Designing Stress Scenarios
We develop a tractable framework to study the optimal design of stress scenarios. A risk-averse principal (e.g, a manager, a regulator) seeks to learn about the exposures of a group of agents (e.g., traders, banks) to a set of risk factors. The principal asks the agents to report their outcomes (e.g., credit losses) under a variety of scenarios that she designs. She can then take remedial actions (e.g., mandate reductions in risk exposures). The principal's program has two parts. For a given set of scenarios, we show how to apply a Kalman filter to solve the learning problem. The optimal design is then a function of what she wants to learn and how she intends to intervene if she uncovers excessive exposures. The choice of optimal scenarios depends on the principal's priors about risk exposures, the cost of ex-post interventions, and the potential correlation of exposures across and within agents.