The impact of social media on firm value following a financial misconduct

(1)

Amsterdam Business School

The impact of social media on firm value following a financial misconduct

Name: Inge Verdouw Studentnumber: 10013695

Thesis supervisor: Ir. Drs. A.C.M. de Bakker Date: 19 June 2016

Word count: 13829

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

(2)

Statement of Originality

This document is written by student Inge Verdouw 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

This study researches the impact of social media on firm value following a financial misconduct by conducting an event study.

Prior research has shown that committing a financial misconduct leads to a negative price reaction for the committing firm. It is also shown that the presence in regular media of this particular event influences this negative reaction. In this study, the focus is on the use of social media for firms to disclose this financial misconduct. This study is motivated by the increasing use of social media and the increasing amount of financial misconducts investigated by the SEC. The selected firms are the firms that are issued by the SEC with an Accounting and Auditing Enforcement Release (AAER) and are listed in the S&P 500, in the period 2011-2015. The findings show that there is a negative price reaction after an AAER is issued. It also shows that the price reaction is larger when firms disclose the AAER on their social media account. There is no evidence found that there is a difference between which social media account is used. However contrary to what was expected, the negative reaction of social media disclosure increased after the SEC recognized social media as a disclosure channel.

(4)

1 Introduction

The use of social media by firms is rapidly increasing. In 2014, 413 firms (83%) of the Fortune 500 have a corporate Twitter account and 401(80%) firms are on Facebook. In both cases, this is an increase compared to 2013, in the case of Twitter an increase of 6% and in the case of Facebook an increase of 10%. While the use of corporate blogs declined in 2014 to 157 firms (31%), a decrease of 3% in the use of this tool compared to 2013 (Barnes & Lescault, 2014). It shows that over the past few years, the use of social media enlarged and firms use it to relate with customers and other stakeholders. Nowadays firms even started with using social media to disseminate and bring attention to financial information.

On 12 August 2013, US billionaire investor Carl Icahn tweeted twice from his Twitter account (@Carl_C_Icahn) about Apple’s share price. In the afternoon, he first posted: ‘’we currently have a large position in APPLE. We believe the company to be extremely undervalued. Spoke to Tim Cook today. More to come.’’ He posted a follow-up tweet four minutes later, stating that he would press Apple to increase its stock buyback, in which the firm purchases its own shares in a bid to boost its value. ‘’Had a nice conversation with Tim Cook today. Discussed my opinion that a larger buyback should be done by now. We plan to speak again shortly.’’ The markets responded almost immediately to his enthusiastic tweets, which totaled just 49 words. What followed was a rise in trading volume from regularly about 200,000 shares changings hands, to more than 829,000 shares. It raised Apple’s shares from $475 to an intra-day high of $494 and closed near $490, around 5% above the opening price (Aymerich-Franch & Carrillo, 2014). This example shows the powerful relation between social media and financial markets.

In addition, regulators started to recognize the growing use of social media. Where the use of social media by firms was previously without any regulatory guidance, this changed in 2013. The Securities and Exchange Commission (SEC) released a guidance on how firms that are active on social media should disseminate financial information. In their guidance, they approved that firms could disseminate corporate announcements that can have a material impact on an investor’s interpretation of the financial information (SEC, 2013).

1.1 Research question

Zhang (2015) found that it is important for firms to use social media in order to reach the target audience and influence consumer-purchasing decisions. Active firms, in adopting new communication media, embrace larger audiences and have greater impacts than less active firms

(7)

disclosure. Despite more and more firms are using social media, not every one of them is convinced to also use social media as a new media to disclose financial information. A lot of them are concerned about the uncertainty of social media and are not ready to adopt new media for information disclosure.

So even though regulators and firms increase their attention regarding the use of social media, there is not much data about the consequences of this use, and if different social media platforms have different impacts. The papers of Blankespoor et al. (2014) and Lee et al. (2015) were one of the first papers that looked into these consequences. Blankespoor et al. (2014) found that firms that use Twitter to disseminate links to firm-initiated press releases experience lower information asymmetry. Lee et al. (2015) found that corporate social media attenuates the negative price reaction to product recall announcements. On the other hand, they found that since social media developed from less to more interactive platforms, the attenuation benefits of social media lessened.

Given the development of social media and the important role it starts to play in disseminating financial information, the subject of this study is the effect of social media after a financial misconduct announcement on firm value. With the research question:

Does corporate use of social media facilitate larger market reactions for financial misconduct announcements?

This research tries to find if the use of social media leads to a smaller/bigger market reaction for social media user’ firms than firms that do not use social media. In addition, it is interesting to see whether there are different results regarding the type of social media channels firms use to disseminate the information. At last, this research divides social media use in two samples, a sample before the SEC guidance on the use of social media and a sample after the guidance, to see whether this guidance led to any differences in market reaction.

1.2 Motivation for this study

The main reason for this study is to better understand the evolving role of social media in the financial accounting disclosure landscape. Two papers discussed during the Financial Accounting Research course in the Master triggered this study. The paper of Lee et al. (2015) is about product recalls and the effect of social media on firm value, where the paper of Kedia & Rajgopal (2011) is about corporate misconduct and the role of the SEC. Lee et al. (2015) suggested extending their research for other negative effects. In this study, the negative effect is committing a financial

(8)

misconduct, and the aim is to capture the effect of those financial misconducts on the market prices.

This study is interesting for scientific and societal knowledge as well. From a scientific perspective, this study expands the existing literature. There have been some studies about an economic event and the role of social media. However, this is the first study that looks at the role of social media after a financial misconduct. This is interesting because the use of social media increases rapidly but also there are more and more financial misconducts detected in the financial reports of firms. When something interesting comes out of this study, it is going to be a motivation for future researchers to look at other financial disclosures and their effect on market prices. From a societal perspective, this study is also interesting. It would be interesting for firms if they have more insight into the consequences of the use of social media. Maybe when they find out that, after multiple types of research about the effects of social media on firm value, they have to adjust their social media policy. Also for investors, it could be interesting if they know more about the market reaction to certain disclosures on social media. Better understand how they should react on a particular event would help them in making better investors decisions. It could give them some insight as well, to determine for how long the effect lasts and if they have to react.

1.3 Structure of this study

The structure of this study is set out as follows: Section 2 gives a research background, discusses the related literature and introduces the hypotheses. This section introduces the chosen research methodology, namely an event study. Section 3 presents the research methodology in more detail with the used formulas and shows the data selection process. Section 4 explains the operationalization of the data and the collection of the data sample. Section 5 shows the empirical results of this event study. In section 6, a sensitivity analysis is performed and the study discusses the results and concludes in section 7.

(9)

2 Theoretical Framework

2.1 Overview

The previous section introduced the aim of this research. This section discusses some fundamental knowledge about financial misconduct and social media. The first three paragraphs explain some essential concepts used in this research and then look into the relevant papers that already have been written about those subjects. Especially it discusses some relevant papers that looked at the market reaction after a financial misconduct announcement without considering social media. The fourth paragraph focusses on explaining the study’s methodology. Finally, the last section formulates the hypotheses that this study aims to test.

2.2 Financial misconduct

This section introduces the definition of a financial misconduct and discusses some relevant papers regarding a financial misconduct. After that, it discusses some relevant papers regarding the consequences of conducting a financial misconduct on firm value.

2.2.1 What is a financial misconduct?

Financial accounting standards are designed rules that firms follow in order to provide shareholders, directors, corporate managers, potential investors and regulatory agencies with information relating to the firms’ financial performance. Nevertheless, often do these standards give the firms some degree of flexibility regarding how to measure and when to recognize the revenues/expenses and assets/liabilities. This could lead to misinterpretation or misuse the standards. A financial misconduct means that firms misrepresent their financial statements. While examining the financial reports, there can be some indications that a financial misconduct has occurred in a firm, but there should also be some consideration about the given flexibility when preparing a financial report. This flexibility and discretion, given by the accounting standards in preparing and presenting the financial information, needs to be considered and weighted. The most common and well-known ways to manipulate the financial statements are accelerating revenue, channel stuffing, bill and hold transactions and, finally changing accounting methods or assumptions (Liang and Tober, 2009).

A lot of accounting literature investigates the quality of firm’s reported earnings. To identify the occurrence of a financial misconduct, certain papers used Accounting and Auditing Enforcement Releases (AAERs). The SEC first issued an AAER in 1982. They issue an AAER during or at the conclusion of an investigation against a firm, an auditor, or an officer for alleged

(10)

accounting and/or auditing misconduct. The releases obtain different insights on the nature of the misconduct, the individuals and entities involved, and the effect on the financial statement (Dechow et al., 2011).

To give some better insight into the kind of data used in this study, it is helpful to look at the events and in, which order those events take place during an SEC enforcement action (Figure1). Almost all enforcement actions start with a ‘trigger event’ that there might be a potential misconduct and gets to the SEC’s attention. Trigger events are for example the disclosure of a firm that there are some deficiencies, restatements, and the departure of the auditor. After such an event, the SEC starts an informal investigation to collect information, and their findings may lead to a formal investigation of the appearance of financial misconduct. If it leads to a formal investigation, the SEC sends a Wells Notice to the firm, to let them know that they mean to start enforcement proceedings. After that, the SEC lays administrative sanctions on the firm or take some civil actions. The case can even be sent to the Department of Justice, which may lead to criminal charges too. The last step of the SEC involves the releasing of their findings and given penalties in their Administrative Proceedings and Litigation Releases. Those releases, give a detailed description about the period the misconduct took place and other relevant information (Karpoff & Lou, 2010).

Figure 1. Timeline of a typical enforcement action, Karpoff & Lou (2010)

2.2.2 Important literature regarding financial misconduct

Dechow et al. (1996) performed a detailed analysis of 2,190 firms that the SEC investigated for misstating earnings between 1982 and 2005. According to their research, the most frequent types of misstatements are the overstatement of revenues and the misstatements of expenses and capitalizing costs. They also investigate the motives for and consequences of earnings manipulation

(11)

misstating their compensation is often sensitive to the firm’s stock price. The reason to misstate is that they want to cover up a decline in financial performance to keep the market valuations at the same level. They also find returns of -25 percent for 13 AAER firms where a problem identified by the auditor or an auditor dismissal represents the first announcement of accounting trouble, compared to -4 percent for 25 AAER firms where an SEC investigation represents the first announcement. This study investigates whether the presence of social media extends this price reaction.

Graham et al. (2008) find that the adverse effect of accounting restatements on firms’ external financing activities last for about three years. Prior research finds that restatement firms undertake various actions to restore investors’ trust and many such actions take the time to implement. Farber (2005) finds that accounting fraud firms improve corporate governance to enhance financial reporting credibility. He finds that it takes fraud firms up to three years to achieve the same quality of corporate governance as non-fraud firms.

2.2.3 Financial misconduct and the market reaction

In their working paper Hranaiova & Byers (2007) research if the market reaction to a restatement announcement changed during some crucial economic events like the Sarbanes-Oxley act, and how this possibly can affect market value, market efficiency, and price volatility. Their outcome is in line with previous research that the market reaction is solely statistically significant on the two days (0, +1) of the restatement announcement. They also find that a negative two-day market reaction is on average two and a half time bigger than a positive market reaction, this counts for the statistically significant reactions. When investors are only interested in those statistically significant market reactions after a restatement announcement, there is a two-day share price decline of 33 percent because of a negative market reaction and a 14 percent increase because of positive market reactions. They also find that SOX and some crucial events surrounding, have significantly and favorable influence on the market reaction to a restatement announcement. Since the presence of SOX, the negative market value reaction to a restatement announcement declined and the uncertainty after an announcement about the restating firm declined as well, this suggests that the investors trust increased.

Wu (2002), also looks at the market reaction and the value of the stock around a restatement announcement but also focus on the characteristics that influence the short-term abnormal returns. This research takes the trust of investors in the restated firms also into account. In her research, Wu finds a powerful negative short-term market reaction to a restatement announcement and the downward pattern is already notifiable in the six-month period before the

(12)

restatement announcement, and the negative consequences of such an announcement can last for approximately four months. The research shows that the market reaction is significant and negative during a three-day price response window around the announcement of the restatement, over -11 percent cumulative abnormal return. She also finds that some quantitative information and qualitative characteristics of a particular restatement play a significant role in the explanation of the short-term market response. This research is interesting because it divides restating firms into two groups, the ones that that show the restated earnings on the day of the announcement and the ones that do not provide restated earnings numbers. In addition, the qualitative characteristics of a specific restatement are grouped on a more accounting perspective instead of an economic perspective. The restatement announcement with the largest losses are the firms with changes in revenue, the firms with fraudulent accounting practices and the firms that do not provide restated earnings numbers.

2.3 Social Media

This section defines social media and looks at the possibilities of information disclosure. It also looks at the disclosure possibilities within traditional media and gives an overview of the first social media papers that writes about social media and market reaction.

2.3.1 What is social media

Before looking at the impact of Social media on firm value, this study needs to define Social media. According to Kaplan & Haenlein (2010), ‘’Social Media is a group of Internet-based applications that build on the ideological and technological foundations of Web 2.0, and that allow the creation and exchange of User Generated Content’’. They provide a classification of Social Media, which groups’ applications currently subsumed under the generalized term into more specific categories by characteristics: collaborative projects, blogs, content communities, social networking sites, virtual game worlds, and virtual social worlds (Table 1).

Table 1. Classification of Social Media by social presence/media richness and self-presentation/self-disclosure, Kaplan&Haenlein (2010)

(13)

They are referring to Web 2.0 and, although Web 2.0 does not refer to a specific technical update of the World Wide Web, there is a set of basic functionalities that are necessary for its functioning. Like RSS, it stands for Really Simple Syndication, what is ‘’a family of web feed formats used to publish frequently updated content, such as blog entries or news headlines, in a standardized format.’’ This led to a platform for the evolution of Social media. The first signal of a social media landscape was the development of blogs. They can come in a multitude of different variations, from personal diaries describing the author’s life to summaries of all relevant information in one specific content area. Many firms start using blogs to update employees, customers, and shareholders on developments they consider important.

Twitter is a microblogging service that allows its users to share short messages up to 140 characters in length with each other. These short messages are tweets and can be sent and retrieved across a wide variety of media including e-mail, text messaging, instant messaging, the internet and other third-party applications (Hughes & Palen, 2009). Started in October 2006, Twitter estimates having 307 million monthly active users (Statista, 2015). It is also on Alexa’s top 500 sites on the web, ranked number 9 (Alexa, 2015).

Facebook is a social network that enables users to connect with people with the same interests and background. It is a different use of data sharing because it enables the sharing of pictures, videos, and other forms of media. Facebook allows users to post messages without supervision

Saxton (2012) states that the use and upcoming use by firms and their environments of social media had a great impact on different disciplines inside the firm. Those disciplines are characterized by the following actions: using it as a disclosure channel, as an investor relations vehicle, and as a tool for demonstrating social responsibility and fostering accountability.

2.3.2 Social Media and disclosure

During the developing technology environment, there is a frequent introduction of new methods for a firm to communicate, such as social media. However, firms are required to report material information through the SEC-mandated disclosure channel to the investing public (e.g., Form-10-K on EDGAR). Nevertheless, since the presence of social media, managers may opt to disclose the material information through the alternative channels. An optional disclosure channel affords managerial discretion over the information content and flow. The way firms disclose the information on these social media channels depends on the channel used. Where a message on Facebook may include many text data, Twitter is limited to 140 characteristics of the text. The

(14)

chosen disclosure technology affects the type and volume of data that is disclosed (Dorminey et al., 2015).

In April 2013, the SEC allowed firms to use social media for corporate announcements that may have a material impact on an investor´s interpretation of the financial information if investors are informed about which social media channels the firm uses (SEC, 2013). If there is not any guidance or formal reporting requirements, it is difficult to determine how firms use social media for material disclosures. A good example of that is the information that Netflix’s CEO reported on his personal social media site. He mentioned there that Netflix users had streamed in excess of 1 billion hours of video. As a reaction to this post, the SEC started an investigation what led to a potential SEC enforcement action. However, it resulted in the formal SEC guidance on the use of social media. Since then Social media (Facebook or Twitter) is an acceptable outlet for material disclosures (Dorminey et al., 2015).

2.3.3 Disclosure through traditional media

Before the existence of social media, firms could not directly communicate to investors on a regular basis. Their way of communication was primarily through information intermediaries, like newspapers/magazines, to disclose firm-initiated information. It seemed that these newspapers were focusing on the more visible firms, and because investors had limited time and resources, firm disclosures did not always reach every investor on a timely basis. This reach of only a portion of investors resulted in information asymmetry. This led to lower market liquidity (Miller, 2006). To look at the effect of social media on firm value, this research looks at market reaction, because many types of research use this measure to look at the impact of press. First, some researches where they look at the ‘old’ communication channels are being discussed, where social media was not available jet. Gordon et al. (2008) examine the effect of press release characteristics on returns in a three-day window centered on the restatement announcement. They find that firms reporting the restatement in the headline experience a larger price decline than other firms. Their research found, just like the research of Palmrose et al. (2004), that restatements initiated by management are usually associated with more negative returns. However, by providing greater amounts of (objective) disclosures, it shows an ongoing ability and willingness to self-monitor. This could mitigate the negative market reaction to restatements.

Palmrose et al. (2004) examine the market reaction to a sample of 403 restatements announced from 1995 to 1999. They find an average abnormal return of -9 percent over a 2-day announcement window.

(15)

Also, Files et al. (2009) looked at the market reaction and find that firms providing less prominent press release disclosure of a restatement have a less negative return at the announcement. However, much of this reward disappears in the next 20 days. They also find that firms providing less prominent disclosure are less likely to be sued for securities fraud, and the decrease in litigation is a permanent reward.

Myers et al. (2013) investigate determinants of restatement disclosure choices and related stock price reactions in the post-Sarbanes-Oxley era. They find that the market reaction to a restatement, which leads to lower reported income, is less negative when the restatements reports in periodic amended SEC filings compared to when it discloses more transparently in an 8-K filing. What is in contrast with the research of Gordon et al. (2008) who states that when you are constantly disclosing objective disclosure, you can mitigate the negative price reaction to a restatement. They also find that firms announce large income-decreasing restatements less rapidly than other restatements. This is in line with the investigation of Bloomfield (2012) who according to the ‘’incomplete revelation hypothesis’’ predicts that firms try to temporarily boost or maintain stock prices by disclosing bad news obscurely because obscure information may be more difficult for some investors to extract.

2.3.4 Disclosure through social media

The past few years the use and influence of social media in de economic landscape is taken into account and several researchers started a research. Two prominent types of research are those of Blankespoor et al. (2014) and Lee et al. (2015).

Blankespoor et al. (2014) looked at the role of Twitter as a complementary communication channel to improve the disclosure of firm-initiated information. They find that dissemination by Twitter during news event windows is associated with lower bid-ask spreads, greater depths, and a higher liquidity ratio after controlling for the information content of the news, the presence of information intermediaries, market conditions, and firm-specific characteristics. They find that this relation declines for high visibility firms because they already had a big communication platform. Final their findings indicate that firms use Twitter to disseminate firm-initiated disclosures, and this helps reduce information asymmetry, especially for those firms that did not have a broad disclosure channel. Their study shows that in addition to the impact of the information within the disclosure, broader dissemination of that information can have real market consequences.

Lee et al. (2015) examine how corporate social media affects the capital market consequences of firm’s disclosure in the context of consumer product recalls. During a consumer product recall, it is really important for the firm to spread the information as soon as possible to a

(16)

wide network of stakeholders, they expect that social media could be helpful in making this happen. They document that on average, social media attenuated the negative price reaction to recall announcements, but this attenuation depends on the level of control the firm has over its social media content. Especially the attenuation benefits of corporate social media, while still significant, lessened when firms used Twitter and Facebook. This is particularly explained by the loss of complete control over their social media content. In their study, the negative price reaction to a recall is attenuated by the frequency of tweets by the firm, while exacerbated by the frequency of tweets by other users.

2.4 Event study

This section explains the method used to perform this study and describes some main components of that study.

The event study method is generally used in the accounting, economics, finance, and management fields to examine the effect on value when a corporate event occur (Park et al., 2004). Within accounting research, an event is often a moment when information is made public to investors and other stakeholders that in their place use that information for their decision making regarding the firm. To measure the value of a firm following a corporate event, studies often use the stock prices. The relevance of using an event study comes from the assumption of an ‘effective market’, what means that the market immediately reacts and shows an effect in the stock price around the event (MacKinlay, 1997).

An event study tries to capture whether a specific event leads to an abnormal stock return. A given stock return model expects an estimated return, however when the observed return differs, that difference is called the abnormal return. The first step in performing an event study is to define the particular event of interest. After that, as the second step, a stock return model is developed (Park, 2004).

There are multiple models obtainable to calculate the return on a stock. They can be divided into two groups namely, economic and statistical models. The main difference between the two groups are the assumptions the models are based upon. Within the economic model, assumptions are made about investors’ behavior and some statistical assumptions, where the statistical model only make statistical assumptions. The most used economic models are the Capital Asset Pricing Model (CAPM) and the Arbitrage Pricing Theory. The most used economic models are the Constant Mean Return Model and the Market Model (MacKinlay, 1997).

(17)

In multiple types of research, the Market Model is used to estimate the expected stock return. They choose this model because it looks at a stocks’ specific past performance and takes into account its sensitivity to general market movements reflected by their own countries’ stock market indexes (Park, 2004).

Third, an estimation window should be determined, in order to estimate the parameters of the model by using only a subset of the data. However, within the Market Model, it is possible that an estimation window contains unusual market movements, for example, political crises, that can significantly influence the parameters. When looking at multi-country studies, it is possible that there occurred some country-specific unusual events. It is really time-consuming and cost-intensive to address all those events. According to Park (2004), extending the estimation window solves this problem, because it reduces the influence of a particular unusual market movement (Park, 2004).

For the Constant Mean Return Model during the estimation window, the mean return of a particular asset is calculated. This model assumes that this mean return also counts for the event window, so they do not consider the market. Where it seems that the Constant Mean Return Model is very simplistic, Brown and Warner (1980) however suggests that the results of this model do not significantly differ from the Market Model.

McWilliams & McWilliams (2011) in their manuscript, try to better the integrity of the researches that are using the event study method. They discuss and give some alternatives to the recommendations by McWilliams and Siegel (2007) in order to perform an event study. They especially give some extra recommendations in order to be consistent with statistical theory, existing research and the accepted practice. For example, they are convinced that a researcher is always obliged to define enough details about how and what sources are addressed to gather the data so that the study can be repeated by another party.

2.5 Hypotheses development

This section uses the information discussed in the previous sections to develop hypotheses to test in this study. To find an answer to the research question ‘’Does corporate use of social media facilitate larger market reactions for financial misconduct announcements?’’ Three main hypotheses and one sub-hypothesis are created. Most of the literature that is discussed in the previous section finds a negative price reaction following the announcement of a financial misconduct. So first, this study needs to test if this is also the case with an issued AAER. What leads to the following hypothesis?

(18)

H1: A financial misconduct announcement leads to a negative price reaction for the firm’s stock price

After this is tested, the presence of social media should be taken into account. It is expected that with the use of social media the negative price reaction increase because the scope of the announcement is even bigger. In other words, the use of social media leads to even more transparent data sharing, what may lead to an even bigger market reaction. This leads to the following hypothesis;

H1a: The price reaction for firms that use social media is larger than for firms that do not use social media following a financial misconduct announcement

According to the study of Lee et al. (2015), there is also a difference in the effect of social media usage regarding, non-interactive and interactive social media. In line with his expectations, it’s expected that interactive corporate social media in this research leads to a more negative price reaction to a financial misconduct announcement because the negative effect can be enlarged by the opinions of the multiple stakeholders on the social media. The hypothesis would then be;

H2: Compared to non-interactive corporate social media, interactive corporate social media has a larger price reaction to a financial misconduct announcement

There is not much literature written about the consequences of the SECs approval of the use of social media as a disclosure channel. However, given the research by Hranaiova & Byers (2007) who looks at SOX implementation as an event, this research would see the SEC approval of the use of social media as a disclosure channel, as an economic event. Because since the SOX adoption the negative effect of a restatement declined, it is expected that this count for the SEC approval as well. This means it leads to more transparency and therefore investor’s trust increase. This leads to the following third and last hypothesis.

H3: Since the SEC’s approval of the use of social media to disclose important firm information to stakeholders, the negative market reaction declined

(19)

3 Methodology

3.1 Overview

As previously mentioned the method used in this study to test the hypotheses is an event study. By performing an event study, the effect of a financial misconduct announcement on the stock price is measured. In this section, the research methodology is further developed by using the theory of Mackinlay (1997) about event studies.

3.2 Steps event study

According to MacKinlay (1997), there is no ‘unique structure’ to follow when conducting an event study, however, he discusses that there is a general flow of analysis that can be followed.

Step 1: Determining the event and relevant event period

In order to perform an event study, the first step to perform is determining the event and relevant event period to evaluate the stock returns in. According to the relevant event period, the period of market reaction following the event should be determined, this is the event window. In this study the event is, conducting a financial misconduct. The event date (day 0) is the date the conducted financial misconduct is announced. However as MacKinlay states and previous research in this field shows, the event window is regularly enlarged to multiple days. Often some days after the announcement are included, to better capture the price effects after the announcement. Also, some days before are included because there is a possibility the market already received some information about the financial misconduct. This study uses an event window similar of the study of Lee et al. (2015), namely [0, +1]. This is in line with the recommendation of Mackinlay to put the day of the announcement and the day after the announcement, in the event window. After the event window is determined, an estimation window should be set. In general, the period before the event window is used for the estimation period. To prevent that unusual market movements influence the parameters, an extended estimation window is used. Mackinlay gives examples in his research for an event window of -120 or -250 days, in this research a 250 days estimation window is used.

(20)

Step 2: Calculate the daily stock returns for the estimation window and the event window The daily change in stock return is given by:

𝑅_𝑖𝑡 =𝑃𝑖𝑡− 𝑃𝑖𝑡−1 𝑃_𝑖𝑡−1 Where:

𝑅_𝑖𝑡 = the return of stock i at day t 𝑃_𝑖𝑡 = the stock price of firm i at day t 𝑃_𝑖𝑡−1 = the stock price of firm i at day t-1

Step 3: Calculate the normal daily returns for the estimation window and the event window In line with previous studies that look at price reactions, this study uses the Market Model where 𝑅_𝑚𝑡 is the market return. In this model there supposed to be a solid linear relation between the market return and the stock return. Through this model, the parameters for the ‘normal expected return’ can be determined to estimate the stock returns for all days in the estimation window. The Market Model is given by:

𝑅𝑖𝑡 = 𝛼𝑖 + 𝛽𝑖𝑅𝑚𝑡 + 𝜀𝑖𝑡 𝐸 (𝜀𝑖𝑡 = 0) 𝑣𝑎𝑟 (𝜀𝑖𝑡) = 𝜎𝜀𝑖 2 Where: 𝑅_𝑖𝑡 = 𝑅𝑚𝑡 = 𝛼_𝑖,𝛽𝑖 & 𝜎𝜀2 =

the return of stock i at day t

the return of the market portfolio at day t the parameters of the Market Model

Step 4: Calculate the expected daily returns for the event window

Fourth, with the parameters 𝛼_𝑖 and 𝛽_𝑖 the expected market return is calculated for the events during the event window. This means that the real event does not have to emerge to be able to calculate the expected return in the event window. This is found on the evolution of the return measured in the estimation window. The expected stock return is given by:

(21)

Where:

𝐾𝑖𝑡 = the expected return on day t of the event window 𝑅_𝑚𝑡 = the return of the market at day t of the event window

𝑎̂𝑖 & 𝛽̂𝑖 = the estimated parameters of the Market Model based on the estimation window Step 5: Calculate the abnormal daily returns for the event window

Fifth, in order to evaluate the event’s influence on firm value, the abnormal returns should be measured. This is the actual return of the stock over the event window minus the expected return of the stock. The abnormal returns are given by:

𝐴𝑅_𝑖𝑡 = 𝑅_𝑖𝑡 − 𝑎̂_𝑖 − 𝛽̂_𝑖𝑅_𝑚𝑡 Where:

𝐴𝑅𝑖𝑡 = the abnormal return of stock i at day t of the event window 𝑅_𝑖𝑡 = the actual return of stock i at day t in the event window Step 6: Calculate the disturbance variance for the event window

In the sixth steps, the disturbance variance for the estimation window is estimated. The disturbance variance is given by:

𝜎̂𝜀𝑖 2 ₌ 1 𝐿1− 2 ∑ (𝑅𝑖𝑡− 𝛼̂ − 𝛽𝑖 ̂ 𝑅𝑖 𝑚𝑡) 2 −1 𝑡=−𝐿1 Where:

𝑅_𝑖𝑡 = the return of stock i at day t of the event window 𝑅𝑚𝑡 = the market return at day t of the event window

𝐿₁ = the length of the estimation window Step 7: Calculate the average abnormal daily returns

Seventh the average abnormal returns is calculated, for N events on day t in the event window. The average abnormal returns are calculated for all days in the pre-, event- and post-window of the firms in the selected sample. This is given by:

(22)

𝐴𝑅_𝑡 = 1

𝑁∑ 𝐴𝑅𝑖𝑡 𝑁

𝑖=1 Step 8: Calculate the cumulative abnormal returns

The eighth step is cumulating the aggregated abnormal returns for the given event window. 𝑡1 is the first day of the event window and, 𝑡2 is the last day of the event window. The aggregated cumulative abnormal returns for any interval (𝑡1, 𝑡2) in the event window is given by:

𝐶𝐴𝑅(𝑡₁𝑡₂) = ∑ 𝐴𝑅_𝑡 𝑡2

𝑡=𝑡1

Step 9: Calculate the variance of the cumulative abnormal returns

In the ninth step, the variance of the aggregated abnormal returns for any interval (𝑡1𝑡2)in the event window is given by:

𝑣𝑎𝑟 (𝐶𝐴𝑅(𝑡1𝑡2)) = 𝑡₂−𝑡₁+ 1 𝑁2 ∑ 𝜎̂𝜀𝑖 2 𝑁 𝑖=1 Step 10: Test the significance

The tenth step is about testing the significance. In this study, the null hypothesis is that the cumulative abnormal return (𝐶𝐴𝑅) is zero. This means that the event (the announcement of an issued AAER) does not lead to a significant impact on firm value. It can be stated as:

𝐻₀ = 𝐶𝐴𝑅 and 𝐻𝑎: 𝐶𝐴𝑅 ≠ 0

The null hypothesis is tested under various significance levels by comparing the outcome of the given equation against the appropriate critical values, the p-values. The test-value 𝜃_𝑖 follows a t-distribution with N-1 degrees of freedom. The test-value 𝜃1 is given by:

𝜃₁ = 𝐶𝐴𝑅(𝜏1, 𝜏2) √𝑣𝑎𝑟 (𝐶𝐴𝑅(𝜏₁, 𝜏₂))

The frequently used p-values are α=10%, α=5% and α=1% that is used in this study as well. Step 11: Testing the differences

At last, a Welch’s t-test is done to test the hypothesis that two selected samples have equal means, assuming the two samples have unequal variances and unequal sample sized. It tests if the

(23)

𝑡 = 𝑋̅1− 𝑋̅2 √𝑠12

𝑁1+ 𝑠₂2 𝑁2 With the degrees of freedom:

𝜈 ≈ (_𝑁𝑠12 1+ 𝑠₂2 𝑁₂) 2 𝑠₁4 𝑁₁2(𝑁₁− 1)+ 𝑠₂4 𝑁₂2(𝑁₂− 1)

(24)

4 Data collection

4.1 Overview

In order to select the right data to perform this study, the data should be operationalized. After that, the right sample can be selected from the available databases or hand collected in order to find answers on the formulated hypothesis.

4.2 Define financial misconduct

This study is aiming to look at fraudulent financial reporting disclosure on social media platforms, and to look at the consequences of this disclosure on firm value. To select a qualified research sample, a sample of firms that committed a financial misconduct should be identified.

This study uses firms that were investigated by the SEC and what resulted in the issuing of an AAER (Accounting and Auditing Releases, or AAERs). Several studies used firms that have been cited by the SEC in its AAERs as a proxy for fraudulent financial reporting [Dechow et al. (1996), Hennes et al. (2008) and Dechow et al. (2011)] and, because of multiple similarities within this study, it is a good proxy for this study as well.

In an AAER, the SEC has taken an enforcement action against a firm or individual it identified as having violated the financial reporting requirements of the Securities Exchange Act of 1934 (Grove and Basilico, 2008). Since 1982, the SEC has issued AAERs during or at the conclusion of their investigation of a firm, auditor, or an officer for alleged accounting and/or auditing misconduct. The releases provide varying degrees of detail on the nature of the misconduct, the individuals and entities involved and their effect on the financial statements. In this study, data is collected from the SEC website and their database is used to hand collect the AAERs the SEC has issued. AAERs are issued for a wide range of violations. A firm is included in the sample if an AAER is identified on the SEC website and that release is related to an accounting fraud committed by the firm. Most of the time a firm only receives one AAER, however often the SEC issued additional AAERs against individual officers, accountants, and occasionally the audit firm. The AAERs are normally issued together, resulting in one reporting event, with multiple AAERs only if individuals and/or the audit firm are being singled out for censure (Nicholls, 2015).

(25)

4.3 Define corporate Social Media use

In this study four social media channels are included – RSS feeds, corporate blogs, Facebook, and Twitter. This is in line with the research of Lee et al. (2015), they were one of the first to look at the impact of social media on stock returns after a negative event disclosure. However, in their study, they were not able to include the feed of Facebook or the presence of messages’ since there was not a good tool available and they were mainly focusing on Twitter. However, in this study, the presence of a Facebook message is included. In order to look at the announcements on the firm’s corporate Facebook a new data collection and extraction application is used, called Netvizz. Netvizz can extract posts and give an insight into the activity around those posts, for example, post source, the number of likes and comments. This study wants to select a dataset including all posts from page administrators within the selected time (the announcement day) to see whether they announced the issued AAER through Facebook. (Rieder, 2013).

In this study the sample of firms are the firms dealing with an AAER, of those firms the social media channels the social media channels are identified. For Facebook and Twitter, the corporate Website is checked whether there is a link to a Facebook and/or Twitter account. For RSS feeds and blogs, it is first determined from the corporate website whether the firm provides RSS feeds or has a blog. These four social media channels are especially useful for answering Hypothesis 2 since it investigates whether there is a different effect of non-interactive and interactive social media. RSS feeds and corporate blogs are classified as non-interactive and, Twitter and Facebook as interactive social media. This study is especially interested in the period from 2006 and onwards because in 2006 Twitter was developed.

4.4 Define market reaction

In order to perform the study, a proper measure of the market reaction should be in place. Market reaction is measured trough stock return, especially the changes in stock return following a financial misconduct. So in order to fully study the data to measure stock return, it should be available. Because of that, this study focuses on firms that are listed in the S&P 500. The daily stock returns of those firms are stated in a database called, Center for Research in Security Prices (CRSP). When performing this study, the available data in CRSP is only until December 2015. Since there are some events where the post-event window is in 2016 we collected the stock prices of those particular firms with the help of Compustat.

The returns on the stocks are going to be compared with the average market return of the S&P 500 index. This data is also available in CRSP. However, with this data, there is the same

(26)

problem that the data in CRSP is only available until December 2015. The extra data that is needed for this study is taken from the yahoo finance website (YAHOO, 2016).

4.5 Data sample

In this study, data is used from different databases made available by Wharton Research Data Services (WRDS). First, the Compustat database is used to gather the S&P 500 index constituents. Second, the daily stock prices of those firms and the market return on the S&P 500 are gathered in the CRSP database. The rest of the data is hand collected, what is further discussed in this section.

In Table 2, the sample selection process is summarized. For this study, the base sample was selected from the S&P 500 index. From the firms on this list, the daily stock returns are available to analyze and used in the event study. The initial sample contains approximately 600 firms that were listed in the S&P 500 during the period 2011-2015. The firms’ names distracted from the CRSP database were then used for the search on the SEC’s website. The website of the SEC according to issued AAERs, were checked to find whether there were firms of the S&P 500 on those lists in the period 2011-2015. This period was selected because it captures the acknowledgment of the SEC, to use social media as a corporate disclosure medium right in the middle, 2013. There were 33 AAERs issued against firms that are on this list. However, from two firms there were no stock prices available during the whole research period so these are eliminated from the sample. There were also four firms that had two AAER’s at the same date, these were removed from the list as well because they could bias the results by giving an extra negative reaction and this study tries to treat all AAERs the same. There is one firm that has two AAER’s, however since the two individual incidents are not in each other’s estimation window, they are both included in the sample.

Since social media is operationalized by the use of a Website, Blog, Twitter and/or Facebook, the presence of those media is checked for the 27 firms. The Facebook, Twitter, and blog accounts are checked by first searching for a link on their corporate website. The collected data shows that 78% use Facebook, 89% use Twitter, 22% use a blog and all firms have a corporate website (Table 3). Then, to see if those firms use these media to disclose the issued AAER, the pages are checked around the announcement date. This shows that 0% used their Facebook account, 13% used their twitter account, 0% used their blog and, 26% used their corporate website to disclose the issued AAER (Table 4).

(27)

Table 2 : Sample selection

AAER issued between 2011-2015 508

AAER issued for firms that are not in the S&P 500 -475

Subtotal 33

Firms with an AAER that are not in the S&P 500 the full time of this study -2

Subtotal 31

Firms that got an AAER twice on the same date are excluded -4 Number of firms with an AAER in the S&P 500 - Final sample 27

Table 3: Sample description – Firms and their social media

Nr. Date Firm Facebook Twitter Blog Website

1 24-3-2011 Ball Corporation 1 1 0 1

2 8-4-2011 Johnson & Johnson 1 1 1 1

3 3-5-2011 Rockwell Automation, Inc. 1 1 1 1

4 3-5-2011 Tenet Healthcare Corporation 0 1 0 1

5 14-6-2011 Huntington Bancshares, Inc. 1 1 0 1

6 29-7-2011 Waste Management 1 1 1 1

7 29-9-2011 Time Warner Inc. 1 1 1 1

8 20-12-2011 Aon Corporation 1 1 0 1

9 25-4-2012 Morgan Stanley 0 1 0 1

10 8-8-2012 Pfizer Inc. 1 1 0 1

11 24-9-2012 Tyco International Ltd. 0 0 0 1

12 24-4-2013 Capital One Financial Corporation 1 1 0 1

13 26-4-2013 Time Warner Inc. 1 1 1 1

14 3-6-2013 PACCAR Inc. 0 0 0 1

15 19-9-2013 JPMorgan Chase & Co. 1 1 0 1

16 10-10-2013 Celgene Corp 0 1 0 1

17 4-12-2013 Fifth Third Bancorp 1 1 0 1

18 9-1-2014 Alcoa Inc. 1 1 0 1

19 8-4-2014 CVS Caremark Corporation 1 1 0 1

20 25-6-2014 Regions Financial Corporation 1 1 0 1

21 27-6-2014 Nvidia Corporation 1 1 1 1

22 7-7-2014 Noble Corporation 0 0 0 1

23 29-9-2014 Bank of America Corporation 1 1 0 1

24 24-2-2015 The Goodyear Tire & Rubber Company 1 1 0 1

26 18-8-2015 The Bank of New York Mellon Corporation 1 1 0 1

27 15-9-2015 Computer Sciences Corporation 1 1 0 1

Total 21 24 6 27

Percentage 78% 89% 22% 100%

(28)

Table 4: Firms that use social media to disclose issued AAER

Nr. Date Firm Facebook Twitter Blog Website

1 24-3-2011 Ball Corporation 0 0 0 0

2 8-4-2011 Johnson & Johnson 0 1 0 1

3 3-5-2011 Rockwell Automation, Inc. 0 0 0 1

4 3-5-2011 Tenet Healthcare Corporation 0 0 0 0

5 14-6-2011 Huntington Bancshares, Inc. 0 0 0 0

6 29-7-2011 Waste Management 0 0 0 0

7 29-9-2011 Time Warner Inc. 0 0 0 0

8 20-12-2011 Aon Corporation 0 0 0 1

9 25-4-2012 Morgan Stanley 0 0 0 0

10 8-8-2012 Pfizer Inc. 0 1 0 1

11 24-9-2012 Tyco International Ltd. 0 0 0 0

12 24-4-2013 Capital One Financial Corporation 0 0 0 0

13 26-4-2013 Time Warner Inc. 0 0 0 0

14 3-6-2013 PACCAR Inc. 0 0 0 0

15 19-9-2013 JPMorgan Chase & Co. 0 0 0 0

16 10-10-2013 Celgene Corp 0 0 0 0

17 4-12-2013 Fifth Third Bancorp 0 0 0 0

18 9-1-2014 Alcoa Inc. 0 1 0 1

19 8-4-2014 CVS Caremark Corp 0 0 0 1

21 27-6-2014 Nvidia Corporation 0 0 0 0

22 7-7-2014 Noble Corporation 0 0 0 0

23 29-9-2014 Bank of America Corporation 0 0 0 0

24 24-2-2015 The Goodyear Tire & Rubber Company 0 0 0 0

26 18-8-2015 The Bank of New York Mellon Corporation 0 0 0 0

27 15-9-2015 Computer Sciences Corporation 0 0 0 0

Total 0 3 0 7

(29)

5 Results

5.1 Overview

This part of the thesis shows the results found during the event study for all four hypothesis. With the help of a special designed Excel file, this study tests the formulas used in this research of Mackinlay (1997). Then the performed T-tests shows whether the results significantly differ to reject or not reject the hypothesis

5.2 Testing Hypothesis 1 and 1a

This study evaluates the market reaction of 27 individual events, namely issued AAERs. The first step is to test whether in general conducting a financial misconduct leads to a negative price reaction. This led to the following hypothesis:

H1: A financial misconduct announcement leads to a negative price reaction for the firm’s stock price

When performing the t-test, it gives significant statistical evidence that a financial misconduct announcement leads to a negative price reaction, with an average cumulative abnormal return of -0,0060 (α=10%) for event window [0,+1], see Table 5. This means that Hypothesis 1 is not rejected. The theoretical framework suggests that for firms that use social media to disclose their issued AAER the price reaction is bigger than firms that do not use social media. This led to the following hypothesis:

H1a: The price reaction for firms that use social media is larger than for firms that do not use social media following a financial misconduct

Table 5, shows the abnormal returns during the event window for all 27 firms. When evaluating those abnormal returns it is noticeable that not all events lead to a negative price reaction during the event window. The biggest positive abnormal return shows after the Celgene corporation filling (0.036), and the biggest negative abnormal return after the filling of Waste Management (-0.036).

(30)

Table 5: Abnormal returns during 2 days event window

Firm Day 0 Day +1

Ball corporation -0,0033 0,0047

Johnson & Johnson 0,0022 0,0086

Rockwell Automation, Inc. -0,0246 -0,0116

Tenet Healthcare Corporation -0,0124 -0,0384

Huntington Bancshares, Inc. -0,0075 0,0362

Waste Management -0,0313 0,0032

Time Warner Inc. -0,0107 0,0010

Aon Corporation 0,0032 0,0002

Morgan Stanley -0,0432 -0,0092

Pfizer Inc. 0,0027 0,0007

Tyco International Ltd. 0,0041 -0,0132

Capital One Financial Corporation 0,0132 -0,0049

Time Warner Inc. 0,0049 -0,0002

PACCAR Inc. -0,0067 0,0100

JPMorgan Chase & Co. 0,0033 0,0016

Celgene Corp 0,0077 -0,0062

Fifth Third Bancorp -0,0044 0,0014

Alcoa Inc. -0,0132 -0,0566

CVS Corporation -0,0128 0,0068

Regions Financial Corporation 0,0035 0,0021

Nvidia Corporation -0,0012 0,0088

Noble Corporation -0,0075 0,0108

Bank of America Corporation 0,0018 0,0056

The Goodyear Tire & Rubber Company -0,0043 0,0016

Regions Financial Corporation -0,0211 0,0200

The Bank of New York Mellon Corporation 0,0089 0,0040

Computer Sciences Corporation 0,0023 -0,0038

AR -0,0054 -0,0006

CAR -0,0054 -0,0060

To find an answer whether social media leads to a bigger price reaction, the cumulative abnormal returns of the sample with 20 firms that do not use social media (Table 6) are compared with the 7 firms that also use their website (Table 7) to disclose the issued AAER. Unfortunately, this research shows that firms do not disclose their AAER filling on Facebook or their blog. Social media use is therefore defined as the disclosure on their website or Twitter. This study shows that the firms that use twitter to disclose the AAER also use their website so there are two samples to test the Hypothesis 1a. The no social media use sample shows that there is no significant statistical evidence that the price reaction of the market differs from those prices that were expected. The social media sample shows that there is significant statistical evidence (α=5%) that the price reaction of firms that use social media is (negatively) larger than the values that were expected.

(31)

Table 6: Abnormal returns during 2 days event window No-Social Media use

Firm Day 0 Day +1

Ball corporation -0,0033 0,0047

Tenet Healthcare Corporation -0,0124 -0,0384

Huntington Bancshares, Inc. -0,0075 0,0362

Waste Management -0,0313 0,0032

Time Warner Inc. -0,0107 0,0010

Morgan Stanley -0,0432 -0,0092

Tyco International Ltd. 0,0041 -0,0132

Capital One Financial Corporation 0,0132 -0,0049

Time Warner Inc. 0,0049 -0,0002

PACCAR Inc. -0,0067 0,0100

JPMorgan Chase & Co. 0,0033 0,0016

Celgene Corp 0,0077 -0,0062

Fifth Third Bancorp -0,0044 0,0014

Nvidia Corporation -0,0012 0,0088

Noble Corporation -0,0075 0,0108

Bank of America Corporation 0,0018 0,0056

The Goodyear Tire & Rubber Company -0,0043 0,0016

Regions Financial Corporation -0,0211 0,0200

The Bank of New York Mellon Corporation 0,0089 0,0040

Computer Sciences Corporation 0,0023 -0,0038

AR -0,0054 0,0017

CAR -0,0054 -0,0037

After that, to test the H1a a Welch’s t-test is conducted. Since the CARavg for no-social media use is -0.0037 and for social media -0.0127 this study tests whether the difference between the two is significantly bigger than zero.

H0: CARavg no-Social Media sample – CARavg Social Media sample = 0 H1: CARavg no-Social Media sample – CARavg Social Media sample > 0

Table 7: Abnormal returns during 2 days event window Social media use

Firm Day 0 Day +1

Johnson & Johnson 0,0022 0,0086

Rockwell Automation, Inc. -0,0246 -0,0116

Aon Corporation 0,0032 0,0002

Pfizer Inc. 0,0027 0,0007

Alcoa Inc. -0,0132 -0,0566

CVS Corporation -0,0128 0,0068

Regions Financial Corporation 0,0035 0,0021

AR -0,0056 -0,0071

(32)

This test shows if the CARavg of stock prices is significantly higher when using social media or if this outcome is just a coincidence. When performing the Welch’s t-test by comparing the results of no-social media and social media use it show significant evidence with a p-value of 0.003 (α=1%) that the difference between the two samples is not equal to zero. The price reaction to social media use is larger than the firms that do not use social media. Therefore, the t-test finds a significantly worse average CAR for firms that use social media than firms that do not use social media.

To conclude, for Hypothesis 1 this research finds significant statistical evidence that the financial misconduct announcement leads to a negative price reaction. It shows for Hypothesis 1a that there is significant statistical evidence that social media use leads to a more negative CAR than firms that do not use social media. The results of the t-test show that there is no reason to reject H1 and H1a.

Table 8: Statistical test Hypothesis 1a

No-Social Media Social Media Welch's Test

CARavg -0,0037 -0,0127

T-value -0,7019 -2,3981 3,8593

P-value 0,2456 (-) 0,0267 (**) 0,0030 (***) 5.3 Testing Hypothesis 2

After testing Hypothesis 1, the results show that there is a significant difference in the price reaction after a financial misconduct in the social media use sample. This study also tries to investigate whether there is a difference between the types of social media a firm uses to disclose the issued AAER. In this study, interactive social media are corporate Facebook and Twitter accounts, and non-interactive social media are corporate websites and blogs. Based on previous research the following hypothesis was formulated:

H2: Compared to non-interactive corporate social media, interactive corporate social media has a larger price reaction to a financial misconduct announcement

Since only seven firms use social media to disseminate the issued AAER, it is hard to find significant evidence within this sample. Of the seven firms that use social media, three firms use also Twitter as a disclosure channel. The sample of seven firms is split in the four firms that use only their website and the three firms that also use Twitter as a disclosure channel. For both samples, the average cumulative abnormal return is calculated, and a t-test performed what led to a t-value and corresponding p-value. The firms that only use their website to disseminate the issued AAER shows that there is no significant statistical evidence that the price reaction of the stock

(33)

there is significant statistical evidence (α=10%) that the price reaction of firms that use social media is (negatively) larger than the values that were expected.

After that, this study performs a Welch’s t-test to test the H2. Since the CARavg for Website use is -0.0083 and for Website and Twitter use -0.0185 it tests if the difference between the two is significantly bigger than zero.

H0: CARavg Website sample – CARavg Website&Twitter sample = 0 H1: CARavg Website sample – CARavg Website&Twitter sample > 0

When performing a Welch’s t-test by comparing the results of Website users with Website and Twitter users, the test shows no significant evidence to reject the H0 and that there is a difference between the two samples. However, the t-statistic has a p-value of 0.152, which is quite close to the 10% significance level. To conclude there is no significant statistical evidence that non-interactive social media leads to a larger price reaction after a financial misconduct. This means that Hypothesis 2 is rejected.

Table 9: Statistical test Hypothesis 2

Website Website and Twitter Welch's Test

CARavg -0,0083 -0,0186

T-value -1,1709 -2,3337 1,7671

P-value 0,1631 (-) 0,0724 (*) 0,1520 (-) 5.4 Testing Hypothesis 3

The last hypothesis tested in this study is to see whether there is an effect regarding the approval of the SEC to use social media as a disclosure channel. Since there is no statistical, evidence found in Hypothesis 2 and because of the limited size of the samples there is no distinction possible between interactive and non-interactive social media regarding this hypothesis. To test if the SEC’s approval influences the market reaction based on previous literature, the following hypothesis was formulated:

H3: Since the SEC’s approval of the use of social media to disclose important firm information to stakeholders, the negative market reaction declined

Again only seven firms use social media to disseminate the issued AAER. Of the seven firms that use social media, four had to disclose an issued AAER before the date the SEC approved social media use as a disclosure channel (04-02-2013). The sample of seven firms is split into the four firms that disclosed their issued AAER before 04-02-2013 on social media and the three firms that disclosed their issued AAER after 04-02-2013 on social media. For both samples, the average

(34)

cumulative abnormal return is calculated, and a t-test performed what led to a t-value and corresponding p-value. The results of firms that use social media before 04-02-2013 show that there is no significant statistical evidence that the issued AAER disclosure on social media leads to negative abnormal returns. The sample with firms that use social media after 04-02-2013 shows that there is significant statistical evidence (α = 10%) that the issued AAER disclosure leads to negative abnormal returns.

After that, a Welch’s t-test is conducted to test the H3. Since the CARavg for the Before SEC sample e is -0.0047 and for the After SEC sample -0.0233 it is in contrast with what the theory expected that after the SEC approval the negative price reaction would decline. Therefore, instead of testing whether the negative market reaction declined is significant, the bigger negative market reaction is tested. Instead of testing whether H1 is smaller than zero it tests of the mean difference between the two is significantly bigger than zero.

H0: CARavg Before SEC sample – CARavg After SEC sample = 0 H1: CARavg Before SEC sample – CARavg After SEC sample > 0

When performing a Welch’s t-test by comparing the results of the Before SEC sample with the After SEC sample, it shows significant statistical evidence (α = 10%) that H0 should be rejected and there is a difference between the two samples. This means that in contrast with the expected outcome in this study, the approval of social media disclosure of the SEC leads to a bigger market reaction. To conclude, this means that H3 is rejected but there is significant evidence that shows a bigger market reaction after the approval of the SEC.

Table 10 Statistical test Hypothesis 3

Before SEC After SEC Welch's Test

CARavg -0,0047 -0,0234

T-value -0,7394 -2,5917 -3,0693

(35)

6 Sensitivity analysis

6.1 Overview

In addition to the results found and discussed in the previous section, some sensitivity analyses are performed. These show the sensitivity of the p-values for the length of the estimation and event window and what the consequences are whether to reject or not reject the hypotheses. In this section, the results of the Market Model will also be compared with the results of the Constant Mean Return Model to see if this leads to any big differences.

6.2 Constant Mean Return Model

This study uses the Market Model (MM) to test the hypothesis by calculating the abnormal returns. However as already mentioned in paragraph 2.4, there is also a Constant Mean Return Model (CMRM) that can be used. In that paragraph, it is also mentioned that according to Brown and Warner (1980) the results of the CMRM are not significantly different from the results of the MM. This statement is checked by performing the same tests for the four hypothesis but then instead of using the MM, the CMRM is used.

When comparing the p-value of the MM with the CMRM it shows that for Hypothesis 1 it does not lead to any differences. It still shows significant statistical evidence (a=10%) that a financial misconduct announcement leads to a negative price reaction for the firm’s stock price.

When comparing the p-values for Hypothesis 1a, there is also little difference, but the statistical evidence shifts to another category of two stars, namely a=5% instead of a=1%. For Hypothesis 2 there was not any statistical evidence within the MM, and with the CMRM this shifts even further away from being significant. Therefore, this does not lead to another outcome as well. Quite interesting is the outcome of the third hypothesis. Where Hypothesis 3 was significant for the MM with a p-value of 0,055 for the CMRM the p-value increases to 0,43900, this is not even close to being significant. On the one hand, this could be interpreted as evidence that the conclusion of Ball and Brown (1980) is not correct. However, since the limited size of the sample, it seems more logic that it has to do with the lack of data. To conclude, besides the third hypothesis, the sensitivity analysis did not lead to significant differences in rejecting or not rejecting the hypotheses.

The impact of social media on firm value following a financial misconduct

Abstract

Contents

1 Introduction

2 Theoretical Framework

3 Methodology

4 Data collection

5 Results

6 Sensitivity analysis