Annual Report Narratives and the Cost of Equity Capital: U.K. Evidence of a U-shaped Relation

Abstract

We hypothesize and test for a U-shaped relation between the cost of equity capital and the level of disclosure in annual report narratives. Using a computer-generated word-count-based index of the level of disclosure in U.K. annual report narratives, we document a negative relation with the cost of equity capital at low levels of disclosure, and a positive relation at higher levels of disclosure, together implying the presence of an optimal level of disclosure. We interpret the positive relation at higher levels of disclosure as evidence of uninformative clutter increasing the cost of equity capital. Additional analyses indicate the presence of both firm-level learning and regulatory corporate reporting initiatives as factors shaping adjustments towards optimum levels of disclosure.

Keywords:

• Clutter
• Disclosure Optimum
• Learning Effects
• Narratives

JEL codes::

• C33
• C36
• G30
• M41

Acknowledgements

Florian Eugster gratefully acknowledges the support of Handelsbanken Wallander stipendium. The authors thank Mahmoud El-Haj, Paul Rayson, and Steven Young, for their work on developing the CFIE annual report narratives database; and Mattias Hamberg (discussant) and the participants of the Accounting Research Seminar at Stockholm School of Economics, the University of Bradford, the 2016 Nordic Accounting Conference, and the 2017 EAA Congress in Valencia, for helpful comments on earlier versions of the paper. We thank the anonymous reviewer and the special issue editors, Reuven Lehavy and Florin Vasvari, for very helpful comments. All remaining errors are the responsibility of the authors.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental Data and Research Materials

Supplemental data for this article can be accessed on the Taylor & Francis website, doi:10.1080/09638180.2019.1707102

Section A) Further details on the disclosure index, Discindex, and its components.

Section B) and Table IA.3 provide details on the estimation of the implied cost of equity capital.

Section C) Correlation table for main regression variables.

Section D) Robustness tests relating to the main regression model.

Section E) Results of using nonparametric regression to infer the functional form.

Table IA.1. Further Details on the Components in Discindex: Fractional Ranks versus Standardised Component Raw Scores.

Table IA.2. Further Details on the Components in Discindex: Year-on-Year Changes in Fractional Ranks versus Year-on-Year Changes in Standardised Component Raw Scores.

Table IA.3. Descriptive Statistics and Results from the Cross-sectional Earnings Regression Models

Table IA.4. Correlations of Main Regression Variables

Figure IA.1. Non-linear Relation between Cost of Equity Capital and Volume of Annual Report Narratives Using Non-parametric Regression

ORCID

Thomas Schleicher http://orcid.org/0000-0002-8553-4532

Notes

1 Within the Business Review firms provide a commentary on corporate objectives, strategy and resources available to deliver those objectives, risk and uncertainties facing the entity, and trends and factors likely to affect the company’s future.

2 RS1 is the only reporting statement the FRC has ever issued.

3 Another argument relates to the joint effect of capital structure when it is possible that more expansive disclosures increase the firm’s cost of capital to the extent that they enhance assessments of the firm’s downside risk (Bertomeu, Beyer, & Dye, 2011).

4 During our sample period, the FRC identified clutter, defined as immaterial disclosures that inhibit users’ ability to identify and understand relevant information, as a growing problem of the U.K. annual report (2009, 2011).

5 While there are numerous studies conducting content analysis of U.K. annual reports, they focus on particular aspects (e.g., intellectual capital, Beattie & Thomson, 2007) or particular aspects (e.g., tone) of particular sections (e.g., chairman’s statement, operating and financial review, Smith & Taffler, 2000; Rutherford, 2005).

6 Within this pool of studies, initial findings by J. E. Smith and Smith (1971) suggest that the readability of financial statement footnotes is poor, while other studies associate variations in annual report readability with the external auditor (Barnett & Leoffler, 1979), the risks facing the company (Courtis, 1986), and financial performance (Subramanian, Insley, & Blackwell, 1993).

7 Annual report disclosure scores are available from the Corporate Financial Information Environment (CFIE) website under http://ucrel.lancs.ac.uk/cfie/annual-report-scores.php.

8 The CFIE dataset is the outcome of a publicly funded collaboration between Lancaster University Management School, Alliance Manchester Business School, and the London School of Economics, and, at present, provides disclosure scores for 10,443 annual reports published during calendar years 2003–2014 by 1989 LSE-listed, non-financial companies. These 10,443 reports are also employed in El-Haj et al. (2020) for an assessment of the disclosure scores’ accuracy which is typically in excess of 95%.

9 The volume of forward-looking and causal reasoning commentary is measured using keyword lists defined by the CFIE project team based on a review of the prior literature, especially Hussainey, Schleicher, and Walker (2003) for forward-looking keywords, and Zhang and Aerts (2015) for causal reasoning keywords. Both keyword lists are available from http://ucrel.lancs.ac.uk/cfie/annual-report-scores.php. The two readability indices are Gunning’s (1968) Fog index and the Flesch-Kincaid (Kincaid, Fishburne, Rogers, & Chissom, 1975) readability index. See also El-Haj, Rayson, Walker, Simaki, and Young (2019).

10 Hou et al. (2012) argue that a cross-sectional model helps to overcome optimistically biased analyst forecasts and potential IBES coverage problems, especially in relation to smaller and financially distressed firms.

11 When we require all four individual models to estimate the average cost of equity capital, then untabulated results remain unchanged.

12 However, a potential problem with firm fixed effects is that it runs the risk of introducing a large number of irrelevant variables into the model thereby inducing a type II error.

13 Requiring financial analyst coverage would reduce our sample even more and we would end up with a sample of 2808 firm-years.

14 A correlation table for the main regression variables is also presented and discussed in the IA.

15 To assess the sensitivity of our findings in Table 3 to the way the components of the index are aggregated, we run the fixed effects regression with all the individual components and all the squared values of the components included as separate variables. The adjusted R-squared of the model increases only marginally, from 0.43 to 0.44. Consistent with the U-shaped hypothesis, the t-value of the average value of the individual (linear) coefficients is −5.41, and the t-value of the average value of the squared term coefficients is 6.33.

16 The IA presents robustness results where our implied Coec estimates are replaced with realized future returns, acknowledging that realized returns are a noisy proxy for Coec (e.g., Campbell, 1991; Elton, 1999; Vuolteenaho, 2002). In addition, the IA addresses the concerns of Wang (2015, 2017) about implied Coec estimates.

17 Changes in sample composition are also unlikely to underlie our results, on account of the large size of the cross-section in our sample (1183 firms in our final sample with a 4.4 average number of years). As a robustness test, we also repeat the analysis for a ‘quasi-balanced’ sample of firms that have at least 10 observations over our sample period. The average observation per firm for this test rises to 10.8. We document similar results for this sample. We find optima that are similar to those reported in Table 3 (e.g., in column (2) we have 0.68 compared to 0.64 and in column (3) we have 0.65 compared to 0.62).

18 Because of the limited number of observations per firm in each window, this analysis focuses on results for industry fixed effects. In untabulated results we recalculate the optima based on a one-year rolling window and, similar to the results in Table 4, the optimal level of disclosure is increasing over time. Furthermore, we observe a significant non-linear effect in all years but 2013. Moreover, when we define separate three-year rolling windows for each ICBN one-digit industry sector, then over 80% of the regressions have a significantly negative slope on Discindex and a significantly positive coefficient on Discindex_squared, and the average optimum is 59% with a standard deviation of 9 percentage points.

19 Average Discindex rises for both firms lying below the optimum from 25% in 2005 to 42% in 2012 and for firms over the optimum from 72% in 2005 to 89% in 2012.

20 This approach is informed by the two-stage approach devised by Core and Guay (1999) but it is not identical. Core and Guay (1999) use the residual of their first-stage model (of stock-based compensation on a set of potential determinants such as firm size, idiosyncratic risk, book to market, a proxy for a potential free cash flow problem, and industry controls) as the key variable in their second-stage model. In our model, we use measures of the extent to which the partially endogenous variable, the disclosure level, differs from the optimum level in the first stage as the main variable in our second stage estimations.

21 We use a three-year window to ensure a reasonable number of observations for the first stage regression and taking into account the possibility that the optimum level can change gradually over time.

22 In an untabulated test, we also added two-period lagged change in Discindex and found that the coefficients on Lag_distance_overdisclosure and Lag_distance_underdisclosure continued to show their predicted sign and remained significant (despite a further reduction in sample size to 1405 observations). Specifically the two coefficients were −0.316 and 0.061, with t-statistics of −8.95 and 2.55. There was no evidence in the data of serial autocorrelation of third order. After controlling for first and second order autocorrelation the coefficient on the three-period lagged change in Discindex is −0.005 with a t-statistic of −0.14.

23 This finding is not sensitive to research design choices especially the definition of the firm-year optimum. For example, we replicated Table 5 column (1) with an optimum based on a one-year rolling window and find similar results. In untabulated tests we also employed industry-specific three-year rolling windows based on the ICBN one-digit industry classification. Again, we obtain similar results to those reported in Table 5 column (1).

24 This result is robust to modeling the incidence of over-disclosure, rather than the amount of overdisclosure, as an alternative operational proxy for clutter.

25 Bushee, Gow, and Taylor (2018) make a similar attempt in that they try to decompose the linguistic complexity of U.S. firm conference calls into an informative component and an obfuscation component. In doing so they assume that the readability of analyst contributed sentences reflect informativeness, but not obfuscation. Using a measure of stock illiquidity as a proxy for information asymmetry, they find that information asymmetry is positively (negatively) associated with the obfuscation (informativeness) component of the conference calls.

26 We use the number of words in the financial statements section of the report rather than the number of words in the narratives section because the latter includes aspects of performance, governance or strategy-related commentary, which we wish to maintain.

27 Gunning (1968) defines Fog as the estimated number of years of formal education a person needs to understand the text on first reading. It is calculated by adding the percentage of words in a text with three or more syllables to the average number of words per sentence, both multiplied by 0.4. For example, a Fog Index of twelve requires the reading level of a high school graduate, that is, someone around 18 years old. Because higher values of Fog indicate text that is more difficult to understand we multiply Fog by minus one before we rank it. This implies that a higher ranking indicates a report that is easier to read, that is, requires fewer years of formal education.

Additional information

Funding

This work was supported by the Economic and Social Research Council (ESRC) under grant ES/J012394/1 (Athanasakou, Schleicher, and Walker) and under grant ES/R003904/1 (Walker), and the Institute of Chartered Accountants in England and Wales (ICAEW) under grant 5-443.