The impact of corporate disclosure through social media on the cost of equity capital : an archival study on US-listed firms

(1)

Amsterdam Business School

The impact of corporate disclosure through social media

on the cost of equity capital

An archival study on US-listed firms

Name : Egehan Yildirim

Student number : 10432574

Date : June 16th₂₀₁₆

Supervisor : ir. drs. A.C.M. de Bakker

Word count : 16,166

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

(2)

Statement of Originality

This document is written by student Egehan Yildirim, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

Technological developments of the last couple of decades made changes in corporate disclosure an interesting topic to conduct research to. This development has also changed everyday life, for example with the rising phenomenon of social media. With a sample of 100 US-listed firms, it is investigated whether corporate disclosure through social media has an effect on firms’ cost of equity capital. Drawing back on previous literature on corporate disclosure and social media, it is hypothesised and found that corporate disclosure through social media has a reducing effect on the cost of equity. However, no significant interaction of this effect with analyst following frequency and corporate governance quality was reported. Finally, the study shows that there is a clear difference between the effects of interactive and non-interactive social media platforms.

Keywords: corporate disclosure, social media, analyst forecast frequency, corporate governance

(4)

4.3 Descriptive statistics ... 30 4.4 Correlation analysis... 33 4.5 Multicollinearity test ... 36 5 Regression results ... 37 5.1 Hypothesis testing ... 37 Hypothesis 1 ... 37 Hypothesis 2 ... 39 Hypothesis 3 ... 41 Hypothesis 4 ... 43 5.2 Sensitivity analysis ... 45 6 Conclusion ... 47 7 References ... 50

(5)

1 Introduction

In recent years, corporate disclosure has been widely investigated, since changes in macroeconomic forces such as the rapid technological innovation and the globalization had created new opportunities for research in this field in the past decades. However, the traditional ways of firms’ communication regarding their performance and corporate governance have been widely studied, and most of their determinants have been confirmed.

Corporate disclosure generally solves for problems related to information asymmetry, with a distinction in information problems and agency problems (Healy & Palepu, 2001). As a result, firms’ true value is possibly not reflected by the share price, which affects investor decisions. Furthermore problems occur as business owners are often not participating (actively) in firms’ management, which creates asymmetry in the knowledge on information between owners and managers.

1.1 Towards social media as a disclosure channel

It was innovation that Healy & Palepu (2001) described as one of the macroeconomic forces that still had unanswered questions. Innovation has not only changed the effect on corporate disclosure, but it has also changed the communication channels that are being used by corporations. Nowadays, a large amount of time spent on social media is seen as the standard. Similar to the increase in the everyday use of social media compared to regular media, the communication of firms also tends towards an increased use of social media (Nair, 2011). Additionally, there are firms that use these outlets as an investor-relations tool (Saxton, 2012). As a result, the SEC provided guidance on the use of social media by investment advisers in 2012:

“Reflecting the fact that many registered investment advisers and their personnel use social media in various forms to communicate with existing and potential clients and to promote their services, the SEC staff recently issued a National Examination Risk Alert providing suggestions for complying with the antifraud, compliance, and recordkeeping provisions of the federal securities laws.

(6)

This is the first time that the SEC staff has provided guidance concerning the use of social media by advisers.”1

The SEC’s definition of social media is: “an umbrella term that encompasses various activities that integrate technology, social interaction and content creation. Social media may use many technologies, including, but not limited to, blogs, microblogs, wikis, photos and video sharing, podcasts, social networking, and virtual worlds.” 2

One year after the provision of this guidance, the SEC recognised social media as an official outlet for corporate announcements. Their requirement is that investors are alerted about these channels:

“The Securities and Exchange Commission today issued a report that makes clear that companies can use social media outlets like Facebook and Twitter to announce key information in compliance with Regulation Fair Disclosure (Regulation FD) so long as investors have been alerted about which social media will be used to disseminate such information.”3

As the SEC started to allow companies to use social media in announcing firm specific information that is in compliance with Regulation Fair Disclosure, companies, amongst whom Netflix was one of the first, have been using social media networks to disclose their information (Lee et al., 2015). Other than using social media for marketing and public relations purposes, it has now also become an outlet of corporate disclosure. Despite the growing use of corporate social media, research focus is mainly on commercial use of social media, and fewer studies have been conducted on the effects on e.g. earnings announcements and investor relations.

1.2 The link to the cost of equity

The consequences of this new outlet of corporate disclosure, its use, and the comparison to traditional ways of corporate disclosure are interesting topics of research. In a large number of studies to the traditional outlets of (voluntary) corporate disclosure, relations with agency, information and corporate governance problems, and cost of equity problems in general have

1_{See :}

http://www.pwc.com/us/en/financial-services/regulatory-services/publications/sec-social-media-guidance.html

2_See:

(7)

been investigated (Yi et al., 2011; Henry, 2010; Bauwhede and Willekens, 2008; Botosan, 1997).

Social media can serve the reduction of information asymmetry though transferring corporate information to stakeholders in a timely manner (Uyar & Boyar, 2015). The nature of its communication is different than those of the traditional outlets, and could therefore have an additional influence on premiums requested by investors as a result of information asymmetry. This paper aims to discover whether corporate disclosure by means of social media platforms by corporations can be ‘game changing’ within the area of cost of capital. As it is intended to obtain a significant finding on the relation between corporate use of social media and the reduction of cost of capital, the following research question is:

Does corporate disclosure of firms through social media result in a decrease of cost of equity?

The main contribution to existing literature is especially on the link between cost of equity and corporate disclosure. Whereas these studies are mainly focused on traditional outlets of corporate disclosure, such as annual and quarterly reports, this paper focuses on a complete new kind of outlet, which has not yet been investigated on its consequences on the cost of equity, and is therefore unique in its field.

The research question is investigated in an archival, quantitative study on 100 US-listed firms for the years 2013 and 2014. The influence of corporate disclosure through social media on the cost of equity are based on the PEG-method, the cost of equity estimation as developed by Easton (2004). Controlling for beta, firm size, return on assets, and leverage, it is first hypothesised and found that the cost of equity decreases by engaging in corporate disclosure through social media. Additionally, the next two hypotheses test whether this effect contains an interaction for analyst forecast frequency and corporate governance quality. Neither of these effects are found. Finally, the study hypothesises and finds that interactive and non-interactive social media platforms have different impacts on the cost of equity. A significant reducing impact on the cost of equity was found for non-interactive social media, but not for interactive social media.

The findings of this paper imply that social media is a source of corporate disclosure that reduces information asymmetry, which results in less risk exposed on the shareholder, therefore resulting in a lower risk premium incorporated in the share price. Hypotheses 2 and 3 being

(8)

insignificant furthermore shows that corporate disclosure through social media cannot be treated homogenously with other channels of corporate disclosure. Furthermore, the last finding suggests that non-interactive social media channels are more decision useful to investors than interactive social media channels, and that firms’ main purpose of the use of interactive social media channels does not yet include communication with investors.

1.3 Structure of the paper

This paper is structured as follows. First chapter two provides the literature study and the resulting hypotheses. In chapter three the methodology is explained. Then, chapter four describes the data, sample, and descriptive statistics. Chapter five covers the analysis. Finally, the conclusion is provided in chapter six.

(9)

2 Literature review & hypotheses

As explained by Healy and Palepu (2001), corporate disclosure generally arises from, and solves for two existing problems in capital markets: the information problem (i.e. adverse selection) and the agency problem. These occurring problems gave rise to the two objectives of (financial) reporting: decision usefulness and stewardship (Wagenhofer, 2014). Both of these problems create costs which are incorporated into the demanded return by shareholder.

This chapter will first describe these two problems leading to corporate disclosure and includes the efficient market theory and the agency theory. Then it discusses existing literature on cost of equity and corporate disclosure, including a separate section on social media. These sections form the basis for the hypothesis development, with which the chapter is ended.

2.1 Information problem – adverse selection

A firm’s insiders and outsiders do not have access to the same business information. Information asymmetry arises when one of the parties in a relationship has favourable information over the other party (Yi et al., 2011). As more specifically explained by Healy and Palepu (2001), managers have superior information over potential outside investors on the true firm value, and the firm’s expected future performance. The authors explain that it is desirable to diminish this gap in knowledge between two parties because it is likely to result in the undervaluation of good projects and the overvaluation of bad projects. These issues arise from the problem that investors cannot distinguish between good and bad projects whenever there is no (high quality) information available4.

The information problem relates to the decision usefulness role of reporting. Wagenhofer (2014, p. 350) describes this role of reporting as “informing the capital market participants.” In his study, he underlines that information is decision useful when the providers of capital are able to assess how a company’s performance was for a certain period. According to Connelly et al. (2010), corporate disclosure affects investors’ decisions by sending out a signal on firm quality and real firm value. Additionally, Korajczyk et al. (1991) point out that during periods in which earnings and other relevant information are being made public, the level of information asymmetry reaches its lowest point. Lang and Lundholm (1996) have a similar finding. They point out that whenever a firm is close to issue securities, there is a change in the

(10)

disclosure quality observed by analysts. These findings imply that investors have a need for information as a result from information asymmetry. This implications is in line with the statement of Healy and Palepu (2001) that the demand for corporate disclosure partly arises from information asymmetry.

Furthermore, Barth et al. (2013) explain that the investors of firms with less transparent earnings may have to acquire private information. The (high) costs resulting from this process will then be incorporated into the demanded return. Therefore, the authors conclude that the cost of capital will be reduced with more transparency, implying that transparent information provide more decision useful information to the providers of capital.

Another factor that is influential for the decision usefulness of corporate disclosure is the timeliness of announcements. General findings regarding this factor are that higher timeliness of earnings results in lower information asymmetry, together with the cost of equity (Francis et al. 2004; Frankel & Li, 2004).

2.2 Efficient market theory

The information problem is directly related to the efficient market theory. According to the efficient market theory, it is believed that the prices of securities traded at capital markets fully reflect all of the information that is publicly known about the underlying instrument, at all times (Basu, 1977). This theory implies that changes in security prices reflect the publicity of new information. Therefore, information asymmetry can be seen as the result of private information that is not available to traders on the capital markets. The amount of undisclosed information is reflected in the risk premium, as investors require a higher risk premium when information asymmetry increases, as explained by Healy & Palepu (2001). They further explain that disclosures are essential for capital markets to function well. Management’s behaviour and a firm’s policies towards corporate disclosure are possible solutions for this problem (Healy & Palepu, 2001; Welker, 1995).

Analyst forecast

Information with substantial consequences on share prices is usually widely known before earnings announcements (i.e. price leads earnings). This makes disclosure through earnings announcements, such as quarterly reports and annual reports, not the only important component of a firm’s information environment. Frankel and Li (2004) add analyser following and

(11)

corporate news as significant sources in their study to the characteristics of firms’ information environment and information asymmetry. This implies that analyst following is related to information asymmetry. Marquardt and Wiedman (1998) explain that these forecasts should include all available information to the market. This implies that an inaccurate forecast will be due to either low quality or low availability of publicly disclosed information by a firm. Therefore, firms with a better information environment will have a smaller gap between publicly known information and private information.

Botosan’s (1997) study looks at the relation between the disclosure level and cost of equity capital. In her study, she includes analyst following intensity, and reports a difference between frequently analysed firms and infrequently analysed firms. Her explanation is that the types of disclosure communicated by firms with infrequent analyst following, are mainly forecast information and key nonfinancial statistics. For firms that are being analysed more frequently, historical summary information was more beneficial, given that their firm already had a strong forecast environment.

Additionally, it is found that a common type of asymmetry in available information is reflected in the dispersion found between predictions of different analysers in their capital market forecasts. Lang et al. (2004) find that whenever there is more accuracy in analyst forecasts, there is less asymmetry in the beliefs of investors about performance of the firms. The potential benefits of better corporate disclosure revealed by these authors are described as the reduced estimation of risk, and a reduction in the cost of capital.

2.3 Agency theory and agency costs

In addition to the asymmetry of information regarding investment choices of potential investors, agency problems also occur. These problems exist as the owners of a firm (i.e. shareholders) are in most cases not part of the firm’s management. The agency theory describes the problems that result from this separation of ownership and control between the principal (e.g. shareholders) and the agent (e.g. management). Jensen and Meckling (1976) state that it is likely that the agent will not always act in line with the principal’s best interest, because these interests are not perfectly aligned with his own. The arisen agency problems relate to the stewardship role of reporting, which Wagenhofer (2014) explains as the evaluation of performance by shareholders. Evaluation of performance, for example by focusing on

(12)

incentives, is necessary to track whether or not management acts in in line with the interest of the shareholders.

Jensen and Meckling (1976) give two explanations for the existence of agency problems. The first of these two explanations is that both parties within a principal-agent relationship are ‘utility maximizers.’ This means that both parties try to maximize their personal returns whenever this is possible. The second explanation is that there is a misalignment between the parties’ interests. Managers may have the incentive to retain information, which makes it more difficult and less effective for investors to monitor them (Karamanou & Vafeas, 2005). The costs that are incurred for the limitation of these problems, thus regarding the reduction of misalignment between management and investors, are the agency costs of equity. The agency problem and arising costs will remain until there is no longer a difference in interest between both parties at all. In addition Jensen and Meckling (1976) underline that agency problems may arise in any situation including two or more people/parties working together. Therefore, these problems do not necessarily occur only in situations with a clear principal-agent relationship.

Agency costs of equity

Agency costs are a result of information asymmetry, which in turn means that firms with a higher degree of information asymmetry are undervalued due to the investors’ required risk premium (Broberg et al., 2010). Therefore, agency costs are being incorporated in the share price. Jensen and Meckling (1976) build the agency costs of equity out of the following components: (1) monitoring costs incurred by the principal, (2) bonding expenditures, and (3) the residual loss. They describe monitoring costs as those costs that are being incurred in order to lower the divergence in interest that exists between the principal and the agent, which is an attempt to limit the activities done by the agent that are not in the shareholders’ best interest. Bonding costs are expenditures made by the principal, in order to guarantee that the agent will not take actions in the principal’s disfavour (Jensen & Meckling, 1976). Finally, these authors state that mitigating the interests between the principal and the agent is only possible to a certain extent and a perfect alignment between interests is not possible. The remaining misalignment is being described as the residual costs.

(13)

Corporate governance

Corporate governance plays a big role in many studies regarding agency cost issues (Kent & Steward, 2008; Karamanou & Vafeas, 2005; Bauwhede & Willekens, 2008; Chen et al., 2003). For example, Chen et al. (2003) state that it is clear that the cost of equity is being influenced by corporate governance. In regard to the agency theory, Kent and Stewart (2008) state that firms that are well-governed increase the frequency of their corporate disclosure. This should lead to the existing ‘information gap’ between management and shareholders being mitigated, which in its turn results in lower agency costs.

Regarding corporate disclosure, Karamanou and Vafeas (2005) find that well governed firms protect their shareholders from surprises by means of their disclosure policy. Furthermore, Bauwhede and Willekens (2008) use the agency theory in their investigation of disclosures on corporate governance in the European Union. In this study, they find that companies disclose corporate governance information so that they can reduce both information asymmetry and agency costs. This finding shows the three-way relationship between information asymmetry, agency costs, and corporate disclosure.

Additionally, Ashbaugh et al. (2004) report a significant association between a firm’s corporate governance and its cost of equity. In their paper, they explain that transparent information and a higher number of disclosures help in the reduction of information risk exposed to the shareholder. Thus, they state that better corporate governance reduces the cost of equity within a firm by reducing the required risk premium. Kelton and Yang (2008) report similar findings. They show that well governed firms have differences in their disclosure transparency, which can be seen as a response to information asymmetry and related agency costs. These firms were more likely to engage in internet financial reporting as an additional outlet than poorly governed firms.

2.4 Corporate disclosure

There are several methods that make disclosure of information to an organization’s stakeholders possible. According to Sengupta (2004), the quarterly earnings announcements are one of the most highly anticipated events and receive the most media and investor attention. Frankel and Li (2004, p. 231) state that also articles in the popular press and corporate news releases are “significant sources of firm-specific information” and that these articles “can proxy for the resources devoted to private information collection.” Their results imply that the

(14)

timeliness of disclosure is important, and that sources other than earnings announcements are of substantial value for the reduction of information asymmetry as well.

Purposes and determinants of corporate disclosure

A big part of the main demand for financial reporting, thus corporate disclosure of business information, arises from information asymmetry according to Healy and Palepu (2001). They explain that problems of information asymmetry, and incentives of managers to retain this information, have negative influences on the efficiency of markets: these problems can be mitigated through disclosure.

Additionally, Grüning (2007) gives three main reasons for firms to provide more information. The first reason is a firm’s desire to improve information intermediation by financial analysts. This will be accomplished when risk estimation and information asymmetry is reduced. Second, the reduction of information asymmetry will also improve stock liquidity, because of the reduction in information asymmetry. Finally, lower information risk should lead to reduced capital cost. Thus, the use of disclosed information should provide investors and other stakeholders the ability to make healthy decisions regarding investments and other areas (Uyar & Boyar, 2015).

It can be assumed that there is a clear link between corporate disclosure and information asymmetry. Frankel and Li (2004) state that a timely disclosure of value relevant information by a corporation, and timely collection of information by related outsiders, will result in the reduction of information asymmetry. In their study, they find that both increased analyst following and increased financial statement informativeness have a negative influence on insider trades and the profit made on them. Broberg et al. (2010) show that information asymmetry leads to agency costs, which in turn means that firms with more information asymmetry are undervalued due to the following investors’ required risk premium. They further explain that a firm’s valuation improves whenever there it is more transparency, but that it also decreases capital cost, and increases the likelihood of investment. In addition, Lambert et al. (2007) find a direct link between the quality of a firm’s disclosures and their cost of capital. They show that higher quality disclosures affects the assessment of, for example, distribution of cash flows by market participants, and that it therefore may result in a lower cost of equity.

Other influential interrelated factors shown by Grüning (2007), are firm size, listing status, industry, and country of origin. These factors prove for the determinants of corporate disclosure

(15)

and its quality. Broberg et al. (2010) add that firms make a cost-benefit trade-off between the positive effects of disclosure and their costs, whenever these disclosures are voluntary.

Social media as an outlet

In a corporate setting, social media platforms are mainly used by firms for sales, marketing, research, and public relations purposes (Larcker et al., 2012). However, the wide range of followers that are being reached on social media show that it can serve more than one purpose. Saxton (2012) describes some advantages of firms’ intensive use of social media. Most importantly, he states that it enables firms to interact with their investors. This interaction makes it possible for these investors to obtain real‐time accounting information. Jung at al. (2014) document that about half of the firms using Twitter, partly use it for the announcement of their earnings. However, this happens inconsistently for good news and bad news.

Social media therefore has also changed the way in which firms communicate with their investors. Miller and Skinner (2015) explain that the recent changes in technology, capital markets, and media have affected firms’ overall disclosure policies. As one of the prominent changes, they give the emergence of social media and its use by companies in the process of corporate disclosure. The authors furthermore state that given the increasing use of social media together with the ability for users to create own content, social media seems to be an important new aspect to conduct research to.

The main difference between traditional outlets, such as newspapers, and social media outlets is that the latter makes it possible for firms to send multiple messages in a short amount of time, straight to (known) followers (Jung et al., 2014). A firm with a corporate social media account makes it possible for both investors and customers to have conversations among themselves and with the firm, while these are observable by all users of the platform (Lee et al., 2015). Corporations can have general social media accounts, but also accounts specifically for investors (Uyar & Boyar, 2015).

Uyar and Boyar (2015) state that one of the main advantages of social media is that it enables corporations to transfer their information to stakeholders in a timely manner, which helps decrease information asymmetry. They also state that social media changes the interaction with stakeholders and herewith has positive influences on transparency. Another advantage they expose is the fact that its usage is relatively inexpensive compared to the traditional outlets. Jung et al. (2014) further elaborate these advantages by explaining that costs drop because of the reduction in dissemination costs of information, and increasing speed and

(16)

flexibility for this dissemination. On the other hand, they state that acquisition costs for the receiver of information (e.g. investor) are being reduced. Saxton (2012) explains that investors may use these types of media as tools to conduct research, hold discussions, and obtain stock information.

Business Wire (n.d.) conduct a study on how firms may use social media successfully. In conjunction with Uyar and Boyar (2015), they state the enhancement of transparency as an important advantage. They explain this by underlining the enlargement of news visibility and the possibilities to better elaborate on (earnings) announcement by using blogs and videos (i.e. making it interactive). Business Wire’s (n.d.) report further finds positive effects on timeliness with the rise of possibilities to live tweet earnings calls, events, or breaking news. This is given to enhance the efficiency of capital markets (Uyar & Boyar, 2015).

In addition, it is important to note that social media cannot be treated homogenously, because of its many variants/platforms. Lee et al. (2015), make a distinction between interactive and non-interactive social media. Interactive social media, such as Twitter and Facebook, allow multiple users to respond to messages and directly interact with other users. In these cases, the firm may lose control over content because it is not always the messenger (Lee et al., 2015). These are the platforms that allow firm-investor interactions, as described by Saxton (2012). On the other side, the non-interactive social media are those where the firm does not lose control (i.e. it always is the provider of the news/announcements). These are outlets such as corporate blogs. Different impacts are documented between the two different types of social media by Lee et al. (2015).

2.5 Empirical studies

Many studies have been conducted on the relationship between corporate disclosure and the cost of equity. This section discusses earlier works on the effect of corporate disclosure on the cost of (equity) capital. Among these studies, Diamond and Verrecchia (1991) were one of the first researchers to investigate this topic. In their study, they conduct research to the causes and consequences of the liquidity of securities. Their results reveal that disclosing corporate information can result into a reduction of a firm’s cost of capital. The explanation for their findings is that corporate disclosure improves the future liquidity of a firm's securities by reducing the price impact, which results in the reduction of the firm’s cost of capital. In addition, they find this impact to be bigger for larger firms.

(17)

In addition to Botosan (1997), Botosan and Plumlee (2002) find that disclosure levels reduce the cost of equity also for frequently analysed firms. Furthermore, the results of their study indicate that disclosure type is critical, as their documented results are different for different outlets of disclosure (e.g. annual reports were found to have a different impact than quarterly reports). They reported both a positive and negative association in their study, as well as no association at all.

Francis et al. (2005) study firms in industries with great external financing needs. They find that firms engaging in more voluntary disclosure benefit from lower cost capital, both for debt and for equity. Contradicting their expectations, the incentives to voluntary disclose are independent from country-level factors. This is contradictory to earlier findings, for example to those of Chen et al. (2003), who report that country-level investor protection is important in reducing the cost of equity. The finding of Francis et al. (2005) thus implies that the reducing effect of disclosure on cost of equity is the same around the world. In a later study, Francis et al. (2008) test whether voluntary disclosures result in a difference in the cost of capital of firms. They report that the level of voluntary disclosure results in lower cost of capital, being unconditional on other factors.

Chen et al. (2003) test the effects of corporate disclosure and other corporate governance mechanisms on the cost of equity. They report that disclosure and other corporate governance mechanisms have a significant negative effect on the cost of equity capital, implying that corporate disclosures decisions are resulting from corporate governance. Furthermore, Chen et al. (2003) find that the reduction in cost of equity resulting from corporate disclosure is weaker when non-disclosure governance mechanisms are included in the regressions, suggesting a moderating effect of corporate governance mechanisms.

However also opposite effects are reported in these studies. For example, Diamond and Verrecchia (1991) report some situations, in which a reduction of information asymmetry is desired and results in increasing cost of capital. An explanation is that information asymmetry may lead to a more rapid exit by brokers. Furthermore, Botosan and Plumlee (2002) find that the cost of equity increases in some cases when disclosures get more timely. They explain that this finding is contrary to theory, but consistent with claims that greater timely disclosures may increase stock price volatility, and therefore result in higher cost of equity.

In addition, cost of capital is found to possibly increase when the disclosures result in a more asymmetric information environment, i.e. their absence means that the information environment is better (Kim and Verrecchia, 1994; Zhang, 2001). Zhang (2001) explains that

(18)

effects of disclosure level on the cost of capital can differ among firms, and it depends on multiple factors that cause variation samples. Given factors are “earnings volatility, variability of liquidity shocks facing investors, the cost of producing information privately, and the cost of disclosure to the firm” (Zhang, 2001, p. 375).

2.6 Hypotheses development

Based on the information in the previous sections, four hypotheses will be tested in this study. All hypotheses are formulated as the alternative hypothesis.

As described in the previous sections, previous literature has shown that information asymmetry, corporate disclosure and the cost of equity are interrelated. Information asymmetry arises investors’ need for information. The cost of equity increases due to information problems and agency problems, including the misalignment of interest between management and investors. Little transparency and low timeliness, for example, make information less decision useful, which in its turn causes investors to require a higher risk premium. Disclosure of corporate information is considered as a solution: its purpose is to reduce the gap of information between management and investor, thus also reducing cost of equity.

However, given the mixed results (e.g. Botosan and Plumlee (2002) and Zhang (2001) find that some disclosure result in higher cost of equity or have no effect at all), it can be assumed that every disclosure channel has a different impact on the cost of equity, and disclosure frequency may have an adverse effect on the cost of equity.

As explained before, social media makes the costs of information acquisition by investors more affordable. Given that also other sources than earnings announcements are of big value in reducing information asymmetry and drawing back on the advantages of social media as an outlet for corporate disclosure, such as increasing transparency, and enabling to provide information in a timely and inexpensive manner, one can predict that using social media as an outlet for disclosure purposes is likely to further reduce the cost of equity. Therefore, hypothesis one is formulated as follows:

H1: Corporate disclosure through social media by corporations results in a lower cost of equity

(19)

analysed infrequently. This relates to the information problem. Firms with an infrequent analysis need greater disclosure on forecast information and key nonfinancial statistics, whereas for intensively analysed firms, financial reports appear to be more important. This is because information asymmetry of the latter group is given to be substantially lower. Also, social media enables these organisations to reach and interact with a wide range of followers. The interactiveness and ease to reach the investor may have positive effects. Therefore, a logical prediction is that usage of social media will be in line with prior findings on firm analysis:

H2: There is a stronger negative association between cost of equity capital and corporate disclosure through social media for non-frequent analysed firms than for frequent analysed firms

Similar to the treatment of firm analyst following frequency is the quality of a firm’s corporate governance. It was found in earlier studies such as those of Chen (2003), that well-governed firms have lower agency costs than firms with lower corporate governance quality, and engage in corporate disclosure more often and with higher quality. It was found that the reduction in cost of equity resulting from corporate disclosure was moderated by governance mechanisms. This implies that the information asymmetry gap is bigger for firms whose corporate governance is of lower quality. Thus, a logical prediction is:

H3: There is a stronger positive association between cost of equity capital and corporate disclosure through social media for poorly governed firms than for well governed firms

As there are different types of social media, differences between each type may be expected as well. Interactive social media make it possible for messengers to communicate with the receiver, which makes it different from traditional media outlets. However, interactive social media’s main purpose for firms is not to communicate with investors. In addition, the firm is not always the messenger in interactive social media channels, and may therefore lose control over the content. Also, different impacts between both types of social media were reported by Lee et al. (2015) before. For these reasons, it can be predicted that both sources of social media create a different impact, also on the cost of equity.

(20)

H4: Corporate disclosure through interactive social media impacts the cost of equity differently than non-interactive social media does

(21)

3 Research methodology

This chapter describes the variables that are included in the later regression models. First the measures are identified and the choice for proxies and control variables is elaborated on. This is followed by a description of the regression models, which test the hypotheses in chapter 5. The chapter ends with table 1, which provides a summary of all variables that are part of the regression analyses.

3.1 Measures

This section describes the dependent, independent, and control variables, which are included in the regression analyses. As the effect on cost of equity is explained, cost of equity is the dependent variable. The independent variables are social media usage – both interactive and non-interactive, analyst forecasts, and corporate governance quality. Furthermore, a number of control variables are included.

Cost of equity – dependent variable

There has been much uncertainty about the way of measuring of cost of equity. The traditional method of determining the cost of equity is by using the Capital Asset Pricing Model (CAPM). However, researchers state that this method is inadequate in some cases, within which also corporate disclosure is incuded (Botosan, 1997; Gebhardt et al., 2001; Botosan, 2006; and Chen et al., 2009). One reason is that this model is focused on primarily ex-post data – such as realized returns – whereas studies on the cost of equity require an ex-ante approach – based on estimates (Gebhardt et al., 2001).

Chen et al. (2009) state that these ex-post methods in estimating the cost of equity may reflect multiple other elements, such as growth opportunities, growth rates forecast differences, and changes in risk aversion of investors. Furthermore, Botosan (1997), notes that the CAPM does not provide an adequate measure for disclosure studies, as “the CAPM assumes that cross-sectional variation in market beta alone drives variation in the cost of capital” (p. 337). Ex-ante measures on estimating the cost of equity on the other hand, determine the cost of equity that is implied in stock prices and earnings per share forecasts (Chen et al., 2009). Gebhardt et al. (2001) furthermore state that the CAPM is imprecise.

An ex-ante measure on estimating the cost of equity that was found to be adequate is the cost of equity estimated based on the Price/Earnings to Growth (PEG) ratio. This method of

(22)

obtaining cost of equity estimates was operationalised first by Easton (2004). One of the possible purposes given by Easton (2004) for the use of his model is to determine the effects of several matters on the cost of equity capital, including disclosure quality.

Although the PEG approaches of obtaining cost of equity estimates are seen as simplistic, it has been recommended in multiple studies. The model was included in the study of Botosan and Plumlee (2005), in which they test the reliability among multiple proxies for the determination of the cost of equity capital. The estimation based on the PEG ratio method was documented among the two most adequate methods. Furthermore, Francis et al. (2004) use the PEG approaches to test the sensitivity of their results and find that PEG approaches are overall scoring higher than residual income models (i.e. implied cost of equity).

The chosen approach is equation (12) from the Easton (2004) study:

𝑟𝑃𝐸𝐺 = √

𝑒𝑝𝑠2−𝑒𝑝𝑠1

𝑃0 (1)

where:

rPEG = expected rate of return, r is fixed constant and r > 0

P0 = current, date t=0, price per share

eps1 = expected, earnings per share, at date t=1

eps2 = expected, earnings per share, at date t=2

Second, equation (11) from Easton (2004) will be used to test the sensitivity to the specific proxy and is being modified for the dividends:

𝑟_{𝑀𝑃𝐸𝐺}2_{− 𝑟} 𝑀𝑃𝐸𝐺( 𝑑𝑝𝑠1 𝑃0 ) − 𝑒𝑝𝑠2−𝑒𝑝𝑠1 𝑃0 = 0 (2) where:

dps1 = expected, dividends per share, at date t=1

Both equations are included in Easton’s (2004) study in order to conduct the empirical analyses.

Common with Easton (2004), forecasts from December were derived for each sample year. Thus, for sample year 2013, the forecasts are from December 20th 2013. The eps1 anddps1

(23)

eps2’s forecasts are two years ahead, thus have December 20th 2015 as their ending date. The

share prices are those at the moment of estimation, which is December 20th of 2013 in the case of the first sample period.

Social media

Data for the independent variable social media were hand collected by the use of several tools, as there are no available datasets. All selected firms’ corporate disclosure messages through social media are being measured with a total of three dummy variables.

As hypotheses 1, 2, and 3 do not distinct between the different types of outlets, dummy variable DUM_SM captures whether or not firms engage in corporate disclosure through any kind of social media outlet. A firm labelled ‘0’ does not use any type of social media outlet for their corporate disclosure, whereas a firm labelled ‘1’ does engage in corporate disclosure through any type of social media, including Facebook, Twitter, and corporate blogs.

In the determination on whether or not a firm uses social media as a disclosure outlet, the research methods of the studies of Jung et al. (2014), Lee et al. (2015), and Uyar and Boyar (2015) are mostly followed. Common with Lee et al. (2015), Facebook and Twitter are treated homogeneously. Both outlets are classified as an interactive source of social media.

Unlike Lee et al. (2015), the classification is based on content, and not on the existence of a corporate social media account. This approach is similar to those of Jung et al. (2014) and Uyar and Boyar (2015). In the classification, first Facebook messages for all firms were downloaded. When it was found that a firm engaged in social media through Facebook, the firm was labelled ‘1’ in both DUM_SM and DUM_INTSM. For firms that did not use Facebook as a corporate disclosure channel, Twitter messages were retrieved. Firm engaging in social media through Twitter, are also labelled ‘1’ in both DUM_SM and DUM_INTSM. The remaining firms were sought for corporate blogs. Firms with a corporate blog meant for corporate disclosure are labelled ‘1’ in DUM_SM and ‘1’ in DUM_NINTSM. Firms with special investor websites, including extensive webcasts and board messages were also labelled ‘1’ in DUM_SM and DUM_NINTSM, although their investor web sites are not explicitly called a corporate blog (e.g. AT&T).

(24)

Intensity of analyst following

For analyst following, the most convenient measure is based on the number of analyst forecasts that are available for a firm’s earnings per share prediction. As described by Botosan (1997), the number of analyses is in line with the gap of information between the firm and investor. Therefore, Botosan’s (1997) measure for the extent of analyst intensity is determined by the number of analyst forecasts reported per year. However, unlike Botosan (1997), this study measures the number of analyst forecasts for earnings per share, rather than those for share prices. This is in line with the proxy used to estimate the cost of equity, in which earnings per share play a big role. The variable DUM_EST is a dummy variable, labelled ‘0’ if a firm has infrequent analyst following and ‘1’ if a firm has a frequent analyst following. Infrequent is defined as lower than the median of earnings forecasts, whereas frequent is defined as equal to the median or a higher value.

Corporate governance quality

The corporate governance quality measure is based on multiple studies to the role of corporate governance in the cost of equity (e.g. Stulz, 1999; Chen and Jaggi, 2000; Klapper & Love, 2004; Ashbaugh et al., 2004; Chen et al., 2009). The chosen measure for corporate governance quality that is most prominent in this research area is board independence and is proxied by ratio of board independence. Ashbaugh et al. (2004), for example, find that the cost of equity and the independence of the board are negatively correlated.

The main motivation is that the board of directors can be seen as management’s most direct monitoring body (Stulz, 1999). Board independence is also included as a proxy for corporate governance quality in many studies, among whom include Chen et al. (2009) and Harford et al. (2012). Furthermore, Chen and Jaggi (2000) note that board independence is associated with disclosure levels. Board independence is reflected by the ratio of total independent board members to total board members:

𝐵𝑜𝑎𝑟𝑑 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒 𝑟𝑎𝑡𝑖𝑜 =𝑇𝑜𝑡𝑎𝑙 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑏𝑜𝑎𝑟𝑑 𝑚𝑒𝑚𝑏𝑒𝑟𝑠 𝑇𝑜𝑡𝑎𝑙 𝑏𝑜𝑎𝑟𝑑 𝑚𝑒𝑚𝑏𝑒𝑟𝑠

The dummy variable DUM_GOV is created and included in the regression analysis. It is labelled ‘0’ when a firm is poorly governed, and labelled ‘1’ when a firm is governed well. Poorly governed firms have a board independence ratio lower than the median board

(25)

independence ratio, whereas well governed firms have a board independence ratio that is equal to or higher than the median board independence ratio.

Control variables

In order to further improve the study, control variables influencing the dependent variable are included. Based on the control variables of Botosan (1997), Botosan and Plumlee (2002), Francis et al. (2005), and Chen et al. (2009), market beta (BETA), firm size (LNTOTASST), return on assets (ROA), and leverage (CRTLEVMKT) are used as the factors to be controlled for, as these were found to have an influence on the cost of equity according to these studies. These studies find a positive effect of beta and firm size, and a negative effect of ROA and leverage on the cost of equity.

Firm size (LNTOTASST) is measured as the natural logarithm of total assets, whereas leverage (CRTLEVMKT) is measured the cube root of the debt-to-market value of equity ratio, because the natural logarithm was too strong of a transformation and made the variable skewed in the opposite direction. In conjunction with Francis et al. (2004), all financial information is from period t-1.

3.2 Regression models

This section describes the regression models for each hypothesis. As the study is a quantitative archival study, regression analyses are conducted on the data. Given the nature of the panel data, the hypotheses are all tested by a Generalized Least Squares (GLS) regression model. This model is a random effects model, which assumes that there are no fixed effects. The present fixed effects in the dataset are eliminated after data is being clustered by year.

Regarding hypothesis 1, the overall effect of corporate disclosure through social media on the cost of equity is tested. The expectation is that information announcements through social media as an outlet, results in lower cost of equity incurred by firms. As described in the previous section, the cost of equity is measured through Easton’s (2004) PEG ratio method, defined as rPEG. For this hypothesis, there is no distinction within the type or specific outlet of

the social media outlets. The regression model for H1 is formulated as follows:

𝑟_𝑃𝐸𝐺 = 𝛽₀+𝛽₁𝐷𝑈𝑀_𝑆𝑀_𝑡,𝑖+ 𝛽₂𝐵𝐸𝑇𝐴_𝑡,𝑖+ 𝛽₃𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑇_{𝑡−1,𝑖} + 𝛽₄𝑅𝑂𝐴_{𝑡−1,𝑖} +

(26)

Hypothesis 2 tests whether the effect of corporate disclosure through social media on the cost of equity is further influenced/moderated by the number of analyst forecasts on the earnings per share for the specific firm. The expectation is that there is a stronger negative association between cost of equity capital and use of social media for corporate disclosure for infrequently analysed firms than for frequently analysed firms. As splitting the sample up into two different sub-samples weakens the model and can result into misleading outcomes, an interaction effect is tested in hypothesis 2. The regression model for H2 is formulated as follows:

𝑟_𝑃𝐸𝐺 = 𝛽₀ + 𝛽₁𝐷𝑈𝑀_𝑆𝑀_𝑡,𝑖 + 𝛽₂𝐷𝑈𝑀_𝐸𝑆𝑇_𝑡,𝑖 + 𝛽₃𝐷𝑈𝑀_𝑆𝑀_𝑡,𝑖× 𝐷𝑈𝑀_𝐸𝑆𝑇_𝑡,𝑖+ 𝛽4𝐵𝐸𝑇𝐴𝑡,𝑖+ 𝛽5𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑇𝑡−1,𝑖+ 𝛽6𝑅𝑂𝐴𝑡−1,𝑖+ 𝛽7𝐶𝑅𝑇𝐿𝐸𝑉𝑀𝐾𝑇𝑡−1,𝑖 + 𝜀 (4)

Hypothesis 3 is similar to hypothesis 2. Instead of the number of analyst forecasts, now it is being tested whether the effect of corporate disclosure through social media on the cost of equity is further influenced/moderated by the specific firms’ corporate governance quality, which is proxied by board independence. The expectation is that there is a stronger positive association between cost of equity capital and use of social media for corporate disclosure for poorly governed firms than for well governed firms. Again, the hypothesis is tested by including an interaction term, rather than splitting up the sample. The regression model for H3 is therefore formulated as follows:

𝑟_𝑃𝐸𝐺 = 𝛽₀+ 𝛽₁𝐷𝑈𝑀_𝑆𝑀_𝑡,𝑖 + 𝛽₂𝐷𝑈𝑀_𝐺𝑂𝑉_𝑡,𝑖+ 𝛽₃𝐷𝑈𝑀_𝑆𝑀_𝑡,𝑖× 𝐷𝑈𝑀_𝐺𝑂𝑉_𝑡,𝑖+ 𝛽₄𝐵𝐸𝑇𝐴_𝑡,𝑖+ 𝛽₅𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑇_{𝑡−1,𝑖} + 𝛽₆𝑅𝑂𝐴_{𝑡−1,𝑖} + 𝛽₇𝐶𝑅𝑇𝐿𝐸𝑉𝑀𝐾𝑇_{𝑡−1,𝑖} + 𝜀 (5)

Finally, hypotheses 4 test for differences between the two types of social media: interactive and non-interactive social media. These include Facebook and Twitter (interactive), and corporate blogs (non-interactive). It is hypothesised that there is a substantial difference in the impact between both types of social media. In order to test for this hypothesis, the dummy variables DUM_INTSM and DUM_NINTSM are included in the regression model. Therefore, the following regression models results:

𝑟_𝑃𝐸𝐺 = 𝛽₀+𝛽₁𝐷𝑈𝑀_𝐼𝑁𝑇𝑆𝑀_𝑡,𝑖+ 𝛽₂𝐷𝑈𝑀_𝑁𝐼𝑁𝑇𝑆𝑀_𝑡,𝑖+ 𝛽₃𝐵𝐸𝑇𝐴_𝑡,𝑖+ 𝛽₄𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑇_{𝑡−1,𝑖} +

(27)

TABLE 1

Summary of variables included in the regression models

Variable Description

RPEG Proxy of the cost of equity, estimated using the PEG method

DUM_SM

Dummy variable labelled 1 if a firm engages in corporate disclosure through social media,

0 otherwise

DUM_INTSM

Dummy variable labelled 1 if a firm engages in corporate disclosure through interactive

social media, 0 otherwise

DUM_NINTSM

Dummy variable labelled 1 if a firm engages in corporate disclosure through

non-interactive social media, 0 otherwise

BETA Beta variable

LNTOTASST Total assets

ROA Return on assets

CRTLEVMKT Cube root of the leverage ratio

DUM_EST

Dummy variable labelled 1 if number of analysts forecasts is higher than or equal to the

median of analyst forecasts, 0 otherwise

DUM_GOV

Dummy variable labelled 1 if board independence ratio is higher than or equal to the median of board independence ratio, 0 otherwise

3.3 Sensitivity analysis

To assess and mitigate for concerns regarding the sensitivity to the used models and proxy for the cost of equity, and therefore accuracy and reliability of the regression outcomes, the analyses are also being conducted by taking equation (2) as the measure for the dependent variable cost of equity. By taking the rMPEG instead of rPEG, the results can be tested for the

sensitivity to the inclusion of the dividend estimations. This approach to test for the sensitivity is also applied by Francis et al. (2004). The resulting regression model is as follows:

𝑟_{𝑀𝑃𝐸𝐺} = 𝛽₀+𝛽₁𝐷𝑈𝑀_𝑆𝑀_𝑡,𝑖 + 𝛽₂𝐵𝐸𝑇𝐴_𝑡,𝑖+ 𝛽₃𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑇_{𝑡−1,𝑖}+ 𝛽₄𝑅𝑂𝐴_{𝑡−1,𝑖} +

(28)

4 Sample and data

This chapter discusses the chosen sample and the collected data. First, the motivation behind the chosen sample will be elaborated on. Then, the data sources and way of data collection follows. Finally, the descriptive statistics are covered.

4.1 Sample

Thus far, the SEC is the only financial services regulatory authority that has provided guidelines on the use of social media for the disclosure of corporate information. These guidelines have resulted in corporations’ ability to officially disclose corporate information through social media. In addition, US listed firms in general have more dispersed ownership in which agency issues occur, in comparison with, for example, Asian firms. Furthermore, US listed firms report in English, which enhances the understandability of the social media content. For these reasons, US publicly traded firms are chosen to conduct this archival study on.

The sample consists of 100 US publicly traded firms, which were randomly selected from 397 eligible firms. 2013 and 2014 were chosen as the sample years, as 2013 was the first year in which the SEC recognised social media as an official disclosure platform. Also, obtainable Twitter messages are restricted to the most recent 3,200 messages per account, which makes it impossible to retrieve messages older than 2013 for a majority of the firms. As a result, the study has been conducted on 200 sample years.

4.2 Data sources and collection

A wide range of data sources have been used in completing the dataset. To ensure that the same firms were obtained in the different outputs, the official tickers were selected in each dataset. Data on the earnings per share and dividend estimations – necessary for the determination of RPEG and RMPEG – and the number of analyst forecasts were derived from the Institutional Brokers' Estimate System (IBES) Summary database, which contains historical analyst forecast. A total of 3,353 firms that have forecasts on EPS1, EPS2, DPS1 in both 2013 and 2014 are found. In order to be consistent with the publication dates of the analyst forecasts, only firms with their fiscal year ending on December 31st were selected. This resulted in the ineligibility of 1,135 firms.

The historical stock prices and beta variables are obtained through the Center for Research in Security Prices (CRPS) database. For 891 firms, the CRPS database did not contain

(29)

the beta variable. Then, information on board independence is obtained from the Institutional Shareholder Services (ISS) Directors database. For 708 firms, the information on the board of directors was insufficient. Financial data needed to compute LNTOTASST, ROA, and CRTLEVMKT are obtained from Compustat North America. Firms with Global Industry Classification Standard (GICS) starting with ‘40’ (i.e. financial institutes) were removed, together with firms inactive in 2016, as it is highly probable that these firms have either aborted their social media account, or have integrated their social media accounts with their acquirer’s. Furthermore, firms with EPS1 higher than EPS2 have to be removed for RPEG and RMPEG to be calculable (see chapter 3.1.1). This resulted in 397 remaining firms, which were then further divided in deciles based on firm size (LNTOTASST). An overview of the sample selection procedure can be seen in table 2.

TABLE 2

Summary of Sample Selection Procedure

Firms with 2013's & 2014's EPS1, EPS2, and DPS1 forecasts in the IBES database 3,353

Eliminate:

Firms with fiscals years not ending on 31/12 - 1,135 Firms without available BETA in the CRSP database - 891 Firms without necessary director information in the ISS Directors database - 708

Inactive firms - 26

Financial institutions - 175

Firms with EPS1>EPS2 - 21

Total remaining firms 397

From each decile, 10 firms were randomly selected, resulting in 100 firms and 200 sample years. Therefore, the data is set to be panel data, as both firm observations and multiple sample years are part of the same database.

As far as research has led, there is no database that includes data on whether or not firms engage in corporate disclosure through social media. Therefore, the necessary data on social media usage are hand collected, and obtained by first searching the corporate web site for the firms’ social media account names. Then, the obtained messages are analysed on the basis of their content. Firms were also searched for possible social media accounts at the investor or corporate news level.

Facebook messages are downloaded through the Netvizz application (Rieder, 2013). Regarding Twitter, all tweets were obtained through the www.allmytweets.net web-application

(30)

and the DMI-TCAT tool (Borra & Rieder, 2014). The analysis of the downloaded posts is common with Uyar and Boyar (2015) and Jung et al. (2014): the obtained messages are searched through for the following key concepts: ‘investor,’ ‘earnings,’ ‘release,’ ‘quarter,’ ‘report,’ ‘corporate,’ ‘announcement,’ ‘financial,’ ‘dividend,’ ‘sustainability,’ fiscal,’ and their synonyms. Corporate blogs are searched for by the use of search engines. These blogs are not analysed on their full content, but the identification of a corporate blog with corporate disclosure as one of its purposes is sufficient for a firm to be labelled ‘1’ in DUM_SM and DUM_NINTSM.

After having gathered the data, the data are imported from MS Excel into the Stata software for further analyses. Due to a high number of outliers, which could distort the sample, the variables ROA and LEV have been winsorised. This means that observations in percentile 1 set equal to the smallest value of percentile 2, and observations in percentile 100 are set equal to the largest value of the percentile 99 (Veeman, 2013). Furthermore, data were clustered by year, on which therefore a Generalized Least Squares (GLS) regression was conducted.

4.3 Descriptive statistics

Table 3 provides an overview of the descriptive statistics for the sample, and includes the number of observations, mean, median, standard deviation, maximums, and minimums for every variable except social media, both in total and per year. In addition to the variables of table 1, the untransformed variables for total assets and leverage are included in this overview.

In general, the means and medians of the variables are relatively close to each other, which indicates that there are not many extreme outliers. However, the outcomes still show great dispersion in the data. RPEG reports a mean of 0.0915, whereas its minimum and maximum respectively are 0.0145 and 0.2997. Furthermore, the data show a wide range of firm size: total assets have a minimum of USD 347 million, and a maximum of USD 277,787 million. The big difference in the minimum ROA and maximum ROA is caused by the 8 sample year observations in which a negative net income is reported.

On average, for each sample firm there are 28,6 analyst forecasts on earnings numbers for both years. The most frequently followed observation contained 65 forecasts, whereas the least frequently followed observation was forecasted thrice. Furthermore, as shown in table 3 panel D, it can be seen that the means of RPEG are near to being equal for both frequently and infrequently analysed firms. For frequently analysed firms, the dispersion in RPEG is higher

(31)

show, frequently analysed firms engage more in corporate disclosure through social media than infrequently analysed firms: both overall as well as through interactive social media.

Moreover, the best governed firm contained only independent directors, while the worst governed firm’s ratio on independent to total directors is 0.412. Also, as shown in table 3 panel D, similar means for RPEG were reported for both well governed and poorly governed firms. However, unlike the frequently analysed firms, well governed firms have a smaller dispersion in the value of RPEG. Additionally, based on the DUM_SM and DUM_INTSM means, it is observed that well governed firms engage more in corporate disclosure through social media than poorly governed firms.

TABLE 3 Descriptive Statistics

Panel A: Full dataset

Variable n Mean Median Std. Dev. Min Max

RPEG 200 0.092 0.083 0.035 0.015 0.300 BETA 200 1.175 1.149 0.451 0.105 2.352 TOTASST 200 21144 6469 44058 347 277787 LNTOTASST 200 8.876 8.775 1.425 5.850 12.535 ROA 200 0.064 0.056 0.051 -0.112 0.202 LEVMKT 200 0.705 0.538 0.584 0.028 3.255 CRTLEVMKT 200 0.833 0.813 0.220 0.305 1.429 ESTIMATES 200 28.6 28 14.1 3 65 GOVNCE 200 0.828 0.875 0.102 0.412 1 DUM_EST 200 0.505 1 0.501 0 1 DUM_GOV 200 0.545 1 0.499 0 1

Panel B: Observations for 2013

RPEG 100 0.088 0.083 0.027 0.026 0.195 BETA 100 1.184 1.139 0.428 0.463 2.352 TOTASST 100 20543 6324 42729 347 272315 LNTOTASST 100 8.834 8.752 1.441 5.850 12.515 ROA 100 0.060 0.052 0.049 -0.112 0.191 LEVMKT 100 0.775 0.554 0.635 0.044 3.255 CRTLEVMKT 100 0.860 0.821 0.226 0.354 1.429 ESTIMATES 100 28.7 28 14.1 5 64 GOVNCE 100 0.827 0.8819 0.102 0.412 0.9231 DUM_EST 100 0.510 1 0.502 0 1 DUM_GOV 100 0.550 1 0.5 0 1

(32)

Table 3, continued

Panel C: Observations for 2014

RPEG 100 0.095 0.083 0.042 0.015 0.300 BETA 100 1.166 1.149 0.475 0.105 2.208 TOTASST 100 21745 6564 45555 430 277787 LNTOTASST 100 8.917 8.789 1.414 6.063 12.535 ROA 100 0.067 0.061 0.054 -0.112 0.202 LEVMKT 100 0.635 0.495 0.522 0.028 2.983 CRTLEVMKT 100 0.806 0.791 0.212 0.305 1.429 ESTIMATES 100 28.5 27.5 14.2 3 65 GOVNCE 100 0.829 0.875 0.102 0.412 1 DUM_EST 100 0.500 0.5 0.503 0 1 DUM_GOV 100 0.540 1 0.501 0 1

Panel D: differences among estimation frequency and corporate governance quality

Variable n Mean Std. Dev. Min Max

Infrequently analysed RPEG 99 0.0914 0.0335 0.0256 0.206 DUM_SM 99 0.3535 0.4805 0 1 Frequently analysed RPEG 101 0.0915 0.0372 0.0145 0.2997 DUM_SM 101 0.6337 0.4842 0 1 Poorly governed RPEG 91 0.0924 0.0378 0.0145 0.2997 DUM_SM 91 0.3516 0.4801 0 1 Well governed RPEG 109 0.0907 0.0333 0.0256 0.2026 DUM_SM 109 0.6147 0.4889 0 1

Table 4 then shows the social media usage by firms. In 2013, out of the 100 firms, 48 firms engaged in corporate disclosure though an interactive channel (either Facebook or Twitter) and 10 firms used only a source of non-interactive social media. A small increase in social media usage as a disclosure channel is observed for 2014, as 51 firms used social media as a disclosure channel during this year. Of these 51 observations, 9 firms had an active corporate blog but did not engage in any kind of corporate disclosure through interactive social media.

(33)

TABLE 4

Descriptive Statistics - Frequency table Social Media Usage

Panel A: Social media overall

2013 2014

n 100 100

0 52 49

1 48 51

Panel B: Interactive social media

2013 2014

n 100 100

0 52 49

1 48 51

Panel C: Non-interactive social media

2013 2014

n 100 100

0 90 91

1 10 9

4.4 Correlation analysis

Table 5 shows the Pearson correlations among the dependent, independent, and control variables that are used in all three empirical models. The Pearson correlation matrix calculates the separate correlation coefficients between all variables used in the regression models. The matrix shows that multiple variables are correlated. For example, there is a strong significant correlation between DUM_SM and DUM_INTSM, valued 0.8247, which could be expected due to the relatively small number of users of non-interactive social media. Although multicollinearity issues may arise when a Pearson correlation coefficient exceeds 0.8 (Botosan, 1997), these issues are not applicable to this dataset, as both variables are not being included in the same regression model. The figures of table 5 therefore indicate that there are no multicollinearity issues, as no other coefficients exceed 0.8. However, section 4.5 does an additional test for multicollinearity issues.

Also, it is found that social media usage has a strong positive correlation with both analyst forecast frequency and corporate governance quality, which suggests that frequently analysed firms and well governed firms are more likely to engage in disclosures through social

(34)

media. These findings are therefore consistent with the earlier assumptions made based on the descriptive statistics.

In contrast to findings of earlier studies, amongst whom Lang et al. (2004), Marquardt and Wiedman (1998), Bauwhede and Willekens (2008), and Ashbaugh et al. (2004), the correlation coefficient suggest that there is no significant relationship between neither forecast frequency nor corporate governance level and the cost of equity, as neither DUM_EST nor DUM_GOV show a significant correlation with RPEG.

Furthermore, all three the dummy variables for social media – overall social media usage, interactive social media usage, and non-interactive social media usage – show a negative relation with RPEG. However, the correlation coefficient for interactive social media is insignificant. An implication of this finding is that both type of outlets are complements rather than substitutes.

The impact of corporate disclosure through social media on the cost of equity capital : an archival study on US-listed firms