Name: Student number: Address: Tel. Number: E-mail: Supervisor: Co-assessor: Word count: 14 February 2018 Gert Kits S3270742 Molenweg 5, 9354 BN, Zevenhuizen 06 – 27032027 g.a.kits@student.rug.nl dr. K. Linke dr. N. Hussain 9.644 (Ch. 1 – 5)
Faculty of Economics and Business | University of Groningen
Master thesis | MSc Controlling
DETERMINANTS OF CORPORATE
DISCLOSURE VIA SOCIAL MEDIA
Abstract
This study examines determinants regarding the extent to which firms disseminate corporate information relevant for their shareholders and other investors via social media. Possible determinants are derived from (voluntary) disclosure literature and tested based on the Twitter use of 63 Dutch-listed firms, using a content analysis approach. Based on the results it could be concluded that profitability, industry, and to some extent leverage are important determinants which influence the level of dissemination of corporate information via social media. Additionally, based on the included control variables, firm size, a companies’ general involvement with social media, and to some extent seasonal effects are explanatory factors towards the dissemination level. The results were quite contradictory however, regarding the expectations derived from (voluntary) disclosure literature. This led to the conclusion that the corporate use of social media cannot be compared very well with voluntary disclosure. Instead, support is found for the assumption that Dutch firms employ impression management strategies in an aim to manage the image of the firm held by the public.
Keywords: social media, Twitter, content analysis, information dissemination, voluntary disclosure, impression management.
Content
1. Introduction ... 1
2. Theoretical background and hypothesis development ... 4
2.1 Background ... 4
2.2 Social media and voluntary disclosure ... 5
2.3 Prior research ... 6 3. Research methodology ... 12 3.1 Sample selection ... 12 3.2 Variable definitions ... 14 3.3 Statistical method ... 17 4. Results ... 18 4.1 Descriptive statistics ... 18 4.2 Regression analysis ... 19
4.3 Additional regression analysis ... 21
5. Conclusions and discussion ... 24
References ... 28
Appendices ... 31
Appendix A. Social media usage among Dutch-listed firms ... 31
Appendix B. Coding scheme used in the content analysis ... 32
Appendix C. List of the used Twitter accounts ... 33
1. Introduction
Over the last decade, social media has become a more and more important topic for companies around the world. The concept provides companies with the ability to communicate directly with a broad audience without intervention of intermediaries, such as the press, and is therefore a valuable tool. Research shows that companies are aware of the value of social media (Harvard Business Review Analytic Services, 2010). Two of the most popular platforms for companies are Twitter and Facebook. In the United States about half of the companies make use of either one of these two platforms (Zhou, Lei, Wang, Fan, & Wang, 2015; Jung, Naughton, Tahoun, & Wang, 2017). Based on a preliminary inquiry among Dutch-listed firms – performed in order to explore initial leads regarding this research – it can be said that nearly all the firms listed on the three main indices (93,3%) make use of a certain form of social media via a corporate social media account. A more detailed overview of social media usage among Dutch-listed firms can be found in appendix A.
As the adoption of social media by companies increased over the past years, this phenomenon gained more attention by academics. Its use and potential has been subject of quite an extensive body of literature. A great deal of this literature focuses on the value of social media for companies with regard to their consumers or employees (e.g. Luo, Zhang, & Duan, 2013; Kietzmann, Hermkens, McCarthy, & Silvestre, 2011; Kaplan & Haenlein, 2010). However, for this study it is another important – if not the most important (Friedman, 1962) – stakeholder that is subject of research, namely: the investor (or similarly: the shareholder).
Prior research has proven that firms quite actively use social media to disseminate relevant corporate information for investors (e.g. Zhou et al., 2015; Jung et al., 2017). In this context, relevant information for investors can be defined as the kind of information that enables investors to evaluate whether the management has managed the firms’ resources in the interests of the (external) investors (Healy & Palepu, 2001). Examples regarding this kind of information can be financial by nature, such as annual reports or earnings announcements, as well as non-financial, such as announcements concerning the launch of a new product or the appointment of a new CEO. Disseminating such information via social media can enhance the reduction of the information asymmetry between the firms’ management and the shareholders, leading to more activity on the capital market (Blankespoor, Miller, & White, 2014; Prokofieva, 2015). The interest of the current literature has for a great part been the effect of this kind of corporate use of social media. However, research regarding factors that influence or explain the use of
social media to disclose corporate information is rather scarce. Up to now, only a few studies are known to explicitly address explanatory variables regarding the level of dissemination of valuable corporate information via social media (Zhou et al., 2015; Yang, Liu, & Zhou, 2016; Jung et al., 2017; Yang & Liu, 2017). Yet, these researches currently provide rather limited or contradicting results regarding possible explanatory factors. Therefore, to extend the literature, this research further examines variables that explain the variation in the level of corporate information relevant for investors or shareholders shared via social media. To accomplish this, the main research question covered by this paper is as follows:
Which determinants influence the level of dissemination of corporate information via social media?
By investigating this question, this study aims to create better understanding towards the way in which firms use social media to share corporate information relevant for investors. This understanding can be useful for both the users of this information as firms that share, or consider to share, relevant information via social media. For the users (e.g. investors or shareholders), a better understanding as to why firms disclose certain information puts them in a better position to assess this specific information and hereby the firm in general. Regarding firms themselves, it is interesting to know how and why other companies use social media to disseminate corporate information, in order to make better informed decisions on whether or not to share certain information.
Moreover, this study further develops the understanding of corporate social media usage from a theoretical perspective. The rich (voluntary) disclosure literature is used in the search for possible determinants concerning the level of corporate information shared via social media, as will be further elaborated in the next chapter. By comparing corporate social media usage with voluntary disclosure literature and testing hypotheses derived from this field of knowledge, this research examines the extent to which this comparison is legitimate. Hereby, besides the main research question, the question whether sharing corporate information via social media can be seen as a form of voluntary disclosure is implicitly covered by this study as well.
As mentioned earlier, only a limited amount of studies are known to have studied corporate social media usage in an aim to explain the difference in levels of corporate information shared by different companies. These studies have only examined corporate social media usage in the United States (Zhou et al., 2015; Jung et al., 2017) and the United Kingdom (Yang et al., 2016; Yang & Liu, 2017). Because of this, this research investigates a different setting, namely The
Netherlands. This study is hereby the first known to examine corporate social media usage towards investors in a Continental European setting, providing an entirely new perspective on this phenomenon.
Additionally, prior research covers data up to the year 2014. Due to the quickly changing world of information technologies and the fact that social media is still a relatively new concept, findings concerning several years ago may be less relevant in the current time. By examining data regarding the years 2015 and 2016, this study therefore provides a more current and relevant perspective on how firms use social media to share corporate information.
The remainder of this paper is structured in the following way. Chapter 2 provides a theoretical background regarding corporate social media usage and elaborates on the presumed link between social media and voluntary disclosure. Based on this link, hypotheses are drawn building on the (voluntary) disclosure literature. In chapter 3, the sample and methodological approach to measure and test the different variables are explained. Chapter 4 discusses the results of the analysis, whereafter chapter 5 elaborates on the conclusions and discussion that can be derived from these results.
2. Theoretical background and hypothesis development
2.1 Background
The corporate use of social media towards investors or shareholders is a relatively new research topic that emerged over the last four years. Corporate disclosure via social media can be seen as a new stream among internet-based reporting literature. This literature originated since the late nineties of the previous century (e.g. Ashbaugh, Johnstone, & Warfield, 1999; Craven & Marston, 1999; Debreceny, Gray, & Rahman, 2002; Xiao, Yang, & Chow, 2004). The main focus of this literature has been the use of corporate websites as a platform to share information relevant for investors. However, with social media, recent technological developments provided a new internet-based reporting venue for companies.
In the current literature concerning the corporate use of social media towards investors, a distinction can be made between studies that address social media regarding so called user-generated content (UGC) and studies investigating a phenomenon that can be labelled as company-generated content (CGC). UGC concerns content generated by “regular people who voluntarily contribute data, information, or media that then appears before others in a useful or entertaining way, usually on the Web” (Krumm, Davies, Narayanaswami, & Chandra, 2008: p. 10). An example of UGC is an analyst informing investors of his or her opinion about the future of a certain share via social media, leading to activity of these investors in the form of replies via the specific medium. As the name does suggests, CGC is about the content that the companies create themselves and disseminate via their corporate social media account or any other online medium. CGC is therefore the counterpart of UGC. Despite the fact that social media in the form of UGC can be very important for companies in that it shows a strong relationship with stock performance (e.g. Chen, De, Hu, & Hwang, 2014; Yu, Duan, & Cao, 2013; Sprenger, Tumasjan, Sandner, & Welpe, 2014) or can influence investor perception (e.g. Trinkle, Crossler, & Bélanger, 2015; Kadous, Mercer, & Zhou, 2017), only CGC is relevant regarding the main question of this research, since the question specifically addresses the way in which firms use social media as a dissemination channel for corporate information.
Among the studies concerning the role of CGC on social media, in terms of disseminating corporate information potentially relevant for investors, another distinction can be made. On the one hand, present literature focuses on the effects of social media usage by firms in terms of stock performance (e.g. Blankespoor et al., 2014; Prokofieva, 2015, Bhagwat & Burch, 2016) or, closely related, investor perception (Cade, 2017). On the other hand, some studies have paid
attention to explaining variation in the level to which firms disseminate important corporate information via social media. Yet, this latter body of literature has been quite scant to date: only four studies are known to specifically address this specific topic (Zhou et al., 2015; Yang, et al., 2016; Jung et al., 2017; Yang & Liu, 2017). Empirical results have therefore been limited and the question as to what influences or can explain the level of corporate disclosure via social media remains far from answered (a review of existing literature will be presented in the third section of this chapter). As already introduced earlier, this research aims to contribute in answering this question by further examining determinants which influence the level of dissemination of corporate information via social media.
2.2 Social media and voluntary disclosure
In the search for possible answers regarding the main research question of this paper, this study builds on the assumption that disseminating relevant corporate information via social media can be seen as a form of corporate (voluntary) disclosure. This assumption may seem obvious, however, it asks for further explanation. Jung et al. (2017) discuss in their paper that the decision to disclose certain important information is in essence something fundamentally different compared to the decision to disseminate this information. They describe the term disclosure to be related to the question whether or not to share certain content (e.g. financial results or new strategic decisions) where the dissemination of the information is related to the way (e.g. press releases, conference calls, or social media) and extent (amount and timing) in which the information is distributed. As Jung et al. (2017) put even stronger: “the dissemination decision goes beyond the disclosure decision and reveals how firms try to control their information environment.” (p. 1). Yet, for this study the (voluntary) disclosure literature is used in the search for answers regarding the main research question. The reasoning behind this will now be explained in more detail.
Voluntary disclosures are defined by Meek, Roberts, and Gray (1995) as “free choices on the part of company managements to provide accounting and other information deemed relevant to the decision needs of users of their annual reports” (p. 555). The ‘free choices’ refer to the choices for firms to share certain information that exceed what is required by laws and regulations. Voluntarily reporting accounting information is based on the demand for such information by shareholders since the ownership and control are in essence separated (Beyer, Cohen, Lys, & Walther, 2010). This separation leads to information asymmetry between the shareholders and the management of the concerned company which in turn will lead to agency problems (Jensen & Meckling, 1976). By providing the shareholders (and investors in general)
with relevant information that surpasses what is required by the applicable laws and regulations, firms tend to lower this information asymmetry.
Voluntary disclosures are traditionally distributed via conference calls, press releases or (their own) websites (Healy & Palepu, 2001). However, with the emergence of social media a new disclosure tool has developed. In their study, Blankespoor et al. (2014) found significant changes on the stock market after a company had provided information about their earnings via social media (Twitter) and hereby proved as one of the first that social media is indeed an effective tool to inform the public and hereby lower the information asymmetry. Furthermore, they proved that this conclusion applied particularly well for the situations where traditional ways of dissemination had failed to lower this asymmetry (smaller/less visible firms).
Based on this conclusion and combining this with the fact that companies are not obliged to share relevant information via social media, this research builds on the assumption that sharing corporate information via social media relevant for investors is a form of voluntary disclosure. By this, it is to be expected to find valuable answers regarding the main research question in the corporate (voluntary) disclosure literature.
2.3 Prior research
As mentioned before, only a limited amount of studies have attempted to explicitly explain variances regarding the corporate use of social media towards investors. The first study known to address this topic was performed by Zhou et al. (2015). In their study they analysed the use of Twitter and Facebook as dissemination channels for disclosures among 9.861 U.S. firms. They found that Twitter ought to be the more preferred platform regarding corporate disclosures compared to Facebook. More interestingly in the context of this study, they discovered that the adoption rate and disclosure rate were different per industry and positively related to firm size. Jung et al. (2017) examined all firms included in the S&P 1500 index and also found the level of corporate use of social media to differ among different industries and firm size to be positively related to the level of dissemination of earning news via social media (Twitter in the case of their investigation). Besides this, they found audience size of the platform (number of followers) to be inversely related to the level of dissemination. Interestingly, they proved that variables that are commonly associated with the use of traditional disclosure outlets (e.g. conference calls, press releases, etc.) such as firm performance, growth and leverage are not related to corporate social media usage. The biggest takeaway of the study of Jung et al. (2017) however, is that they found evidence for the fact that firms are more likely to communicate
earnings news when this news is positive. Based on this evidence they inferred that firms try to control their information environment by strategically disseminating earnings news via social media. They investigated three possible incentives for firms to engage in strategic disseminating: higher litigation risk, lower investor sophistication, and the size of its social media audience. For all three incentives they found supportive evidence.
Despite being more focused on the effects of firms’ tweeting behaviour, concurrent work by Bhagwat and Burch (2016) also found support for the fact that firms (in the U.S.) strategically use Twitter to distribute earnings news. They found that “firms tweet strategically by increasing the frequency of financial tweets during and after the release of positive earnings results, suggesting they are aware of the impact such tweeting can have” (p 23). Furthermore, they proved that this strategic dissemination increased the magnitude of the announcement return: stock price reaction to positive earnings news will be more positive, and negative stock price reaction to negative earnings news will be more negative.
Similar results regarding strategic dissemination on social media were found in the U.K. by Yang and Liu (2017). Like the previously mentioned studies, this study also investigated Twitter and concluded that “firms tend to omit negative earnings news by posting a significantly lower volume of negative earnings-related tweets than positive earnings-related tweets” (p. 689). Furthermore, they found significant evidence for the assumption that improving performers (companies with increased profit before taxation relative to the prior year) are more likely to disseminate earnings news via Twitter in comparison to firms dealing with a declining performance. Hereby they concluded that firms try to construct a positive public image, which is known from impression management theory originating from (social) psychology (Goffman, 1959).
Lastly, Yang et al. (2016) investigated – again in the U.K. – whether corporate governance plays a role regarding a firms’ decision to disclose and disseminate corporate information on social media. Twitter usage was examined for corporate disclosures in the form of earnings announcements and narrative earnings tweets. To find the effects of corporate governance they investigated the influence of board size, board independence, gender diversity, and board effectiveness. Only board independence turned out not to be significantly related. Their findings suggest that better functioning corporate governance mechanisms increase the likelihood of adopting social media and helps enhance the dissemination of financial information. Besides these main variables of interest concerning corporate governance, Yang et al. (2016) found
average board age, firm size, profitability, and industry (all control variables) also to be significantly related to the extent of financial information shared via social media.
Concluding this review of the relevant prior literature, it can be said that only some factors influencing the corporate use of social media have been researched and proved yet. These factors include: firm size, industry, audience size of the platform, firm performance (however, inconsistent outcomes), direction of the news (positive news vs. negative news) and corporate governance (board composition and effectiveness). Furthermore, it is important to notice that prior research has been limited to studies in the U.S. and U.K. and empirical results are therefore quite narrow. Yet, the national setting is often found to be a quite influential factor regarding corporate (voluntary) disclosure due to cultural or legislative differences between countries (e.g. Meek et al., 1995; Jaggi & Low, 2000; Archambault & Archambault, 2003; Hope, 2003). It is therefore likely that different results may appear when investigating a different setting in comparison to the prior research. Based on these two conclusions, this research aims to contribute to existing literature by further investigating possible determinants regarding the level of dissemination of corporate information via social media in a new setting, namely The Netherlands. Building on this aim, the following section will elaborate on hypotheses that are tested by this study.
2.4 Hypothesis development
As introduced in the second section of this chapter (2.2), empirical and theoretical evidence regarding corporate (voluntary) disclosure is used in the development of multiple hypotheses. This body of literature has produced some quite extensively tested hypotheses which may provide answers as to what variables are able to explain the decision for firms to disseminate corporate information relevant for investors via social media.
Profitability
Profitability already has been investigated to some extent regarding its effect on the level of corporate disclosure via social media, as mentioned in the previous section. However, the results were inconsistent: Jung et al. (2017) did not find a relationship for data covering companies in the U.S., where Yang et al. (2016) did find evidence for a positive relationship in the U.K.. Besides this, Yang and Liu (2017) found that improving performers are more likely to use Twitter for disseminating earnings news. This ambiguity in the empirical results asks for further investigation and therefore the variable profitability is included in this research.
The voluntary disclosure literature explains a relationship between profitability and disclosure based on agency theory and signalling theory (Watson, Shrives, & Martson, 2002; Broberg, Tagesson, & Collin, 2010). When business is going well for a company, the management of the regarding company is more likely to share corporate information in order to ensure its position and remuneration (Singhvi & Desai, 1971; Inchausti, 1997).
As mentioned in the previous section (2.3), prior research finds that firms are more likely to disseminate positive information via social media and hereby try to control their information environment (Bhagwat & Burch, 2016; Jung et al., 2017; Yang & Liu, 2017). Looking at this conclusion, profitability can be seen in the same light since more profitable companies are likely to have more positive news or information to disclose. Therefore, it is possible that the explanation found in the disclosure literature is also applicable for dissemination via social media. In order to investigate this inference, and hereby trying further support the claim that firms try to control their information environment, the following hypothesis is tested:
H1a: Profitability is positively related to the level of dissemination of corporate information via social media.
This first hypothesis covers a rather static approach towards the effect of profitability (i.e. performance) on the level of dissemination. Additionally, it is interesting to look at a more dynamic measure as this potentially more accurately captures the tendency regarding to the performance of the firm. Drawing on the work of Yang & Liu (2017) and in line with the previously mentioned corporate disclosure literature this research further investigates the influence of profitability on the corporate use of social media by testing the following hypothesis:
H1b: Improving performers disseminate more corporate information via social media compared to declining performers.
Leverage
A second firm characteristic relevant in the context of this research is leverage. Jung et al. (2017) found this variable not to be related to corporate social media usage towards investors, in contrast to the existing literature regarding voluntary disclosure and disclosure outlets. Due to this inconsistency, it is interesting to examine the influence of this variable again.
Leverage is a quite prominent factor in the voluntary disclosure literature regarding its influence on voluntary disclosure behaviour by companies. Empirical research to date provides quite
strong evidence for the existence of a positive relation between the (relative) amount of debt and voluntary disclosure (e.g. Malone, Fries, Clarence, & Jones, 1993; Ismail & Chandler, 2005; Barako, Hancock, & Izan, 2006; Broberg et al., 2010). These results are consistent with the agency theory, predicting that firms with a higher amount of debt have to deal with increased agency costs (Jensen & Meckling, 1976). These increased agency costs will lead companies to provide more information in order to lower these costs.
Yet, it needs to be said that existing research is not entirely unanimous and this relation is therefore subject to some discussion (e.g. Meek et al., 1995; Watson et al., 2002). However, the positive relationship between the amount of debt and voluntary disclosure is more frequently and more recently observed. Therefore, this relationship is applied for this research in the context of corporate use of social media by investigating the following hypothesis:
H2: The debt ratio is positively related to the level of dissemination of corporate information via social media.
Industry
The third determinant that is examined by this study is industry. Prior research discovered a difference in corporate disclosure via social media among different industries (Zhou et al., 2015; Yang et al., 2016; Jung et al., 2017). This is not a strange outcome when comparing this to the (voluntary) disclosure literature. For example, Meek et al. (1995) and Watson et al. (2002) both concluded that the extent of disclosure varies among different industries.
Literature provides several reasons for this phenomenon. Healy and Palepu (2001) mention that competition influences the decision to disclose information since such disclosures may damage the competitive position of companies (known as the proprietary cost hypothesis). Since the level of competition may vary from one industry to another, it can be expected that the level of voluntary disclosure may vary as well. Another explanation mentioned in the literature is referred to as the bandwagon effect, which means that when one dominant firm in an industry shows high levels of disclosure, other firms may follow (Cooke, 1991; Watson et al., 2002). Based on this empirical and theoretical evidence, it is therefore interesting to examine whether differences in corporate disclosure via social media exist among industries in the Netherlands. In order to do this, the following hypothesis is tested:
H3: The level of dissemination of corporate information via social media varies between different industries.
Ownership
Building further on the effect of corporate governance on corporate social media usage, as discovered by Yang et al. (2016), it is interesting to investigate in what way ownership may influence the level of corporate information shared via social media. Jung et al. (2017) proved that the dissemination behaviour of firms differs depending on the level of investor sophistication, which highlights the possible influence of ownership on corporate disclosure via social media.
In order to investigate the importance of ownership, this study specifically tests the influence of ownership concentration. Ownership concentration and its effect on corporate (voluntary) disclosure in general has been studied quite extensively to date. García-Meca and Sánchez-Ballesta (2010) performed a meta-analysis regarding 19 empirical studies concerning ownership concentration and its effect on voluntary disclosure and concluded that ownership concentration is negatively related to the level of voluntary disclosure. This outcome supported the idea that, as they state themselves, “firms with high levels of ownership concentration are expected to have less information asymmetry, due to dominant shareholders having access to the information they need and, as a consequence, disclosing less information to the market” (p. 620). By this, the effect of ownership concentration can also be explained by the agency theory as mentioned earlier in the elaboration of the hypotheses concerning profitability and leverage. This hypothesis can be argued to be particularly relevant in the context of social media. Since social media consists of platforms which are easily accessible for small or individual investors, it can be very relevant for companies with dispersed ownership. Therefore, building on the hypothesis derived from the voluntary disclosure literature, it is likely that firms of which the ownership is more concentrated disclose less corporate information via social media. To test this assumption, this study investigates the following hypothesis:
H4: Ownership concentration is negatively related to the level of dissemination of corporate information via social media.
3. Research methodology
3.1 Sample selection
In order to analyse the determinants of the level of dissemination of corporate information via social media, this research investigates the use of social media by Dutch-listed non-financial companies for the years 2015 and 2016. The decision to investigate Dutch firms is elaborated in the previous chapter (section 2.3). Specifically listed firms are examined since it is assumed that for these companies sufficient information is available to properly conduct this research. Financial firms are not included in the sample due to the fact that the financial characteristics of these firms deviate substantially compared to non-financial firms, leading to possible biased results if included (Raffournier, 1995).
Dutch non-financial firms that are listed on one of the three main indices (AEX, AMX and ASCX) are included in the sample, leading to an initial sample consisting of 66 firms. The sample data includes data for the years 2015 and 2016, because these years are the two most recently ended years. Investigating the two most recent years is important since the current state of corporate social media usage among firms is the main interest of this study. The development of the corporate use of social media over the years is not within the scope of this research. Since corporate social media use towards investors is still a relatively new and emerging concept, it is likely that data originating from years prior to 2015 may blur the view of the current state of social media usage by companies towards investors. Data covering years prior to 2015 are therefore not included in this research. As companies usually report on a quarterly basis, quarterly data are sufficiently available. By this, quarterly data are used to create a more detailed dataset (in comparison to yearly data).
Based on a preliminary inquiry regarding the usage of different social media platforms among Dutch-listed firms (also mentioned in the first chapter), it can be concluded that LinkedIn and Twitter are the most prominently adopted platforms among Dutch-listed firms (see appendix A). Respectively 86,7% and 82,7% of the Dutch-listed firms have adopted LinkedIn and Twitter. Due to this clear prominence, it would be very interesting to look at both platforms. However, data collection efforts in a master thesis are constrained by available time resources, therefore only Twitter is investigated. Based on existing software, Twitter messages (tweets) can be retrieved efficiently. In contrast, for LinkedIn there is no software known to be available and therefore retrieving LinkedIn messages would be a time-consuming activity.
During the preliminary research it became clear that firms often have multiple Twitter accounts (e.g. different accounts for different business units or divisions). For this research, only the general overarching account or the account dedicated to corporate or investor news is selected (if present) and added to the sample. For each of the 66 listed firms the publicly available and official Twitter account or official investor relation account is searched for by consulting the corporate websites or manually by using the search engine of Twitter (see appendix C for an overview of the used Twitter accounts). Subsequently, all tweets and retweets disseminated by the account for the years 2015 and 2016 are retrieved by using an application (i.e. extension) for the internet browser Google Chrome, called ‘Twlets’. Due to a restriction of Twitter, a maximum of 3.200 (re)tweets could be recovered per company. This led to a limited amount of observations concerning seven companies (for these companies the limit of 3.200 tweets was exceeded and were either partly or completely left out of the sample). Additionally, one company merged within the timeframe of this research (Ahold Delhaize). Since data on tweets prior to this merger were not available (the pre-merger Twitter account could not be recovered), data on this company were limited to post-merger observations.
Besides missing data regarding Twitter messages, some quarterly data missed in the consulted databases. Table 1 provides a detailed overview of the missing data and the total amount of observations (quarters) and firms that were ultimately included in the analysis.
The used dataset consists of both Twitter-adopting as non-adopting companies. The fact that a company does not actively use Twitter is also a relevant observation regarding the aim of this research and these observations are therefore included in the dataset. Ultimately, a total of 25.880 (re)tweets from 48 different Twitter-adopting companies have been retrieved and used for analysis.
Table 1. Sample data
Observations Firms Initial sample (total listed firms on AEX, AMX, and ASCX) 600 75 Initial sample, excluding financial firms 528 66 Missing data concerning Twitter messages -43 -2 Missing data concerning financial statements (based on Compustat and Orbis) -43 -1 Missing data concerning ownership (based on Orbis) -17 -Ultimate sample 425 63
Of which concerning Twitter-adopters 304 48* Of which concerning non-adopters 121 15* * Based on the last sample period (Q4 2016)
3.2 Variable definitions
3.2.1 Level of dissemination (dependent variable)
In order to determine the level of dissemination of corporate information via social media, a quantitative content analysis is conducted based on the retrieved tweets. For this study, a content analysis is especially useful since the data concern a large quantity of textual information. By conducting a content analysis, the messages disseminated via Twitter are analysed systematically and efficiently. The aim of this analysis is to determine whether a message contains information relevant for investors or shareholders. In order to achieve this, it is crucial to properly set up criteria based on which information can be determined as relevant. A content analysis succeeds or fails based on the validity of the analytical constructs (Krippendorff, 2004), and therefore the used criteria are of great significance for this study.
At first, it is essential to elaborate on what information is considered corporate information. As introduced in the first chapter, for this research relevant corporate information concerns information that enables investors to evaluate whether the management has managed the firms’ resources in the interests of the external owners (Healy & Palepu, 2001). To further operationalise this definition, this research builds on a checklist in order to determine what information can be considered corporate information relevant for investors or shareholders. In the corporate (voluntary) disclosure literature a checklist is often used to assess the extent or quality of the disclosure policy of a company (e.g. Meek et al., 1995; Botosan, 1997; Eng & Mak, 2003; Broberg et al., 2010). Such a checklist consists of a comprehensive set of disclosure related items, commonly divided into three categories: strategic information, financial information and non-financial information. Based on the items included in the checklist, these studies give a company a disclosure score or index as a measure for the level or quality of the (voluntary) disclosure.
In their study, Zhou et al. (2015) used a checklist developed by Meek et al. (1995) as a starting point for their analysis of Twitter usage among companies for disseminating corporate information. For this research a comparable approach is taken. The framework of Meek et al. (1995) is adopted to derive specific English and Dutch keywords relating to corporate disclosure based on the three information categories (strategic, non-financial, and financial information). These keywords together form a coding scheme (Rose, Spinks, & Canhoto, 2015) used to analyse the Twitter messages in order to determine whether a message contains corporate information. A comprehensive list of the used keywords is presented in the appendix section (appendix B). Microsoft Excel is used to perform this analysis since this program makes
it possible to efficiently assess a considerable amount of data in a relatively easy way. If the Twitter message contains one or more of the keywords of one of the three information categories (STRAT_TW, NONFIN_TW, and FIN_TW), it is assumed that the message contains information relevant for investors or shareholders and is counted. Ultimately, the total number of messages per quarter containing one or more keywords is used as a measure for the level of dissemination of corporate information via social media (DISS_TW).
In order to assess and improve the validity of the applied coding scheme and the approach (content analysis) in general, a sample of the analysed messages has been manually re-evaluated to test whether they are correctly included (counted) or excluded (not counted). By this, Krippendorff (2004) is followed in an aim to assess as to what he more specifically refers to as the semantic validity (correct inclusion or exclusion). Randomly selected tweets have been assessed on relevant corporate information in a two-step approach, based on human judgement of the researcher jointly with the framework of Meek et al. (1995). At first, 70 tweets (35 coded to contain corporate information, 35 coded not to) have been assessed. Based on this assessment, the initial coding scheme has been adjusted. For example, keywords such as ‘FY’ and ‘EPS’ that were initially included were excluded, and ‘Board’ was changed to ‘Board of Directors’. It turned out that these keywords were very common and not specific to corporate information. Besides altering or excluding initial keywords, some keywords were included based on the manual assessment. Examples are ‘Partnership’, ‘CFO’ (and ‘Chief Financial Officer’) and ‘1Q15’ (and variants). In the second step, an increased random sample of 140 messages was assessed (70 coded to contain corporate information, 70 coded not to) and the coding scheme was slightly adjusted once more based on the outcome of this assessment. Ultimately, 90,7% of the messages that were included in the second sample were considered to be rightfully coded based on the final version of the coding scheme. This equals a margin of error of 9,3%. Since Krippendorff (2004) considers an error of 16% remarkably good for an example study, an error of only 9,3% was considered sufficient for this research.
3.2.2 Independent variables
Profitability and leverage are measured based on the Return on Equity ratio (ROE) and the debt-to-assets ratio (D_A). For debt, the total liabilities are used instead of long-term debt due to the fact that data on the long-term debt were quite often not available. Following Yang and Liu (2017), the change in net income before tax is used to determine whether a company can be considered an improving performer. If the change in net income before tax is positive, a
dummy variable (IMPR_PERF) takes the value ‘1’. If the change is negative the dummy variable will take the value ‘0’. For these three variables, the data is retrieved from the Compustat database.
In order to test the effect of industry on the dependent variable, the companies will be divided based on the Global Industry Classification Standard, as retrieved from the Compustat database. The industry codes will be translated into ten dummy variables based on the different sectors used in the standard1: Energy (IND_ENERGY), Materials (IND_MAT), Industrials (IND_IND),
Consumer Discretionary (IND_CD), Consumer Staples (IND_CS), Health Care (IND_HC), Information Technology (IND_IT), Telecommunication Services (IND_TC), Utilities (IND_UTIL), and Real Estate (IND_RE). For the statistical analysis, the Industrials industry is used as a reference variable, since this is the most frequently observed industry.
As García-Meca and Sánchez-Ballesta (2010) mention in their work, various methods are used in existing literature to measure ownership concentration. A variant of one of the two most prominent methods mentioned in this paper is used for this study: the percentage of shares hold by the three largest shareholders. Quarterly data concerning the direct percentage of shares held by the shareholders were derived from the Orbis database to measure the ownership concentration (OWN_CONC). The data on ownership in the Orbis database are known to sometimes contain some flaws concerning double records. The data is therefore manually checked and adjusted for obvious double records. Additionally, Orbis often includes sum-items such as ‘Public’ or ‘Employees and Former employees’ for the total amount of shares held by respectively the public or employees and former employees. Records regarding these items are also excluded since these do not concern one specific shareholder.
3.2.3 Control variables
The results will be controlled for by three additional variables. Firstly, firm size has been proven to influence the level of dissemination in prior work by for example Zhou et al. (2015) and Jung et al. (2017). By this, it is possible that firm size influences the level of dissemination and therefore the results are controlled for the effects of this variable. The variable firm size (SIZE) is included in the analysis by taking the natural log of the total asset of the regarding company. Secondly, as the dataset is designed based on quarterly data, the results of the analysis will be controlled for possible quarterly or seasonal effects. By including three dummy variables (Q2,
Q3 and Q4; Q1 is used as a reference variable), the results will be controlled for the effects of the different quarters (calendar year).
Lastly, two variables concerning the general activity and reach of the Twitter account of the concerning company are accounted for in the analysis, in order to control for a companies’ general involvement with social media. However, this can only be done for the companies that have adopted Twitter, since the required data is logically not available for firms that have not adopted Twitter. The general activity (overall activity of the company on Twitter, not specifically relating to disseminating corporate information) is captured by measuring the total amount of tweets in the regarding quarter (TOT_TW). Building on the results of Jung et al. (2017), a control variable is included concerning the audience size i.e. reach of the Twitter account of the regarding firm. Jung et al. (2017) found that firms with fewer followers are more likely to use Twitter for earnings information. They implicitly linked this result to the idea that companies with more followers are more likely to be a “retail customer-facing” company and therefore use their social media mainly for this group of stakeholders. Ideally, the historic number of followers per Twitter account should be able to proxy the audience size of the regarding company (in line with Jung et al., 2017). However, such data are not easily accessible. Instead the average amount of retweets (per quarter) is used. Since Jung et al. (2017) found a significant relation between the number of followers and the amount of retweets, it is assumed that the average amount of retweets per tweet per quarter (AVG_RT) should be a decent proxy of the audience size.
3.3 Statistical method
In order to test the hypothesised effects of multiple independent variables regarding profitability, leverage, industry, and ownership on the level of dissemination of corporate information via Twitter (dependent variable based on an interval scale), a multiple regression analysis is applied for this research. Possible effects of the previously mentioned control variables are accounted for by including these variables in the analysis as well.
4. Results
4.1 Descriptive statistics
Table 2 shows the descriptive statistics of the used dataset. Since the statistical analysis, as elaborated in the previous chapter (3.3), is sensitive for outliers, the data are checked for extreme values. Data concerning variables which contain extreme values are winsorised based on a 2,5% threshold (two-tailed). The impact of this procedure on the data is included in appendix D. As mentioned in the previous chapter, data concerning TOT_TW and AVG_RT cover only firms that have adopted Twitter and therefore only 304 observations are available for these variables.
Table 2. Descriptive statistics (full sample)
N Minimum Maximum Sum Mean
Std. Deviation DISS_TW 425 0 72 4126 9,708 16,250 STRAT_TW 425 0 35 1243 2,925 5,471 NONFIN_TW 425 0 52 2189 5,151 10,060 FIN_TW 425 0 25 1140 2,682 4,602 ROE 425 -30,674 47,305 951,776 2,239 9,211 IMPR_PERF 425 0 1 163 0,384 0,487 D_A 425 0,058 1,094 232,106 0,546 0,183 IND_ENERGY 425 0 1 29 0,068 0,252 IND_MAT 425 0 1 49 0,115 0,320 IND_IND 425 0 1 106 0,249 0,433 IND_CD 425 0 1 40 0,094 0,292 IND_CS 425 0 1 60 0,141 0,349 IND_HC 425 0 1 28 0,066 0,248 IND_IT 425 0 1 62 0,146 0,353 IND_TC 425 0 1 5 0,012 0,108 IND_UTIL 425 0 0 0 0,000 0,000 IND_RE 425 0 1 46 0,108 0,311 OWN_CONC 425 0,84 100 14957,32 35,194 23,968 SIZE 425 2,733 12,947 3298,243 7,761 1,960 Q1 425 0 1 96 0,226 0,419 Q2 425 0 1 105 0,247 0,432 Q3 425 0 1 104 0,245 0,430 Q4 425 0 1 120 0,282 0,451 TOT_TW 304 0 612 25880 85,132 112,481 AVG_RT 304 0 41,806 1844,385 6,067 8,750
4.2 Regression analysis
A regression analysis consisting of six models has been performed to test the hypotheses mentioned in chapter two. Four models cover the analysis of the four variable groups (profitability, leverage, industry, and ownership) including the control variables concerning firm size and seasonal effects. A fifth model includes all variables and serves to verify the results from the first four models. The analysis regarding the first five models concerns the full sample. An additional sixth model has been tested on a subset of the full sample covering only firms that have adopted Twitter. This additional model has been tested in order to be able to control for the variables concerning the general activity of the firms on Twitter (TOT_TW) and the reach i.e. audience size of the Twitter account (AVG_RT). The results of the analysis are summarized in table 3.
Based on the results presented in table 3, it can be concluded that regarding the effect of profitability on the level of dissemination of corporate information via social media (H1a and H1b) only the ROE turns out to be a significant predictor in the models 1 and 5. Model 6 does not show significance regarding the variable ROE, which indicates the importance of the excluded firms that did not adopt Twitter concerning this variable. However, the significant coefficients for ROE are in an opposite (negative) direction as supposed by H1a. Hereby, no support is found in regard of this hypothesis. For IMPR_PERF the coefficients are insignificant consistently through all three applicable models (1, 5, and 6). By this, H1b is not supported by the data. Ultimately, it cannot be concluded that profitability is positively related to the level of dissemination of corporate information via social media, nor that improving performers disseminate more corpore information via social media.
Regarding the effect of leverage on the level of dissemination via social media (H2) it can be concluded that there are no significant results when looking at models 2 and 6. In contrast, model 5 does show a significant positive coefficient concerning the variable D_A. The difference in outcomes between the models 2 and 5 can be explained by the fact that model 5 is more sophisticated and hereby allows for more effects than only firm size and seasonal effects. Similar to the provided explanation for the different results concerning the variable ROE, the difference between the models 5 and 6 can possibly be explained by the fact that firms that not adopted Twitter were excluded from the analysis concerning model 6. Since the coefficient for D_A is significant in just one of the three models and at the 10% level, little evidence is found in support of H2.
Table 3. Regression analyses (summary)
***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels using a two-tailed t-test. Variables IND_IND and Q1 were used as reference variables. Companies in the utilities industry (IND_UTIL) were not present in the sample and therefore excluded from the analysis. All models were tested for multicollinearity by calculating the Variance Inflation Factor. All factors were below the value 2,609, indicating that the results are not influenced by multicollinearity.
Models 3, 5, and 6 show quite comparable results concerning the effect of different industries on the level of dissemination of corporate information via social media (H3). Consistent through all three models, the industry variables IND_ENERGY, IND_CD, IND_TC, and to some extent IND_RE show significant and negative coefficients. This indicates that the industries related to
Dependent variable: Sample: Prediction Model: 1 2 3 4 5 6 (Constant) -19,390 *** -19,607 *** -16,478 *** -18,352 *** -20,688 *** -8,355 ** ROE + -0,221 *** -0,362 *** 0,021 IMPR_PERF + -0,682 -0,242 -0,075 D_A + 4,809 7,642 * 3,883 Industry: IND_ENERGY (-) -14,821 *** -17,267 *** -8,753 *** IND_MAT (-) 0,904 0,073 1,420 IND_CD (-) -10,875 *** -11,170 *** -5,563 ** IND_CS (-) -1,711 -1,087 1,199 IND_HC (-) -4,570 -7,314 ** -1,633 IND_IT (-) -3,252 -2,225 -4,635 ** IND_TC (-) -12,201 * -13,929 ** -11,840 ** IND_RE (-) -12,921 *** -12,973 *** -2,849 OWN_CONC – 0,001 -0,016 -0,030 SIZE 3,878 *** 3,459 *** 3,986 *** 3,632 *** 4,209 *** 1,871 *** Q2 0,999 0,966 1,023 0,978 1,229 -0,356 Q3 -1,314 -1,245 -1,575 -1,269 -1,485 -1,847 Q4 -0,597 -0,318 -0,438 -0,295 -0,619 -3,292 * TOT_TW 0,100 *** AVG_RT 0,219 ** N 425 425 425 425 425 304 Adjusted R2 0,199 0,187 0,283 0,185 0,314 0,688 F-Statistic 18,529 20,533 14,948 20,212 13,104 38,045 Significance 0,000 0,000 0,000 0,000 0,000 0,000 DISS_TW Full sample Only Twitter Adopters
these variables disseminate less corporate information via social media in comparison to the reference industry (IND_IND). IND_MAT, IND_CS, IND_HC, and IND_IT do not (consistently) show significant results, implying that companies included in these industries do not significantly differ from the reference industry regarding the amount of corporate information shared via social media. Based on the finding that four industries differ significantly from the other five industries, it can be concluded that the level of dissemination of corporate information via social media does vary between different industries. By this, support is provided for H3.
Lastly, table 3 does not show any significant results for the effect of ownership on the level of dissemination via social media (OWN_CONC). The hypothesised negative effect of ownership concentration (H4) could therefore not be supported by the data, as presented in the columns concerning the models 4, 5, and 6.
The variable SIZE, included in the analysis as a control variable, turns out to be strongly significant consistent through all six models. This implies that, next to the variables concerning profitability and industry, firm size has strong explanatory power regarding the level of dissemination of corporate information via social media. The results regarding the variables that were included to control for possible seasonal effects (Q2, Q3, and Q4) are in general not statistically significant. Only with the sixth model the coefficient for Q4 turned out to be significant. Yet, this outcome is only significant at the 10% level. All in all, the significant result is not convincing and the influence of any seasonal effects can therefore be ignored. At last, model 6 contains two additional control variables concerning the general activity of a company on Twitter (TOT_TW) and the reach of the concerning Twitter-account (AVG_RT). Both variables show significant and positive results. Furthermore, regarding model 6, the explanatory power (adjusted R2) of the model is substantially increased in comparison to model 5 (0,688
versus 0,314). This highlights the high explanatory power of the two control variables TOT_TW and AVG_RT relative to the extent to which firms disseminate corporate information relevant for their shareholders and investors via social media.
4.3 Additional regression analysis
Since the results were in general quite contradictory towards the hypotheses (only one of the four hypotheses could be convincingly supported), an additional regression analysis was performed. In an aim to gain further insight into how companies use social media to disseminate corporate information, the four hypotheses were tested in the context of particular types of corporate information. As described in chapter 3, the framework of Meek et al. (1995) is used
as a theoretical foundation for determining whether a tweet contains corporate information. This framework distinguishes between three types of information: strategic information, non-financial information, and non-financial information. By using this framework, it was therefore possible to measure the level of information disseminated via social media specifically concerning strategy (STRAT_TW), non-financial information (NONFIN_TW), and financial information (FIN_TW). Using these three variables as dependent variables, an additional regression analysis was performed to test whether the hypothesis may apply for the specific information types. The results of this analysis are presented in table 4. The results presented in this table concern the full sample (the control variables TOT_TW and AVG_RT are hereby excluded), as this is the main interest of this study. Besides, the analysis concerning solely Twitter-adopting firms did not provide interesting new insights worth reporting.
The results portrayed in table 4 are in general quite in line with the main outcomes of the previous analysis: ROE is consistently significant and negatively related to the dependent variable, clearly varying and significant results between different industries are perceived, and a significantly positive effect of SIZE (control variable) is discovered. Therefore, when comparing the outcome of tables 3 and 4, no substantial differences were discovered regarding the inferences concerning the variable categories profitability and industry and the control variable concerning firm size.
However, when comparing the significant outcome for D_A in table 3 (model 5) with the results of table 4, it can be concluded that this significant relation specifically applies for strategical information (model 1). The models 2 and 3 (table 4) do not show significant outcomes regarding the effect of D_A on non-financial or financial information.
Furthermore, the control variables regarding the seasonal effects (Q2, Q3, and Q4) do play a significant role concerning financial information (model 3). Since the coefficient of variable Q4 is significant and negative, it can be concluded that firms disseminate less financial information in the fourth quarter of the calendar year. This is in contrast to the results of the previous analysis, as portrayed in table 3, where seasonal effects were not present.
Table 4. Additional regression analyses (summary)
***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels using a two-tailed t-test. Variables IND_IND and Q1 were used as reference variables. Companies in the utilities industry (IND_UTIL) were not present in the sample and therefore excluded from the analysis. Since only the full sample is analysed, the control variables TOT_TW and AVG_RT are excluded from the analysis. All models were tested for multicollinearity by calculating the Variance Inflation Factor. All factors were below the value 1,755, indicating that the results are not influenced by multicollinearity. Dependent variable: Sample: Prediction Model: 1 2 3 (Constant) -6,896 *** -11,999 *** -4,434 *** ROE + -0,112 *** -0,213 *** -0,072 *** IMPR_PERF + 0,050 -0,307 -0,216 D_A + 2,761 * 3,939 1,165 Industry: IND_ENERGY (-) -4,641 *** -10,860 *** -3,257 *** IND_MAT (-) -0,118 0,786 0,257 IND_CD (-) -3,250 *** -6,780 *** -1,985 ** IND_CS (-) -0,248 -0,617 -0,044 IND_HC (-) -1,934 * -4,518 ** -1,244 IND_IT (-) 0,046 -1,597 -0,554 IND_TC (-) -2,159 -8,881 ** -4,338 ** IND_RE (-) -3,681 *** -7,481 *** -3,232 *** OWN_CONC – -0,007 -0,003 -0,013 SIZE 1,307 *** 2,365 *** 1,135 *** Q2 0,236 1,144 -0,718 Q3 -0,726 -0,675 -0,761 Q4 0,004 0,117 -1,172 ** N 425 425 425 Adjusted R2 0,247 0,273 0,246 F-Statistic 9,710 10,969 9,645 Significance 0,000 0,000 0,000
STRAT_TW NONFIN_TW FIN_TW
5. Conclusions and discussion
This study has investigated determinants regarding the extent to which Dutch-listed firms disseminate corporate information relevant for their shareholders and other investors via social media. Theories known from the (voluntary) disclosure literature, such as agency theory, signalling theory, and the proprietary costs hypothesis, were adopted leading to four hypotheses concerning the expected influence of profitability, leverage, industry, and ownership.
Despite the fact that for the 75 firms listed on the three major Dutch indices (AEX, AMX, and ASCX) LinkedIn turned out to be the most adopted social media platform (86,7%), this study focused on the slightly less adopted platform Twitter (82,7%). Twitter messages of the firms that adopted Twitter were retrieved and analysed on corporate information potentially relevant for shareholders and investors. For this analysis, the framework of Meek et al. (1995) was adopted to derive a coding scheme, consisting of a list of keywords. The keywords were divided into strategic information, non-financial information and financial information based on the framework and were used to perform a quantitative content analysis. Ultimately, the amount of Twitter messages containing corporate information were counted and used as a measure for the level of dissemination of corporate information via social media.
Regression analysis was used in order to examine whether the hypothesised independent variables did relate to the level of dissemination of corporate information via social media. Besides the variables derived from the hypotheses, additional variables were included in the analysis to control for firm size, seasonal effects, and a firms’ general involvement with social media (general activity and reach). Based on the results from the analyses, only the expectation that the level of dissemination varies between different industries could be convincingly (high level of significance) confirmed. Additionally, some evidence (low level of significance) was discovered for the positive relationship between leverage and the level of dissemination of corporate information via social media. This relation was specifically present regarding the level of strategical information shared via social media, as additional analysis pointed out. The results concerning the effect of profitability (ROE) were surprising in regard to the hypotheses, in that the relation turned out to be significantly negative instead of positive. Ownership concentration did turn out not to be significantly related to the level of dissemination.
Besides these four main variables of interest, control variables concerning firm size, a firms’ general activity on Twitter, and the reach of the Twitter account turned out to be strongly significant and positively related to the level of dissemination of corporate information via
social media. Seasonal effects were only present for financial information share via social media, in that firms tend to disseminate significantly less financial information in the fourth quarter compared to the first quarter (calendar year).
Coming back on the main research question as mentioned in the first chapter of this paper, it can be concluded that profitability, industry, and to some extent leverage are important determinants which influence the level of dissemination of corporate information via social media. Additionally, based on the included control variables, firm size, a companies’ general involvement with social media, and to some extent seasonal effects are explanatory factors towards the dissemination level.
This study adds to current literature by pointing out that the theories applied in this research, derived from the (voluntary) disclosure literature, turned out to be rather limitedly applicable. Especially agency theory and signalling theory do not provide plausible explanations for the extent to which firms disseminate corporate information via social media since the results of the analyses were not in line with the hypotheses concerning profitability and ownership (concentration). By this, it can be argued that firms’ corporate use of social media concerning information dissemination towards shareholders and other investors cannot be compared very well with voluntary disclosure. This inference underlines the statement of Jung et al. (2017) that disclosure and dissemination are essentially two different concepts, as mentioned in the second chapter (2.2).
Furthermore, the observed negative and significant relationship between profitability and the level of dissemination of corporate information via social media is contradicting compared to prior research. Prior research either was not able to find a relation (Jung et al., 2017) or did find a positive relation (Yang et al., 2016; Yang & Liu, 2017). By this, the perceived negative relation is quite remarkable and therefore an interesting outcome. It may imply that firms that do not perform so well (have low levels of profitability) feel the urge to disseminate more corporate information via social media in an aim to explain or disguise dissappointing results. This idea is in line with the conclusion of Yang and Liu (2017), and hereby impression management theory (Goffman, 1959), that firms use social media in an attempt to manage the image of the firm held by the public. As Yang and Liu (2017) describe in their paper, impression management strategies can either be defensive or assertive (Tedeschi & Melburg, 1984). With their finding that better performing firms are more willing to disseminate corporate information they find support for the latter strategy. However, based on the negative relation between
profitability and the level of dissemination observed in this research, it is likely that Dutch-listed firms employ defensive impression management strategies.
Lastly, the results concerning the control variables regarding a firms’ general activity and the reach of the social media platform are interesting as well, when taking into account that Jung et al. (2017) found a negative relation between audience size and the extent to which firms use social media (Twitter) for earnings announcements. By this, the observed positive relation undermines the explanation of Jung et al. (2017) that firms that reach a greater audience via social media are probably more retail customer-oriented and therefore are less likely to use social media to disseminate corporate information. Instead, it can be said that Dutch companies that in general are more active on social media and serve a larger audience, are more active in disseminating corporate information via social media as well.
Like any research, this research is exposed to some limitations, despite all attempts to take these away or minimise the effects. Firstly, in comparison to prior research, the used sample size is relatively small. For example, Jung et al. (2017) used all companies in the S&P 1500 index, where this study examined only 63 firms. Next to that, the limit set by Twitter, which limits the maximum amount of retrievable tweets to 3.200, led to the exclusion of some observations. Since companies for which this limit led to an obstacle do quite intensively use Twitter i.e. social media, it would have been interesting to see which results would have been discovered with this data included. Thirdly, the content analysis employed in this study is prone to limitations as well. A content analysis in general comes with validity concerns, since it does not pay attention to the context of the assessed keyword. Yet, by assessing and improving the applied coding scheme in a two-step approach, attempts have been made to maximise the validity of this analysis. Lastly, the data derived from the Orbis database regarding ownership (concentration) were manually modified for double records and sum-items to increase the validity. However, the validity of this data may still be questioned due to the fact it cannot be guaranteed that all issues are filtered out. Besides, the manual modification could also raise questions regarding the reliability of the data.
This study provides some interesting leads for future research. At first, as discussed earlier in this chapter, this research provided surprising results regarding the relation between profitability and the amount of corporate information shared via social media. These contrasting results towards prior research indicate that there is yet a lot unknown about this relation and the extent to which firms possibly use social media to manage the impression of the public. This hereby calls for future research. Next to that, industry was perceived as an important
determinant concerning the level of dissemination of corporate information via social media. However, similar to prior research, this study does not provide an explanation as to what may explain the variations of the dissemination levels between the different industries. This leaves an interesting area yet unexplored. Lastly, the preliminary inquiry used to shape this study indicated that LinkedIn was the most prominently adopted social media platform regarding Dutch-listed firms. However, due to several reasons, Twitter was chosen as subject of research. For future research it would therefore be interesting to investigate in what way firms use LinkedIn to provide their shareholders and other investors with corporate information.
References
Archambault, J. J., & Archambault, M. E. (2003). A multinational test of determinants of corporate disclosure. The International Journal of Accounting, 38, 173-194. Ashbaugh, H., Johnstone, K. M., & Warfield, T. D. (1999). Corporate Reporting on the
Internet. Accounting Horizons, 13 (3), 241-257.
Barako, D. G., Hancock, P., & Izan, H. Y. (2006). Factors Influencing Voluntary Corporate Disclosure by Kenyan Companies. Corporate Governance: an international review, 14 (2), 107-125.
Beyer, A., Cohen, D. A., Lys, T. Z., & Walther, B. R. (2010). The financial reporting environment: Review of the recent literature. Journal of Accounting and Economics, 50 (2-3), 296-343.
Bhagwat, V., & Burch, T. R. (2016). Pump it Up? Tweeting to Manage Investor Attention to Earning News. Unpublished paper. Retrieved from
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2382962
Blankespoor, E., Miller, G. S., & White, H. D. (2014). The Role of Dissemination in Market Liquidity: Evidence from Firms' Use of Twitter. The Accounting Review, 89 (1), 79-112.
Botosan, C. A. (1997). Disclosure Level and the Cost of Equity Capital. The Accounting Review, 72 (3), 323-349.
Broberg, P., Tagesson, T., & Collin, S.-O. (2010). What explains variation in voluntary disclosure? A study of the annual reports of corporations listed on the Stockholm Stock Exchange. Journal of Management & Governance, 14, 351-377.
Cade, N. L. (2017). Corporate social media: How two-way disclosure channels influence investors. Unpublished paper. Retrieved from
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2619249
Chen, H., De, P., Hu, Y., & Hwang, B.-H. (2014). Wisdom of Crowds: The Value of Stock Opinions Transmitted Through Social Media. The Review of Financial Studies, 27 (5), 1367-1403.
Cooke, T. E. (1991). An assessment of voluntary disclosure in the annual reports of Japanese corporations. The International Journal of Accounting, 26 (3), 174-189.
Craven, B. M., & Marston, C. L. (1999). Financial reporting on the Internet by leading UK companies. The European Accounting Review, 8 (2), 321-333.
Debreceny, R., Gray, G. L., & Rahman, A. (2002). The determinants of Internet financial reporting. Journal of Accounting and Public Policy, 21, 371-394.
Eng, L., & Mak, Y. (2003). Corporate governance and voluntary disclosure. Journal of Accounting and Public Policy, 22, 325-345.
Friedman, M. (1962). Capitalism and Freedom. Chicago : University of Chicago Press. García-Meca, E., & Sánchez-Ballesta. (2010). The Association of Board Independence and
Ownership Concentration with Voluntary Disclosure: A Meta-analysis. European Accounting Review, 19 (3), 603-627.
