The evolution of 10-K textual disclosure: Evidence from Latent Dirichlet Allocation

Corporate disclosures in 10-K filings have become significantly longer, more redundant, and less readable over the past two decades. This study investigates the drivers of these trends, specifically examining whether they stem from changes in firm fundamentals or from evolving regulatory requirements.

The authors analyze 75,991 10-K filings from 1996 to 2013. They employ Latent Dirichlet Allocation (LDA), a machine-learning technique, to identify 150 latent topics within the text. By tracking the prevalence of these topics over time and linking them to specific regulatory events—such as the implementation of SFAS 157 (fair value), SOX Section 404 (internal controls), and SEC Item 1A (risk factors)—the researchers quantify how much of the observed increase in disclosure length and complexity is attributable to these specific mandates.

The study finds that the median 10-K length more than doubled during the sample period. Remarkably, three specific topics—fair value/impairment, internal controls, and risk factor disclosures—account for virtually the entire increase in length. These topics also explain the observed trends in other textual attributes, such as increased boilerplate, redundancy, and stickiness, as well as decreased readability and specificity. The authors demonstrate that these trends are robust even when controlling for firm-level economic factors, and they observe similar patterns across diverse industries. Furthermore, firms for which these disclosures were less relevant often provided the most boilerplate and redundant text, suggesting that regulatory compliance, rather than firm-specific information needs, is the primary driver of the observed disclosure evolution.

This research provides a systematic, objective method for "opening the black box" of corporate text. By linking aggregate textual trends to specific regulatory origins, the findings offer critical evidence for regulators and standard setters currently evaluating the effectiveness of disclosure frameworks. The study suggests that while regulation has successfully increased the volume of information, it has also contributed to a significant rise in disclosure "overload" that may hinder efficient price discovery.