“The relationship between corporate forward-looking disclosure and stock return volatility”

The study assesses corporate forward-looking disclosure by measuring four attributes, namely disclosure quantity, disclosure coverage, disclosure concentration and disclosure quality, through a sample of 34 listed firms in the Bahrain Bourse from 2014 to 2017. The study also investigates the relationship between these attributes and stock return volatility. Regression analysis has been employed with five different models to examine the relationship between the four attributes of corporate forward-looking disclosure and stock return volatility. The main finding of this study agrees with the results of Bravo et al. (2009) who found that the selection of a specific disclosure index could influence crucially the results of the analysis. In addition, stock return volatility has a statistically significant negative association with the three attributes of forward-looking disclosure, namely disclosure quantity, disclosure coverage and disclosure quality. In contrast, it has a non-significant association with the fourth attribute of forward-looking disclosure, disclosure concentration. This study provides a novel contribution to disclosure quality studies by being the first study to examine forward-looking disclosure quality attributes in the Kingdom of Bahrain.


INTRODUCTION
Disclosure novels have attracted a great interest in accounting literature.Theoretical arguments in literature (Lang & Lundholm, 1993;Cormier et al., 2010) suggested that the increase of corporate disclosure, in particular, disclosure quality, has a positive influence on capital markets in different ways, such as it reduces cost of capital, information asymmetry, and stock return volatility (SRV).However, questions on the measurement of disclosure quality and types of the information disclosed are still open (Bravo et al., 2009).Hussainey (2004) classified information disclosed in corporate annual report into "backward-looking information" and "forward-looking information".The first one refers to disclosures on the past financial results."Forward-looking information is the class of information that refers to future forecasts and current plans that enable different users to assess a future corporate performance" (Aljifri & Hussainey, 2007, p. 882)."The role of corporate forward-looking disclosure in capital markets is today crucial since the economic environment is too dynamic to rely on historical in-formation only" (Menicucci, 2013(Menicucci, , p. 1667)).Such disclosure enables users to predict company's future financial performance (Athanasakou & Hussainey, 2014).Bravo (2016, p. 123) stated that "forward-looking information has become crucial, since historical information could be insufficient for investors.Both organizations and researchers have stated the significance of forward-looking information in order to improve the forecasts about a company and ease decision-making processes in capital markets".Prior studies offered answers to the question why disclosure can affect SRV.For example, Bushee and Noe (2000) pointed out that more disclosure leads to reduced information asymmetries, consequently, decreases surprises about a firm and helps to make its stock price have low volatility.Easley and O'Hara (2004) showed that disclosure quality affects corporate stock volatility and its cost of capital.
The current study has two objectives.First, it measures a specific type of disclosure, forward-looking disclosure (FLD), by assessing four attributes, namely disclosure quantity, disclosure coverage, disclosure concentration and disclosure quality, for a sample of 34 Bahraini listed firms from 2014 to 2017.Second, it examines the effect of the four attributes of FLD on SRV.
The importance of this study builds on the unique demand for forward-looking information and its impact on critical matters such as SRV.Such importance has two streams.First, several studies documented the importance of future information for investors (AICPA, 1994;FASB, 2001;ICAEW, 2002).For example, AICPA (1994) identifies five categories of information that companies should disclose in their financial reports, such as "the management's analysis of financial and nonfinancial data; information on managers and stakeholders; forward-looking information; and finally company background".Other professional bodies (IASB, 2010;ICAEW, 2000ICAEW, , 2002) ) argued that different users of financial reports need future information that helps them to improve their expectations about business performance.International Accounting Standards Board (IASB) in its study titled "Management Commentary: A Framework for Presentation" pointed out that "forward-looking information might present an over-optimistic picture of the entity" (IASB, 2010, par.BC 39), the IASB points out that "management should disclose the assumptions used in providing forward-looking information" (IASB, 2010, p. 18).Beretta and Bozzolan (2008) argued that FLD can help in explaining future earnings, therefore, such disclosure is considered useful for users of companies' financial reports.The authors found a significant positive relationship between the quality of corporate disclosure and the analysts' forecasts for a sample of Italian firms.
Second stream reflects the association between FLD and SRV.Prior studies such as Hussainey and Walker (2009) provided evidence that the quality of disclosure can play a critical role to improve stock market decisions and to provide better expectations about future earnings.In the light of scarcity of studies on emerging markets, the current study has a high value, since it is based on one of these markets, namely the Kingdom of Bahrain as a member of the Gulf Cooperation Council (GCC).
This study contributes to the current literature on FLD by assessing such disclosure in Bahraini capital market.To the best of the authors' knowledge, this study is the first to assess such disclosure in the Bahraini capital market , as well as it investigates the effect of FLD on SRV.The results of our study imply practical implications for a number of interested parties, such as managers, investors and regulators.
The paper is organized as follows.Section 1 presents an overview on agency theory.Section 2 reviews the relevant literature and develops the hypotheses of the study.Section 3 presents background on Bahraini capital market.The research method is provided in section 4.Last section shows the empirical analysis of the study.

OF THE STUDY)
Different theories can be used to explain managers' motivations for voluntary disclosures.The present study adopted agency theory to explain the potential association between the four attributes of FLD and SRV.From an agency perspective (Core, 2001 Beretta and Bozzolan (2008), the current study adopted four attributes to measure FLD, namely quantity, coverage, concentration and quality.

Disclosure quantity attribute
The current study followed Bravo (2016, p. 125) who measured the quantity disclosure (QUTD) "as the amount of forward-looking information disclosed by companies taking into account only number of units (sentences), as a coding unit, with forward-looking information".Every sentence with forward-looking information is considered (Mousa & Elamir, 2018).The current study used "a simple index that only captures absolute quantity of disclosure" that was suggested by Bravo et al. (2009, p. 264), as shown in the following equation: where i F is number of sentences with forward-look- ing information disclosed by company .i i Max is the maximum number of sentences with forward-looking information disclosed by company i across the sample.
i Min is the minimum number of sentences with forward-looking information disclosed by company i across the sample (Bravo et al.,  2009, p. 264).

Disclosure coverage attribute
Prior studies, such as Beattie et al. (2001Beattie et al. ( , 2002aBeattie et al. ( , 2002bBeattie et al. ( , 2004)) Width depends on both the coverage of relevant topics (or subtopics) of the framework and the dispersion of disclosure across different topics (or subtopics)".The current study used the approach of Beretta and Bozzolan (2008, p. 344) to measure disclosure coverage (COVD), which "ranges from 0 to 1 and assumes its maximum value when a company makes disclosure over each of the topics (subtopics) considered".

Disclosure concentration attribute
At the same time, Beretta and Bozzolan (2008, p. 344) suggested concentration of disclosure (COND) as another dimension that should be considered in measuring disclosure quality.Beretta and Bozzolan (2008, p. 344) pointed out that COND "refers to how concentrated disclosed items are and corresponds to the standardized entropy index (COND)" where ij P -number of information disclosed in sub-topic j divided by total disclosure of compa- ny , i st -number of topics (or sub-topics), ln is a natural logarithm" (Beretta & Bozzolan, 2008, p. 344; as quoted also by Bravo et al., 2009, p. 260).

Disclosure quality attribute
Disclosure quality (DQA) is measured as the average of the above three attributes (QUTD, COVD and COND) as: .

Corporate forward-looking disclosure and stock return volatility
The topic of SRV has attracted the attention of many stakeholder groups in financial markets, as well as researchers and professional associations.Based on the theoretical framework of the study, an agency perspective expects a significant association between SRV and FLD, since it is associated with improving the anticipation of future earnings and reducing information risk, consequently, FLD affects SRV.In the current study, FLD was measured by four attributes (QUTD, COVD, COND and DQA) therefore, the following hypotheses (H) are formulated: H1: There is a significant relationship between QUTD and SRV.
H2: There is a significant relationship between COVD and SRV.
H3: There is a significant relationship between COND and SRV.
H4: There is a significant relationship between DQA and SRV.

WHY BAHRAINI CAPITAL MARKET
The As the current study used two approaches to analyze forward-looking information, namely QDA Miner software package analysis and the manual content analysis, Pearson and Spearman correlation analyses had been adopted to evaluate the linear correlation between the two approaches.Strong significant positive correlations at 1% level were found between the two types of analyses (Pearson correlation is 0.92 and Spearman correlation is 0.91).Such results provide evidence on the reliability of using the QDA Miner software package.

Definition of the study's variables
To investigate the relationship between FLD and SRV as the second aim of the study, multiple regression analyses were conducted.The dependent variable, SRV was measured similar to Bravo (2016).On the other hand, four independent variables were included (QUTD, COVD, COND, DQA) to reflect the attributes of FLD.Moreover, in line with several studies, seven control variables were selected to include in the regression models to control for potentially omitted relationships, namely leverage, foreign ownership, financial performance of the firm, the firm age, firm size and independence of the board.

SAMPLE AND DATA COLLECTION
By the end of 2017, 43 companies were listed in the "Bahrain All Share Index" as the main index of the BHB.Table 2 shows the distribution of these firms by sectors.The current study applied a number of criteria to include any company in the sample: (1) companies had to be Bahraini firms that were listed on BHB from 2014 to 2017 continuously; (2) availability of complete annual reports.In addition, closed company sector and non-Bahraini companies are excluded.After applying previous

Hotels and tourism 4
Closed company 2 Non-Bahraini companies 1 Total 43

Descriptive statistics
The results of the descriptive statistics for the current study are shown in Table 3, the four attributes of FLD, and seven control variables.COND has maximum mean (0.937) among four attributes of FLD, while QUTD has minimum mean (0.533).
With respect to the standard deviation, QUTD has the highest variation (0.254) among them, while COVD (0.054) has the lowest variation.

The assessment of FLD across the sample of the study
To achieve the first objective of our study, FLD with four attributes was assessed through 34 listed firms in BHB (from 2014 to 2017), as shown in Table 4 and Tables B1, B2 and B3 in Appendix B.
To investigate the effect of using the four attributes of FLD on the rank-orderings of companies, the ranking of companies (year by year) was presented in Tables 4 Bahrain Bank) comes first in all four indices.CINAMA comes at the bottom when using COVD index, while the same company comes 17 for QUTD and 18 for both COND and DQA indices.
In Table B3, AUB bank has score 1 for three indices (QUTD, COND and DQA), while it has score 2 in COND.However, some companies have a high score in one index and, at the same time, they have a low score in other indices.Finally, it should be noted that through the ranking in Table 4 and Tables B1, B2 and B3 (see Appendix B for more details), banks and financial firms ranked first in the rankings order within the four different indices of FLD (QUTD, COVD, COND and DQA).Our main finding of the study agrees with the results of Bravo et al. (2009) who found that the selection of a specific disclosure index influences crucially the results of the analysis.DQA.The lowest correlation is -0.04, which exists between SRV and COND.QUTD has significant positive correlations at the 1% level with the three attributes of FLD (COVD, COND and DQA).Table 5 reveals that SRV has significant negative correlations (at the 5% level) with QUTD and DQA (-0.17

Regression analysis
To test the hypotheses developed earlier in this study, different statistical models were performed to examine problems, such as multicollinearity and heteroscedasticity.Results revealed that these problems do not exist for all the models.To study the effect of the variables of the study on SRV, Table 6 presents five models, including four attributes of FLD and seven control variables.The following models were proposed, in which SRV is a function in all of these variables:  Model 1 presents seven control variables for SRV based on previous literature, which have been considered.Model 1 is not statistically significant and it has R 2 (18%).Model 2 considers QUTD as an independent variable, and control variables.This model explains the effect of QUTD individually on SRV beyond the control variables.The R 2 in model 2 is 27%.There is an increase about 9% compared with model 1.QUTD variable has a negative association with SRV (significant at 5% level).Findings of model 2 reveal that the increase of QUTD results in an incremental reduction in SRV.These findings confirm theoretical perspective of agency theory, which expected that the more the increase of QUTD, the more the decrease in SRV, which reflects the impact of FLD on financial markets.Hence, hypothesis H1 is accepted.This finding agrees with the results reported by Mousa and Elamir (2018).
In Table 6, model 3 includes COVD plus the control variables.The model is statistically significant (at the 5% level).The regression analysis shows how this variable, COVD, alone helps to explain the changes in SRV beyond that of the control variables.The explanatory power for this model is 26% with an increase 8% than model 1.In addition, the new independent variable, COVD, has a significant negative effect on SVR.This finding supports the hypothesis H2, consequently, it is accepted.
Concerning  (Bravo, 2016).In line with agency theory, FLD can be a useful tool to reduce information asymmetry or agency costs, which can play a unique role in having an impact on stakeholders' perception from the stakeholders' perspective.The overall results of the current study support that FLD has significant effects on capital markets and helps to reduce SRV.Our results are consistent with prior studies such as Sahore and Verma (2017) and Jayshree (2012) who argue that more disclosure of information helps investors to take reliable decisions and avoids confusion.Unclear information or no information often leads to wrong decisions.

CONCLUSION
The current study has measured FLD by considering four attributes (QUTD, COVD, COND and DQA) in annual reports for a sample of listed firms in BHB from 2014 to 2017.The study's results revealed that firms have different score in each index.Consequently, their rankings differ in the four indices related to FLD attributes, which supports the argument on using different disclosure indices impacts on the results of disclosure studies.Banks and financial firms obtained the first 17 positions in the four FLD indices in most cases.This is due to the vital role played by banks and financial firms in the economies of countries and the importance they represent to a large number of investors.Therefore, especially banks are subjected to strict control by governments and international legislation.Moreover, the current study investigated the relationship between the four attributes of FLD and SRV.The main findings of the regression analyses showed significant negative relationships between SRV and three attributes of FLD (namely QUTD, COVD and DQA), which supported the hypotheses H1, H2 and H4, in contrast, H3 was rejected, because, in model 4, the coefficient of the variable COND (-4.265) is not statistically significant with SRV.
This study contributes to the current literature on FLD by assessing FLD in Bahraini capital market.It implies practical implications for a number of interested parties, such as managers, investors and regulators.Since several studies have documented the importance of future information for different stakeholder groups, this study meets the unique demand for FLD and its impact on critical matters, such as SRV for these groups.
The study is not free of limitations.Firstly, the sample size is small, which can be increased in future research by including other countries.The results of the study cannot be generalized to other countries.Since each country has different economic status and regulations.Finally, the study has used content analysis, which is inevitably subjective.

Table 1 .
Definitions of the study's variables SRV is measured as one plus the natural logarithm of the standard deviation of daily stock returns".iftheannualreport of company i discloses information about the subtopic, 0 otherwise(Beretta & Bozzolan, 2008, p. 344as quoted also byBravo et al., 2009,  p. 260) where 1 ij INF = P -number of information disclosed in sub-topic j divided by total disclosure of company , i st -number of topics (or sub-topics), ln is a natural logarithm (Beretta & Bozzolan, 2008, p. 344, as quoted also by Bravo et al., 2009, p. 260).

Table 2 .
Summarized firms' distribution by sectors

Table 3 .
Descriptive statistics , B1, B2 and B3 based on the values of each index.It can be noted that rank-orderings of companies differ among different indices.For example, in Table4, AUB (United Ahli Bank) comes first in the ranking when using the quantity index (QUTD) and it comes second in other indices, while CPARK (non-financial company) comes number 24 when using QUTD index, number 30 in COVD index, number 34 in COND index and number 18 in DQA index.In the same vein, in TableB1(see Appendix B),, BANDER comes number 31 in the three indices (QUTD, COND and DQA) and it comes number 22 with COVD.CPARK comes 20 in the ranking when using both QUTD and DQA indices, while it comes 34 and 27 in the COVD and COND indices, respectively.In TableB2, in 2015, NBB (National Note: Stock return volatility (SRV), disclosure quantity (QUTD), disclosure coverage (COVD), disclosure concentration (COND), and disclosure quality (DQA), financial leverage (LEV), foreign ownership (FOWN), firm performance (ROA), the age of the firm (AGE), independence of the board (BoD), firm size (FSIZE), type of industry (TYPE).Number of firms 34 covering the period from 2014 to 2017 (136 firm-year observations).

Table 5
presents the correlation coefficients of the variables of the current study.Most correlations are statistically significant.The highest correlation is 0.97, which exists between QUTD and

Table 6 .
model 4, COND variable plus the control variables are considered.The model is significant at the 10% level.The regression analysis shows how COND helps to explain the changes in SRV beyond that of the control variables.Regression analysis of the study Notes: 1.Stock return volatility (SRV), disclosure quantity (QUTD), disclosure quality (DQA), disclosure coverage (COVD) and disclosure concentration (COND), financial leverage (LEV), foreign ownership (FOWN), firm performance (ROA), age of the firm (AGE), independence of the board (BoD), firm size (FSIZE), type of industry (TYPE)]. 2. Number of firms 34 covering the period from 2014 to 2017 (136 firm-year observations).3.* Significant at the 0.10 level (two-tailed); ** at the 0.05 level (twotailed); *** correlation is significant at the 0.01 level (two-tailed).increase in DQA will decrease the SRV.A significant negative relationship (at 5% level) was reported between DQA and SRV.This finding supports the hypothesis H4 that was developed earlier in the study.Our results are consistent with The current study suggests several trends for future studies.For example, studying the effect of other factors, such as economic and corporate governance factors, on FLD can be a promising avenue.Other directions are exploring the effect of legal environments and stockholders' rights on FLD.