Integrated reporting and analyst forecast accuracy

(1)

Integrated Reporting and Analyst Forecast Accuracy

Name: Stella Ravesteijn Student number: 10325786

Thesis supervisor: Dr. G. Georgakopoulos Date: 21 June 2017

Word count: 13453

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

Statement of Originality

This document is written by student Stella Ravesteijn who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This study provides evidence on the relationship between Integrated Reporting and analyst forecast accuracy. Specifically, this study examines if the forecast error is smaller for firms that adopted Integrated Reporting, mandatory Integrated Reporting and assurance on Integrated Reporting. The study is conducted using a unique hand collected database. Overall, the results suggest that Integrated Reporting does not increase analyst forecast accuracy, the fact that Integrated Reporting is mandatory does not influence analyst forecast accuracy and assurance on Integrated Reporting does not impact the accuracy of analyst forecasts. Integrated Reporting does therefore not provide an advantage for analyst when predicting earnings.

Keywords Analyst Forecast Accuracy, Assurance on Integrated Reporting, Integrated Reporting, Mandatory Integrated Reporting

(4)

List of equations

Equation 1. Forecast error ... 29

Equation 2: General empirical model ... 31

Equation 3: Hypothesis 1 regression model ... 31

Equation 4: Hypothesis 2 regression model ... 32

(6)

List of figures

Figure 1. Value creation process (IIRC, 2013b, p. 13) ... 15 Figure 2. Financial and information flows in a capital market economy (Healy & Palepu, 2001, p. 408)... 16 Figure 3. Contrasting Models of the Corporation: The Stakeholder Model (Donaldson & Preston, 1995, p. 69) ... 2

(7)

List of tables

Table 1. Literature overview ... 25

Table 2. Sample mutations ... 29

Table 3. Variables ... 31

Table 4. Number of unique observations for IR per fiscal year ... 33

Table 5. Descriptive statistics for the variable IR per fiscal year ... 34

Table 6. Number of unique observations for MANIR per fiscal year ... 34

Table 7. Descriptive statistics for the variable MANIR per fiscal year ... 35

Table 8. Number of unique observations for ASSURIR per fiscal year ... 35

Table 9. Descriptive statistics for the variable ASSURIR per fiscal year ... 36

Table 10. Descriptive statistics for regression variables ... 36

Table 11. Pearson correlation matrix ... 37

Table 12. Regression results of FORERROR and IR ... 39

Table 13. Regression results of FORERROR and MANIR ... 40

(8)

1 Introduction

In this thesis a research about the influence of Integrated Reporting on analyst forecast accuracy will be conducted. In this section the motivation and relevance (§ 1.1) will first be discussed followed by the aim, research question and sub questions (§ 1.2). Concluding the methodology and results (§1.3) followed by the research structure (§ 1.4) will be provided.

1.1 Motivation and relevance

“Integrated Reporting is the way to achieve a more coherent corporate reporting system, fulfilling the need for a single report that provides a fuller picture of organizations’ ability to create value over time.”

(IFAC, 2017, p. 1) The International Integrated Reporting Council (IIRC) was formed in August 2010, the first report was published in 2002, and recently in December 2013 the first guideliness have been released (Cheng, Green, Conradie, Konishi, & Romi, 2014, p. 90; Jensen & Berg, 2012, p. 300). Integrated Reporting is therefore a relative new concept in reporting. However, according to Dumay, Bernardi, Guthrie, and Demartini (2016, p. 178), no research robustly established the benefits of Integrated Reporting. They state that there are opportunities for future research into Integrated Reporting with the emphasis on its history, emerge and future (Dumay et al., 2016, p. 178). Other research also state that there is the need for further research into the value relevance of Integrated Reporting and if, and how, the financial markets react on Integrated Reporting (Villiers, Rinaldi, & Unerman, 2014, pp. 1059-1062). Cheng et al. (2014, p. 99) claim that there is little known about the effects of Integrated Reporting with the emphasis on how Integrated Reporting affects stakeholders and the companies preparing them. Cheng et al. (2014, pp. 91-92) identified a range of potential research areas relateting to the development and implementation of Integrated Reporting. They state that the effects of Integrated Reporting and the extent of these effects on the capital market should be investigated. They specifically mention that research about the effects of Integrated Reporting on analyst following or analyst forecast accuracy is needed (Cheng et al., 2014, p. 100).

Analyst forecasts are used by various financial market participants for resource allocation decisions (Altınkılıç, Balashov, & Hansen, 2013, p. 2550; How, Phung, & Verhoeven, 2005, p. 67). Analyst forecast accuracy is often used as proxies of the quality of the information that in published by analyst (Chen, Lingmin, & Yuanyuan, 2016, p. 217; Coën, Desfleurs, & L'Her, 2009). Both the agency theory and the theory of political economy are related to Integrated Reporting and analyst

(9)

forecasts. The agency problem arises when the managers have more information available than the shareholders, there is information asymmetry (Jensen & Meckling, 1976, p. 309; Akerlof, 1970, p. 489). To limit the information asymmetry, principals monitor agents via financial and non-financial reporting, corporate disclosure is therefore critical for the functioning of an efficient capital market (Healy & Palepu, 2001, p. 406). Iinvestors often use institutions, such as analyst forecasts, to limit the quality uncertainty of these investments opportunities (Akerlof, 1970, p. 499). CSR reporting and Integrated Reporting both increase information disclosure and therefore decrease information asymmetry, the increase in information disclosure should results in more informative analyst forecasts (Healy & Palepu, 2001, p. 417).

Research about the effects of disclosure of non-financial information and the publication of stand-alone corporate social responsibility (CSR) reports on analyst forecast accuracy has been conducted (Dhaliwal, Radhakrishnan, Tsang, & Yang, 2012). Dhaliwal et al. (2012, p. 753) found that the disclosure of non-financial information improved analyst earnings forecasts accuracy. However, there has not been any research conducted about the effects of Integrated Reporting on analyst forecast accuracy. This thesis will therefore contribute to the limited literature on this subject.

1.2 Aim, research question and sub questions

1.2.1 Aim

This study aims to provide better insights in the effects of Integrated Reporting on analyst forecast accuracy.

1.2.2 Research question and sub questions

Based on the aim of the research, the following research question is formulated: “How does Integrated Reporting influence analyst forecast accuracy?”

This research question will be answered by answering the following sub questions: - What is Integrated Reporting?

- What are analyst forecast?

- What are the theories related to Integrated Reporting and analyst forecasts?

(10)

1.3 Methodology and results 1.3.1 Methodology

In order to examine the relation between Integrated Reporting and analyst forecast accuracy, the publication of Integrated Reporting is determined via the GRI Sustainability Disclosure Database. Furthermore the model of Lang and Lundholm (1996, p. 476) is used for the determination of analyst forecast accuracy via the forecast error. Three different samples from the GRI Sustainability Disclosure Database are used to examine the effects of Integrated Reporting, mandatory Integrated Reporting, and assurance on Integrated Reporting. The GRI Sustainability Disclosure Database collects information about Integrated Reports starting 2010 and information about assurance on sustainability and Integrated Reports is collected starting 2012. The Institutional Brokers Estimate System (I/B/E/S) Database collects data about analyst forecasts and currently consists of data until September 2016. The first and second sample therefore includes data from 2010 until 2016 and the third sample consists of data from 2012 until 2016. The first sample, related to voluntary Integrated Reporting, consists of 9,112 observations from 2,790 unique firms from the period 2010-2016. The second sample, related mandatory Integrated Reporting, includes companies located in South-Africa and who publish an Integrated Report and consists of 847 observations from 435 unique firms from the period 2010-2016. The third sample, related to assurance on Integrated Reporting, includes companies that provide some sort of assurance on their Integrated Report and consists of 623 observations from 366 unique firms from the period 2012-2016. OLS regressions are used to determine the relationship between the dependent, independent and control variables.

1.3.2 Results

The OLS regressions results implicate that the publication on an Integrated Report results in less accurate analyst forecasts. Furthermore, the results suggest that there is no effect on analyst forecast accuracy for mandatory or voluntary adoption of Integrated Reporting. The last regression results suggest that there is no effect on analyst forecast accuracy for assurance on Integrated Reporting. Possible explanations for these results include the (un)usefulness of information in Integrated Reports, the cost of information processing, the guidance for Integrated Reporting practices, and the type, scope, and level of assurance on Integrated Reporting.

(11)

1.4 Research structure

This thesis is structured as follows. Section 2 provides the literature review and hypotheses development. Section 3 explains the research methodology, in section 4 the results will be presented and the conclusion and limitations are presented in section 5.

(12)

2 Literature review and hypotheses development

In the following section the concept of Integrated Reporting (§ 2.1) and the concept of analyst forecast (§ 2.2) will be explained. Subsequently different theories related to Integrated Reporting and analyst forecasts (§ 2.3) will be discussed and the hypotheses about the relationship between Integrated Reporting and analyst forecast accuracy (§ 2.4) will be developed.

2.1 Integrated Reporting

2.1.1 Corporate Social Responsibility (CSR)

The growing gap between rich and poor on both national and international level has led to critique from activist on corporate organizations (Waddock, 2008, p. 88). But not only human rights violations are a topic of interest, environmental impact, such as climate change has been attributed to human (industrial) activity (IPCC, 2008). According to the Intergovernmental Panel on Climate Change (IPCC, 2008), changing development paths can make a major contribution to climate change mitigation and adaptation. Even though corporate organizations have impact on economic, social and environmental level they seem to concentrate on financial results and reporting while many stakeholders are affected by their actions (Waddock, 2008, p. 88). These stakeholders have increasing expectations about the responsibilities of corporate organizations (Hahn, 2012, p. 717; Frias-Aceituno, Rodríguez-Ariza, & Garcia-Sánchez, 2014, p. 56). Organizations must therefore balance the need to deliver profits to shareholders and the need to mitigate their negative impact/increase their positive impact on stakeholders (Moan, Lindgreen, & Swaen, 2009, p. 71).

As a response to the increase in expectations, a range of different instruments have been developed to improve, evaluate, communicate, and report on these socially responsible practices (Golob & Bartlett, 2007, p. 1). Not only the expectations about responsibilities have increased, according to Deegan (2002, p. 302), the interest into various social and environmental issues has also increased. Corporate organizations have embraced this new demand of stakeholders and responded by providing various stand-alone non-financial, sustainability reports or corporate social responsibility (CSR) reports to complement their financial reports (Simnett, Vanstraelen, & Chua, 2009, p. 938). CSR reporting is a practice that is adopted by a growing majority of organizations and this is promoted by the Global Reporting Initiative (GRI). To achieve global adoption of CSR reporting the GRI provides global standards on sustainability reporting, the GRI Standards, which is the most widely applied sustainability framework in the world (GRI, 2017).

The GRI standards contain three universal standards and three series of topic specific standards which cover: economic, environmental, and social impacts. The first universal standard,

(13)

GRI 101 – Foundation, contains information on how to use the standards, introduces the ten reporting principles and contains information on how to prepare reports. The materiality principle from this standard is used to identify which topics in the organization have the most impact and influence on stakeholders. Based on this materiality principle specific topics are selected and the relevant topic specific standards are selected and applied. The second universal standard, GRI 102 – General Disclosures, is designed for reporting contextual information about the organization and its reporting practices. The last universal standard, GRI 103 – Management Approach, provides guidelines about how to report on the handling of material topics. This standard is used together with the topic specific standards to explain the topic’s materiality and where its impact occurs (GRI, 2017).

2.1.2 Integrated Reporting

Jensen and Berg (2012, p. 299) conclude that separate financial and CSR reports only make sense when financial and non-financial aspect would occur independently in the organization. However, when CSR is incorporated in the strategy, these aspects are interrelated and should therefore not be reported in separate reports. A report that combines both financial and non-financial aspects would therefore represent a more comprehensive representation of organizations (Jensen & Berg, 2012, p. 299). In August 2010, the International Integrated Reporting Council (IIRC) was formally formed by the Accounting for Sustainability Project and the GRI and they (further) developed the concept of Integrated Reporting (Cheng et al., 2014, p. 90; Humphrey, O'Dwyer, & Unerman, 2017). Users of these Integrated Reports would be able to better understand the cause and effect relationships between financial and sustainability performance (Krzus, 2011, p. 271). Integrated Reporting has several advantages including, but not limited to: more in line information provision with investor’s needs; more accurate non-financial information available to data providers; key users have more confidence; improved decision making, use of resources, risk management and identification of opportunities; improved stakeholder engagement and public image; and lower cost of capital (Frias-Aceituno et al., 2014). The concept of Integrated Reporting was not new but the IIRC made a difference with their global ambitions and reach (Humphrey et al., 2017, p. 40).

Integrated Reporting is voluntary in most countries but there are initiatives to make it mandatory (de Leo & Vollbracht, 2011, p. 70). Today, South Africa is the only country where Integrated Reporting is a mandatory practice. According to the King III principle, starting March 2010 all companies listed at the Johannesburg Stock Exchange are obliged to publish an Integrated Report or to publically explain why they deviated from the King III principle (IRC, 2016). In France the first steps towards mandatory Integrated Reporting are also taken. Companies are

(14)

required to, in accordance with the Grenelle II Act of 2012, mandatory report on environmental and social issues (Cheng et al., 2014, pp. 93-94).

In 2011 the IIRC launched the discussion paper and the pilot programme, an opportunity for businesses and investors to provide input and feedback about the development of the International Integrated Reporting Framework (IIRC, 2013a, p. 1). The discussion paper offered proposals for the development of a framework for Integrated Reporting and how this framework might be established and adopted (Humphrey et al., 2017, p. 41). The voluntary pilot program consisted of two parts: the Business Network and the Investor Network. In the Business Network over 100 firms, located in 25 countries, participated. These firms were the early adopters of the Integrated Reporting principles and participated in the program for three years. During these three years the participating firms provided the IIRC with structured feedback on the Framework while also creating momentum for the implementation of Integrated Reporting (IIRC, 2013a, pp. 1-2). In the Investor Network over 35 providers of financial capital, originating from 12 countries, participated. Providers of financial capital are the primary audience of Integrated Reporting because they will use the information provided in Integrated Reports to make decisions about their capital allocation (IIRC, 2014a, p. 1). The IIRC therefore included them in the development of the International Integrated Reporting Framework and participants were asked to provide insights into their needs from Integrated Reporting. The result of consultation and testing by firms and investors in the pilot programme has led to the introduction of the International Integrated Reporting Framework in 2013 (IIRC, 2014b, p. 2). The discussion paper of 2011 and the concept of Integrated Reporting resulted in a high level of interest and support (Humphrey et al., 2017, p. 44). However, according to Humphrey et al. (2017, p. 40), the vague conceptual underpinnings needed clariﬁcation. Based on the feedback on the discussion paper and results of the pilot programme, the IIRC developed the International Integrated Reporting Framework.

2.1.3 The International Integrated Reporting Framework

The International Integrated Reporting Framework provides the conceptual underpinning to help organizations to implement Integrated Reporting (IIRC, 2014b, p. 2). This framework takes a principal-based approach to balance flexibility and individual circumstances with information needs and the degree of comparability across organizations (IIRC, 2013b, p. 4). According to the IIRC (2013b, p. 4), the purpose of the framework is to provide guiding principles, content elements, and to explain the fundamental concepts that underpin them. The guiding principles underpin preparation, content, and presentation and consists of the seven principles: strategic focus and future orientation; connectivity of information; stakeholder relationships; materiality;

(15)

conciseness; reliability and completeness; and consistency and comparability (IIRC, 2013b, p. 5). The seven content elements, which are fundamentally linked but not mutually exclusive, are: organizational overview and external environment; governance; business model; risks and opportunities; strategy and resource allocation; performance; outlook; and basis of presentation (IIRC, 2013b, p. 5). The framework identifies six fundamental concepts, categories of capitals used by organizations: financial, manufactured, intellectual, human, social and relationship, and natural (IIRC, 2017). The link between the fundamental concepts and content elements is represented in the value creation process. This process is illustrated in figure 1. Value creation process (IIRC, 2013b, p. 13), and is created to give investors and stakeholders more insight in the value creation process of the organization.

The IIRC claims that if more organizations would adopt the International Integrated Reporting Framework there will be more effective resource allocation; end incentive systems that perpetuate short-term thinking and decisions making; result in reporting that is more aligned with modern day business models; lead to investors that are better able to understand a company and its prospects better so they can manage investment risk, validate decisions and assess companies forward looking information; lead to a system of capital allocation that is better aligned to the long-term goals of business and society; and build trust in organizations (IIRC, 2017).

(16)

2.2 Analyst forecast

In this section analyst forecasts and the quality of these forecasts will be discussed in more detail.

2.2.1 Analyst forecast

Analyst process information without bias or delay and analyze market-relevant disclosures (Griffin, 2003, p. 479). They act as intermediaries by receiving and processing information from businesses for investors (Schipper, 1991, p. 105) and therefore add value to the capital market (Healy & Palepu, 2001, pp. 416-417). Capital flows from investors, directly or via financial intermediaries, to organizations and information flows from organizations, directly or via information intermediaries, to investors. Financial analysts are an example of information intermediaries (Healy & Palepu, 2001, pp. 408-409). The flows of capital and information and the function of financial and information intermediaries are schematically presented in figure 2 Financial and information flows in a capital market economy (Healy & Palepu, 2001, p. 408).

Most information used by analyst for their forecasts are directly provided by the firm (Lang & Lundholm, 1996, pp. 467-468). Schipper (1991, p. 105) states that since analysts are sophisticated, informed users, they can be viewed as representatives of the group to whom (financial) reporting is and should be addressed to. Analysts enhance the efficiency of the stock market (Griffin, 2003, p. 480) and according to Brown (1996, p. 40) the primary use of analyst forecasts of earnings is to make investment decisions. Analyst forecasts form a reliable basis for

(17)

determining the expected future returns and expected cash flows of organizations and these forecasts are used by various financial market participants for resource allocation decisions (Altınkılıç et al., 2013, p. 2550; How et al., 2005, p. 67). On average analyst forecasts are informative because the market reacts stronger when analyst forecasts ae presented (Frankel, Kothari, & Weber, 2006, pp. 51-52).

2.2.2 Analyst forecast quality

Analyst forecast accuracy and analyst forecasts dispersion are often used as proxies of the quality of the information that in published by analyst (Chen et al., 2016, p. 217; Coën et al., 2009). In other words, the quality of their reports is determined by the degree of certainty that exists about the firm’s earnings prior to their release (Alford & Berger, 1999, pp. 219-220). Brown (1996, p. 40) concludes that users of analyst forecasts prefer more accurate forecasts and according to Dreman and Berry (1995, p. 30), accurate analyst forecasts are essential for most stock valuation models.

Analyst forecasts play important economic roles (Altınkılıç et al., 2013, p. 250). On average analyst forecasts perform better than time-series methods (Conroy & Harris, 1987, p. 725) and according to Kross, Ro, and Schroeder (1990, p. 461) and Healy and Palepu (2001, pp. 416-417), analyst forecasts are more accurate because time-series models are materially related to the historical variability in the earnings time series. Analyst forecasts also provide better quality forecasts of the earnings than less sophisticated agents (De Bondt & Thaler, 1990, pp. 52-53). Analyst forecasts are however not without flaws. De Bondt & Thaler (1990, p. 57) conclude that a pattern of overreaction, which is present in most professional predictions, is also present in the forecasts of analysts. Overreaction results in forecasts that are too extreme to be considered rational.

2.2.3 Determinants analyst forecasts accuracy

There are several factors that influence analyst forecast accuracy. According to Hayes (1998, p. 299), analyst determine their effort, the amount on information that they will gather, on their expectation on the commission that they will generate from that information. But when an analyst has lower forecasts accuracy than peers, the analyst is more likely to turn over (Mikhail, Walther, & Willis, 1999, p. 185). Mikhail et al. (1999, p. 199) therefore suggest that analyst concentrate on providing accurate forecasts.

Beside these internal factors, which come from the analyst himself/herself, there are also several external factors, characteristics of organizations. These external factors that influence analyst forecast accuracy include: IFRS adoption (Tan, Wang, & Welker, 2011, pp. 1352-1353),

(18)

firm profability (Hwang, Jan, & Basu, 1996, p. 19), and firm size (Hwang et al., 1996, p. 19; Kross et al., 1990, pp. 465-466). These factors will implemented in the empirical design of this study and will be discussed in more detail in section 3.2.2, Control variables.

However, despite the research that has been conducted about analyst forecasts consensus, the degree to which analysts share a common belief, the combination of individual analysts decision making process remain hidden in a black box (Ramnath, Rock, & Shane, 2008, p. 35).

2.3 Theories related to Integrated Reporting and analyst forecasts

There are several theories related to Integrated Reporting and analyst forecasts, such as the agency theory and the theory of political economy. In this section these theories will be discussed in more detail.

2.3.1 Agency theory and information asymmetry

Jensen and Meckling define a principal-agent relationship as: “A contract under which one or more persons (the principals) engage another person (the agent) to perform some service on their behalf which involves delegating some decision making authority to the agent.” (1976, p. 308) When a principal lacks knowledge, ability, or skill to perform a task, they delegate this task to an agent (Linder & Foss, 2015, pp. 344-345). The principal however has a disadvantage, he/she does not have perfect knowledge about the agent’s knowledge, ability, or skill. This disadvantage, ex ante, in the contracting stage, is known as ‘adverse selection’ or ‘hidden information’ (Linder & Foss, 2015, p. 345). But not only is the agent’s ability not (perfectly) observable, the motives and actions of the agent are also not (fully) observable by the principal. This disadvantage, ex post, when the agent carries out the delegated tasks, is known as ‘moral hazard’ or ‘hidden action’ (Linder & Foss, 2015, p. 345). These principal-agent problems arise because both the principal and the agent act to maximize their utility, which may lead to actions of the agent, that are not in the best interest of the principal (Jensen & Meckling, 1976, p. 308). As Chow (1982, p. 273) stated, most firms are characterized by a separation of ownership and management. This results in a principal-agent relationship where the shareholders act as principals and the managers act as agents (Jensen & Meckling, 1976, p. 309; Grossman & Hart, 1983, p. 10). The shareholders cannot monitor the actions of the managers but can only observe the outcome (Grossman & Hart, 1983, p. 10). Because the managers usually do not own big portion of shares and their actions cannot be monitored, they may not allocate firm’s resources in the best interest of shareholders (Chow, 1982, p. 273). These issues are associated with the agency problem and arise because the managers have

(19)

more information available than the shareholders, there is information asymmetry (Jensen & Meckling, 1976, p. 309; Akerlof, 1970, p. 489).

By monitoring the actions of the agent and by providing the agent with contracts that establish incentives which are both beneficial for the agent and the principal, information asymmetry can be (partially) solved (Holmström, 1979, p. 74; Jensen & Meckling, 1976, p. 308). It is however impossible for the principal to solve the information asymmetry problem at zero cost (Jensen & Meckling, 1976, p. 308). Principals can monitor agents via financial and non-financial reporting (Healy & Palepu, 2001, p. 406). According to Healy and Palepu (2001, p. 406), corporate disclosure is critical for the functioning of an efficient capital market. Providing additional information, such as CSR reports or Integrated Reports that show agent’s action, can be used to improve the welfare of both the principal and the agent (Holmström, 1979, p. 75)

Another solution to the agency problem are information intermediaries (Healy & Palepu, 2001, pp. 409-410). The role of information intermediaries is schematically presented in figure 2. As Akerlof (1970, p. 488) stated, investors often use market statistic to determine the quality of investment opportunities. They use institutions to limit the quality uncertainty of these investments opportunities (Akerlof, 1970, p. 499).

CSR reporting and Integrated Reporting both increase information disclosure and therefore decrease information asymmetry. The increase in information disclosure also results in more informative analyst forecasts (Healy & Palepu, 2001, p. 417). Summarizing, disclosures and institutions mitigate information asymmetry and therefore reduces the agency problem (Healy & Palepu, 2001, p. 407), in other words, in theory Integrated Reporting and analyst forecast (accuracy) should reduce the agency problem.

2.3.2 Political economy

Another theory associated with Integrated Reporting is the theory of political economy. “Political economy is the study of the interplay of power, the goals of power wielders, and the productive exchange system.” (Zald, as cited in Gray, Kouhy, & Lavers, 1995, p. 52). Political economy includes market exchanges and analyses the relationships between social, institutional, and the economy (Gray et al., 1995, p. 52). Without the considerations about the political, social, and institutional framework in which economic activities take place, economic issues cannot be investigated since society, politics, and economics are interrelated (Deegan, 2002, p. 292). Gray et al. (1995, p. 52) state that legitimacy theory and stakeholder theory are two overlapping perspectives within the political economy framework. These theories will be discussed in more detail in the sections below.

(20)

2.3.2.1 Legitimacy theory

Legitimacy theory is derived from the notion of a social contract between organizations and members of society (Deegan, 2002, pp. 292-293). All social institutions operate via an expressed or implied social contract (Shocker & Sethi, 1973, p. 97). Society provides organizations with resources and in return these organizations perform socially desired actions and distribute economic, social, and/or political benefits (Deegan, 2002, pp. 292-293; Guthrie & Parker, 1989, p. 344; Shocker & Sethi, 1973, p. 97). The organizations survival and growth are dependent on these social contracts (Guthrie & Parker, 1989, p. 344; Shocker & Sethi, 1973, p. 97; Deegan, 2002, p. 293). Legitimacy theory states that organizations need to act in accordance with these social contracts and that they need to establish congruence between the social values associated with their activities and socially acceptable behavior (Dowling & Pfeffer, 1975, p. 122; Deegan, 2002, pp. 292-293; Suchman, 1995, p. 574). Legitimacy is therefore dependent on the beliefs of society (Suchman, 1995, p. 574; Deegan, 2002, p. 296). Legitimacy is a dynamic concept and changes in socially acceptable behavior or events which impact the reputation of an organization, influence the (perceived) legitimacy of that organization (Deegan, 2002, p. 296).

Since legitimacy enhances stability and comprehensibility of organizational actions, management aims to limit the legitimacy gap (Suchman, 1995, p. 574). The legitimacy gap is the incongruence between an organizations actions and society’s perception of what is socially acceptable behavior (O'Donovan, 2002, pp. 346-347). As discussed in paragraph 2.1, stakeholders have increasing expectations about the responsibilities of corporate organizations (Hahn, 2012, p. 717; Frias-Aceituno et al., 2014, p. 56). Environmental performance and CSR are therefore part of socially acceptable behavior. Social contracts impose pressure on organizations to not only take responsibility for effects on stakeholders with direct economic ties, but also take responsibility for environmental and social impacts (Mobus, 2005, p. 499; Shocker & Sethi, 1973, p. 98).

There are several strategies organizations can implement to reduce the legitimacy gap and to reduce the imposed pressure (O'Donovan, 2002, pp. 346-347). The strategies for gaining, extending, maintaining, or repairing legitimacy differ (O'Donovan, 2002, p. 349). Lindblom (as cited in Deegan, 2002, p. 297) mentions four different strategies namely: inform publics about actions; change the perceptions of publics, but not change its actual behavior; manipulate perception by deflecting attention; and change external expectations of its actions. These strategies can all be implemented via disclosure of information. Voluntary disclosure, such as Integrated Reporting, is used to communicate the actions of an organization to gain legitimacy and to reduce external pressure (Mobus, 2005; O'Donovan, 2002, p. 351). According to Deegan (2002, p. 297), corporate social and environmental disclosure strategies are linked to legitimizing intentions of

(21)

organizations. An organization needs to disclose social information to able society to assess whether they act according to the social contract (Guthrie & Parker, 1989, p. 344). When organizations do not disclose information about their actions, their audiences and society will not be informed whether or not the organization is performing socially desired actions. Organizations cannot gain legitimacy when they do not disclose (O'Donovan, 2002, pp. 345-346) and they therefore choose to disclose more information when the need for legitimacy increases (Brown & Deegan, 1998, p. 33).

2.3.2.2 Stakeholder theory

Stakeholder theory, stakeholder model and stakeholder management have been defined in numerous ways (Donaldson & Preston, 1995, p. 66). According to Freeman (1984), stakeholders are defined as: “Any group or individual who can affect or is affected by the achievement of the firm’s objectives.” (as cited in Clement R. W., 2005, p. 255). In other words, stakeholders have ownership, rights or interest in an organization’s actions (Clarkson, 1995, p. 106). Some stakeholders are for example: the organization’s shareholders, its employees, customers, suppliers, community organizations, environmentalists, governments, special interest groups, media, and even competitors (Clement R. W., 2005, p. 255). Freeman (1984) defines stakeholder theory as: “Managers must pay attention to any group or individual who can affect or is affected by the organization’s purpose, because that group may prevent (the ﬁrm’s) accomplishments.’’ (as cited in Brower & Mahajan, 2013, p. 314). Most definitions of stakeholder theory include the principle that an organization must take the interest of people or groups who they have an impact on with

(22)

their actions (Frederick et al., 1992 as cited in Mainardes, Alves, & Raposo, 2011, p. 228). Stakeholder theory and some possible stakeholders are schematically presented in figure 3. Contrasting Models of the Corporation: The Stakeholder Model (Donaldson & Preston, 1995, p. 69). The arrows run from the firms to its stakeholders and vice versa because the firm influences its stakeholders and stakeholders influence the firm (Donaldson & Preston, 1995, p. 68).

The stakeholder-manager relationship are dependent on three attributes of a stakeholder: the degree power, the degree power legitimacy, and/or the degree power urgency (Mitchell, Agle, & Wood, 1997, p. 864). The degree of power is according to Weber (1947) dependent on: “the probability that one actor within a social relationship would be in a position to carry out his own will despite resistance.” (as cited in Mitchell et al., 1997, p. 865). Legitimacy, as mentioned in section 2.3.2.1, is based on socially desired behavior. The degree of urgency is based on the need for (immediate) attention of a stakeholder (Mitchell et al., 1997, pp. 866-867). Organizations need to communicate the efforts they exerted to keep the support of stakeholders. They communicated via disclosure of information, including CSR reporting and Integrated Reporting (Deegan & Blomquist, 2006, p. 349). As Brower and Mahajan (2013, p. 315) conclude, reporting on CSR is difficult and costly to imitate, organizations that therefore report on such practices build support from stakeholders, which is critical for its success. Stakeholder theory therefore encourage organizations to “consider their responsibilities toward several stakeholders with the goal of integrating economic, social, and environmental concerns into their strategies, their management tools, and their activities, going beyond simple compliance.” (Russo & Perrini, 2010, p. 208).

2.4 Relationship between Integrated Reporting and analyst forecast accuracy

In this part of the thesis the hypotheses about the relationship between Integrated Reporting and analyst forecast accuracy will be developed. An overview of prior literature that is used in the development of these hypotheses can be found at the end of this paragraph in table 1. Literature overview.

2.4.1 Hypothesis 1

Frankel et al. (2006) investigated cross-sectional determinants of the informativeness of analyst research. Their research concludes, amongst other things, that analyst forecasts are less informative when information processing costs are higher (Frankel et al., 2006, pp. 31-32). Since Integrated Reporting is designed to make information provision less complex and more informative about the relationship between financial and sustainability performance the information processing costs should decline and the analyzed information should be more informative and thus more accurate

(23)

(Jensen & Berg, 2012, p. 299; Krzus, 2011, p. 271). The reliability of analyst forecasts, their accuracy, is a proxy of information production about a firm (Frankel et al., 2006, p. 52). Kross et al. (1990, p. 462) also conclude that analyst forecasts improve as more information becomes available. Lang and Lundholm (1996, p. 467) provide evidence that firms with more informative disclosure policies have more accurate forecasts and less dispersion among individual analyst. Hope (2003, pp. 264-265) also concludes that the quantity of firm-level disclosures is positively associated with analyst forecast accuracy and according to Bhat, Hope, and Kang (2006, p. 730), corporate governance transparency also positively influences analyst forecast accuracy. Firm-specific effects are the main determinants of forecast accuracy (Coën et al., 2009, p. 469). Since firm-specific information is given in Integrated Reports the forecast accuracy should be effected by the adoption of Integrated Reporting. Dhaliwal et al. (2012) investigated the effect of stand-alone CSR reports on analyst forecast accuracy. They find that the publication of CSR reports results in less forecast errors, in other words higher forecast accuracy (2012, pp. 752-753). Based on these studies the following hypothesis is formulated:

Hypothesis 1: The implementation of Integrated Reporting is positively associated with analysts forecast accuracy.

As mentioned in section 2.1.2, Integrated Reporting is a voluntary practice in most countries with the exception of South Africa (IRC, 2016). Starting March 2010 all companies listed at the Johannesburg Stock Exchange are obliged to publish an Integrated Report or to publically explain why they deviated from the King III principle.

The effect of mandatory adoption of Integrated Reporting in South Africa on forecast accuracy has been investigated by Bernardi and Stark (2016). They found that after mandating Integrated Reporting the forecast accuracy increased but contribute this to the increase in disclosure levels (Bernardi & Stark, 2016, p. 15). Besides the research of Bernardi and Stark (2016) there has been little research into the effects of mandating Integrated Reporting on analyst forecasts. There is however extensive literature about the effects of mandating reporting standards on analyst forecast accuracy with emphasis on the mandatory implementation of International Financial Reporting Standards (IFRS). Prior literature finds that forecast errors decrease, and thus forecast accuracy increases, with the mandatory adoption of IFRS (Byard, Li, & Yu, 2011; Glaum, Baetge, Grothe, & Oberdörster, 2013; Horton, Serafeim, & Serafeim, 2013; Tan et al., 2011). The improvement of the forecast accuracy as a result of the introduction of these international accounting standards may be due to an increase the quality of disclosures. However, this disclosure

(24)

effect only explains a small portion of the improvement in forecast accuracy (Glaum et al., 2013, p. 80). Glaum et al. (2013, p. 80) find that there is no difference in the forecast accuracy of early adopters (voluntary adoption) of IFRS and the late adopters (mandatory adoption) of IFRS. However, Horton et al. (2013, p. 389) find that the improvement in forecast accuracy is greater for mandatory adopters relative to non-adopters and voluntary adopters. This indicates that the mandatory part of IFRS, and not the disclosure part, is the cause of the increase in forecast accuracy. Based on these studies the following hypothesis is formulated:

Hypothesis 2: The mandatory adoption of Integrated Reporting is positively associated with analyst forecast accuracy

The quality of financial statements is externally assured by an audit. The accuracy, fairness, and conformity with principles is assessed by an independent party (Choi, Kang, Kwon, & Zang, 2005, pp. 37-38). Pﬂugrath, Roebuck, and Simnett (2011, p. 241) state that externally assured information is perceived as more credible and reliable than non-assured information. The increase in quality of information as a result of external assurance may therefore result in less analyst forecast errors and more accurate forecasts (Behn, Choi, & Kang, 2008, p. 330). Behn et al. (2008, p. 347) and Choi et al. (2005, p. 38) confirm this and find a positive relation between general audit quality and forecast accuracy. According to Pﬂugrath et al. (2011, p. 240) assurance on CSR reports result in an increase in the perceived credibility of these reports. More specifically, Casey and Grenier (2015, p. 122) find that analyst forecast errors decrease when CSR reports are assured. Based on these studies the following hypothesis is formulated:

Hypothesis 3: External assurance of Integrated Reporting is positively associated with analyst forecast accuracy

(25)

Table 1. Literature overview

Reference Topic Importance Sample origin Years Proxy dependent variable

(Behn, Choi, & Kang, 2008) Association between audit quality and properties of analyst’ annual earnings forecasts

Research looks at quality

of external assurance I/B/E/S, Compustat 1996 - 2003 Forecast accuracy: (Lang & Lundholm, 1996) (Bernardi & Stark, 2016) Association between mandatory

adoption of Integrated Reporting in South Africa and analyst earnings forecast Research looks at mandatory adoption of Integrated Reporting Bloomberg, I/B/E/S, Compustat

2008 -2012 Forecast accuracy 1: (Lang & Lundholm, 1996) Forecast accuracy 2: difference between actual earnings per share and median analyst

forecast earnings per share, divided by actual earnings per share (Bhat, Hope, & Kang, 2006) Association between corporate

governance transparency and accuracy of analyst forecasts

Research looks at the availability of ﬁrm-speciﬁc non-financial information

I/B/E/S,

Compustat 1992 - 2002 Forecast accuracy: (Lang & Lundholm, 1996) (Byard, Li, & Yu, 2011) Association between mandatory

IFRS adoption on financial analysts’ information environment Research looks at mandatory adoption of accounting standards I/B/E/S, Compustat, Datastream

2003 - 2006 Forecast accuracy: (Lang & Lundholm, 1996)

(Casey & Grenier, 2015) Investigation of the market for

assurance of CSR reporting Research looks at CSR assurance I/B/E/S, Compustat, Corporate Register database

1993 - 2010 Forecast error: difference between absolute value of actual earnings and the mean forecast, deﬂated by beginning stock price

(Choi, Kang, Kwon, & Zang,

2005) Association between the quality of external audit and earnings predictability of firms

Research looks at quality

of external assurance I/B/E/S, Bank of New York Guide to Depository Receipts

(26)

(Coën, Desfleurs, & L'Her,

2009) Determinants of earnings forecast accuracy Research looks at firm-speciﬁc effects I/B/E/S 1990 - 2006 Forecast accuracy: forecast error calculated as difference between forecasted earnings per share and actual earnings per share

(Dhaliwal, Radhakrishnan,

Tsang, & Yang, 2012) Association between the disclosure of non-financial information and analyst forecast accuracy

Research looks at the disclosure of non-financial information I/B/E/S, Compustat, Corporate Register, CRSP

(Frankel, Kothari, & Weber,

2006) Determinants of informativeness of analyst research Research looks at the usefulness of financial reports and information processing costs

I/B/E/S, CDASpectrum, CRSP,

Compustat

1995 - 2002 Analyst informativeness: absolute abnormal stock price reaction on the dates analyst release forecast

revisions (no restriction on forecast revision)

(Glaum, Baetge, Grothe, &

Oberdörster, 2013) Association between international accounting standards and the accuracy of analysts’ forecasts

Research looks at mandatory introduction of accounting standards I/B/E/S, Financial reports from Germany

(Hope, 2003) Association between forecast accuracy and level of disclosure, and association between forecast accuracy and degree of enforcement of standards

Research looks at the

annual report disclosure I/B/E/S, CIFAR (1993, 1995)

1991 & 1993 Forecast accuracy: (Lang & Lundholm, 1996)

(Horton, Serafeim, &

Serafeim, 2013) Determinants of IFRS of improvement in the information environment of ﬁrms

Research looks at mandatory adoption of accounting standards

I/B/E/S,

Annual reports 2001 - 2007 Forecast error: absolute difference between actual earnings and consensus forecast, deﬂated by absolute actual earnings

(27)

(Kross, Ro, & Schroeder,

1990) Association between analyst earnings forecasts and firm characteristics Research looks at the information collection and dissemination activities of analysts Compustat, Value Line Investment Survey

1973 - 1981 Forecasts error: difference between forecasted earnings per share and actual earnings per share

(Lang & Lundholm, 1996) Association between disclosure

policy and analyst behavior Research looks at the informativeness of the firm's disclosure policy

I/B/E/S, Compustat, FAF reports

1985 - 1989 Forecast accuracy: absolute forecast error calculated as the difference between actual earnings per share and median analyst forecast of earnings per share, deflated by stock price

(Pﬂugrath, Roebuck, &

Simnett, 2011) Association between assured CSR reports and ﬁnancial analysts’ perceived credibility of the information reported

Research looks at CSR

assurance Experiment with analysts as participants

Not

applicable Not applicable

(Tan, Wang, & Welker, 2011) Association between mandatory adoption of (IFRS) and ﬁnancial analysts

Research looks at mandatory adoption of accounting standards

I/B/E/S,

(28)

3 Research methodology

In the following section the sample selection (§ 3.1) will be discussed. Subsequently, the empirical design (§ 3.2) for testing hypotheses will be explained.

3.1 Sample selection 3.1.1 Research design

In order to answer the research question, an archival research approach with publicly available databases is applied. Data about firms that use and publish Integrated Reports is collected and combined with data about analyst forecasts to determine whether Integrated Reporting influences analyst forecast accuracy. Only data from public-listed companies is used since there is more information available about these firms than private firms and analyst often cover these firms.

The data about the use of Integrated Reporting is collected from the GRI Sustainability Disclosure Database. The GRI Sustainability Disclosure Database is a collection of sustainability reports, and is available to all members of the public. This database currently holds almost 38,400 reports of 10,175 organizations about sustainability and Integrated Reporting. This research includes data from 2010 until 2016 since the GRI Sustainability Disclosure Database collects data about Integrated Reports starting 2010. The GRI Sustainability Disclosure Database does not provide company identifiers other than the company name. The sedol- and cusip-codes (unique company identifiers) are therefore hand collected in order to match the GRI data with data about analyst earnings forecast and financial controls. This extensive data collection process creates a unique dataset.

In line with previous studies, included in table 1, the data about analyst earnings forecast and actual earnings is collected from the Institutional Brokers Estimate System (I/B/E/S) Database. The data of other financial variables for control purposes are collected from Compustat Global and Compustat North America.

3.1.2 Sample

Table 2 provides the sample selection from the GRI Sustainability Disclosure Database and the executed mutations to create the final sample. The GRI Sustainability Disclosure Database contains 34,435 observations from the period 2009 until 2016. This sample is split in firms located in North America and firms located in all other countries (global). The sedol-codes are hand collected for the global firms and the cusip-codes are hand collected for the North American firms. The sedol- or cusip-code could not be collected for 13,555 observations and 2,568 firms have

(29)

missing information or consist of duplicate observations, these observations are therefore excluded from the sample. The GRI Sustainability Disclosure Database is matched with the Compustat and I/B/E/S database via the sedol- and cusip-codes. A total of 6,701 firms could not be matched with the Compustat database and a total of 1,806 firms could not be matched with the I/B/E/S database. These observations are therefore excluded from the final sample. The 693 observations of 2009 are excluded since the GRI Sustainability Disclosure Database collects data about Integrated Reports starting 2010. The final sample contains 9,112 observations total, 1,912 observations from firms located in North America and 7,200 observations from firms located in other countries.

Table 2. Sample mutations

3.2 Empirical design 3.2.1 Main variables

Several studies determined the analyst forecast accuracy by determining by the changes in forecast error, see table 1. In line with these prior studies the dependent variable (DV) used in this research is the forecast error (FORERROR), see equation 1. The forecast error will be measured as the absolute value of analyst forecast error, deflated by stock price (Lang & Lundholm, 1996, p. 476).

Equation 1. Forecast error

(30)

𝐸𝑃𝑆𝑡 = 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑝𝑒𝑟 𝑠ℎ𝑎𝑟𝑒 𝑖𝑛 𝑝𝑒𝑟𝑖𝑜𝑑 𝑡

𝐴𝐹𝑡 = 𝑚𝑒𝑑𝑖𝑎𝑛 𝑎𝑛𝑎𝑙𝑦𝑠𝑡 𝑓𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝑜𝑓 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑝𝑒𝑟 𝑠ℎ𝑎𝑟𝑒 𝑖𝑛 𝑝𝑒𝑟𝑖𝑜𝑑 𝑡

𝑃𝑡 = 𝑝𝑟𝑖𝑐𝑒 𝑝𝑒𝑟 𝑠ℎ𝑎𝑟𝑒 𝑖𝑛 𝑝𝑒𝑟𝑖𝑜𝑑 𝑡

The independent variable for the first hypothesis (IV1) is Integrated Reporting (IR) and will be

measured via a dummy variable equal to 1 if Integrated Reporting is applied and 0 otherwise. The independent variable for the second hypothesis (IV2) is mandatory Integrated Reporting (MANIR)

and will be measured via a dummy variable equal to 1 if the firm is listed at the Johannesburg Stock Exchange and an Integrated Report is published and 0 otherwise. The independent variable for the third hypothesis (IV3) is external assurance on Integrated Reporting (ASSURIR) and will be

measured via a dummy variable equal to 1 if there is external assurance provided and an Integrated Report is published and 0 otherwise. An overview of the used variables is presented in table 3. Variables.

3.2.2 Control variables

There is a need to control for external factors, as discussed in section 2.2.3, that influence analyst forecast accuracy to ensure that the observed effects can be allocated to the adoption of Integrated Reporting, the publication of mandatory Integrated Reporting, or assurance on Integrated Reporting. The first factor that need to be controlled for is IFRS adoption, as discussed in section 2.4.3, accuracy of foreign analyst forecasts is greater for organizations in countries that mandate IFRS than for organizations in non-IFRS countries (Byard et al., 2011; Glaum et al., 2013; Horton et al., 2013; Tan, Wang, & Welker, 2011). The control variable (CV), IFRS adoption (IFRS) is therefore included. Another factor is firm profitability, forecasts about firms that make a loss are less accurate than for firms that make a profit (Hwang et al., p. 19). Hwang et al. (1996, p. 19) also find that accuracy is lower for small firms than for large firms. Kross et al. (1990, pp. 465-466) therefore control in their research for firm size. The control variables, firm profitability (FPROF) and firm size (FSIZE) are therefore included. An overview of the used control variables is presented in in table 3. Variables.

(31)

Table 3. Variables

3.2.3 Model specifications

The hypotheses are tested in the following general empirical model of Hope (2003, p. 250):

Equation 2: General empirical model

𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝑎𝑐𝑐𝑢𝑟𝑎𝑐𝑦= 𝑓(𝐼𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑒𝑑 𝑅𝑒𝑝𝑜𝑟𝑡𝑖𝑛𝑔, 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠)

3.2.3.1 Model hypothesis 1

In order to test the first hypothesis and determine the effect of Integrated Reporting on analyst forecast accuracy, the following OLS (Ordinary Least Squares) regression model is used:

Equation 3: Hypothesis 1 regression model

Δ%𝐹𝑂𝑅𝐴𝐶𝐶𝑅𝐶𝑌(𝑖,𝑡)= 𝛽0+ 𝛽1𝐼𝑅(𝑖,𝑡)+ 𝛽2𝐹𝑃𝑅𝑂𝐹(𝑖,𝑡)+ 𝛽3𝐹𝑆𝐼𝑍𝐸(𝑖,𝑡)+ 𝛽4𝐼𝐹𝑅𝑆(𝑖,𝑡)+ 𝜀(𝑖,𝑡)

i = firm t = year

ε(i, t) = random noise term

(32)

3.2.3.2 Model hypothesis 2

In order to test the first hypothesis and determine the effect of mandatory adoption of Integrated Reporting on analyst forecast accuracy, the following OLS-regression model is used:

Δ%𝐹𝑂𝑅𝐴𝐶𝐶𝑅𝐶𝑌(𝑖,𝑡)= 𝛽0+ 𝛽1𝑀𝐴𝑁𝐼𝑅(𝑖,𝑡)+ 𝛽2𝐹𝑃𝑅𝑂𝐹(𝑖,𝑡)+ 𝛽3𝐹𝑆𝐼𝑍𝐸(𝑖,𝑡)+ 𝛽4𝐼𝐹𝑅𝑆(𝑖,𝑡)+

𝜀(𝑖,𝑡)

i = firm t = year

Table 3 provides the variable definitions presented in this equation 3.2.3.3 Model hypothesis 3

In order to test the first hypothesis and determine the effect of external assurance of Integrated Reporting on analyst forecast accuracy, the following OLS-regression model is used:

Δ%𝐹𝑂𝑅𝐴𝐶𝐶𝑅𝐶𝑌(𝑖,𝑡)= 𝛽0+ 𝛽1𝐴𝑆𝑆𝑈𝑅𝐼𝑅(𝑖,𝑡)+ 𝛽2𝐹𝑃𝑅𝑂𝐹(𝑖,𝑡)+ 𝛽3𝐹𝑆𝐼𝑍𝐸(𝑖,𝑡)+

𝛽4𝐼𝐹𝑅𝑆(𝑖,𝑡)+ 𝜀(𝑖,𝑡)

i = firm t = year

Table 3 provides the variable definitions presented in this equation

3.2.4 Hypotheses testing

In order to determine the effect of Integrated Reporting, mandatory Integrated Reporting and assurance on Integrated Reporting on the analyst forecast accuracy, regressions on the forecast error including multiple control variables are performed. An ordinary least square (OLS) regression model is executed for the forecast error, taking linearity of the dependent and independent variables into account. To extract the outliers from the sample, the variables forecast error (FORERROR) and firm size (FSIZE) are winsorized at the 1st_{and 99}th_{percentile. The validity of}

the regression is ensured by checking whether the residuals of the regressions are normally distributed, linear and independent. A pairwise correlation matrix is created to check for multicollinearity among the variables. Homoscedasticity of the residuals is tested and corrected.

(33)

4 Results

In the following section the results of the empirical tests are explained. A summary of the descriptive statistics (§ 4.1) and the results of hypotheses test (§ 4.2) are outlined.

4.1 Descriptive statistics

Table 4 provides the distribution of the sample used for Integrated Reporting (IR) sorted by year from the period of 2010 until 2016. The growing popularity of CSR reporting results in an increase each year of listed firms that registered their report in the GRI Sustainability Disclosure Database. When the data for this regression was collected, the I/B/E/S database contained data until September of 2016, the sample therefore only contains 78 firms with reports that have a fiscal period end before September 2016. The sample consists of 9,112 observations of 2,790 unique firms.

Table 4. Number of unique observations for IR per fiscal year

Table 5 provides the descriptive statistics related to Integrated Reporting (IR) over the sample period 2010 until 2016. Overall, 9% of the reports in the sample is an Integrated Report with a distribution in relation to the mean, standard deviation, of 0.2904.

(34)

Table 5. Descriptive statistics for the variable IR per fiscal year

Table 6 provides the distribution of the sample used for mandatory Integrated Reporting sorted by year from the period of 2010 until 2016. The sample consists of 847 observations of 435 unique firms.

Table 6. Number of unique observations for MANIR per fiscal year

Table 7 provides the descriptive statistics related to mandatory Integrated Reporting (MANIR) over the sample period 2010 until 2016. Overall, almost 25% of the reports in the sample are of firms situated in South Africa and are obligated to publish an Integrated Report. The distribution in relation to the mean, standard deviation, is 0.4328.

Fiscal year N Percentage 2010 91 10.74% 2011 133 15.70% 2012 141 16.65% 2013 152 17.95% 2014 132 15.58% 2015 192 22.67% 2016 6 0.71% Total 847 100%

Number of unique observations

This table provides the distribution of the number of observations for mandatory Integrated Reporting per fiscal year. The sample consists of 847 observations of 435 unique firms. The sample is from the period of 2010 until 2016.

(35)

Table 7. Descriptive statistics for the variable MANIR per fiscal year

Table 8 provides the distribution of the sample used for assurance on Integrated Reporting sorted by year from the period of 2012 until 2016. The observations of 2010 and 2011 are excluded since the GRI Sustainability Disclosure Database collects data about assurance starting 2012. The sample consists of 623 observations of 366 unique firms.

Table 8. Number of unique observations for ASSURIR per fiscal year

Table 9 provides the descriptive statistics related to assurance on Integrated Reporting (ASSURIR) over the sample period 2012 until 2016. Overall, a little over 51% of the reports in the sample provide some sort of assurance on Integrated Report. The distribution in relation to the mean, standard deviation, is 0.5004.

(36)

Table 9. Descriptive statistics for the variable ASSURIR per fiscal year

Table 10 provides descriptive statistics for the key variables used. The variable forecast error (FORERROR) and firm size (FSIZE) are transformed with the use of a logarithm because of the positive skewness in the sample data. The coefficient of the dependent variable forecast error is in line with the median. Suggesting that the sample is normally distributed. The standard deviation of 1.8660 indicates that the variable has a wide distribution in relation to the mean. In addition, 12% of the firms included in the sample report a loss and the majority of the firms in the sample period (58%) have adopted IFRS.

Table 10. Descriptive statistics for regression variables

Table 11 provides a Pearson correlation matrix of the forecast error in relation to the variables used in the regression. This correlation test is used to test whether correlations between variables exist, and to assess the explanatory power of a variable. As shown in table 11, none of the variables in the sample have a very large level of correlation. Furthermore, there is a slight

(37)

significant negative correlation between firm size (FSIZE) and the other variables used in the regression. The relation between IR, MANIR and ASSURIR are not of importance since these variables will not be used in the same regression.

Table 11. Pearson correlation matrix

4.2 Results of hypotheses test 4.2.1 Regression results Integrated Reporting

Table 12 presents the results of the OLS regression of the forecast error (FORERROR) and Integrated Reporting (IR). The fit of the regression model is tested through the F-test. The results related to IR and FORERROR regression in table 12 depict a significant F-statistic of 432.75 (p-value of 0.0000). This indicates that at least one predictor variable significantly influences the dependent variable FORERROR. However, the adjusted R-squared of 0.1593 indicates that only 15.93% of the variation in FORERROR is explained by the variables used in the model. Hypothesis 1: The implementation of Integrated Reporting is positively associated with analysts forecast accuracy.

The expectation derived from hypothesis 1 was that the variable IR would be positively significant associated with FORERROR. The regression however shows that IR is negatively (coefficient -0.1177) and significantly at 5% level (p-value 0.0489) associated with FORERROR. This result indicates that the forecast error of firms which publish an Integrated Report is greater than the forecast error for firms that do not publish an Integrated Report. The results from the

(38)

regression are therefore not in line with the expectation and the first hypothesis is therefore rejected.

Lang and Lundholm (1996, p. 472) found that the usefulness of disclosures is positively associated with the accuracy of analyst forecasts, our result may therefore indicate that providing an Integrated Report does not provide analysts with useful information. Dhaliwal et al. (2012, pp. 752-753) found that the publication of CSR reports results in less forecast errors, the results from this research may therefor indicate that the combination of non-financial and financial information in an Integrated Report does not provide useful information for analysts. Another possible explanation for the negative significant relationship between Integrated Reporting and forecast error, are higher information processing costs. Integrated Reporting is designed to make information provision less complex but since Integrated Reporting is not yet widely implemented analysts may find the Integrated Reports confusing (Jensen & Berg, 2012, p. 299; Krzus, 2011, p. 271). This could lead to higher information processing costs and lower forecast accuracy (Frankel et al., 2006, pp. 31-32).

The coefficient of the first control variable, firm profitability (FPROF) is positively (coefficient 1.8814) and significantly (p-value 0.0000) associated with FORERROR. This implies that firms that reported a loss have a higher forecast error in comparison with firms that reported a profit. This result is in line with the results of Hwang et al. (1996, p. 19) who found that loss firms are associated with lower analyst forecast accuracy. The coefficient of the second control variable, firm size (FSIZE) is negatively (coefficient -0.0580) and significantly (p-value 0.0000) associated with FORERROR. This implies that large size firms have a lower forecast error in comparison with small size firms. This result is also in line with the results of Hwang et al. (1996, p. 19) who found that large firms are associated with higher analyst forecast accuracy. The coefficient of the third control variable, implementation of IFRS (IFRS) is positively (coefficient 0.5549) and significantly (p-value 0.0000) associated with FORERROR. This implies that firms that implemented IFRS have a lower forecast error in comparison to firms that did not adopt IFRS. This result is in line with the results of Glaum et al. (2013) who found that firms that adopted IFRS are associated with higher analyst forecast accuracy.

(39)

Table 12. Regression results of FORERROR and IR

4.2.2 Regression results mandatory Integrated Reporting

Table 13 presents the results of the OLS regression of the FORERROR and mandatory Integrated Reporting (MANIR). The results in table 13 depict a significant F-statistic of 55.52 (p-value of 0.0000). This indicates that at least one predictor variable significantly influences the dependent variable FORERROR. However, the adjusted R-squared of 0.1984 indicates that only 19.84% of the variation in FORERROR is explained by the variables used in the model.

Hypothesis 2: The mandatory adoption of Integrated Reporting is positively associated with analyst forecast accuracy

The expectation derived from hypothesis 2 was that the variable MANIR would be positively significant associated with MANIR. The regression however shows that MANIR is negatively (coefficient -0.0659) and insignificantly (p-value 0.6061) associated with FORERROR. This result indicates that there no difference between with the forecast error of firms which are obligated to publish an Integrated Report and firms that voluntary publish an Integrated Report. The results from the regression are therefore not in line with the expectation and the second hypothesis is therefore rejected.

Since there is little research into the effects of mandating Integrated Reporting on analyst forecast the argumentation for the second hypothesis was based on the effects of the mandatory implementation of International Financial Reporting Standards (IFRS). The results of the regression however indicates that the effects of mandating Integrated Reporting are not similar to

Integrated reporting and analyst forecast accuracy