Assurance on sustainability information and environmental performance:
an empirical analysis.
Master thesis, combined Msc Accountancy and Controlling University of Groningen, Faculty of Economics and Business (FEB)
June 22, 2020 Jaap Winia Student number 3749789 Kwelderstraat 23 8931AW Leeuwarden tel.: 06 57155615 e-mail: j.d.winia@student.rug.nl Supervisor: dr. T.A. Marra Word count: 8362
Acknowledgement: A word of thanks to the other three master students for the pleasant cooperation during the period in which the thesis was written. A special word of thanks goes out to my supervisor, dr. T.A. Marra for the feedback, guidance and advice during the process of writing this thesis, despite
Abstract
Given the fact that assurance on non-financial disclosure is a relatively unexplored area in research, this thesis takes a look into the possible internal benefits of assurance, investigating the relationship between assurance on sustainability disclosure and corporate environmental performance. Using a predefined coding scheme 2,149 assurance reports were scored by an assemblage of 28 bachelor and master students. Further data regarding companies’ environmental performance was collected using Thomson Reuters Eikontm database. To test the hypotheses a mixture of different regressions analyses were performed on the accumulated panel data. The results suggest that companies with higher scored assurance on their sustainability disclosure are more likely to have lower environmental performance. This effect is further increased if the company is shareholder oriented. These unexpected findings suggest that further research is necessary into the contemplated relation between assurance and environmental performance.
1. Introduction
September 20th 2019 marked the start of a week of global climate strikes which over six million people participated in, a lot of whom were students. Topics which were brought to the attention of the public include rising sea levels, air pollution and plastic waste. These topics were supporting the overall message: urgent action is necessary to stabilise climate change (Laville and Watts, 2019).
Madrid, December 13th 2019, the United Nations Climate Change Conference (UNFCCC COP 25) conclude their talk on the climate change process and the operationalization of the Paris Agreement (United Nations, 2015) and to prepare for 2020, the year in which countries have to update their climate action plans. Focus areas include technology, finance and the transparency of climate action.
Besides countries, companies were also represented at COP 25. Representatives for a group of 177 companies, with a combined global market capitalisation of 2.8 trillion USD and CO2 emissions equivalent to France, agreed to adhere to the climate targets set in the Paris Agreement and set out to reach emission neutrality no later than 2050 (United Nations Global Compact, 2019).
Stakeholders, like investors, customers, employees and governmental bodies, can and have caused companies to adapt sustainable goals and take action to achieve these goals. An example of stakeholders’ power to enforce companies’ adherence to certain goals is the case of plastic straws. Because oceans are polluted with copious amounts of plastics, multiple parties agree that plastic straw should be banned by companies (Gibbens, 2019). Companies have changed their operation due to subjugation to stakeholders expectations and requirements. However, a recent situation at McDonalds resulted in a great deal of negative media attention when their new paper straws are even harder to recycle, ergo creating more
waste than their previous plastic straws (Hanburn, 2019). Such negative media attention can have a substantial negative, and sometimes business destroying, impact on companies.
Another example of companies’ actions related to stakeholder perception is the Swedish energy company Vattenfall. Vattenfall portrayed itself as a climate champion winning the climate greenwash award in 2009 (Corporateeurope.org, 2019), while four of its power plants are listed in the top 30 most polluting power plants in Europe (CAN Europe, WWF, HEAL, EEB and Klima Allianz Germany, 2014). This led to mass criticism from stakeholders. Vattenfall recently started their media campaign “fossil-free living possible within one generation” (Vattenfall, 2019), showing that even one of the most environmental unfriendly industries, as measured by combination of greenhouse gasses emitted (Ritchie and Roser, 2017), can improve their organisation in a way that reduces the overall environmental footprint substantially (Vattenfall, 2018).
Because stakeholders pressure companies to increase sustainability they want to receive information regarding companies’ goals, actions and results in this area; in other words, to receive evidence of the company’s legitimacy (Deegan, 2002). To show this legitimacy regarding environmental activity towards stakeholders, companies have started using sustainability reports in addition to their annual report. Stakeholders, e.g. investors, can view what actions companies undertake to achieve certain goals. Sustainability reporting is a form of non-financial voluntary disclosure which has become somewhat of a standard when it comes to themes regarding the environment (Kolk, 2004). Companies use this form of disclosure to achieve both internal and external benefits (Globalreporting, 2019).
Simnett et al. (2009) argue that assurance on such voluntary statements increases the credibility of information conveyed. Furthermore, sustainability disclosure is deemed to be of higher quality when assured (Moroney et al., 2011). Both of these reasons along with the cost aspect make assurance a beneficial aspect to the decision making process of stakeholders. With the increasing interest from stakeholders in sustainability reporting, the need for assurance becomes more apparent too. Again, stakeholders want to be able to trust on information regarding the companies sustainability actions, assurance can provide means of increase in reliability (Kolk, 2007).
However, not only investors are interested in the assurance on sustainability disclosure. Companies themselves can use sustainability reporting to achieve internal benefits (Globalreporting, 2019), i.e. business strategy development (Zorio et al., 2012). On top of this, O’Dwyer et al. (2011) mention that the inclusion of external assurance on sustainability disclosure can lead to internal learning processes and cultural change. The findings of Park and Brorson (2005) support this statement that companies seek assurance to improve both internal reporting systems and actual performance.
This aforementioned motivation for companies to use the assurance on sustainability disclosure to achieve internal benefits is a relatively unexplored area in current literature and research. Currently assurance on non-financial sustainability disclosure is a relatively new practice and therefore not very
developed (Junior et al., 2013). This voluntary disclosure of non-financial information is not uniform and subjugated to a variety of different standards. As a result, reports diverge heavily from another. This makes it hard to provide a uniform assurance statement resulting in variation in quality of assurance. Current studies in the subject of assurance on sustainability disclosure (or voluntary disclosure in general) mostly direct the attention towards external benefits, like increasing legitimacy and credibility of the report to please stakeholders, e.g. investors and reduce information asymmetries (O’Dwyer et al., 2011; Kolk, 2007). However, besides research into the external benefits of assurance on sustainability disclosure, there appears to be a lack of research in exploring the internal benefits of assurance. These internal benefits could, as acknowledged before by O’Dwyer et al. (2011) and the Global reporting initiative (2019), be presented in a wide variety of areas within companies. A further link between internal benefits and corporate environmental performance will be explained in the theoretical framework and hypothesis development section of this paper.
In this research I will investigate the relationship between assurance on sustainability reporting and corporate environmental performance. Hence, my research question is:
To what extent does assurance on sustainability reporting affect corporate environmental performance?
The remainder of this paper is structured as follows. In the theoretical framework and hypothesis development section all relevant theory applicable to thesis study will be described, based on this the hypothesis are developed. After this the research methodology is explained. Section 4 presents the regression results. The final part of this research paper will consist conclusion and interpretation of the results, closing with this research’s limitations and possible avenues for future research.
2. Theoretical framework and Hypothesis development
As mentioned in the introduction, the goal of this research is to answer the research question if and how assurance on sustainability reporting affects corporate environmental performance. This chapter connects the research question to existing literature.
2.1. Theory behind assurance
First, as previously mentioned, this paper will look at two justifications which firms use in assuring their sustainability disclosure.
The first one being the external benefits (Globalreporting, 2019). A wide range of studies already showed interest in the external benefits (Zorio et al., 2013; O’Dwyer et al., 2011; Junior et al., 2014;
(Deegan, 2002) and how the firm is held accountable towards different stakeholders (Guthrie et al., 2006) respectively. An et al. (2011) highlight the similarities across these theories with their integrated framework. Both legitimacy and accountability are key concepts within these theories.
The second reason for firms to start providing assurance with their sustainability reporting can be internal benefits (Globalreporting, 2019). Edgley et al. (2010) interviewed multiple accountant assurors to find that assurance has benefits/adds value for management, specifically in the form of improvements of the internal management systems giving management an overview of where they are standing and pinpointing deficiencies. In addition to the assurors point of view, the KPMG survey (2013) contains the results of interviews with several high ranking individuals within companies that score most highly on corporate responsibility reporting (according to their quality criteria). Some of these practitioners mention that assurance helps them improve the clarity and credibility of corporate responsibility reporting. While others observe that assurance has helped them focus on materiality and improve the strength of internal systems. An interview based research paper from Park and Brorson (2005) shows several reasons why companies would seek third-party assurance on their sustainability reports. While most companies use external credibility as main motivation, some pioneers stated that they sought out assurance to get credibility compared to what the firm has accomplished. This allows the companies to position themselves as leader on the subject of environmental management. Several companies pointed out that they invested in assurance to improve both their reporting system and actual performance. Jones and Solomon (2010) add to this by identifying that external assurance can be used to improve quality of reporting and identify possible gaps regarding current activities.
Gürtürk and Hahn (2016) theorize that potential internal benefits for assurance seeking companies heavily depends on the type of assuror. The choice for a “Big Four” accounting firm as assuror might have negative impact on the learning abilities of the company, this because of the uniformity and standardized approach when it comes to assurance statement. These accounting approaches usually do not take any company specific situations into account. Therefore, an assurance engagement executed by a sustainability expert might have a more positive impact on the learning aspect of the assurance engagement.
Companies can make use of external assurance to enhance decision making and bring forth organisational change (Gürtürk and Hahn, 2016). Furthermore, research from Adams and Frost (2008)
show that sustainability related factors have a higher likelihood of being included in the decision-making process of managers when they have access to larger sets of sustainability data. The reliability of the data is paramount in the previous drawn example (Gürtürk and Hahn, 2016), assurance can enhance this by means of improvements to the current internal information system (Edgley et al., 2010).
The possible internally oriented benefits of assurance on sustainability reporting has mechanisms that are inclined towards behavioural decision theory (Cohen and Simnett, 2015). This theory entails how users (in this case the firms’ managers) interpret and use the assurance to obtain firm benefits.
The above can be linked to the evidence O’Dwyer et al. provide in their 2011 case study about determinants of corporate environmental reporting. They find that assurance on sustainability reporting is positively related to elements of learning and evolutionary improvements in firms’ organizational management systems and reporting in general.
External assurance enables companies to anticipate approaching material social and environmental issues (Adams and Frost, 2008; O’Dwyer, 2011), this can improve the company’s capabilities of attending these concerns. Adams and McNicholas (2007) alleges that sustainability related performance (e.g. social-, environmental-, or governance performance) can improve when a sustainability report is published, because this might result in better integration of sustainability specific issues into planning and decision making. However the effect of assurance on this relation is not explored in Adams and McNicholas’ (2007) research paper. To summarize, based on aforementioned papers and rationalization, assurance on sustainability disclosure can be connected to sustainability related performance, whether this is in the form of improved management systems, reporting systems or actual performance regarding a subject categorized under sustainability (e.g. social performance or environmental performance).
Quality of assurance on sustainability reporting is a highly subjective matter. One aspect that can have an impact on quality is whether the assuror is provided by an auditor or a sustainability consultant/expert. The auditors’ profession generally provides high quality assurance, due to high ethics and international applicable standards (Pflugrath et al., 2011). However, there is no conclusive evidence regarding the level of quality when looking at different types of assurance providers. Zorio et al. (2013) find that quality is significantly higher when assurance is provided by an auditor (instead of a sustainability consultant/expert) while others claim that (sustainability) specialists provide a more complete and fair statement (Hodge et al., 2009).
This paper applies content analysis, using a coding scheme of Gürtürk and Hahn (2016), which will be explained in further detail in the methodology section (3.2.2. Independent variables). Quality of assurance has been used as a measurement in previous academic research (Jones and Solomon, 2010; Zorio et al., 2013; Cohen and Simnett, 2015). The coding scheme of Gürtürk and Hahn (2016) can be used to indicate the quality of a certain assurance report. The coding scheme as depicted in Gürtürk and Hahn (2016) is an adaptation from Perego and Kolk (2012), which in turn finds its origins within a paper of O’Dwyer and Owen (2005). In O’Dwyer and Owen’s research (2005) the content of the report is compared to recommended minimum contents of assurance statements according to three sets of guidelines (AA1000 1999, FEE 2002 and GRI 2002). The scores following are a mere representation of
describe what is being measured by this scheme1. Two assurance statement can obtain the same score but to the reader be of different levels of quality, for instance when different criteria in the coding scheme of Gürtürk and Hahn (2016) are directly compared to each other. An example of this is when scoring criteria number 5. Report date is compared with number 14. Assurance standard used (Appendix A). Despite having the same scoring scale (0 or 1) information contained within this score is different. The coding scheme simply indicates whether different aspects are present or not and summarizes this into a number which can be seen as a quantification of the extensiveness of the assurance statement.
The inclusion or exclusion of certain elements can indicate whether the assurance report is aimed towards obtaining internal benefits, for instance when there is no addressee mentioned, i.e. management of the company or a specific stakeholder. This can indicate that the main purpose of the assurance is fully aimed to benefit the management of the company, e.g. internal oriented benefits of assurance
(O’Dwyer and Owen, 2005). O’Dwyer and Owen continue in their paper (2005) by constantly referring to this analysis of assurance statements as the extensiveness of the content, more specific the extent of the assurance statements’ content addresses key elements of guidelines (AA1000, FEE and GRI). Following the rationale described above, I will acclimate to the description of extensiveness of assurance on sustainability disclosure as introduced by O’Dwyer and Owen (2005), instead of the quality of assurance used by other authors like Gürtürk and Hahn (2016) and Perego and Kolk (2012).
Extensiveness of assurance also constitutes a better fit in describing the effect assurance can have on internal and external stakeholders (Gürtürk and Hahn, 2016; Zorio et al., 2013). Other research is looking at the effect of assurance on the sustainability information as a whole, e.g. whether sustainability information is assured or not (O’Dwyer et al., 2011; Reimsbach et al., 2018, Braam et al., 2016), but not at the content of the assurance itself. The Gürtürk and Hahn coding scheme (2016) can be used to relate the extensiveness of assurance itself to possible internal effects.
Following the arguments presented above, the first hypothesis is as follows:
Hypothesis 1: Extensiveness of assurance on sustainability reporting has a positive effect on corporate environmental performance.
2.2. Stakeholder orientation
As previously mentioned stakeholders can have a substantial degree of influence over a company
(Deegan, 2002). External benefits of assurance on sustainability information could consist of the ability to enhance the relation with these stakeholders (Braam et al., 2016; Jones and Solomon, 2010), increase legitimacy (O’Dwyer et al., 2011) and credibility (Pflugrath et al., 2011). The origin of these external
1 The Oxford Dictionary describes ‘Quality’ as “a standard of something when it is compared to other things
alike” (Oxford Dictionary, n.d.), therefore the coding scheme is unable to give a description of quality when merely distributing points based on the content of assurance statements.
benefits can be traced back to what is most likely the legitimacy- or stakeholder theory (An et al., 2011). This in its part is closely related to stakeholder orientation, in other words whether a company is more predisposed towards stakeholders or shareholders (Simnett et al., 2009). The stakeholder oriented firms are most likely to perceive pressure from third parties as described above, which is one of the main drivers of their sustainability reporting and respectively assurance thereon (KPMG, 2013; Junior et al., 2014). For this reason, assurance is more likely to be beneficial for the relation between the company and its stakeholders, e.g. external benefits.
On the contrary, companies who are situated on the other end of the spectrum e.g. shareholder oriented are less likely to perceive the same pressure from stakeholders. Following this logic, the rationale for companies that can be identified as predominantly shareholder oriented to get their sustainability disclosure assured more likely has its origin in internal benefits. In other words, a feasible reason for shareholder oriented companies to get assurance can be to improve their own internal systems, performance, reporting (O’Dwyer et al., 2011; Park and Brorson, 2005; Jones and Solomon, 2010;
Edgley et al., 2010).
Based on the reasoning above, I argue that companies that are more oriented towards their shareholders might have larger benefits from the extensiveness of assurance statements on their corporate environmental performance than companies that are stakeholder oriented. I, therefore, argue that the relation contemplated in hypothesis 1 is stronger for companies that are more oriented towards shareholders. This moderating effect is illustrated in the conceptual model in Figure 1.
Hypothesis 2: The effect of extensive assurance on corporate environmental performance will be stronger (weaker) when the firm is increasingly oriented towards shareholders (stakeholders).
+ + + Extensiveness of assurance on sustainability reporting Corporate environmental performance Industry sensitivity Figure 1 Conceptual model Stakeholder orientation
2.3. Industry sensitivity
Previous studies show that differences in reporting and assurance between industries exist (KPMG, 2013; Garcia et al., 2017; Patten, 2002). KPMG (2013) reports that industries more subjective to environmental and social risk (Simnett et al., 2009) show overall better reporting quality. Furthermore, a more recent study by Garcia et al. (2017) shows that ESG (environmental, social and governance) performance in sensitive industries is superior in comparison to less sensitive industries (empirical evidence from BRICS emerging markets). Sensitive industries are defined as industries more susceptible to social taboos, moral questioning and political pressure (Garcia et al., 2017; Richardson and Welker, 2001). Simnett et al. (2009) argue that firms who are situated in industries of extraction and/or depletion of natural resources, i.e. industries with a large environmental footprint can be categorized as sensitive industries. The increased performance in addition to subsequent disclosure and assurance on sustainability from companies within these sensitive industries can be the result of public/community concerns (e.g. pressure from external stakeholders) regarding the environment (Simnett et al., 2009). The company is responsible for the effect it has on, for instance, the environment. Companies might make use of assurance to further increase their environmental performance and enhance their reputation and legitimacy.
Based on this argumentation, I argue that companies operating in sensitive industries might have larger benefits from extensiveness of assurance statements on their corporate environmental performance than companies that are operating in insensitive industries. In short, the relation proposed in hypothesis 1 is stronger for companies operating in sensitive industries. This moderating effect of industry sensitivity is illustrated in the conceptual model in Figure 1.
Hypothesis 3: The effect of extensive assurance on corporate environmental performance will be stronger (weaker) when the firm is operating in a(n) (in)sensitive industry.
3. Methods 3.1. Sample
In order to test the aforementioned hypotheses, a sample of 2,149 observations was compiled over a period from 2013 to 2019 for 843 listed companies worldwide. Data was collected from companies’ sustainability disclosure, whether this was in form of an autonomous sustainability report or an annual report with integrated sustainability information. The main source of these reports was the Global Reporting Initiative Sustainability Disclosure database (https://database.globalreporting.org/). Companies who uploaded a report in the period from 2015 to 2017 are expected to do so in any subsequent years. Where those reports were not uploaded to the GRI Sustainability Disclosure database their availability was investigated at the respective corporate website.
The data has been collected by 28 students including myself who are either engaged in their bachelor or master thesis at the RUG. The first selection of reports were collected by the 24 bachelor students. This included all sustainability reports directly downloaded from the GRI Sustainability Disclosure database, these reports span a period of 2013 to 2018. The four master students (including myself) collected reports for subsequent reporting years in relation to the latest reports present in the dataset comprised of the reports collected by the bachelor students. Most of these reports were not uploaded to the GRI Sustainability Disclosure database, thus we had to individually gather the reports on companies’ websites. Further explanation as to what information was collected from these reports and how this two-step approach impacts the data integrity is given under 3.2.2. ‘Independent variables’.
For the companies’ financial and environmental information the Thomson Reuters Eikontm database (https://eikon.thomsonreuters.com/index.html) was used and the reports were matched against this data. The data consists of a diverse arrangement of industries and countries. These aspects are of importance in answering the second and third hypotheses regarding stakeholder orientation and industry sensitivity respectively. A further description is given under 3.2.2 ‘Independent variables’.
The result is an unbalanced panel dataset containing a total of 2,149 observations and an average of 2.1 year observations per company with a minimum of 1 and a maximum of 5 years.
Orientation Total Year
2013 2014 2015 2016 2017 2018 2019
Stakeholder (code law) 1,507 2 327 369 405 344 58 2
Shareholder (common law) 642 1 111 160 164 148 57 1
Total 2,149 3 438 529 569 492 115 3
Panel B: Industry characteristics across the years of observations
Industry a Total Year
2013 2014 2015 2016 2017 2018 2019
Chemicals 129 0 27 32 38 27 5 0
Coal 17 0 3 4 5 5 0 0
Metals & Mining 161 0 33 45 40 29 14 0
Oil & Gas 120 0 31 26 30 28 5 0
O & G Related Equipment and Services 18 0 4 4 6 3 0 1
Paper & Forest Products 18 0 5 4 4 4 1 0
Uranium 1 0 0 1 0 0 0 0
Other 1,685 3 335 413 446 396 90 2
Total 2,149 3 438 529 569 492 115 3
Panel A: Country characteristics across the years of observations Table 1
Summary statistics on sample (2013-2019)
a
Descriptive statistics on the sample companies is presented in Table 1. Panel A shows the descriptive statistics by the company’s orientation depending on the legal system in the country of origin (3.2.2.). Panel A demonstrates that most of the observations are recorded between 2014 and 2017. Panel B is a more in-depth view into the industry in which the companies operate, with the industries considered to be sensitive in nature (3.2.2.). Again this table shows that the period from 2014 to 2017 contain the most observations.
3.2. Variables
3.2.1. Dependent variable: Corporate environmental performance
In line with previous CSR related studies, I measure corporate environmental performance (CEP) by ESG factor, obtained from the Thomson Reuters Asset4 Database (Garcia et al., 2017; Sassen et al., 2016; Cheng et al., 2014). This database is one of the most comprehensive when it comes to ESG information, using more than 450 different ESG metrics divided between 3 main pillars; environmental, social and governance (Refinitiv, 2020). The Database covers more than 9,000 companies globally. The data is collected by research analysts who are trained to collect the data from companies’ sustainability disclosure. This occurs on average in line with companies’ disclosure pattern, thus once a year. The ESG score is a weighted sum of the pillars following the magnitude matrix. The three pillars are given a percentile rank score benchmarked against Thomson Reuters Business Classification2 (TRBC) for all environmental and social categories and against the county where the company is situated for governance categories.
The environmental pillar score measures the company’s performance on environmental subjects: resource use, i.e. water, fossil fuels and energy; innovation and emissions, i.e. waste, carbon dioxide and emission policies. I primarily focus on this particular pillar since the research is directed towards the relation between extensiveness of assurance statements on sustainability disclosure and corporate environmental performance. Companies in the sample receive an environmental score, following the Thomson Reuters environmental pillar, from 0.00 to 1.00 for each firm year. The dependent variable is abbreviated to CEP, see Table 2
3.2.2. Independent variables
To assess the extensiveness of assurance on sustainability disclosure, the 28 students (explained in 3.1. ‘Sample’) scored the assurance reports following the Gürtürk and Hahn coding scheme (2016), which is as described in the theoretical framework an adaptation of an earlier set of coding rules from Perego and Kolk (2012). The coding scheme consist of 23 separate criteria that can be scored from 0 up to 3
2 TRBC is the global, comprehensive, industry classification system owned and operated by Thomson Reuters. The classification system is broken down into 5 levels; activities, industries, industry group, business sector and economic sector.
with a potential maximus score of 33 (see Appendix A). For all of the 2,149 observations the assurance statement, within the respective sustainability reporting of the company, was analysed following the coding scheme.
The coding scheme, shown in Appendix A, includes some points which are not that apparent to score in the first place. For instance scoring criteria number 1. Title versus number 17. Completeness. The obvious existence of a Title is more straight forward to score as opposed to the Completeness of the assurance statement. This makes some points of the coding scheme more subjective to personal interpretation then others. The paper of Gürtürk and Hahn (2016) was provided as extra guidance to limit the room for interpretation. Furthermore, referring to the two-step procedure of data gathering as described in 3.1. ‘Sample’, the scoring process of the 4 master students (which is separate from the bachelor scoring process) was preceded by a pre-scoring of 8 reports to increase reliability. These 8 reports were selected form various assurance providers to achieve a varied sub-sample. The assurance statements were scored by each student individually. Afterwards the differences were discussed and consensus regarding interpretation and scoring was achieved. Continuous discussion among group members helped to enhance internal validity. This independent variable’s abbreviation is labelled EXT, Table 2.
Further, the additional variables use to answer the second and third hypothesis are moderating variables on the main relation. First, in order to measure the stakeholder orientation I follow previous research methods where stakeholder orientation is measured by the legal system associated with the country where the company is established (Simnett et al., 2009). Companies in common law countries are believed to be more shareholder oriented. On the antithesis, companies in code law countries supposedly are more stakeholder oriented (Ball et al., 2000). Following this methodology, to measure the stakeholder orientation of the company, I use a dummy variable for code law versus common law country as a proxy for stakeholder orientation. This dummy variable is labelled STAKEHOLDER as summarized in Table 2.
Second, I make use of the TRBC codes to define the sensitive industries. These are already determined in previous research (Richardson and Welker, 2001; Deegan and Gordon, 1996; Simnett et al., 2009; Garcia et al., 2017). These industries, often classified as environmental- or social sensitive, include: mining, oil and gas, paper, chemicals, uranium and metals. This list with industries was cross checked against a NAICS code document of environmentally sensitive industries (Office of Management and Budget, 2017) to ensure that the industries mentioned before are also present in this, predominantly, North American classification system. Sensitivity of industries will be used as a dummy variable, INDUSTRY in Table 2.
3.2.3. Control variables
Following previous research (Braam et al., 2016; Zorio et al., 2013; Garcia et al., 2017; Simnett et al., 2009; Barnett and Salomon, 2012; Surroca et al., 2010; Mishra and Modi, 2013), I adopt the following control variables: Size of the company, profitability of the company, leverage of the company and the cashflow of the company. Most studies investigating voluntary disclosure include company size as a control variable (Clarkson et al., 2011), larger companies depicted to benefit from economies of scale. Size of the company will be measured following the approach of Braam et al. (2016) and Zorio et al. (2013), by the logarithm of the company’s total assets. The ROA (return on assets) will be used as the
measure for the company’s profitability, as it is a commonly accepted proxy for profitability. Companies with higher profitability have increased flexibility to engage in sustainability in general. Leverage of the company is measured by total (non-current) liabilities divided by total assets. Studies indicate that leverage possibly influences management decision making (Barnett and Salomon, 2012). Lastly,
Table 2
Variables definition and theoretical foundation
Variable Definition Theoretical foundation
CEP Corporate environmental performance obtained from Thomson Reuters asset4. The environmental pillar score measures the company’s performance on resource use, innovation and emissions.
(Thomson Reuters Asset4)
Garcia et al., 2017; Sassen et al., 2016; Cheng et al., 2014.
EXT Extensiveness of assurance is being measured by application of the Gürtürk and Hahn coding scheme (2016). The assurance statements are scored on 23 different points, from 0 to 33.
Gürtürk and Hahn, 2016; O’Dwyer and Owen, 2005.
STAKEHOLDER Dummy variable, 1 if the company is located in a common law country (shareholder oriented), 0 if company is located in a code law country (stakeholder oriented). (Thomson Reuters Datastream)
Simnett et al., 2009; Ball et al., 2000.
INDUSTRY Dummy variable, 1 if industry is considered sensitive, 0 if otherwise.
(Thomson Reuters Eikon)
Richardson and Welker, 2001; Deegan and Gordon, 1996;
Simnett et al., 2009; Garcia et al., 2017.
SIZE Size is measured by the logarithm of the company’s total assets.
(Thomas Reuters Worldscope Fundamentals)
Braam et al., 2016; Zorio et al., 2013.
ROA Return on assets is used to measure profitability and financial performance. (Thomas Reuters Worldscope Fundamentals)
Garcia et al., 2017; Zorio et al., 2013; Braam et al., 2016; Simnett er al., 2009.
LEVERAGE Calculated by dividing total (non-current) liabilities by total assets.
(Thomas Reuters Worldscope Fundamentals)
Barnett and Salomon, 2012;
Simnett et al., 2009; Braam et al., 2016.
FCF Calculated by dividing cashflow from operations by the company’s total assets. (Thomas Reuters Worldscope Fundamentals)
Surroca et al. (2010) argue that cash flows influence the ability of the company to engage in corporate social responsibility, whereas higher cash flows amounts to higher accessibility to CSR. The cashflow from operations will be normalized by the company’s total assets (Mishra and Modi, 2013).
3.3. Regression model
The choice of a suitable regression model starts with the characteristics of the dataset. The data used can be seen as cross-sectional time-series data (or panel data), as it contains a multitude of companies over a period spanning from 2013 till 2019. There are three methods to analyse panel data: Pooled ordinary least squares (OSL), fixed effects model or random effects model.
As first step in determining the appropriate model I ran the both the Lagrange Multiplier test by Breusch and Pagan (1980) and the modified Wald test to determine if the pooled OLS method is an appropriate model for regression. These tests describe whether any heteroskedasticity exists in the regression, entailing that variance in error terms might be caused by independent variables. Both test statistics show that the probability of heteroskedasticity being present is highly significant (P < 0.01). This renders the pooled OLS model obsolete as it is based on the assumption of homoskedasticity.
The next step was to run the Hausman test statistic to determine the appropriate regression model. This test examines whether correlation exists between individual effects and other regressors in the model, if this correlation exists the random effects model is no longer suitable due to violation of a Gauss-Markov assumption. The test shows that p < 0.01, thus this significant value infers the use of the fixed effects model.
The main objective of this research is to check for a relation between the extensiveness of assurance and corporate environmental performance. To test the earlier described hypotheses the following general model was used:
𝐶𝐸𝑃𝑖𝑡 = 𝛽0+ 𝛽1𝐸𝑋𝑇𝑖𝑡−1+ 𝛽2𝑆𝑇𝐴𝐾𝐸𝐻𝑂𝐿𝐷𝐸𝑅𝑖+ 𝛽3𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌𝑖+ 𝛽4𝑆𝐼𝑍𝐸𝑖𝑡+ 𝛽5𝑅𝑂𝐴𝑖𝑡
+ 𝛽6𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸𝑖𝑡+ 𝛽7𝐹𝐶𝐹𝑖𝑡 + 𝜀𝑖𝑡
(Model 1)
To test the first hypothesis, the model was used in a fixed effects setting, following the test results as described earlier. Where CEP is the score of the environmental pillar from Thomson Reuters Asset4 ESG data. EXT is the lagged score of the extensiveness of assurance. This score is lagged because assurance, and sustainability disclosure in general, is usually provided in the months after the company’s fiscal year end. For example, company X presents sustainability disclosure accompanied by assurance on the 2018 fiscal year in April 2019. To test the relation between the assurance presented in 2019 and the companies’ environmental performance in 2018 would go against the causal relationship and, above all, common sense. The lagging of the extensiveness of assurance variable also limits reverse causality
causal order. A time fixed effect term is added to the equation in the form of year dummies as control variables, after I ran a test for time-fixed effects on the original equation. This tests if the coefficients for all years are jointly equal to zero, which in this case has to be rejected in favour of time-fixed effects.
To test the second and third hypothesis dummy variables were used for country and industry respectively (STAKEHOLDER and INDUSTRY in the model), as depicted in Table 2. The regression model will be changed from the fixed effects because the time invariance of the independent (dummy) variables. These dummy variables are expected to have been the same for all observed years for each individual company. For this reason, the second and third hypotheses were regressed following a between estimation approach. This entails that the model uses a regression based on group means. Because of the time invariant characteristics of the variables, differences could only be found between companies instead of within the companies observed. I will still control for time fixed effects. Because the step away from the fixed effects model in favour of a between estimation I can also control for industry specific effects using the TRBC codes. These industry codes are time invariant therefore these could not be used in the fixed effects model. The stakeholder and industry dummy have been scored as described in Table 2. The between regression examines the independent relation of EXT on CEP and STAKEHODLER on CEP followed by an interaction term of the two (Model 2).
𝐶𝐸𝑃𝑖𝑡 = 𝛽0+ 𝛽1𝐸𝑋𝑇𝑖𝑡−1+ 𝛽2𝑆𝑇𝐴𝐾𝐸𝐻𝑂𝐿𝐷𝐸𝑅𝑖 + 𝛽3𝐸𝑋𝑇𝑖𝑡−1∗ 𝑆𝑇𝐴𝐾𝐸𝐻𝑂𝐿𝐷𝐸𝑅𝑖+ 𝛽4𝑆𝐼𝑍𝐸𝑖𝑡
+ 𝛽5𝑅𝑂𝐴𝑖𝑡+ 𝛽6𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸𝑖𝑡+ 𝛽7𝐹𝐶𝐹𝑖𝑡 + 𝜀𝑖𝑡
(Model 2)
This was similar for the INDUSTRY variable (Model 3).
𝐶𝐸𝑃𝑖𝑡 = 𝛽0+ 𝛽1𝐸𝑋𝑇𝑖𝑡−1+ 𝛽2𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌𝑖 + 𝛽3𝐸𝑋𝑇𝑖𝑡−1∗ 𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌𝑖+ 𝛽4𝑆𝐼𝑍𝐸𝑖𝑡 + 𝛽5𝑅𝑂𝐴𝑖𝑡
+ 𝛽6𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸𝑖𝑡 + 𝛽7𝐹𝐶𝐹𝑖𝑡 + 𝜀𝑖𝑡
(Model 3)
4. Results
4.1. Descriptive statistics
Table 3 gives a summary of the descriptive statistics for the dependent, independent and control variables (panel A, B and C respectively). Panel A shows the mean, median and standard deviation of the CEP scores are 64.514, 67.890 and 20.621 respectively. Sassen et al. (2016) show a comparable value for the mean CEP score (63.54), while their median and standard deviation are substantially higher (73.46 and 29.21). The mean CEP score is higher for companies in sensitive industries (65.420),while lower for firms in shareholder oriented (common law) countries (62.199). The slight increase of the mean CEP score in sensitive industries can be anticipated following the proposed effects mentioned in 2. ‘Theoretical framework and Hypothesis development’: companies considered sensitive perceiving pressure as a result of public concerns. The lower mean CEP score in shareholder oriented (common
law) countries could be attributed to a lesser degree of stakeholder interest and mostly profitability driven companies.
EXT scores of the sample range between 1 and 32. This is in line with the scoring system used as described in section 3.2.2. The maximum score according to the criteria of Gürtürk and Hahn is 33. 14 observations with the EXT score of zero were removed as this is most likely an error or indicative of the lack of assurance in general. The mean of the EXT scores is 17.638 and standard deviation 4.489, which are comparable to the those found in Gürtürk and Hahn study (2016) (16.86 and 5.41). The higher standard deviation could be explained due to the fact that their research only includes 61 observations. Again the mean scores are (as with CEP scores) slightly higher (18.022) for companies in sensitive industries and slightly lower (17.048) for companies in shareholder oriented countries. Again, as with CEP, similar reasoning as described above can explain this deviation of means in EXT scores.
Panel B also includes the two dummy variables STAKEHOLDER and INDUSTRY since these are independent dummy variables for hypotheses two and three. The mean for STAKEHOLDER, e.g. country, is 0.299. In the Simnett et al. study (2009) the mean is 0.559 across 40,993 observations, however their sample is somewhat drawn towards common law countries with larger European and Asian countries (i.e. Russia and China) not present. The mean for industry sensitivity is 0.216 which is comparable to the Braam et al. (2016) industry sensitivity mean of 0.282 on a far smaller sample of only 209 observations.
Variable Mean Median Std. dev. Min. Max. Skewness N
Panel A: Dependent variable
CEP 64.514 67.890 20.621 0 98.270 -0.830 2149
CEP (Shareholder orientation) 62.199 65.985 21.794 0 98.200 -0.712 642 CEP (Industry sensitivity) 65.420 67.450 16.649 0 94.650 -0.709 464 Panel B: Independent variables
EXT 17.638 17 4.489 1 32 -0.022 2149
EXT (Shareholder orientation) 17.048 17 4.796 2 32 0.055 642
EXT (Industry sensitivity) 18.022 18 4.578 2 32 -0.137 464
STAKEHOLDER 0.299 0 0.458 0 1 0.879 2149
INDUSTRY 0.216 0 0.412 0 1 1.381 2149
Panel C: Control variables
SIZE 18.640 18.176 2.759 12.496 27.123 0.596 2149 ROA 5.083 4.190 7.882 -70.080 128.420 4.103 2149 LEVERAGE 34.403 31.890 23.288 0 286.120 1.927 2149 FCF 0.080 0.072 0.072 -0.329 0.827 1.809 2149 Table 3 Descriptive statistics
Table 4 presents the correlation coefficients between the dependent, independent and control variables as displayed in table 3 under panel B and C. The correlation matrix nor the variance inflation factors (VIF) suggest any problems regarding multicollinearity. A VIF greater than 10 indicates multicollinearity issues between the variables, the highest test result found in the sample was 1.48, however.
4.2. Regression results
The results of the regression examining the effect of extensiveness of assurance on corporate environmental performance is shown in Table 5. Model 0 gives the results from the controls only fixed effect regression. Showing that for the full sample of 2,149 observations across 843 companies the control variables SIZE (β 2.700; p < 0.05) and LEVERAGE (β 0.0328; p < 0.1) have both a positive and significant effect on CEP.
Model 1 shows the regression using fixed effects on a sample 1306 observations across 612 companies, the minimum number of observations per company being 1 and the maximum 4. Model 1 presents the regression results of CEP as a function of EXT, following the first hypothesis. The model shows that the effect of the independent variable EXT is negatively related to companies environmental performance (β -0.117; p < 0.1). This result indicates that companies who obtain more extensive assurance on their sustainability disclosure are more likely to achieve lower environmental performance. This is contradicting the first hypothesis, which states that companies with extensiveness of assurance would be positively related to environmental performance. The variable SIZE appears to be positively related to environmental performance (β 2.644; p < 0.1), suggesting that larger companies perform better in environmental areas. The fit of the model, depicted by the R2 coefficient, is rather low (0.101) when compared to research that use the CEP variable in similar manner (Braam et al., 2016; Sassen et al., 2016; Garcia et al., 2017). The R2 coefficient in these articles’ regressions sits around 0.40-0.50 on average. This indicates that Model 1 only presents limited explanation for the observed outcome.
1 2 3 4 5 6 7 8 1 CEP 1.000 2 EXT -0.041 1.000 3 STAKEHOLDER -0.095*** -0.086** 1.000 4 INDUSTRY -0.010 0.049 0.014 1.000 5 SIZE 0.014 0.121*** -0.203*** -0.102*** 1.000 6 ROA 0.027 -0.053 0.063* -0.050 -0.106*** 1.000 7 LEVERAGE 0.076** -0.059* 0.071* -0.097*** -0.001 -0.132*** 1.000 8 FCF 0.040 -0.031 0.053 0.091** -0.202*** 0.523*** -0.213*** 1.000 Correlation matrix Table 4
The second model shows the results of the between estimation regression of CEP as a function of EXT, STAKEHOLDER and an interaction term of the two. Controlling both for time fixed effects and industry specific effects, as described in 3.3.‘Regression model’. Model 2 follows the second hypothesis, utilizing the exact same sample as used in Model 1. The model shows that EXT has a negative although not significant effect on CEP (β -0.284; p > 0.1). The effect of STAKEHOLDER on CEP is negative and significant (β -19.06; p < 0.01). The interaction term, which measures the moderating effect, is positive and significant (β 0.840; p < 0.05). This indicates that the effect of extensiveness of assurance on corporate environmental performance (as described in the first hypothesis), is stronger for companies in shareholder oriented countries. Albeit stronger for a negative coefficient, thereby contradicting hypothesis 2, which states that shareholder oriented companies are more likely to achieve higher environmental performance as an effect of the extensiveness of assurance. Furthermore the control variables ROA, LEVERAGE and FCF are significant at the 0.05 level. The goodness of fit of Model 2 is higher than that of Model 1 with an R2 of 0.164, indicating that the model has higher explanatory power. However, Model 2 uses a different method of regression as Model 1, making it difficult to compare the two side by side.
Variables Fixed effect Fixed effect Between estimation Between estimation
Model 0 Model 1 Model 2 Model 3
EXT -0.117* (0.0681) -0.284 (0.258) 0.130 (0.236) STAKHOLDER -19.06*** (7.278) STAKHOLDER x EXT 0.840** (0.416) interaction term INDUSTRY 16.38 (18.04) INDUSTRY x EXT -0.336 (0.478) interaction term SIZE 2.700** (1.106) 2.644* (1.590) 0.536 (0.336) 0.730** (0.331) ROA -0.0102 (0.0352) 0.0169 (0.0380) -0.354** (0.154) -0.366** (0.155) LEVERAGE 0.0328* (0.0185) 0.0194 (0.0243) 0.0902** (0.0373) 0.0913** (0.0376) FCF 6.328 (5.627) -1.079 (7.104) 42.59** (17.80) 43.08** (17.97) Constant 5.734 (21.07) 12.15 (30.17) 49.30* (25.37) 26.73 (27.33) Observations 2149 1306 1306 1306 R-squared 0.120 0.101 0.164 0.148 Number of DSID2 843 612 612 612 Controlled for:
Time fixed effects yes yes yes yes
Industry specific effects no no yes yes
Regression results with CEP as dependent variable Table 5
Significance levels indicated with *** p<0.01, ** p<0.05, * p<0.1. Standard errors in parentheses.
Model 3 shows the results of the regression examining the moderating effect of sensitive industries on the relation between extensiveness of assurance and corporate environmental performance, the third hypothesis. The results in the model show no significant relationships between EXT and CEP and the influence of INDUSTRY thereon. Showing no apparent evidence to support hypothesis 3, e.g. the effect of extensiveness of assurance on environmental performance is not significantly influenced by industry sensitivity. All of the control variables show to be significant where p < 0.05. The individual coefficients of these control variables are consistent with those in Model 0 (controls only model) and Model 3. The fit of the model is slightly lower as that of Model 2, this indicates that the explanatory power of the model including INDUSTRY is lower than the model including STAKEHOLDER.
4.3. Additional tests and robustness checks
To investigate the robustness of the findings described above two additional tests have been constructed. Both of these tests make use of a different sample in the regression to check if the results are consistent across different sections of the total sample. Both tests are focussed on the main relation contemplated under hypothesis 1.
The first additional test is a change in sample to adhere for any bias or problems related to the unbalanced characteristics of the panel set (Wooldridge, 2010). For this test the sample for the regression has been reduced to firms that are consistently present throughout the original sample (Sassen et al., 2016). This includes firms with observations across four or more subsequent years. The regression observes firms with 4 and 5 years per firm, resulting in 3 and 4 observations per firm due to the lagged characteristic of the EXT variable. This test into firms with larger subsequent years observations (more than 4) makes for an excellent additional check in finding within-group variation. This test follows a similar regression as used in Model 1, using the firm fixed effects and, as before, controlling for time-fixed effects with a year dummy.
The regression results from the first additional test are shown in Table 6, Model 4. This model shows the regression using fixed effects on a sample of 748 observations across 244 companies. The relationship between EXT and CEP is consistent with the results found under the initial hypothesis (Table 5, Model 1), with a slightly greater significant negative coefficient for EXT (β -0.168; p < 0.05). Implying that higher extensiveness of assurance on sustainability disclosure is related to lower environmental performance. Strengthening the results found in Model 1. In contrast to the original regression in Model 1, the effect of SIZE on CEP is not significant. The fit of the model (0.110) is comparable to the fit of the regression of model 1 (0.101).
The second test reduces the original sample to reports scored by the master students only. As described in section 3.2.2. ‘Independent variables’ assurance statements were scored by 28 students from both bachelor and master programs. Because the process of scoring these statements can be subjective to personal interpretation, a regression is ran on a sample consistent of assurance statements scored by the
four master students, including myself. As described before, this group had frequent contact on scoring issues and consensus was reached on these problems. For this reason, the extensiveness of assurance scores are expected to be more consistent.
The sample consists of two year observations per company due to the lagged independent variable EXT. This entails that the regression technique used changes from a fixed effects panel data regression to a normal OLS regression model using the same control variables as used in the original regression, testing the first hypothesis. The use of OLS regression also allows control for industry specific effects. This is added to the regression in an identical way as in Model 2 and 3, using the TRBC codes. Additionally I control for any time-fixed effect with the inclusion of a year dummy.
The result of this regression are shown in Table 6, model 5. The relation between the independent variable EXT and CEP is showing to be negative, consistent with the original model. However this relationship lacks significance (β -0.109; p > 0.1). The model further shows that the control variables SIZE and LEVERAGE have a significant positive effect on the corporate environmental performance (SIZE: β 1.576, p < 0.01; LEVERAGE: β 0.169, p < 0.01). This indicates that both larger companies and companies with a higher leverage are more likely to have higher environmental performance. The R2 of the model is 0.262, which is substantially higher than Model 1 presents. This could be caused by the difference in sample size (313 observations) or, most likely, due to the usage of the OLS regression method as oppose to a fixed effects regression model.
Variables Fixed effect OLS
Model 4 Model 5 EXT -0.168** (0.0808) -0.109 (0.264) SIZE 0.761 (2.098) 1.576*** (0.466) ROA 0.00895 (0.0462) 0.229 (0.267) LEVERAGE -0.000903 (0.0274) 0.169*** (0.0566) FCF -2.024 (9.311) 10.43 (22.68) Constant 54.72 (39.84) 31.92* (17.93) Observations 748 313 R-squared 0.110 0.262 Number of DSID2 244 313 Controlled for:
Time fixed effects yes yes
Industry specific effects no yes
Significance levels indicated with *** p<0.01, ** p<0.05, * p<0.1. Standard errors in parentheses.
See Table 2 for variable definitions. Table 6
These additional robustness tests show that most of the results are similar to the results in model 1, regarding the original sample, albeit lacking significance in individual coefficients, as mentioned before. The similarities of the first test support results depicted in model 1. The second test using ordinary least squared regression shows similar coefficients while displaying differences in statistical significance, however this could be explained due to the fact that sample is rather small (313 observations).
5. Discussion
5.1. Conclusion and interpretation
The main objective of this thesis is to answer the following research question:
To what extent does assurance on sustainability reporting affect corporate environmental performance?
To investigate this relationship a sample of 843 globally distributed companies from the period 2013 till 2019 was used in a regression analysis. A total of 2,149 assurance statements were scored and incorporated into the independent variable extensiveness of assurance. This variable is a quantification of content present in the assurance statement according to the predetermined coding scheme of Gürtürk and Hahn (2016). Put simpler, a number indicating that predisposed criteria are present in the assurance statement. The expected effect of the extensiveness of assurance on corporate environmental performance was positive, based on internally oriented benefits of assurance
My results indicate that a significant relationship exists between the extensiveness of assurance and corporate environmental performance. However, in contradiction to the expectations, this main relation is found to be negative. In other words, companies that have higher scores on extensiveness of assurance are more likely to have slightly lower environmental performance. Consequently, hypothesis 1 has to be rejected in favour of the null hypothesis. This result is unpredicted and therefore calls for additional research into this relation.
Reasons for this unexpected result could be the caused by a couple of ways. The first reason being the model used. Model 1 follows fixed effects regression, additionally controlling for time-fixed effects. As explained in the results section this makes this model unable to account for industry specific effects, since the industry variables contain time invariant characteristics. The second and the third model both do control for these industry specific effects, resulting in better fits of the models (R2 coefficient is 0.164 and 0.148 respectively). Looking at another reason for the discrepancy between hypothesized relation and result. An argument could be made that internal effects of assurance are simply not as strong and significant as external effects (O’Dwyer et al., 2011). Or, continuing this rationale, that these internal effects are not necessary translated into actual environmental performance.
The moderating effect of stakeholder orientation, contemplated in the second hypothesis, is shown to be positive and significant, indicating that firms which operate in shareholder countries perceive a stronger negative relation between extensiveness of assurance on sustainability reporting and corporate environmental performance. This subsequently provides the evidence that supports hypothesis 2, hence I reject the null hypothesis in favour of the alternative hypothesis. Hypothesis 2 was expected to strengthen the positive relation between extensiveness of assurance on sustainability disclosure and environmental performance. This effect has, however, not been compromised due to the fact that the relation in hypothesis 1 appears to be negatively related, e.g. the null hypothesis is rightfully rejected.
These results are mostly consistent with the results of first hypothesis, despite the fact that the extensiveness of assurance lacks significance the coefficient remains negative. Model 2 indicates that companies in shareholder oriented countries have less incentive to perform in environmental areas, a reasonable argument behind this would be that these companies perceive less pressure from third parties i.e. different stakeholders like environmental organizations.
The third and last relation tested is whether or not a company is operating in a sensitive industry has any moderating effect on the relationship between extensiveness of assurance on sustainability reporting and corporate environmental performance. The expectation was that companies in sensitive industries perceived a stronger effect. However the results are not significant and a solid verdict regarding this hypothesis remains inconclusive. For this reason hypothesis 3 has to be rejected.
In general the findings in this thesis indicate that the extensiveness of assurance is negatively associated with the environmental performance of a company. Stakeholder orientation is shown to have a positive moderation on this aforementioned relationship, while the moderating effect of industry sensitivity remains insignificant. However, comparison between the 3 model remains an issue, since the method used in Model 2 and 3 is different from the one used in Model 1. A detailed description is given in the results section.
Although only one of the initial three hypotheses has been confirmed by means of empirical analysis, this thesis contributes to existing academic research by indicating that a relation exists between the contents of assurance on sustainability disclosure and corporate environmental performance. Circling back to the research question, my results indicate that assurance does seem to have a minor negative effect on corporate environmental performance. However, the results found in this thesis indicate that more research is necessary to gain more understanding regarding this relation. Some directions for future research will be mentioned in the next section following the limitations of this thesis.
5.2. Limitations
assurance statement. This could be enhanced by, for example, interviews with managers. A second point of interest is the imperfect method of measuring extensiveness of assurance. Some drawbacks of the method as used have been mentioned already in section 3.2.2, breaking down in two characteristics of this limitation. The first one being the framework of Gürtürk and Hahn (2016) itself. The problem of giving two separate items the same score even though these might not coincide with the same amount of ‘assurance quality’ (quality as Gürtürk and Hahn (2016) describe it), for this reason I choose to refrain from this description and instead use the term extensiveness of assurance. However this does not take away the fact that all items in the model are weighted the same.
The second problem is that 28 different people were involved in the content analysis following the aforementioned framework. This has the positive side that many assurance statements have been analysed and scored. However this method it subject to some problems as well. While the four master students had close contact regarding differences in interpretation on some points or statements to improve consistency in data. I cannot guarantee the same level of consistency for the data gathered by students from the bachelor programme.
Despite these limitations, the goal of this thesis was to explore the effects of assurance on sustainability disclosure on corporate environmental performance. While there is no conclusive evidence to confirm all the hypotheses, results indicate that such a relation exists and that future research is needed to get a more comprehensive view on this relation.
References
Adams, C. A. and Frost, G. R. (2008). Integrating sustainability reporting into management practices. Accounting forum, 32(4), pp.288-302.
Adams, C. A. and McNicholas, P. (2007). Making a difference. Accounting, Auditing & Accountability Journal, 20(3), pp.382-402.
An, Y., Davey, H. and Eggleton, I. (2011). Towards a comprehensive theoretical framework for voluntary IC disclosure. Journal of Intellectual Capital, 12(4), pp.571-585.
Ball, R., Kothari, S. P. and Robin, A. (2000). The effect of international institutional factors on properties of accounting earnings. Journal of accounting and economics, 29(1), pp.1-51.
Barnett, M. L. and Salomon, R. M. (2012). Does it pay to be really good? Addressing the shape of the relationship between social and financial performance. Strategic Management Journal, 33(11), pp.1304-1320.
Braam, G. J., de Weerd, L. U., Hauck, M. and Huijbregts, M. A. (2016). Determinants of corporate environmental reporting: The importance of environmental performance and assurance. Journal of cleaner production, 129, pp.724-734.
Breusch, T. S. and Pagan, A. R. (1980). The Lagrange multiplier test and its applications to model specification in econometrics. The review of economic studies, 47(1), pp.239-253.
CAN Europe, WWF, HEAL, EEB and Klima Allianz Germany (2014). Europe’s Dirty 30. [online] Brussels, Belgium: CAN Europe, WWF, HEAL, EEB and Klima Allianz Germany, p.10.
Available at: http://awsassets.panda.org/downloads/dirty_30_report_finale.pdf [Accessed 30 Dec. 2019].
Cheng, B., Ioannou, I. and Serafeim, G. (2014). Corporate social responsibility and access to finance. Strategic management journal, 35(1), pp.1-23.
Clarkson, P. M., Overell, M. B. and Chapple, L. (2011). Environmental reporting and its relation to corporate environmental performance. Abacus, 47(1), pp.27-60.
Cohen, J. R. and Simnett, R. (2015). CSR and assurance services: A research agenda. Auditing: A Journal of Practice & Theory, 34(1), pp.59-74.
Corporateeurope.org. (2019). Climate Greenwash Winner Revealed. [online] Available at:
https://corporateeurope.org/en/2009/05/climate-greenwash-winner-revealed [Accessed 30 Dec. 2019].
Deegan, C. (2002). Introduction. Accounting, Auditing & Accountability Journal, 15(3), pp.282-311.
Deegan, C., & Gordon, B. (1996). A study of the environmental disclosure practices of Australian corporations. Accounting and business research, 26(3), 187-199.
Edgley, C. R., Jones, M. J., & Solomon, J. F. (2010). Stakeholder inclusivity in social and
environmental report assurance. Accounting, Auditing & Accountability Journal, 23(4), pp.532-557.
Freeman, R. (2015). Strategic management. Cambridge: Cambridge University Press, pp.pp. 52-82.
Garcia, A. S., Mendes-Da-Silva, W. and Orsato, R. J. (2017). Sensitive industries produce better ESG performance: Evidence from emerging markets. Journal of cleaner production, 150, pp.135-147.
Gibbens, S. (2019). A brief history of how plastic straws took over the world. [online] Nationalgeographic.com. Available at:
https://www.nationalgeographic.com/environment/2018/07/news-plastic-drinking-straw-history-Global Reporting Initiative (2016). GRI 102: General Disclosures. GRI Standards. [online] https://www.nationalgeographic.com/environment/2018/07/news-plastic-drinking-straw-history-Global Reporting Initiative, pp.pp. 41-42. Available at:
https://www.globalreporting.org/standards/media/1037/gri-102-general-disclosures-2016.pdf [Accessed 30 Dec. 2019].
Globalreporting. (2019). Benefits of reporting. [online] Available at:
https://www.globalreporting.org/information/sustainability-reporting/Pages/reporting-benefits.aspx [Accessed 30 Dec. 2019].
Gürtürk, A. and Hahn, R. (2016). An empirical assessment of assurance statements in sustainability reports: smoke screens or enlightening information?. Journal of Cleaner Production, 136, pp.30-41.
Guthrie, J., Petty, R. and Ricceri, F. (2006). The voluntary reporting of intellectual capital. Journal of Intellectual Capital, 7(2), pp.254-271.
Hanburn, M. (2019). McDonald's says that the paper straws that many UK customers hate can't actually be recycled. [online] Business Insider Nederland. Available at:
https://www.businessinsider.nl/mcdonalds-says-its-paper-straws-cant-be-recycled-2019-8?international=true&r=US [Accessed 30 Dec. 2019].
Hodge, K., Subramaniam, N. and Stewart, J. (2009). Assurance of Sustainability Reports: Impact on Report Users' Confidence and Perceptions of Information Credibility. Australian Accounting Review, 19(3), pp.178-194.
Jones, M. and Solomon, J. (2010). Social and environmental report assurance: Some interview evidence. Accounting Forum, 34(1), pp.20-31.
Junior, R., Best, P. and Cotter, J. (2014). Sustainability Reporting and Assurance: A Historical Analysis on a World-Wide Phenomenon. Journal of Business Ethics, 120(1), pp.1-11.
Kolk, A. (2004). A decade of sustainability reporting: developments and significance. International Journal of Environment and Sustainable Development, 3(1), p.51.
Kolk, A. (2007). Sustainability, accountability and corporate governance: exploring multinationals' reporting practices. Business Strategy and the Environment, 17(1), pp.1-15.
KPMG (2013). The KPMG Survey of Corporate Responsibility Reporting 2013. [online] pp.pp. 40-45. Available at: https://assets.kpmg/content/dam/kpmg/pdf/2015/08/kpmg-survey-of-corporate-responsibility-reporting-2013.pdf [Accessed 30 Dec. 2019].
Laville, S. and Watts, J. (2019). Across the globe, millions join biggest climate protest ever. [online] the Guardian. Available at: https://www.theguardian.com/environment/2019/sep/21/across-the-globe-millions-join-biggest-climate-protest-ever [Accessed 29 Dec. 2019].
Mishra, S. and Modi, S. B. (2013). Positive and negative corporate social responsibility, financial leverage, and idiosyncratic risk. Journal of business ethics, 117(2), pp.431-448.
Moroney, R., Windsor, C. and Aw, Y. (2011). Evidence of assurance enhancing the quality of voluntary environmental disclosures: an empirical analysis. Accounting & Finance, 52(3), pp.903-939.
O’Dwyer, B. (2011). The case of sustainability assurance: Constructing a new assurance service. Contemporary Accounting Research, 28(4), pp.1230-1266.
O'Dwyer, B. and Owen, D. L. (2005). Assurance statement practice in environmental, social and sustainability reporting: a critical evaluation. The British Accounting Review, 37(2), pp.205-229.
O’Dwyer, B., Owen, D. and Unerman, J. (2011). Seeking legitimacy for new assurance forms: The case of assurance on sustainability reporting. Accounting, Organizations and Society, 36(1), pp.31-52.
Office of Management and Budget (2017). North American Industry Classification System. [online] pp.pp.1-963
Park, J. and Brorson, T. (2005). Experiences of and views on third-party assurance of corporate environmental and sustainability reports. Journal of Cleaner Production, 13(10-11), pp.1095-1106.
Patten, D. M. (2002). The relation between environmental performance and environmental disclosure: a research note. Accounting, organizations and Society, 27(8), pp.763-773.
Perego, P. and Kolk, A. (2012). Multinationals’ Accountability on Sustainability: The Evolution of Third-party Assurance of Sustainability Reports. Journal of Business Ethics, 110(2), pp.173-190.
Pflugrath, G., Roebuck, P. and Simnett, R. (2011). Impact of Assurance and Assurer's Professional Affiliation on Financial Analysts' Assessment of Credibility of Corporate Social Responsibility Information. AUDITING: A Journal of Practice & Theory, 30(3), pp.239-254.
Refinitiv (2020). Environmental, social and governance (ESG) scores from Refinitiv. [online] pp.pp.1-25
