Inferring accounting conservatism from textual analysis of 10-K forms

Purpose Following an extensive body of research that analyzes accounting conservatism using financial statement data, we examine the text of corporate disclosures to determine whether textual analysis can provide insight into the degree and motivation behind conservative accounting practices. Design/methodology/approach This study applies natural language processing (NLP) techniques to assess the degree of accounting conservatism in a large sample of 10-K filings from 1999 through 2023. Findings Our NLP measures of accounting conservatism, which employed the “log odds ratio informative Dirichlet prior,” generally align with conventional measures regarding conservatism, text uncertainty and tone. Still, the readability of 10-K reports and their implications for analyst forecast accuracy differ. We conclude that supplementing traditional accounting conservatism measures with NLP techniques can enhance our understanding of conservatism in Securities and Exchange Commission (SEC) filings. Originality/value We conclude that traditional accounting conservatism measures, supplementing them with NLP techniques, can enhance our understanding of conservatism in SEC filings.