The link between sustainability practices and the cost of debt

(1)

The Link Between Sustainability Practices and the Cost of Debt

Name: Bing Borawitz

Student number: 10784756

Thesis supervisor: Dr. Alexandros Sikalidis Date: June 25, 2018

Word count: 13,623

MSc Accountancy & Control, specialization Control

(2)

2

Statement of Originality

This document is written by student Bing Borawitz 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)

3

Abstract

This research examines the link between the sustainability practices of companies and the cost of debt. Specifically, this research examines whether more enhanced sustainability practices are associated with a lower average cost of debt, a lower short-term cost of debt, and a lower long-term cost of debt. According to the legitimacy theory and the stakeholder theory, enhanced sustainability practices should lead to a licence to operate from society. This licence to operate enhances future performance, which leads to a lower default risk. This lower default risk should lead to a lower cost of debt. The results of this quantitative study indeed suggest that more enhanced sustainability practices lower the average cost of debt, the short-term cost of debt, and the long-term cost of debt. The findings of this study have implications for the development of Integrated Reporting and provide new insights in the limited available fragmented literature. Moreover, the findings suggest how companies can lower their cost of debt.

(4)

4

1. Introduction

This research examines the relationship between the sustainability practices of companies and their cost of debt. Specifically, this research examines whether companies with more enhanced sustainability practices have a lower cost of debt. The link between the sustainability practices of companies and the financial performance of companies is extensively researched in the past (Goss & Roberts, 2011). However, the literature review of Renneboog, Ter Horst, and Zhang (2008) found that the relationship between sustainability practices and specifically the cost of debt is not comprehensively researched. On contrary, the link between sustainability practices and the cost of equity capital is more extendedly researched (Dhaliwal, Tsang, & Yang, 2011, El Ghoul, Guedhami, Kwok, & Mishra, 2011).

Goss and Roberts (2011) examined as one of the few the relationship between the sustainability practices of companies and the cost of debt. They found that companies with below average sustainability practices, have a higher cost of debt than companies with average or above average sustainability practices. More recently, Hoepner, Oikonomou, Scholtens, and Schröder (2016) also researched the link between sustainability practices and the cost of debt. Contrary to the findings of Goss and Roberts, they found that the sustainability practices of firms had no influence on the cost of debt. So, the limited available literature is fragmented. Therefore, it is interesting to re-examine the link between sustainability and the cost of debt in this thesis. Also, the research of Goss and Roberts (2011) and Hoepner et al. (2016) examined the effect of sustainability on the cost of debt of only bank loans. To get a more comprehensive view of the overall effect of sustainability practices, this thesis examines the effect of sustainability practices on the aggregated cost of debt. The aggregated cost of debt is the average cost of debt of all types of debt together.

So, this thesis first examines the effect of sustainability practices on the average cost of debt. Then, a distinction is made between short-term debt and long-term debt, and the effect of sustainability practices on the short-term cost of debt and the long-term cost of debt is separately examined. I expect that more enhanced sustainability practices reduce the average cost debt, the long-term cost of debt, and the short-term cost of debt. The sustainability practices of companies are measured using the overall sustainability score. This overall sustainability score is available through the MSCI KLD STATS database. The cost of debt is defined as the interest costs divided by the average debt during the year. The interest costs

(6)

6

and average debt are respectively available through Compustat – Capital IQ or can be calculated with Compustat – Capital IQ data.

After winsorizing and trimming the samples, 11,545 observations are left for the average cost of debt sample, 1,422 observations are left for the short-term cost of debt sample, and 987 observations are left for the long-term cost of debt sample. All observations are from 2003 till 2013. Using Ordinary Least Square (OLS) regression, I found that more enhanced sustainability practices reduce the average cost of debt of companies. More specifically, more enhanced sustainability practices reduce both the short-term cost of debt and the long-term cost of debt.

This research makes several contributions. First, as earlier mentioned, this research is extending the prior literature by examining the effect on the average cost of debt instead of only examining the effect on the cost of debt of bank loans. Second, as also earlier mentioned, this study is giving some clarity in the fragmented literature. Third, the findings of this thesis say something about the opinion of debtholders towards the sustainability practices of companies. The opinions of the debtholders have important implications for the development of Integrated Reporting and say something about the importance of Integrated Reporting. Because this research finds that more enhanced sustainability practices are associated with a lower cost of debt, it can be concluded that debtholders do care about sustainability. Therefore, the development of Integrated Reporting could be seen as important and useful, because debtholders will use the Integrated Reports to assess the sustainability practices of companies. Fourth, the results of this research are relevant for companies, because they now know how they can reduce their cost of debt, and so their profitability.

The remainder of this thesis is structured as follows. Chapter 2 provides a literature review. The hypotheses are discussed in chapter 3. Chapter 4 discusses the sample and model used to test the hypotheses. In chapter 5 the results are discussed. Finally, chapter 6 provides the conclusion of this research.

(7)

7

2. Literature Review

2.1 The Concept of Sustainability

In 1987, Brundtland and Khalid published The Brundtland Report, also known as Our Common

Future. This report defined corporate sustainability as a way of operating that meets the needs

of present generations without compromising the ability of future generations to meet their own needs. The Brundtland Report was published because of the increasing public environmental awareness. The public environmental awareness was a result of several nuclear disasters, for example the Chernobyl disaster. Corporate sustainability can be categorized in economic sustainability, environmental sustainability and social sustainability (Dyllick & Hockerts, 2002). Economic sustainability focusses for example on profit and cost savings, environmental sustainability focusses for example on pollution preventions and natural resources use, and social sustainability focusses for example on education and standards of living. To achieve overall sustainability, it is important that companies focus on all three categories of sustainability (Gladwin, Kennelly, & Krause, 1995).

Closely related to the concept of corporate sustainability is the concept of Corporate Social Responsibility (CSR). The most used definition of CSR is the definition of the Commission of the European Communities (Dahlsrud, 2008). They define CSR as “a concept whereby

companies integrate social and environmental concerns in their business operations and in their interaction with their stakeholders on a voluntary basis” (Commission of the European

Communities, 2001, p. 6). CSR is seen as a process, that balances the economic, environmental and social activities of the company, which ultimately leads to the accomplishment of corporate sustainability. Today, the terms corporate sustainability and Corporate Social Responsibility are seen as synonyms (Van Marrewijk, 2003). Therefore, I will use the terms corporate sustainability and CSR also interchangeably in this thesis.

Recently, there was a growing worldwide academic interest in CSR. As a result of this, there is a lot of CSR literature available nowadays. But the CSR literature is very fragmented (Aguinis & Glavas, 2012). First of all, this is because CSR is studied through different perspectives. For example, the effect of CSR on financial performance, on human resource management, and on operations is researched through respectively a financial, a human resources, and an operational perspective. So CSR is researched in various disciplines. Secondly, the CSR researches all have a different level of analysis (Aguinis & Glavas, 2012).

(8)

8

This means that the effects of CSR on certain disciplines are researched at for example an institutional, organizational, or individual level.

2.2 The Concept of Sustainability Reporting

Companies are enforced by law to report their financial position every accounting period. Since 2005, listed companies in many countries around the world have been mandated to report their financial statements in line with the International Financial Reporting Standards (Daske, Hail, Leuz, & Verdi, 2008). These International Financial Reporting Standards (IFRS) contain principles which guide how the balance sheet, the profit or loss statement, the statement of changes in equity, and the statement of cash flows should be reported. The standard setting body, the International Accounting Standard Board, expects that the worldwide application of IFRS enhances for instance the comparability of financial statements of different companies (Daske et al., 2008).

However, there are not yet mandatory accounting standards that describe how companies should report about their sustainability practices. But there are voluntary applicable guidelines that prescribe how companies should report about their CSR practices. Since around 2000, the guidelines of the Global Reporting Initiative (GRI) have been seen as the most developed guidelines regarding sustainability reporting, and these voluntary guidelines are therefore mostly applied nowadays (Brown, De Jong, & Lessidrenska, 2009). The GRI developed 36 standards that prescribe how a company should report about sustainability related themes like biodiversity, child labour, emissions, and discrimination (GRI, 2016). Companies voluntary report about their CSR because this results in more stakeholder engagement and a better reputation (EY, 2009). But the reported CSR performance is only credible for sustainability report users when these reports are assured by auditors.

Therefore, the Institute of Social and Ethical Accountability, also known as AccountAbility, and the International Audit Assurance Standard Board (IAASB), developed respectively the AA1000 Assurance Standard (AA1000AS) and the International Standard on Assurance Engagements (ISAE 3000) (Junior, Best, & Cotter, 2014). These two standards provide both a framework for the assurance of sustainability reports. Regarding the assurance of CSR reports, two types of assurance are possible, namely limited assurance and reasonable assurance (EY, 2015). When applying limited assurance, the auditors only investigate whether the CSR reports are in accordance with the relevant reporting standards. When applying

(9)

9

reasonable assurance, the auditors also investigate whether the CSR reports are in accordance with the relevant reporting standards, but also investigate whether the information in the reports is free from material misstatements.

The International Integrated Reporting Council (IIRC) is attempting to change the fact that there are only voluntary accounting standards regarding sustainability reporting. They are attempting to change this by trying to institutionalize that Integrated Reporting is a practice that is critical to the relevance and value of corporate reporting (Humphrey, O’dwyer, & Unerman, 2017). Integrated Reporting is a form of corporate reporting whereby companies report on their financial performance and CSR practices in one report. Humphrey et al. examined how the IIRC is trying to get support for Integrated Reporting. By studying documentation of the IIRC, by conducting interviews with IIRC members and by participating in the IIRC network, Humphrey et al. found that the IIRC builds on their social network to get support for Integrated Reporting. The IIRC has a broad network, because its members are subtracted from several areas in the accounting society. Also, Humphrey et al. found that the efforts of the IIRC to institutionalize Integrated Reporting are associated with the efforts of the IIRC to mobilize the notion among investors and stakeholders that long-term matters. For a long time, capital providers have prioritized the short-term investment horizon over the long-term investment horizon. So the IIRC tries to create a long-term investment horizon among capital providers. When the capital providers have a long-term vision, there is need for sustainability reporting, Integrated Reporting becomes institutionalized, and then Integrated Reporting can become the international corporate reporting norm (Humphrey et al., 2017).

2.3 CSR and the Cost of Equity

Besides debt capital, the company can also finance its activities with equity capital. The cost of equity is mostly estimated with the capital asset pricing model (CAPM). According to CAPM, the cost of equity capital is determined by the extent to which the company is exposed to market risk (Fama & French, 1997). Market risk, also known as systematic risk, is the risk that a portfolio with all shares available at the market has. This market portfolio has only undiversifiable risk, because the portfolio could not be further diversified because it already contains all shares available. So, the cost of equity capital of a company is determined by the risk of their shares that could not be diversified with a more comprehensive share portfolio, with more shares of other companies (Berk & DeMarzo, 2014).

(10)

10

Dhaliwal, Zhen Li, Tsang, and Yang examined the relationship between CSR reporting and the cost of equity capital. By examining whether sustainability reporting reduces the cost of equity capital, they could find an explanation for the increasing disclosure of CSR reports (Dhaliwal et al., 2011). The researchers state that, although they examine the effect of sustainability reporting on the cost of equity capital, CSR reporting could have an effect on a lot of other aspects. For instance on the cost of debt, what is examined in this thesis. The research of Dhaliwal et al. found that companies which have a relatively high cost of equity capital in a certain year, have the tendency to disclose a report about their sustainability practices the next year. When these sustainability reports report a superior CSR performance in the next year, these reporting companies indeed have a lower cost of equity capital. The research also found that companies which publish sustainability reports that report a comprehensive sustainability performance, get more attention from analysts. Moreover, the analysts are able to make more accurate forecasts of the companies that disclose CSR reports. Finally, firms that publish sustainability reports have more urge to issue new stock than companies that do not publish CSR reports. And the stock issuances of sustainability reporting companies raise more capital than the issuances of non-sustainability reporting firms (Dhaliwal et al., 2011).

Besides Dhaliwal et al. (2011), El Ghoul, Guedhami, Kwok, and Mishra (2011) also examined the effect of sustainability on the cost of equity capital. In line with the research of Dhaliwal et al., Guedhami et al. also found that companies with more comprehensive sustainability practices have a lower cost of equity capital. They also found that companies which operate in the tobacco or nuclear power industry, have a higher cost of equity capital. Concluding, the disclosure of CSR reports has a lot of advantages, and ultimately it lowers the cost of equity capital. This thesis examines whether the sustainability reports also lower the cost of the other source of capital, namely the cost of debt capital.

2.4 Characteristics of Debt

The previous paragraph described the research of Dhaliwal et al. (2011) and Guedhami et al. (2011), which examined the relationship between CSR and the cost of equity. Dhaliwal et al. (2011) stated that sustainability practices of companies could have a different effect on the cost of debt than on the cost of equity, because of the fact that debt has a different payoff function than equity. This difference is because debtholders have a fixed claim on the company, while equityholders are the residual claimants (Kanda, 1992, Admati, DeMarzo,

(11)

11

Hellwig, & Pfleiderer, 2018). This means that when the firm is performing extraordinary, the debtholders still only receive the amount that they initially borrowed to the company. The equityholders receive the money that is left in the company after the debtholders have been repaid, which can be huge by extraordinary performance.

Because of the difference between the payoff functions of debt and equity, some conflicts of interests between debtholders and equityholders can arise. First of all, the sharing problem can arise (Kanda, 1992, Berk & DeMarzo, 2014). This means that the company first borrows money from a certain debtholder, and after that borrows money from another debtholder. In the case of a default of the company, the law treats the first debtholder and the second debtholder equally, which means that both debtholders receive only part of the amount that they originally borrowed to the company. So because the company chose to obtain additional capital, the invested amount of the first debtholder becomes more risky, while the first debtholder still receives the originally agreed interest payments. The second debtholder knows that the company also has the first debtholder as debtholder, and can ask a higher interest rate because of the risk.

Second, the risk-alteration problem can arise (Kanda, 1992, Berk & DeMarzo, 2014). The risk-alteration problem lies in the fact that companies have the legal form limited liability, which means that they do not have to repay their debt in case of a default. Because of this, the company borrows a lot of money and then takes extensive risks. The company, that acts in the interest of the equityholders, takes extensive risks because only taking extensive risks can be profitable for the equityholders. When they do not take extensive risks, all money that a company generates will go to the debtholders because of the high leverage of the company, and the equityholders will get nothing. When the company does take extensive risks, and the risk taking fails, there is only money left in the company to repay part of the debtholders. The equityholders get no money, but also do not have to repay money to the debtholders because of the limited liability of the company. When the company takes extensive risks, and the extensive risk taking is successful, the company generates a lot of money. Then the company repays the debtholders and the equityholders receive what is left. So the risk-alteration problem means that the company, that acts on behalf of the equityholders, takes extensive risks because only by taking these extensive risks, there is a chance that the equityholders will receive some money.

(12)

12

Companies take debt because of the tax advantages of debt (Graham, 2000, Fama & French, 2005, Kemsley & Nissim, 2002). Debt has a tax advantage, because the interest costs paid on debt can be subtracted from the taxable profit. The tax advantage is known as the tax shield, and can be calculated by multiplying the interest costs paid on debt by the corporate tax rate. When there is debt, the equityholders receive less money, because part of the money is going to the debtholders, but the equityholders and debtholders together receive more money because of the tax shield. So by classifying part of the capital providers as debtholders, the projects of the company generate more money because less taxes are paid to the authorities.

2.5 The Determinants of the Cost of Debt

The interest rate which a company should pay on their debt depends on three factors (Merton, 1973). First, the interest rate depends on the required rate of return on riskless debt. Mostly, the rate of return on riskless debt is equal to the rate of return on governmental bonds. Second, it depends on the conditions of the debt, for example the maturity date. A debt that has to be repaid after ten years, has a higher interest rate compared to a debt that has to be repaid after only one year. Third, the interest rate depends on the probability that the firm will be unable to satisfy the requirements of the debt. So, this is the probability that a company will default. When the companies have a bigger chance to default, capital providers do want a compensation for this risk, for example by means of a higher interest rate. The research of Longstaff, Mithal, and Neis (2005) found that the probability of a default is the biggest determinant of the cost of debt. The interest rate is important for companies, because it determines how much interest costs have to be paid every year. This has consequences for the liquidity, because money that is paid to debtholders cannot for example be used to invest in new projects.

Sengupta (1998) examined the relationship between the quality of the disclosed financial statements and the cost of debt. The study measured the quality of the disclosures by using evaluations of financial analysts. Sengupta (1998) found that higher quality disclosures are associated with a lower cost of debt. The argument for this is that companies which disclose high quality information are perceived to have a lower likelihood that they conceal unfavourable information, and therefore have a lower default risk. The research also found that disclosing high quality information is very important for companies that have very volatile stocks. On this way these companies can reduce the cost of debt. The research of Hail

(13)

13

(2002) also found that higher quality disclosures are associated with a lower cost of debt. Francis, Khurana, and Pereira (2005) examined the relation between the cost of debt and the disclosure levels of companies. They found that companies with high disclosure levels, so companies that voluntary disclose additional data, have a lower cost of debt capital and also a lower cost of equity capital. This is in line with the research of Francis, Nanda, and Olsson (2008), which also examined the relationship between the cost of debt and the disclosure levels of companies.

2.6 The Legitimacy Theory and the Stakeholder Theory

Nowadays, companies are focussing on sustainability. The CSR reporting area is developing, and there is already a lot of academic literature available regarding sustainability. But why do companies care about sustainability? According to Gray, Kouhy, and Lavers (1995), there are two main theories that could explain why organisations focus on sustainability. These are the legitimacy theory and the stakeholder theory, which are discussed below.

The legitimacy theory states that an organisation is legitimate when there is congruence between the social norms associated by the activities of the company and the norms of acceptable behaviour in the larger social system of which the organisation is part of (Dowling & Pfeffer, 1975, Magness, 2006). When there is incongruence between the activities of the company and the norms of acceptable behaviour of society, there will be a threat to the organisational legitimacy. This means that the company will be punished by society for their behaviour with legal, economic or other social sanctions. This will have a negative effect on the future continuity of the company. So, the companies have to act in line with the interests of society, to get a licence to operate. On first sight, this seems in contrast with a neo-classical economic perspective, because a neo-classic economic perspective states that companies should focus only on creating shareholder wealth. Expenditures for society, for example expenditures on social aspects, reduce shareholder wealth in a neo-classic economic perspective (Ruf, Muralidhar, Janney, & Paul, 2001). But when further investigating the legitimacy theory, it turns out that the expenditures on social aspects are creating shareholder value, because on this way the company is getting legitimacy from society, which will result in future continuity of the company. When there is continuity in the future, the company is able to make profits in the future, which is in favour of the shareholders.

(14)

14

The stakeholder theory states that an organisation will act to the demands of their stakeholders. So instead of acting to the norms of one larger social system as the legitimacy theory states, the stakeholder theory states that the organisation will act to the norms of several stakeholder groups (Jensen, 2010, Laplume, Sonpar, & Litz, 2008). Deegan and Blomquist (2006) state that every stakeholder group has different preferences. The company cannot fulfil the needs of every stakeholder group. So the power that a stakeholder group has is important for a stakeholder group, because it determines whether it is able to push the company in a direction they like. At last, Deegan and Blomquist (2006) emphasised that the legitimacy theory and the stakeholder theory are largely overlapping each other and so are complementing.

(15)

15

3. Hypothesis

From the literature review it becomes clear that society is considering sustainability more and more important. According to the legitimacy theory, companies should also consider sustainability as important. On this way, there is congruence between the activities of the company and the norms of society. The company gets a licence to operate from society, which has a positive effect on the future continuity of the company and so on the future performance of the company.

The stakeholder theory argues that the company should act in line with the demands of their stakeholder groups. According to Deegan and Blomquist (2006), the company will act in line with the demands of the most powerful stakeholder groups. Because debtholders are the main source of external finance, they could be seen as a powerful stakeholder group. So, the company will act in line with the demands of the debtholders. When taking into account the legitimacy theory and the related licence to operate from society, debtholders do care about this licence. They do so because this licence has a positive effect on the future continuity and the future performance of the company, and so in the end on the chance that the debt will be repaid. So the debtholders want that the company gets a licence to operate from society. This is possible when the company is focussing on sustainability, because then there is congruence between the activities of the firm and the norms of society.

So when combining the legitimacy theory and the stakeholder theory, enhanced sustainability practices should lead to future continuity and in the end future performance, and the enhanced sustainability practices of the company are supported by the debtholders. But do CSR practices indeed enhance the performance of a company? Ruf et al. (2001) examined in their research whether a change in corporate social performance is related to a change in financial accounting measures. The research found that the change in corporate social performance was positively related to sales growth in the short-term and return on sales in the long-term. Of course, some short-term accounting measures report a decreasing financial performance, because of the expenditures related to the enhancement of the sustainability practices (Baird, Geylani, & Roberts, 2012). Al-Tuwaijri, Christensen and Hughes (2004) examined the relation between environmental practices and economic performance of firms in the long run. They found that more environmental practices were associated with a higher economic performance. Their findings are in accordance with the research of López, Garcia, and Rodriguez (2007), who also examined this link.

(16)

16

Overall, it can be concluded that enhanced sustainability practices indeed enhance future performance, and that the described relations related to the legitimacy theory and stakeholder theory are plausible.

Because of the positive effect of sustainability on firm continuity and firm performance, there is less risk that a company is not able to repay the debt in the future. As Merton (1974) stated, one determinant of the cost of debt is the risk of a credit default. According to Longstaff et al. (2005), the probability of a credit default is even the biggest determinant of the cost of debt. Because there is less risk of a credit default, I expect that the cost of debt reduces. So, enhanced sustainability practices have a positive effect on future performance, and this increased future performance lowers the risk of a credit default and thus lowers the cost of debt. The hypothesis is as follows:

H1: More enhanced sustainability practices reduce the average cost of debt.

As stated in chapter two, enhanced sustainability practices are related to economic, environmental and social aspects.

The research of Ruf et al. (2001) found that enhanced sustainability practices increase the financial performance measure sales growth in the short-term. Blaird et al. (2012) found that some financial performance measures report a decreasing financial performance in the short-term, because of the expenditures related to the enhancement of the sustainability practices. However, Blaird et al. (2012) state that investors, like debtholders, focus on financial performance measures like sales growth, because these measures are more future orientated than financial performance measures that directly include the CSR related expenditures. So, because debtholders focus on the future orientated performance measures, enhanced sustainability practices are associated with enhanced financial performance in the short-term. Because of this, the risk of a credit default diminishes in the short run, and so the short-term cost of debt (Merton, 1974, Longstaff et al., 2005).

The researches of Ruf et al. (2001), Al-Tuwaijri et al. (2004) and López et al. (2007) found that enhanced sustainability practices were associated with increased financial performance in the long-term. Because of the increased financial performance, the credit default risk in the future diminishes. The company is able to repay the debt in the future. Therefore, I expect that more enhanced sustainability practices reduce the cost of long-term

(17)

17

debt (Merton, 1974, Longstaff et al., 2005). Summarized, the hypotheses related to the short-term cost of debt and the long-short-term cost of debt are as follow:

H2a: More enhanced sustainability practices reduce the short-term cost of debt.

(18)

18

4. Sample and Empirical Design

4.1 Sample

To examine the effect of the sustainability practices of companies on the cost of debt, I obtained data that was all available through the Wharton Research Data Services. The MSCI ESG KLD STATS database provided data about the sustainability practices of companies. The MSCI database has gathered data since 1991, but I chose to only use data between 2003 and 2013. On this way, I get the best data coverage. The data of the MSCI database is only gathered from companies from the United States of America. Before merging the MSCI dataset with other relevant data and before removing the outliers, but after deleting the missing data, the MSCI sample consists of 33,962 observations of 8,610 companies.

The Compustat – Capital IQ database provided the data regarding the cost of debt variable and all control variables. In general, the data obtained from Compustat was, just like the data from the MSCI database, from 2003 till 2013. However, Compustat measures the value of some variables at the end of the year, while I needed the value of the variables at the beginning of the year. Therefore, for some variables I gathered data from 2002 till 2012. On this way, by using the end-year value of a variable in yeart-1_{, I could get the begin-year value}

of the variable in yeart_.

The data from the MSCI database and the Compustat database have been merged. After this merge and after winsorizing and dropping the outliers, regarding the sample for the first hypothesis, 11,545 observations from 2,604 companies between 2003 till 2013 are left. Regarding hypothesis 2a, 1,422 observations from 412 companies are left. Regarding hypothesis 2b, 987 observations from 395 companies are left. The samples of hypotheses 2a and 2b are smaller than the sample that is used to test hypothesis 1. This is because a lot of companies report about their total interest cost paid on debt, so the interest cost paid on short-term debt together with the interest cost paid on long-term debt, and report about their total debt outstanding, so the amount of short-term debt together with the total amount of long-term debt. But only relatively a few companies in the database specify which part of the total interest cost paid are attributable to short-term debt and which part to long-term debt.

(19)

19 4.2 Model

In order to test the effect of sustainability practices on the cost of debt, I developed three regression models. According to Dhaliwal et al. (2011), using regression models are appropriate to assess the effect of sustainability on the cost of capital. In the following section, I will discuss three models that are equivalent to the models used in Dhaliwal et al. (2011).

4.2.1 The Average Cost of Debt

To assess the impact of sustainability practices on the average cost of debt, so the long-term cost of debt and the short-term cost of debt together, I use the following model:

COD_AVERAGEi,t = β0 + β1SUSi,t + β2FIRM_SIZEi,t + β3LEVi,t + β4LITi,t + β5MKT_RISKi,t + β6LTGi,t +β7ROAi,t + β8BIG4i,t + εi,t

The variables of the model are defined as follows:

COD_AVERAGE is the average cost of debt, defined as the total interest cost paid on both

short-term debt and long-term debt during the year, divided by the average of the sum of the short-term debt and long-term debt during the year (Pitmann & Fortin, 2004).

SUS is the overall sustainability score of the company according to the MSCI database. The

overall sustainability score can be calculated by subtracting the CSR weaknesses from the CSR strengths (Kim, Park, & Wier, 2012, Dhaliwal et al., 2011).

FIRM_SIZE is a control variable for firm size. Firm size is measured by taking the logarithm

natural from the market value of equity at the beginning of the year (Dhaliwal et al., 2011).

LEV is a control variable for leverage. Leverage is measured by dividing the total debt, so

long-term debt and short-term debt, by the total assets (Goss & Roberts, 2011, Dhaliwal et al., 2011).

LIT is a control variable for litigation risk. Litigation risk is a dummy variable, and equals 1 if

there is litigation risk (Dhaliwal et al., 2011).

MKT_RISK is a control variable for market risk. Market risk is measured by taking the

(20)

20

LTG is a control variable for long-term growth. Long-term growth is measured using the

market-to-book ratio (Goss & Roberts, 2011, Dhaliwal et al., 2011).

ROA is a control variable for the return on assets. ROA is measured as income before

extraordinary items divided by the total assets at the beginning of the year (Dhaliwal et al., 2011).

BIG4 is a control variable to assess whether a company is audited by a Big 4 firm or not. This

dummy variable equals 1 if a firm is audited by a Big 4 firm (Karjalainen, 2011).

4.2.2 Short-term and Long-term Cost of Debt

To assess the impact of sustainability practices on the short-term cost of debt, I use the following model:

COD_SHORTi,t = β0 + β1SUSi,t + β2FIRM_SIZEi,t + β3LEVi,t + β4LITi,t + β5MKT_RISKi,t + β6LTGi,t +β7ROAi,t + β8BIG4i,t + εi,t

COD_SHORT is a measure for the short-term cost of debt. The interest rates that the

companies pay on short-term debt are directly retrieved from the Compustat database. All other variables in the above model are the same as described in the model of the average cost of debt in paragraph 4.2.1.

Finally, to assess the impact of sustainability practices on the long-term cost of debt, I use the following model:

COD_LONGi,t = β0 + β1SUSi,t + β2FIRM_SIZEi,t + β3LEVi,t + β4LITi,t + β5MKT_RISKi,t + β6LTGi,t +β7ROAi,t + β8BIG4i,t + εi,t

COD_LONG is the long-term cost of debt, defined as the total interest cost paid on long-term

debt during the year, divided by the average amount of long-term debt during the year (Pitmann and Fortin, 2004). All other variables in the above model are equal to the model of the average cost of debt and the model of the short-term cost of debt.

(21)

21 4.3 Main Variables

When reviewing relevant literature, the cost of debt is measured in two ways. First of all, the cost of debt is measured by the absolute amount of interest costs paid on debt (Sengupta, 1998). Second, the cost of debt is measured by the interest rate paid on debt (Pitmann and Fortin, 2004). The interest rate can be calculated by dividing the interest costs paid on debt by the total amount of debt. So, the interest rate measures the interest cost paid on debt relative to the amount of debt. On this way, the interest costs paid on various amounts of debt can be compared. As becomes clear from the previous paragraph, I measure the cost of debt in this research using the interest rate.

The MSCI KLD STATS database measures the performance of companies on CSR practices. According to Chatterji, Levine, and Toffel (2009) is the MSCI database the most used database to operationalize the sustainability practices of companies. The database assesses the sustainability performance of the companies by using for example company disclosures, government databases, academic literature, and NGO datasets. Also, the MSCI database invites companies to participate in the data verification process (MSCI, 2015). The database measures the performance of the companies in thirteen categories. These categories are environmental, community, human rights, employee relations, diversity, product, corporate governance, alcohol, firearms, gambling, military, nuclear power, and tobacco (MSCI, 2015). In each of the first seven categories are positive performance indicators and negative performance indicators. The last six categories are controversial business involvement indicators (MSCI, 2015). In the overall assessment of the sustainability performance of the companies, I do not include the performance in the last six categories, because these categories do not pertain to the companies’ discretionary activities (Kim et al., 2012).

If a company is in line with a positive performance indicator it will score a “1”. When a company is not in line with a positive performance indicator it will score a “0”. But when the company is in line with a negative performance indicator, it will also scores a “1”. And when the company is not in line with a negative performance indicator, it will score a “0” on that performance indicator (MSCI, 2015). To get the CSR performance in a certain category, I subtract the total score on negative performance indicators from the total score on positive performance indicators. To get the total overall sustainability score of a company, I sum up together the total score of the seven categories (Kim et al., 2012).

(22)

22 4.4 Control Variables

First of all, I will control for firm size. According to Agustini (2016) is firm size negative related to the cost of capital. Bigger firms have a lower risk than smaller firms, so capital providers will borrow capital at a lower interest rate to larger firms than to smaller firms. Second, I include a control variable for leverage. From the theoretical part it became clear that when a company has a higher leverage, the sharing problem could arise. Therefore, capital providers have a higher risk, and will provide capital against a higher interest rate. So, leverage is positively related to the cost of debt. A lot of researches include leverage as a control variable, for example the research of Anderson, Mansi, and Reeb (2003).

Third I include the control variable litigation risk. Litigation risk is the risk of having lawsuits in the future. Potential lawsuits can have an impact on future performance and future continuity, and in the end thus the cost of debt. The dummy variable in this research equals 1 if a company operates in a high-litigation industry. High litigation industries are industries with the SIC codes of 2833-2836, 3570-3577, 3600-3674, 5200-5961, and 7370 (Dhaliwal et al., 2011). Fourth, I control for market risk. According to Sengupta (1998), companies that face higher market risks have a higher cost of debt. So market risk and the cost of debt are positively related. The measure of market risk, the logarithm natural of the assets, is in some other researches also used as measure for firm size (Anderson et al., 2003, Dhaliwal et al., 2003).

Fifth, I include the control variable long-term growth. This variable is measured using the market-to-book ratio. The ratio says something about the future potential of the company, about the growth opportunities. A higher market-to-book ratio is associated with a lower cost of debt (Chen and Zhao, 2006). Sixth, the control variable return on assets is included. This measure says something about the profitability of the company (Dhaliwal et al., 2011). When the company is profitable, more money is available to engage in sustainability practices. These sustainability practices lower the cost of debt because of the licence to operate from society. So, more profitable companies are associated with a lower cost of debt. Finally, I include the control variable Big 4 auditor. Big 4 auditors are auditors from Deloitte, Ernst and Young, KPMG, and PricewaterhouseCoopers. According to Karjalainen (2011) are Big 4 audits associated with a decreased cost of capital.

(23)

23 4.5 Method

After merging all data, outliers had to be removed from the datasets to be able to run appropriate regression analyses. To test the three hypotheses, three different datasets were created. The three datasets differ from each other, because they contain respectively data about the average cost of debt (COD_AVERAGE), the short-term cost of debt (COD_SHORT), and the long-term cost of debt (COD_LONG). In the dataset of the average cost of debt, I winsorized all variables, except the variable COD_AVERAGE, at the 1st_{and 99}th_{percentiles to}

remove outliers. The variable COD_AVERAGE contained much more outliers than the other variables in the dataset, especially on the upper bound of the dataset. Therefore, to also get reliable descriptive statistics for this variable, I chose to trim the variable COD_AVERAGE at the 1st_{, 95}th_{, 96}th_{, 97}th_{, 98}th_{, and 99}th_percentiles.

To remove the outliers from the short-term cost of debt dataset, I winsorized all variables at the 1st_{and 99}th_{percentiles. To remove the outliers from the long-term cost of}

debt dataset, I winsorized all variables, except the variable COD_LONG, at the 1st_{and 99}th

percentiles. The variable COD_LONG contained, like the variable COD_AVERAGE, a lot of outliers on the upper bound of the dataset. To get appropriate descriptive statistics, I trimmed the variable COD_LONG at the 1st_{, 98}th_{, and 99}th_percentiles.

After winsorizing and trimming the data, I used ordinary least square regression (OLS) to estimate the coefficients for the variables. In the additional analysis of this research, a control variable for credit rating (CR) is included in the average cost of debt sample. To test the impact of this credit rating, again an ordinary least square regression (OLS) is run.

(24)

24

5. Results

5.1 Descriptive Statistics

In table 1, the observations per year for respectively the average cost of debt sample, the short-term cost of debt sample, and the long-term cost of debt sample are shown. The table shows that for all three samples, the number of observations per year increased each year. Except for the last three years of the short-term cost of debt sample. In these last three years, the number of observations per year decreased compared to the previous years. This is because the number of invalid observations increased in the last three years of this sample.

Table 1. Observations per year

Year COD_AV. Percentage COD_SH. Percentage COD_LO. Percentage

2003 514 4.45% 102 7.17% 14 1.42% 2004 585 5.07% 123 8.65% 15 1.52% 2005 575 4.98% 122 8.58% 22 2.23% 2006 682 5.91% 122 8.58% 49 4.96% 2007 823 7.13% 136 9.56% 52 5.27% 2008 891 7.72% 127 8.93% 54 5.47% 2009 1441 12.48% 208 14.63% 113 11.45% 2010 1482 12.84% 231 16.24% 119 12.06% 2011 1483 12.85% 214 15.05% 156 15.81% 2012 1485 12.86% 37 2.60% 147 14.89% 2013 1584 13.72% 0 0.00% 246 24.92% Total 11,545 100.00% 1,422 100.00% 987 100.00%

Notes: This table provides the number of observations per year for the average cost of debt sample (COD_AV), the short-term cost of debt sample (COD_SH), and the long-term cost of debt sample (COD_LO), for the years 2003 till 2013.

Table 2, table 3, and table 4 contain the descriptive statistics for respectively the average cost of debt sample, the short-term cost of debt sample, and the long-term cost of debt sample. First, the descriptive statistics in the three tables show that the mean short-term cost of debt is 0.85%, while the mean average cost of debt and the mean long-term cost of debt are around 6%. This seems reasonable in practice, because long-term debt has higher interest rates than short-term debt (Merton, 1973). Second, the three tables show that the mean sustainability score of the companies in the three samples differ slightly from zero, but are around zero. This

(25)

25

means that all companies aggregated, engage in approximately as much negative sustainability practices as positive sustainability practices. Third, around 90% of all companies in the three samples are audited by a Big 4 auditor. The mean firm size of a company is about equal in the three samples, and the companies in the average cost of debt sample and the long-term cost of debt sample are on average more levered than the companies in the short-term cost of debt sample.

Table 2. Descriptive statistics average cost of debt sample

Variables Mean Median St. dev Minimum p25 p75 Maximum

COD_AVER. 0.0637 0.0611 0.0267 0.0098 0.0462 0.0775 0.1615 SUS -0.2601 -1.0000 2.6767 -6.0000 -2.0000 1.0000 10.0000 FIRM_SIZE 7.2772 7.1274 1.6214 3.9710 6.0843 8.2556 11.7273 LEV 0.2574 0.2263 0.1889 0.0000 0.1158 0.3550 0.9138 LIT 0.1942 0.0000 0.3956 0.0000 0.0000 0.0000 1.0000 MKT_RISK 7.6024 7.4772 1.6325 4.1475 6.4598 8.6232 12.0218 LTG 2.7155 1.9873 3.7668 -12.6706 1.2663 3.2354 23.4229 ROA 0.0371 0.0444 0.1087 -0.5484 0.0096 0.0856 0.3092 BIG4 0.8942 1.0000 0.3075 0.0000 1.0000 1.0000 1.0000

Notes: This table provides the descriptive statistics for the average cost of debt sample (N = 11,545).

Table 3. Descriptive statistics short-term cost of debt sample

COD_SHORT 0.0085 0.0000 0.0185 0.0000 0.0000 0.0000 0.0780 SUS -0.4044 -1.0000 2.1087 -4.0000 -2.0000 1.0000 7.0000 FIRM_SIZE 6.8650 6.5749 1.5548 4.1030 5.7489 7.7445 11.8233 LEV 0.0646 0.0000 0.1274 0.0000 0.0000 0.0690 0.6136 LIT 0.4163 0.0000 0.4931 0.0000 0.0000 1.0000 1.0000 MKT_RISK 6.3449 6.0165 1.6579 3.3522 5.1374 7.3974 11.0536 LTG 4.0706 2.8569 5.5118 -10.5127 1.7866 4.6088 41.5152 ROA 0.0389 0.0706 0.1937 -0.8167 0.0120 0.1303 0.3961 BIG4 0.8509 1.0000 0.3563 0.0000 1.0000 1.0000 1.0000

Notes: This table provides the descriptive statistics for the short-term cost of debt sample (N = 1,422).

Finally, in the short-term cost of debt sample are relatively more firms with litigation risk compared to the other two samples. The mean values of the market risk variable and the return on assets variable do not differ much across the three samples. The average long-term

(26)

26

growth rate of the companies in the short-term cost of debt sample are higher than the average long-term growth rate in the other two samples.

Table 4: Descriptive statistics long-term cost of debt

COD_LONG 0.0581 0.0570 0.0238 0.0000 0.0440 0.0706 0.1555 SUS 0.6282 0.0000 2.8102 -4.0000 -1.0000 2.0000 10.0000 FIRM_SIZE 7.8675 7.7392 1.5570 4.5153 6.8259 8.9972 11.8848 LEV 0.2477 0.2100 0.1718 0.0033 0.1114 0.3612 0.7765 LIT 0.1733 0.0000 0.3787 0.0000 0.0000 0.0000 1.0000 MKT_RISK 8.7073 8.5381 1.9465 4.5215 7.2893 9.9329 14.1167 LTG 2.2480 1.6312 2.2059 0.1419 1.0569 2.5765 14.8644 ROA 0.0261 0.0190 0.0840 -0.5115 0.0075 0.0533 0.2303 BIG4 0.9392 1.0000 0.2391 0.0000 1.0000 1.0000 1.0000

Notes: This table provides the descriptive statistics for the long-term cost of debt sample (N = 987).

Table 5, table 6, and table 7 supply the Pearson correlations between the cost of debt, the sustainability scores, and the control variables in respectively the average cost of debt sample, the short-term cost of debt sample, and the long-term cost of debt sample. Most variables are correlated, and most correlations are significant at the 1% level. Moreover, in each of the three tables, the correlation coefficient between market risk (MKT_RISK) and firm size (FIRM_SIZE) has the highest value. Of all correlation coefficients in each of the three tables, this correlation coefficient is most closely to 1. A correlation coefficient of 0 means no correlation, while a correlation coefficient of 1 (or minus 1) means perfect correlation. The correlation coefficient of market risk and firm size is more than 0.80 in all three tables, which mean that the two variables are closely correlated. There is a significant relationship between these two variables, because the proxy in this research for market risk (the logarithm natural of the assets) is in other researches often used as proxy for firm size (Anderson et al., 2003, Dhaliwal et al., 2003).

To test whether the level of correlation between the variables is not to severe, I test for multicollinearity using the Variance Inflation Factor (VIF). According to O’brien (2007), a VIF of 10 or higher is associated with serious multicollinearity problems. I found that the average cost of debt sample, the short-term cost of debt sample, and the long-term cost of debt sample have a mean VIF of respectively 1.88, 2.47, and 2.34. Because no sample has a

(27)

27

mean VIF of 10 or higher, it can be concluded that there are no multicollinearity problems in the samples. For the VIF of each individual variable, see Appendix B.

Table 8, table 9, and table 10 provide the Spearman correlations between the cost of debt, the sustainability scores, and the control variables. In contrast to the Pearson correlation, the Spearman correlation is based on the rank orders of the values of the variables in the sample instead of on the absolute values. From table 8, table 9, and table 10 can be distracted that most variables are significant correlated, and that the significant correlations are mostly significant at the 1% level. The correlations between the variables market risk and firm size are again in all the three tables most closely to one. As I said above, the logarithm natural of the assets is often used as a proxy for firm size.

(28)

28 Table 5. Pearson correlations average cost of debt sample

COD_AV. SUS FIRM_SI. LEV LIT MKT_RI. LTG ROA BIG4

COD_AVERAGE 1 SUS -0.1322*** 1 FIRM_SIZE -0.2165*** 0.3516*** 1 LEV 0.0034 -0.0573*** -0.0024 1 LIT -0.0348*** 0.0633*** -0.0024 -0.0695*** 1 MKT_RISK -0.1104*** 0.3113*** 0.8384*** 0.0789*** -0.1303*** 1 LTG -0.0702*** 0.0809*** 0.1267*** -0.0396*** 0.0577*** -0.0520*** 1 ROA -0.1890*** 0.0884*** 0.2658*** -0.1517*** -0.1025*** 0.1692*** 0.0674*** 1 BIG4 -0.0726*** 0.1035*** 0.3151*** 0.0890*** 0.0214** 0.2480*** 0.0491*** 0.0475*** 1

Notes: This table shows the Pearson correlations between the dependent variable, the independent variable and the control variables in the average cost of debt sample (N = 15,545). * mean significant at the 10% level, ** mean significant at the 5% level, *** mean significant at the 1% level.

Notes: This table shows the Pearson correlations between the dependent variable, the independent variable and the control variables in the short-term cost of debt sample (N = 1,422). * mean significant at the 10% level, ** mean significant at the 5% level, *** mean significant at the 1% level.

Table 6. Pearson correlations short-term cost of debt sample

COD_SH. SUS FIRM_SI. LEV LIT MKT_RI. LTG ROA BIG4

COD_SHORT 1 SUS 0.0762*** 1 FIRM_SIZE 0.2467*** 0.3180*** 1 LEV 0.5466*** 0.1768*** 0.3151*** 1 LIT -0.1779** -0.0498* 0.0494* -0.1509*** 1 MKT_RISK 0.4163*** 0.3073*** 0.8854*** 0.4855*** -0.0296 1 LTG -0.0459* 0.025 0.0388 -0.0375 0.0397 -0.1076*** 1 ROA 0.0714*** 0.1402*** 0.3178*** 0.0334 -0.1160*** 0.2708*** -0.0753*** 1 BIG4 0.1078*** 0.0649** 0.2098** * 0.1458*** 0.0131 0.2519*** -0.0016 -0.0874*** 1

(29)

29 Table 7. Pearson correlations long-term cost of debt sample

COD_LO. SUS FIRM_SI. LEV LIT MKT_RI. LTG ROA BIG4

COD_LONG 1 SUS -0.2079*** 1 FIRM_SIZE -0.1741*** 0.4417*** 1 LEV 0.1043*** -0.1013*** -0.1187*** 1 LIT 0.1425*** -0.009 -0.0286 0.0109 1 MKT_RISK -0.2894*** 0.4263*** 0.8324*** -0.1888*** -0.2371*** 1 LTG 0.0708** 0.0425 0.1110*** 0.1415*** 0.1902*** -0.2295*** 1 ROA -0.1640*** 0.0354 0.1934*** -0.0686** 0.0517 0.0545* 0.2283*** 1 BIG4 -0.0192 0.1052*** 0.2679*** -0.0247 0.038 0.1756*** 0.0435 0.0847*** 1

Notes: This table shows the Pearson correlations between the dependent variable, the independent variable and the control variables in the long-term cost of debt sample (N = 987). * mean significant at the 10% level, ** mean significant at the 5% level, *** mean significant at the 1% level.

Table 8. Spearman correlations average cost of debt sample

COD_AV. SUS FIRM_SI. LEV LIT MKT_RI. LTG ROA BIG4

COD_AVERAGE 1 SUS -0.1255*** 1 FIRM_SIZE -0.1960*** 0.2564*** 1 LEV -0.0012 -0.0499*** 0.0748*** 1 LIT -0.0528*** 0.0419*** -0.0203** -0.1056*** 1 MKT_RISK -0.0741*** 0.2294*** 0.8221*** 0.1570*** -0.1360*** 1 LTG -0.1515*** 0.1101*** 0.2696*** -0.0179* 0.1077*** -0.0402*** 1 ROA -0.2043*** 0.0700*** 0.3171*** -0.1697*** 0.0097 0.0503*** 0.4225*** 1 BIG4 -0.0480*** 0.1022*** 0.3355*** 0.1131*** 0.0214** 0.2445*** 0.1228*** 0.0893*** 1

Notes: This table shows the Spearman correlations between the dependent variable, the independent variable and the control variables in the average cost of debt sample (N = 15,545). * mean significant at the 10% level, ** mean significant at the 5% level, *** mean significant at the 1% level.

(30)

30 Table 9. Spearman correlations short-term cost of debt sample

COD_SH. SUS FIRM_SI. LEV LIT MKT_RI. LTG ROA BIG4

COD_SHORT 1 SUS 0.0944*** 1 FIRM_SIZE 0.3191*** 0.2215*** 1 LEV 0.7572*** 0.1111*** 0.3768*** 1 LIT -0.1947*** -0.0764*** 0.0271 -0.1646*** 1 MKT_RISK 0.5011*** 0.2202*** 0.8495*** 0.5720*** -0.0591** 1 LTG -0.1785*** 0.0908*** 0.1932*** -0.1675*** 0.0617** -0.1247*** 1 ROA -0.0084 0.1357*** 0.3747*** -0.037 -0.0208 0.2111*** 0.3624*** 1 BIG4 0.1184*** 0.0580** 0.2194*** 0.1690*** 0.0131 0.2569*** 0.0053 -0.1244*** 1

Notes: This table shows the Spearman correlations between the dependent variable, the independent variable and the control variables in the short-term cost of debt sample (N = 1,422). * mean significant at the 10% level, ** mean significant at the 5% level, *** mean significant at the 1% level.

Table 10. Spearman correlations long-term cost of debt sample

COD_LO. SUS FIRM_SI. LEV LIT MKT_SI. LTG ROA BIG4

COD_LONG 1 SUS -0.2246*** 1 FIRM_SIZE -0.0983*** 0.4088*** 1 LEV 0.1029*** -0.0661** -0.0488 1 LIT 0.1311*** -0.0261 -0.0342 -0.0119 1 MKT_RISK -0.2165*** 0.3875*** 0.8129*** -0.1688*** -0.2356*** 1 LTG 0.1056*** 0.0291 0.1201*** 0.1172*** 0.2172*** -0.2826*** 1 ROA 0.0690** -0.0113 0.1390*** 0.0223 0.2178*** -0.1808*** 0.4986*** 1 BIG4 -0.0214 0.0995*** 0.2652*** 0.0037 0.038 0.1705*** 0.0463 0.1163*** 1

Notes: This table shows the Spearman correlations between the dependent variable, the independent variable and the control variables in the long-term cost of debt sample (N = 987). * mean significant at the 10% level, ** mean significant at the 5% level, *** mean significant at the 1% level.

(31)

31 5.2 Regression Results

This section shows the results of the OLS regressions executed to test the hypotheses. This section consists of three sub-sections, where respectively the regression results of the average cost of debt sample, the short-term cost of debt sample, and the long-term cost of debt sample are discussed. Table 11 provides the regression results. When in paragraph 5.2.1, 5.2.2, or 5.2.3 is spoken about significant, I mean significant at the 1% level.

5.2.1 Regression Result Average Cost of Debt

Table 11 shows that the model used to test hypothesis 1, has a F-value of 134.39, with a corresponding p-value of 0.0000. This means that at least one variable in the model is not equal to zero and so is explaining the variation in the dependent variable. The adjusted R-squared shows that 8.5% of the variation in the average cost of debt is explained by the independent variables. The CONSTANT of the model is 0.0799724, and this value is statistical significant. The coefficient of the variable SUS is -0.0007002, with a corresponding p-value of 0.0000. This means that SUS has a significant negative impact on COD_AVERAGE. So, enhanced sustainability practices of firms are reducing the average cost of debt, which means that hypothesis H1 should not be rejected.

The control variable FIRM_SIZE is significant negative related to COD_AVERAGE, which means that larger firms have a lower cost of debt. This is in line with Agustini (2016). The control variable MKT_RISK is significant positive related to COD_AVERAGE, and the control variable ROA is significant negative related to COD_AVERAGE. This means that a company with a higher market risk has a higher cost of debt, and that a company with a higher return on assets has a lower cost of debt. This is both in line with respectively Sengupta (1998) and Dhaliwal et al. (2011). The control variables LTG and BIG4 are negatively associated with COD_AVERAGE. This signifies that companies that have long-term growth or are assured by a Big 4 auditor, have a lower cost of debt. This is in line with respectively Chen and Zhao (2006) and Karjalainen (2011), but the found negative coefficients for LTG and BIG4 are not significant. Finally, according to Anderson et al. (2003), leverage (LEV) should be positively related to the cost of debt. According to Dhaliwal et al. (2011), litigation risk (LIT) should be positively related to the cost of debt. However, this research found a negative significant relationship between LEV and COD_AVERAGE, and a negative relationship between LIT and

(32)

32

COD_AVERAGE. The negative coefficient of LIT was only significant at a significance level of 5%.

Table 11. Regression results

COD_AVERAGE COD_SHORT COD_LONG

SUS -0.0007002*** (0.000) -0.0005211*** (0.008) -0.0011427*** (0.000) FIRM_SIZE -0.0055852*** (0.000) -0.0042111*** (0.000) 0.0056595*** (0.000) LEV -0.0057879*** (0.000) 0.0560064*** (0.000) 0.0040846 (0.331) LIT -0.0013039** (0.038) -0.0032988*** (0.000) 0.0029392 (0.144) MKT_RISK 0.0036053*** (0.000) 0.0062454*** (0.000) -0.0064374*** (0.000) LTG -0.0000080 (0.904) 0.0001643** (0.030) -0.0005977 (0.135) ROA -0.0338741*** (0.000) 0.0019538 (0.376) -0.0539141*** (0.000) BIG4 -0.0002140 (0.794) -0.0004345 (0.706) 0.0005684 (0.852) CONSTANT 0.0799724*** (0.000) -0.0050334** (0.015) 0.0710238*** (0.000) N 11,545 1,422 987 R-squared 0.0852 0.3661 0.1531 Adj. R-squared 0.0846 0.3625 0.1461 F-value 134.39 102.02 22.09 Prob > F 0.0000 0.0000 0.0000

Notes: This table reports the results of the regressions using respectively the average cost of debt (COD_AVERAGE), the short-term cost of debt (COD_SHORT), and the long-term cost of debt (COD_LONG) as dependent variables. The definitions of the variables are given in paragraph 4.2.1. Below each coefficient, the p-value of that coefficient is shown between brackets. The observations (N) on which the regressions are based, are from the years 2003 till 2013. In the bottom of the table, the R-squared, the adjusted R-squared, the F-value, and probability that the F-value is significant are given. Because paragraph 5.1 states that FIRM_SIZE and MKT_RISK are strongly correlated, the regression in this table is redone in Appendix C, without the control variable FIRM_SIZE.

*** indicate significance at the 1% level. ** indicate significance at the 5% level. * indicate significance at the 10% level.

(33)

33 5.2.2 Regression Result Short-term Cost of Debt

Table 11 also shows the regression results of the short-term cost of debt sample. The F-value of the model that is used to test hypothesis 2a, is 102.02, with a corresponding p-value of 0.000. So, the model contains at least one variable that is explaining the variation in the short-term cost of debt. The adjusted R-squared shows that 36% of the variation in the short-short-term cost of debt is explained by the model. This is more than the R-squared of the model used to test the effect of sustainability on the average cost of debt. The model has a constant of -0.0050334, which is only significant when testing at the 5% significant level.

The variable SUS has a coefficient of -0.0005211, with a corresponding p-value of 0.008. This means that SUS is significant negative related to COD_SHORT. So, companies with more enhanced sustainability practices, have a lower short-term cost of debt. This is in line with hypothesis 2a, so hypothesis 2a should not be rejected. The control variable FIRM_SIZE is significant negative related to COD_SHORT. This means that bigger firms are associated with a lower short-term cost of debt, which is in line with Agustini (2016). The coefficients of the variables LEV en MKT_RISK are both significant positive related to COD_SHORT, which means that companies that are more leveraged or are exposed to extensive market risk, have a higher cost of short-term debt. This is in line with researches of respectively Anderson et al. (2003) and Sengupta (1998).

Like in the average cost of debt model, the coefficient of the variable BIG4 is negative, in line with the research of Karjalainen (2011). However, the coefficient indicates again that the relationship between COD_AVERAGE and BIG4 is not significant. Where research of Chen and Zhoa (2016) and research of Dhaliwal et al. (2011) indicates that companies with respectively long-term growth or a high return on assets are associated with a lower cost of debt, this research found contrary that long-term growth (LTG) and return on assets (ROA) are positive associated with the cost of debt. However, the association between the short-term cost of debt and the return on assets is not significant, and the relation between the short-term cost of debt and the long-short-term growth is only significant when tested at a 5% significance level. Finally, where research of Dhaliwal et al. (2011) suggests that a higher litigation risk is associated with a higher cost of debt, this model indicates that litigation risk (LIT) is significant negative related to the cost of debt.

(34)

34 5.2.3 Regression Result Long-term Cost of Debt

From table 11 it becomes clear that the model that is used to test the effect of sustainability on the long-term cost of debt has a F-value of 22.09, with a corresponding p-value of 0.0000. At least one of the independent variables is explaining the effect on the long-term cost of debt. The table states that 15% of the variation in the long-term cost of debt is explained by the model. The model has a significant constant of 0.0710238.

The variable SUS has a coefficient of -0.0011427, with a corresponding p-value of 0.0000. There is a significant negative relationship between SUS and COD_LONG, which means that more enhanced CSR practices of companies are associated with a lower cost of long-term debt. This is in line with hypothesis 2b, which therefore should not be rejected. ROA has a significant negative coefficient, which means that more profitable companies have a lower cost of debt. This is in line with research of Dhaliwal et al. (2011). According to Anderson et al. (2003) and Dhaliwal et al. (2011), companies with a higher leverage (LEV) or with a higher litigation risk (LIT), should have a higher long-term cost of debt. This research indeed found a positive relationship between respectively leverage and long-term cost of debt, and litigation risk and long-term cost of debt. However, the positive relationships were not significant.

The coefficient of LTG is negative, indicating that companies with extensive long-term growth are associated with lower long-term debt costs. This is in line with research of Zhen and Zhao (2006). However, the coefficient is not significant negative. The coefficient of FIRM_SIZE is significant positive, indicating that larger firms have a higher cost of debt. This finding is contrary to the findings of Agustini (2016). The variable MKT_RISK is significant negative related to COD_LONG. This is not in line with the findings of the research of Sengupta (1998), because this research found that companies that are exposed to major market risk should have a higher cost of long-term debt. Finally, the coefficient of the variable BIG4 is positive, indicating that companies audited by a Big 4 company should have a higher cost of debt. This is not in line with the research of Karjalainen (2011), which found that companies audited by Big 4 companies should have a lower cost of debt. However, like in the average cost of debt sample and the short-term cost of debt sample, the coefficient of the variable BIG4 was not significant.

The link between sustainability practices and the cost of debt