Effects of integrated reporting on the cost of capital and analysts’ forecasts errors: A study of Johannesburg Stock Exchange listed mining firms

Orientation: Integrated reporting (IR) has gained traction over the last decade. Although IR became mandatory for all firms listed on Johannesburg Stock Exchange (JSE) in 2010, the International Integrated Reporting Council (IIRC) issued the IR framework in December 2013.Research Purpose: The study seeks to investigate the effects of IR on the cost of equity capital and analysts’ forecast errors for the mining firms listed on JSE.Motivation for the study: A large part of empirical evidence highlights benefits of IR; however, some studies still find no link between the quality of integrated reports and economic benefits for the reporting firm. It is against this backdrop that the study investigates effects of integrated reports on the cost of equity capital and analysts’ forecast errors.Research approach, design and method: We use a quantitative research design to test effects of IR on the cost of equity capital and analysts’ forecast errors. We study used a panel regression to analyse relationship amongst IR, cost of equity capital and analysts’ forecast errors.Main findings: The study found a significant negative relationship between IR scores and cost of equity and analysts’ forecast errors.Practical or managerial application: The findings of the study could incentivise managers in other jurisdictions where IR is not mandatory. Furthermore, findings may contribute to the existing discourse on firm-based benefits related to the quality of IR.Contribution or value addition: The study contributes to the body of knowledge with regard to possible benefits associated with compliance with the IR reporting framework.

its dominance in South Africa's political, social and economic landscape (MineralsSA 2018). In 2017, this sector employed 464 667 people and contributed R312 billion to GDP (MineralsSA 2018). The listed mining firms are, therefore, a reasonable representation of the South African economy. It is against this backdrop that this study investigates the effect of IR on the cost of equity capital (COE) and analysts' forecast errors (FCERROR) for the JSE listed mining firms.

Review of literature
This study is premised on the agency theory. The theory seeks to identify and solve the agency problem between owners and managers of the firm (Jensen & Meckling 1976;Panda & Leepsa 2017;Ross 1973). For example, the managers may have good strategic plans about investments that would yield good returns in the end and in the process decide to adjust the dividend policy. However, if the shareholders do not have such information, the strategic plan could result in shareholders disposing their shares if they were expecting more dividends. Some of the ways to minimise agency problem (Adhariani & De Villiers 2019;García-Sánchez & Noguera-Gámez 2017;Vitolla, Raimo & Rubino 2020a) include a firm's desire to meet the needs of providers of financial capital. To this end, the reporting firm provides strategic information on the business model and its ability to seize opportunities and mitigate risks and explain the creation of value over time.
The inclusiveness nature of IR and its inclination towards providers of capital make agency theory relevant to the study (IIRC 2013, Flower 2015, Vitolla et al. 2020b. Theories provide the basis upon which reporting and disclosures are developed (Urquiza et al. 2010). The agency theory acknowledges that there is an agency problem between principals (shareholders) and agents (executives) (Shapiro 2005) and seeks to mitigate this problem and its associated costs. By addressing agency problems, firms are likely to experience a reduction in estimation risk and information asymmetry (Diamond & Verrecchia 1991;Verrecchia 2001;Vitolla et al. 2019;Zhou et al. 2017). As posited by Serafeim (2015), firms that adopt strategies to reduce information asymmetries improve their opportunities of attracting funds from the capital market. Such improvement is attributed to the reduction of agency costs. The integrated thinking culture in itself forces the firm to think big and strategically in their daily operations. The conciseness, connectivity, future orientation, risk mitigation and opportunistic approach of IR are possible contributing factors that could enable a firm attract funding from the capital market.
Empirical evidence from previous literature suggests that there is a negative relationship between the high-quality disclosures and FCERROR (Barth, Kasznik & McNichols 2001;Bradshaw, Miller & Serafeim 2009;Dhaliwal et al. 2012;Flores et al. 2019;Lehavy, Li & Merkley 2011;Zhou et al. 2017). Similarly, studies by Zhou et al. (2017), Vitolla et al. (2019) and Flores et al. (2019) find a link between high quality disclosures and reduction in FCERROR. The evidence from previous studies by Dhaliwal et al. (2012), Zhou et al. (2017) and Bernardi and Stark (2018) suggests that financial analysts do not only use financial information when predicting future cash flows, but they also rely on the non-financial information presented in the form of voluntary disclosures. The findings highlight the relevance of non-financial information to the analysts. Provision of more complete data would thus reduce FCERROR (Dhaliwal et al. 2011). Moreover, as highlighted by Serafeim (2015), analysts are likely to adopt a favourable view of organisations that provide more information.
Some studies attribute reduction of COE to the quality of non-financial disclosures (Chang et al. 2017;El Ghoul et al. 2011;Hail 2002;Suto & Takehara 2018;Vitolla et al. 2019;Zhou et al. 2017). These studies suggest benefits such as improved forecasts of cash flows, improvement of investor loyalty and mitigation of agency costs. Previous studies suggest a theoretical link between the extent to which disclosures affect non-diversifiable risk and COE (Botosan 1997;Johnson 2020;Mazzotta & Veltri 2014;Petrova et al. 2012;Poshakwale & Courtis 2005). Similarly, Lambert, Leuz and Verrecchia (2007) suggest that the assessed level of risk associated with future cash flows can be influenced by accounting information. As an improved and concise disclosure mechanism, IR is expected to help firms reduce information asymmetry and estimation risk. Information that is more relevant will reduce the risk premium. Serafeim (2015) argues that data asymmetry between an organisation's executives and external financial specialists creates uncertainty. The bigger the potential information gap between management and investors, the more probable it is that investors will require a higher rate of return to compensate for the bigger risk. The impact of information asymmetry is greater for long-term investors as their returns depend on future cash flows and sustainability accounting. Because IR takes cognisance of the long-term outlook, it makes provision for information about long-term growth prospects. This is crucial for investors who wish to plant seeds (financial capital) with the aim of obtaining a bountiful harvest (returns). According to Serafeim (2015), firms that provide more information on social, environmental and governance issues tend to outperform (in financial capital terms) those that do not produce or produce fewer Environmental, Social and Governance (ESG) reports. This suggests that firms with superior ESG activities have easier access to investors' financial capital. Furthermore, it is posited by Serafeim (2015) that organisations that reveal more ESG data enjoy a lower cost of capital and improved access to funds.
Previous research also shows that disclosures can limit the non-diversifiable risk, which would result in a decrease in the cost of capital. When a firm produces transparent disclosures, investors tend to show greater support and willingness to invest, which will improve its liquidity. In the https://www.jefjournal.org.za Open Access same vein, Zhou et al. (2017) investigated the impact of integrated reports' level of compliance with the IR framework on the implied cost of capital and FCERROR dispersion for all firms listed on JSE. The researchers developed a scorecard to assess the level of alignment with the framework. The study found a significant negative relationship between integrated reports' level of alignment with the framework and the implied cost of capital and FCERROR.
Despite the theoretical inclination towards negative relationship between the quality of disclosures and COE, it is not uncommon to encounter literature that suggests the opposite. Richardson and Welker (2001) found that the greater extent of disclosures leads to an increase in COE. Dirman (2019) and Eriandani, Narsa and Irwanto (2019) found a positive relationship between the non-financial disclosures and COE, which suggests the undesirable effect of voluntary disclosures. Moreover, Mulyati (2017) found information included in disclosures to be irrelevant to the decisions of investors, suggesting no effect of the cost of capital. Furthermore, Ellis, Fee and Thomas (2012) suggest that managers are faced with a trade-off between benefits and risk disclosures of exposing the firm to the competitors who could capitalise on disclosed information. This potential exposure to competitors is referred to as proprietary costs (Ellis et al. 2012;Guidry & Patten 2012;Peters & Romi 2014). More disclosures could also work against the reporting firm by revealing risks to the market participants. For example, Baboukardos and Rimmel (2016) investigated the impact of adopting IR and found that firms' net asset value declined after the adoption of IR. This was attributed to more reliable risks disclosures about financial sustainability. One can thus deduce that it is possible for the COE to increase because of adopting IR, depending on the nature of the new information and how the markets react to it.

Methodology Data
The study aims to assess the impact of IR on the cost of equity and financial FCERROR. Therefore, a quantitative design is appropriate. The annual integrated reports were obtained from firms' websites. Board size and board independence were sourced from McGregor BFA Library database. We collect socially responsible investment (SRI) data from the FTSE/JSE responsible investment index, which is available on JSE website. The rest of metrics (firm size, market-to-book ratio, leverage, beta and COE, FCERROR and volatility) was sourced from the Bloomberg and McGregor BFA Library databases. We then arranged dependent and independent variables in a panel data format, from 2013 to 2018. The year 2013 is a good starting point because the international IR framework was issued in December 2013 (IIRC 2013). According to the category list from JSE website, there were 58 mining firms listed on JSE (2015). Nevertheless, for the purpose of the study, some firms were omitted because they either did not publish integrated reports or did not adopt the IR framework. Therefore, total treatment securities group covered 37 JSE listed mining firms over a period of 6 years (2013-2018) listed on JSE. However, in some instances, the data required for the panel regression model were not available from Bloomberg and McGregor BFA Library databases. Thus, in the final analysis, there were 166 observations to test the effect of IR on the cost of equity. This was followed by 218 observations for assessing the cost of equity before and after IR framework. Lastly, 58 observations were recorded for the analysts forecast errors.

Model specification
Similar to related studies by Dhaliwal et al. (2011), Zhou et al. (2017, we adapt panel data regression to assess the relationship between the variables (independent and dependent). In order to customise the model to the South African mining industry, we include sustainability reporting and corporate governance. All firms listed on JSE are also expected to publish sustainability and corporate governance reports. We estimate multivariate equations to achieve the study's objectives. To this end, the relationship between the cost of equity and IR is estimated as follows: The regression model as adapted from Dhaliwal et al. (2011).
In order to probe the possible effect of IR on the cost of equity, firms with high integrated scores (30 and above) were selected. For the purpose of the study, firms with high scores were considered to be in full compliance with the IIRC framework (IIRC 2013). Although IR became mandatory for all JSE listed firms, the IR framework was only pronounced in December 2013. It is on this basis, firms would be expected to produce quality and comparable integrated reports after the pronouncement of IR framework. Hence, we regress balance panel data for years 2007-2012 and 2013-2018. The effect of once off events, such as global financial crisis, is mitigated by the use of longitudinal data. We create a dummy variable for periods before and after (BAIR), and it is coded as zero (0) and one (1), respectively. Accordingly, the following regression is estimated:

Cost of equity capital
There are different views when it comes to selecting best proxy for the cost of capital. This can be attributed to possible errors because of measurement assumptions (Barth et al. 2017;Vitolla et al. 2019;Zhou et al. 2017 (2017), Cox and Britten (2019) argued against the appropriateness of using CAPM and Fama-French Three Factor models, citing shortcomings such as incomplete market proxies, volatility of market and restriction of market liquidity. Moyo and Mache (2018), however, investigated variances between financial models in relation to CAPM. Moyo and Mache (2018) found that Gordon Dividend and Ohlson and Juettner-Nauroth (2005) models yielded results, which were close to that of CAPM. In order to mitigate biasness and the effect of measurement errors, our final proxy for the cost of equity is the average of CAPM and Gordon Dividend Valuation models.

Analysts' forecast errors
As stated in the study by Zhou et al. (2017), FCERROR refer to the difference between the average estimates made by analysts over a period of 12 months and actual earnings. Bissessur and Veenman (2016) attribute analysts forecast errors to uncertainty about firms' prospects. Bakker et al. (2020) found a significant negative relationship between mandatory IR and analysts forecast errors. Such findings suggest that financial analysts do not only rely on conventional financial statements but also consider nonfinancial disclosures. Therefore, if financial analysts consider IR, the actual earnings should not be too far from estimated earnings. We collected data for FCERROR from Bloomberg and McGregor BFA Library databases. Adapted from Bissessur and Veenman (2016), Zhou et al. (2017) and Bakker et al. (2020), the formula for analysts forecast errors is presented as follows: where: • FCERROR = Analysts' forecast errors • EPS = Earnings per share Integrated reports Pistoni, Songini and Bavagnoli (2018) suggest that the quality of integrated reports is assessed by including contents elements, assurance and reliability. Similar to Dhaliwal et al. (2011) study on voluntary disclosures, a negative coefficient was expected between the IR score and the dependent variables. Whilst there is no available database for IR scores, the study adopted the same approach from other related studies by Zhou et al. (2017) and Vitolla et al. (2019) in which the <IR> framework content elements are used as an assessment tool. Barth et al. (2017) used the proprietary data from EY excellence awards data, whilst Cosma, Soana and Venturelli (2018) used the events study methodology in which they assessed the quality of integrated reports on the basis of awards ceremonies. Owing to the limited access to proprietary data, a self-constructed scorecard was developed.
Although there is an inherent biasness in the self-constructed assessment tool, this factor was mitigated by the adoption of IIRC content elements. The scoring of integrated reports was thoroughly scrutinised against each point of the content elements. Moreover, in terms of robustness and consistency, firms with high scores were compared with those who were top performers on the performance awards by Ernst & Young and Institute for Chartered Secretaries. Consistent with the IR Framework, firms with higher scores also featured prominently in EY Excellence Awards. The data for the regression models were prepared in three phases. Firstly, the annual IR scores for each company were documented in a spreadsheet. Similar to Lee and Yeo (2016), Dube (2017), a scorecard was developed to record the quality levels of integrated reports. To this end, we make use of spreadsheet tool to assess and record each company's integrated reports based on the elements of IR Framework (IIRC 2013). The selected dimensions and their respective scores were (1) Organisational overview and external environment (5), (2) Governance (5), (3) Business model (10), (4) Risks and opportunities (3), (5) Strategy and resource allocation (5), (6) Performance (5) and (7) Outlook (5). Therefore, the minimum and maximum scores for each firm were 0 and 38, respectively. Zero is a theoretical score, for instance, in which the purported integrated report does not meet any requirement of the content elements of the framework. Each point of content elements met by the reporting firm was awarded one. The total scorecard is shown in Appendix 1.
The annual integrated reports were obtained from the companies' websites.
Secondly, the annual COE and FCERROR data were obtained from the Bloomberg and McGregor BFA Library databases.
The data in respect of the control variables were also obtained from the Bloomberg terminal. All data pertaining to the first two phases were then captured in an excel spreadsheet. The third and final step is an analysis of data through STATA 15 statistical software.

Control variables
The previous literature shows that certain variables are correlated to the COE and analyst' forecast errors (Barth et al. 2017;Conway 2019;Dhaliwal et al. 2011Dhaliwal et al. , 2012Zhou et al. 2017). Likewise, we control the board size (BRDSIZE), which is measured as a total number of both executive and non-executive directors. Alabdullah, Nor and Ries (2018) . King Code of Corporate Governance suggests that independence leads to objectivity in execution of mandated duties (IoDSA 2020). Firm size (SIZE) is measured as a natural logarithm of the total assets, because total assets is widely used as a proxy for firm size. Botosan and Plumlee (2005) found an inverse correlation between the size of the firm and the COE. Socially responsible investment is a variable for sustainability reporting. The JSE FTSE SRI makes data available, pertaining to the firms that met the minimum ESG rating requirements (FTSE 2019). Constituents are those firms that have complied with minimum ESG rating requirements in a period. Zero (0) is coded where the firm is not included in the constituency list, whereas one (1) refers to being a constituent.
Beta measures the security risk in relation to the overall market risk (Sharpe 1964

Descriptive statistics
In order to examine the link between dependent variables (cost of equity and FCERROR), we begin with descriptive statistical analysis.

Correlation
The pairwise correlation is presented in Tables 3 and 4. Table 3 shows the results of the pairwise correlation coefficient between COE and the independent variables (IR_SCORE, BRDSIZE, INDEP, SRI, LEV, BETA, SIZE and MB). As expected, there is a negative relationship between quality of IR and cost of equity. However, the Pearson correlation between the cost of equity and the IR score is insignificant. This is in contrast with results found by Vitolla et al. (2019). There is a significant correlation between the COE and INDEP and LEV variables. This seems to suggest that board's independence and firm's leverage contribute to reducing the COE. The remainder of the control variables does not have significant correlation with the COE. The correlation results (untabulated) for the pre-and post-IR framework also show similar patterns. Table 4 presents correlation results of the analysts' forecast error (FCERROR). Similar to Dhaliwal et al. (2012) and Zhou et al. (2017), the FCERROR has a significant (p = 0.0249) negative correlation with IR_SCORE. The same is true for BRDSIZE, INDEP and SRI. Notably, the volatility of security (VOL) has significant (p = 0.01) negative correlation with analysts forecast errors. This highlights a significant impact of volatility on financial FCERROR (Dichev & Tang 2009;Zhou et al. 2017). The remainder of the control variables (LEV, BETA, SIZE and MB) has insignificant correlation with the FCERROR.
Prior to regression analysis, we performed the diagnostics test to check for the normal assumptions of linear regression (GURU 2019). To this end, we tested for multicolinearity (VIF factor), heteroscedasticity (Breusch -Pagan), normality (PR-Kurtosis) and serial correlation (Breusch Godfrey test). The results (untabulated) primarily show that the normal assumptions of regression are met. After performing diagnostic tests, we run the Hausman test to arrive at an informed decision in selecting the appropriate model between fixed and random effects. The Hausman test provides guidance on the choice between fixed effects and random effects regression models (Hausman & Taylor 1981). The Hausman test results (not included in tables) show that the random effect model was appropriate for the first two models (COE), whilst the fixed effect was selected for the third model (FCERROR).

Regression analysis
Panel regressions were used to test the effect of the independent variables (IR_SCORE, BRDSIZE, INDEP, SRI, SIZE, LEV, MB and BETA) on the dependent variables (Cost of Equity and Analysts' forecast error). Tables 5, 6 and 7 present results from regression analyses. Table 5 shows the results of random effects of generalised least square (GLS) regression processed with Stata 15 statistical software. Similar to previous studies by Zhou et al. (2017) and Vitolla et al. (2019), there is a negative coefficient and significant (p = 0.043) negative relation between the cost of equity and the IR score (IR_SCORE). The results suggest that as the firm improves IR, the cost of equity tends to decrease. As indicated in the literature   review section, some of the costs of equity can be attributed to the agency problem. In this context, the integrated report is linked to reducing risks related to agency costs. BRDSIZE and BETA are also observed to be negatively associated with COE although not significant. The other variables, which significantly affect cost of equity, are SIZE (p = 0.001) and LEV (p = 0.000). Table 6 presents results of pre-and post-IR regression analysis. Compared to the first regression (Table 5), the results in Table 6 are significant and more pronounced. Table 5 includes 37 mining firms that produced IR reports, whereas Table 6 only uses a sample of firms with integrated scores above 30. These firms with high scores were included in a longitudinal panel over 6 year's pre-and post-IR framework. The coefficient of 'BAIR' represents before and after IR framework. The 'BAIR' coefficient is both negative and significant (p = 0.000). Similar to related studies, the results confirm that the adoption of IR may assist a reporting firm in reducing the cost of equity (Johnson 2020;Vitolla et al. 2019;Zhou et al. 2017). Sustainability reporting (SRI) also has a significant (p = 0.10) negative relationship with cost of equity. This indicates that firms who meet the ESG disclosure requirements may realise reduced cost of equity. Such findings highlight the importance of sustainability reporting, which could support the importance of sustainability accounting in IR framework (Flower 2015). LEV (p = 0.01) and MB (p = 0.05) also show an inverse significant relationship with the cost of equity. Table 7 above presents fixed effects regression analysis of the relationship between analysts' forecast error and the independent variables. Similar to the study on the effect of voluntary disclosures on FCERROR by Dhaliwal et al. (2012), there is a significant (p = 0.1) inverse relationship between the IR score and forecast errors and negative coefficient of -0.956. The results confirm the findings of related previous studies (Dhaliwal et al. 2012;Flores et al. 2019;Vitolla et al. 2019). Despite a limited number of observations owing to unavailability of data, the results indicate that IR may assist in minimising analysts forecast errors. The variables that significantly influence FCERROR include company board size (BRDSIZE, p = 0.06) and market-to-book value ratio (MB, p = 0.06).

Conclusion
This study investigated the effects of IR on the COE as well as FCERROR. The research was motivated by the prevalent discourse on whether or not IR offers any benefits to the reporting entity. In this regard, we examined whether firms with higher IR scores experienced a reduction in the COE. Secondly, an analysis of cost of equity was made for before and after IR framework. Furthermore, the study investigated if higher IR scores lead to low FCERROR. To this end, we prepared three regression analyses in order to observe interplay between dependent variables (COE and FCERROR) and IR scores.
After controlling for other variables that could influence the outcome, the study found a significant negative relationship between higher IR scores and cost of equity. Notably, results were more pronounced for before and after IR framework cohort. The more pronounced results could be attributed to more observations and higher quality of integrated reports. We also noted that firms with higher scores generally are bigger (SIZE) than those with lower IR scores. The quality of IR could be largely influenced by the size of the firm. Similarly, the study found significant relationship between higher IR scores and FCERROR. Despite limited data for  Note: Robust standard error estimates were obtained from STATA in order to deal with heteroscedasticity and autocorrelation. **, significant at the 0.05 level (two tailed); ***, significant at the 0.001 level (two tailed). s.e., standard error, BAIR for before and after IR framework, IR_SCORE, integrated reporting score; BRDSIZE, board size; INDEP, independence of directors; SRI, sustainability reporting index; LEV, financial leverage; BETA, market beta; SIZE, firm size; MB, market-to-book value. Note: Robust standard error estimates were obtained from STATA in order to deal with heteroscedasticity and autocorrelation. *, Significant at the 0.1 level (two tailed); **, significant at the 0.05 level (two tailed); ***, Significant at the 0.001 level (two tailed).
probing analysts forecast errors, the results still confirm the possible positive contribution of IR towards minimising analysts forecast errors. Similar to previous literature by Zhou et al. (2017), Vena, Sciascia and Cortesi (2020) and Vitolla et al. (2020b), results of this study confirm that IR may lead to reduction of COE and FCERROR.
The study was limited to only mining firms listed on JSE. Some of the other limitations included unavailability of metrics on databases, particularly, data for analysts' forecasts errors. Future studies could cover a larger sample in order to produce more compelling evidence. Furthermore, the scope of measurement of markets rewards could be larger. Other economic outcomes to be investigated include financial performance, accessibility to capital and firm value. Moreover, given an increase in the number of jurisdictions that have made IR mandatory, it would be interesting to investigate whether its purported benefits are prevalent in other settings. Incorporating questionnaires and interviews with accountants, executives, investors and some shareholders to corroborate the evidence would also provide a broader perspective on the benefits of IR.

Implications for policy makers and researchers
Albeit a small sample, results of the study provide support for IR. Studies of this nature are relevant, particularly, in the corporate reporting space where there are ongoing discourses about the benefits of IR. Integrated reporting framework provides a paradigm shift in reporting, especially the integrated thinking phenomenon and connectedness of both financial and non-financial reporting sections. Policy makers could be influenced by more empirical evidence to support or reject financial gains of adopting IR. This is particularly true in settings where IR is not mandatory. In terms of research, findings may contribute to the existing discourse of firm-based benefits associated with IR.