decisive management action. But the crisis had put the bank’s market-risk models to the test, revealing considerable issues with the accuracy of their projections. In response, the chief risk officer launched a 3-year program to transform the bank’s market-risk tools and supporting infrastructure, with a focus on redesigning the economic-capital model, the Value-at-Risk (VAR) model, and stress-testing frameworks. The program—steered by senior managers from the risk, front-office, finance, operations, and IT functions—became one of the bank’s highest-priority initiatives.
The bank asked McKinsey to lead certain aspects of the transformation—specifically, the redesign of the VAR model and stress-testing frameworks—as well as to help set up and scale up the program, run the program-management office, and ensure stakeholder alignment.
The overhaul of the VAR model involved a major shift for the client: from Monte Carlo simulations and sensitivity analysis based on data residing in the risk function’s systems to historical simulations and full revaluation based on data residing in front-office systems (which is the model already used by most banks). Among other benefits, historical simulations create a tangible link between VAR and real-world events, while full revaluation allows for more accurate assessment of complex products with nonlinear payoff profiles and large exposure to “tail” events. The McKinsey team structured the project by creating four working groups: methodology, market data, IT implementation, and process. An early pilot—in which the team tested the new model—served as a proof of concept.
To revamp the bank’s stress-testing approach, the McKinsey team implemented a framework that defined a wide range of “shocks” that could affect the economy. To create a customized stress-test scenario, client managers mix and match these shocks at different severity levels via a user interface. Earlier, setting up and running a stress-test scenario had been a manual process that took up to 2 weeks. The new framework allowed the client to produce stress-test scenarios in 1 day—and thus to use stress tests as a tool for day-to-day risk management.
In addition to defining the methodology for the VAR model and the stress-testing framework, the McKinsey team worked with the bank’s IT project managers to establish an operating model and implement IT systems to support the new tools. One major goal was to ensure that the bank used the same systems when calculating risk measures across the entire portfolio, resulting in a consistent market-risk outlook.
Throughout the project, the team kept all stakeholders informed of important developments and ensured they were aligned on program goals. In addition, the team helped the bank address interdependencies between the market-risk transformation program and other initiatives.
After setting up the program and supporting implementation for 12 months, the McKinsey team transitioned leadership of the project-management office to internal change agents at the client.
The bank used the new historical-simulation model for initial VAR estimates in a successful pilot program involving a portfolio of interest-rate products. The new model reduced process and system complexity. It also enabled front-office managers and risk managers to assess the incremental effect of individual trades, facilitating the selection of optimal hedging strategies. Pending regulatory approval for the new model, within 2 years, the bank plans to be using historical simulations to assess all elements of market risk for the entire portfolio.
Implementation of the stress-testing framework is progressing as planned, with high involvement from the market-risk managers for all asset classes. The new IT components and operating model are already partly in use and will be further developed over the next 2 years.
We are continuing to work with the client on related projects, including a review of its pricing-model control processes and implications for model risk. (The appropriate use of pricing models is crucial to the accuracy of VAR estimates.)