Qualitative Disclosure and Changes in Sell‐Side Financial Analysts' Information Environment

Abstract

We examine a routine and timely disclosure, earnings press releases, to determine the extent to which several novel qualitative elements of such disclosures are associated with changes in sell‐side financial analysts' information environment. Using a comprehensive set of GARCH‐based (generalized autoregressive conditional heteroscedasticity) proxies, we examine how disclosure readability's components, across‐document textual similarity, and within‐document lexical diversity alter analysts' information environment. We find that readability in the form of shorter sentences, textual similarity, and lexical diversity are strongly related to decreases in analysts' uncertainty. Further, shorter sentences and lexical diversity improve both public and private information precision, whereas similarity affects solely analysts' private information precision. While the GARCH‐based proxies allow us to alleviate concerns regarding potentially spurious inferences (Sheng and Thevenot 2012), we note as a caveat that such an estimation restricts our inferences to large, stable, and heavily followed firms. These findings should be of interest to analysts who may wish to explore the latent information contained within the qualitative elements of disclosure, regulators who direct the form and content of disclosure, and academics who study the use (and possible misuse) of various forms of information and its presentation.

Linguistic Complexity in Firm Disclosures: Obfuscation or Information?

“…Since Li [2008], a growing literature uses Fog to examine the relation between linguistic complexity of mandatory disclosures (e.g., 10-Qs or 10-Ks) and a variety of firm outcomes, such as investment efficiency, bid-ask spreads, delayed pricing of accounting information, voluntary disclosure, and short-sale constraints (see Loughran and McDonald [2016] for a review). Prior work also examines the relation between Fog and the decisions of external parties, such as trading of unsophisticated investors (Miller [2010], Lawrence [2013]), analysts' outputs (Lehavy, Li, and Merkley [2012], Bozanic and Thevenot [2015]), and disagreement among credit rating agencies (Bonsall and Miller [2017]). 2 Despite the widespread use of the Fog index, there is an underlying concern about its construct validity because Fog potentially commingles informative text with uninformative text (see, e.g., Lang and Stice-Lawrence [2015]).…”

IRS Attention

“…The cosine textual similarity of firm i's tax footnote in year t as compared to firm i's tax footnote in year t − 1 (Brown and Tucker [2010], Bozanic and Thevenot [2015]).…”

Quantifying differential interpretation of public information using financial analysts’ earnings forecasts

“…We expect that investors will differ more in their interpretations of the earnings release when the disclosure is less readable. This prediction follows from the work of Li (2008), who shows that annual reports that are more difficult to read are associated with lower earnings, and Bozanic and Thevenot (2014), who find that the readability of earnings press releases is associated with 12 In addition, we examine the question of whether higher levels of differential interpretation are related to lower returns around the earnings announcement, as was documented by Berkman et al (2009). We find that all opinion divergence proxies are related negatively to the three-day abnormal return centered on the earnings announcement, although the results are statistically insignificant for MATO and DTO, statistically significant at the 5% level for LKP and ChangeLD, and statistically significant at the 1% level for LDI, SUV and BAspread.…”