Textual Classification of SEC Comment Letters
Review of Accounting Studies, Forthcoming
61 Pages Posted: 2 Aug 2014 Last revised: 7 Nov 2019
Date Written: November 5, 2019
Abstract
This study examines the impact of SEC comment letters on future financial reporting outcomes and earnings credibility. Naive Bayesian classification identifies comment letters associated with future restatements and write-downs. An investor attention-based quantitative measure of importance, using EDGAR downloads, is also predictive of these outcomes. Disclosure-event abnormal returns, revenue recognition comments, and the number of letters in a conversation appear to be useful quantitative metrics for classifying importance in certain settings. This study also documents trends in comment letter topics over time, and identifies topics associated with the textual and quantitative classifications of importance, providing insights into the factors drawing investor attention and which relate to future restatements and write-downs. Innocuous comment letters are associated with improvements in earnings credibility following comment letter reviews.
Keywords: SEC comment letters, text classification, topic modeling, financial performance, restatements, write-downs, internal control weaknesses, investor attention, Naive Bayes, Latent Dirichlet Allocation
JEL Classification: M41, G14, G18
Suggested Citation: Suggested Citation