A machine learning approach to risk disclosure reporting

Max Resende () and Alexandre Ferreira ()
Additional contact information
Max Resende: Federal University of Santa Catarina
Alexandre Ferreira: Save Advisers LLC

Economics Bulletin, 2021, vol. 41, issue 2, 234-251

Abstract: It is widely recognized that corporate annual reports play a key role in financial markets. Given the debate on risk analysis, this paper applies a machine learning statistical technique called Latent Dirichlet Allocation (LDA) in order to classify companies risks reported on 10-k SEC Form from 2006 to 2017 and applies a predictive logit model to assess the idiosyncratic risks of individual firms and relate it to firm-specific characteristics, such as market equity, total assets, among others. Among several results, it was verified that non-diversifiable risks, such as tax, competition, insurance, intellectual property and government behave similarly throughout all the industries, whereas Financial Statements concerns appear to be temporary. Moreover, market equity, total assets and the firm's age are predictive of all risks and firms for which the risk is captured are smaller on average, present lower market equity and total assets besides been younger and slightly less profitable when compared to traditional firms.

Keywords: Risk factors, Topic modeling; LDA, 10-k SEC Form (search for similar items in EconPapers)
JEL-codes: C6 G3 (search for similar items in EconPapers)
Date: 2021-04-09
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://www.accessecon.com/Pubs/EB/2021/Volume41/EB-21-V41-I2-P22.pdf (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:ebl:ecbull:eb-20-00810

Access Statistics for this article

More articles in Economics Bulletin from AccessEcon
Bibliographic data for series maintained by John P. Conley ().