Does Corporate Governance Influence Voluntary Disclosure? Evidence from India

The purpose of this study is to examine the influence of different corporate governance (CG) attributes on voluntary disclosures (VD) made by 100 companies listed on the Bombay Stock Exchange (BSE) in their annual reports. To this end, the paper uses appropriate panel data regression technique, whereby the results indicate that three CG attributes—board independence, board gender diversity, and its risk management committee—have significant influence on VD. In particular, board independence is found to have weak negative influence on VD while its gender diversity and risk management committee indicate strong positive influence on VD. The other CG attributes, specifically the board size, role duality, ownership concentration, audit committee independence, and nomination and remuneration committees, do not reveal any significant influence on VD. Overall, the finding suggests that one of the conventional attributes of CG, i.e. board independence, acts with VD as an alternate control mechanism to reduce agency costs and protect investor interests. Meanwhile, VD coexists with some of the latest CG attributes, including board gender diversity and its risk management committee, to monitor managers. The results of this paper should be relevant to regulators, practitioners, and other market participants in the Indian context, as well as other emerging markets with similar institutional settings.


INTRODUCTION
In view of information asymmetries in the capital market, voluntary disclosure (VD) is often used by managers, over and above what is required by regulations (Healy & Palepu, 2001;Morris & Tronnes, 2018). Prior literature provide evidence that, decision to disclose information voluntarily is driven by several underlying motives such as reducing cost of capital (Botosan, 1997), improving liquidity (Kim & Verrecchia, 1994), increasing stock compensation (Brockman et al., 2010), getting better analyst coverage (Lang & Lundholm, 1993;Shehata, 2014). Further, literature also highlights different cost associated with VD, such as proprietary costs (Dye, 1985), India employing single conventional CG attribute, the present study examines the effectiveness of some major CG reforms in India by employing some prominent attributes of CG, assists in evaluating the effectiveness of these reforms in promoting transparency and disclosure and highlights the scope of further improvement.

METHODS
The sample used for testing the hypotheses developed in the study includes top 100 non-financial and nonutility companies based on market capitalization listed at BSE as on 31 st March 2014. The study excludes financial and utility companies because of their different regulations and reporting requirements (e.g. Banking Regulation Act, 1949, Electricity Act, 2003 are the regulations under which financial and utility firms are excluded from the sample). The study is based over a period of four years i.e., from 2013-2014 to 2016-17 as the major corporate reforms such as Companies Act, 2013, SEBI's Revised Clause 49, 2014 were implemented during this period. The necessary information regarding VD and CG have been generated through content analysis of annual reports of respective companies and information relating to control variables have been collected from corporate database called Capitaline Plus.
To capture the dependent variable, voluntary disclosure index (VDI) is used as a surrogate measure for the information disclosed by companies voluntarily in their annual reports over the study period. The following steps were undertaken in formulating the VDI: 1) Based on the extensive review of relevant literature (Meek et al., 1995;Botosan, 1997;Ho & Wong, 2001;Eng & Mak, 2003;Gul & Leung, 2004;Lim et al., 2007;Patelli & Prencipe, 2007;Akhtaruddin et al., 2009;Rouf, 2011;Charumathi & Ramesh, 2015), a preliminary list of 131 items was prepared; 2) To ensure that the items are discretionary, they are checked against applicable Indian Regulations such as Companies Act, 2013, SEBI (Listing Obligation and Disclosure Requirement) Regulation, 2015, Converged Indian Accounting Standards (Ind AS), 2016, that affect the reporting requirements of sample companies over the study period; 3) Following the prior studies (Adams & Hossain, 1998;Barako et al., 2006), the list was further scrutinized by three practicing Chartered Accountants who are associated with the institutions that influence corporate disclosure in India to ensure that all items are voluntary. Their feedback is subsequently used in modifying the list of VD items.
Following the above steps, a final list of 69 items is derived, which are further distributed over eight categories. Subsequent to the formulation of list of VD items, one of the contentious issues in literature is scoring of disclosure items. Most of the earlier approaches to scoring are unweighted, indicating only presence or absence of such information without accounting for its quality (Meek et al., 1995;Lim et al., 2007); Akhtaruddin et al., 2009;Charumathi & Ramesh, 2015). To measure disclosure quality, some studies have employed weighted index whereby they mostly followed two approaches for scoring of disclosure itemsfirstly, assigning more score to items disclosed in a comprehensive manner based on the amount of information disclosed (Eng & Mak, 2003;Gul & Leung, 2004) and secondly, assigning more weightage to the items disclosed in quantitative terms (Botosan, 1997;Patelli & Prencipe, 2007). Since VD includes both financial and nonfinancial items it may not be possible for companies to disclose all items only in quantitative terms as there are some non-financial disclosure such as corporate outlook, policy, strategy, etc., which cannot be expressed in quantitative terms but the importance of such information cannot be undermined. Thus, to capture quality of disclosure, this study uses a weighted index whereby a combination of both the prior approaches of scoring have been used and a score of '0' is assigned for absence of information, '1' for partial disclosure of information and '2' for extensive disclosure. Then, each firm's VDI score is calculated as percentage of actual disclosure score obtained against the maximum score.

= ∑ =1 100
Where, 'Nj' is the maximum expected score, 'j' refers to company, 'i' stands for VD items and 't' refers to time.
To capture VD quality 'Xij' assumes the score of '0-2'.  There are a number of factors for large firms to disclose more information voluntarily such as to reduce information gap as the firm gets bigger, information asymmetry problem emerges (Jensen & Meckling, 1976). Further, they are considered to be more sensitive to political cost (Watts & Zimmerman, 1978) and litigation cost (Skinner, 1994). Financial Leverage (LEV) Ratio of total debt by equity share capital plus reserve Firms with higher debt attempt to reduce monitoring costs through better VD (Gul & Leung, 2004;Barako et al., 2006).

Profitability (PROF)
Ratio of earnings before interest and tax to equity share capital Managers of highly profitable firms disclose more information voluntarily in order to avail different compensation arrangement as well as to signal the market (Cerbioni & Parbonetti, 2007;Kaur et al., 2016). Auditor Type (BIG4) '1' for companies audited by BIG4 audit firms otherwise '0' Big4 audit firms generally ensure that their clients reporting practices should be of higher quality as their reputation is primarily associated with it and thus they encourage more VD (Liu, 2015;Nahar et al., 2016).

Liquidity (LIQ)
Ratio of total current assets to total current liabilities Firms with poor liquidity position might opt for more VD in order to justify their liquidity status (Wallace et al., 1994).
To examine the influence of CG attributes on VD, prior works have mostly used pooled ordinary least square (OLS) regression model. However, pooled OLS regression model does not take into consideration the uniqueness/heterogeneity existing within each cross-section unit or time and thus fails to control for potential omitted bias due to firm-specific characteristic related to VD. Moreover, this study also conducted Breusch-Pagan Lagrange Multiplier (LM) test to know the suitability of using pooled OLS regression whereby the Chi-Square value = 286.82, p-value = 0.000, indicates that pooled OLS model is not appropriate for the data set.
Further, to choose appropriate panel data model [Fixed Effect Model (FEM) or Random Effect Model (REM)], the outcome of Hausman specification test is considered whereby the results advocates in favor of FEM. While REM assume that each firm's intercept value are random drawn from a large population of firms, FEM allow each firm to have its own intercept value, thereby controls the effect of time-invariant factors so that the net effect of predictors on the outcome variable can be assessed over time. The functional form of FEM is: Where, αi is the unknown intercept for each firm and uit denotes the error term.
The following model is employed to examine the influence of different CG attributes on VD after controlling the influence of different firm characteristics.
Where, β1 … β13 are the slopes of CG attributes and firm characteristics, αi is the intercept for each firm, uit is the error term, 'i' = 1, …, 100 sample firms; 't' = 2014-2017.  Table 4 shows the distribution of sample firms by year grouped in quartile of VDI_score. Considering the percentage of firms lying in different quartile every year, there exists less variation among them and thus it can be inferred that VDI_score of the sample firms are somehow normally distributed. However, the percentage of firms falling under 4 th quartile over the years is lowest indicating a general tendency among sample firms to disclose less information. Table 5 reports Pearson correlation coefficients for all variables. Consistent with the prior studies, the correlation matrix indicates that VDI is positively correlated with some firm specific characteristics like firm size, leverage and big4 audit firms while negatively correlated with liquidity. The findings also show significant positive correlation between VDI and BS while NRC is negatively correlated with VD. These correlation statistics are consistent with the prior studies (Ahmed & Courtis, 1999;Barako et al., 2006;Akhtaruddin et al., 2009;Liu, 2015) and thus provide some support for the VDI used in this study.   The presence of multicollinearity is tested using both Variance Inflation Factor (VIF) and correlation matrix. VIF for all variables are calculated and the highest VIF obtained is 1.45 which is much below the threshold limit of 10. In addition, the Pearson correlation matrix (Table 5) gives the highest correlation coefficient of 0.48 suggesting that multicollinearity is not a cause of concern in this study. Table 6 presents the regression results. Since the outcome of Hausman test shows that Chi-Square statistic is highly significant, hence the results obtained from FEM are considered for interpretation.  i) *** significant at 1% level; ** significant at 5% level; * significant at 10% level. ii) Normality of the error term was examined by using Jarque-Bera test. The null hypothesis of normality of error term cannot be rejected for both FEM and REM as the Chi 2 statistic of [109.01 (p-value = 2.10)] and [33.25 (p-value = 6.08)] respectively are insignificant in both the cases.

RESULTS AND DISCUSSION
Nevertheless, the results obtained from REM are also somewhat consistent with the FEM results. The observed value of overall R-Square and highly significant value of F-statistic purports in favour of goodness of fit of the model. The results also indicate that size of board is positively associated with VD; however, this lacks statistical significance at conventional level, thus rejecting H1. This finding suggests that though large board is expected to bring greater pool of human resources in developing countries (Jackling & Johl, 2009), in Indian context large proportion of family owned firms leads to mere increase in board size due to appointment of family members who does not actively contributes towards its overall decision making (including VD decision).
Contrary to the expectation, proportion of INDs on board is found to have inverse influence on VD. This finding is consistent with Eng & Mak (2003); Gul & Leung (2004); Barako et al. (2006) suggesting that INDs act as a substitute for VD. Another explanation for this finding in case of Indian companies might be due to the prevalence of closely held ownership structure whereby substantial owners generally develops ties with INDs and thus they may not be considered independent in real sense. Moreover, CG reforms in India over the past few years emphasized on implementing more strict norms for classifying INDs advocates in favour of this reasoning.
Though, coefficient of role duality is consistent with its expected negative impact, the hypothesis is not statistically significant. This is parallel with the findings of Ho & Wong (2001); Arcay & Vázquez (2005); Barako et al. (2006); Liu (2015) suggesting that separating both the positions does not necessarily improves VD. Further, in case of sample firms, since most of them (67%) have already adopted a separate leadership structure, in can be inferred that there are limited instances for this attribute to have statistical significance.
Consistent with the expected H4, female representation on board reveals strong positive influence on VD. Though, women representation on board for the sample companies is relatively meagre as compared to developed countries like Norway, Spain, Iceland, Australia, Denmark, Germany etc. yet this finding suggests that this minimum percentage of women directors are given equal opportunity to play active role in the decision making process and thus their strong monitoring behaviour assists in reducing information asymmetry through better VD. Moreover, in Indian context an upward trend in participation of women on corporate board can be noticed over the past few years as discussed under descriptive statistics section indicating the interest of practitioners to increase more women involvement in their decision-making process.
Regarding ownership concentration, though its coefficient is negative, it lacks statistical significance at conventional level, thus rejecting H5. One possible reason for such insignificant finding might be due to passive attitude of substantial owner with respect to disclosure as they have more efficient and timely channels of obtaining relevant information (Eng & Mak, 2003;Donnelly & Mulcahy, 2008). The findings also reveal insignificant influence of ACI on VD, thus discarding H6, however supports the findings of Ho & Wong (2001);Othman et al. (2014). This result cast doubt about the motive for formation of an independent audit committee because if it is actually formed for the purpose of overseeing firm's reporting process, it is ought to assist in reducing information gap through better disclosure. Unfortunately, in case of India active participation of substantial owners in corporate decision making alleviates information asymmetry problem, whereby formation of an independent audit committee is often viewed as compliance requirement rather than genuine necessity which might account for such insignificant impact.
In relation to H7, NRC does not reveal any significant influence on VD extending support to the finding of Allegrini & Greco (2013). This finding suggests that in Indian context, the functioning of NRC is undermined due to frequent intervention of dominant shareholders and promoters in selection of directors and KMPs and fixing their remuneration. Finally, consistent with the expectation in H8, existence of RMC reveals strong positive influence on VD. This finding suggests that as RMC is formed with members of the board and senior executives having better understanding of business complexities and special expertise in risk management, it assists them in maintaining better transparency and disclosure and thus, upholding resource dependency perspective (Ali et al., 2018). With regard to control variables, only firm size is found to have statistically significant positive influence on VD, thus supporting agency theory and positive accounting theory argument.

CONCLUSION
The study aims to investigate the influence of some major CG reforms pertaining to board attributes and ownership concentration on VD after controlling the influence of some firm characteristics for top listed companies on BSE over 2014-2017. Eight research hypotheses stating the expected influence of different CG attributes on VD are examined, based on comprehensive data set which are manually collected from annual report of sample companies as well as from corporate database 'Capitaline Plus'. The findings reveal that board independence, board's gender diversity and risk management committee have significant influence on VD. In particular, board independence is found to have weak negative influence on VD while board's gender diversity and risk management committee indicates strong positive influence on VD.
The remaining CG attributes such as board size, role duality, ownership concentration, audit committee independence and nomination and remuneration committee does not have any significant influence on VD. Overall, finding suggests that, one of the conventional proxies of CG i.e., board independence has negative influence on VD, while some recently introduced CG mechanisms like board's gender diversity, risk management committee positively influences VD. Thus, it can be inferred that, the direction of impact of CG on VD depends on the type of CG mechanism considered, as under conventional CG mechanism, level of VD decreases, because it is considered as costly affair together with the presence of such mechanism while some latest mechanisms of CG boosts VD as they emphasize more on transparency and disclosure in order to maintain investors' confidence.
The result of this paper should be of relevance to regulators, practitioners and other market participants in Indian context and other emerging market having similar institutional setting. Firstly, the positive influence of gender diversity on VD extends support to the recent regulatory initiative of mandating minimum one women director on board. It further encourages regulators and practitioners to increase current levels of women representation of board in order to promote transparency among Indian corporate. Secondly, the coexistence of risk management committee and VD suggest the regulators to mandate formation of separate risk management committee for all listed companies so as to have better and more transparent risk management process for welfare of all stakeholders concerned. Thirdly, the negative impact board independence on VD suggest the regulators that though companies are appointing independent directors to fulfil statutory requirement, they are not functioning in their true spirit due to convergence of owner-manager interest. Finally, the finding also draws a clear picture about ineffectiveness of other CG mechanisms in promoting transparency and thus signalling the need for further improvement of CG structure in Indian context.
The study is based on annual reports as the only one avenue of VD, without considering other ways of VD such as websites, press release, etc. constitutes a limitation. Further, this study employs some variables of CG only whereas ownership structure related variables such as family ownership, managerial ownership are also prominent and need be included in future to know their impact on VD. Moreover, since prior literature provide evidence for positive relationship between VD and firm value, studies in future can extend the linkages between CG, VD and firm value.