The market response to integrated reporting and the moderating effect of ownership concentration

(1)

The market response to integrated

reporting and the moderating effect of

ownership concentration

Evidence from a worldwide sample

⁕

Master thesis, MSc Accountancy

University of Groningen, Faculty of Economics and Business

January, 21st 2019 TEUN SCHAKEL S3504840 Peizerweg 68-38 9726 JM, Groningen tel: +31 6 21 46 36 13 e-mail: t.schakel@student.rug.nl Word count: 7,659

(2)

2

The market response to integrated

reporting and the moderating effect of

ownership concentration

Evidence from a worldwide sample

ABSTRACT: In response to the academic call to expand research on integrated reporting to

other countries, this study investigates the market response to integrated reporting, using a worldwide sample. Building on the agency theory, integrated reporting can reduce information asymmetry between an organization and its shareholders, leading to a reduction in cost of capital. This study takes into account the moderating role of ownership concentration, since organizations with high levels of ownership concentration, experience a second agency problem between controlling shareholders and minority shareholders, leading to greater information asymmetries and higher costs of capital. In a worldwide sample of 16 countries and 5,341 firm-year observations from 2012-2017, results show that integrated reporting is negatively associated with cost of capital, and this association is positively moderated by ownership concentration. This moderating effect tends to be stronger for countries characterized by high levels of ownership concentration. The results of this study suggest that organizations with high levels of ownership concentration, could potentially reduce the information asymmetry with integrated reporting, subsequently rewarded by a reduction in the cost of capital.

Keywords: integrated reporting, cost of capital, ownership concentration, agency theory,

(3)

3

I. INTRODUCTION

ver the last decades, numerous corporate scandals have been brought to light, attracting worldwide public attention. Two well-known examples are the Deepwater Horizon oil spill caused by BP in 2010, or humiliating working conditions in a Chinese factory of Apple. Events like these can harm a company’s image, but more importantly, inflict damage on the planet and the lives of many people. Because of this intersection of business and society, companies are now held at a higher standard by the public. Business operations should be socially responsible and high transparency to stakeholders is demanded.

This transparency is also endorsed by Uyar (2016), who argues that the disclosure of nonfinancial information is growing increasingly in its importance to stakeholders. Nonfinancial information may include information regarding sustainability, human capital, and social relations. To satisfy stakeholders, companies may decide to issue stand-alone sustainability reports (Berthelot et al., 2012). Although these reports provide additional information to stakeholders on top of the annual report, not everybody is convinced about the value relevance of nonfinancial information in this sustainability report (Accounting for Sustainability & GRI, 2012; Ioana & Adriana, 2014). Sometimes, even a contradicting effect is discovered. Some organizations only release sustainability reports or make their reports less readable, in order to cover their actions, which is called ‘greenwashing’ (Marquis et al., 2016; Wang et al., 2018). It is hard for investors to see through this, because the information in the sustainability report is not connected to financial outcomes. Another concern is the timing, since sustainability reports are often published in a late stage, compared to annual reports (Serafeim, 2015). In other words, there is a lack of coherence between the sustainability report and the annual report, which could potentially lead to information asymmetry between the organization and its stakeholders, or not contribute to the reduction of information asymmetry.

Integrated Reporting (IR), a new way of principle-based corporate reporting, could solve this issue (Cheng et al., 2014). According to the International Integrated Reporting Council, an integrated report can be defined as: “a concise communication about how an organization’s strategy, governance, performance and prospects, in the context of its external environment, lead to the creation of value over the short, medium and long term” (IIRC, 2013, p. 7). Companies can achieve this by reporting in accordance with the principles of the IIRC Framework (IIRC, 2013). The added value of IR is that it challenges organizations to combine material non-financial information with financial outcomes. This should lead to coherence between nonfinancial information of the sustainability report and financial information of the

(4)

4

annual report (Baboukardos, 2018; Baboukardos & Rimmel, 2016), while IR can also contribute to an increase in firm value (Barth et al., 2017). Although not everybody endorses the usefulness and relevance of IR and the IIRC (Flower, 2015; Thompson, 2015), most responses are positive. The Federation of European Accountants (2015) already mentioned IR as the next step in corporate reporting. According to the IIRC (2018) an increasing amount of companies adopted IR in their reporting process over the last years.

There are many reasons why this study contributes to the academic literature. First of all, the current literature on IR is limited, because it is a relative new phenomenon in the accounting research area. Just six years ago, in 2012, the first empirical studies were published (Velte & Stawinoga, 2017). Additional research on IR is therefore necessary, in order to expand the current scientific knowledge. Secondly, although Zhou et al. (2017) and Barth et al. (2017) already investigated the relation between IR adoption and the cost of capital, their study is limited to firms with a listing on the South African stock exchange, where the adoption of IR is mandatory since 2010, through the King III Report of Corporate Governance (Lee & Yeo, 2016). This study instead uses a sample with countries where the implementation of IR is still voluntary, and examines the effect of ownership concentration. Although it is an empirical question whether voluntary integrated reports result in a reduction of cost of capital, earlier research from a survey indicates that communicating information in a voluntary setting can reduce a firm’s cost of capital (Graham et al., 2005). A third reason is the limitation of the geographic difference in current studies. Although some researchers use worldwide samples, most research is applied to stock listed companies in South Africa (Velte & Stawinoga, 2017). It is therefore not possible to control for national characteristics. Multiple researchers address the need to expand the IR research to other countries, to ensure that country specific results can be analyzed and compared (de Villiers et al., 2017; Vaz et al., 2016; Velte & Stawinoga, 2017). As far as known, only one multi-country study investigated the relationship between IR and the cost of capital (García-Sánchez & Noguera-Gámez, 2017a). The current study is different from García-Sánchez & Noguera-Gámez (2017a), because it uses a different proxy to measure IR, which enables to indicate the extent of IR, and investigate a relatively larger sample. On top of that, this study takes into account the influence of ownership concentration. A final contribution of the current study to existing literature is that the outcome could potentially influence the willingness of firms to adopt IR.

The purpose of this study is to investigate whether IR (i.e. the extent of practising IR) has an impact on the firm’s cost of capital, which is of interest, because it can indicate the capital market’s reaction to this new information (Zhou et al., 2017). A negative relation with

(5)

5

cost of capital is expected, which would be in line with García-Sánchez and Noguera-Gámez (2017b), who demonstrated that IR can reduce the information asymmetry between the organization and its stakeholders (i.e. the adoption of IR should lead to a decrease in the cost of capital). It is interesting to examine if ownership concentration will influence this effect, because earlier research indicates a positive relationship between ownership concentration and information asymmetry (Fan and Wong, 2002; Jiang et al., 2011). Moreover, Jensen and Berg (2012) found that firms with high ownership concentration are less willing to make IR part of their reporting system. IR could have a stronger impact on companies with high ownership concentration, because those companies experience greater information asymmetries. In the current study, IR is hypothesized to have a negative relation with the cost of capital, while this relation is positively influenced by the level of ownership concentration.

The remaining of this paper is structured as follow. In the next chapter the most important theories related to this topic are presented, followed by the hypotheses. Chapter three gives a description of the methodology of this study. In the fourth chapter, the results are analyzed and presented. This paper ends with a brief conclusion of the study and a discussion of the limitations of this research.

II. THEORY

Concepts of interest

This paper uses different concepts of interest that are important for developing the hypotheses. In this paragraph, a brief description is presented of integrated reporting, cost of capital, and ownership concentration, consecutively.

Integrated reporting. IR is a new way of principle-based corporate reporting, aimed to

combine material nonfinancial information with financial information, presented in a composed overview (IIRC, 2013). The current literature on IR is scarce, but different academics and institutions have attempted to define the concept of it. According to the IIRC (2013, p. 2), the publication of an integrated report is driven by the process of integrated thinking, defined as “the active consideration by an organization of the relationships between its various operating and financial units and the capitals that the organization uses or affects”. Integrated thinking eventually leads to integrated reporting (IIRC, 2013). Perego et al. (2016) describe integrated reports as a value creation story to the stakeholders, using the six capitals of the IIRC Framework: (1) financial, (2) manufactured, (3) intellectual, (4) human, (5) social and

(6)

6

relationship, and (6) natural. Others see IR as an integrated presentation of nonfinancial and financial information (Olivier et al., 2016), by connecting information that was previously separated from each other (Cho et al., 2013; Middleton, 2015). Combining definitions from current literature, IR has a twofold objective. Firstly, it creates value to the stakeholders by providing information about the short, middle, and long term performance of the organization. Secondly, it aims to combine nonfinancial information and financial information together in one composed report.

Cost of capital. Different measures can be used to determine the market’s reaction to

new information, such as analysts’ forecast error and forecast dispersion (Arping & Sautner, 2013; Zhou et al., 2017), or bid-ask spreads and Tobin’s Q (Barth et al., 2015). A common and important measure to determine the reaction of the market to voluntary disclosure is the cost of capital (COC). For example, Zhou et al. (2017), and García-Sánchez and Noguera-Gámez (2017a) found positive market reactions to organizations that issued an integrated report through a reduction in the cost of capital. Other academics that investigated the value relevance of accounting information (Easley & O’Hara, 2004; Lambert et al., 2007; Apergis et al., 2011), corporate social responsibility information (Dhaliwal et al., 2011), or in general the level of voluntary disclosure (Diamond & Verrecchia, 1991; Botosan, 1997; Francis et al., 2008) also used cost of capital as a proxy for the market reaction.

Ownership concentration. Within organizations, ownership can be dispersed, when

many shareholders own a minority of the shares, or more concentrated, when a minority of the shareholders holds a substantial proportion of the shares. The level of ownership concentration is different among countries, as La Porta et al. (1998) state that countries operate within different legal systems (i.e. English-origin, French-origin, German-origin, and Scandinavian-origin). La Porta et al. (1998) prove this, as they found a strong negative correlation between ownership concentration and the legal protection of investors. Ownership concentration can also be a driver of information asymmetry, since it creates a second agency problem between controlling shareholders and minority shareholders (Fan & Wong, 2002; Barroso Casado et al., 2016).

Relevant theories

At the market level of IR, the main theories used in previous archival research are the stakeholder theory and legitimacy theory (Velte & Stawinoga, 2017). This paragraph explains why this study uses the agency theory, instead of the stakeholder theory and legitimacy theory.

(7)

7

Stakeholder theory focuses on meeting goals of a broad group of stakeholders, instead of only maximizing shareholder value (An et al., 2011). From this perspective, IR can be used to communicate about nonfinancial information combined with financial information, in order to manage accountability to, or meet the goals of, multiple stakeholders. According to the legitimacy theory, organizations operate under the requirements of a social contract with society (Velte & Stawinoga, 2017). From a legitimacy theory perspective, an organization could implement IR as a legitimating strategy, in order to meet the requirements of the ‘social contract’ (Fernando & Lawrence, 2014).

Although the stakeholder theory and legitimacy theory are widely used in IR research, they are not suitable to investigate the market reactions of shareholders to the extent of IR, but more to determine why organizations actually engage in IR. The purpose of this study is not to investigate why organizations engage in IR, but rather on what the consequences are of practising IR. Therefore, the agency theory looks more suitable to investigate the market response, because it is mainly focused on shareholders.

Agency theory. According to the agency theory, a firm is managed by a manager

(agent) who will make decisions to maximize his utility at the expense of the shareholders (principal) (Jensen & Meckling, 1976). This is partly a result of the information asymmetry between the manager, who relies on inside information, and the shareholders, who rely on publicly available information. Shareholders have to make monitoring costs in order to make sure that the manager is operating in the best interest of shareholders. Assuming they are risk averse, shareholders demand a higher return for their invested capital, which increases the cost of capital of the company (Hughes et al., 2007). Although new and ‘better’ information not always leads to a lower cost of capital (Johnstone, 2015), it is widely accepted that organizations can lower the cost of capital by strengthening the information environment (i.e. disclosure of more relevant information that can influence investment decisions) (Apergis et al., 2011; Dhaliwal et al., 2011; Easley et al., 2002; Easley & O’Hara, 2004; Elliot & Jacobson, 1994; Lambert et al., 2007; Leuz & Verrecchia, 2000). This can be achieved by providing more relevant information about the performance of the organization, both financial and nonfinancial (Healy & Palepu, 1993).

The theoretical structure of this paper, specifically the hypothesis development, continues in the following paragraph, building further on agency theory.

(8)

8

Hypotheses development

From an agency theory perspective, the implementation of IR can contribute to reducing information asymmetry between an organization and its capital market (Burke & Clark, 2016; García-Sánchez & Noguera-Gámez, 2017b). A reason for this reduction of information asymmetry is that IR reports are less voluminous than the combination of sustainability reports and annual reports, by connecting nonfinancial information with financial information, and only presenting information that is material. It is easier for the capital market to absorb and interpret information in an easy-to-read format, and as a result, lower costs have to be made with respect to the absorption of information. Previous research already demonstrated that the readability of a report (i.e. absorbability of the information) can influence the decision making process of shareholders (Hooghiemstra et al., 2017), the demand for information intermediaries (Lehavy et al., 2011), and the possibility to monitor management (Luo et al., 2018), which all affects the level of agency costs and in the end the cost of capital. Moreover, IR reports have a stronger focus on long term oriented information, which makes it more relevant for decision-making for investors. Slack and Tsalavoutas (2018) interviewed financial capital providers regarding the extent to which IR can contribute to decision-making. Although these results indicate scepticism, other studies demonstrate the opposite. For a sample of South African listed firms, Baboukardos and Rimmel (2016) and Bernardi and Stark (2018) provided evidence that an IR approach leads to more value relevant accounting information. In support of that claim, García-Sánchez & Noguera-Gámez (2017a) found that companies that adopting IR are rewarded with a lower cost of capital. The authors concluded that integrated information is value relevant to investors. Based on the aforementioned literature, it is hypothesized that the voluntary adoption of IR will reduce information asymmetry, agency costs, and in the end a firm’s cost of capital.

H1: The extent of integrated reporting is negatively related to a firm’s cost of capital.

To add a level of complexity to the relationship of IR and information asymmetry, Jensen and Berg (2012) showed that the perception of integrated information to be value relevant depends on the level of ownership concentration. Their study demonstrates that the willingness to adopt IR tend to be lower for companies with a higher degree of ownership concentration. Controlling shareholders benefit less from an improved information environment compared to minority shareholders, because they have potential access to private information, and therefore rely less on publicly available information (Fan & Wong, 2002; Jiang et al., 2011). In this case, an agency conflict arises between the controlling shareholder

(9)

9

and the minority shareholders, known as the second agency problem or principal-principal conflict (Barroso Casado et al., 2016; Dharwadkar et al., 2000). The controlling shareholder has an interest to withhold private information from the minority shareholders, which may lead to an increase in information asymmetry. This relationship is demonstrated by Fan and Wong (2002), who showed that high ownership concentration is associated with low earnings infomativeness, and Jiang et al. (2011), who found support for their claim that the level of ownership concentration is positively related to information asymmetry. Other evidence comes from France, where Zhao and Millet-Reyes (2007) investigated the relation between ownership structure and accounting information content. They argue that large shareholders have a “lack of incentive to report timely and relevant earnings to outside (minority) investors” (Zhao & Millet-Reyes, 2007, p. 1). The greater the information gap between controlling shareholders and minority shareholders, the higher costs of monitoring will be (Serafeim, 2015). Ownership concentration appears to be an important driver of agency costs (Morellec et al., 2018).

In the end, investors will require compensation for the estimation risk of private information, affecting a firm’s cost of capital (Diamond & Verrecchia, 1991; Easley & O’Hara, 2004). García-Sánchez & Noguera-Gámez (2017a) already claimed that firms with greater information asymmetries benefit the most of a reduction in the cost of capital, after the adoption of IR. Assuming greater information asymmetries are present in firms with higher ownership concentration (i.e. between large and small investors), it is hypothesized that the reduction in the cost of capital through IR, is moderated by the level of ownership concentration (i.e. the higher the level of ownership concentration, the greater the reduction in the cost of capital by IR).

H2: The relationship between the extent of integrated reporting and the cost of capital is positively influenced by the level of ownership concentration.

III. METHODOLOGY

This study contains archival research on panel data and focusses on stock listed companies during the time-frame 2012-2017. By using a world-wide sample, it is possible to control for national characteristics between different countries, like legal environment. The data with respect to the extent of IR was collected from the ASSET41 database, where professional analysts collected information about economic and ESG aspects. To collect data with respect

(10)

10

to ownership concentration and companies’ cost of capital, respectively the ORBIS2_and

I/B/E/S3 database was used. All control variables were extracted from Datastream.

Sample selection

When selecting countries for the sample, I choose to follow La Porta et al. (1998), who showed that ownership concentration varies by legal origin. With companies from Canada, India, South Africa, United Kingdom, and United States (English-origin), Brazil, France, Italy, Netherlands, and Spain (French-origin), Germany, Japan, and South-Korea (German-origin), and Denmark, Finland, Norway, and Sweden (Scandinavian-origin), I ensure that all legal systems are represented in the sample, an approach that corresponds with Santos et al. (2014). Table 1 shows how the observations are distributed among the selected countries and legal origins. Some countries (e.g. United States, Japan) are strongly represented in the sample, in contrast to others.4 However, other studies with samples of multiple countries show similar distributions among countries compared to my sample (Santos et al., 2014; García-Sánchez & Noguera-Gámez, 2017a), as those countries have a relatively larger amount of public listed companies. I started by selecting all public listed companies (72,505) in the ORBIS database, and deleted the data from not-selected countries (36,709). Consistent with prior research, I excluded all financial institutions (10,997), since those companies are attributed to other regulation (Gul & Leung, 2004). After that, I deleted all companies below 1 billion market capitalization (21,427), because data from large companies is better represented in the databases, which enabled me to investigate a larger sample. Eventually, I had 3,372 companies (20,232 firm-year observations) left. Merging the ORBIS data with I/B/E/S and Datastream remained me with a total sample of 5,341 firm-year observations. From all 17 countries within the sample, India did not meet the requirements of all control variables, which forced me to delete this country from the sample. An overview of the sample selection is presented in Table 2.

2_{Retrieved by Bureau van Dijk analysts, a Moody’s Analytics Company.} 3_{Providing analyst forecast information, part of WRDS.}

(11)

11

Table 1

Sampling distribution by country and legal origin.

Country N Percentage IR OWNCONCEN

Canada 912 4.51 57.05 30.82 India 1,236 6.11 72.79 45.03 South-Africa 264 1.30 82.43 49.41 United Kingdom 1,356 6.70 70.78 31.67 United States 8,334 41.19 39.22 33.14 English-origin 12,102 59.81 48.36 34.33 Brazil 486 2.40 68.63 57.82 France 708 3.50 83.79 61.25 Italy 294 1.45 73.26 56.64 Netherlands 354 1.75 72.55 44.00 Spain 288 1.42 80.93 46.69 French-origin 2,130 10.52 76.69 55.18 Germany 696 3.44 71.59 61.92 Japan 3,648 18.03 58.74 22.71 South-Korea 822 4.06 58.67 43.53 German-origin 5,166 25.53 60.61 30.12 Denmark 168 0.83 78.34 44.50 Finland 168 0.83 87.51 42.46 Norway 156 0.77 76.28 62.91 Sweden 342 1.69 78.99 47.00 Scandinavian-origin 834 4.12 80.73 48.78 Total 20,232 100.00 Table 2 Sample selection Amount

Total amount of public listed companies in ORBIS 72,505

Less: Companies from not-selected countries (36,709)

Less: Companies considered as financial institutions (10,997)

Less: Companies below 1 billion market capitalization (21,427)

Final amount of companies 3,372

Starting amount of firm-year observations (2012-2017) 20,232 Less: Incomplete data after merging ORBIS, Datastream, and I/B/E/S (14,891)

(12)

12

Dependent variable

The dependent variable in this research that measures the market reaction to IR is the cost of capital. I calculated the cost of capital by using Easton’s PEG-ratio (2004), which includes positive earnings forecasts per share from financial analysts, divided by the firm’s share price. In contrast to García-Sánchez and Noguera-Gámez (2017a), who use five- and four-year ahead forecasted earnings per share, I chose to use two- and one-year ahead forecasted earnings per share (Ashbaugh-Skaife et al., 2009), because of the following two reasons. First of all, five- and four-year ahead earnings forecasts include more forecast bias, as it is predicted with a longer horizon, leading to less accurate predictions. Secondly, data of two- and one-year ahead earnings forecasts was significantly better represented in the I/B/E/S database, and five- and four-year ahead forecasted earnings per share including the long-term-growth ratio, were poorly available. The calculation of the cost of capital is as follow:

r

PEG =

√

ⅇps₂−ⅇps₁

P₀

Test variables

The test variables in my study are the extent of integrated reporting (IR) and the level of ownership concentration (OWNCONCEN). To measure the extent of IR, I used the ‘Corporate Governance: Integration/vision and strategy’ (‘CGVS’) score from the ASSET4 database. This CGVS score ranges between 0 and 100 and is described as follow. “The overall level of internal integration measures a company’s management commitment and effectiveness toward the creation of an overarching vision and strategy integrating financial and non-financial aspects. It reflects a company’s capacity to convincingly show and communicate that it integrates the economic, social and environmental (ESG) dimensions into its day-to-day decision-making process.” (Maniora, 2017, p. 12). I have chosen to use the CGVS score, because of the following three reasons. First of all, both Maniora (2017) and Serafeim (2015) use it as a proxy to measure the extent of IR. Serafeim (2015) states that the ASSET4 data is collected by specially trained research analysts. Secondly, using the CGVS score allows me to investigate a large sample, which would not be the case if I hand collected the data and applied content analysis (García-Sánchez & Noguera-Gámez, 2017a; Zhou et al., 2017), or used an external source (Barth et al., 2015). According to De Villiers et al. (2017), content analysis is restricted to small samples and experiences coders that understand the objectives of IR, and external sources can only be used

(13)

13

for South African samples. Thirdly, it is practically not feasible to apply time consuming research methods, such as content analysis on a large sample, after hand collecting the data.

I followed La Porta et al. (1998) to determine the level of ownership concentration. I collected data from ORBIS about the three largest shareholders of each company, and computed the combined ownership stake of these shareholders. Various other researchers use La Porta’s method as well, when measuring the level of ownership concentration (e.g. Fan & Wong, 2002; Jensen & Berg, 2012).

Control variables

There are various other variables that could influence a firm’s cost of capital. To control for these effects, several control variables are included in the regression model. Most control variables are based on Dhaliwal et al. (2011) and García-Sánchez and Noguera-Gámez (2017a), because they also investigated the association between voluntary disclosure and the cost of capital. An overview of the variables is presented below (see also Table 3).

Fama and French (1992) found a positive relation between the cost of capital and firm leverage. A possible explanation is that companies that use external funding experience higher agency costs (García-Sánchez & Noguera-Gámez, 2017a). Therefore, leverage (LEV) is included as a control variable. Following Dhaliwal et al. (2006), leverage is computed by the ratio of total debt divided by total assets. Fama and French (1992) also found firm size to be negatively related with expected market returns. On the other hand, larger companies feel more outside pressure to disclose more information, and from an agency perspective, they need more external funding, leading to greater disclosure (García-Sánchez & Noguera-Gámez, 2017a). A control variable for firm size (FSIZE) is therefore added to the model, calculated by the logarithm of the total assets. In her study on disclosure level and the cost of equity capital, Botosan (1997) argues that for firms with low analyst following, greater disclosure is related to a lower cost of equity capital. Therefore, a control variable for the number of analyst following (AFOLLOW) is added as well. Other financial ratios that could influence the cost of capital are return on assets (ROA) and market-to-book ratio (MBR) (García-Sánchez & Noguera-Gámez, 2017a). De Villiers et al. (2017) endorse the importance of using control variables that reduce measurement issues related to the test variable IR. I follow up their advise by adding a control variable for ESG performance (PERF) and for ESG disclosure (CSR and GRI). As a last control variable, I include the firm’s beta (BETA) to control for systematic risk, thereby following Dhaliwal et al. (2011).

(14)

14

According to De Villiers et al. (2017) it is also important for IR studies to control for industry fixed-effects, since firms in different sectors are affected by environmental issues differently. Other researchers that investigate the market reaction to IR control for industry fixed-effects as well (Barth et al., 2017; García-Sánchez & Noguera-Gámez, 2017a). This causes me to include a control variable for industry (IND). An overview of the industry types within the sample is presented in Appendix A. Finally, because my study involves panel data over multiple countries, I also add dummy variables for countries (COUNTRY) and firm years (FYEAR).

Empirical model

At first sight, performing a Hausman test suggested that I should use a fixed-effects model in favour of a random-effects model, as the test delivered me significant results. However, because I want to include the effects of BETA, IND, and COUNTRY as well, I chose to use a random-effects model. This enables me to control for these additional random-effects, which will be impossible if I use a fixed-effects model, because the fixed-effects model omit the constant variables BETA, IND, and COUNTRY from the model.

COCi,t = β0 + β1IRi,t + β2OWNCONCENi,t + (β3IRi,t * OWNCONCENi,t ) +

β4LEVi,t + β5FSIZEi,t + β6AFOLLOWi,t + β7ROAi,t + β8MBRi,t +

β9PERFi,t + β10CSRi,t + β11GRIi,t + β12BETAi,t + β13INDi +

(15)

15

Table 3

Variable description

Measurement Database

Dependent variable

Cost of capital (COC) Calculated by the PEG ratio of Easton (2004) WRDS (I/B/E/S)

Test variables

Ownership concentration (OWNCONCEN)

Combined ownership stake of three largest shareholders of each company (La Porta et al., 1998)

ORBIS

Integrated reporting (IR) CGVS score ranging between 0 and 100 (CGVS*)

Datastream (ASSET4)

Control variables

Leverage (LEV) Calculated as total debt / total assets (WC08236*)

Datastream

Firm size (FSIZE) Calculated as the logarithm of total assets (WC02999*)

Datastream (Worldscope)

Analysts following (AFOLLOW) Number of analysts following the company WRDS (I/B/E/S) Return on assets (ROA) Calculated as net income / total assets

(WC08326*)

Datastream (Worldscope)

Market-to-book ratio (MBR) Calculated as market value equity / book value equity (MTBV*)

Datastream

ESG performance (PERF) Weighted average of environmental performance (ENVSCORE*), social performance (SOCSCORE*), corporate governance performance (CGVSCORE*)

Datastream (ASSET4)

Sustainability report (CSR) Dummy variable which equals 1 if the company issues a separate sustainability report, and 0 if not (CGVSDP026*)

Datastream (ASSET4)

GRI guidelines (GRI) Dummy variable which equals 1 if the company reports in accordance with GRI framework, and 0 if not (CGVSDP028*)

Datastream (ASSET4)

Firm risk (BETA) Beta for firm (WC09802*) Datastream (Worldscope)

Industry (IND) Industry type, determined by one-digit SIC code.

ORBIS

(16)

16

IV. EMPIRICAL RESULTS

Descriptive statistics

As presented in Table 4, the mean cost of capital (COC) is 0.097, indicating that the average cost of capital of a firm in the sample is 9.7 percent. The mean cost of capital of United States companies in the sample is 11.6% (41.19% of the observations). This corresponds to Dhaliwal et al. (2011), who reported a mean cost of capital of 11.9% for United States companies. Further on, the average ownership concentration (OWNCONCEN) of the sample is 35.6 percent. With a strong presence of English- and German-origin countries, this is similar to the results of La Porta et al. (1998). A further specification of these ownership concentration data is presented in Table 1. The last variable of interest (IR) shows a mean of 55.42, which signifies the average integrated reporting score of a company within the sample. A standard deviation of 32.48 indicates there is a high variation within the sample when it comes to the extent of integrated reporting, with the lowest score for United States (39.22), and the highest score for Finland (87.51) (see Table 1). A notable detail is that the average IR score for companies does not increase over the years, but in contrast does slightly decrease over the years (i.e. 2012: 57.67;

2013: 57.92; 2014: 57.89; 2015: 54.12; 2016: 54.10: 2017: 52.46). For one variable (GRI) the

ASSET4 database did not provide enough data, so this control variable was dropped from the model.

Table 5 presents the correlations between the different variables, in order to detect multicollinearity issues. All control variables are strongly correlated to the dependent variable (COC). Consistent with prior research (García-Sánchez & Noguera-Gámez, 2017a), there is no significant correlation between leverage (LEV) and the number of analysts (AFOLLOW) and market-to-book ratio (MBR). I calculated the variance inflation factor (VIF) of each variable to detect problems of multicollinearity. At first sight, there seem to be no occurrences of multicollinearity, because all VIF factors are less than 10. A VIF factor that is more than 10 indicates there are serious problems of multicollinearity. However, Table 5 shows that the variables PERF and CSR are strongly correlated (0.807 & 0.854) to the test variable IR, indicating a problem of multicollinearity. I therefore excluded variables PERF and CSR from the model.

(17)

17

Table 4

Descriptive statistics

Variable N Mean Std. Dev. Min Max

COCa _8,298 _0.097 _0.097 _0.002 _0.531 OWNCONCEN 17,019 0.356 0.226 0.000 1.000 IR 11,934 55.42 32.48 8.590 95.42 LEVa _16,993 _0.282 _0.187 _0.007 _0.912 FSIZE 19,484 7.107 1.227 0.602 11.47 AFOLLOW 13,777 13.47 8.331 1.000 54.50 ROAa _18,994 _4.899 _11.20 _-55.79 _32.72 MBRa _18,180 _3.526 _5.457 _-14.85 _36.81 PERF 11,988 57.19 24.41 4.62 96.51

CSR Dummy Dummy Dummy Dummy Dummy

GRIb _Dropped _Dropped _Dropped _Dropped _Dropped

BETAa _17,808 _1.05 _0.71 _-0.34 _3.63

a_{Variables are winsorized at the 1}st_{and 99}th_{percentile, using mean +/- three times standard deviation.}

b_{Variable GRI was excluded from the sample, because the data in ASSET4 does not meet the requirements to} run a large sample.

(18)

Table 5

Pearson correlation matrix

Variable 1 2 3 4 5 6 7 8 9 10 11 12 1 COC 1.000 2 IR -0.042*** _1.000 3 OWNCONCEN 0.072*** _0.060*** _1.000 4 IR * OWNCONCEN 0.022* _0.629*** _0.729*** _1.000 5 LEV 0.091*** _-0.059*** _0.028*** _-0.019* _1.000 6 FSIZE -0.403*** _0.354*** _-0.119*** _0.148*** _-0.066*** _1.000 7 AFOLLOW -0.032*** _0.345*** _0.067*** _0.264*** _-0.012 _0.324*** _1.000 8 ROA -0.311*** _0.057*** _-0.005 _0.039*** _-0.110*** _0.146*** _0.093*** _1.000 9 MBR -0.066*** _-0.080*** _0.030*** _-0.050*** _0.006 _-0.177*** _0.027*** _0.087*** _1.000 10 PERFa _0.074*** _0.807*** _0.020** _0.487*** _-0.008 _0.101*** _0.410*** _0.108*** _-0.031*** _1.000 11 CSRa _-0.087*** _0.854*** _0.026*** _0.523*** _-0.085*** _0.364*** _0.278*** _0.085*** _-0.080*** _0.673*** _1.000 12 BETA 0.244*** _-0.113*** _0.006 _-0.078*** _0.060*** _-0.178*** _-0.024*** _-0.264*** _0.049*** _-0.054*** _-0.128*** _1.000 ***_,**_{, and}* _{coefficients are significant at a 1, 5, and 10 percent level respectively.}

Variable GRI was excluded from the sample, because the data in ASSET4 does not meet the requirements to run a large sample. a _{Variables PERF and CSR are highly correlated ( > 0.7 ) with IR, indicating multicollinearity problems.}

(19)

19

Basic analysis

The first hypothesis, guided by the agency theory, stated that IR is expected to reduce information asymmetry between the company and its shareholders, thereby reducing agency costs, and as a result the firm’s cost of capital. Hence, it was hypothesized that a negative relationship exists between the extent of IR and cost of capital. Table 6 shows that the IR coefficient of the first hypothesis is negative (-0.000), indicating a negative relationship exists between the extent of IR and the cost of capital. Firms in this dataset practising IR are rewarded with a reduction in the cost of capital. However, the p-value (0.524) proves this relationship is insignificant, and therefore no conclusions can be drawn from this result, rejecting the first hypothesis.

The second expectation was that ownership concentration would moderate the relation between IR and the cost of capital, explained by the idea that ownership concentration influences the level of information asymmetry within a company, creating a second agency problem (i.e. the higher the ownership concentration, the higher the information asymmetry). Hence, it was hypothesized that the level of ownership concentration would positively influence the relationship between IR and cost of capital. As further can be observed from the table, the market reaction to IR is stronger for companies that are characterized by higher levels of ownership concentration, since the interaction term of the second hypothesis is negative (-0.001) and highly significant at a 1 percent level (0.003). Hence, the second hypothesis is accepted, claiming that the level of ownership concentration positively influences the relationship between IR and cost of capital. The results show that in this sample ownership concentration increases the cost of capital, but that IR mitigates this effect, indicating that the market response to IR is stronger for firms with higher levels of ownership concentration, explained by the phenomenon that these firms experience greater information asymmetries.

Summarizing, the first tested relationship proves to be negative but insignificant, and the second hypothesis is negative and highly significant at a 1 percent level. Running both hypotheses in a simple OLS regression delivers me the same results (see Appendix B). The following paragraphs provide additional analyses that should add more robustness to the results of the basic analysis. First, a fixed-effects model is performed, in order to control for endogeneity. Second, an alternative measure is calculated for the cost of capital, using realized returns in a random-effects model. In the last additional test, the results of all countries are analyzed and compared, running country dummies in a simple OLS regression.

(20)

20

Table 6

Results of random-effects regression model

Hypothesis 1 Hypothesis 2

Coefficient t Coefficient t

Variable (Std. Err.) (p-value) (Std. Err.) (p-value)

IR -0.000 (0.000) -0.350 (0.726) 0.000 (0.000) 2.610 (0.009)*** OWNCONCEN 0.049 (0.012) 3.670 (0.000)*** IR * OWNCONCEN -0.001 (0.000) -3.730 (0.000)*** LEV 0.021 (0.007) 3.010 (0.003)*** 0.024 (0.007) 3.290 (0.001)*** FSIZE -0.007 (0.003) -2.580 (0.010)** -0.009 (0.003) -3.130 (0.002)*** AFOLLOW 0.001 (0.000) 4.130 (0.000)*** 0.001 (0.000) 4.550 (0.000)*** ROA -0.002 (0.000) -18.460 (0.000)*** -0.002 (0.000) -17.660 (0.000)*** MBR -0.002 (0.000) -8.490 (0.000)*** -0.002 (0.000) -7.620 (0.000)*** BETA 0.021 (0.002) 11.480 (0.000)*** 0.020 (0.002) 10.170 (0.000)*** Constant 0.110 (0.021) 5.230 (0.000)*** 0.113 (0.023) 4.980 (0.000)***

GRIa _Dropped _Dropped _Dropped _Dropped

CSRb _Dropped _Dropped _Dropped _Dropped

PERFb _Dropped _Dropped _Dropped _Dropped

R-squared 0.363 0.376

Observations (N) 5,802 5,341

Industry fixed-effects Yes Yes

Country fixed-effects Yes Yes

Year fixed-effects Yes Yes

***_,**_{, and}*_{coefficients are significant at a 1, 5, and 10 percent level respectively.}

a_{Variable GRI was excluded from the sample, because the data in ASSET4 does not meet the requirements to} run a large sample.

(21)

21

Robustness test with fixed-effects model

As described previously, a random-effects model was chosen in favour of the fixed-effects model, because it enables to include effects for industry and country in the regression. One disadvantage of the random-effects model compared to the fixed-effects model is that it does not control for endogeneity (Nikolaev & Van Lent, 2005). Endogeneity occurs when variables that are not observable in the model are correlated with explanatory variables that are included in the model. Because the fixed-effects model does partially control for endogeneity and the Hausman test turned out to be significant, I performed an additional robustness test with a fixed-effects regression.

Table 7 shows that the IR coefficient of the first hypothesis is barely changed (0.000 compared to -0.000), and still insignificant (0.962), and the interaction term of the second hypothesis is still negative (-0.001) and highly significant at a 1 percent level (0.002). This robustness test indicates that the findings of the basic analysis are robust to the use of a fixed-effects model, since the test shows similar results as the basic analysis. However, the results should be interpreted with caution, because this robustness test does not include industry- and country fixed-effects. Moreover, this robustness test has a significantly lower r-squared compared to the basic analysis, indicating that less variance of the dependent variable can be explained by the independent variables.

(22)

22

Table 7

Robustness test with fixed-effects model

IR 0.000 (0.000) 0.050 (0.962) 0.000 (0.000) 1.050 (0.292) OWNCONCEN 0.042 (0.018) 2.400 (0.016)** IR * OWNCONCEN -0.001 (0.000) -3.130 (0.002)*** LEV 0.082 (0.017) 4.760 (0.000)*** 0.093 (0.017) 5.340 (0.000)*** FSIZE -0.012 (0.012) -1.010 (0.313)*** -0.016 (0.013) -1.280 (0.200) AFOLLOW 0.002 (0.000) 3.490 (0.000)*** 0.001 (0.000) 2.580 (0.010)** ROA -0.002 (0.000) -10.640 (0.000)*** -0.002 (0.000) -10.350 (0.000)*** MBR -0.001 (0.000) -2.580 (0.010)** -0.001 (0.000) -2.110 (0.035)* Constant 0.138 (0.090) 1.540 (0.124) 0.159 (0.092) 1.720 (0.085)*

R-squared 0.200 0.213

Firm fixed-effects Yes Yes

(23)

23

Alternative measures cost of capital (COC)

There are alternative methods to measure the dependent variable of this study, the cost of capital. Other used measures are the CT model by Claus and Thomas, the GLS model by Gebhardt et al., the Finite Horizon model developed by Gordon and Gordon, and the Ohlson and Juettner-Nauroth model (Guay et al., 2011). As mentioned in the previous chapter, the elements forecasted earnings per share five- and four-year-ahead are poorly available in I/B/E/S, as well as the long-term growth rate. Since these alternative models all include one of these elements, it is not possible to perform one these models.

However, alternative ways to measure the dependent variable are not limited to the previous mentioned methods. Different academics use realized returns as a sensitivity test for the cost of capital (Francis et al., 2008; Guay et al., 2011; Zhou et al., 2017). As described by Zhou et al. (2017), one advantage of using realized returns is that it has fewer measurement errors than the implied cost of capital. Moreover, realized returns are much more available in the databases, which enables me to investigate a large sample with this measure. An additional test with realized returns is therefore performed, in order to test the sensitivity of the relationships in the random-effects model. The alternative proxy is calculated by the realized earnings per share divided by the firm’s annual share price. Data for realized returns and share price were retrieved from ORBIS and ASSET4 respectively. In accordance with the main analysis, a random-effects regression model was performed.

Table 8 shows that the results of the sensitivity test are similar to those of the basic analysis, meaning that both relationships remain negative. However, both hypotheses are insignificant, indicating that conclusions cannot be drawn from the results. Compared to the main analysis, this analysis uses a considerable larger sample, and a lower r-squared, which could possibly explain the differences.5_{But overall, the results of the sensitivity test are}

consistent with the results of the main analysis.

(24)

24

Table 8

Results of sensitivity test

IR -0.000 (0.000) -0.480 (0.634) -0.000 (0.000) 0.630 (0.530) OWNCONCEN 0.001 (0.003) 0.460 (0.644) IR * OWNCONCEN -0.000 (0.000) -1.020 (0.308) LEV 0.010 (0.002) 5.930 (0.000)*** 0.010 (0.002) 5.810 (0.000)*** FSIZE 0.001 (0.001) 1.790 (0.073)* -0.001 (0.001) 0.870 (0.384) AFOLLOW 0.000 (0.000) 2.400 (0.016)** 0.000 (0.000) 3.250 (0.001)*** ROA -0.001 (0.000) -22.690 (0.000)*** -0.001 (0.000) -21.490 (0.000)*** MBR -0.000 (0.000) -3.860 (0.000)*** -0.000 (0.000) -3.930 (0.000)*** BETA 0.004 (0.000) 7.210 (0.000)*** 0.003 (0.000) 6.540 (0.000)*** Constant -0.005 (0.005) -0.890 (0.372) -0.001 (0.006) -0.110 (0.910)

R-squared 0.141 0.139

Country fixed-effects Yes Yes

(25)

25

Country analysis

As mentioned in the introduction of this paper, one important advantage of this study is that it contains a worldwide sample, making it possible to compare differences among countries. In this country analysis, I compared the results of all countries within the sample, using a simple OLS regression with country dummies. Since Finland has the highest extent of IR (87.51) among all countries within the sample, I chose for the country analysis on the first hypothesis to compare the results of all other countries against the result of Finland. Consistent with La Porta’s (1998) analysis of corporate governance characteristics among countries, Japan scores lowest on ownership concentration (22.71%). This made me decide to compare all other countries against the result of Japan, for the country analysis on the second hypothesis.

Concerning the first hypothesis, it was expected that there exist a negative relation between the extent of IR and the cost of capital. Since Finland has the highest score on IR within the sample, I expect that all other countries show a weaker effect on the cost of capital, because those countries have a lower extent of IR. Table 9a and 9b show that the coefficient of all countries, except Italy, for the first hypothesis is negative, indicating a weaker relation between IR and cost of capital, compared to Finland. This suggests that the higher the extent of IR, the higher the reduction in the cost of capital. The results of this analysis are significant for 10 out of the 16 countries. Only for Canada, Italy, Netherlands, Spain, and Germany, the coefficient is insignificant, so no conclusions can be drawn from those countries.

Regarding the second hypothesis, it was predicted that ownership concentration would positively influence the relationship between IR and the cost of capital. Because Japan has the lowest level of ownership concentration, I expect that all other countries show a stronger moderating effect, as those countries have higher levels of ownership concentration. Table 9a and 9b confirm this expectation, since all countries show a positive coefficient, compared to Japan. With all values significant at a 1 percent level, further support is provided for hypothesis 2, claiming that ownership concentration positively influences the relation between IR and the cost of capital.

(26)

26

Table 9a

Results of country analysis (English- and French-origin countries)

Hypothesis 1

(IR)

Hypothesis 2

(IR*OWNCONCEN)

(Std. Err.) (p-value) (Std. Err.) (p-value)

English-origin: Canada -0.022 (0.014) -1.560 (0.119) 0.106 (0.007) 14.160 (0.000)*** South-Africa -0.082 (0.014) -5.840 (0.000)*** 0.048 (0.007) 6.880 (0.000)*** United Kingdom -0.054 (0.013) -4.030 (0.000)*** 0.073 (0.008) 8.900 (0.000)*** United States -0.055 (0.013) -4.220 (0.000)*** 0.071 (0.006) 12.520 (0.000)*** French-origin: Brazil -0.025 (0.015) -1.700 (0.089)* 0.098 (0.010) 9.940 (0.000)*** France -0.037 (0.014) -2.690 (0.007)*** 0.095 (0.008) 11.290 (0.000)*** Italy 0.013 (0.019) 0.730 (0.467) 0.146 (0.017) 8.710 (0.000)*** Netherlands -0.012 (0.015) -0.780 (0.435) 0.100 (0.010) 10.450 (0.000)*** Spain -0.008 (0.016) -0.500 (0.620) 0.108 (0.012) 8.780 (0.000)***

***_,**_{, and}*_{coefficients are significant at a 1, 5, and 10 percent level respectively.} a_{Starting point for hypothesis 1, since Finland has the highest extent of IR (87.51).}

(27)

27

Table 9b

Results of country analysis (German- and Scandinavian-origin countries)

Hypothesis 1

(IR)

Hypothesis 2

(IR*OWNCONCEN)

(Std. Err.) (p-value) (Std. Err.) (p-value)

German-origin: Germany -0.023 (0.014) -1.630 (0.103) 0.100 (0.010) 10.540 (0.000)*** Japanb _-0.134 (0.014) -9.240 (0.000)*** - - South-Korea -0.145 (0.015) -9.400 (0.000)*** -0.012 (0.003) -3.720 (0.000)*** Scandinavian-origin: Denmark -0.071 (0.019) -3.800 (0.000)*** 0.066 (0.018) 3.690 (0.000)*** Finlanda _- _- _0.127 (0.015) 8.670 (0.000)*** Norway -0.071 (0.017) -4.070 (0.000)*** 0.054 (0.013) 4.100 (0.000)*** Sweden -0.104 (0.013) -7.700 (0.000)*** 0.023 (0.006) 3.910 (0.000)***

***_,**_{, and}*_{coefficients are significant at a 1, 5, and 10 percent level respectively.} a_{Starting point for hypothesis 1, since Finland has the highest extent of IR (87.51).}

(28)

28

V. DISCUSSION & CONCLUSION

The purpose of this study is to investigate the capital market’s reaction to the extent of IR, measured by the cost of capital and CGVS score. Substantiated by the agency theory, IR could help in reducing the information asymmetry between the firm and its shareholders. This effect could be stronger for firms with high levels of ownership concentration, since ownership concentration can influence the level of information asymmetry. The following research question is therefore relevant: “How does the capital market react to the extent of IR, and what is the moderating role of ownership concentration?”

Findings

To get an answer on the research question, two different hypotheses were developed. I firstly hypothesized that the extent of IR is negatively related to a firm’s cost of capital. This means that firms with a higher extent of IR are rewarded with a reduction in cost of capital. Secondly, I hypothesized that the level of ownership concentration moderates this relationship, meaning that the higher the level of ownership concentration, the higher the reduction in the cost of capital. Both hypotheses were tested, using a random-effects regression model.

The first random-effects regression suggests that IR is associated with a reduction in the cost of capital (i.e. the higher the extent of IR, the lower the cost of capital). However, since the results are insignificant, it is impossible to draw conclusions from it. The results do not correspond with García-Sánchez & Noguera-Gámez (2017a), who found a significant negative relation, possibly explained by the alternative proxy that is used in my study to measure the extent of IR. Where I used the CGVS score to measure the extent of IR, García-Sánchez & Noguera-Gámez (2017a) used dummies to indicate the adoption of IR. A second possible explanation is the use of two- and one-year ahead forecasted earnings per share, instead of five- and four-year ahead forecasted earnings per share, to measure cost of capital. A third explanation, is the use of a later timeframe (2012-2017) in my study, compared to García-Sánchez & Noguera-Gámez (2017a) (2009-2013). The early mover effect, which means that the capital market reacts stronger to companies that are first movers in adopting IR, could occur less in the later timeframe, since the market reaction to IR can diminish over time (Arguelles et al., 2015).

The second random-effects regression shows a negative coefficient for the moderator, and on top of that, it appears to be significant at a 1 percent level. Given this significance level, it is safe enough to conclude that in this sample the level of ownership concentration moderates

(29)

29

the relationship between IR and cost of capital. The results suggest that IR is associated with a reduction in the cost of capital, and that this effect is higher for firms with higher levels of ownership concentration. Moreover, an additional analysis specified on all countries within the sample shows that the moderating effect is stronger for countries that are characterized by high ownership concentration, providing more support for the moderating effect. This result is in line with prior research, claiming that firms with greater information asymmetries benefit the most with a reduction in the cost of capital (García-Sánchez & Noguera-Gámez, 2017a), assuming that firms with higher levels of ownership concentration experience greater information asymmetries (e.g. Fan & Wong, 2002; Zhao & Milles-Reyes, 2007; Jiang et al., 2011).

The results of this study remained the same after performing a robustness check with a fixed-effects models, and using an alternative measure for calculating the cost of capital, by running a larger sample with realized returns.

Implications

There are several theoretical implications in this study. Firstly, this study contributes to the literature by using a worldwide sample, making it possible to compare differences between countries, which is recommended by different researchers (e.g. de Villiers et al., 2017; Vaz et al., 2016; Velte & Stawinoga, 2017). As already mentioned in the introduction of this paper, current research on IR is focusing mostly on stock listed firms in South Africa (Velte & Stawinoga, 2017). The results of those studies are not directly generalizable to other countries, as other countries may operate under different institutional settings. Another advantage compared to other studies (Barth et al., 2017; Zhou et al., 2017), is the inclusion of companies that operate in a voluntary setting.

Secondly, this study contributes to the literature by involving the level of ownership concentration in the analysis, which is recommended by García-Sánchez & Noguera-Gámez (2017a). As far as I know, no other studies that investigated the market reaction to IR use the effect of ownership concentration. The current study complements this literature gab by using ownership concentration as a possible moderator in the market reaction to IR. The study shows that the market reaction to IR is stronger for companies that are characterized by higher levels of ownership concentration, and subsequently, greater information asymmetries. This study shows that organizations can profit from this finding, by implementing IR to lower the information asymmetry, as they are eventually rewarded with a reduction in the cost of capital.

(30)

30

A third important advantage of this study is that it investigates a relatively large sample of firm-year observations, compared to other studies that investigate the market reaction on IR (Barth et al., 2017; García-Sánchez & Noguera-Gámez, 2017a; Zhou et al., 2017). Generally, the greater the sample size, the more precise unknown parameters can be estimated.

Limitations and future research

Some limitations were identified in this study that potentially could have biased the results. These limitations are discussed below to serve as recommendations for future studies investigating market reactions to IR.

Firstly, it can be stated that the CGVS score that is used in this study to measure the extent of IR has some limitations, as it also measures the extent of integrated thinking within a company. However, because IR has not been applied for a long period of time yet, proxies to measure its level are scarce. De Villiers et al. (2017) acknowledge this issue, but at the same time claim that the CGVS score is a sufficed measure, because no other measures are available that can be applied on databases. Moreover, it enables the researcher to investigate a larger sample of companies, since alternative measures are restricted to content analysis. Future studies could attempt new proxies to measure the extent of IR.

Secondly, given the limited amount of South African companies, this study cannot exactly explore the differences between a mandatory setting and a voluntary setting of IR. A major reason is that for this study a sample was chosen of companies above 1 billion market capitalization, which are not abundant in South Africa. Future studies could compare the differences between mandatory and voluntary settings, using a more balanced sample with smaller companies, including a larger amount of South African companies.

Thirdly, the timeframe of this study does not allow to identify possibly early mover effects, because the period is scattered over the years 2012 to 2017. Although this period makes it able to compare differences over time, it does not include the early adaptor period of IR. Future research could focus on investigating a broader timeframe, in order to identify possibly early mover effects. With this information, studies over different timeframes can be compared with each other.

Lastly, this study does not take into account the impact of assurance on integrated reports. With the increasing adoption of IR, firms also tend to seek opportunities to get assurance on their integrated reports by an external auditor. At this moment of writing, assurance on integrated reports is limited. However, Atkins and Maroun (2015) argue that the need for this IR assurance is really there, after investigating the opinions of institutional

(31)

31

investors. Future research could include the effects of assurance on integrated reports, when investigating the market reaction to IR.

Conclusion

Answering the research question on how the capital market reacts to IR, this study shows that IR tends to be negatively associated with cost of capital, suggesting that an increase in the extent of IR can result in a reduction of the cost of capital. Moreover, this study indicates that ownership concentration positively moderates this association, where this effect tends to be stronger for firms and countries characterized by higher levels of ownership concentration. Firms can use this information in managerial decisions on how to lower the cost of capital. This information is especially relevant for organizations with high levels of ownership concentration, since those companies have greater information asymmetries, and as a result higher costs of capital.

(32)

32

REFERENCES

Accounting for Sustainability, Global Reporting Initiative. (2012). The value of extra-financial disclosure. What investors and analysts said. Radley Yeldar. London.

An, Y., Davey, H., & Eggleton, I.R.C. (2011). Towards a comprehensive theoretical

framework for voluntary IC disclosure. Journal of Intellectual Capital, 4(12), 571-585.

Apergis, N., Artikis, G., Eleftheriou, S., & Sorros, J. (2011). Accounting information and cost of capital: A theoretical approach. Modern Economy, 2(4), 589-596.

Arguelles, M.P.M., Balabat, M., & Green, W. (2015). Is there an early-mover market value effect for signalling adoption of integrated reporting? Working paper: New South Wales.

Arping, S., & Sautner, Z. (2013). Did SOX section 404 make firms less opaque? Evidence from cross-listed firms. Contemporary Accounting Research, 30(3), 1133-1165.

Ashbaugh-Skaife, H., Collins, D.W., Kinney, W.R., & Lafond, R. (2009). The effect of SOX internal control deficiencies on firm risk and cost of equity. Journal of Accounting

Research, 47(1), 1-43.

Atkins, J., & Maroun, W. (2015). Integrated reporting in South Africa in 2012: perspectives from South African institutional investors. Meditari Accountancy Research, 23, 197 -221.

Babourkados, D. (2018). The valuation relevance of environmental performance revisited: The moderating role of environmental provisions. The British Accounting Review, 50(1), 32-47.

Baboukardos, D., & Rimmel, G. (2016). Value relevance of accounting information under an integrated reporting approach: A research note. Journal of Accounting and Public

Policy, 35(4), 437-452.

Barroso Casado, R., Burkert, M., Dávila, A., & Oyon, D. (2016). Shareholder protection: The role of multiple large shareholders. Corporate Governance: An International Overview,

24(2), 105—129.

Barth, M.E., Cahan, S.F., Chen, L., & Venter, E.R. (2017). The economic consequences associated with integrated report quality: Capital market and real effects. Accounting,

(33)

33

Bernardi, C., & Stark, A.W. (2018). Environmental, social and governance disclosure, integrated reporting, and the accuracy of analyst forecasts. The British Accounting

Review, 50(1), 16-31.

Berthelot, S., Coulmont, M., & Serret, V. (2012). Do investors value sustainability reports? A Canadian study. Corporate Social Responsibility and Environmental Management,

19(6), 355-363.

Botosan, C.A. (1997). Disclosure level and the cost of equity capital. The Accounting Review,

72(3), 323-349.

Burke, J.J., & Clarke, C.E. (2016). The business case for integrated reporting: Insights from leading practitioners, regulators, and academics. Business Horizons, 59(3), 273-283. Cheng, M., Green, W., Conradie, P., Konishi, N., & Romi, A. (2014). The international

integrated reporting framework: Key issues and future research opportunities. Journal

of International Financial Management and Accounting, 25(1), 90-119.

Cho, S.Y., Lee, C., & Pfeiffer, R.J. (2013). Corporate social responsibility performance and information asymmetry. Journal of Accounting and Public Policy, 32(1), 71-83. De Villiers, C., Venter, E.R., & Hsiao, P.K. (2017). Integrated reporting: Background,

measurement issues, approaches and an agenda for future research. Accounting &

Finance, 57(4), 937-959.

Dhaliwal, D.S., Heitzman, S., & Li, O.Z. (2006). Taxes, leverage, and the cost of equity capital.

Journal of Accounting Research, 44(4), 691-723.

Dhaliwal, D.S., Zhen Li, O., Tsang, A., & Yang, Y.G. (2011). Voluntary nonfinancial disclosure and the cost of equity capital: The initiation of corporate social responsibility reporting. The Accounting Review, 86(1), 59-100.

Dharwadkar, R., George, G., & Brandes, P. (2000). Privatization in emerging economies: An agency theory perspective. The Academy of Management Review, 25(3), 650-669. Diamond, D.W., & Verrecchia, R.E. (1991). Disclosure, liquidity, and the cost of capital. The

Journal of Finance, 46(4), 1325-1359.

Easley, D., Hvidkjaer, S., & O’Hara, M. (2002). Is information risk a determinant of asset returns? The Journal of Finance, 57(5), 2185-2221.

Easley, D., & O’Hara, M. (2004). Information and the cost of capital. The Journal of Finance,

59(4), 1553-1583.

Easton, P.D. (2004). PE ratios, PEG ratios, and estimating the implied expected rate of return on equity capital. The Accounting Review, 79(1), 73-95.