The Ability of SEC Investigations to Detect Misconduct

We provide a point estimate of the probability that an SEC investigation detects misconduct. To do so, we develop a method that applies Bayes’ Rule to cases where two separate SEC investigations examine the same company at the same time. The method gives a point estimate of 0.370, which means that when a company is committing misconduct, each of the two SEC investigations has a 37.0% chance of detecting it. When we relax the method’s identifying assumption, we show that each investigation has at most a 61.4% chance of detecting the mis conduct. In additional tests, we modify the method to estimate different detection probabilities for investigations with different characteristics. With this modification, we estimate that the SEC’s detection probability significantly increases when its staff are less busy and when the investigation is conducted from the DC office. Both results indicate that detection improves when investigations have more resources.