Internal Controls Disclosure and the Cost of Equity
Written by: Frank Sander Muller Student number: S1923862
Supervised by: Niels Hermes
Co-assessed by: Reggy Hooghiemstra
Abstract
This thesis examines the relationship between the amount of internal controls disclosure disclosed in firms’ annual reports and the cost of equity capital. Following a theoretical framework, I propose that the level of voluntary disclosure relates negatively to the cost of
equity. Moreover, I propose that investor protection and legal enforcement relate negatively to the cost of equity, and that these two variables strengthen the relationship of
internal controls disclosure and the cost of equity. Using data from a sample of 2,529 firm year observations from 29 countries, I find a positive relationship between the level of
internal controls disclosure and the cost of equity.
University of Groningen
Faculty of Economics and Business
Nettelbosje 2
9747 AE, Groningen, the Netherlands Tel: +31 50 363 8900
f.s.muller@rug.nl
Uppsala University Department of Economics
S:t Olofsgatan 10B 753 12 Uppsala, Sweden +46 18 471 00 00 frank.muller.4532@student.uu.se
2
1. Introduction
The link between accounting information and the cost of capital of firms is one of the most fundamental issues in accounting (Lambert, Leuz, Verrecchia, 2006). There is mounting evidence that the level of corporate disclosure plays a role in equity valuations (e.g., Botosan and Plumlee, 2002;
Francis, Nanda and Olsson, 2008). There appears to be a direct and an indirect effect which influence equity valuations. The direct effect occurs because investors face less estimation risk when more is disclosed, whereas the indirect effect influences equity valuations through the firm’s change in real decisions as a result of disclosing certain information (Lambert et al., 2007).
The level of corporate disclosure is determined by the level of mandatory disclosure and the level of voluntary disclosure. A firm can influence the level of voluntary disclosure by disclosing information which is not required by law and is on a voluntary basis. Given the empirical evidence that the level of voluntary disclosure has capital market consequences for the firm’s stock (e.g. Healy, Hutton and Palepu, 1999; Gelb and Zarowin, 2000; Botosan and Plumlee, 2002), this topic is interesting to both practitioners and researchers. This might explain why a great deal of research in this area exists. The research on one specific aspect of voluntary disclosure, voluntary disclosure with regard to the firm’s internal controls, is however not as rich as research on voluntary disclosure in general. This is especially the case in regulatory environments outside the United States, where there the disclosure of information on the internal controls is not mandated by the government. Since 2008, Canada requires an effectiveness statement from management on the internal control system. And in the European Union, a comply-or-explain principle (meaning a firm has to comply with the internal control regulations or explain why it does not comply with these regulations) with regard to corporate governance regulation is effective since 2006. There is currently no consensus on which internal control regulations are preferred. Hermanson (2000) showed in his survey study that financial statement users agree that voluntary management reports on internal controls disclosure are important, because these reports motivate the management to improve internal controls, encourage audit committees to increase their oversight of the system, and provide additional, useful information for decision making.
This study attempts to provide empirical results on the effect of disclosing information with regard to a firm’s internal controls from a cost of equity perspective. I study countries where the disclosure of internal controls is up to management’s discretion and not prescribed by the government and therefore I exclude the United States and Canada.
In this study I investigate whether the level of internal control disclosure affects the implied cost of equity capital. I posit, in line with prior research in the area of voluntary disclosures (e.g., Botosan, 1997; Piotorski, 1999; Botosan and Plumlee, 2002), that the amount of internal control disclosures relates negatively to equity valuations. In addition, I investigate the interaction effect of investor
3 protection and the enforcement of the legal system with internal controls disclosure on firms’ cost of equity. Prior research has indicated that firms operating in countries with strong legal institutions which favor investors can obtain financing at better terms (La Porta, Lopez-de-Silanes, Shleifer, Vishny, 2000), hence these firms are expected to have a lower cost of equity capital. Legal institutions comprise the legal protection by law and the enforcement of these laws in practice. The effect of strong legal protection on paper can be diminished by poor enforcement of these laws in practice. Moreover, Lopez-de-Silanes, La Porta, Shleifer, and Vishny (1998) argue that a strong system of legal enforcement could substitute for weak rules since active and well-functioning courts can step in and rescue investors abused by the management. In this research, I also investigate the interactive effect of internal controls disclosure with investor protection and legal enforcement. I argue that the latter two factors improve the ability of investors to exercise control and when investors are able to exercise control (i.e. they are involved in more decisions), there is a larger need for information, because investors want to take informed decisions. In contrast, an investor who is not able to exercise control is more likely to be involved in less decisions taken within the firm, resulting in a lower need for information. This implies that an investor who is able to exercise control benefits more from a higher disclosure level than an investor who is less able to exercise control.
My regression analyses show that the level of internal controls disclosure relates positively to the implied cost of equity. With regard to the interaction effects, the regression analyses provide mixed results, suggesting that the proposed effects are not present in practice or have a very small effect that cannot be observed in my dataset.
In section 2, I outline the theoretical connection between internal control disclosure, investor protection, and legal enforcement and the cost of equity capital. Consequently, in section 3, I present the data used and the research design. I show the empirical results in section 4, before revealing the conclusions, limitations, and suggestions for future research.
2. Literature review
2.1. Internal control disclosures
Internal controls is broadly defined as “a process effected by an entity’s board of directors, management and other personnel, designed to provide reasonable assurance regarding the achievement of objectives in three categories” (COSO, 1992). The three categories to which these objectives relate are i. the effectiveness and efficiency of operations, ii. the reliability of financial reporting, and iii. the compliance with applicable laws and regulations. Internal controls are about managing the risks a firm faces (Spira and Page, 2002).
Disclosing information with regard to the internal controls has largely been voluntary. The passing of the Sarbanes Oxley Act of 2002, however, made disclosure of internal controls mandatory in the
4 United States in a response to a number of corporate scandals. The act has two provisions with regard to internal controls. Section 302 of the Act requires that chief executive officers and chief financial officers evaluate the design and effectiveness of internal controls on a quarterly basis. Moreover, section 404 of the Act requires an audit of the management’s evaluation of internal controls and the effectiveness of these on an annual basis.
Still, outside the United States law does not prescribe corporate disclosure of internal controls, thereby making disclosure on internal controls largely voluntary. Accordingly, firm management faces the decision of whether or not to disclose information with regard to the firm’s internal controls. The decision of voluntarily disclosing information is embedded in a larger theoretical framework of voluntary disclosure in general, which shall be discussed now.
2.2. Voluntary disclosures
The central premise in the disclosure literature is that asymmetric information among differentially informed capital market participants inhibits investment in a firm’s equity (Francis et al., 2008).
Consequently, a firm which issues equity is motivated to commit to a high level of disclosure in order to lower its cost of equity, and thereby lowering its overall cost of capital. There are however also costs related to disclosing information. These costs do not merely consist of costs directly associated by disclosing information, but also costs which are the indirect effect of disclosing information, such as the spilling over of strategic information and the potential costs of litigation. Despite the costs however, most previous research holds that firms that disclose more should have a lower cost of capital. Moreover, Healy and Palepu (2001) have identified four motives – based on existing literature – for managers to voluntarily disclose information, and two motives against disclosing information.
These shall be discussed now.
Capital markets transactions hypothesis. If information asymmetries between the management and outside investors cannot be resolved, firms experience issuing public equity or debt costly for the shareholders (Myers and Majluf, 1984). Managers can anticipate capital market transactions by reducing this information asymmetry, and thereby reducing the information risk for investors. One way of reducing this information asymmetry is by voluntarily disclosing information. Consequently, the firm can reduce its cost of capital by disclosing information (Merton, 1987).
Corporate control contest hypothesis. Managers are being held accountable for current stock performance by boards of directors and investors. Moreover, CEO turnover is positively related to poor stock performance (Warner, Watts and Wruck, 1988; Weisbach, 1988). Because of this, managers want to reduce the likelihood of undervaluation and to explain poor earnings performance, and do this by using corporate disclosures.
5 Stock compensation hypothesis. This hypothesis is based on two arguments and on the premise of managers receiving stock-based compensation plans, such as stock options. The first argument for this hypothesis is that managers, who are involved in stock trading (thus, also managers who possess stock- based compensation plans), have to comply with insider trading restrictions. Because of these restrictions on insider trading, managers are forced or incentivized to make voluntary disclosures.
Second, managers do not want to bear risk with regard to their stock-based compensation plans.
Therefore, it is likely that managers want stock prices to be an accurate estimate of firm value. In this way, it is more likely for stock compensation to be an efficient form of compensation.
Management talent signaling hypothesis. Because a firm’s market value is a function of investors’
perceptions of the managers’ ability to anticipate and respond to future changes in the firm’s economic environment, Trueman (1986) argues that talented managers want to make voluntary earnings forecasts prior to the actual earnings announcement in an attempt to build a good reputation as a manager. Early and accurate voluntary earnings forecasts makes investors more favorably assess the manager’s ability to anticipate economic environment changes and to adjust production schemes accordingly.
Litigation cost hypothesis. This is the first of the two arguments which provides a motive for managers not to disclose information. Shareholders can hold managers accountable for their disclosure decisions, and this influences the managers’ disclosure decisions in two ways. The first effect is that litigation can potentially reduce the managers‘ incentives to disclose information, because of the risk of disclosing wrong information. This is especially the case for forward-looking information.
Second, managers can be litigated when they disclose inadequate information or untimely, therefore managers have an incentive not to disclose information.
Proprietary cost hypothesis. This hypothesis argues that firms’ decisions to disclose information is influenced by the effect on the firm’s strategic position in the product market (e.g., Verrecchia, 1983;
Darrough and Stoughton, 1990; Wagenhofer, 1990; Feltham and Xie, 1992; Gigler, 1994). By voluntarily disclosing information, a firm gives away information which might attract new entrants or provide valuable information to the current competitors. The degree to which this motive is present depends on characteristics of the competition in the industry.
The previously mentioned motives for disclosing information are taken from a manager’s point of view, who eventually decides upon what to disclose. These motives do not all relate to the equity valuations, for example the management talent signaling hypothesis is not an argument for disclosing information in order to reduce the cost of equity capital. Shareholders, who eventually determine the equity valuations, are more interested in the economic consequences of voluntary disclosure. There are a number of studies which have investigated these economic consequences. These studies have
6 shown that there are three potential economic effects; improved liquidity for the firm’s stock in the capital markets, increased following by financial analysts and reductions in the cost of capital.
Improved stock liquidity. Voluntary disclosures reduce the information asymmetries among informed and uninformed investors. Consequently investors of firms with a high level of disclosure have confidence that stock transactions occur at a fair price, thereby increasing the stock’s liquidity and cost of capital (Diamond and Verrecchia, 1991; Kim and Verrecchia, 1994). The findings of Healy et al. (1999) suggest that indeed firm’s disclosure strategies affect the speed with which information gets reflected in prices. Moreover, Leuz and Verrecchia (2000) find that firms listed on an exchange with high disclosure requirements have lower bid-ask spreads than firms listed on an exchange with lower disclosure requirements.
Increased information mediation. Bhusan (1989) and Lang and Lundholm (1996) argue that if relevant information of the management is disclosed, the cost of acquiring this information for analysts is lower, consequently the demand will be higher. Lang and Lundholm (1993) find that firms which disclose more information have larger analyst following, less dispersion in analyst forecasts and less volatility in forecast revisions. Assuming that higher analyst following implies less information asymmetry, increased information mediation will ultimately lead to a lower cost of capital.
Reduced cost of capital. Barry and Brown (1986) argue that when disclosure is imperfect, investors bear risk in forecasting future returns on their investment. If investors cannot diversify this risk, they will demand a higher return on their investment for bearing such risk. Therefore, firms with higher levels of disclosure should have a lower cost of equity capital. In the following section I shall elaborate on this matter.
2.3. The relationship between voluntary disclosure and cost of equity: empirical evidence
Theoretical research indicating a negative relationship between voluntary disclosure and the cost of equity is grounded in two streams of arguments. The first argument is that higher disclosure enhances the stock market liquidity, which in turn leads to a reduced cost of capital through reduced transaction costs or increased demand for the firm’s securities (Demsetz, 1968; Copeland and Galai, 1983; Glosten and Milgrom, 1985; Amihud and Medelson, 1986; Diamond and Verrecchia, 1991). Secondly, from a theoretical perspective, greater disclosure reduces the estimation risk which arises from estimates of the parameters of an asset’s return or payoff distribution by investors. Hence, when information is low, uncertainty regarding the true parameters exist. If this estimation risk cannot be diversified by investors, they require a compensation for this element of risk (Klein and Bawa, 1976; Barry and Brown,
7 1985; Coles and Loewestein, 1988; Handa and Linn, 1993; Coles, Loewestein and Suay, 1995; Clarkson, Guedes and Thompson; 1996).
Empirical research in this area is not as rich as theoretical research, as reliable estimation methods of firms’ cost of capital have not been developed until relatively recently. Moreover the results are mixed. One of the first to empirically investigate the relationship is Botosan (1997). She found that a negative relationship exists for firms with a low analyst following, for firms with a high analyst following there was no significant relationship. In their paper, Botosan and Plumlee (2002) found that greater disclosure in the annual report reduces the cost of equity capital, however other (timely) publications, such as quarterly reports increase the firms’ cost of capital. This result is striking. It may, however, reflect executives’ concerns who noticed that the stock price responds to publications and when a firm has many timely publications, the stock price will be subject to more volatility. This, in turn, will lead to a higher cost of equity capital. Thus, the authors basically find that the frequency of corporate voluntary disclosure relates negatively to the cost of equity capital. Finally, Lambert et al. (2007) find a negative relationship, caused by a direct effect (more disclosure reduces the estimation risk for investors, consequently investors require a lower risk premium and thus a lower return) and an indirect effect (high quality disclosures have a positive impact on the firm’s real decisions, which in turn leads to a lower cost of capital).
A result not in line with the previously mentioned is provided in the paper of Francis et al. (2008).
They found that the unconditional relationship between voluntary disclosure and cost of capital is negative. However, they also found that the level of voluntary disclosure is positively related to earnings quality. Earnings quality, in turn, is negatively related to the cost of equity. Moreover, they find that earnings quality has a larger effect on the cost of equity than the level of voluntary disclosure has. And when earnings quality was included in the model, the effect of voluntary disclosure on the cost of equity reduced and in some instances was not significant. This result gets explained by the finding that firms with better earnings quality have more voluntary disclosures. Thus, the relationship between voluntary disclosure and earnings quality is complementary; a firm with a better earnings quality is likely to have more voluntary disclosures and a lower cost of equity capital. This explains why Francis et al. (2008) find that the effect of voluntary disclosures on cost of capital reduces or disappears completely, when earnings quality is included in the model as well.
Assuming that the voluntary disclosure of internal controls is not very deviant of voluntary disclosures in general, one would expect a negative relationship of internal controls disclosure on the cost of equity capital. Therefore I propose hypothesis 1, thereby testing whether internal control disclosure does not differ from voluntary disclosures in general.
8 Hypothesis 1: The level of internal controls disclosure is negatively related to the firms’ cost of equity.
2.4. Investor protection
The principle reason that investors provide financing to firms is because investors want a return on their investment. The means to get this return is, in a way similar to a contract, the acquisition of certain control rights vis-à-vis the assets of the firm (Hart, 1995). Investors rely on these control rights when they want their investment to materialize and this can be quite essential as investors face the risk that controlling shareholders and management appropriate returns (La Porta et al., 2000). When terms of the contract are violated by the management, the investors can go to court to enforce their rights. The nature of the legal obligations managers have towards the investors, as well as the differences in how courts interpret and enforce these obligations differ around the world.
Investor protection collectively refers to those features of the legal, institutional, and regulatory environment that facilitate financial contracting between inside owners and outside investors (Himmelberg, Hubbard, Love, 2004). The legal protection of investors is expressed in several practices.
One of these is the possibility for investors to vote by mail, because of this, the voting right of mainly small shareholders gets protected, as they are the least likely to attend shareholders meetings, due to the relative cost of attending such. Another example of legal protection comes from the requirement of certain countries that all ordinary shares carry one vote per share.
As investor protection improves, it requires more effort and becomes more costly for management to divert cash flows (Shleifer and Vishny, 1997; La Porta et al., 2000; Shleifer and Wolfenzon, 2000). Due to the higher costs of expropriation by the management in the situation of high (low) legal protection, they are less (more) inclined to do so, which results in a lower (higher) risk of being mistreated for the investors. When investors face less risk, they require a lower return on their investment, resulting in a lower cost of equity of the firm. Therefore I hypothesize:
Hypothesis 2: The level of investor protection is negatively related to the firms’ cost of equity.
As the empirical results on the relationship between voluntary disclosure and cost of equity capital are mixed I try to explain the variation in the results by the interaction effect of investor protection. As previously mentioned, informed shareholders require a lower return on their investment because of the reduced information risk. Investors also require this information to compare different investment opportunities. Apart from that, shareholders also build upon this information when they exercise their control. Assuming that a shareholder who is facing a decision wants to acquire information, the
9 investor will have a lower cost for obtaining this information when the information has been disclosed publically. Hence, when an investor is able to exercise control (and thus in need of information), voluntary disclosures are more valuable compared to the situation whereby an investor is not able to exercise control. In other words, the relevance of internal controls information increases as investor protection increases. Consequently, I propose that the relationship of voluntary internal controls disclosure is stronger when investors are to a higher extent protected and thus better able to exercise their power.
Hypothesis 3: When investor protection is high (low), the effect of voluntarily disclosing information on the cost of equity is stronger (weaker).
2.5. Legal enforcement
As Modigliani and Perotti (2000) mentioned back in 2000 already, legal enforcement is an issue which has often been neglected in literature, and it still is. The rights of the shareholders need to be enforced in order to be effective. For example, a country may have very good legal provisions for investors, however, when the insiders mistreat the investors in ways that are explicitly prohibited by law, it may be difficult for investors to “claim” their right.
According to the “law and economics perspective” regulations of financial markets are unnecessary (considering there are no transaction costs), because of the existence of financial contracts between issuers and investors. From this perspective, investors crucially rely on the legal enforcement and are not dependent on regulations protecting them. Hence, as Lopez de Silanes et al.
(1998) note, a strong system of legal enforcement can substitute for weak rules.
Whether investors rely on the legal regulations or the contracts they set up with the firms, the value of these forms of protection for the investors crucially relies on the legal enforcement. The level of legal enforcement differs per country. Courts can be unable or unwilling to invest the time and money required to understand the (sometimes complex) contracts investors set up with firms. This often is the case in poorer countries where the quality of enforcement is generally lower (La Porta, López de Silanes, Shleifer and Vishny, 1997). Moreover, courts at times operate slowly, are subject to political pressures and can be corrupt, thereby hindering the protection of investors.
In conclusion, investors can rely on the country regulations or contracts they set up with firms, or a combination of these. The value of these forms of protection is crucially dependent on legal enforcement. Therefore I hypothesize the following:
Hypothesis 4: The level of law enforcement is negatively related to the firm’s cost of equity.
10 Moreover, an investor whose rights are enforced well is actually able to exercise control. Using the same line of reasoning as why investor protection influences the relationship between cost of equity capital and internal controls disclosure, I argue that the effect of internal controls disclosure on cost of equity capital is influenced by the degree of law enforcement. This argument relies on the expectation that to an investor who is able to exercise control the disclosed information is more useful.
Hypothesis 5: When law enforcement is high (low), the effect of voluntarily disclosing information on the cost of equity is stronger (weaker).
3. Data and methods
3.1. Sample and data collection
This study relies on manually collected information about the disclosure practices of 1,078 listed firms in 29 countries during the years 2005 to 2007 with analyst following. The United States are excluded from this sample, as internal controls disclosure is mandated by the Sarbanes Oxley Act. The sample is diverse and firms have been chosen in highly developed countries as well as less developed countries and in different legal systems.
The sample contains non-financial firms with available Compustat data as of 2005, 2006, and 2007.
A stratified approach has been applied to the collection of the internal controls disclosure data, where typically 15 large firms (mostly blue chip firms) for each country were collected and then other firms got added randomly, until 10% of the country’s amount of listed firms were included. This approach is similar to the approach applied by the Center for International Financial Analysis and Research (CIFAR).
However CIFAR includes relatively more large firms (Francis et al., 2008), whereas my sample contains small and medium-sized enterprises as well.
3.2. Dependent variable
The cost of equity has often been measured in the asset pricing literature by the ex post realized return (e.g., Foerster and Karolyi, 1999; Errunza and Miller, 2000). However, more recent literature has criticized this method (e.g., Fama and French, 1997; Elton, 1999). Elton (1999), for example, shows that realized return is a poor and potentially biased proxy for the cost of equity. Moreover, Fama and French (1997) argue that the single-factor, capital asset pricing model as well as the Fama-French three-factor model arrive at imprecise cost of equity estimates. Finally, these measures also capture changes in market expectations about firms’ future cash flows (Bekaert and Harvey, 2000).
11 Due to these limitations, I employ a widely used alternative cost of equity proxy (e.g., Botosan and Plumlee, 2005; Dhaliwal, Heitzman, and Li, 2006; Hail and Leuz, 2009; Pastor, Sinha, Swaminathan, 2008; Chen, Chen, Wei, 2009; Boubakri, Guedhami, Mishra, Saffar, 2012), i.e. the ex ante rate of return implied in the contemporaneous stock price and analyst forecast data (RAVG). This proxy is based on accounting valuation models. I estimate the cost of equity based on four models: Gebhardt, Lee, and Swaminathan (2001), Ohlson and Juettner-Nauroth (2005), Claus and Thomas (2001), and Easton (2004), denoted as rGLS, rOJ, rCT, rES, respectively. These four models differ primarily in their assumptions about growth rates, forecast horizons, the incorporation of industry effects, and inputs. Moreover, the models are either based on the residual income valuation model or on abnormal earnings growth valuation models. As there is no consensus about which of the four models produces the most accurate cost of equity measure, following Dhaliwal et al. (2006), Hail and Leuz (2006), Ogneva, Subramanyan, and Raghunandan (2007), and Boubakri et al. (2012), I employ an equally weighted average of the four models. Averaging the four models’ cost of equity estimates reduces the possibility of spurious results stemming from a particular cost of equity capital model (Dhaliwal, et al., 2006). The cost of equity estimates of the individual models possibly exhibit different associations with a given risk proxy (Boubraki et al., 2012). GLS estimates, for example, are often negatively associated with long-term growth rates, whereas OJ estimates are often positively associated with long-term growth rates (Dhaliwal et al., 2006). Moreover, Botosan and Plumlee (2005) show that GLS estimates are inconsistently associated with some risk proxies, whereas Guay, Kothari, and Shu (2005) find that GLS estimates are the best predictor of expected returns.
Furthermore, the dependent variable relies on analysts’ forecast of the earnings per share. This data is obtained from the Institutional Brokers’ Estimate System (IBES). Not all firms are followed by analysts, which reduced the available data set on internal control disclosure. The data obtained from IBES is the one-and two-year ahead earnings per share forecast and the long-term growth rate. If the latter is not available for a certain firm, then the long-term growth rate got based on the difference in the earnings per share forecasts. In appendix A, I describe the way I calculated this long-term growth rate. Additionally, in appendix A, I summarize the four models, describe their key assumptions and data requirements, as well as explaining how I implemented the models and which numerical procedures I used to iteratively solve the models.
Table 1 shows the average of each cost of equity measure per country, as well as the average of all four methods. The overall average cost of equity estimate is 13.0%, which is similar to values reported by prior research (e.g., Boubraki et al., 2012; Hou, van Dijk, and Zhang, 2012). Moreover, contrary to for example Boubraki et al. (2012), maximum values for the cost of equity estimates in my sample do not exceed 100%, as I stopped iterations at this level. However, the four individual estimates
12 Table 1. Average cost of equity capital estimates per country (%)
Country Observations 𝑟𝐺𝐿𝑆 𝑟𝑂𝐽 𝑟𝐶𝑇 𝑟𝐸𝑆 𝑟𝐴𝑉𝐺
Australia 113 8.9 18.3 12.7 16.4 14.1
Austria 29 9.8 16.8 11.4 13.0 12.8
Brazil 89 10.5 17.1 14.0 14.0 13.9
Czech Republic 1 3.8 17.4 12.0 16.6 12.4
Denmark 83 8.9 14.7 12.1 13.6 12.3
Finland 121 10.1 17.2 11.6 15.3 13.5
France 198 8.9 13.9 9.4 12.7 11.2
Germany 213 9.2 17.3 11.0 15.6 13.3
Greece 83 9.0 16.1 13.9 14.6 13.4
Hungary 26 8.7 20.8 15.0 13.4 14.5
India 103 8.9 17.1 15.9 12.8 13.7
Indonesia 57 11.4 18.3 20.0 15.7 16.4
Italy 112 7.5 16.4 10.7 14.0 12.1
Japan 373 4.6 12.8 7.5 12.3 9.3
Malaysia 88 9.1 16.5 13.6 13.9 13.3
Mexico 48 10.5 14.1 14.7 13.1 13.1
New Zealand 22 8.2 13.9 9.5 12.4 11.0
Poland 40 8.7 23.6 8.8 21.0 15.5
Russian Federation 6 16.1 16.4 15.3 13.1 15.2
Singapore 32 8.8 18.1 12.9 15.2 13.8
South Africa 55 15.0 21.0 19.3 16.8 18.0
South Korea 62 8.5 19.8 15.3 16.1 14.9
Spain 59 7.9 12.8 10.7 11.6 10.7
Sweden 95 9.4 16.0 10.8 14.3 12.6
Switzerland 121 7.8 12.9 9.0 11.8 10.4
Taiwan 12 11.2 11.4 10.7 11.3 11.2
Thailand 82 12.1 17.4 14.2 15.4 14.8
Turkey 64 13.1 23.1 17.3 16.3 17.5
United Kingdom 243 25.5 13.9 12.0 12.6 16.0
Sample weighted average 10.3 15.9 11.8 13.9 13.0
This table presents the average implied cost of equity capital estimate, as well as the average of the four implied cost of equity capital estimates for each of the countries included in the sample. For the sample weighted average, weights were determined based on the number of observations per country. The average number of observations per country is 91.
do come close to this artificial limit. The minimum values for 𝑟𝐺𝐿𝑆 , 𝑟𝑂𝐽, 𝑟𝐶𝑇, 𝑟𝐸𝑆, and 𝑟𝐴𝑉𝐺, are respectively 0.001, 0.003, 0.002, 0.002, and 0.030, these values are generally lower than for example Boubraki et al. (2012), yet the sample used for this study is larger. Finally, all the cost of equity estimates correlate positively to each other, as can be seen in table 2. In line with prior research, 𝑟𝑂𝐽
and 𝑟𝐸𝑆 are highly correlated (e.g., El Ghoul, Guedhami, Kwok and Mishra, 2011). 𝑟𝐺𝐿𝑆 generally has lower correlations than the other implied cost of equity estimates and correlates most with 𝑟𝐶𝑇, this is also in line with prior research (e.g., El Ghoul et al., 2011).
13 Table 2. Implied cost of equities correlation matrix
Probability 1 2 3 4 5
1 RAVG
2 RCT 0.51
3 RGLS 0.42 0.25
4 ROJ 0.87 0.20 0.03
5 RES 0.87 0.24 0.04 0.89
This table presents the correlation coefficients between four implied cost of equity estimates and the average of these four. RAVG is the average of the four implied cost of equity estimates. RCT is the implied cost of equity calculated with the Claus and Thomas (2001) method. RGLS is the implied cost of equity calculated with the Gebhardt et al. (2001) method. ROJ is the implied cost of equity calculated with the Ohlson and Juetner-Nauroth (2005) method. RES is the implied cost of equity calculated with the Easton (2004) method.
Boldface type indicates statistical significance at the 1% level.
Despite the improvements over other cost of equity measures, such as the ex post realized returns, the analyst-forecast based implied cost of equity measure suffers from some limitations as well. First, they require the assumption that analyst forecasts are reasonable proxies for the market’s expectations of future earnings, which is not always the case (e.g., Frankel and Lee, 1998; Easton and Sommers, 2007). Often, analysts’ forecasts are overly optimistic when compared to current accounting data (Easton and Sommers, 2007). However the extent to which analysts are overly optimistic varies and Frankel and Lee (1998) find evidence that analysts tend to be more overly optimistic for firms with high stock to book price ratio. Second, due to the data requirement of availability of analyst forecasts, the sample only includes firms with analyst following. Third, the models suffer from measurement error due to the analyst forecasts which are slowly updated. In an attempt to overcome the last issue, I use analyst forecast ten months ahead of the measurement moment, following Hail and Leuz (2009).
3.3. Independent variables
3.3.1. Voluntary internal control disclosure
Voluntary disclosure studies have generally relied on the disclosure index developed by CIFAR (e.g., Francis, Khurana, Pereira, 2005; Hope, 2003). However, the CIFAR disclosure index comprises voluntary as well as mandatory disclosure items. When including mandatory disclosure, I would measure a country-level characteristic, which is more difficult to change for a firm (a firm would have to relocate or cross-list, and it is debatable whether firms would go that far to change their level of mandatory disclosure). Therefore, the CIFAR disclosure index is an inappropriate measure of internal controls disclosure for this study. Moreover, the CIFAR data is relatively old, as it has been collected in the 1990s.
14 This thesis relies on unique, hand-collected data on voluntary internal control disclosures, collected in the years after the change in demand of corporate disclosure as a consequence of large financial accounting scandals. Only just very few studies have researched internal control disclosures.
These studies have used self-constructed disclosure indices. Following prior research, I use a self- constructed disclosure index (ICD) developed by Hooghiemstra, Hermes and Emanuels (2013). Seven items have been selected to be included in this index. This selection reflects public policy reports about corporate governance and internal control (e.g., COSO 2004, FEE, 2005; IFAC, 2006) and a comprehensive review of prior research on internal controls. Consequently, annual reports were examined in order to identify the presence or absence of these disclosure items. Because this index is based on annual reports, it is very appropriate for this study, as this is a key source of information for investors. Finally, firms were assigned one point for the presence of each item in the annual report. All items are equally weighted, thus, consisting of seven items, the index measures values ranging from 0 to 7.
3.3.2. Investor protection
Following prior research (e.g., Engelen and van Essen, 2010; Boulton, Smart, Zutter, 2011), my proxy for investor protection (IP) is the anti-director rights index (ADRI) measure developed by Djankov, La Porta, Lopez-de-Silanes and Shleifer (2008). This index represents a measure of legal protection of minority shareholders against expropriation by corporate insiders. Six items have been selected for the inclusion in this index. These items are regulations protecting minority shareholders, such as the right to vote by mail, the right to call a meeting for shareholders owning less than 10% of the capital and the right for shareholders to hold the first opportunity to buy newly issued stock. The commercial code laws of the countries got examined for the presence of each of these six items. A country got assigned one point for the presence of a regulation protecting minority shareholders or half a point for the presence of a substituting regulation, which partly protects minority shareholders, of each item in the country code law. For this index all items are equally weighted, and therefore each country got assigned a value ranging from 0 to 6. The index is available for 72 countries, among which all of the countries that are included in my dataset.
3.3.3. Legal Enforcement
In their article, Lopez de Silanes et al. (1998) created an enforcement index. This index is based on five variables, namely the efficiency of the judicial system, rule of law, corruption, risk of expropriation (outright confiscation or nationalization) and risk of contract repudiation by the government. As the first three variables more directly relate to the enforcement of regulations, and the latter two deal with a more general stance of the government towards business, I decided to base my legal enforcement measure only on the first three variables, thereby following Leuz, Nanda and Wysocki (2003). The efficiency of the judicial system is an assessment of the “efficiency and integrity of the legal
15 environment as it affects business, particularly foreign firms.” The rule of law variable presents an assessment of the law and order tradition in the country produced by the country risk rating agency International Country Risk (ICR). Corruption is ICR’s assessment of the corruption present in the government, whereby a lower score indicates the presence of briberies. All three variables can have values ranging from 0 to 10. The legal enforcement measure is an equally weighted average of these three variables. The construction of the legal enforcement measure (ENF), based on these three variables, can be found in appendix B.
3.3.4. Control variables
Using solely these variables would bias the results, as prior research has found firm and country characteristics which affect the cost of equity capital. I conduct regression analyses controlling for traditional risk and country factors.
Consistent with prior research with regard to the cost of equity capital, I use a country level control variable (INFL) to control for differences in the inflation rate (e.g., Hail and Leuz, 2009). Analysts’
forecasts are expressed in nominal terms and local currency. This implies that the analysts’ estimates for the forecasted earnings per share reflect the expected inflation rates. I obtained inflation rates for the different countries from World Bank databases for the three years in which I have observations.
Furthermore, I control for the traditional risk proxy firm size (SIZE), as prior research suggests that firm size negatively relates to the cost of equity, thus larger firms generally have a lower cost of equity capital (e.g., Botosan and Plumlee, 2002). To measure firm size I used the natural logarithm of the firm’s total assets. Total assets is measured in U.S. dollars. This data is obtained from ORBIS.
Additionally, I control for the firm’s financial leverage (FLEV). The firm’s financial leverage position reflects the risk it takes, the more leveraged a firm, the higher the risk. Investors recognize and price this risk, hence firms with a higher leverage are expected to have a higher cost of equity capital, as previous research has shown (e.g., Hail and Leuz, 2009). Financial leverage is measured as the firm’s outstanding liabilities divided by the total assets the firm possesses. This data is obtained from ORBIS.
Prior research has also shown that the firm’s cost of equity is influenced by the performance (ROE) of the firm (e.g., Berger and Patti 2005, Hillman and Dalziel 2003). Shareholders show a preference, reflected in the cost of equity capital, toward better performing firms. The return on equity (ROE) measures the firm’s efficiency of generating profits for every unit of shareholders’ equity. ROE reflects the firm’s profitability, operating efficiency, and the financial leverage, as ROE is the product of a firm’s profit margin, total assets turnover and the equity multiplier. The ROE is obtained from ORBIS.
3.4. Statistical analysis
My empirical model to analyze the association between internal controls disclosure, investor protection and legal enforcement on the one hand and the cost of equity on the other, using Ordinary Least Squares regression analyses, is as follows:
16
RAVGi,j,k = 𝛽0 + 𝛽1ICDi,j,k + 𝛽2 IPj,k + 𝛽3 ENFk + 𝛽4 ICD*IPi,j,k + 𝛽5 ICD*ENFi,j,k + 𝛽6
SIZEi,j,k+ 𝛽7 FLEVi,j,k + 𝛽8 ROEi,j,k + 𝛽9 INFLj,k + 𝛽10 Y2005i,j,k + 𝛽11 Y2006i,j,k
+ Ԑi,j,k
In this equation, RAVGi,j,k is the average of the four implied cost of equity capital estimates of firm i for year j in country k; Ԑi,j,k is the error term; and all other variables have been defined already.
In the alternative measures and estimation techniques section, I check for robustness by using different measures for the dependent and independent variables. Moreover, in this section, I estimate the above model with a lagged dependent variable, I perform separate regression analyses for all three years separately, and I use panel estimation techniques in an attempt to find fixed effects.
4. Results
4.1. Descriptive analysis
Table 3 shows the correlations between the dependent and independent variables used in the regression model. The correlation coefficients are generally low, which indicates that the likelihood of multicollinearity issues is low. One coefficient value exceeds 0.7, namely the correlation between enforcement and inflation. The reason for this correlation is unclear and prior research does not report this correlation. Moreover, a correlation of 0.49 between the dummy year variables exists. Based on these correlation coefficients I do not expect a large impact from multicollinearity issues.
Table 3 Correlation matrix
Probability 1 2 3 4 5 6 7 8 9 10
1 RAVG
2 SIZE -0.18
3 FLEV 0.11 0.31
4 ROE -0.06 0.05 0.06
5 INFL 0.26 -0.15 -0.03 0.07
6 Y2005 -0.09 -0.03 0.01 -0.02 -0.08
7 Y2006 -0.04 0.00 0.00 0.00 0.06 -0.49 8 ICD 0.11 0.06 0.03 0.04 0.03 -0.08 0.01 9 IP 0.05 0.03 -0.07 -0.01 0.03 0.01 0.02 0.19
10 ENF -0.18 0.02 0.02 -0.05 -0.77 0.03 0.01 0.05 -0.02
17 This table presents the correlation coefficients between the dependent and independent variables. RAVG represents the average of four implied cost of equity capital estimates. SIZE is the natural logarithm of the firm's total assets. FLEV is the amount of financial leverage, as measured by total liabilities divided by total assets. ROE measures firm performance by the return on equity. INFL is the country level inflation rate, obtained by Worldbank. Y2005 (Y2006) is a year dummy, which takes the value of 1 in 2005 (2006) and 0 otherwise. ICD measures to which extent management voluntarily discloses internal controls information, with values ranging from 0 (no items disclosed) to 7 (all items disclosed). IP is the anti-directors right index developed by Djankov et al. (2008). ENF is the level of legal enforcement based on Lopez de Silanes et al. (1998), for further information with regard to the construction of this measure, I refer to the Appendix B.
The correlation coefficients are presented below the diagonal. Boldface type indicates significance at the 1% level.
4.2. The effect of internal controls disclosure, investor protection and enforcement on cost of equity
In this section I empirically analyze the effect of internal controls disclosure, investor protection and legal enforcement on cost of equity capital. Table 4 presents the coefficients of the different regression variables and the standard errors (in parentheses) from ordinary least squares regressions. In models (1), (2), and (3), I present analyses that include the set of control variables and ICD, IP, and ENF, respectively. According to model (1), there appears to be a significantly (p < 0.01) positive association between internal controls disclosure and the cost of equity capital. This result is in contrast with hypothesis 1, and therefore this hypothesis is rejected. A possible explanation for a positive relationship here is that internal controls possesses more strategic information than general voluntary disclosures. As Deumes and Knechel (2008) pointed out, disclosures of internal controls disclosures could provide competitors with information about the firm’s key risks, how the firm manages these and provides information on the firm strategy. This can hurt the firm’s position in product markets so much, that the benefit of the reduced information asymmetry between investors and managers is not worth disclosing this kind of information. Investors might perceive the disclosure of strategic information bad for the firm performance and penalize the firm in the return they demand on their investment. Hence, the disclosure of voluntary internal controls might reduce the estimation risk for investors, but this effect gets overshadowed by the spilling over of crucial information on how the firm manages its key risks to competitors. Consequently, this result suggests that the general voluntary disclosure literature is not applicable to the voluntary disclosure of internal controls.
In model (2), I find that the level of investor protection also shows a significantly (p < 0.1) positive relationship with the cost of equity capital. This result, too, contradicts the proposed hypothesis, namely hypothesis 2. The coefficient, as well as the statistical power of this relationship, however, is weaker than the previous. So the presence of this relationship is less certain compared to the previous relationship. Additionally, in model (3), the relationship between enforcement and cost of equity
18 capital is weakly positive and not statistically significant. Therefore, it seems like the two variables are not related. In model (4), which includes the direct effect of all three independent variables on the cost of equity capital, the previous results are being reinforced; a significantly (p < 0.01) positive relationship between internal controls disclosure and the cost of equity capital, a significantly (p < 0.1)
Table 4. Ordinary Least Squares Regressions of return on equity on internal controls disclosure. investor protection and legal enforcement
(1) (2) (3) (4) (5) (6) (7)
Intercept 17.063*** 17.430*** 17.971*** 16.407*** 21.059*** 18.161*** 21.518***
0.967 1.049 1.421 1.487 1.517 1.952 2.283
SIZE -0.638*** -0.625*** -0.635*** -0.666*** -0.623*** -0.662*** -0.650***
(0.064) (0.065) (0.063) (0.063) (0.064) (0.063) (0.063) FLEV 0.061*** 0.062*** 0.066*** 0.067*** 0.062*** 0.066*** 0.067***
(0.007) (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) ROE -0.017*** -0.017*** -0.017*** -0.017*** -0.017*** -0.017*** -0.017***
(0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) INFL 0.610*** 0.610*** 0.658*** 0.600*** 0.596*** 0.596*** 0.588***
(0.051) (0.051) (0.079) (0.080) (0.051) (0.080) (0.080) Y2005 -1.487*** -1.634*** -1.864*** -1.772*** -1.517*** -1.733*** -1.778***
(0.289) (0.291) (0.284) (0.283) (0.290) (0.283) (0.282) Y2006 -1.311*** -1.367*** -1.515*** -1.484*** -1.344*** -1.450*** -1.514***
(0.277) (0.279) (0.272) (0.270) (0.277) (0.271) (0.271)
ICD 0.371*** 0.359*** -0.889*** 0.225 -1.107**
(0.070) (0.070) (0.333) (0.329) (0.469)
IP 0.212* 0.300** -1.044*** -0.887***
(0.126) (0.130) (0.317) (0.341)
ENF 0.022 -0.042 -0.127 -0.102
(0.099) (0.099) (0.188) (0.188)
ICD * IP 0.308*** 0.322***
(0.080) (0.085)
ICD * ENF 0.020 0.017
(0.040) (0.040)
# 2529 2529 2461 2461 2529 2461 2461
R-squared 0.132 0.123 0.141 0.154 0.137 0.152 0.159 This table presents the results based on Ordinary Least Squares (OLS) regression. In all of the regressions included in this table RAVG is the dependent variable, which represents the average of four implied cost of equity capital estimates. SIZE is the natural logarithm of the firm's total assets.
FLEV is the amount of financial leverage, as measured by total liabilities divided by total assets.
ROE measures firm performance by the return on equity. INFL is the country level inflation rate, obtained by Worldbank. Y2005 (Y2006) is a year dummy, that takes the value of 1 in 2005 (2006) and 0 otherwise. ICD measures to which extent management voluntarily discloses internal controls information, with values ranging from 0 (no items disclosed) to 7 (all items disclosed). IP is the anti- directors right index developed by Djankov et al. (2008). ENF is the level of legal enforcement based on Lopez de Silanes et al. (1998), for further information with regard to the construction of this measure, I refer to the Appendix B.
***, **, and * indicate significance at the 1%, 5%, and 10% level respectively. Standard errors are in parenthesis.
19 positive relationship between investor protection and the cost of equity capital, and no significant relationship between legal enforcement and the dependent variable.
In the following set of regressions (i.e., models (5) and (6) in table 4) I assess the interaction effect of internal controls disclosure and investor protection and legal enforcement on the cost of equity capital. In model (5), the coefficient of internal controls disclosure is significantly (p < 0.01) negative, which is in line with hypothesis 1. The coefficient for investor protection in this model is also significantly (p < 0.01) negative, which is in accordance with hypothesis 2, namely that the level of investor protection relates negatively to the cost of equity capital. Moreover, hypothesis 3, which states that the combined effect of investor protection and internal controls disclosure has a stronger negative relationship to the cost of capital, does not find support in this model. On contrary, the relationship turns out to have a significant (p < 0.01) positive relationship to the cost of equity. This implies that when investor protection is low (high), the negative relationship of internal controls disclosure on the cost of equity is stronger (weaker). This could mean that a substitution effect between the two variables is present. Hence, when investor protection is high the negative effect of voluntarily disclosing information on the cost of equity becomes weaker. Moreover, the reasoning that an investor with control is more in need of information does not find support. Rather, internal controls information is more relevant to an investor who has little control, as the results suggest. Perhaps, investors with relatively more control have a lower need for internal controls information, because this kind of investor might have more knowledge of the internal controls as they are more involved in the firm. Hence, an investor who has less control has a higher need to obtain internal controls information from the annual report. In conclusion, the results of model (5) could imply that internal controls disclosure as well as investor protection relate negatively to the cost of equity, however the two substitute each other.
In model (6) I analyze the relationship between internal controls disclosure and legal enforcement, and these two combined, with the cost of equity capital. In this model, it turns out that none of these relationships are significant. Hence, in model (6), I do not find support for hypothesis 1, 4, and nor for hypothesis 5.
Additionally, in model (7) I analyze all of the previously mentioned relationships in one model. In this model, the relationship between internal controls disclosure and the cost of equity capital is significantly (p < 0.05) negative, which is in line with model (5) and supports hypothesis 1, however this result is striking with model (1) and (4). Investor protection appears to have a significantly (p <
0.01) negative association with the cost of equity capital. This is in line with hypothesis 2 and model (5), however this result is striking with models (2) and (4). Moreover, in this model I find a weak negative association between legal enforcement and the cost of equity capital, however this relationship has little statistical power and is not significant. The interaction effect of internal controls
20 disclosure and investor protection has a significantly (p < 0.01) positive association with the cost of equity capital, this effect was observed in model (5) already and seems to be confirmed by this model.
This information leads me to confirm that the opposite of the hypothesized effect can be found empirically, thus the interaction effect of internal controls disclosure and investor protection associate significantly (p < 0.01), positively to the cost of equity capital. Finally, this model confirms model (6) in the way that it shows a non-significant positive relationship between internal controls disclosure combined with legal enforcement and the cost of equity capital.
Lastly, all of the control variables find empirical support for their expected association with the cost of equity capital. Firm size shows a significant (p < 0.01), negative coefficient in models (1) to (7), implying that larger firms have a lower cost of equity capital. This result confirms findings in prior research (e.g., Botosan and Plumlee, 2002). Financial leverage, which is the degree to which a firm is debt-financed, shows a significant (p < 0.01), positive coefficient in the models (1) to (7). This indicates that shareholders investing in firms with a larger, relative amount of debt financing are exposed to a higher risk, compared to investing in firms that attract relatively less debt capital. Hence, as prior research already suggested (e.g., Hail and Leuz, 2009), shareholders require a higher return on their investment when the financial leverage is higher. Firm performance, as measured by the return on equity, shows a significant (p < 0.01), negative relationship to the cost of equity capital in models (1) to (7). This implies that investors require a lower return on their investment when they invest in firms with a higher firm performance than when they invest in firms with lower firm performance. This confirms the research by, e.g., Berger and Patti, (2005) and Hillman and Dalziel (2003). Finally, the country-level variable inflation shows a significant (p < 0.01), positive association towards the cost of equity capital in models (1) to (7). This result is in line with Hail and Leuz (2009).
5. Alternative measures and estimation techniques
5.1. Alternative measures for ICD, IP, and ENF
Because the relationships between ICD, IP, and ENF and the cost of equity were not in line with the hypothesized relationship based on theory, or not consistent in the different models, I analyzed the same set of relationships, replacing their measures by alternative measures.
First, I replaced the dependent variable for the four individual cost of equity measurement methods. This provided me with non-consistent results with regard to the dependent variables. This is probably due to the fact that the individual models rely on different assumptions and structurally exhibit different associations with certain risk proxies (Boubraki et al., 2012). Therefore, in line with prior research (e.g., Dhaliwal et al., 2006; Hail and Leuz, 2006; Ogneva et al., 2007; Boubraki et al.,
21 2012), I support my initial measure of cost of equity, namely the average of the four different estimation methods.
Next, I replaced the independent variables by alternative measures in models (8) to (14). The results of this are shown in table 5. I replaced the ICD measure for a disclosure dummy variable, which equals 1 if the firm’s ICD score is greater than the sample’s median score and 0 otherwise.
Table 5. Ordinary Least Squares Regressions of return on equity on internal controls disclosure, investor protection and legal enforcement
(8) (9) (10) (11) (12) (13) (14)
Intercept 17.889*** 17.032*** 17.880*** 16.039*** 1.709*** 1.794*** 1.636***
(0.946) (1.025) (1.036) (1.116) (1.716) (1.052) (1.263) SIZE -0.628*** -0.664*** -0.613*** -0.655*** -0.669*** -0.626*** -0.656***
(0.064) (0.063) (0.065) (0.064) (0.063) (0.064) (0.064) FLEV 0.061*** 0.066*** 0.061*** 0.064*** 0.065*** 0.608*** 0.064***
(0.007) (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) ROE -0.017*** -0.016*** -0.017*** -0.016*** -0.017*** -0.017*** -0.017***
(0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) INFL 0.611*** 0.694*** 0.647*** 0.765*** 0.685*** 0.625*** 0.763***
(0.0511) (0.055) (0.069) (0.073) (0.055) (0.069) (0.074) Y2005 -1.528*** -1.868*** -1.597*** -1.768*** -1.782*** -1.516*** -1.765***
(0.290) (0.282) (0.290) (0.282) (0.283) (0.290) (0.289) Y2006 -1.324*** -1.532*** -1.343*** -1.537*** -1.514*** -1.311*** -1.532***
(0.277) (0.271) (0.278) (0.271) (0.270) (0.277) (0.271)
ICD 0.775*** 0.794*** 0.273 0.403 0.253
(0.236) (0.233) (1.024) (0.409) (1.028)
IP 0.338*** 0.316** 0.233 0.266
(0.130) (0.13) (0.184) (0.187)
ENF 0.157 0.337* 0.114 0.242
(0.205) (0.204) (0.270) (0.269)
ICD * IP 0.133 0.085
(0.233) 0.241
ICD * ENF 0.331 0.160
(0.306) (0.308)
# 2529 2469 2529 2469 2469 2529 2469
R-squared 0.125 0.143 0.121 0.149 0.148 0.125 0.149
This table presents the results based on Ordinary Least Squares (OLS) regression. In all of the regressions included in this table, RAVG is the dependent variable, which represents the average of four implied cost of equity capital estimates. SIZE is the natural logarithm of the firm's total assets.
FLEV is the amount of financial leverage, as measured by total liabilities divided by total assets.
ROE measures firm performance by the return on equity. INFL is the country level inflation rate, obtained by Worldbank. Y2005 (Y2006) is a year dummy, that takes the value of 1 in 2005 (2006) and 0 otherwise. ICD is a dummy variable, which measures to which extent management voluntarily discloses internal controls information, with values of 1 if the firm has an ICD score higher than the median score and 0 otherwise. IP is the revised anti-directors right index by Spamann (2010). ENF is the level of legal enforcement based on the rule of law index developed by Kaufmann et al. (2009).
***, **, and * indicate significance at the 1%, 5%, and 10% level respectively. Standard errors are in parenthesis.
22 Models (8) and (11) yield similar results as models (1) and (4), namely they associate a significantly (p
< 0.01) positive relationship between internal controls disclosure and the firm’s cost of equity capital.
Where model (5) and (7) exhibited a negative relationship between ICD and cost of equity capital, this relationship loses statistical power and is positive in models (12) and (14).
To test the model’s sensitivity to my choice of investor protection variables I replaced these by the revised investor protection measure by Spamann (2010). In model (9) the result of model (2) gets confirmed, i.e. there appears to be a significant (p < 0.01), positive relationship between investor protection and the cost of equity capital. Therefore, I can conclude that hypothesis 2 is not supported by the empirical analysis.
In the models presented in table 5, I considered the rule of law measure by Kaufmann, Kraay and Mastruzzi (2009) as an alternative measure for legal enforcement. In model (10) it again appears like there is a positive relationship between legal enforcement and the cost of equity. However again this relationship is not backed up by statistical power and therefore I conclude that the two do not possess a strong relationship. The same thing can be observed with regard to the interaction effect of internal controls disclosure and enforcement on the cost of equity in models (13) and (14).
5.2. Alternative estimation techniques
Finally, in this section I describe the results of a number of sensitivity checks with regard to the estimation methods. First, I will present the results of regressing the three years separately. Then, I will analyze the effect of replacing the dependent variable by a lagged dependent variable. Finally, I will examine the data for fixed or random effects.
In models (15) to (21), which are presented in table 6, I regress the same models as in table 4 (i.e.
models (1) to (7)). However, this time I regress the models for every year separately to investigate whether the found effects deviate over time. To save space and provide a better overview I put the regressions of the separate years in one table. For example, in the column of model (15) one can observe the outcome of three regressions, namely a regression model which includes internal controls disclosure in 2005, one in 2006, and another regression of this relationship in 2007. In order to save space, the table omits the results of the control variables. All control variables showed significant relationships, apart from some instances where the performance measure ROE did not yield significant results. In model (15), the result that internal controls disclosure positively relates to the cost of equity capital finds confirmation; in all three years there appears to be a significantly (respectively, p < 0.01, p < 0.01, p < 0.1) positive relationship.
The results with regard to investor protection differ in model (16) in comparison with model (2).
The outcome which suggests that a positive relationship exists between investor protection and the cost of equity capital does not find support in all of the three years, as the results show that relationship
23 in 2005, 2006, and 2007 appears to be, respectively, significant (p < 0.05), negative, positive, and significant (p < 0.01), positive. Thus, although interpreting the results it seems more likely that a negative relationship exists, the empirical evidence is mixed.
Similar to model (3), in model (17) I investigate the direct relationship of legal enforcement with the cost of equity capital. In model (3), I observed a positive relationship without strong enough statistical power. The same applies to 2005 and 2006 in model (17), whereas in 2007 the coefficient becomes negative, however, again without strong statistical power.
Table 6. Ordinary Least Squares Regressions of return on equity on internal controls disclosure, investor protection and legal enforcement
(15) (16) (17) (18) (19) (20) (21)
ICD 2005 0.425*** 0.483*** -1.138** 0.523 -1.093*
(0.096) (0.083) (0.466) (0.436) (0.627)
ICD 2006 0.378*** 0.346*** -0.444 0.459 -0.850
(0.103) (0.106) (0.498) (0.490) (0.708)
ICD 2007 0.299* 0.239 -0.835 -0.101 -1.183
(0.158) (0.165) (0.720) (0.708) (1.005)
IP 2005 -0.378** -0.047 -1.884*** -1.376***
(0.186) (0.164) (0.420) (0.404)
IP 2006 0.082 0.196 -0.848* -1.023*
(0.187) (0.198) (0.480) (0.526)
IP 2007 0.804*** 0.673** -0.239 -0.206
(0.262) (0.285) (0.709) (0.779)
ENF 2005 0.132 0.037 0.063 0.014
(0.147) (0.146) (0.257) (0.255)
ENF 2006 0.073 0.002 0.045 0.112
(0.156) (0.156) (0.280) (0.280)
ENF 2007 -0.034 -0.083 -0.293 -0.325
(0.193) (0.193) (0.408) (0.410)
ICD * IP 2005 0.394*** 0.375***
(0.111) (0.003)
ICD * IP 2006 0.204 0.328**
(0.118) (0.131)
ICD * IP 2007 0.260 0.235
(0.176) (0.190)
ICD *ENF 2005 -0.005 0.001
(0.051) (0.052)
ICD *ENF 2006 -0.010 -0.021
(0.059) (0.059)
ICD *ENF 2007 0.053 0.059
(0.087) (0.087)
Controls Yes Yes Yes Yes Yes Yes Yes
Total obs. 2529 2529 2461 2461 2529 2461 2461
