Is corporate environmental disclosure associated with firm value? A multi- country study of Gulf Cooperation Council firms

Several studies have found a relationship between corporate social and environmental disclosure and firm value or accounting profitability. Where environmental disclosure has been the focus, though, only single-country studies have been published; and most of the previous research concerns the developed world. This study examines the association between corporate environmental disclosure (CED) and firm value (FV) in the Gulf Cooperation Council (GCC) countries, where CED has been increasing from its previous low base. Findings from a multi-country sample of 500 firm-year observations using a 55-item unweighted environmental disclosure index suggest that CED is significantly and positively related to FV as measured by Tobin’s Q (TBQ). The relationship is robust to using a weighted version of the disclosure index, individual countries and environmental disclosure sub-indices. Some evidence of a positive relationship between CED and return on assets (ROA) is also found, but even where statistically significant, the relationship is much weaker than in the case of TBQ. For empirical and theoretical reasons, we recommend that future studies pay greater attention to market-based proxies, if possible, when investigating the value relevance of CED in both developed and developing countries. Our results suggest that both managers and policymakers in GCC countries should take a positive view of expanded CED.


Is corporate environmental disclosure associated firm value?
A multi-country study of Gulf Cooperation Council firms

Introduction
In a world of climate change, natural resource constraints and other socio-environmental pressures, corporate sustainability has been increasingly pushed to the forefront of corporate decision-making and communication. Corporate Environmental Disclosure (CED)defined here as the provision of information to external parties about an organisation's environmental policies, activities and performancehas become an important source of insights into the efficiency and effectiveness of corporate sustainability strategies (Deegan, 2002;D'Amico, Coluccia, Fontana, & Solimene, 2016;Shahab, Ntim, Chengang, Ullah, & Fosu, 2018). Ideally, CED should include crucial environmental matters and their influence on businesses' future position and performance, uncertainties and risks, material items of expense or income, and environmental policies (Brammer & Pavelin, 2008;Iatridis, 2013;Shahab, Ntim, Yugan, Ullah, & Ye, 2020). Such matters are likely to be of interest to a wide range of users including, increasingly, investors that are concerned about environmental sustainability, either for its own sake or because of its business implications. High-quality CED can also play a symbolic role as an indicator of corporate transparency, leading to enhanced corporate reputation (Deegan & Blomquist, 2006;Hassan & Romilly, 2018;Haque & Ntim, 2018).
There is considerable literature on whether environmental or socially responsible performance enhances firms' financial performancethe question of the so-called 'business case'. (See Brooks and Oikonomou (2018) for a useful recent review of some of the important themes in, and conclusions from, this broad literature.) As that literature has developed in sophistication, greater attention has been paid to which particular elements of environmentally or socially responsible behaviour are associated with improved firm financial performance. CED might be seen as either complementary to corporate environmental practices, in that it provides information about them, or, alternatively, a further example of 'good' practice in itself. However, in either case, whether symbolically (as transparency) or substantively (in providing relevant information for shareholders and other external stakeholders), the provision of environmental information has been found to be associated with enhanced firm value (FV) of corporations (e.g. Broadstock, Collins, Hunt, & Vergos, 2018;Cormier & Magnan, 2013;Clarkson, Fang, Li, & Richardson, 2013;Haque & Ntim, 2020;Plumlee et al., 2015;Iatridis, 2013).
environmental disclosure index, we are effectively addressing all three of these limitations.
Furthermore, one of the benefits of multi-country studies is that they enable an identical methodology to be applied to more than one country, which is more efficient and more reliable than trying to compare findings based on single-country studies. A further advantage is that they enable a larger sample to be created when researching smaller stock markets, assuming suitable statistical controls are then put in placeas is the case in this study.
We examine five of the six countries that are full members of the GCC (Kuwait, Oman, Qatar, Saudi Arabia and the UAE), 2 a body formed in 1981 to advance economic development and cooperation in the region. The nations share many cultural as well as economic characteristics, being 'Arab', Muslim-majority countries (Hampden-Turner & Trompenaars, 2005;Hofstede, Hofstede, & Minkov, 2015). The GCC provides an ideal context for our study since its substantial economic growth has been achieved primarily through environmentally sensitive industries. The five selected GCC countries together hold 45% of global oil reserves ( Al-Shammari, Brown, & Tarca, 2008), and they also suffer from high environmental pollution, with the UAE considered to be the most polluted country in the world in relation to small particulate matter (World Bank, 2015). While there is evidence indicative of some varying levels of adoption and use of CED in the GCC and the wider MENA (the Middle East and North Africa) region (Gerged, Cowton, & Beddewela, 2018), whether CED in the GCC region is associated with firm value (FV) is as yet unknown. That 1 Formally, The Cooperation Council for the Arab States of the Gulf, but still usually known by its original name. 2 Bahrain has been omitted for data accessibility reasons.
is the question addressed by this research. In doing so, our study contributes to the existing literature as follows. First, it provides new evidence on the value relevance of CED using a detailed environmental disclosure index that has been sufficiently developed to provide a comprehensive understanding of the financial consequences of a company's decision to report on its environmental activities. Second, we offer new empirical evidence about the CED-FV nexus from an underresearched developing region, namely the GCC region. Third, our study provides, to the best of our knowledge, the first multi-country and/or regional investigation of the relationship between environmental disclosure and firm value.
The remainder of the paper is organised as follows. The next section provides further background and theoretical perspective, reviews previous studies that have examined the relationship between CED (or related disclosure types) and FV and develops the hypotheses. Section 3 outlines the research method, and Section 4 presents the findings. Section 5 provides discussion and conclusions.

Environmental regulations, reporting and developments in GCC countries
While the GCC region was once known for negligible levels of CED by listed firms (Eljayash, James, & Kong, 2012), recent research indicates growth in disclosure across the region (Gerged et al., 2018;Eljayash et al., 2012 (Broadstock et al., 2018) and so, in turn, reap business gains or at least stave off challenges to their legitimacyto the benefit of firm value.

Neo-institutional theoretical framework for environmental disclosure
With its understanding of the way in which firms deal with different types of pressures, neoinstitutional theory provides a suitable conceptual narrative for understanding the context of CED.
Di Maggio and Powell (1983) have identified three specific types of institutional isomorphic pressure, denoting the differing levels of conformance expected of organisations by external stakeholders; mimetic, normative and coercive. Coercive isomorphism would compel substantive engagement in certain practices as a result of their being required by powerful external stakeholders, such as a country's national government through legislation, while normative isomorphism would result from a need to align organisational practices with the collective societal norms of expected behaviours as promoted by institutional stakeholders such as NGOs or professional accounting bodies. The influences described in section 2.1 provide examples of both coercive and normative institutional pressure. In relation to CED in the GCC region, in the absence of coercive or normative pressures, mimetic isomorphism is more likely. This is a type of comparative behavioural pressure, pressing organisations to follow the CED practices of their competitors in order to level the playing field and thereby maintain their competitive advantages within the organisational field. Organisational conformance arising out of adhering to these institutional pressures would ultimately enable organisations to attain legitimacy from salient institutional (and other) stakeholders.
Prior literature has adopted various theoretical perspectives to examine corporate engagement in CED activities, including stakeholder, legitimacy and agency theories (Reverte, 2009). Nevertheless, neoinstitutional theory provides the most substantive explanation of the influence of external factors upon CED, and the subsequent organisational performance impacts (Brammer, Jackson, & Matten, 2012;Campbell & Hollingsworth, 1991;Campbell, 2007;Ioannou & Serafeim, 2012). In this regard, institutional conformance, specifically in relation to CED, has been found to influence the market value of firms (Cormier & Magnan, 2017), due to reduced transaction costs (North, 1990). Therefore, we draw on neo-institutional theory to emphasise the influential role played by external institutions 8 in engendering CED and its organisational outcomesincluding the possibility of enhanced FV, as reflected in the hypotheses we develop.

Empirical literature review and hypotheses development
In our review of the empirical literature, we focus on the value relevance of environmental-related and similar disclosures. CSD index that included 21 environmental items, but they only investigated the relationship between overall disclosure and FV. Likewise, Malik and Kanwal (2018) only examined the overall CSD-FV nexus, although the disclosure index they adopted from Bayoud, Kavanagh and Slaughter (2012) included seven environmental items.
INSERT In terms of measuring disclosure, the use of an index method has become very common in disclosure studies in general, tending to displace content analysis methods such as the counting of words or sentences (Malik & Kanwal, 2018). Table 1 shows that the use of an index to measure disclosure is by far the most popular method in the studies that we review. The one exception amongst CED-FV studies (Panel A2) is Broadstock et al. (2018), which has a much narrower focus on greenhouse gas emissions. Generally, studies employ a simple disclosure index that uses a binary dummy variable to indicate the presence or absence of some item of information, although Khlif et al. (2015) attempt to quantify the quality of the information by giving higher scores for financial, quantitative and qualitative disclosures respectively. Bloomberg ESG scores are also used, by Zuraida et al. (2018); they are useful for some purposes, but they do not provide much detail about environmental issues.
Researchers, therefore, tend to hand-collect data if they are compiling a disclosure index to cover a good range of issues within a particular category. In some cases (e.g. Plumlee et al., 2015), the contents of stand-alone sustainability (or similar) reports are analysed, but in most cases, it is the corporate annual report -which is a company's main accountability mechanismthat is analysed.
The studies in Table 1 use a variety of proxies to measure the financial consequences of disclosure, which may account for some of the variations in resultsalthough there is a general finding of a positive association between environmental disclosure and firm value (e.g. Broadstock et al., 2018;Clarkson et al., 2013;Plumlee et al., 2015;Iatridis, 2013). Some studies of the impact of disclosure employ just a single proxy (e.g. Khlif et al., 2015), but most use two or three. Although there is little or no discussion of the differences between them or their relative merits, the proxies can be categorised as either market-based or accounting-based. Many studies use both types. The accounting-based proxies generally relate a profit figure to a balance sheet denominator, for example, ROA (return on assets) or ROE (return on equity). Given the nature of the financial statements from which they are drawn, the accounting-based proxies can be viewed as backwards-looking. They also do not measure FV as such, since they are a single-period measure of profitability. On the other hand, market-based measures can be seen not only as relating to FV but also as more forward-looking, since the share price is expected to reflect expectations about the future effects of actions and policies, including those that are reflected in, or relate to, environmental disclosure.
We use a market-based measure as our principal dependent variable because of its forward-looking nature and ability to capture firm value. Although others are used (e.g. cost of equity capital), Tobin's Q (TBQ) is common (see Table 1); we follow suit. For our supplementary, accounting-based proxy we use ROA, in common with many other studies; unlike ROE, it does not reflect how the assets are financed, only how they are used. The two measures give rise to two versions of our hypothesis.

H1:
There is a positive relationship between CED and market value (TBQ) in the GCC region. H2: There is a positive relationship between CED and profitability (ROA) in the GCC region.
In the next section, the research design, including sampling criteria, research methods and analysis, will be discussed.

Sample
The sample for this study is based on all 405 non-financial companies that are listed on the stock exchanges of the five selected GCC countries and have complete data for five years (2010)(2011)(2012)(2013)(2014). The financial sector is excluded for several reasons. First, its effects on the environment are primarily indirect (Thompson & Cowton, 2004). Second, financial firms, such as banks and insurance companies, are heavily regulated, which could differently influence their performance and disclosure practices (Guest, 2008;Huang & Wang, 2015;Yermack, 1996) and the relationship with FV.
Additionally, excluding financial firms is in line with much previous literature (e.g. Baber, Liang & Zhu, 2012;Haniffa & Hudaib, 2006;Ntim, 2016;Siregar & Utama, 2008). Based on stock exchange definitions, the sampled companies are divided into two broadly defined sectors, Industrial 4 and Services, 5 since the nature of a sector can influence CED, including in the MENA region (Gerged et al., 2018), and it is likely also to have an impact on the relationship with firm value.
Earlier studies of social and environmental disclosure show that, in addition to the industry sector, firm size also tends to have a considerable impact on firm disclosures ( Beattie, McInnes, & Fearnley, 2004;Lang & Lundholm, 1993;Ntim, 2016;Oyelere, Laswad, & Fisher, 2003). There are various options for dealing with this, including the selection of the largest and smallest firms from a stratified population. Following Ntim (2016) (and see Gerged et al. (2018)), we select the five largest and the five smallest firms (based on the average of their total assets over the five-year period) from each sector within each of the five selected GCC countries. 67 Therefore, the final sample comprises 100 listed companies over five years, resulting in 500 firm-year observations. The collection of five years of data permits the running of a panel data analysis, which provides opportunities for much more robust insights into relationships than using, for example, cross-sectional analysis. Table 2 provides an overview of the selection process and the financial characteristics (log of total assets) of the resulting sample.

Variables and data
The data for the research variables are hand-collected from annual reports, 8 Ntim, 2016;Williams, 1999). We use an index method, as do most of the studies shown in Table 1.
There are two particular decisions to be made regarding the use of a disclosure index. First, and most important, is the choice of particular disclosure items. The aim was to develop a disclosure index that was both sufficiently comprehensive and granular to meet the aims of the study. The most comprehensive CED studies tend to have been conducted in the developed world, so they were used as the initial basis (Wiseman, 1982;Gray, Kouhy & Lavers, 1995;Hackston & Milne, 1996).
However, the appropriateness of Western CED instruments to measure CED within the different socio-cultural contexts of developing countries has been criticized (e.g. Gray & Kouhy, 1993;Bebbington et al., 1994;Baydoun & Willett, 1995;Belal, 2001;O'Donovan, 2002), so the draft disclosure index was developed further by checking for additional disclosure items used in CED studies in developing countries, including MENA countries (e.g., Hossain et al., 2006;Islam & Deegan, 2010;Akrout & Othman, 2013;Ullah et al., 2014). A pilot study of Saudi Arabian companies was then conducted, which resulted in the addition of a few items, such as the influence of Islamic principles. This process resulted in a relatively long list of 55 environmental disclosure items which, given the 500 firm-years of observations, provides a total of 27,500 data points measuring the independent variable.
The second decision relates to the kind of index to be usedunweighted or weighted. The use of an unweighted index has become the norm in annual report studies because it avoids the subjectivity entailed in weighting individual items differently (Ahmed & Courtis, 1999). In this approach, an item scores one if it is disclosed and zero otherwise when a particular firm-year is analysed. However, a concern might be that certain categories of the disclosure are given undue weight because more items fall within them. The process for calculating the alternative, a weighted index, is first to calculate the individual sub-index scores and then to award them each an equal weight (20% in this case), thereby effectively adjusting the weighting of the individual disclosure items, depending on which category they fall into. Following some previous studies (Ntim et al., 2013, Elghuweel, Ntim, Opong, & Avison, 2017Ntim, 2016), this is used as a robustness check for the results based on the unweighted index.
To ensure the reliability of the content analysis, ten annual reports from the Tadawul stock market in Saudi Arabia were independently coded by two investigators. The Cohen's kappa coefficient of agreement was 0.79, which is at the high end of the satisfactory range of 0.7-0.8 (Beattie & Thomson, 2007;Krippendorff, 2004;Milne & Adler, 1999). Additionally, Cronbach's α was used to assess the reliability of the measurement of CED. The Cronbach's α for the sub-indices was 0.79, which again lies at the top of the 0.7-0.8 range considered to be satisfactory (Bland & Altman, 1997). Table 3 outlines how the variables (grouped into dependent, independent and control 9 variables) were operationally defined.

INSERT TABLE 3 HERE
To test the main and supplementary research hypotheses about the impact of CED on FV, we employ a set of panel data technologies: fixed-effects and two-stage least squares (2SLS), and generalised method of moment (GMM). The findings are presented in the next section. In a region of increasing, but still limited, CED, it is perhaps no surprise that the mean figure for EDI is not high and the minimum score for most of the sub-indices is zero. However, the possibility of high CED in GCC nations, even if not widespread, is shown by the high maximum scores. The variation in the independent variables relating to CED is confirmed by the material standard deviations, implying the potential role of CED in explaining variations in firms' market value. It is also notable that the mean scores for the five sub-indices show a significant degree of variation, which leads to a difference between the unweighted (EDI) and weighted (WEDI) versions of the overall disclosure index. Since the sub-indices with fewer items tend to score more highly, WEDI is Although growth in CED is apparent among the GCC countries, and although not all the items in the disclosure research instrument will be relevant to some companies, the levels of disclosure still appear to be low. Other studies might use different disclosure indices and so not be strictly comparable, but the phenomenon observed is consistent with findings of prior CED studies in developing economies and contrasts with studies in developed countries. For example, on the one hand, Shahab et al. (2020) reported   Table 4 showed some differences between the independent variables, EDI and WEDI, both forms of correlation between them are strongly positive, which suggests that the results from using the two versions of the CED variable are unlikely to differ. However, although positive and significant, the correlation between the dependent variables, TBQ and ROA, is less than 0.2.

Descriptive statistics
Therefore, the choice of the dependent variable is likely to matter, and the development of the separate market-based and accounting-based forms of the hypotheses is supported. Both TBQ and ROA are positively correlated with EDI at the 1% level of significance, but it is notable that the relationship is stronger in the case of TBQ. The correlations suggest that the hypotheses may have merit, but many of the control variables also show significant correlation with the dependent variables, so multivariate analysis is warranted.

Multivariate analysis
A fixed-effects model has been applied to undertake the primary regression analysis in our study.
Using the fixed-effects estimation addresses statistical concerns that might not be tackled employing an ordinary least squares (OLS) method. For example, it enables us to control for unobservable firmspecific heterogeneities across time that are expected to be constant, yet may have an influence on the relationship between the predictor and the outcome variable and might not be identified using an OLS method (Gujarati, 2003;Wooldridge, 2013). Consequently, we begin our regression analysis by estimating a fixed-effects model which is specified as follows: In this equation, FV is the measure of firm value (i.e., TBQ or ROA). The equation is written for unweighted CED (EDI), but it can also be written for the weighted form (WEDI). We control for firm-level factors: firm size (SIZE), leverage (LEV), industry (INDUS) and auditor type (BIG4); and further by gross domestic product (GDP) per person in the GCC country concerned. We also include country dummy variables.
The appropriateness of using a fixed-effects rather than a random-effects estimation was checked using the Hausman Test, which confirmed that the unobserved firm-specific variables were insignificantly related to those of the other corporations in the sample of our study. The results of our four models, which include firm-level characteristics and other control variables, are shown in Table   6. The four models represent the different combinations of the two dependent (TBQ and ROA) and the two independent variables (EDI and WEDI).

INSERT TABLE 6 HERE
All four models, whose adjusted R-squared vary between 0.30 and 0.52, show a significant positive relationship between CED and FV, consistent with both our hypotheses (H1, TBQ; H2, ROA) and robust to the form of disclosure index used (EDI or WEDI). However, there is a notable difference, depending on which proxy is used as the dependent variable; it was noted earlier that, while positive and significant, the correlation between TBQ and ROA was not high (see Table 5). Although there is evidence of a positive association between CED and ROA, it is only at the 5% level of significance in the case of WEDI and only marginal, at the 10% level of significance, in the case of our prime disclosure proxy, EDI. However, in the case of both versions of the environmental disclosure index, the positive relationship with our principal measure of FV (TBQ), is significant at the 1% level of significance. More importantly, the relationship is stronger for TBQ than for ROA. It will be recalled that TBQ was chosen as our principal measure because of its conceptual superiority. It can incorporate not only the information contained in ROA at time t, but also more besides, including the previous trend in ROA, together with anticipated performance and any shareholder valuation of environmental actions for their own sake as reflected in the share price.
Our results may be compared with the five single-country studies in Panel A2 of Table 1 In contrast, Nor et al. (2016) suggest an ambiguous effect of CED in Malaysia; specifically, whereas CED is positively related with ROE, it has a non-significant association with ROA. It might be surmised that this is because of the developing country context, where the institutional environment is likely to be less developed in some respects. However, it is notable that Nor et al. did not use a market-based proxy for firm value but only accounting measures, which we found, as theorised, are less likely to reflect value-relevant information. This interpretation is consistent with the results of another study in Malaysia, in which Iatridis (2013) indicates that high-quality CED is value relevant using a variant of the Ohlson (1995) valuation equation for the year-end stock price, which is a market-based approach that has similarities to using TBQ.
Drawing on neo-institutional theory to interpret our results, it appears that managers are adapting and developing their CED policies (a growth in disclosure is evident) and responding positively to the requirements of powerful influencers, with firm value being enhanced as a result. As explained earlier, even though CED is not mandatory in GCC countries, governments have been starting to exert coercive pressures on companies to adopt more environmentally-sensitive policies, and environmental pressure groups and NGOs have been increasingly exerting normative pressures. 11 In other words, managers in the GCC region appear to be positively interacting with a changing institutional environment. However, this is unlikely to be a simple matter; pro-active and sustained corporate participation in environmental initiatives is likely to be required for companies to increase their legitimacy and enhance their firm value, perhaps pragmatically by gaining valuable resources, such as low-cost capital (Ntim, 2016;Suchman, 1995).
Although not the central focus of the study, some of the control variables for firm-specific characteristics also have statistically significant associations with FVnotably firm size (SIZE) and, at least for the models involving TBQ, leverage (LEV). However, the type of auditor (BIG4) does not affect the relationship between CED and FV. Dummy variables were also included for the five individual countries. Only one of the 20 coefficients was marginally significant (Kuwait, Model 1, at 10%), which underscores the legitimacy of studying the five countries together on this occasion. Thus, although the level of CED by country differs, the relationship with FV does not do so significantly.
Using the five categories of environmental disclosure contained in our 55-item disclosure index, we probe further into the impact of CED on FV. The results of estimating fixed-effects models for the five sub-indices of the EDI, using TBQ and ROA, respectively, are presented in Models 1 to 10 in Table 7.

INSERT TABLE 7 HERE
The models in Table 7 confirm the pattern seen previously, with the relationship between CED and TBQ much stronger than the relationship with ROA. Indeed, only one sub-index is significantly related to ROA, and then only marginally, namely 'other' environmental disclosure. It is also striking that all the sub-indices are statistically significantly related to TBQ, usually (except for 'other') at the 1% level. Thus, all the categories in our disclosure instrument seem to play a role in explaining FV as proxied by our primary measure, TBQ. Since the results are not being driven by a particular element of CED, the models in effect provide a robustness check on the value relevance of environmental disclosure in general. We outline further robustness checks next.

Extra robustness checks
Arellano and Bond (1991) argue that panel data techniques may not be reliably estimated by the use of a fixed-effects estimator only, since the regressor is, by nature, not determinedly exogenous. Using our primary proxy for CED, EDI, the current research, therefore, employs both 2SLS and 2-step GMM estimators as robustness checks to make sure that the primary results of estimating a fixedeffects model are not severely influenced by the possible occurrence of endogeneity problems (Blundell and Bond, 1998). We use the Durbin-Wu-Hausman test to detect the potential occurrence of endogeneity of individual regressors. From a theoretical perspective, the explanatory variable should not be correlated with the error term (residuals), and the Durbin-Wu-Hausman test determines whether the residuals are correlated with the explanatory variable (Ullah, Akhtar & Zaefarian, 2018).
The result of conducting a Durbin-Wu-Hausman test indicates that the CED variable is endogenous rather than exogenous, and thus our results presented in Table 6 might be biased. Overall, the findings of the Durbin-Wu-Hausman test suggest that endogeneity is a major concern in our regression model.
Consequently, we believe that the use of both 2SLS and dynamic GMM regression models is appropriate to address the endogeneity concerns.
The operational definitions for all variables are as presented in Table 3. FVit-1 indicates one year lag of our dependent variable FV (previous year's FV) as proxied by TBQ, and FVit-2 denotes a second lag of the dependent variable, which represents FV two years previously. These lagged variables are considered as explanatory variables in our two-step GMM estimation. By including lags of FV (the dependent variable in our study), the dynamic GMM method controls for endogeneity by internally transforming the data where a variable's past value is subtracted from its present value' (Roodman, 2009, p.86). In doing so, the number of observations is decreased, and the internal transformation process improves the efficiency of the GMM estimation (Wooldridge, 2016). Furthermore, to avoid potential data loss due to the internal transformation, Arellano and Bover (1995) recommend the use of the two-step GMM model. Thus, Roodman (2009) states that, by using a two-step GMM model, researchers can prevent unnecessary data loss and provide more consistent and efficient estimates for the included coefficients.
The Sargan test and the Arellano-Bond are post-estimation tests and have been used in our study to determine whether the dynamic GMM model is valid or not and whether the instruments (lags of FV) are correctly specified or not (see Table 8). A crucial assumption for the validity of the dynamic GMM estimates is that instruments (the lagged dependent variables) are exogenous (see Ullah et al., 2018). If the results of these pre-estimation tests turn out to be insignificant, it means that the included instruments in the GMM specifications are exogenous; thus, the instruments we use in this study are valid. Overall, a two-step dynamic GMM model is believed to be an ideal method to overcome any endogeneity issues in our research.  Table 8 show the results of estimating the GMM models. Again, the positive relationship between CED and TBQ is significant at the 1% level, whereas there is not a statistically significant relationship with ROA. Thus, we continue to find strong support for the relationship between CED and FV as proxied by TBQ, but we cannot confirm the finding of a positive relationshipwhich was relatively weak anywaybetween CED and ROA.

Discussion and conclusion
To the best of our knowledge, this is the first multi-country study of the effect of corporate environmental disclosure (CED) upon firm value (FV). It finds that CED is significantly and positively related to FV as measured by Tobin's Q (TBQ) in the Gulf Cooperation Council (GCC), an economically and environmentally important set of countries. Drawing on neo-institutional theory, our findings suggest that, even though all CED in the region is voluntary (i.e. not directly subject to coercive isomorphism), the broader changes that are taking place in terms of government environmental activities and NGO initiatives are probably providing a degree of normative influence that not only encourages increased disclosurea process that is likely to be reinforced by mimetic isomorphismbut also helps to build an environment in which such disclosure enhances corporate reputation and legitimacy amongst stakeholders, thus increasing the market value of companies.
Our empirical evidence is broadly in line with the results of some prior studies (e.g., Broadstock et al., 2018;Clarkson et al., 2013;Plumlee et al., 2015;Iatridis, 2013), where the disclosure of environmental information is positively and significantly associated with market-based outcome proxies. Our findings are robust to various statistical tests, and the relationship applies across both the individual GCC countries and all the component disclosure sub-indices, which themselves provide a level of detail absent from most similar studies. Some evidence of a positive relationship between CED and return on assets (ROA) is also found, but even where statistically significant, the relationship is much weaker than in the case of TBQ.
In focusing upon the GCC, the study is a relatively rare example that examines the relationship between CED and FV in the context of the developing world. Further studies might examine how CED comes to be reflected in TBQ. However, it is worth noting that, for this study, none of the relevant CED was mandatory, although there are increasing signs of normative pressures (DiMaggio & Powell, 1983) in the region as CED continues to grow (Eljayash et al., 2012;Gerged et al., 2018).
In terms of future research, given the speed and enormity of climate change and given that our analysis only goes as far as 2014, it would be worth repeating the study with more recent data at some point.
Although relationships between variables are more likely to be stable over time than the levels of the variables themselves (Bell, Bryman, & Harley, 2018;Cowton, 2019), replication of the current study could determine whether the apparent increased interest of investors and other stakeholders in environmental issues accentuates the positive relationship that we have discovered between environmental disclosure and firm value in GCC countries.
Vastly increased concern about the climate also means that environmental issues are now, if they were not before, too important to be subsumed within studies of CSD (corporate social disclosure) or A further possible development regarding the research instrument lies in how it is used. Although binary coding proved to be effective for the purposes of the current study, it may be regarded as a limitation, and a future study of the region (or elswhere using a similar research instrument) might employ an ordinal coding method that recognises a distintion between qualitative, quantitative and financial forms of environmental disclosure.
In terms of the proxy used to capture the effects of CED, we suggest that researchers should in future focus upon market-based measures such as Tobin's Q (TBQ), assuming there is a reasonably welldeveloped stock market present. TBQ is conceptually superior to a single-period accounting-focused measure such as ROA (used here as a supplement), ROE or EPS (as used in some previous studiessee Table 1) since it can in principle capture any information contained in the accounting measure as well as any contained in past trends, together with other information about future expectations and any valuation by shareholders of environmental actions for their own sake. This theoretical superiority is borne out by our empirical findings, with TBQ featuring more strongly than ROA at all stages of the analysis. Perhaps the results of the literature will appear to be less 'mixed' (cf. Nor et al. (2016), who relied on accounting-based measures) if, in future studies, outcome variables based on accounting profitability are omitted or reduced to subsidiary status. At the very least, our results and argument strongly suggest that greater thought should be given to the choice of outcome proxy in future studies.
Further research building on this study could examine other countries, or collections of countries, using the methodological as well as substantive insights that we have presented. It might also be useful to examine financial institutions, which have a less direct impact on the environment but still have a significant role to play (Thompson & Cowton, 2004;cf. Platonova et al., 2018).
As for implications for policy and practice, the current results suggest that managers can take a positive view of opportunities to expand CED and that policymakers considering the introduction or extension of mandatory CED should consider not only that managers might have less to fear than some might think but also that such disclosure appears to be value-enhancing. Given our findings, resistance to change in GCC countries should not be predicated on a belief or claim that environmental disclosure is irrelevant to users.
Clarkson, P. M., Fang, X., Li, Y., Richardson, G. (2013). The relevance of environmental disclosures: Are such disclosures incrementally informative? Journal of Accounting and Public Policy, 32(5), 410-431. Cormier, D., Magnan, M. (2013). The economic relevance of environmental disclosure and its impact on corporate legitimacy: an empirical investigation. Business Strategy and the Environment, 24 (6)  1 The Industrial group of sectors includes: oil and gas; glass and ceramics; textiles; pharmaceutical and medical; leather and clothing; tobacco and cigarettes; chemical; paper and cardboard; printing and packaging; food and beverages; mining and extraction; engineering and construction; and electrical. 2 The Services group of sectors includes: hotels and tourism; healthcare; educational; transportation; media; utilities; real estate and resorts; and technology and communications. 3 This line combines the Industrial and Services groups to give the total sample for a country/GCC. 4 Log transformation of $US total assets figures. SUB-EDI5 Environmental 'other' sub-index comprising 11 items.

Panel C: Control variables (firm-level and country-level)
SIZE Firm size as measured by the natural log of total assets. LEV Leverage, as measured by the ratio of debt to total assets.

INDUS
Type of sector, measured by dummy variable based on the Industry Classification Benchmark (ICB). BIG4 Type of auditor, measured by dummy variable, equals 1 if a firm is audited by a Big4 auditing firm, 0 otherwise. GDP The natural log of gross domestic product per capita, measured in GBP.  Table 3.  Table 3.  100 This table presents the findings of estimating ten fixed-effects models based on the five sub-indices for all firm-years. Standard errors in parentheses. *Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level. Variables are defined in Table 3. The robustness tests are a two-stage least squares (2SLS) model and a two-step GMM estimation. *Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level. Standard errors in parentheses. Variables are defined in Table 3. In conducting a 2 SLS regression model, the industry dummy variable has been employed as an Instrumental Variable (see Abdelfattah & Aboud, 2020).