We provide a systematic, data-driven methodology for choosing test scenarios among a set of potential scenarios. The test scenarios can be used to accurately estimate measures of tail risk of financial institutions, such as conditional value at risk (CVaR), and can also simultaneously be used for stress testing, that is, to identify conditions for large losses. We validate the methodology on historical data used in stress tests by the Commodity Futures Trading Commission and the Federal Reserve and connect it to the design of experiments methodology with a risk-based objective. The methodology does not require detailed knowledge of financial institutions’ portfolios and can aid regulators in evaluating the resilience of multiple institutions using uniform risk assessment standards. This paper was accepted by Giesecke Kay, finance. Supplemental Material: The electronic companion and data files are available at https://doi.org/10.1287/mnsc.2024.06126 .
Arora et al. (2026) studied this question.