Textual Changes in 10-Ks and Stock Price Crash Risk: Evidence from Neural Network Embeddings

Yahya Yilmaz,Doron Reichmann,Milan Reichmann
DOI: https://doi.org/10.2139/ssrn.4613815
2023-01-01
SSRN Electronic Journal
Abstract:Previous research attributes stock price crash risk to managerial bad news hoarding. Contrary to this notion, we find evidence that stock price crash risk is determined by investor inattention to textual changes in corporate disclosures. Using a large sample of 10-K filings, we estimate neural network embeddings to quantify the degree of textual changes in successive 10-Ks. We find that changes in 10-Ks have a positive and economically meaningful impact on one-year-ahead stock price crash risks. Our results suggest that investor inattention to textual changes in 10-Ks can have broader capital market consequences than previously documented.
What problem does this paper attempt to address?