Content area
Full text
Prior to the global financial crisis of 2008, stress testing was typically used as an adjunct to statistical approaches to risk measurement and management (e.g., Value at Risk [VaR] and ex ante tracking error) to quantify the profit and loss (P&L) associated with potential tail events. The perceived mathematical sophistication of these models lulled some risk managers into overlooking the limits of relying on models calibrated exclusively with historical data to forecast future market risk.
The extreme market moves during the financial crisis exposed the limitations of standard risk models and highlighted the need to augment their insights. Post-financial crisis, market risk has become increasingly difficult to forecast as prolonged monetary policy intervention by central banks and sudden political shocks have overtaken economic fundamentals or technical data in driving financial markets.1 These policy shocks have triggered sudden regime shifts and breakdowns in historical relationships between market variables.
Given the unpredictability in the current market environment, scenario analysis provides a critical complement to VaR and other related statistical risk measures. Scenario analysis forces risk managers to think about what may happen in the future, create direct and explicit links between changes in the macroeconomic environment and financial markets, and then apply them to portfolio exposures to determine hypothetical investment outcomes. In contrast to purely statistical or risk models, scenario analysis has the singular virtue of being forward looking, even at the risk of being less scientific. This has led regulators to increasingly emphasize scenario analysis (i.e., stress tests) as an important element of the supervisory process.2 However, this virtue proved to be a double-edged sword because, unlike more traditional historical approaches, there is no established standard or framework for constructing scenarios. In fact, the greatest challenge in stress testing is how to effectively define and generate hypothetical yet plausible stress scenarios. Furthermore, research on best practices in scenario generation has been limited.3
In this article, we describe a market-driven scenario (MDS) framework designed to mitigate the subjective and often ad hoc nature of hypothetical scenario generation. In the MDS framework, economic forecasts and market views are collected from a wide number of firm constituents, including risk management, investor, and economic research teams. These views are then distilled using a disciplined process that incorporates...