Informational Content and Assurance of Textual Disclosures: Evidence on Integrated Reporting

Abstract

This paper examines the economic benefits associated with textual attributes and the external assurance of integrated reporting (IR), an innovative form of corporate disclosure that connects financial and environmental, social and governance (ESG) information in a single report. We investigate the setting of South Africa, where IR has been mandatory since 2010 for listed companies. We find that IR readability is associated with a higher market valuation, conciseness with higher stock liquidity and tone bias with less dispersed analysts’ estimates. Results suggest that market participants appreciate IRs that are readable, short and focused, as well as hint at tone management strategies targeting analysts. We also show that assurance on IR moderates the negative effects of poor textual attributes: if firms publish IRs that are difficult to read but assure them, this compensates for the negative influence of reading difficulty on a market value; if long IRs are assured, this dampens the negative effect of verbosity on liquidity; if firms assure IRs, analysts’ forecast dispersion is lower, therefore suggesting that assurance acts as a credibility-enhancing mechanism for external users. Finally, we show that textual attributes and assurance matter for broader audiences interested in the ESG dimensions of a firm’s performance.

Keywords:

• Textual analysis
• Assurance
• Integrated reporting
• Capital market effects

JEL codes:

• M40
• M41
• M42
• G3
• M14

Acknowledgements

We thank the Editors of this Special Issue and two anonymous reviewers for their constructive remarks and guidance. We acknowledge the helpful suggestions from Andrea Bafundi, Claudia Imperatore and seminar participants at Aalto University, Norwich Business School, Oxford Saïd Business School, University of Halle-Wittenberg. We also appreciate comments from participants at the Cass and CeFARR conference, 15 September 2017, Cass Business School; 41 European Accounting Association Conference, Bocconi University, 30 May–1 June 2018; IX Financial Reporting Workshop, University of Bologna, 14–15 June 2018; XIV International Accounting Symposium, Fundación Ramón Areces, 25–29 June 2018; 30 International CSEAR congress, St Andrew University, 28–30 August 2018; Sidrea Annual Conference, Verona University, 13–14 September 2018.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental Data and Research Materials

Supplemental data and research materials for this article can be accessed on the Taylor & Francis website, doi: 10.1080/09638180.2019.1677486.

Online Appendix includes:

• Table OA.1. Overview of published IR studies on economic consequences of JSE-listed firms in South Africa.

• Table OA.2. Protocol for content analysis of IR assurance reports.

• Table OA.3. Textual characteristics, assurance and additional economic effects.

ORCID

Ariela Caglio http://orcid.org/0000-0002-5829-9784

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

1 Some examples of recent papers in the emergent research stream on IR comprise Perego, Kennedy, and Whiteman (2016), de Villiers, Venter, and Hsiao (2017), Melloni et al. (2017), Reimsbach et al. (2017), Velte and Stawinoga (2017), and Kannenberg and Schreck (2019).

2 In this paper, we use the acronym IR for both ‘Integrated Report’ and ‘Integrated Reporting’.

3 For example, under the EU Directive 2014/95 enforced from January 1, 2017, all public interest entities with more than 500 employees have to report information concerning environmental and social performance, human rights, bribery matters and diversity policy (Directive 2014/95/EU, 2014).

4 The disclosure of an IR is de facto mandatory in South Africa; however, ‘the board of directors, in its collective decision-making, could decide to apply the recommendation differently or apply another practice and still achieve the objective of the overarching corporate governance principles of fairness, accountability, responsibility and transparency. Explaining how the principles and recommendations were applied or the reasons for non-application results in compliance’ (IoD, 2009).

5 By comparing how participants react to disclosures either with or without an auditor’s attestation, several experimental studies examine the effects of auditing on credibility. Both bankers (e.g., Libby, 1979) and financial analysts (e.g., Hodge, 2001) rate audited disclosures as more credible than unaudited disclosures. Several archival studies also corroborate this rationale (DeFond & Zhang, 2014). For instance, Behn, Choi, and Kang (2008) show that higher audit quality is associated with less dispersed and more accurate earnings forecasts by sell-side analysts.

6 Archival studies about the effects of sustainability assurance on investor decisions and capital markets point at similar effects, although the evidence is quite sparse. Dhaliwal, Li, Tsang, and Yang (2011) find that the reduction in the cost of equity capital associated with first-time CSR report adopters is approximately two times higher when the CSR report is externally assured. Similarly, contrary to Dhaliwal, Radhakrishnan, Tsang, and Yang (2012), Casey and Grenier (2015) document that firms with an assured CSR report have a lower cost of equity capital, analyst forecast error and dispersion than do firms that issue unassured CSR reports.

7 Muslu et al.’s (2019) proxy for CSR disclosure quality aggregates the decile rankings of six textual attributes, i.e., optimism/pessimism, readability, length, numerical and horizon content.

8 We argue that South Africa provides a suitable institutional setting to explore the credibility-enhancing role of auditing beyond the alleged value-enhancing mechanisms prompted by the introduction of a new mandatory disclosure (H1 of this study). The IR assurance market in South Africa ensures a high level of heterogeneity for two main reasons. First, the interpretative construction of an IR by preparers because of limited prescriptive disclosure requirements in the IIRC framework. Second, the difficulty of defining specific auditing procedures required to assess an IR’s assurance, for instance, whether or not an IR deals adequately with the different types of capital transformations and the interconnections between these (Simnett & Huggins, 2015).

9 Since 2011, considering the top 100 companies listed on the JSE (EY, 2015), EY has annually organized the ‘EY Excellence in Integrated Reporting’ awards. The awards are assigned to those firms whose IRs are best aligned with the IIRC framework principles. A panel of three independent coders involved in the EY awarding process since its first edition carry out a separate content analysis of each IR by using a pre-agreed mark plan. A mark out of ten is awarded for each of the seven Guiding Principles (i.e., strategic focus and future orientation, connectivity of information, stakeholder relationships, materiality, conciseness, reliability and completeness and, last, consistency and comparability). A mark out of ten is awarded for each of the eight Content Elements (i.e., organizational overview and external environment, governance, business model, risks and opportunities, strategy and resource allocation, performance, outlook and, finally, basis of presentation and preparation). Marks are also awarded for the extent to which the IR incorporates the IIRC framework’s principles dealing with how value is created with reference where relevant to the six ‘capitals’. The coding process results in a public-announcement in which each of the 100 companies is classified as ‘Excellent’, ‘Good’ ‘Average’ or ‘Progress to be made’.

10 Barth et al. (2017) state that ‘the final ranking is based on a combination of the average of these scores, overall perceptions and extensive discussions surrounding the final rankings for each company’ (EY, 2015). Despite their reassurance that EY rankings are not prone to the coders’ bias, they report that EY prefers not to disclose the detailed score sheet ‘as a consequence of the intention to change the mark plan on annual basis and the subjectivity involved in its use’ (Barth et al., 2017, p. 49). EY further emphasizes that the ‘ranking process is important, as the scoring process is subjective and scores may differ, based on the adjudicators’ impressions at the time’ (EY, 2015, p. 25). Such key remarks highlight the potential threat due to human coding, that subjective interpretation and biased measurement might be reflected in EY scores.

11 Through the issuance of a syntactically complex corporate report, the omission of information content and the obfuscation of bad news (concealment), managers might tend to manipulate an IR to preserve their own and their firm’s reputation. Melloni et al. (2017) found that to manage public impressions, IR early adopters employ quantity and syntactical reading ease manipulation, as well as thematic content and verbal tone manipulation, as deliberate reporting choices.

12 Recent ESG controversies are reflected in the latest complete period. The default value of all controversy measures is 0. For example, if the benchmark consists of 6 companies, 4 with a value of 0 and two with a value of 1 (polarity here is negative, so the higher the number is, the worse the polarity), then the formula for companies with no controversies is (2 + 4/2)/6 = 67%, and for companies with one controversy, (0 + 2/2)/6 = 17%. All recent controversies are counted in the latest closed fiscal year, and no controversy is double counted. Controversies are benchmarked by industry group. For instance, the last completed fiscal year for a company is Dec 31, 2015. If there is one controversy on May 1, 2016 and one controversy on May 1, 2017, both are accounted under recent controversies and included in the scoring for FY2015. Once FY 2016 is completed, the two recent controversies are moved to FY2016, but the one on May 1, 2016 is moved to the normal controversy, while the one from 2017 remains under a recent but accounted status in FY2016. When FY2017 is completed, it is removed from a recent status in 2016 and moved to normal DP in 2017. Except for management departures, all other controversies are quantitative (Thomson Reuters, 2019).

13 Before computing all the indexes, removing tables, graphs, titles, and headings, we convert the pdf into text; we also remove numbers, symbols, URLs, special characters and stop words (Lewis, Yang, Rose, & Li, 2004; Li, 2008; Loughran & McDonald, 2016).

14 This measure is calculated as follows: Fog = (words_per_sentence + percent_of_complex_words) * 0.4. The index indicates the number of years of formal education a reader of average intelligence would need to read and understand the text with such a word-sentence workload. In particular, the relation between the Fog and reading ease is as follows: Fog  ≥ 18 (unreadable); 14–18 (difficult); 12–14 (ideal); 10–12 (acceptable); and 8–10 (childish).

15 This index is measured as follows: Flesch-Kincaid Grade = (11.8*syllables per word) + (0.39 *words per sentence) -15.59. The third and final measure is the Flesch Grade Level, which has also been used in previous accounting studies (e.g., Subramanian, Insley, & Blackwell, 1993; Laksmana, Tietz, & Yang, 2012; De Franco et al., 2015).

16 McLaughlin (1969) developed its formula as a more precise and more simply calculated substitute for the Gunning Fog index (an approximate formula requires the counting of the words of three or more syllables in three 10-sentence samples, the estimation of the count’s square root (from the nearest perfect square), and the final addition of 3).

17 Indeed, a distinctive feature of DICTION is that its use of normative values for comparative purposes allows the application of dictionaries specifically tailored for particular types of disclosures (including corporate financial reports). In our analysis, we consider the optimism score with a ‘corporate financial report’ as a normative profile because it carries the advantage of being designed specifically for use in a financial context. We use ‘standardize scores’, meaning that DICTION extrapolates each particular text to a 500-word norm equivalent (which is the basic unit of analysis) so that input texts of any length can be measured consistently.

18 The optimism score is computed using this formula: DICTION Optimism = [praise + satisfaction + inspiration]−[blame + hardship + denial]. The DICTION Certainty score is computed by using this formula: [tenacity + leveling + collectives + insistence]−[numerical terms + ambivalence + self-reference + variety].

19 The ranking criteria range from items that convey factual data (such as title, addressee, name and location of assuror) to items that capture the characteristics of the assurance engagement and require an interpretation of the principles stated in the standards (such as materiality, completeness, and responsiveness to stakeholders).

20 Through some t-tests, we find that IR assurance is significantly more likely for the following: firms belonging to environmentally sensitive industries, such as oil and gas, basic materials, utilities; larger firms (where size is measured as total assets); and less profitable ones (where profitability is measured as ROA). On the contrary, there are no significant differences between financial and non-financial firms. In addition, we find that 71% of the assurance providers are Big 4 and 74% are accounting firms. Firms belonging to environmentally sensitive industries and larger firms are more likely to choose a Big 4 firm as assurance provider.

21 We also calculate correlation coefficients. Our (untabulated) analysis shows that there are statistically significant relationships among the IRs’ textual characteristics, external assurance adoption and our dependent variables. The reading difficulty (measured by the Fog, Flesch and Smog indexes) is negatively correlated with Tobin’s Q. Length (measured by the number of words and characters) is negatively associated with Tobin’s Q and the bid-ask spread and positively with the analysts’ forecasts dispersion. Length is additionally positively associated with assurance. Optimism is negatively associated with forecasts dispersion, whereas certainty is negatively correlated with assurance adoption.

22 It is important to mention that to assess the robustness of our multivariate analyses, we run some additional tests (namely, the Fog index, the Flesch index, the Smog index for Readability; total number of words and total number of characters for Length; optimism and certainty scores, for Tone), where we use the single items instead of the three factors. The (untabulated) results of these supplementary analyses are largely consistent with the results of our main multivariate analyses.

23 To understand if the modelled relationships vary for financial companies, we re-estimate our main models by using financial industry specific variables as controls. Drawing on Laeven and Levine (2007), we include in our models the following: net interest income to total operating income, deposits to liabilities ratio, equity to assets ratio and income diversity. We re-run our analyses with these new control variables on the sub-sample of financial firms included in our main sample. The results are largely consistent with our main analyses, while we also gain some additional insights from these tests. IR length seems to be appreciated by investors, as we find a positive and significant association between length and Tobin’s Q. Moreover, reading difficulty increases the analysts’ forecast dispersion, while assurance moderates the negative impact of reading difficulty.