Do Fraudulent Firms Engage in Disclosure Herding?

The following post comes to us from Gerard Hoberg of the Department of Finance at the University of Maryland and Craig Lewis of the Finance Area at Vanderbilt University.

In our paper, Do Fraudulent Firms Engage in Disclosure Herding?, which was recently made publicly available on SSRN, we present two new hypotheses regarding the strategic qualitative disclosure choices of firms involved in potentially fraudulent activity. First, these firms have incentives to herd with industry peers in order to escape detection. Second, these firms have incentives to locally anti-herd with the same peers on specific aspects of disclosure consistent with achieving fraud-driven objectives. We use text-based analysis of firm disclosures and compare disclosures across firms involved in SEC enforcement actions to benchmarks based on industry, size and age, and also to each firm’s own disclosure before and after SEC alleged violations.

We hypothesize that firms involved in potentially fraudulent activity face tensions when providing qualitative disclosures to the Securities and Exchange Commission, the agency tasked with enforcing anti-fraud laws. Our focus is on the Management’s Discussion and Analysis section of the 10-K, which is where managers have a high level of discretion to describe the key issues facing their firms and to describe their performance in detail. A primary motive is to escape detection, and managers who assume that the SEC is less likely to scrutinize disclosures that resemble industry peers, or that such disclosure is less likely to raise red flags, have incentives to herd with industry peers. On the other hand, the same objectives that lead managers to commit fraud may also provide incentives to anti-herd in their disclosure from industry peers. However, these latter incentives are likely more localized, and anti-herding would be predicted only on disclosure dimensions that might help managers to achieve these objectives.

We find especially strong support for the localized anti-herding hypothesis, and moderately strong support for the industry herding hypothesis. Indeed the same firms can choose disclosure strategically that both resembles industry peers overall, yet deviates from industry peers on a small number of important dimensions. These results provide strong support for the conclusion that strategic disclosure incentives are strong for firms involved in SEC accounting and auditing enforcement actions (AAERs). We can separately measure both types of disclosure because industry herding is directly tested by considering similarities between a firm’s raw disclosure and that of its industry, size and age matched peers. Localized anti-herding is then measured using the clustering properties of abnormal disclosures (raw disclosure adjusted for that of industry-size and age peers).

Our results are particularly striking along two dimensions. First, our identification examines firms involved in AAERs both relative to firms not involved in AAERs, and also relative to the same firms involved in AAERs but for the years before and after the AAER-influenced years as indicated in the AAER filing announcements. These results suggest that disclosures are revised frequently as a firm evolves from a pre-AAER firm, to a firm involved in AAER actions, to a firm that has been revealed as allegedly committing fraud. Content analysis reveals much granularity regarding the discussions firms disclose over this cycle. In particular, firms involved in fraud focus more than average on complexity including acquisitions and international words to potentially conceal fraudulent activity. These firms also discuss issues relating to uncertainty, litigation, and speculative statements more than their peers do. After firms are revealed by the SEC to allegedly have committed fraud, their disclosure is revised materially and focuses more than average on describing the SEC investigation itself in great detail.

The full paper is available for download here.

Both comments and trackbacks are currently closed.