Textual analysis and international financial reporting: Large sample evidence

Abstract

We examine annual report text for over 15,000 non-US companies from 42 countries over the period 1998-2011, focusing on the length of disclosure, presence of boilerplate, comparability with US and non-US firms, and complexity. We find that textual attributes are predictably associated with regulation and incentives for more transparent disclosure and are correlated with economic outcomes such as liquidity, institutional ownership, and analyst following. Using mandatory IFRS adoption as an exogenous shock, annual report disclosure improved in the sense that quantity of disclosure increased, boilerplate was reduced, and comparability increased relative to both US and non-US firms. Firms with the greatest improvements in financial reporting experienced the greatest improvements in economic outcomes around IFRS adoption.

Readability of 10‐K Reports and Stock Price Crash Risk

“…Other studies that use the Fog Index to measure readability include those of Allee and DeAngelis (), Biddle et al (), Bonsall and Miller (), Bozanic and Thevenot (), Bushee et al (), Callen et al (), Guay et al (), Laksmana et al (), Lang and Stice‐Lawrence (), Lawrence (), Lee (), Lehavy et al (), Lo et al (), Lundholm et al (), Merkley (), Miller (), Nelson and Pritchard (), and Rogers et al () for corporate disclosures; De Franco et al () and Hsieh and Hui () for financial analysts’ reports; and Dougal et al () for news articles.…”

The Economics of Disclosure and Financial Reporting Regulation: Evidence and Suggestions for Future Research

“…Many important disclosures such as the Management Discussion & Analysis, 10‐K footnote disclosures, and conference call transcripts are qualitative, text‐based, and narrative in nature, which previously made it difficult to use them. However, recent advances in text analysis, computational linguistics, and natural language processing allow us to construct new measures for narrative disclosures, some of which have quality dimensions (e.g., Li [, ], Loughran and McDonald [], Lang and Stice‐Lawrence []). Text‐based proxies can be applied broadly (e.g., to the entire 10‐K) or more narrowly (e.g., to an earnings announcement or a particular part of the 10‐K), so the earlier discussion of the tradeoffs between narrower and broader measures applies here as well.…”

A Review of the IFRS-Adoption Literature

“…In all of the periods, firm fixed effects appear to be the least frequently used relative to either country or industry fixed effects. Only four of the sixty-four studies, including those by Daske et al (2008); Horton, Serafeim, and Serafeim (2013) ;Chen, Young, and Zhuang (2013);and Lang and Stice-Lawrence (2015), include firm fixed effects in their main regression models. This is perhaps due to the large loss in degree of freedom that results when firm fixed effects are included, which lowers the power of tests.…”