Evaluating the information in the Federal Reserve stress tests

We find evidence that the Federal Reserve stress tests (CCAR and DFAST) produce information about the stress-tested firms as well as other, non-stress-tested banking companies. Although standard event studies do not always show abnormal returns for the stress-tested sample on average, we argue that such tests are ill-suited for this sort of information event. Using a different empirical approach, we show that around stress test announcement dates, the absolute value of the cumulative abnormal returns (|CAR|) of stress-tested bank holding companies averages almost 3 percent. Cumulative abnormal trading volumes are more than 1 percentage point higher than a market model would predict. Absolute value abnormal returns and volumes are higher for more levered and riskier firms. We explore several theoretical hypotheses outlined in Goldstein and Sapra (2014) but find no evidence of negative welfare costs associated with the disclosure of stress test results.