An Analysis of the Effect of Information Activism on Capital Markets: Investor Behavior and Divergent Market Conditions

Home , Closing Bell, Fast Money (talk show), Jim Cramer, Mad Money

A dissertation submitted to the Kent State University Graduate School of Management in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Laura K. Rickett

May 2011

Dissertation written by

Laura K. Rickett

B.S.B.A., Bowling Green State University, 1989

M.B.A., The University of Akron, 1997

Ph.D., Kent State University, 2011

Approved by

______Chair, Doctoral Dissertation Committee Dr. Pratim Datta ______Members, Doctoral Dissertation Committee Dr. Alan Brandyberry ______Dr. Indrarini Laksmana ______Dr. Linda Zucca

Accepted by

______Doctoral Director, Graduate School of Management Dr. Murali Shanker ______Dean, Graduate School of Management Dr. Frederick Schroath

ACKNOWLEDGEMENTS

I wish to thank the many people in my life who provided encouragement, support, and sacrifice in order to make the completion of this dissertation and the doctoral program possible. I would like to begin by thanking my dissertation chair, Dr. Pratim Datta, whom without his insight and encouragement, the conceptualization of this dissertation would have not been possible. He not only provided support and motivation throughout this process, but was a constant source of inspiration through his many achievements in research and otherwise. I also owe a deep gratitude to my dissertation committee members. As a committee member, Dr. Indrarini Laksmana was instrumental in providing key guidance in the methodological development. In addition, she provided essential feedback and challenges to greatly improve this dissertation. I also want to thank my other committee members Dr. Alan Brandyberry and Dr. Linda Zucca who also were extremely supportive and offered critical feedback on my drafts as well as crucial insight. I especially want to thank all my committee members for their overall dedication, sacrifice, and commitment in helping me to complete this dissertation in such a timely manner and often on tight schedules particularly given their other important obligations. I will forever be grateful to you all. I wish to also thank the friends and family members which are unfortunately too many to name. My friends and family who provided encouragement, helped with our children, or were just there to listen, I can’t thank you enough as you were instrumental in my completion of this dissertation and the doctoral program. In particular, I thank my husband Todd and our four amazing children, Ellie, Hanna, Grace, and Jake, whom without you I could not achieve anything. You all make my life so joyous and you are my greatest gifts. Todd, I thank you for your unwavering support, even when times were tough and also for the many sacrifices you have made to allow me to achieve this accomplishment. You, at times, do more than any Dad or Mom I know and I would not want to venture on this journey without you. Thanks to our children for their understanding when I was not there for them as much as I would like and for understanding at a young age the importance of sacrifice and commitment. Thanks especially to Jake who came into our life the first year of the Ph.D. program and although many thought it impossible to take on this challenge with a new baby, you always made me smile even when at times I felt discouraged. You and your sisters gave me the inspiration to keep pushing forward and to never give up. Finally, I want to thank my parents, James and Sherry, for instilling in me the value of hard work, dedication and believing in myself. They always taught me that you can achieve anything if you work hard enough and this is proof of that belief. I want to dedicate this dissertation to my late father who I know would be proud and I thank him and my mother for providing a good example of a strong work ethic and values. Above all else, I thank God who guides me each day and whom I call on constantly for strength and guidance and offer thanks for my many blessings.

TABLE OF CONTENTS

CHAPTER 1 INTRODUCTION…………………………………………...………………...….... 1 1.1 Overview……………………………………………………………………………. 1 1.1.1 CaseinPoint………………………………………………………………………... 1 1.1.2 Demand and Growth of Infomediaries……………………………..……………….. 2 1.1.3 A Downside of Infomediation………………………………………………………. 3 1.1.4 Infomediaries: Two Sides of the Coin……………….…………………………….... 6 1.2 Information Activism Defined...…………………………………………………...... 7 1.2.1 Activism……………………………………………………………………………... 7 1.2.2 Activism in Accounting……………………………………………………………... 8 1.2.3 Reliance on Information Activists…………………………………………………... 9 1.2.4 Aspects of Information Activism…………………………………………………..... 9 1.3 Continued Motivation: Bridging the Information Divide………………………….... 10 1.3.1 Information: Beyond the Financial Statements…………………………………….. .. 10 1.3.2 Importance of Market Conditions………………………………………………….... 12 1.3.3 Sophisticated vs. Unsophisticated Investors……………………………………….... 13 1.3.4 A Changing Capital Market………………………………………………………..... 13 1.4 Research Objectives………………………………………………………………..... 14 1.4.1 Modus Operandi…………………………………………………………………...... 15 1.5 Contributions………………………………………………………………………... 16

CHAPTER 2 RELATED PRIOR LITERATURE & THEORETIAL FOUNDATION…….… 19

2.1 Signal Theory & Information Asymmetry………………………………….……...... 19 2.2 Shareholder Activism……………………………………………………….………. 20 2.3 Media Coverage……………………………………………………………………... 22 2.4 Investor Behavior……………………………………………………………………. 23 2.5 Information Intermediaries and Online Stock Recommendations…...... 26 2.6 Risk & Loss Aversion……………………………………………………………...... 28 2.7 Literature Synopsis……………………………………………………….…….…… 29

CHAPTER 3 HYPOTHESES DEVELOPMENT & RESEARCH MODEL………….…...... 30

3.1 Investor Behavior………………………………………………………………...... 30 3.1.1 Price Reaction………………………………………….……………….………...... 32 3.1.2 Trading Volume………………………………………………………………..….... 33 3.1.3 Sentiment of Information Activism…………………………………………….…... 35 3.2 Moderating Effects………………………………………………………………..... 35 3.2.1 Investor Sophistication………….…………………………………………….…….. 36 3.2.2 Market Condition..………………………………………………………………...... 37 3.2.3 Information Asymmetry…………………………………………………………….. 38 3.2.4 Earnings Quality……………………………………………………………….….... 39 3.3 Research Model…………………………………………………………………...... 40 CHAPTER 4 RESEARCH DESIGN…………………………………………....……………...... 42 4.1 Sample Time Periods and Sample Selection………………………………………... 42 4.1.1 Sample Time Periods……………………………………………………………...... 42 4.1.2 Sample Selection…………………………………………………………………..... 43 4.1.3 Other Data Sources……………………………………………………………….. … 46 4.2 Sample Characteristics & Confounding Events…………………………………...... 47 4.2.1 Sample Characteristics…………………………………………………………….... 47 4.2.2 Confounding Events……………………………………………………………...... 47 4.3 Methodological Framework……………………………………………………….... 49 4.3.1 Cumulative Abnormal Return – Univariate Analysis…………………………...…... 49 4.3.2 Abnormal Trading Volume – Univariate Analysis………………………...……….. 51 4.3.3 Information Activism Sentiment – Univariate Analysis……………………...…….. 52 4.3.4 Moderating Effects – Univariate Analysis……………………………………....….. 52 4.3.5 Crosssectional Regressions – Multivariate Analysis…………………………...….. 53 4.3.6 Regressions Functions and Variable Definitions………………………………...... 54 4.3.7 Earnings Quality………………………………………………………………...….. 59

iii

CHAPTER 5 EMPIRICAL RESULTS………………………….…………………………..….. 62 5.1 Univariate Results……………………………………………………………...... 6.2 5.1.1 Effect of Information Activism on Investor Behavior – Returns…………..……..... 62 5.1.1.1 Returns – Overall (H1a)…………………………………………………………..... 62 5.1.1.2 Returns – Sentiment (H1c)…………………………………………………………. 62 5.1.1.3 Returns – Investor Sophistication (H2)…………………………………………….. 63 5.1.1.4 Returns – Market Condition (H3)………………………………………………….. 65 5.1.1.5 Returns – Information Asymmetry (H4)………………………………………….... 66 5.1.1.6 Returns – Earnings Quality (H5)……………………………………………….….. 67 5.1.2 Effect of Information Activism on Investor Behavior – Trading Volume…….…... 67 5.1.2.1 Trading Volume – Overall (H1b)………………………………………………….. 67 5.1.2.2 Trading Volume – Investor Sophistication (H2)…………………………………... 68 5.1.2.3 Trading Volume – Market Condition (H3)…….………………………………….. 69 5.1.2.4 Trading Volume – Information Asymmetry (H4)…………………………………. 69 5.1.2.5 Trading Volume – Earnings Quality (H5)…………………………………………. 70 5.2 Multivariate Results………..……………………………………………………..... 71 5.2.1 Intensity…………………………………………………………………..………... 71 5.2.2 Sentiment……………………………………………………………………..……. 73 5.2.3 Moderating Effects……………………………………………………………...…. 75 5.2.4 Control Variables………………………………………………………………….. 77 5.2.5 Unexpected Results………………………………………………………………... 77

CHAPTER 6 SUMMARY & CONCLUSION, LIMITATIONS, & FUTURE RESEARCH... 80 6.1 Summary and Conclusion……………………………………………………...... 80 6.1.1 Univariate Summary……………………………………………………………...... 81 6.1.2 Multivariate Summary……………………………………………………………... 82 6.2 Contributions…………………………………………………………………..…... 84 6.3 Limitations………………………………………………………………………..... 85 6.4 Future Research…………………………………………………………………..... 86 REFERNCES…………………………………………………………………………...……….... 93

LIST OF FIGURES

Figure 1 Basic Model…………………………………………….…….. 88

Figure 2 Research Model……………………………………………… 88

Figure 3 Dow Jones Industrial Average……………………………… 89

Figure 4 Average Monthly Returns Bull/Bear Periods……………… 90

Figure 5 Summary of Hypotheses & Findings……………………….. 92

LIST OF TABLES

Table 1 Event Data Initial Sample Sources…………………….……. 102

Table 2 Descriptive Statistics – Initial Sample………………………. 103

Table 3 Final Sample Selection Detail – Univariate Analysis………. 104

Table 4 Univariate Analysis – CAR (All Events) ……………………. 105

Table 5 Univariate Analysis – CAR (Cramer) ………………….……. 107

Table 6 Univariate Analysis – CAR (Blog) ………………….……….. 109

Table 7 Univariate Analysis – ABVOL (All Events) …...………...... 111

Table 8 Univariate Analysis – ABVOL (Cramer) ……………………. 113

Table 9 Univariate Analysis – ABVOL (Blog) ……………………….. 115

Table 10 Descriptive Statistics for Multivariate Analysis Variable…. 117

Table 11 Pearson & Spearman Correlation Matrix……………….… 119

Table 12 INTENSITY Regression Results – CAR…………... ……...... 120

Table 13 INTENSITY Regression Results – ABVOL...……... ……...... 122

Table 14 SENTIMENT Regression Results – CAR...……….. ……...... 124

Table 15 SENTIMENT Regression Results – ABVOL..…….. ……...... 126

CHAPTER 1

INTRODUCTION

1.1 Overview

1.1.1 CaseinPoint

In March of 2009 a weeklong battle between The Daily Show host, Jon Stewart, and the host of CNBC’s Mad Money , Jim Cramer, culminated in a thrashing of Cramer when he appeared on Stewart’s show for the faceoff. Stewart accused the former hedge fund manager of providing distorted financial advice and prioritizing the entertainment factor in his show over objective investment guidance (Lloyd 2009). Furthermore, Stewart’s harsh criticism of CNBC in general went so far as to describe the network’s misrepresentation of the financial crisis as “some sort of crazy onceinalifetime tsunami that nobody could have seen coming” as “disingenuous at best and criminal at worst.” Stewart cited several scenarios where CNBC was dead wrong on the advice offered regarding key debacles in the financial meltdown. Stewart argued that CNBC should be a source of enlightenment particularly during such turbulent economic times, when in fact CNBC completely dropped the ball during the financial meltdown. Cramer fought back by admitting that he wasn’t perfect and has made mistakes. He argued that although he strives to make his investment show entertaining, he also provides quality investment advice to his viewers.

Stewart called for Cramer and other similar broadcasts to practice responsible journalism since they portray themselves as a source of superior investment advice which many viewers may rely upon. Furthermore, Cramer often prompts his audience to act in a particular manner with regard to

buying or selling a particular security. This dissertation examines the market reaction to various

infomediary sources which advocate that investors take a specific action with regard to investment

decisions.

1.1.2 Demand and Growth of Infomediaries

There is an increasing demand for financial investment information provided by various

information intermediaries (Healy & Palepu 2001). Investors often rely on such information as a

result of information asymmetry. Information asymmetry exists when investors lack the superior

information held by managers and therefore face challenges in making optimal investment

choices. The supply of this information has increased through considerable growth in a variety of

information intermediary sources. The Internet has exploded with growth in the blogosphere

(Cheng 2007) as well as the expansion of cable news networks such as CNBC. Thousands of

viewers tune in daily to cable investment news programs. CNBC, with an average viewership of

310,000 according to Neilson (Hempel 2008), broadcasts investment news programs around the

clock. Programs including “Street Signs ,” “ Closing Bell ,” “ Fast Money ,” and “ Mad Money,” are

often associated with major swings in the market. One example is the observed spikes of Jim

Cramermentioned stocks on Mad Money, often referred to as the “CNBC Effect,” or in this case

the “Jim Cramer Effect” (Cooper 2008). Recent academic literature provides support for this

effect (Engelberg et al. 2009; Neumann & Kenny 2007).

The growth seen in Internet communication platforms such as the blogosphere 1 is

astounding and these forms of communication are now often seen as “mainstream” media sources

(Winn 2009). Surveys indicate that approximately 50% of US Internet users are blog readers (94.1

million in 2007) and about 12% of US Internet users are bloggers (22.6 million in 2007)

(www.emarketer.com ). Technorati, a blog Internet search engine, which indexes more than 112.8

million blogs, reported results of a survey of 2,900 bloggers. The survey revealed that many in the blogosphere relied on the uptotheminute information about the financial crisis and some even

suggest that the blogosphere actually contributed to a sense of panic and exacerbated the financial

crisis. Survey respondents also indicated that both politics and business are among the fields most

impacted by the blogosphere which will continue its transformation of these fields in the future

(Sussman 2009). Recent academic literature has examined whether investors consider financial blogs an important source of information and find evidence that capital markets respond to

recommendations provided on financial blogs (Fotak 2008). Also a number of studies examine

similar Internet communication platforms such as stock message boards and overall provide

evidence that the market appears to react to such recommendations (Tumarkin & Whitelaw 2001;

Das & Chen 2001; Antweiler & Frank 2004).

1.1.3 A Downside of Infomediation

To what extent do investors rely on the investment advice offered by commentators like

Jim Cramer and financial bloggers? Particularly during fragile economic periods such as that

1“Blogosphere” is defined as “all of the blogs on the Internet as a collective whole.” “Blog” is a contraction of the term “Web Log.” A “blog” is defined as “a Web site that contains an online personal journal with reflections, comments, and often hyperlinks provided by the writer.” ( www.merriamwebster.com ).

leading up to the financial crisis in 2008, investors may rely heavily on financial advice. Days before the collapse of Bear Stearns in March 2008, Jim Cramer urged investors not to move their money from the investment banking giant (Gomstyn 2008). Following the downfall of Lehman

Brothers in October 2008, the largest bank failure in history, Cramer warned investors that they should take any funds needed within the next five years out of the stock market immediately

(Celizic 2008). It has been speculated whether statements such as these by Jim Cramer and others, could have contributed to the financial crisis and whether CNBC severely misrepresented one of the most devastating financial crises in history (Burrough 2008). CNBC’s “investotainment” is often characterized as capitalizing on the anxiety, greed, and emotions of wealthy investors by offering hope when the market is volatile (Hempel 2008). The CNBC network may face criticism at times, similar to the disapproval directed by Stewart, but in a continually uncertain economy the network remains popular (Stelter and Arango 2009) and investors appear to continue to tunein possibly in hopes of growing their investments and averting losses.

While many sources of investment news are presented as objective analyses, others often appear biased, which brings into question to what extent do investors rely on information intermediaries? Some suggest that the investment media may have even played a role in the fall of

Bear Stearns during March 2008. Although executives of the former investment banking giant admit they made mistakes and that the firm was weakened by the mortgage crisis, it is difficult to understand how it all unraveled so quickly and so unexpectedly (Burrough 2008). Former Bear

Stearns CEO, Alan Schwartz, adamantly maintains that a group of market speculators who stood to profit from its demise launched a premeditated attack on Bear Stearns. This attack was initiated by a rumor about liquidity concerns at Bears, which was then fueled by the investment

news media and eventually resulted in artificial panic. Apparently these unnamed speculators, reportedly investigated by the SEC, constructed a complex scheme compelling a number of major

Wall Street firms to hold up trades with Bears, and then leaked the news to the media (Burrough

2008). William Cohen who describes the firm’s meltdown in a new book, House of Cards: A Tale of Hubris and Wretched Excess on Wall Street, points out that the 247 cable news networks’ constant swirling chatter and rumoring is dangerous in a market that is strictly a confidence game.

Burrough (2008) put it best when describing statements made by CNBC correspondent Charlie

Gasparino, “Publicly speculating on a firm’s liquidity is akin to shouting “Fire!!!” in a crowded theater; in catastrophic cases it can trigger panic selling (Hamilton 2008). It risks, in other words, becoming a selffulfilling prophecy.” CNBC’s continued coverage of Bear’s liquidity was anything but skeptical (Burrough 2008). These scenarios bring about the question: To what extent do the opinions and commentary offered in financial news broadcasts, such as those on CNBC and other sources of investment information, affect capital markets? Shortly after the fall of Bear

Stearns, the investment banking giant Lehman Brothers, completely collapsed in September 2008 and filed bankruptcy. Although the collapse of Lehman Brothers was likely caused by factors such as heightened borrowing costs, inaction by regulators, and abusive shortselling practices, a prevailing theme in the collapse is referred to as “…a storm of fear enveloping the entire investment banking field …” (Woellert & Onaran 2008; Mullins 2008). Stirring up this storm were investment news networks such as CNBC and its programs like Jim Cramer’s Mad Money as

well as the financial Internet blogosphere which communicates investment opinions and

commentary rapidly across the Internet.

One final scenario is illustrated by the Citicorp sell off in November 2008, where apparently due to irrational doom related to the continuing financial crisis in a market driven by the news media, the stock price plunged by more than 64% over three days (Bawden 2008). The

CEO, as well as a number of analysts at the time, argued that the Citigroup’s share price problems were due more to emotions, fear, and rumormongering rather than to the bank’s actual financial condition. One analyst even went so far as to suggest that “…it would take a depression every bit as large and long as the 1930’s debacle to shake the company’s viability” (Bawden 2008).

Bawden (2008) further suggests that as a result of commentary and similar information spreading across various communication channels, panic contributed to Citigroup’s stock slide. These scenarios highlight the importance of examining how infomediaries urge investors to take a particular action and investigating the associated market response.

1.1.4 Infomediaries: Two Sides of the Coin

These narratives suggest that there is an abundance of investment information, opinion, and commentary, promulgated by investment news programs which affect investor behavior and thus capital markets. Additionally, there are many other sources of investment information which investors appear to react to, such as the financial blogosphere. While much of this investment information is objective financial analysis intended to educate investors, often the investment information includes predisposed opinions intended to sway investors; this is defined in this study as information activism . This research serves to introduce the concept of information activism, identify key sources, and examine its downstream effects on investor behavior and capital markets.

1.2 Information Activism Defined

We often hear the phrase “Information is power” or “Information is our most valuable asset.” In the case of investor decision making, these are both true because relevant information about a stock’s future performance gives the investor the power to make optimal investment choices. Accurate information is the most valuable asset that an investor has in the selection of profitable investments (Graham & Dodd 2009). However, a key concern in capital markets is information asymmetry, where the average investor does not have perfect information about the expected return on an investment. The wellknown “lemons problem” (Akerlof 1970) plagues most markets where sellers (firms) have more information about the quality of a product (stock or investment) than buyers (investors), thereby increasing the risk of adverse selection. Although corporate disclosures in the form of financial reporting and regulatory filings attempt to reduce information asymmetry in capital markets, there remains a demand for information intermediaries who engage in private information production to uncover managers’ superior information (Healy

& Palepu 2001). Therefore, since investors realize they may lack relevant information for their investment decisions, they seek out information intermediaries who will provide them with investment advice. At times, these financial information intermediaries appear to provide more opinionbased rather than factbased financial analysis. Consequently, these experts may serve as activists when providing investment advice and attempt to sway investors.

1.2.1 Activism

Activism in the traditional sense is often thought of as organized protests or an effort to draw attention to a social or political issue. However, in recent years the Internet has fostered these

activities by providing a strategic communication platform which enables individuals to disseminate realtime information and share their views via the blogosphere (Kahan & Kellner

2004).

Marketing research examines consumer activism whereby antibrand communities take on social activist roles related to issues like workplace equality, corporate domination, environmentalism, and marketing propaganda (Hollenbeck & Zinkham 2006). More recently, research has begun to examine the impact of the Internet on social action and the opportunities provided by the World Wide Web which allow strategies and coalition building to be developed without the restrictions of space and time (Hollenbeck & Zinkham 2006).

1.2.2 Activism in Accounting

Activism has been examined in accounting literature in the form of shareholder activism.

Groups of unions or shareholders target firms with proposals related to various corporate decisions such as compensationrelated issues (Ertimur et al. 2008), corporate governance preferences (Del

Guercio et al. 2008), and more recently accounting choices (Ferri and Sandino 2009). Shareholder activism is carried out with shareholder proposals presented at the company’s annual meeting or in a private negotiation situation (Karpoff 2001). Activists generally seek to promote their own personal agendas such as social responsibility or corporate governance reform by filing shareholder resolutions. In any event, research on various forms of activism has come to include the study of direct action, in support of or opposition to, a given cause or issue.

1.2.3 Reliance on Information Activists

Investors have come to rely on key sources of easily accessible information provided by information intermediaries, not only due to imperfect information as a result of information asymmetry, but also due to value relevance, timeliness, and complexity issues with published financial reports. Financial statements may be too complex for the average investor or may not be available in a timely manner (Francis and Schipper 1999). Therefore, investors tend to look toward other sources of information which are easily accessible. Rather than solely providing objective investment analysis, these sources of investment information often actively support or oppose certain investment decisions. These sources of investment information may even persuasively and actively promote or discredit a stock with the intent of offering investment advice or emphasizing specific achievements or failures of a firm. When information is presented as to considerably influence or prompt market participants to act in a particular manner, this is known as information activism. Thus far, this discussion has centered around investment news broadcasts as the information source; however other outlets such as Internet financial blogs and even the firm itself can also be included. Information activism is defined as intentional action stemming from formal and informal sources which provides supplemental communication intended to form and/or sway investor behavior for one or more firms or industries.

1.2.4 Aspects of Information Activism

The primary sources of information activism examined in this study are 1) CNBC financial news program Mad Money hosted by Jim Cramer; and 2) the SeekingAlpha financial Internet blog

(www.SeekinAlpha.com .) This research examines the differential effects of these sources of

information activism on investor behavior in an effort to discern the importance of each source and type of information activism.

While the financial news press can also be a source of information activism, the content of the financial news is often included in the other sources, and therefore will not be specifically examined. However, significant new releases are discussed later as an important control variable in this study.

1.3 Continued Motivation: Bridging the Information Divide

1.3.1 Information: Beyond the Financial Statements

Investors spend considerable time evaluating information for investment decisions.

Investors are at a distinct disadvantage due to information asymmetry in capital markets, because they do not have all relevant information necessary for optimal investment decisions. Although financial statements can reduce information asymmetry, they are not without critical limitations.

Even though financial statements are issued at least annually, and normally quarterly, there can be a considerable lapse between one reporting date and the next. Even during short lapses between reporting dates, changes in a firm’s financial position can occur and may not be revealed to the public before the next reporting date. Therefore, investors are unable to make continual assessments of a firm based on financial statements alone. Sutton et al. (2009) tracked information used by investors and find that virtually all sophisticated investors and 79% of unsophisticated investors examined the firm’s website for current summary information which is outside of the

financial statements. These results suggest that investors search for other sources of information outside financial statements, such as easily accessible financial news networks and Internet blogs.

According to FASBs (Financial Accounting Standards Board) Statement of Financial

Accounting Concepts No. 1: “Financial reporting should provide information that is useful to present and potential investors and creditors and other users in making rational investment, credit, and similar decisions…” (FASB 1978). Although studies examining the relationship between accounting information and security prices generally find that regulated financial reports provide new and relevant information to investors (Kothari 2001), a number of studies have revealed evidence of a decline in the level of relevance of earnings and other financial statement items

(Chang 1999; Lev & Zarowin 1999; Brown et al. 1999). Academics and practitioners alike have expressed concerns about the decreasing value relevance of the financial reporting system (Francis and Schipper 1999). The focus is often on the content of what is reported. Critics suggest that the financial reporting model fails to recognize and measure essential economic assets which produce shareholder value. Lev and Zarowin (1999) find that the relevance of financial information is decreasing and attribute their results to the importance of unreported intangible assets such as innovation and human resources, as well as the inability of the financial reporting model to reflect the changing business environment. Concerns also focus on the issue of when financial information is reported, or more specifically the timeliness of financial reporting, and the extent to which more timely competing information is required to compliment the financial statements

(Francis and Schipper 1999). Investors often look to information intermediaries to provide supplemental information for investment decisions (Healy and Palepu 2001).

1.3.2 Importance of Market Conditions

As Investors look to information intermediaries for guidance as a result of financial statement limitations and complexities, they may particularly rely on these sources of information during uncertain economic periods or unstable market conditions. Academic literature offers support for this implication. The concept of risk aversion (Friedman and Savage 1948) implies that when faced with comparable return alternatives, agents tend to choose those which are less risky. The ArrowPratt measure of risk aversion 2 is an essential construct of asset pricing models.

Studies provide evidence that this measure is counter cyclical; in an economic expansion risk

aversion is low and in a recession risk aversion is high (Campbell 1996; Campbell & Cochrane

1999; Rosenberg & Engle 2002). During an economic downturn, when investors have a higher

risk aversion and wish to avoid risk, they are more likely to rely on information provided by

infomediaries in order to avoid risky investment decisions. Similarly, the principle of Loss

Aversion found in Prospect Theory (Kahneman and Tversky 1979), lends support as to why

investors would rely more heavily on infomediaries during uncertain economic periods. Loss

aversion refers to the notion that losses and disadvantages have a greater impact on preferences

than gains and advantages (Tversky & Kahneman 1991). Investors may be more concerned with potential losses they could sustain in an economic downturn and search more ardently for

information to help them avoid such losses. Kaplanski (2004) suggests that investors experience

greater uncertainty in bearish markets and tend to overestimate downside risk leading to investors’

reliance on infomediaries for investment decisionmaking guidance.

2 Pratt (1964) and Arrow (1964) separately developed the constructs of absolute and relative risk aversion providing a method to measure the degree of risk aversion of an agent.

1.3.3 Sophisticated vs. Unsophisticated Investors

Unsophisticated investors, or investors with less time, resources, and experience than sophisticated investors, may rely on other sources of information for investment decisions, not only because of untimely published financial statements, but also as a result of the growing complexity of financial reporting. Sutton et al. (2009) find that unsophisticated investors approach the task of predicting future firm performance with a much smaller set of information than sophisticated investors. This places unsophisticated investors at a distinct disadvantage. In addition, financial statements and specifically footnote information may be too complex for the unsophisticated investor to interpret. Unsophisticated investors may often seek other sources of information. According to Joe et al. (2007), unsophisticated investors appear to react negatively to media exposure of board ineffectiveness while sophisticated investors act as if they anticipate the effect of this media exposure. This suggests that unsophisticated investors do not consider the same information set as sophisticated investors. Barber and Odean (2008) discover that individual

(unsophisticated) investors are more likely to purchase stocks that capture their attention in the news, suggesting that it is difficult for unsophisticated investors to evaluate the worth of every stock. Alternatively, sophisticated investors are armed with more resources, time, and years of experience, and can continuously monitor a wide range of stocks. Unsophisticated investors appear to require information beyond the financial statements.

1.3.4 A Changing Capital Market

Remarkable growth has taken place in the availability of investment news and information outside the financial statements. From the growth of investment programs such as those on CNBC

to the plethora of financial blogs found on the Internet, investors have an abundance of information at their disposal. We have witnessed rapid progress in technological innovation leading to profound changes in capital markets (Healy & Palepu 2001). These technological innovations have created new channels for investor communication and have resulted in capital markets operating at an accelerated pace. The Internet allows investors to easily obtain information and provides a channel for firms to communicate rapidly with investors and financial intermediaries. How is the investment information portrayed by these various investment sources and how do investors react to these information intermediaries? Although many investment news sources aim to educate investors with objective analyses, much of this information seems to be geared toward swaying investors in favor of a commentator’s or blogger’s personal opinion rather than educating investors.

1.4 Research Objectives

There are a number of anecdotal cases that imply that information activism affects financial markets. There are also a number of studies which provide evidence that information activism affects capital markets and investor behavior (Engelberg et al. 2009; Neumann & Kenny

2007; Fotak 2008). This study builds upon and expands the empirical evidence of this effect, examines the differential effects of two primary sources of information activism, and identifies the specific downstream effects of information activism on capital markets and investor behavior. In order to gain insight into how information activism affects capital markets, factors related to investor behavior, such as trading volume and price reaction, are analyzed on the days surrounding information activism events.

Investor behavior has been studied by examining the associated levels of trading volume and price responses (Beaver 1968; Bamber 1986; Bae & Jo 1999). Trading volume, or the number of shares of stock traded during a given time period, has been examined frequently in prior research to assess investor reaction to information releases (Beaver 1968; Bamber 1986; Bae & Jo

1999). An increase in trading volume generally signifies interest in a stock. Trading volume is often used to determine whether an event has “information content” (Bae & Ho 1999). Stock price responses to various events and information releases, has also been extensively examined in prior research. Research suggests that if an event has informational content, then stock prices will respond accordingly (Karpoff 1987). The effect of information activism on trading volume and price response is investigated in this dissertation.

1.4.1 Modus Operandi

Two sources of information activism are examined in order to detect the effect on investor behavior: 1) CNBC Investment News program, Jim Cramer’s Mad Money ; and 2) Financial

Internet Blog, Seeking Alpha , a financial blog aggregator which contains the largest collection of

financial blogs in the world (McIntyre & Allen 2009). Different attributes of information activism

within each source are also considered in this research. The number of instances of information

activism for a given firm during the period of analysis has been summarized. The direction of the

information activism is also evaluated. For example, is the information activism bullish (positive)

or bearish (negative)? Finally, important moderating factors are considered. For example, how

does the reliance on information activism by investors vary under different market conditions? As

described earlier, it is expected that investors will rely more heavily on information activism

during unstable economic conditions. Other important moderating variables which are expected to affect investor’s reliance on information activism are earnings quality, investor sophistication, and degree of information asymmetry. It is expected that increased levels of information asymmetry and low investor sophistication will lead investors to rely more on information activism. Similarly, when earnings quality is low, investors will likely search for alternate sources of investment decisionmaking information and will tend to rely more on information activism.

1.5 Contributions

This dissertation introduces the term information activism and relates it to understanding investor behavior in capital markets. This study examines information activism from two primary information sources in order to delineate the effects on capital markets. This effect is explored by analyzing the behavior of investors using a number of measurement methods to understand the impact of information activism on capital markets. While prior literature focuses on a narrow set of information sources such as a few broadcasts on CNBC (Busse & Green 2002; Engelberg et al.

2009; Neumann & Kenny 2007) or stocks receiving attention in the news (Barber Odean 2007), this study examines two primary sources of information activism and explores the relationship between each type and the effect on capital markets. Related prior literature studies a limited set of outcomes (Busse & Green 2002; Barber & Odean 2007), while this research explores the downstream effects of information activism by discovering the effect of multiple sources of information activism on investor behavior in order to fully appreciate the consequences in capital markets.

Furthermore, this research introduces important moderating effects related to information activism on investor behavior, which have not been considered in prior studies. One important contribution is the examination of the impact of information intermediaries during divergent market conditions (i.e. bull and bear markets), which has been examined little in prior research.

This study predicts that investors will rely more on information activism during bearish market periods due to risk and loss aversion and examines this reaction during a significant downturn in the economy. This research also builds upon literature regarding investor sophistication and predicts that individual or unsophisticated investors who seem to have a considerable demand for information outside the financial statements will react more strongly to various sources of information activism than sophisticated investors. Similarly, this research provides further evidence regarding the relevance of financial statements and whether investors rely on alternate sources of information in making investment decisions.

This dissertation contributes to the debate regarding the importance of media and investment news in the role of swaying investor behavior and/or affecting capital markets (e.g.

“Cramer Effect”). There is anecdotal evidence that these investment news outlets have an impact on financial markets as well as several prior research studies that offer evidence concerning this effect. This study provides further evidence of a market reaction. The results can also shed light on the demand, necessity, or importance of supplemental information (e.g. to financial statements) in shaping investor decisionmaking.

Additionally, research addresses questions related to the rapid progress in technological innovation which have lead to profound changes in capital markets (Healy and Palepu 2001).

Technology has created new channels for investor communication and has accelerated the pace at which capital markets operate. The Internet makes it easier for investors to obtain financial and investment information and allows firms to communicate rapidly with investors and financial intermediaries. Therefore, it is important to examine these new methods by which investors obtain information and the effect it has on their decisions. Finally, this study contributes to ongoing research in the areas of investor behavior and information asymmetry.

CHAPTER 2

RELATED PRIOR LITERATURE & THEORETICAL FOUNDATION

This study draws upon a number of relevant research areas. Prior work in information asymmetry and signal theory lays the foundation and theory as to why there is a demand for information intermediaries in the area of financial investment information. Significant research on shareholder activism provides a setting for activism research in accounting as well as methods and measures employed. Additionally, research on media coverage describes factors examined which are associated with information releases relevant to this research. Pertinent research on investor behavior provides insight into measurements, methods, and results to further guide this study.

Theoretical and empirical research related to loss and risk aversion provides a basis as to why

investors may rely more heavily on information activism during unstable economic periods.

Finally, research related to information intermediaries and stock recommendations is considered

relevant to this study.

2.1 Signal Theory & Information Asymmetry

Signal theory was originally introduced in the context of a job search (Spence 1973)

whereby prospective employees with college degrees provide a signal to employers regarding their

skills. Information activism provides signals about a firm’s or an industry’s performance in

capital markets. Additionally, substantial research conducted in the accounting field is framed in

information asymmetry since accounting involves the transmission of an enterprise’s information

from those who have it to those who need it for decisionmaking. These signals are assumed to

reduce information asymmetry in capital markets which comes about when investors are differentially informed about a firm’s value; investors with superior information can trade profitably at the expense of less informed investors. To compensate for expected losses, uninformed investors demand a return premium that increases in the risk of trading with privately informed investors (O’Hara 2003).

Prior research suggests a negative association between disclosure quality and information asymmetry (Healy and Palepu 2001; Heflin et al. 2005; Welker 1995). Also, Brown et al. (2004) study corporate conference calls and find that voluntary disclosures lead to longterm reductions in information asymmetry. Although disclosures appear to reduce the level of information asymmetry, a number of studies question the usefulness of financial statements (Lev & Zarowin

1999) and the value relevance of the financial reporting system (Francis & Schipper 1999).

Therefore, investors often rely on information outside of the financial statements (Sutton et al.

2009) and may look toward information intermediaries such as information activists to provide supplemental information.

2.2 Shareholder Activism

Prior literature in accounting and finance has primarily examines activism in the context of shareholder activism whereby shareholders target firms with proposals related to various corporate decisions. These corporate decisions include compensationrelated issues (Ertimur et al. 2008), corporate governance preferences (Del Guercio et al. 2008), and accounting choices (Ferri and

Sandino 2009). These studies examine the effect of shareholder activism on a number of outcomes. The literature explores whether these proposals have been successful in pressuring

firms to take actions desired by shareholders, or whether these proposals are associated with other measures of “success” including accounting measures, operating performance, and firm value

(Karpoff 2001). These activists are often successful when they leverage publicity about the issue and engage in networking activities (Logsdon & Van Buren III 2009). Additionally, activists engage in ongoing communication with corporations to deal with social and governance issues

(Logsdon & Van Buren III 2009).

The effect of shareholder activism on share value (e.g. abnormal returns) is examined in a number of studies (e.g. Gillian and Starks 2000; Prevost & Ramesh 2000). Other studies examine whether improvements in accounting performance measures (e.g. ROA, ROS) are associated with shareholder activism (e.g. Prevost & Ramesh 2000; Wahal 1996). Further studies investigate changes in the targeted firm’s operations (e.g. sales growth, capital expenditures) (e.g. Del Guercia

& Hawkins 1999) and actions of firms targeted by shareholder activism (e.g. Del Guercia and

Hawkins 1998; Martin et al. 2000). The effect on governance activities such as CEO appointments

(Del Guerico et al. 2008) and executive compensation decisions (Ertimur et al. 2008) are also investigated. More recently Ferri and Sandino (2009) examine the effect of shareholder activism on accounting choices and investigate whether or not shareholder proposals persuade firms to expense employee stock options. In any event, shareholder activism typically involves shareholder persuasion of corporate management to take certain actions or adopt policies which are desired by the shareholders.

Activism, in the context of this research setting, is conducted by those who provide investment information. Whether this information is provided on investment news programs or

financial blogs, it is often presented in a manner designed to persuade investors to take specific action with respect to an investment in a firm’s stock. For example, on Mad Money , Jim Cramer

may tell viewers to buy shares of Ford because he really likes the vision of Ford’s CEO. A blogger may write an article to encourage investors to sell their shares of a given stock because he

envisions a downturn in the market. This advice may or may not involve any specific fundamental

financial analysis of said firm, but often represents the opinion of the host or author as to the

direction investors should take regarding an investment. This is a form of activism because there is

an attempt to persuade the investor to take certain actions with respect to their investing decisions.

2.3 Media Coverage

Research conducted in the area of media coverage can provide insight about the effect of

information activism on investor behavior. Joe et al. (2007) examine the effect of media exposure

of board ineffectiveness on corporate governance, investor trading behavior, and security prices.

Results suggest that media exposure of weak boards of directors encourages the targeted firms to

take corrective actions to strengthen their governing boards. Furthermore, while unsophisticated

investors appear to react negatively to those firms with perceived ineffective boards, sophisticated

investors seem to anticipate the corrective action of the target firms. These results suggest that particularly in the case of unsophisticated investors, information activism may affect investor behavior.

Barber and Odean (2008) test and find support for their proposition that individual

(unsophisticated) investors are more likely to buy stocks that capture their attention in the news.

Furthermore, this study suggests that sophisticated investors are less likely to purchase attention

grabbing stocks due to their ability to closely monitor a wider range of stocks as a result of more time and resources. Barber and Odean (2008) examine three proxies for attention grabbing events: news, unusual trading volume, and extreme returns. Joe (2003) conducts an experimental study to explore why auditors are more likely to issue goingconcern opinions when the client has received negative press coverage prior to issuing the audit opinion. Her results suggest that auditors tend to overreact to redundant information because the press coverage does not reveal any new information. Furthermore, prior literature in this area has largely examined media press coverage (Barber & Odean 2007; Joe 2003; Joe et al. 2007) or a small number of broadcasts

(Busse & Green 2002) and has ignored many other available sources in the current information age. This dissertation aims to examine a broader set of information sources and the effect on capital markets.

2.4 Investor Behavior

A number of studies examine investor behavior. Barber and Odean (2008) examine how investors react to stocks which receive attention in the news. They propose that attention is a major factor in determining which stocks individual investors buy and their results provide strong evidence that individual investors are more likely to buy attentiongrabbing stocks. The authors argue that attention is a scarce resource for unsophisticated investors because individual investors have a limited amount of time to spend on searches for stocks to purchase. Theoretical guidance is taken from the notion of bounded rationality which suggests that we have cognitive and temporal limits to how much information we can process. In order to make stock searches more manageable choice sets must be narrowed. On the other hand, sophisticated or institutional

investors devote more time to searching for stocks, and consequently attention is not a scarce resource for this group of investors. This suggests that information activism may have more of an effect on the behavior of unsophisticated investors than sophisticated investors.

Barber and Odean (2008) measure investor behavior by examining net buying reaction

(buy/sell tradeorder imbalances) and provide a proxy for attentiongrabbing stocks by including stocks which are mentioned in the news, experience high abnormal trading volume, or offer extreme oneday returns. Relevance or reach of attention events is gauged by examining the effects on trading volume and returns. Important news about a firm often results in significant positive or negative returns. The authors find that individual or unsophisticated investors are net buyers on highvolume days, following days of extreme positive and negative oneday returns, and when stocks are in the news. However, institutional or sophisticated investors do not exhibit buying behavior associated with attentiongrabbing stocks.

Grinblatt and Keloharju (2000) classify investors as either momentum investors who tend to buy past winners and sell past losers, or contrarian investors. The authors analyze a unique comprehensive data set from the central register of shareholdings for Finnish stocks in the Finnish

Central Securities Depository (FCSD). The primary focus of their study is to discover which investor group exhibits momentum versus contrarian behavior. The results indicate that behavioral patterns with respect to past returns are consistent across various investor classifications. For example, investors who tend to exhibit momentum (contrarian) behavior seem to do so for a wide variety of past return horizons. The strong behavior patterns observed by the researchers suggest that investor behavior is common to a large proportion of investors in the category, rather than as a

result of a statistical anomaly. The authors measure the relationship between investor behavior and past returns by examining whether the buy ratio of past winning stocks exceeds the buy ratio of past losing stocks. Specifically, the more sophisticated the domestic investor and the greater the stock investment, the less contrarian the investment strategy. Domestic institutional investors, with a sophistication level and size between those of foreign investors and household investors in

Finland, follow an investment strategy midway between the strategy pursued by foreign investors and extreme contrarian domestic household investors.

Busse and Green (2002) investigate the price and trading volume effect of analysts’ reports of individual stocks appearing on segments of the Morning Call and Midday Call programs on

CNBC. The authors find that prices respond within minutes of the stock being mentioned on these programs. Their evidence suggests that positive analyst reports are reflected in prices beginning seconds after a stock is mentioned and are fully incorporated into prices within one minute, while the response to negative analyst reports is larger but more gradual lasting around fifteen minutes.

Interestingly, results provide less evidence of a price response to positive reports during the

Morning Call segment which may suggest that this information is not relevant or may have been previously revealed to the market. Results indicate that trading profitability is highest within the first five seconds of the analyst report and steadily declines and is no longer evident beyond 15 minutes.

The study by Busse and Green (2002) provides evidence of a semistrong efficient market whereby active traders cannot generate profits as a result of news which is widely publicized unless they act immediately. Furthermore, a significant increase in trading intensity is observed

following the broadcast of positive analyst reports, increasing from an average of 49 to 286 trades and 84 to 377 trades for the Morning and Midday Call programs respectively. In order to investigate whether increases in trading activity is attributable to viewers trading on the information broadcast in these segments, the authors note that more buyerinitiated trades should follow positive reports and sellerinitiated trades should follow negative reports. Therefore, the researchers examine order imbalances in the minutes surrounding the report and find that order imbalances shift with the sentiment of the reports and are significant in all scenarios.

2.5 Information Intermediaries and Online Stock Recommendations

Recent work examines the role of information intermediaries in providing information to capital market participants that is new, useful, and beyond firm disclosures. Bushee et al. (2010) evaluate whether press coverage affects the degree of information asymmetry around earnings announcements. They conclude that press coverage has a significant effect on firms’ information asymmetry around earnings announcements through reductions in bidask spreads and improvements in depth. Their study implies that the wide distribution of information by the press appears to have a bigger impact on information asymmetry than the quantity or quality of information provided. Additionally, Pownall and Simko (2005) examine the information intermediary role of short sellers when alternative information sources are lacking (i.e. low analyst coverage). They find a significant, negative market response to spikes in short sales for firms covered by no more than one analyst. Their results suggest that investors view short sellers as information intermediaries when the information environment is limited.

Other recent work focuses on investors’ reactions to stock recommendations on Internet blogs and examines market reaction to stock email spam and stock message boards. Fotak (2008) examines the effect of 500 buy and sell stock recommendations from the SeekingAlpha Internet financial blog on abnormal returns and trading volume in the days surrounding the blog recommendation posting. Results indicate that the market reacts more strongly to short recommendations, that blog postings tend to include liquid stocks issued by large firms, and that characteristics of the blogger such as education and perceived skill appear to be associated with the reaction. Furthermore, blogs appear to offer genuine information rather than attempting to carry out schemes such as “pumpanddump,” as displayed by the absence of price reversals in the twenty days following the blog posting.

Additional research related to the online environment involves investor reaction to stock message board postings and stock email spam, and generally finds evidence of a relationship.

However, the direction of this association seems to be inconsistent. Tumarkin and Whitelaw

(2001) examine stock message postings on the Raging Bull discussion forum and find that abnormally high message activity is associated with abnormal returns and increased trading volume. However, the message board activity does not seem to predict industry adjusted returns or abnormal trading volumes. In a study by Dewally (2003), message board stock recommendations are analyzed and results indicate that stocks appear to be recommended following sharp increases in price, although there appears to be no market reaction to these recommendations. In contrast,

Antweiler and Frank (2004) examine stock message board postings on both Yahoo! and Raging

Bull . They find that the postings have a significant predictive content related to trading volume

and price volatility, and between message level of bullishness and trading volume. Recent research

examining market reaction to stock email spam provides evidence of a strong reaction for factors such as price, volume, volatility, and intraday spread in the days surrounding the stock spam (e.g.

Böhme & Holtz 2006; Frieder and Zittrain 2006; Hanke & Hauser 2008; Nelson et al. 2009).

However, shortterm reversals are observed in some results.

2.6 Risk & Loss Aversion

The theories on risk aversion and loss aversion are relevant to this research as these theories support investor rationale in seeking supplementary information, particularly during unstable economic conditions. Risk aversion, originally put forward by Friedman and Savage

(1948), suggests that agents tend to choose less risky alternatives when faced with returns which are comparable. In 1964, Pratt and Arrow separately developed a measure of the degree of risk aversion of an agent for absolute and relative risk aversion. This measure is generally an essential component of most asset pricing models and has been employed in a wide range of research on investor preferences. A number of studies suggest that the ArrowPratt relative risk aversion is counter cyclical (Campbell 1996; Campbell & Cochrane 1999; Rosenberg & Engle 2002). This implies that risk aversion is high in recessionary periods and low in expansionary periods.

Therefore, during an economic downturn or recession, investors with high risk aversion are more likely to rely on information activism in an effort to avoid risk related to investment decisions.

Prospect theory offers the principle of loss aversion which was first introduced by

Kahneman and Tversky (1979). Under the principle of loss aversion, losses and disadvantages affect preferences more strongly than gains and advantages (Tversky & Kahneman 1991).

Therefore, investors would be highly concerned with impending losses they may experience

during an economic downturn or recession which would cause them to seek out information, and therefore be more likely to consider information activism during unstable economic conditions.

Similarly, Kaplanski (2004) proposes that investors are faced with greater uncertainty in bearish market conditions causing them to overestimate downside risk. This supposition may also lead investors to rely more on information activism during bearish markets.

2.7 Literature Synopsis

A variety of literature streams are relevant to this study. These research areas provide theoretical guidance and a basis for understanding the expected market reaction to information activism as well as measurement tools with which to gauge this reaction. Signal theory and the concept of information asymmetry offer theoretical support as to why investors may often look toward infomediaries for financial direction and why investors react to information activism.

Research on shareholder activism presents a basis for establishing the notion of information activism where investors are prompted to take particular action and suggests measures by which this reaction can be examined. Studies in the area of media coverage and investor behavior also indicate measures often used to discover how capital markets react to information releases. Recent work in the area of information intermediaries and online stock recommendations add to knowledge concerning investor behavior and market reactions. Finally, the theories of risk and loss aversion provide a strong case for the conditions under which investors will most likely rely on information activism.

CHAPTER 3

HYPOTHESES DEVELOPMENT & RESEARCH MODEL

The basic model examined in this research involves the effect of information activism on capital markets by way of investor behavior; it appears in Figure 1. This study argues that if information activism is relevant, then various measures useful in gauging investor behavior should respond accordingly. A variety of factors which have been studied in related literature regarding investor behavior are analyzed (section 3.1). However, these relationships do not exist in a vacuum and other contingent factors may enhance or diminish these associations (section 3.2).

The specific hypotheses tested in this study are developed based on supporting literature which provides the reasoning for each hypothesis. The final research model appears Figure 2 and is described in section 3.3.

3.1 Investor Behavior

Information asymmetry is a key issue in capital markets where managers and other insiders have privileged information about the true value of the firm, while many investors do not have all relevant information necessary to make optimal investment decisions. Therefore, less informed investors demand compensation for bearing the risk of expected losses associated with trading amid privately informed investors (O’Hara 2003). A problem associated with information asymmetry is adverse selection (Akerlof 1970) which occurs when investors purchase inferior investments. Managers and insiders who have relevant information about the future performance of the firm exploit this information at the expense of less informed investors by managing or

biasing the information released to investors. This makes it more difficult for investors to carry

out effective investment decision making and results in adverse selection. A solution to the

adverse selection problem is signaling (Akerlof 1970), originally proposed by Spence (1973).

Signaling allows information to be transferred to the less informed party in a situation where there

is information asymmetry. Signals are often sent by information intermediaries (Fombrun &

Shanley 1990) such as broadcast media and publications, as well as the ever growing abundance of

information provided by the Internet. While these information sources are generally seen as

reducing information asymmetry, they can also lead to an over abundance of information making

it difficult for investors to identify relevant information for decision making. Investors look for

unique sources of information to bridge the gap between too little and too much information.

Investors need to be able to identify which information is most relevant so that they may make the best investment decisions.

Often investors look toward sources of information activism in an effort to guide them

toward the relevant investment information. Investors watch investment news broadcasts on

CNBC as evidenced by the growing viewership (Hempel 2008). Also, according to a Nielson

report “Global Faces and Networked Places,” twothirds of the world’s Internet population visit

social networking or blogging sites. This accounts for almost 10% of all Internet time which is

growing at more than three times the rate of overall Internet growth. Business is among the fields

most impacted by the blogosphere and this trend is expected to continue in the future (Sussman

2009). Consequently, investors are increasingly watching and ‘clicking’ in an effort to acquire pertinent information regarding investment decisions.

This study examines the effect of information activism on capital markets. Investors are one of the fundamental actors in capital markets and therefore the effect of information activism on investor behavior is analyzed. This study explores whether investors react to information activism. As described earlier, information activism prompts investors to act in a particular manner (e.g. buy a stock) rather than solely informing investors in an objective manner. Thus, the effect of information activism on investor behavior in capital markets is examined to determine the extent to which investors react to information activism. If information activism has information content, then investors’ reactions will be associated with instances of information activism. A number of measures consistent with prior research are used as a proxy for investor reaction and therefore, individual hypotheses are developed for each of these measures. However, the overall premise suggests the main hypothesis as follows:

H1 (premise): Information activism influences investor behavior.

Subhypotheses are detailed below regarding the specific variables employed to evaluate investor behavior.

3.1.1 Price Reaction

The early research of Fama (1970) characterized an efficient capital market as one in

which security prices reflect all available information. The response of stock prices to various

events and information releases has been extensively examined in prior research. While trading

volume reflects investors’ activity by providing a sum of all trades, security prices reveal the

average of investors’ beliefs (Bamber 1986). Price is seen as reflecting changes in expectations of

the market as a whole whereas trading volume is seen as reflecting changes in expectations of individual investors (Bae & Ho 1999). An examination of the degree of volume reaction in

conjunction with degree of price change leads to better inferences about the information process.

Therefore, in this study of the effect of information activism on capital markets through an

examination of investor reactions by way of changes in stock price associated with information

activism, provides important evidence. A number of studies suggest that if an event has

informational content, then stock prices will respond accordingly (Karpoff 1987).

An appropriate method to assess price response to an information release is by examining

abnormal return. Abnormal return is the return generated over a time period that is different from

the expected return. If information activism is important and considered by market participants,

then it follows that instances of information activism are associated with abnormal returns. The

following hypothesis is proposed:

H1a: Information activism is associated with abnormal returns.

3.1.2 Trading Volume

Trading volume or the number of shares of stock traded during a given time period, has been examined often in prior research for purposes of assessing investor’s reaction to an

information release (Beaver 1968; Bamber 1986; Bae & Jo 1999). High trading volume generally

indicates more interest in a stock. Trading volume is often used to determine whether an event has

“information content” (Bae & Ho 1999). The role of information and its influence on the trading behavior of investors is the focus of these studies. Trading volume appears to be the most

observable indicator that financial accounting information affects investor behavior (Cready &

Hurtt 2002). Trading volume reflects investors’ activity by providing a total of all market trades

(Bamber 1986) or as described by Beaver (1986), trading volume reflects investors’ idiosyncratic reactions. Trading volume has been found to be associated with information releases (Karpoff

1986; Bamber et al. 1997). If an information release is relevant and has information content, then it should be associated with a reaction to that information release. Information activism represents a form of information release to investors and trading volume is an important indicator of investors’ reaction to information activism.

Significant spikes in trading volume often indicate that some kind of important news is taking place. Trading volume has primarily been employed to assess the information content of annual earnings announcements (Beaver 1968; Morse 1980, 1981; Bamber 1986; Lee 1992).

Hirshlefifer et al. (2003) studied investor actions following positive and negative earnings surprises. More recent studies look at other forms of information releases and the effect on trading volume. For example, Antweiler and Frank (2004) study the informativeness of Internet Stock message boards and assess investor reaction by examining trading volume. In addition, Barber and Odean (2008) examine investor reaction to attentiongrabbing stocks, or stocks in the news, by observing trading volume. Busse and Green (2002) also assess investors’ reaction to stocks mentioned in analyst reports on CNBC by analyzing trading volume. The results of these studies suggest that trading volume is an appropriate method to gauge investor reaction to information releases. If investors react to information releases such as those associated with information activism then an observed change in trading volume should occur. Therefore, the following hypothesis is proposed:

H1b: Information activism is associated with abnormally high trading volume.

3.1.3 Sentiment of Information Activism

Since information activism is characterized by information that supports or opposes a particular stock investment intended to sway investors in favor of or against an investment, an essential component of this study is to evaluate investors’ reactions to the inclination of the information activism. Busse and Green (2002) assess whether investors react to the sentiment of analyst reports on CNBC by examining trade order imbalances. Their results suggest that buyer initiated trades follow positive reports on the broadcast while the opposite is true for negative reports. Numerous studies have examined the price reaction to recommendations by analysts and have found that favorable (unfavorable) changes in individual analyst recommendations are associated with positive (negative) prices at the time of the announcement (e.g. Stickel 1995;

Womack 1996). Consequently, if investors react to positive (negative) information activism then an increase (decrease) in price would be observed. Thus, the following hypothesis is proposed:

H1c: Positive information activism is associated with higher prices, while negative

information activism is associated with lower prices.

3.2 Moderating Effects

This research recognizes that a number of contingent factors further describe the primary

relationships described in section 3.1 regarding the effect of information activism on investor behavior. This study will examine how the hypothesized reactions vary according to investor and

firm characteristics, as well as market conditions. Therefore, important moderating variables are considered in this study based upon existing literature related to investor behavior.

3.2.1 Investor Sophistication

A number of prior studies examine investor level of sophistication or type of investor

(individual or institutional) by exploring the size of stock trades. For example, Lee (1992) identifies small traders as those who place orders of less than $10,000 and finds that they are net buyers subsequent to positive and negative earnings surprises. Similarly, Hirshleifer et al. (2008) also provide evidence that individual investors are net buyers following both positive and negative earnings surprises. Barber and Odean (2008) point out that less sophisticated investors have less time and resources to expend on information searches regarding investment decisions. Evidence in their study supports the notion that less sophisticated investors are more likely to react when stocks capture their attention in the media. Furthermore, Joe et al. (2007) find that unsophisticated investors react to negative media regarding board governance effectiveness, while sophisticated investors, often armed with more time and information, appear to anticipate this news.

These studies suggest that less sophisticated investors tend to react more to information events than do sophisticated investors. This may be due to unsophisticated investors’ lack of time, resources, or expertise in assessing available financial information. Consequently, unsophisticated investors rely on other sources of information associated with information activism. If unsophisticated investors are inclined to rely on easily available sources of financial investment information, then they may have a stronger reaction to information activism than do sophisticated investors. Therefore, the following hypotheses are proposed:

H2: The effect of information activism on prices and trading volume will be greater for

unsophisticated investors than for sophisticated investors.

3.2.2 Market Condition

As discussed earlier, the theories of loss aversion and risk aversion suggest that in unstable economic periods, investors will tend to rely more on information activism in order to avoid losses anticipated during these market conditions. Risk aversion suggests that when agents are faced with comparable investment alternatives, agents have a tendency to select those which are less risky

(Friedman and Savage 1948). Research on the ArrowPratt measure of risk aversion indicates that relative risk aversion is counter cyclical. In economic expansionary periods, risk aversion is low and in recessionary periods risk aversion is high (Campbell 1996; Campbell & Cochrane 1999;

Rosenberg & Engle 2002). Consequently, during unstable economic conditions when investors have a high aversion to risk and wish to avoid it, they are more likely to rely on information activism in an attempt to avoid highrisk or unprofitable investments.

Furthermore, the principle of loss aversion from prospect theory (Kahneman & Tversky

1979) strengthens the notion that investors may rely more heavily on information activism during unstable economic periods. According to the principle of loss aversion, losses and disadvantages have a greater influence on an agent’s preferences than gains and advantages (Tversky &

Kahneman 1991). Thus, investors who are concerned with potential losses sustained in an economic downturn will likely search for information to help them avoid these losses and will rely on information activism. Finally, Kaplanski (2004) suggests that investors experience greater uncertainty in bearish markets and tend to overestimate downside risk. As a result, investors will

likely rely on information activism during bearish markets. Therefore, it is expected that the effect

of information activism on investor behavior will be stronger during bearish market periods than

during bullish market conditions. The following hypothesis is proposed:

H3: The effect of information activism on prices and trading volume will be greater

during bearish market conditions than during bullish market conditions.

3.2.3 Information Asymmetry

Information asymmetry in earlier discussions provides a basis as to why investors seek

sources of investment information. Information asymmetry is of particular interest in capital

markets due to the lack of perfect information about the expected return on a given investment

and resulting in the well known “lemons problem” (Akerlof 1970). In capital markets, sellers

(firms) have more information about the quality of a product (security investment) than do buyers (investors), resulting in an increased risk of adverse selection. Though corporate disclosures attempt to reduce information asymmetry in capital markets, investors continue to demand relevant information for decision making. Consequently, for firms with higher levels of information asymmetry, investors are more likely to react to information activism. Therefore, the following hypothesis is proposed:

H4: The effect of information activism on prices and trading volume will be greater for

firms with high information asymmetry than for firms with low information

asymmetry.

3.2.4 Earnings Quality

There is growing concern regarding firms’ quality of earnings, or the extent to which reported earnings reflect actual earnings and are useful for predicting future earnings. Research suggests that poor earnings quality increases the need for investors to acquire alternate information when making investments decisions. Lev and Thiagarajan (1993) and Abarbanell and Bushee

(1997, 1998) find that earnings quality signals, derived from detailed financial statement information, can help predict future earnings. To the extent that earnings quality is low and not useful in predicting future earnings and performance, investors may search for other sources of information to assess the future performance of a firm.

Research suggests that nonearnings financial information such as financial statement disclosures, board independence, and ownership structure (Francis et al. 2006) is required as a result of poor earnings quality. Research suggests that information outside the financial statements is required by investors when a firm exhibits low earnings quality and investors search for alternate methods to predict the future performance of a firm. Therefore, earnings quality may determine whether investors will search for or be swayed by other information related to a firm, such as information activism. It is expected that a firm’s level of earnings quality will moderate the response to information activism by investors. The following hypothesis is proposed:

H5: The effect of information activism on prices and trading volume will be greater for

firms with low earnings quality than for firms with high earnings quality.

3.3 Research Model

Based on the prior hypotheses development and supporting research, a detailed research model is provided in Figure 2. This model indicates that the primary relationship examined in this dissertation is the effect of information activism on investor behavior. Information activism is examined in terms the effect of the event of information activism on investor behavior regarding stock trading volume and price (returns), and is detailed further in Chapter 4. The research model in Figure 2 also indicates that two constructs of information activism are examined; 1) the number of instances of information activism or Intensity ; and 2) the Sentiment or direction (positive or

negative) of information activism. These constructs are detailed in Chapter 4.

Additionally, this research model depicts the four primary moderating effects which are

examined in this dissertation. Based on the hypotheses development, investor sophistication is

expected to negatively affect the relationship between information activism and investor behavior

as more sophisticated investors are expected to rely less on information activism. Market

condition is also a moderating effect examined in this dissertation. A bearish market is expected

to strengthen the relationship between information activism and investor behavior because during

a down market, investors looking to avoid losses are more likely to rely on information activism.

Information asymmetry is also expected to moderate the effect of information activism on investor behavior. High information asymmetry is expected to strengthen the relationship between

information activism and investor behavior as investors search for more information to make

investing decisions. Finally, earnings quality is also expected to moderate the relationship between information activism and investor behavior. As earnings quality increases the effect of

information activism on investor behavior is expected to be smaller because high earnings quality will result in investors relying less upon information activism.

CHAPTER 4

RESEARCH DESIGN

4.1 Sample Time Periods and Sample Selection

4.1.1 Sample Time Periods

As described in Chapter 3, market condition is expected to be an important factor related

to the effect of information activism on investor behavior. In order to examine the varying

effects of information activism on investor behavior during divergent market conditions, two

separate bull and bear market periods are identified. A market period is considered a bull market

when share prices are generally rising, while a bear market is when share prices are falling quite

sharply and are expected to fall further. In order to classify a month as a bull or bear month, the

market return in that month is compared with the median market return over the entire period.

The month is classified as a bull (bear) month if the monthly market return is higher (lower) than

the median market return. This classification method has been employed in a number of other

studies where “up and down market” definitions are utilized (e.g. Fabozzi & Francis 1977, 1979;

Bhardwaj & Brooks 1993).

According to the Dow Jones Industrial Average (DJI), the financial meltdown began with a

sharp decline in the last week of September 2008 and the recovery began around the beginning of

March 2009 (Figure 3). This time period is particularly ripe for the study of information activism

as it provides a volatile economic market period in which to measure the effect of information

activism on capital markets. This time period is unique and therefore the results of this study can

provide insight regarding the importance of information activism during critical economic periods.

Three consecutive bear months during the economic meltdown and three consecutive bull months

following the economic meltdown (recovery) are selected for data collection. Figure 4 summarizes

monthly returns for the period leading up to, during, and following the financial meltdown. Each

month selected for inclusion in the bear/bull period is labeled accordingly. An associated mark

appears in the chart included at the bottom of Figure 4 in order to visualize where the months and

the bull/bear periods fall on the DJI during this time period. The threemonth bear sample period

during the economic crisis includes September, October, and November of 2008 as this period

represents three consecutive months where the average monthly return is below the median return

for the entire period. Similarly, a threemonth period following the economic crisis, which

includes March, April, and May 2009, has been identified as the bull sample period because this period contains three consecutive months where the average monthly return is above the median

monthly return for the entire period. These two threemonth bull/bear periods are examined in

order to select a sample of firms which experienced information activism events. This sample

selection is described in the section 4.1.2.

4.1.2 Sample Selection

The sample for this study includes firms which were mentioned on Jim Cramer’s Mad

Money program and on the SeekingAlpha (http://seekingalpha.com ) financial blog. Mad Money

was chosen due to its popularity by CNBC’s viewership as well as the approach taken by its host

who often calls upon investors to take a particular action, for instance buying or selling a

particular stock, rather than solely providing investment analysis for a firm. A recap of the

stocks mentioned on Mad Money can be found at http://www.madmoneyrecap.com/recap archiveindex.shtml . A daily recap is listed by date and each recap includes a brief summary of

the show and then categorizes stocks mentioned (including ticker symbol) as Cramer being bullish or bearish on the stock. Therefore, the date, ticker symbol, and bullish/bearish stance (for

simplicity and consistency this classification is referred to as buy/sell) were recorded for each air

showing date during the two threemonth bull/bear sample periods.

The initial sample is detailed in Table 1 and contains 2,084 event observations from Mad

Money (Cramer) . Of these 2,084 observations which are split almost equally between the two threemonth bull (1,040) and bear (1,044) periods, 62.30 percent (1,299) were buy recommendations, while 37.67 percent (785) were sell recommendations. Table 2 provides the overall sample descriptive statistics for the univariate analysis. Of the 1,040 (1,044) event observations collected from the Cramer broadcasts during the bull (bear) sample period represent

479 (488) unique firms of which 270 (269) firms had one event, while 209 (219) firms had

multiple events. Table 2 also provides the descriptive statistics for the primary variables of

interest, cumulative abnormal return ( CAR ) and abnormal trading volume ( ABVOL ), which is

discussed further in section 4.2.1.

Additionally, for each firm a tally is constructed for the total number of times the firm is

mentioned during the sample period ( PRG#). This variable is included in the multivariate

analysis (section 4.3.5) and is a measure of the frequency of information activism and is referred

to as Intensity . According to Table 10 which provides the descriptive statistics for the

multivariate analysis, the mean number of times a firm was mentioned on Cramer (PRG#) is

1.61. A measure is also constructed for the ratio of buy recommendations to total number of mentions during the sample period for each firm ( PRG%B). This measure provides the percentage of buy recommendations for each firm during the two sample periods and represents

the Sentiment of information activism. Table 10 shows that the mean percentage of buy

recommendations to total mentions for those firms mentioned on Cramer (PRG%B) is 37.13 percent. These variables are analyzed as the primary independent variables of interest in the

multivariate analysis (4.3.5).

SeekingAlpha (www.SeekingAlpha.com ) is a blog aggregator for financial blogs and provides links to more than 200 financial blogs which offer investment advice. SeekingAlpha is included as one of the “Best 25 Financial Blogs” ( www.time.com ) which examines over 100 financial blog websites and tracks posts for several weeks. Blogs are evaluated for wellwritten content as well as frequency of postings and readership. SeekingAlpha was chosen as a data source due to the large number of blogs links provided as well as its popularity and high standards.

In order to collect information activism observations from SeekingAlpha, search terms including “Bearish,” “Bullish,” “Buy,” and ”Sell” were used to identify stocks recommended on the blog and to remain consistent with firms mentioned on Mad Money . The data captured includes firm name, ticker, whether the blogger was bearish or bullish on the stock or provided a buy or sell recommendation. The characteristics of the 579 event observations obtained from the

SeekingAlpha blog (Blog ) is offered in Table 1. Of this total 39.55 percent (229) and 60.45 percent (350) occurred in the bull and bear market periods respectively, while 78.24 percent

(453) and 21.76 percent (126) represent buy and sell recommendations respectively. According to Table 2 the 229 (350) observations collected from the SeekingAlpha Blog during the bull

(bear) sample period represent 195 (289) unique firms of which 165 (248) firms had one event, while 30 (41) firms had multiple events.

Similar to the data captured for Mad Money , a tally for each firm was constructed for the total number of times the firm was mentioned on SeekingAlpha during each sample period

(BLG#). This Intensity variable is included in the multivariate analysis (see section 4.3.5) as a measurement of the frequency of information activism for a given firm. According to Table 10 which provides the descriptive statistics for the multivariate analysis, the mean number of times a firm was mentioned on SeekingAlpha (BLG #) was 0.45. The ratio of the number of buy recommendations to total number of mentions during the sample period for each firm ( BLG%B) was also calculated . This percentage is analyzed as a measure of Sentiment of information activism in the multivariate analysis (4.3.5). The mean percentage of buy recommendations to total recommendations for firms mentioned on SeekingAlpha (BLG%B) was 28.82 percent as provided by Table 10.

4.1.3 Other Data Sources

Daily closing prices adjusted for splits and dividend payments, equalweighted market index (NYSE, Amex, NASDAQ), daily trading volume data, outstanding shares, bid and ask price, and market capitalization are obtained from the Center for Research in Security Prices (CRSP) database. These data items are described in more detail in section 4.3. Various financial data items described further in section 4.3 as well as earnings release and SEC filing dates needed to remove

confounding event observations (see 4.2.2) are obtained from Standard & Poor’s Research Insight

Compustat . In order to eliminate confounding event observations for significant news around the information activism events (see 4.2.2) in the form of news articles were identified by searching for news on the days surrounding the information activism event using the LexisNexis Academic

database.

4.2 Sample Characteristics & Confounding Events

4.2.1 Sample Characteristics

Table 2 provides additional descriptive statistics for the event observations collected from

the two data sources during the bull and bear market periods. Descriptive information regarding

cumulative abnormal return ( CAR ) (defined in section 4.3.1) and abnormal trading volume

(ABVOL ) (defined in section 4.3.2), as well as market capitalization, is presented. Regarding the

Cramer (Blog) sample, the mean CAR is 0.0010 (0.0040) and 0.0121 (0.0103) during the bull and bear periods respectively. The mean ABVOL for the Cramer (Blog) sample is 19.34 (109.39) and 21.83 (57.29) during the bull and bear periods respectively. Mean market capitalization for the Cramer (Blog) sample is $23,654,742.57 ($15,902,629.33) and $23,883,325.34

($28,808,145.56) during the bull and bear periods respectively

4.2.2 Confounding Events

In order to remove confounding events which could distort the results, three types of events were considered for each of the 2,663 event observations: 1) SEC filings (e.g. 8K, 10K, etc.) around the event date; 2) Earnings announcements around the event date; and 3) Significant

news around the event date. Using Standard & Poor’s Research Insight Compustat, SEC filing dates were obtained for each firm in each sample. These dates were examined and compared to the event dates for each firm in each sample. If the SEC filing occurred around the event date meaning the day before the event, the day of event, or the day after the event (1, 0, +1), then that event was eliminated from the sample. The process of removing observations with confounding events is depicted in Table 3. Similarly, earnings announcement dates were generated from

Standard & Poor’s Research Insight Compustat for firms in each sample and a comparison of these dates with the event dates was conducted. If the earnings announcement date was around the event date (1, 0, +1) then the event observation was removed from the sample. Finally a news search was conducted using the LexisNexis Academic database for each firm in the sample and significant news around the event dates was identified. News was deemed significant if it represented significant changes in performance (i.e. “Profits Drop 30%.”), major events such as mergers, layoffs, substantial litigation, share offerings, etc., or significant product introductions

(e.g. Apple ipad). Events which had significant news around the event (1, 0, +1) were eliminated from the sample. Overall the search for confounding events yielded 319 confounding event observations in the Cramer sample and 79 confounding event observations in the Blog sample.

After removing all confounding event observations, a clean sample size of 1,765 and 500 from the

Cramer and Blog samples respectively, was obtained. An additional 102 (56) event observations from the Cramer (Blog) sample were dropped due to missing data values. The resulted in a final sample size of 2,107 event observations of which 1,663 were for the Cramer sample and 444 were for the Blog sample, for purposes of the univariate analysis (4.3.1).

4.3 Methodological Framework

4.3.1 Cumulative Abnormal Return Univariate Analysis

The methodological framework employed for the univariate analysis is the event study as described by Brown and Warner (1980, 1985). An event study examines the reaction of variables such as a firm’s stock price and trading volume around a firm specific event. In this study the event date is the date of the broadcast airing or the date of the blog posting and is established as

“day 0.” Since the Mad Money broadcasts air after market closing and blog posts may also occur after the close of the market or might not be read immediately, the market reaction is expected to occur on “day +1.” Therefore a twoday event window of day 0 and day +1 is analyzed.

Additionally, a 95day estimation period ending 20 days prior to the event day is utilized for purposes of estimating such parameters as normal or expected returns and trading volumes. In addition to the shortwindow measure examined in the univariate testing, a longwindow measure is constructed for each of the threemonth bull/bear market periods which will be analyzed as the dependent variable in the multivariate analysis described in section 4.3.5.

In order to test H1a as described in section 3.1.1 which suggests that information activism is associated with abnormal returns; the following methodology (Brown & Warner 1980, 1985) is utilized:

For each stock i in the sample, the return R for time period t, where t = day 0 or day +1, is

calculated as:

Rit = rit + eit (1)

Where:

rit = the normal, expected, or predicted return based a particular model of expected returns

eit = the difference between the observed return and the predicted return (also referred to as

abnormal return AR it ):

eit = AR it = Rit rit (2)

Here eit can be described as the difference between the return which is conditional on the event or

information activism and the expected return which is unconditional on the event, also referred to

as abnormal return. The estimation model for expected or normal return is discussed next.

A model of normal or expected returns is used to estimate rit . Commonly employed is a basic market model using the CRSP equalweighted market index and can be expressed as

follows:

rit = α i + βiRmt + ε it (3.1)

Where:

t = 95…20, the estimation period

αi = a constant term for the ith stock

βi = the market beta of the ith stock

Rmt = the market return

The parameters of the model are estimated using the timeseries data from the estimation period that precedes each individual event. The estimated parameters are matched with the actual returns

during the event period (0, +1). Abnormal returns (AR it ) in equation (2) are calculated based on the actual returns during the event period and the estimated coefficients from the estimation period as follows:

ɸ AR it = Rit ( Ȥi + iRmt + ε it ) (3.2)

Cumulative abnormal returns ( CAR ) is calculated and analyzed for the event window (days

0 to t+1) surrounding the information activism event. Cumulative abnormal returns are summed over the event window as follows:

CAR it = ƉAR it (4)

The mean of the distribution of CAR is tested with a null hypothesis that CAR for days 0 to +1 is equal to zero.

4.3.2 Abnormal Trading Volume – Univariate Analysis

In order to test H1b as described in section 3.1.2 which suggests that information activism leads to abnormally high trading volume, abnormal trading volume (ABVOL) is examined around the information activism event (days 0 to t+1). To estimate abnormal trading volume, share turnover is used as a measure of daily trading volume and is defined as the ratio between the number of shares traded during the day over the number of shares outstanding (Lo & Wang 2000;

Stickel & Verrecchia 1994). Abnormal trading volume is calculated as the difference between the daily share turnover and the median share turnover (Bamber 1986; 1987). Median share turnover is calculated over the estimation period of 95 days ending 20 days prior to the information

activism event. The mean of the distribution of cumulative abnormal trading volume (ABVOL ) is tested with a null hypothesis that cumulative abnormal trading volume during the event window is equal to zero. Similar to CAR, in addition to the shortwindow measure examined in the univariate testing, a longwindow measure for ABVOL is constructed for each of the threemonth bull/bear market periods which will be analyzed as the dependent variable in the multivariate analysis described in section 4.3.5.

4.3.3 Information Activism Sentiment – Univariate Analysis

In order to test whether positive (negative) information activism is associated with higher

(lower) prices as predicted by H1c as discussed in section 3.1.3, the sample is partitioned into information activism events which are positive (bullish or buy recommendations) and those which are negative (bearish or sell recommendations). Abnormal returns are analyzed to determine if positive information activism is associated with higher prices by examining whether the returns are significantly positive, and whether returns are significantly negative for negative information activism.

4.3.4 Moderating Effects – Univariate Analysis

In order to test the moderating effects in the univariate analysis for investor sophistication

(H2), Information asymmetry (H4), and earnings quality (H5) the dataset is partitioned using the median measurement for each of these variables (see specific variable definitions in section 4.3.6).

For example, an observation is classified as: 1) High investor sophistication if it is above the median measurement and low investor sophistication otherwise (H2); 2) High information

asymmetry if it is above the median measurement and low information asymmetry otherwise (H4); and 3) High earnings quality if it is above the median measurement and low earnings quality otherwise (H5). The sample is also partitioned into two market condition categories (H3) based on when the information activism event took place, in either a bull or bear market (see discussion of bull/bear market in section 4.1.1). For each partitioned sample, a ttest is performed in order to determine if the CAR and cumulative abnormal trading volume (ABVOL ) over the 2day event

window (0, t+1) differ between the two groups (partitioned using each moderating variable).

4.3.5 Crosssectional Regressions – Multivariate Analysis

In order to further test hypotheses H1a, H1b, and H1c regarding the main effects of

information activism Intensity and information activism Sentiment on cumulative abnormal return

(CAR ) and cumulative abnormal trading volume (ABVOL ), a multivariate analysis is employed.

Monthly cumulative abnormal returns and trading volume over the threemonth bull and bear periods are examined in crosssectional regressions. The moderating effects of investor sophistication, market condition, information asymmetry, and earnings quality (H2, H3, H4, and

H5) are also examined. The dependent variable is the cumulative monthly abnormal returns (CAR )

(threemonth window for each bull and bear period) and cumulative monthly abnormal trading volume (ABVOL ) (same threemonth windows) for each firm in the sample. The independent variables are the information activism measures for Intensity (#PRG and #BLG ) and Sentiment

(PRG%B and BLG%B ) both described in section 4.1.2, for each firm in the sample, as well as

empirical proxies for each moderating variable described in the hypothesized relationships (H2,

H3, H4, and H5), interactive terms, and appropriate control variables. The regression models and variable definitions are described next.

4.3.6 Regression Functions and Variable Definitions

Two regression equations (5) and (6) are used to analyze information activism Intensity

and are detailed below. The dependent variable in each equation is the respective measure of

investor behavior (market reaction) for cumulative abnormal returns ( CAR ) and cumulative

abnormal trading volume ( ABVOL ) over the threemonth bull/bear period window. Primary

independent variables of interest are the associated measures of Intensity for each source of

information activism ( PRG# and BLG# ) and the moderating variable ( SOPH, BEAR, INS , and

EQ ) interaction terms. Control variables ( PRESS and SIZE ) are also included. Each variable is

defined below the regression equations.

CAR = α + β1PRG# + β2BLG# + β3SOPH + β4BEAR + β5INS + β6EQ +

β7PRG#*SOPH + β8PRG#*BEAR + β9PRG#*INS + β10 PRG#*EQ +

β11 BLG#*SOPH + β12 BLG#*BEAR + β13 BLG#*INS + β14 BLG#*EQ + β15 PRESS +

β16 SIZE + ε (5)

ABVOL = α + β1PRG# + β2BLG# + β3SOPH + β4BEAR + β5INS + β6EQ +

β7PRG#*SOPH + β8PRG#*BEAR + β9PRG#*INS + β10 PRG#*EQ +

β11 BLG#*SOPH + β12 BLG#*BEAR + β13 BLG#*INS + β14 BLG#*EQ + β15 PRESS +

β16 SIZE + ε (6)

Two regression equations (7) and (8) are used to analyze information activism Sentiment

and are detailed below. The dependent variable in each equation is the respective measure of

investor behavior (market reaction) for cumulative abnormal returns ( CAR ) and cumulative

abnormal trading volume ( ABVOL ) over the threemonth bull/bear period window. Primary

independent variables of interest are the associated measures of Sentiment for each source of

information activism ( PRG%B and BLG%B ) and the moderating variable (SOPH, BEAR, INS ,

and EQ ) interaction terms. Control variables ( PRESS and SIZE ) are also included. Each variable

is defined below the regression equations.

CAR = α + β1PRG%B + β2BLG%B + β3SOPH + β4BEAR + β5INS + β6EQ +

β7PRG%B*SOPH + β8PRG%B*BEAR + β9PRG%B *INS + β10 PRG%B*EQ +

β11 BLG%B*SOPH + β12 BLG%B*BEAR + β13 BLG%B*INS + β14 BLG%B*EQ +

β15 PRESS + β16 SIZE + ε (7)

ABVOL = α + β1PRG%B + β2BLG%B + β3SOPH + β4BEAR + β5INS + β6EQ +

β7PRG%B*SOPH + β8PRG%B*BEAR + β9PRG%B *INS + β10 PRG%B*EQ +

β11 BLG%B*SOPH + β12 BLG%B*BEAR + β13 BLG%B*INS + β14 BLG%B*EQ +

β15 PRESS + β16 SIZE + ε (8)

Where:

Dependent variables :

CAR = cumulative monthly abnormal return for the firm over the longrun three

month window using the market model described in section 4.3.1. and an

estimation period of 12 months ending one month prior to the 3month bull (bear)

market period.

ABVOL = cumulative monthly abnormal trading volume, calculated as the difference

between the median daily share turnover over the threemonth bull/bear market

period and the median share turnover calculated over the estimation period of 12

months ending one month prior to the beginning of the threemonth bull (bear)

market period.

Independent Variables :

PRG# = number of times a firm is mentioned on the Mad Money investment news

program during the threemonth bull/bear time period.

PRG%B = ratio of buy recommendations to total mentions on the Mad Money

investment news program during the threemonth bull/bear time period.

BLG# = number of times a firm is mentioned on the SeekingAlpha financial blog during

the threemonth bull/bear time period.

BLG%B = ratio of buy recommendations to total mentions on the SeekingAlpha financial

blog during the threemonth bull/bear time period.

Moderating Variables :

SOPH = investor sophistication for which the most common proxy is provided by

institutional holdings (Hand 1990; Walther 1997; Bartov et al. 2000) is measured

as the percentage of the company’s aggregate number of shares held by

institutions to common shares outstanding for the quarter preceding the bear

period (2 nd quarter ended June 30, 2008), and the bull period (4 th quarter ended

December 31, 2008). Institutional holders are those investment managers having a

fair market value of equity assets under management of $100 million or more.

BEAR = dummy variable for the market condition; 1 = Bear market; 0 = Bull market

assigned based on when the firm was mentioned in the information activism event

and whether the month in which it occurred was designated as a bull or bear

month.

INS = information asymmetry for which a widely regarded measure is bidask spread

(Glosten & Milgrom 1985; French & Roll 1986; Guo et al. 2004) is calculated as

the absolute value of the difference between the closing bid and ask prices, scaled

by the mean of the bid and ask (Guo et al. 2004). The mean bidask spread is

computed for one year preceding the Bear period (June 2007 to July 2008) in this

study which is prior to the financial meltdown in order to use a stable measure of

information asymmetry.

EQ = earnings quality is measured using the DichowDichev (2002) model. See

section 4.3.7 for further discussion of this measure and associated equations

Interaction Terms :

PRG#*SOPH, BLG#*SOPH, PRG%B*SOPH and BLG%B*SOPH

= interaction terms based on PRG# x SOPH, BLG# x SOPH, PRG%B x SOPH

and BLG%B x SOPH ; the coefficient is expected to be negative if less

sophisticated investors rely more heavily on information activism.

PRG#*BEAR, BLG#*BEAR, PRG%B*BEAR and BLG%B*BEAR

= interaction terms based on PRG# x BEAR, BLG# x BEAR , PRG%B x BEAR and

BLG%B x BEAR; the coefficient is expected to be positive if investors are

expected to rely more heavily on information activism during unstable economic

conditions, such as a bear market period.

PRG#*INS, BLG#*INS, PRG%B *INS and BLG%B*INS

= interaction terms based on PRG# x INS, BLG# x INS, PRG%B x INS, and

BLG%B x INS; the coefficient is expected to be positive if increased information

asymmetry is expected to be associated with a stronger investor reaction to

information activism.

PRG#*EQ, BLG#*EQ, PRG%B*EQ and BLG%B*EQ

= interaction terms based on PRG# x EQ, BLG# x EQ, PRG%B x EQ and BLG%B

x EQ; this metric (see 4.3.7) represents the variation in the mapping of earnings

into operating cash flows, and large (small) values correspond to poor (good)

earnings quality; if poor earnings quality is associated with a stronger market

reaction to information activism then the coefficient is expected to be positive .

Control Variables :

PRESS = dummy variable set to 1 if the firm had any significant news, earnings

announcements, or SEC filings around the information activism event during the

threemonth bull (bear) market period.

SIZE = market value of common stock; a control variable for the wellknown firmsize

effects (Stickel 1995); Firmsize effect refers to the notion that less information is

generally available for small firms and therefore market reactions to news events

are often stronger. The logarithm of the market value of common stock is

calculated for the most recent quarter preceding each bull and bear period.

4.3.7 Earnings Quality

Various measures of earning quality have been employed throughout the literature.

Francis et al. (2004) identify a number of important earnings attributes utilized by studies such as accruals quality, earnings persistence, predictability, smoothness, value relevance, timeliness, and conservatism. Information risk, which results from the uncertainty in investment payoff based on information available to investors (Easley & O’Hara 2004), is relevant to this research.

As mentioned earlier in 3.2.4, investors may search for alternate sources of information as a result of financial statement limitations and the inability to accurately predict future earnings due to poor earnings quality. Francis et al. (2004) note that accruals quality has the most direct link to

information risk. Accruals quality represents the variation in the mapping of earnings into operating cash flows, a key element of the investment payoff structure which is important to investors. A measure of accruals quality is provided by Dechow and Dichev (2002) in a model which relates current accruals to lagged, current, and future cash flows from operations and is as follows:

TCA j,t / Assets j,t = φ0,j + φ1,j (CFO j,t1 / Assets j,t ) + φ2,j (CFO j,t / Assets j,t )

+ φ3,j (CFO j, t+1 / Assets j,t ) + w j,t (9)

Where:

TCA j,t = firm j’s total current accruals in year t, = (CAj,t – CL j,t – Cash j,t +

STDEBT j,t );

Assets j,t = firm j’s average total assets in year t and t1;

CFO j,t = cash flow from operations in year t, calculated as net income before

extraordinary items (NIBEOI) less total accruals (TA),

Where:

TA j,t = CA j,t – CL j,t – Cash j,t + STDEBT j,t – DEPN j,t

Where:

CA j,t = firm j’s change in current assets between year t1 and year t;

CL j,t = firm j’s change in current liabilities between year t1 and year t;

Cash j,t = firm j’s change in cash between year t1 and year t;

STDEBT j,t = firm j’s change in debt in current liabilities between year t1 and

year t;

DEPN j,t = firm j’s depreciation and amortization expense in year t.

For each firmyear, equation (9) is estimated using a rolling tenyear window. These estimations

yield ten firm and yearspecific residuals wj,t,t = t9…..t, which form the basis for the earnings

quality metric, EQ j,t = σ( ŵj,t), equal to the standard deviation of firm j’s estimated residuals. Year t is the most recent year nearest to the periods examined in this study which is the fiscal year ending 2009. Large (small) values of EQ correspond to poor (good) earnings quality.

CHAPTER 5

EMPIRICAL RESULTS

5.1 Univariate Results

5.1.1 Effect of Information Activism on Investor Behavior Returns

5.1.1.1 Returns – Overall (H1a)

The results of the univariate analysis conducted on the effect of information activism on investor behavior as measured by cumulative abnormal returns ( CAR ) are provided in Table 4 where the results are tabulated for All Events , and in tables 5 and 6 where the results are provided for the individual Cramer and Blog samples respectively. In order to avoid distortions from outliers the return data has been winsorized at the .01 and .99 levels. The Overall results for All

Events in Table 4 provide support for H1a, that information activism is associated with abnormal returns, as the mean CAR of 0.00619 is significant (t = 4.43; p <.0001). The Overall results for the

Cramer sample in Table 5 also support H1a as the mean CAR of 0.00650 is significant (t = 4.15; p

<.0001). H1a is not supported using only the Blog sample (Table 6) as the Overall mean CAR of

0.00466 is not significant (t = 1.55; p = 0.1226). This may be a result of the small sample size.

5.1.1.2 Returns – Sentiment (H1c)

The results for All Events using the combined samples in Table 4 also provide support for

H1c, that positive information activism leads to higher prices, as depicted under Sentiment in

Table 4. The mean CAR for Buy recommendations is positive (0.00975) as expected and

significant (t = 6.13; p <.0001). Although the mean CAR of 0.00052 for Sell recommendations is

negative as predicted, it is not significant (t = 0.19; p = 0.8456). However, the difference in

means between the Buy and the Sell mean CARs of 0.01027 is significant (t = 3.50; p = 0.0005)

indicating that Sentiment may be a factor related to the effect of information activism on price.

Using the Cramer sample (Table 5), H1c continues to be supported. Similar to the overall results,

the mean CAR for Buy recommendations of 0.01119 is positive as expected and significant (t =

6.29; p <.0001). Again, the mean CAR of 0.00105 for Sell recommendations is negative as

expected, but is not significant (t = 0.36; p = 0.7170). However the difference between the mean

CAR for Buy and Sell recommendations of 0.01224 is significant (t = 3.61; p = 0.003). This provides evidence that investors may respond differently to the Sentiment of information activism.

Once again the results for H1c using the Blog sample (Table 6) do not provide evidence of support

as the mean CARs are not significant, nor is the difference between the Buy and Sell

recommendations.

5.1.1.3 Returns – Investor Sophistication (H2)

The results for All Events in the combined sample which are summarized in Table 4 provide limited support for H2, that the effect of information activism on prices will be greater for

unsophisticated investors (Low ) than for sophisticated investors (High ). According to H2 it is

expected that the mean CAR for unsophisticated investors ( Low ) will be greater than the mean

CAR for sophisticated investors ( High ). In analyzing the results for All Events and buy

recommendations, while the mean CAR for Low investor sophistication of 0.01121 (t = 5.55; p

<.0001) is greater than the mean CAR for High investor sophistication of 0.00771 (t = 3.14; p =

0.0018) as expected, the difference between the mean CAR for High and Low investor

sophistication of 0.0035 for buy recommendations is not significant (t = 1.09; p = 0.2748). When analyzing the sell recommendations, while the mean CAR for Low investor sophistication of

0.0053 (t = 1.25; p = 0.2125) is greater than the mean CAR for High investor sophistication of

0.00500 (t = 1.32; p = 0.1862) as expected, neither mean CARs are significant. However the

difference between the Low investor sophistication and the High investor sophistication of

0.01031 (t = 1.81; p = 0.0702) for the sell recommendations is significant. This suggests that

investor sophistication may be important with regard to the effect of information activism on price.

When analyzing the results regarding investor sophistication (H2) using the Cramer

sample (Table 5) for the both the buy and sell recommendations, the mean CARs for Low investor sophistication are higher than the mean CARs for High investor sophistication as expected.

However in the case of the buy recommendations, while the mean CARs of 0.01213 (t = 5.44;

p <.0001) and 0.00926 (t = 3.30; p = 0.0011) for the Low and High investor sophistication

respectively are both significant, the difference between the two mean CARs of 0.00287 (t = 0.80; p = 0.4211) is not significant. While examining the sell recommendations, it can be noted that the mean CARs of 0.00535 (t = 1.18; p = 0.2396) and 0.00465 (t = 1.15; p = 0.2531) for Low and

High investor sophistication respectively are not significant, nor is the difference between the two mean CARs of 0.01000 (t = 1.64; p = 0.1012). These results do not support the notion that the level of investor sophistication influences the effect of information activism on price reaction. Finally, when analyzing the level of investor sophistication using the Blog sample (Table 6), only the mean

CAR for buy recommendations and Low investor sophistication of 0.00780 (t = 1.79; p = 0.0755)

is significant and greater than the mean CAR for High investor sophistication ( buy

recommendations) as expected. However the other mean CARs are not significant, nor are the

differences in the means between the High and Low investor sophistication for both the buy and sell recommendations. Therefore, using the Blog and Cramer samples alone, there is no support

for H2.

5.1.1.4 Returns – Market Condition (H3)

The Univariate results provide some support for H3, particularly for positive information

activism events or buy recommendations. In fact, H3 is consistently supported using the

combined sample (Table 4), and the individual Cramer (Table 5), and Blog (Table 6) samples

when analyzing the buy recommendations. H3 predicts that the effect of information activism on prices will be greater during bearish market conditions than during bullish market conditions.

Upon considering the buy recommendations for the results for All Events in Table 4, while the

mean CAR during the Bear period of 0.01730 (t = 6.93; p <.0001) is significant and greater than

the Bull mean CAR of 0.00079 (t = 0.46; p = 0.6465) as expected, the Bull mean CAR is not

significant. However the difference between the Bear mean CAR and the Bull mean CAR of

0.16510 (t = 5.45; p <.0001) is significant, indicating that for buy recommendations, investors

appear to rely more on information activism during Bear markets than during Bull markets.

Similar results are observed when considering the buy recommendations for the Cramer

sample (Table 5), where the mean CAR for the Bear period of 0.02069 (t – 7.02; p <.0001) is

greater than the Bull market mean CAR of 0.00235 (t = 1.22; p = 0.2221) as expected and the

difference of 0.01830 (t = 5.35; p <.0001) is also significant, providing further support for H3.

When considering the information activism events occurring on the Blog (Table 6) , H3 continues

to be supported for buy recommendations. Both mean CARs for the Bear and Bull periods of

0.00932 (t = 2.09; p = 0.0374) and 0.00593 (t = 1.66; p = 0.0994) respectively, are significant.

The Bear CAR is also greater than the Bull CAR as expected and the difference between the Bear and Bull CARs of 0.01525 (t = 2.67; p = 0.0079) is significant. This provides additional support for

H3, that investors rely more on information activism during Bear market periods than during Bull

market periods, in this case particularly for positive information activism or buy

recommendations. When considering the sell recommendations for all three samples, there is no

support for H3.

5.1.1.5 Returns – Information Asymmetry (H4)

The results regarding H4, which suggests that the effect of information activism on price

will be greater for firms with High information asymmetry than for firms with Low information

asymmetry, are not supported in the univariate analysis. In the sample for All Events (Table 4)

regarding buy recommendations, the mean CAR for High information asymmetry of 0.00814 (t =

2.82; p = 0.0051) is significant but is not greater than the mean CAR for Low information asymmetry of 0.00998 (t = 5.39; p <.0001), which is also significant. Furthermore, the difference between the two mean CARs of 0.00184 (t = 0.55; p = 0.5857) is not significant. Regarding buy recommendations and the Cramer sample (Table 5), both High and Low information asymmetry

mean CARs are significant and the mean CAR for the High information asymmetry events of

0.01236 (t = 3.63; p = 0.0003) is greater than the mean CAR for Low information asymmetry

events of 0.01076 (t = 5.26; p <.0001). However, similar to the sample for All Events , the

difference between the two means of 0.0016 (t = 0.41; p = 0.6805) is not significant, providing no support for H4. Furthermore, the results for the buy recommendations from the Blog sample

(Table 6) provide no support for H4 as the mean CARs are not significant, nor is the difference between the means. Finally, with respect to sell recommendations, no support is offered for H4 as the individual mean CARs are not significant, nor are the differences in mean CARs between High

and Low information asymmetry.

5.1.1.6 Returns – Earnings Quality (H5)

Regarding H5, which suggests that the effect of information activism on price will be

greater for firms with Low earnings quality than for firms with High earnings quality, little support

is evident in the univariate analysis. In most cases for all three samples, the mean CAR for Low

earnings quality is not greater than the mean CAR for High earnings quality, as predicted in H5.

Furthermore, while some individual mean CARs are significant, the difference in mean CARs is

not significant with the exception of the sell recommendations in the All Events sample. However,

the individual mean CARs are not significant, nor is the Low earnings quality mean CAR greater

than the High earnings quality mean CAR . Therefore, H5 appears to be unsupported in the

univariate analysis regarding price reaction to information activism.

5.1.2 Effect of Information Activism on Investor Behavior Trading Volume

5.1.2.1 Trading Volume – Overall (H1b)

The results of the univariate analysis performed on the effect of information activism on

investor behavior as measured by cumulative abnormal trading volume ( ABVOL ) are provided in

Table 7 ( All Events ), Table 8 ( Cramer events ), and Table 9 ( Blog events). In order to avoid distortions from outliers, the trading volume data has been winsorized at the .01 and .99 levels.

The overall results with respect to each sample provide support for H1b, that information activism is associated with abnormal trading volume, as the mean ABVOLs of 22.29 (t = 25.72; p <.0001),

20.57 (t = 29.90; p <.0001), and 41.40 (t = 5.67; p <.0001) for All Events , Cramer events, and

Blog events respectively, are all significant. Additional analysis was performed where these samples were split according to buy and sell recommendations and results remained significant.

Notably, in all three sample categories, the sell recommendations exhibit a stronger reaction with respect to abnormal trading volume than do buy recommendations. These results suggest that information activism appears to be associated with investor behavior through abnormal trading volume.

5.1.2.2 Trading Volume – Investor Sophistication (H2)

Regarding investor sophistication and H2, the mean ABVOL for Low investor sophistication is predicted to be greater than the mean ABVOL for High investor sophistication.

Unfortunately, the results provide no support for this prediction. While a number of the individual mean ABVOLs in each of the sample segments are significant, the differences between the mean

ABVOLs for Low and High investor sophistication are not significant. Regarding the sell recommendations for the All Events sample, while mean ABVOL for Low investor sophistication appears to be greater than the mean ABVOL for High investor sophistication as predicted by H2, the differences in means are not significant. With regard to buy recommendations, the mean

ABVOL in two of the three sample segments ( All Events and Cramer events) for the Low investor

sophistication does not appear to be greater than the mean ABVOL for High investor sophistication. Furthermore, the difference in means is not significant for any of the sample segments. These results do not provide support for H2, that the effect of information activism on trading volume is greater for unsophisticated investors than for sophisticated investors.

5.1.2.3 Trading Volume – Market Condition (H3)

H3 predicts that the effect of information activism on trading volume will be greater during

Bear market conditions than during Bull market conditions. The univariate results in each of the sample segments provide little support for this prediction. While a number of the individual mean

ABVOLs are significant, only one of the differences in mean ABVOL is significant for buy recommendations in the Cramer sample as reported in Table 8. Furthermore, only the buy recommendations appear to have a mean ABVOL which is greater for the Bear market events than for the Bull market events. In the Cramer sample (Table 8) for the buy recommendations, the mean ABVOL for the Bear market period of 19.97 (t = 19.10; p <.0001) is significant and greater than the mean ABVOL for the Bull market period of 16.05 (t = 17.54; p <.0001), which is also significant. In addition, the difference between the Bear and Bull mean ABVOLs of 3.92 (t = 2.84; p = 0.0047) is also significant. For the sell recommendation events in all three sample segments, the mean ABVOL for the Bear period events do not appear to be greater than the mean ABVOLs for the Bull period events, nor are the differences in mean ABVOLs significant. Therefore, there is only minor support for H3, where the moderating effect of market condition is examined for the effect of information activism on trading volume.

5.1.2.4 Trading Volume – Information Asymmetry (H4)

The univariate results regarding information asymmetry provide little support for the prediction in H4, which suggests that the effect of information activism on trading volume will be greater for firms with High information asymmetry than for firms with Low information asymmetry. Again, while many of the individual mean ABVOLs are significant, few of the differences in mean ABVOLs between High and Low information asymmetry are significant. Only the Cramer sample segment for the buy recommendations offers support for H4. Here, the mean

ABVOL for High information asymmetry firms of 20.73 (t = 14.32; p <.0001) is significant and greater than the mean ABVOL for Low information asymmetry firms of 16.78 (t = 21.57; p

<.0001), which is also significant. Furthermore, the difference in mean ABVOL between High and

Low information asymmetry firms of 3.95 (t = 2.41; p = 0.0165) is significant. Therefore, the prediction provided by H4 appears to only be supported for buy recommendations in the Cramer sample.

5.1.2.5 Trading Volume – Earnings Quality (H5)

A fair amount of support is offered in the univariate analysis regarding H5, which predicts that the mean ABVOL associated with information activism will be greater for firms with Low earnings quality than for firms with High earnings quality. In Table 7, when considering results for

All Events , results provide support for H5. The mean ABVOL for buy recommendations and Low earnings quality of 17.67 (t = 14.09; p <.0001) is greater than the ABVOL for High earnings quality of 13.08 (t = 12.69; p <.0001), and both are significant. Furthermore, the difference between the Low and High earnings quality mean ABVOLs for buy recommendations of 4.59 (t =

2.83; p = 0.0048) is also significant. Similarly the mean ABVOL for Sell recommendations and

Low earnings quality of 29.37 (t = 7.40; p <.0001) is greater than the ABVOL for High earnings

quality of 17.76 (t = 6.94; p <.0001), and both are significant. The difference between the Low

and High earnings quality mean ABVOLs for sell recommendations of 11.61 (t = 2.46; p = 0.0144)

is also significant.

Similar findings are evident in Table 8 for Cramer events, where all mean ABVOLs are

significant and the mean ABVOLs for Low earnings quality appear to be greater than the mean

ABVOLs for High earnings quality. The difference in means is also significant. Regarding the Blog

sample (Table 9), support is provided for H5 with regard to buy recommendations only, as the

mean ABVOL for Low earnings quality of 16.18 (t = 8.48; p <.0001) is significant and greater than

the mean ABVOL for High earnings quality of 12.11 (t = 8.09; p <.0001), which is also significant.

The difference between the mean ABVOLs of 4.07 (t = 1.68; p = 0.0945) is also significant.

Regarding sell recommendations, while the individual mean ABVOLs are significant, the mean

ABVOL for Low earnings quality does not appear to be greater than the mean ABVOL for High

earnings quality, nor is the difference in mean ABVOLs significant. Taken together, the results in

the univariate analysis regarding the effect of information activism on trading volume provide

some support for the moderating effect of earnings quality (H5).

5.2 Multivariate Results

To further test the hypotheses put forth in this dissertation, a multivariate regression

analysis using the equations detailed in section 4.3.6 is conducted. Table 10 provides the basic

descriptive statistics with regard to the variables utilized in regression equations (5), (6), (7), and

(8). Also, the Pearson and Spearman correlation matrices are offered in Table 11. Diagnostic tests

were run on the multivariate data to examine the various OLS assumptions. No significant issues were noted with regard to heteroscedasticity, multicollinearity, or autocorrelation.

5.2.1 Intensity

In order to further test the Intensity of information activism on investor behavior, multivariate regression equations (5) and (6) are analyzed for which results are provided in Tables

12 and 13. The model in equation (5) examines CAR as the dependent variable and equation (6) examines ABVOL as the dependent variable. Both models include Intensity variables of PRG# and

BLG# as the primary variables of interest as well as the associated interaction terms for the

moderating effects as described in section 4.3.6.

The models provided in Tables 12 and 13 for both equation (5) and (6) appear to be robust

as indicated by the significant F statistics of 3.34 (p <.0001) and 8.71 (p <.0001) respectively and

adjusted R 2 of 5.29 % and 15.53%, respectively. Regarding the primary variables of interest for

Intensity (PRG# and BLG#), neither variable is significant in the CAR model for equation (5)

(Table 12). However when the variable EQ 3 is eliminated from the model to allow for a larger sample size (Equations 5.1 & 5.2), and the EXCHG variable is added as a control variable for the stock exchange in which the firm is traded (Equation 5.2), the coefficient for PRG# (p = 0.0280; p

= 0.0328) is significant. This provides support for H1a that information activism is associated with abnormal returns for the Cramer sample (PRG#). Regarding Intensity for the Blog sample (BLG#), the coefficient is not significant.

3 The EQ variable results in a significant reduction in sample size due to the large number of years of data needed to calculate this variable using the DechowDichev (2002) model (see section 4.3.7). When removing this variable from the model, sample size increases from 672 to 1,089 (see N in Tables 12, 13, 14, and 15).

The effect of information activism on trading volume (ABVOL) is provided in Table 13

using equation (6). Although the coefficient for PRG# (p <.0001, p = .0006, p = 0.0020 for

equations (6), (6.1), and (6.2) respectively is significant, the coefficient is negative which may

require some further explanation as to why these results are obtained. In the original hypothesis

discussion for H1b, it is expected that the intensity of information activism, in this case the

number of times a firm is mentioned on Cramer’s program, will be associated with abnormal

trading volume . In this case the results in equation (6), (6.1), and (6.2) are actually suggesting that

there is a negative relationship between the number of times a firm is mentioned and abnormal

trading volume. In other words the more a firm is mentioned on Cramer’s program the less

abnormal trading volume takes place or it actually leads to the stock being traded less. Possible

explanations for this result may include the fact that if a stock is receiving a lot of attention on

Cramer’s program over the threemonth periods examined in this study, investors may actually

trade the stock less as they may assume that any additional increase in returns was earned when

the initial attention was given to the stock. Particularly, as a stock is mentioned more and more

during a given period, investors may recognize that it is receiving increased attention and therefore

not trade that stock as much assuming any return earnings were realized when the stock was first

mentioned. Also, as the stock receives more information activism, it may eventually drive up the price as was noted in the CAR analysis (Table 12), making the stock less attractive and therefore

resulting in less trading.

With regard to the Intensity variable for the Blog data source ( BLG#) and the effect on abnormal trading volume, no support was provided in equations (6) and (6.1) where the coefficient for BLG# is not significant. However, the coefficient for BLG# in equation (6.2) is significant (p =

0.0391), but similar to the PRG# coefficients in equations (6.1) and (6.2), the BLG# coefficient in equation (6.2) is negative. The explanations provided with regard to the PRG# variable would also apply for the observed negative and significant coefficient for BLG#.

5.2.2 Sentiment

In order to further test the Sentiment of information activism on investor behavior, multivariate regression equations (7) and (8) are analyzed. Tables 14 and 15 summarize the results of these equations. The model in equation (7) includes CAR as the dependent variable and equation (8) includes ABVOL as the dependent variable. Both models include Sentiment variables of PRG%B and BLG%B as the primary variables of interest as well as the associated interaction terms for the moderating variables as described in section 4.3.6. Equation (7.1), (7.2), (8.1), and

(8.1) again modify the original equations for equation (7) and (8) by removing the EQ variable

[equations (7.1) and (8.1)] and in equations (7.2) and (8.2) removing EQ and adding EXCHG a control variable for the stock exchange on which the firm is traded. Similar to the models for

Intensity, the models provided for Sentiment in Tables 14 and 15 for both equation (7) and (8) appear to be robust as indicated by the significant F statistics of 3.16 (p <.0001) and 3.49

(p<.0001) respectively and adjusted R 2 of 4.90 % and 5.61%, respectively. When the original equations are modified (Equation 7.1, 7.2, 8.1, and 8.2) as described above, they continue to present a good fit as noted by the Fstatistics and adjusted R 2.

The regression equations constructed to examine Sentiment do not yield promising results.

Although, all three CAR equations (7), (7.1), and (7.2) seem to provide a good fit as indicated by their F statistics (3.16, p <.0001; 4.40, p <.0001; 3.86, p <.0001) in Table 14, the coefficients for

Intensity (PRG%B and BLG%B) are not significant. The ABVOL equations (8), (8.1) and (8.2)

(Table 15) are also robust similar to the CAR equations, as indicated by their respective F statistics

(3.49, p <.0001; 5.79, p <.0001; 49.61, p <.0001) and associated adjusted R 2 values (5.61%,

5.41%, 41.69%). While the coefficient for PRG%B (p = 0.0007) in equation (8.1) and the

coefficients for PRG%B (p = 0.0063) and BLG%B (p = 0.0320) in equation (8.2) are significant,

the coefficients are negative in contrast to their predicted sign with regard to H1c. These results

would follow the same possible explanations as described in section 5.2.1 regarding the negative

and significant coefficients for the Intensity variables for PRG# in equations (6), (6.1) and (6.2)

and for BLG# in equation (6.2).

5.2.3 Moderating Effects

Overall, the multivariate regression analysis provides slight support for the moderating

effects put forth in this study. The moderating effects are mostly seen in the multivariate analysis

regarding the Intensity (Table 12 and 13) of information activism. Support is provided for the

moderating effect of market condition or the BEAR variable, which according to H3, the effect of

information activism on price ( CAR ) and trading volume ( ABVOL ) will be greater for firms during

bear economic time periods than during bull periods. In equation (5) and (5.1) (Table 12), the moderating effects of market condition on the Intensity of information activism from the Blog

sample, or BLG#*BEAR (p = 0.0954; p = 0.0920) is significant and positive as expected. These

results provide some evidence that investors may react more strongly to information activism

during bear market periods than during bull market periods for the firms mentioned on the Blog.

Regarding the moderating effect of market condition on trading volume ( ABVOL ) and the Intensity

of information activism using equation (6) (Table 13), although there is no support for the information activism on Blogs (BLG#*BEAR) as was noted previously for the CAR analysis (Table

12) , there is support offered for the Intensity of information activism on Cramer (PRG#*BEAR).

The coefficient for PRG#*BEAR (p = 0.0491) is significant and positive as expected suggesting

that investors appear to respond more to the Intensity of information activism on Cramer during

bear market periods than during bull market periods.

Support is also offered for the moderating effect of the level of Information Asymmetry

(INS) and the effect of the Intensity of information activism on trading volume (ABVOL) in Table

13. In all three equations (6), (6.1), and (6.2) the moderating effect of Information Asymmetry for the effect of information activism on trading volume for the Cramer sample (PRG#*INS) (p

<.0001; p = 0.0537; p = 0.0062) is significant, large, and positive as expected. This provides

support for H4, which proposes that the reaction of information activism on trading volume will be greater for firms with high information asymmetry than for firms with low information asymmetry. Further evidence of the moderating effect of Information Asymmetry is provided in

the multivariate analysis regarding Sentiment of information activism and the effect on trading

volume ( ABVOL) for the Cramer sample as shown in Table 15. The coefficient for PRG%B*INS

(p = 0.0212; p = 0.0893) in equations (8) and (8.2) is significant, large, and positive as expected according to H4. Both of these results regarding Information Asymmetry strengthen the argument that the effect of information activism on investor behavior, as measured by price and trading volume reaction, is greater for firms with high information asymmetry than for firms with low information asymmetry.

Finally, slight support is offered in the multivariate analysis regarding the moderating effects of Earnings Quality on the effect of Sentiment of information activism on abnormal returns

(CAR) in Table 14 . In equation (7), the predicted moderating effect of earnings quality on the effect of the Sentiment of information activism on abnormal returns for the Blog sample

(BLG%B*EQ ) (p = 0.0370) is significant, large, and positive as expected. H5 proposes that for firms with low earnings quality, the effect of information activism on trading volume will be greater than for firms with high earnings quality. Since the larger the EQ variable, the lower the earnings quality (see section 4.3.7), and if in fact H5 is true, then the coefficient for the EQ

interaction terms will be positive. Therefore, some support for this prediction is provided for

Sentiment of information activism for the Blog sample.

5.2.4 Control Variables

In most all equations, the control variables for SIZE and EXCHG are significant,

suggesting that these control variables may aid in removing the associated effects of firm size and

stock exchange on which the security is traded. In three equations (7.1), (7.2), and (8.2), the

PRESS variable was important. So while it is necessary to identify confounding events such as

SEC filings, earnings announcements, and major news for the event study in the univariate analysis, this information may not be significant when analyzing the effects of information activism on investor behavior via price and trading volume in the multivariate model.

5.2.5 Unexpected Results – Moderating Effects

A few of the multivariate analysis yielded unexpected results with regard to moderating effects in contrast the original hypothesis development and therefore require some further discussion. Unexpected results in the multivariate analysis were evident with regard to the moderating effects of Information Asymmetry and the effect of the Intensity of information activism on returns (CAR ) in Table 12 and for the moderating effects of Investor Sophistication and effect of the Intensity of information activism on trading volume (ABVOL) in Table 13. In equation (5.1) and (5.2) the coefficients for PRG#*INS are significant and large, but are negative in contrast to the predictions made in H4, which were supported in some of the models as described in section 5.2.3. These results require some further investigation, particularly since the predictions of H4 are supported by other models put forth in this study. However, these unexpected results where the coefficients are significant, but negative, may suggests that at times primarily with regard to price reaction (CAR) and the effect of the Intensity of information activism for the Cramer sample, the opposite effect is experienced. This means that when

Information Asymmetry is high, there is less of a reaction to information activism. If the firm is less transparent with regard to information provided to investors ( high information asymmetry), then why would investors rely less on infomediaries such as information activism on Cramer?

One explanation is that if information asymmetry is high, then investors may not be looking to

Cramer for information about the firm, but may look toward other information sources.

Unexpected results were also observed with regard to the moderating effects of Investor

Sophistication on the effect of the Intensity of information activism on trading volume (ABVOL)

as detailed in Table 13. In this analysis, the coefficients for the interaction term for the Intensity of

information activism and Investor Sophistication for both samples ( PRG#*SOPH and

BLG#*SOPH) for all three models in equations (6), (6.1), and (6.2) is significant and large, but is positive in contrast to the predicted sign proposed in H2. H2 suggests that the effect of information

activism on prices and trading volume will be greater for unsophisticated investors than for

sophisticated investors. Since the proxy for investor sophistication is the percentage of

institutional holdings, the higher the variable for SOPH, the smaller the reaction on CAR and

ABVOL, which means the coefficient for the interaction terms for SOPH will be negative. These results are in direct opposition to the expected results, suggesting that more sophisticated investors will trade more in response to the Intensity of information activism. It seems unlikely that the more sophisticated an investor, the more trading they will do in response to information activism.

It may just be that when certain securities are receiving consistent attention, institutional owners are trading, not necessarily in response to information activism, but in general because a stock is experiencing a lot of activity and attention in the news and when institutional investors hold more shares, their trading activities will be observed. An alternate possible explanation is that some other event is triggering sophisticated investors to trade, and although the information activism may prompt unsophisticated investors to trade, the effect from some other event causing sophisticated investors to trade is greater and therefore the effects for the unsophisticated investor cannot be observed using this analysis.

CHAPTER 6

SUMMARY & CONCLUSIONS, LIMITATIONS, & FUTURE RESEARCH

6.1 Summary & Conclusions

This dissertation introduces the concept of information activism and examines the effect of information activism on capital markets by analyzing investor behavior through price and volume reactions to instances of information activism. It is important to gain further insight and understanding of information activism due to the abundance of financial commentary and supplemental information available to investors via sources such as financial cable news networks and Internet financial blogs. These sources of information activism often prompt investors to take a particular action with regard to investment decisions. Information activism sources are frequently criticized as appearing biased and as having the potential to exacerbate already unstable economic conditions, such as the financial crisis in 2008. Therefore, it is relevant to assess the extent to which investors appear to react to information activism.

This dissertation draws upon prior research studies and theoretical tenets to suggest important moderating effects which may determine the extent to which investors rely on information activism. Specifically, it is hypothesized that the effect of information activism on investor behavior via price and trading volume is greater for unsophisticated investors, for firms with high information asymmetry and low earnings quality, and during bearish economic conditions. A univariate analysis is conducted to directly examine the main hypotheses as well as the hypotheses related to moderating effects. Additionally, a multivariate analysis where these

relationships are examined crosssectionally is employed. A summary of the findings related to each type of analysis (univariate and multivariate) for each of the hypothesized relationships as well as the overall effect of information activism on investor behavior is provided in Figure 5.

6.1.1 Univariate Summary

In the univariate analysis, a final sample of 2,107 firm events is examined. This sample size varies slightly in the individual analyses due to missing data values related to variables needed for H2, H4, and H5. In examining the overall effect of information activism on investor behavior via price ( CAR ) and trading volume ( ABVOL ), support is provided for H1a and H1b, suggesting that information activism does appear to be associated with investor behavior through price and trading volume. Regarding the Sentiment of information activism, support is provided in

the univariate analysis for H1c for buy recommendations, that positive information activism

appears to be associated with higher prices, while no support is offered to suggest that negative

information activism (sell recommendations) is associated with lower prices. Although some of

the results are mixed for each of the hypotheses related to the moderating effects in the univariate

results, all but one of the hypotheses related to moderating effects associated with the effect of

information activism on investor behavior are supported at some level. The moderating effect for

Investor Sophistication related to the effect of information activism on price ( CAR ) and trading volume (ABVOL), for H2, is not supported. H3, concerning the moderating effect of Market

Condition on the effect of information activism on price reaction ( CAR ), is supported for buy recommendations and for the effect on trading volume ( ABVOL), mixed support is offered for buy recommendations. H4, which proposes that Information Asymmetry will moderate the effect of

information activism on investor behavior is not supported for price reaction (CAR), but offers

mixed support for buy recommendations with regard to trading volume reaction ( ABVOL). Finally,

in the univariate analysis, the moderating effect of earnings quality, as detailed in H5, is

unsupported with respect to the effect of information activism on price (CAR), but is supported for

the buy recommendations and the effect on trading volume (ABVOL) , and indicates mixed support

for sell recommendations.

6.1.2 Multivariate Summary

In the multivariate analysis, a sample of 672 firms (1,089 firms when the earnings quality

variable is dropped due to missing data) was examined using the equations detailed in Chapter 4.

These multivariate regressions provide a crosssectional analysis of both the Intensity and

Sentiment of information activism as well as the hypothesized moderating effects of investor

sophistication, market condition, information asymmetry, and earnings quality. Slight support is provided for Intensity and the effect on price (CAR) for the Cramer sample [PRG# in Equation

(5.1) and (5.2)] suggesting that the quantity of information activism directed toward a firm may be

associated with investor behavior. No support is provided for the Sentiment of information

activism and suggests that the attitude of the information activism (positive or negative) for a

given firm does not appear to be related to the effect on investor behavior. Similarly, no support is provided for the notion that investor sophistication is related to the effect of information activism

on investor behavior in the multivariate analysis. The remaining hypothesized relationships

related to moderating effects provide mixed findings and suggest that market condition,

information asymmetry, and earnings quality may be important with respect to the effect of information activism on investor behavior.

Taken as a whole the results from both the univariate and multivariate analyses provide a moderate degree of evidence that investors appear to react to information activism related to investment decisions. First regarding the main effects, the univariate event study provides support

(H1a & H1b ), while in the multivariate analysis minor support is offered only for H1a, that the

Intensity of information activism on the Cramer program is associated with abnormal returns. No support is found in the multivariate analysis for the effect of Sentiment of information activism on abnormal returns. No support is offered in the multivariate analysis regarding the effect of the

Intensity or the Sentiment of information activism on abnormal trading volume.

The results regarding a number of the moderating effects expected to be important for the primary relationship between information activism and investor behavior, indicate that market condition, information asymmetry, and earnings quality may influence this relationship.

Particularly, the moderating effect of market condition appears to influence the effect of the

Intensity of information activism on abnormal returns for the Blog sample. Market condition also appears to be an important moderator with respect to the effect of the Intensity of information activism on abnormal trading volume for the Cramer sample. In the multivariate results, there is evidence of a moderating effect of information asymmetry on the effect of the Intensity of information activism on abnormal trading volume for the Cramer sample. Finally regarding moderating effects on the Sentiment of information activism, only earnings quality and information asymmetry were supported. Earnings quality appears to moderate the effect of the

Sentiment of information activism on abnormal returns for the Blog sample. Similarly, information asymmetry appears to moderate the effect of the Sentiment of information activism on abnormal

trading volume for the Cramer sample.

6.2 Contributions

This dissertation contributes to the literature by introducing the construct of information

activism as it relates to understanding investor behavior in capital markets. This research examines

information activism from two primary sources, whereas previous studies primarily focus on one

narrow set of information sources. The effect of information activism on capital markets is

investigated by using two measurement methods, Intensity and Sentiment to understand the impact of information activism on capital markets.

In addition, this research puts forth important moderating effects related to information activism on investor behavior, which have not been considered in prior studies. An important contribution is the examination of the impact of information intermediaries during divergent market conditions (i.e. bull and bear markets), which has been examined little in prior research.

This study predicts that investors rely more on information activism during bearish market periods due to risk and loss aversion and examines this reaction during a significant downturn in the economy as a result of the financial crisis in 2008. Although slight support is provided regarding this assertion, the results suggest that market conditions may be important in understanding the effect of information activism on investor behavior. This research also provides further evidence regarding the relevance of financial statements and whether investors rely on alternate sources of information in making investment decisions. Earnings quality was examined as a moderating

factor related to the effect of information activism on investor behavior. Results provide some support that earnings quality may be important in understanding whether investors rely on information activism and low earnings quality may lead investors to search for and react to other sources of information for their investing decisions.

Finally, this research contributes to the debate regarding the role of the media and investment news in swaying investor’s decisions (e.g. “Cramer Effect”). While anecdotal evidence and prior studies suggest that these investment news outlets have an impact on investor behavior and capital markets, this study offers further evidence of such a reaction. Additionally, the results of this study shed light on the increasing demand, necessity, and importance of supplemental information (e.g. to financial statements) in shaping investor decisionmaking.

6.3 Limitations

Consistent with most research, this dissertation has several limitations that should be considered with regard to the results. One of the most important limitations for this study is that a control group of firms which are not mentioned in information activism events, is not analyzed for comparison purposes, in order to fully examine the effect of information activism on investor behavior. Obtaining such a control group would likely be difficult as the firms in this sample are selected based on the firm being mentioned on either Cramer’s show or the financial Blog . It would be a complex selection process to identify firms and ensure that they were not mentioned as part of an information activism event for these sources or in one of the many other available sources of information activism. Therefore, the interpretation of these results should be carefully considered due to this limitation.

A further limitation is that these results are based on a limited sample of events which were collected during a specific time period in order to scrutinize investor reaction during divergent market conditions. Selecting firms during this particular time period could bias the results. Also, the sources used to collect events related to information activism could also lead to some bias, as there are other sources of information activism which could be investigated. Alternate sample time periods and information activism sources should be considered for future research endeavors, as described in section 6.4. Additionally, if a firm was mentioned only on the last day of the sample period and therefore still included in the analysis, this may bias the results because the reaction for this event would not be considered in the analysis.

Another primary limitation is that of the proxies used to measure various factors which cannot be measured directly, such as investor sophistication, information asymmetry, and earnings quality. This is a common limitation in financial research, but is mitigated by the use of commonly used proxies which have generally been accepted in prior research. However, there are alternatively acceptable proxies for some of these factors which could be considered for future research.

6.4 Future Research

Future research should address the limitations mentioned in section 6.3 by examining information activism in ordinary economic conditions rather than the extreme bull and bear economic conditions employed in this study. This choice of time periods was important for this study in order to assess the effect of information activism on investor behavior during divergent market conditions. However, a study during normal economic conditions would advance research

knowledge regarding the relationship between information activism and investor behavior.

Similarly, alternate proxies for factors such as investor sophistication, information asymmetry, and earnings quality could be employed in future research to examine these moderating effects.

Furthermore, when examining the relationships in the univariate analysis, to determine high versus low investor sophistication, information asymmetry, and earnings quality, the sample was partitioned by splitting it at the median. This analysis should be examined further by looking at different partition categories such as a comparison between the top tercile and bottom the two terciles for these moderating variables, rather than just partitioning the sample at the median. This would allow future studies to examine alternate metric categories for these moderators.

Additionally, future research could utilize alternate methodologies to examine the effect of information activism on investor behavior. These include survey methods as well as experimental design work. These methods would allow other factors to be considered or specific variables to be directly manipulated by using subjects acting as investors and making investment decisions while exposing them to information activism. Future studies should provide avenues to investigate the unexpected results observed in this study, particularly with regard to investor sophistication. Not only should additional methods to measure investor sophistication be considered, but also alternate theories of how investor sophistication moderates the effect of information activism on investor behavior should be examined. Finally, future research should investigate other relevant theories in behavioral finance in order to consider additional important moderating factors which may

determine the extent to which investors rely on information activism.

FIGURE 1 Basic Model

Information Investor Activism Behavior

FIGURE 2 Research Model

FIGURE 3 Dow Jones Industrial Average

Figure 4 Average Monthly Returns Bull/Bear Periods

Average Monthly Return Month (Median = 0.72%) Bull/Bear Mark Oct07 0.25% Nov07 4.01% Dec07 0.80% Jan08 4.63% Feb08 3.04% Mar08 0.03% Apr08 4.54% May08 1.42% Jun08 10.19% Jul08 0.25% Aug08 1.45% Sep08 6.00% Bear during meltdown ------Oct08 14.06% Bear during meltdown ------Nov08 5.32% Bear during meltdown ------Dec08 0.60% Jan09 8.84% Feb09 11.72% Mar09 7.73% Bull following meltdown ------Apr09 7.35% Bull following meltdown ------May09 4.07% Bull following meltdown ------Jun09 0.63% Jul09 8.58% Aug09 3.54% Sep09 2.27%

Figure 4 (continued) Monthly Return Bear/Bull Periods

FIGURE 5

Summary of Hypotheses & Findings 4

Univariate Multivariate Hypothesis Buy/Sell CAR ABVOL CAR ABVOL H1a: Information activism is associated with abnormal returns. N/A Mixed N/A Mixed N/A Information activism is associated with abnormally high H1b: N/A N/A Supported N/A Unsupported trading volume. Positive information activism is associated with higher Buy Mixed N/A H1c: prices, while negative information activism is associated Unsupported Unsupported with lower prices. Sell Unsupported N/A The effect of information activism on prices and trading Buy Unsupported Unsupported H2: volume will be greater for unsophisticated investors than Unsupported Unsupported for sophisticated investors. Sell Mixed Unsupported The effect of information activism on prices and trading Buy Supported Mixed H3: volume will be greater during bearish market conditions Mixed Mixed than during bullish market conditions. Sell Unsupported Unsupported The effect of information activism on prices and trading Buy Unsupported Mixed H4: volume will be greater for firms with high information Unsupported Mixed asymmetry than for firms with low information asymmetry. Sell Unsupported Unsupported The effect of information activism on prices and trading Buy Unsupported Supported H5: volume will be greater for firms with low earnings quality Mixed Unsupported than for firms with high earnings quality. Sell Unsupported Mixed

4 Mixed results indicate that one or more of the individual analysis results related to the hypothesis provide support; however, in other cases the particular hypothesis was not supported. In the case of the univariate analysis, it indicates that one or more of the individual sample components ( All Events, Cramer, or Blog) was supported and/or one of the recommendation categories ( buy or sell) was supported. In the case of the multivariate analysis mixed indicates that one or more of the equations examined for the hypothesis provided support.

REFERENCES

Abarbanell, J. S. & Bushee, B. J. (1997, Spring). Fundamental Analysis, Future Earnings, and Stock Prices. Journal of Accounting Research, 85(1), 124.

______. (1998, January). Abnormal Returns to a Fundamental Analysis Strategy. The Accounting Review, 73(1), 1945.

Akerlof, G. A. (1970). The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism. Quarterly Journal of Economics, 84(3), 488–500.

Antweiler. W. & Frank, M. Z. (2004). Is All that Talk Just Noise? The Information Content of Internet Stock Message Boards. Journal of Finance, 59, 1259 – 1294.

Arrow, K.J. (1964). The Role of Securities in the Optimal Allocation of Riskbearing. Review of Economic Studies , 31, 91–96.

Bae, S. C. & Jo, H. (1999, September). The Impact of Information Release on Stock Price Volatility and Trading Volume: The Rights Offering Case. Review of Quantitative Finance and Accounting, 13(2), 153169.

Bamber, L. S. (1986). The Information Content of Annual Earnings Releases: A Trading Volume Approach. Journal of Accounting Research, 24(1), 4056.

______, Barron, O., Stober, T. (1997). Trading Volume and Different Aspects of Disagreement Coincident with Earnings Announcements. The Accounting Review, 72(4), 57597.

Barber, B., & Odean, T. (2008, March). All That Glitters: The Effect of Attention and News on the Buying Behavior of Individual and Institutional Investors. Review of Financial Studies , 21(2), 785818.

Barron, O. E., Harris, D. G., & Stanford, M. (2005, April). Evidence that Investors Trade on Private EventPeriod Information around Earnings Announcements. The Accounting Review, 80(2), 403421.

Bartov, E., Radhakrishnan, S., & Krinsky, I. (2000). Investor Sophistication and Patterns in Stock Returns after Earnings Announcements. The Accounting Review, 75(1), 4363.

Bawden, T. (2008, November). Citigroup Chief Blames Rumors for Stock Slide. http://business.timesonline.co .

Beaver, W. (1968). The Information Content of Annual Earnings Announcements.

Empirical Research in Accounting, Journal of Accounting Research (Supplement ), 6792

Bhardwaj, R. & Brooks, L. (1993). Dual Betas from Bull and Bear Markets: Reversal of the Size Effect. Journal of Financial Research , 16(4), 269 283.

Böhme, R. & Holz, T. (2006). The Effect of Stock Spam on Financial Markets. Working Paper, Dresden University of Technology.

Burrough, B. (2008, August). Bringing Down Bear Stearns. www.vanityfair.com .

Bushee, B. J., Core, J. E., Guay, W., & Hamm, S. J. W. (2010, March). The Role of the Business Press as an Information Intermediary. Journal of Accounting Research, 48(1), 1 – 19.

Busse, J. A. & Green, C. (2002). Market Efficiency in Real Time. Journal of Financial Economics , 65, 415437.

Brown, S., Hillegeist, S. A. & Lo, K. (2004, September). Conference Calls and Information Asymmetry. Journal of Accounting & Economics, 37(3), 343366.

______, Lo, K. & Lys, T. (1999, December). Use of R 2 in Accounting Research: Measuring Changes in Value Relevance over the Last Four Decades. Journal of Accounting and Economics, 28(2), 83115.

Brown, S. & Warner, J. (1980). Measuring Security Price Performance. Journal of Financial Economics, 8, 205258.

______(1985). Using Daily Stock Returns: The Case of Event Studies. Journal of Financial Economics, 14(1), 331.

Campbell, J. (1996). Consumption and the Stock Market: Interpreting International Experience. Swedish Economic Policy Review, 3, 251–299.

Campbell, J. & Cochrane, J. (1999). By Force of Habit: a Consumption Based Explanation of Aggregate Stock Market Behavior. Journal of Political Economy, 107, 205–251.

Celizic, M. (2008, October). Did ‘Mad Money’ host turn into Chicken Little? www.msnbc.com .

Chang, J. (1999). The Decline in the Value Relevance of Earnings and Book Values. Working paper, University of Pennsylvania.

Cheng, J. (2007, April). Blogosphere Growth Slowing Considerably.

http://artechnica.com.

Cooper, I. (2008, January). The Jim Cramer Effect. www.wealthdaily.com .

Cready, W. & Hurtt, D. (2002, October). Assessing Market Response Using Return and Volume metrics. The Accounting Review , 77(4), 891–909.

Das, S. R. & Chen, M. Y. (2007, September). Yahoo! For Amazon: Sentiment Extraction from Small Talk on the Web. Management Science , 53(9), 1375 – 1388.

Dechow, P. & Dichev, I. (2002). The Quality of Accruals and Earnings: The Role of Accrual Estimation Errors. The Accounting Review 11 (Supplement), 3559.

Del Guercia, D. & Hawkins, J. (1999). The Motivation and Impact of Pension Fund Activism. Journal of Financial Economics , 52, 293340.

Del Guerico, D., Wallis, L., & Woidtke, T. (2008, October). Do Board Members Pay Attention when Institutional Investors “Just Vote No”? CEO and Director Turnover Associated with Shareholder Activism. Journal of Financial Economics, 90(1), 84103.

Dewally, M. (2003, July). Internet Investment Advice: Investing with a Rock of Salt. Financial Analysts Journal, 59(4), 65 – 77.

Easley, D. & O’hara, M. (2004). Information and the Cost of Capital. Journal of Finance, 59, 1553–83.

Engelberg, J., Sasseville, C., & Williams, J. (2009, July). Market Madness? The Case of Mad Money. Working paper, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=870498.

Fabozzi, F. J. & Francis, J. C. (1977). Stability Tests for Alphas and Betas over Bull and Bear Market Conditions. The Journal of Finance , 32(4), 10931099.

Fama, E. (1970). Efficient Capital Markets: a Review of Theory and Empirical work. Journal of Finance, 25, 383417.

______, & French, K. R. (1993). Common Risk Factors in the Returns on Stocks and Bonds. Journal of Financial Economics, 33(1), 3 – 56.

Financial Accounting Standards Board, (1978). Statement of Financial Accounting Concepts No. 1: Objectives of Financial Reporting by Business Enterprises. Stanford, Conn: FASB.

Ferreira, E. J., Sinha, A. & Varble, D. (2008). LongRun Performance Following Quality

Management Certification. Review of Quantitative Finance and Accounting, 30, 93109.

Ferri, F. & Sandino, T. (2009, March). The Impact of Shareholder Activism on Financial Reporting and Compensation: The Case of Employee Stock Options Expensing. The Accounting Review , 84(2), 433466.

Fombrun, C. & Shanley, M. (1990). What’s in a Name? Reputation Building and Corporate Strategy. Academy of Management Journal, 33(2), 233258.

Fotak, V. (2008). The Impact of Blog Recommendations on Security Prices and Trading Volumes. Working Paper, Oklahoma University.

Francis, J. & Schipper, K. (1999, September). Have Financial Statements Lost Their Relevance? Journal of Accounting Research , 37(2), 319352.

______, Olsson, P. & Weber, J. (2006). Earnings Quality. Foundations and Trends in Accounting, 1(4), 259328.

______, LaFond, R., Olsson, P. M., & Schipper, K. (2004). Costs of Equity and Earnings Attributes. The Accounting Review , 79(4), 9671010.

Frieder, L. & Zittrain, J. (2007, March). Spam Works: Evidence from Stock Touts and Corresponding Market Activity. Working Paper, Purdue University.

Friedman, M. & Savage, L. P. (1948). The Utility Analysis of Choices involving Risk. Journal of Political Economy , 56, 279304.

French, K., & Roll, R. (1986). Stock Return variances: The Arrival of Information and the Reaction of Traders. Journal of Financial Economics, 17, 526.

Gillan, S. L. & Starks, L. T. (2000, August). Corporate Governance Proposals and Shareholder Activism: the Role of Institutional Investors. Journal of Financial Economics, 57(2), 275305.

Givoly, D., Hayn, C.K., & Katz, S. P. (2010, January). Does Public Ownership of Equity Improve Earnings Quality? The Accounting Review, 85(1), 195–225.

Glosten, L. & Milgrom, P. (1985). Bid, Ask, and Transaction Prices in a Specialist Market with Heterogeneously Informed Traders. Journal of Financial Economics , 14, 71 100.

Gomstyn, A. (2008, March). Should You Stay Away from Jim Cramer? www.abcnews.go.com

Graham, B. & Dodd, D. (2009). Security Analysis: Principles and Techniques. New York: McGrawHill.

Grinblatt, M. & Keloharju, M. (2000). The Investment Behavior and Performance of Various Investor Types: a Study of Finland’s Unique Data Set. Journal of Financial Economics, 55, 4367.

Guo, R., Lev, B., & Zhou, N. (2004, May). Competitive Costs of Disclosure by Biotech IPOs. Journal of Accounting Research , 42, 319355.

Hand, J. R. M. (1990). A Test of the External Function Fixation Hypothesis. The Accounting Review, 65, 740763.

Hanke, M. & Hauser, F. (2008). On the effects of stock spam emails. Journal of Financial Markets, 11, 57 83.

Hamilton, J. (2008). Market Crisis Focus on Short Selling: SEC Adopts Rules to Curb Abusive Practices. www.cch.com.

Healy, P. M. & Palepu, K. G. (2001). Information Asymmetry, Corporate Disclosure, and the Capital Markets: A Review of Empirical Disclosure Literature. Journal of Accounting and Economics, 31, 405440.

Heflin, F. L., Shaw, K. W. & Wild, J. J. (2005, winter). Disclosure Policy and Market Liquidity: Impact of Depth Quotes and Order Sizes. Contemporary Accounting Research, 22(4), 829865.

Hempel, J. (2008, April). CNBC Feels Your Pain. http://money.cnn.com .

Hirshleifer, D. A., Myers, J. N., Myers, L. A. & Teoh, S. H. (2008, November). Do Individual Investors Cause PostEarnings Announcement Drift? Direct Evidence from Personal Trades. The Accounting Review , 83(6), 15211550.

Hollenbeck, C. R. & Zinkhan, G. M. (2006). Consumer Activism on the Internet: The Role of Antibrand Communities. Advances in Consumer Research, 33(1), 479485.

Joe, J. R. (2003, March). Why Press Coverage of a Client Influences the Audit Opinion. Journal of Accounting Research, 41(1), 109133.

______, Henock, L. & Robinson, D. (2009). Managers’ and Investors’ Responses to Media Exposure of Board Ineffectiveness. Journal of Financial &Quantitative Analysis, 44 (3), 579605.

Kahan, R. & Kellner, D. (2004). New Media and Internet Activism: from the ‘Battle of

Seattle’ to Blogging. New Media & Society, 6(1), 8795.

Kahneman, D. & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica , 47(2), 263292.

Kaplanski, G. (2004, December). Traditional Beta, Downside Risk Beta and Market Risk Premiums. The Quarterly Review of Economics and Finance, 44(5), 636653.

Karpoff, J. M. (1986). A Theory of Trading Volume. Journal of Finance , 41(5), 1069 1087.

______(1987). The Relation between Price Changes and Trading Volume: A Survey. The Journal of Financial and Quantitative Analysis , 22(1), 109126.

______(2001, August). The Impact of Shareholder Activism on Target Companies: A Survey of Empirical Findings. Working Paper, University of Washington.

______, Malatesta, P. H. & Walkling, R. A. (1996, November). Corporate Governance and Shareholder Initiatives: Empirical Evidence. Journal of Financial Economics, 42(3), 365395.

Kothari, S.P. (2001, September). Capital Markets Research in Accounting. Journal of Accounting and Economics, 31(1), 105–231.

Lee, C.M.C. (1992). Earnings News and Small Traders. Journal of Accounting & Economics , 15(2/3), 265302.

Lev, B. & Thiagarajan, S. R. (1993, Autumn). Fundamental Information Analysis. Journal of Accounting Research , 31(2), 190215.

Lev, B. & Zarowin, P. (1999, September). The Boundaries of Financial Reporting and How to Extend Them. Journal of Accounting Research , 37(2), 353385.

Lloyd, R. (2009, March). Jon Steward Chastises Jim Cramer on ‘The Daily Show.’ Los Angeles Times. http://articles.latimes.com/2009/mar/14/entertainment/et cramer14.

Lo, A.W., and Wang, J. (2000). Trading Volume: Definitions, Data Analysis, and Implications of Portfolio Theory. Review of Financial Studies, 13(2), 257300.

Logsdon, J. M., & Van Buren III, H. J. (2009, July). Beyond the Proxy Vote: Dialogues Between Shareholder Activists and Corporations. Journal of Business Ethics, 87, 353 365.

Lyon, J. D., Barber, B. M., & Tsai, C. (1999, February). Improved Methods for Tests of

LongRun Abnormal Stock Returns. The Journal of Finance, 59(1), 165201.

Martin, J. D., Kensinger, J. W. & Gillan, S. L. (2000, January). Value Creation and Corporate Diversification: The Case of Sears, Roebuck & Co. Journal of Financial Economics, 55(1), 103137.

McIntyre, D. A. & Allen, A. C. (2009, January). TwentyFive Best Financial Blogs. www.Time.com .

Morse, D. (1980, December). Asymmetrical Information in Securities Markets and Trading Volume. Journal of Financial and Quantitative Analysis, 15(5), 11291148.

Morse, D. (1981). Price and Trading Volume Reaction Surrounding Earnings Announcements: A Closer Examination. Journal of Accounting Research, 19(2), 374 383.

Mullins, L. (2008, October). Richard Fuld: Why Lehman Brothers Went Under. www.usnews.com .

Nah, S., Veenstra, A. S., & Shah, D. V. (2006, October). The Internet and AntiWar Activism: A Case Study of Information, Expression, and Action. Journal of Computer Mediated Communication, 12(1), 230247.

Nelson, K. K., Price, R. A., & Rountree, B. R. (2009, March). Why do Investors Pay Attention to Stock Spam? Working paper, Rice University.

Neumann, J. J. & Kenny, P. M. (2007, December). Does Mad Money Make the Market Go Mad? The Quarterly Review of Economics and Finance, 47(5), 602615.

O’Hara, M. (2003, August). Presidential Address: Liquidity and Price Discovery. The Journal of Finance, 58(4), 13351354.

Peress, J. (2008, March). Media Coverage and Investors’ Attention to Earnings Announcements. Working paper, INSEAD and London School of Economics.

Pratt, J.W. (1964). Risk Aversion in the Small and in the Large. Econometrica, 32, 122– 136.

Pownall, G. & Simko, P. J. (2005, July). The Information Intermediary Role of Short Sellers. The Accounting Review, 80(3), 941966.

Prevost, A. K. & Ramesh P. R. (2000, April). Of What Value are Shareholder Proposals Sponsored by Public Pension Funds? The Journal of Business , 73(2) 177204.

Rosenberg, J. V. & Engle, R. F. (2002). Empirical Pricing Kernels. Journal of Financial Economics, 64, 341 – 372.

Spence, A. M. (1973) Job Market Signaling. Quarterly Journal of Economics , 87, 355– 374.

Stelter, B. & Arango, T. (2009, March). CNBC Thrives as Hosts Deliver News with Attitude. www.nytimes.com .

Stickel, S.E. & Verrecchia, R. (1994). Evidence that Trading Volume Sustains Stock Price Changes. Financial Analysts Journal, 50(6), 5767.

______(1995). The Anatomy of the Performance of Buy and Sell Recommendations. Financial Analysts Journal, 51, 2539.

Sussman, M. (2009, October). State of the Blogosphere 2009. www.technorati.com .

Sutton, S., Arnold, V., Bedard, J., & Phillips, J. (2009, January). Where do Investors Prefer to Find Nonfinancial Information? Working Paper , University of Central Florida.

Tumarkin, R. & Whitelaw, R. (2001). News or Noise? Internet Postings and Stock Prices. Financial Analyst Journal, 57(3), 4151.

Tversky, A. & Kahneman, D. (1992). Advances in Prospect Theory: Cumulative Representation of Uncertainty. Journal of Risk and Uncertainty, 5(4), 297323.

Wahal, S. (1996, March). Pension Fund Activism and Firm Performance. Journal of Financial and Quantitative Analysis, 31(1) 123.

Walther, B. (1997). Investor Sophistication and Market Earnings Expectations. Journal of Accounting Research , 35(2) 157179.

Welker, M. (1995, spring). Disclosure Policy, Information Asymmetry, and Liquidity in Equity Markets. Contemporary Accounting Research, 11(2), 801827.

Winn, P. (2009, August). State of the Blogosphere 2008. www.technorati.com .

Woellert, L. & Onaran, Y. (2008, October). Fuld Blames Lehman’s Fall on Rumors, ‘Storm of Fear. www.bloomberg.com .

Womack, K. (1996). Do Brokerage Analysts’ Recommendations have Investment Value? The Journal of Finance, 51, 137167.

100

101

TABLE 1

Event Data Initial Sample Sources

Cramer Blog All

Bull Bear Total Cramer Bull Bear Total Blog Number Percent Number Percent Number Percent Number Percent Number Percent Number Percent Number Percent

Buy 692 66.54 607 58.14 1,299 62.33 155 67.69 298 65.89 453 78.24 1,752 65.79 Sell 348 33.46 437 41.86 785 37.67 74 32.31 52 34.11 126 21.76 911 34.21

Total 1,040 100.00 1,044 100.00 2,084 100.00 229 100.00 350 100.00 579 100.00 2,663 100.00

102

TABLE 2 Descriptive Statistics – Initial Sample

Cramer Blog Bull Bear Bull Bear Observations Total Events 1,040 1,044 229 350 Unique Firms 479 488 195 289 Firms One Event 270 269 165 248 Firms Multiple Events 209 219 30 41 Events with Available Data 844 819 175 269 CAR Mean 0.001034 0.012123 0.004028 0.010306 Std. Dev 0.055692 0.070840 0.048875 0.070839 Min 0.293358 0.263565 0.216690 0.233765 Q1 0.024270 0.025715 0.029804 0.024931 Median 0.002971 0.014224 0.008939 0.004291 Q3 0.023395 0.049271 0.018417 0.042992 Max 0.294900 0.290139 0.228308 0.369096 ABVOL Mean 19.340617 21.829325 109.39 57.29 Std. Dev 29.557910 23.361053 578.77 235.00 Min 0 0 0.33 0.08 Q1 4.186960 5.400080 3.30 4.89 Median 9.638221 12.349984 9.49 12.29 Q3 21.195241 25.866064 24.07 23.01 Max 193.884430 171.932458 5,330.48 2,700.02 Market Capitalization Mean $ 23,654,742.57 $ 23,883,325.34 $ 15,902,629.33 $ 28,808,145.56 Std. Dev. 40,499,979.51 42,582,397.20 27,559,484.97 58,003,400.03 Min 0 0 0 10,350.00 Q1 1,796,981.99 2,034,983.88 1,251,855.05 614,332.41 Median 6,524,366.38 7,592,809.49 4,974,281.12 4,679,326.79 Q3 23,546,525.07 24,743,869.71 14,995,278.56 26,176,402.76 Max 336,524,996.00 403,522,106.00 145,218,284.00 391,004,540.00 CAR is the cumulative abnormal return over the information activism event window (0,+1 ). ABVOL is abnormal trading volume calculated as the difference between the daily share turnover and median share turnover. Market Capitalization is the closing prices multiplied by the shares outstanding.

103

TABLE 3 Final Sample Selection Detail – Univariate Analysis

Panel A – Cramer Market Condition Bull Bear Total Total Event Observations 1,040 1,044 2,084 Less : Observations with Confounding Events 164 155 319 Clean Sample 876 889 1,765 Less : Events with Missing Data 32 70 102 Total Sample Cramer 844 819 1,663

Panel B – Blog Market Condition Bull Bear Total Total Event Observations 229 350 579 Less : Observations with Confounding Events 40 39 79 Clean Sample 189 311 500 Less : Events with Missing Data 14 42 56 Total Sample Blog 175 269 444

Total Sample Events a 1,019 1,088 2,107 a Note: Some sample event observations (N) in separate univariate analyses will vary due to different data requirements and the resulting missing data.

104

Table 4 Univariate Analysis – CAR All Events Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Overall (H1a) 2,107 0.00619 4.43 <.0001 0.06408 0.29163 0.02542 0.00260 0.03760 0.36910

Sentiment (H1c) Buy 1,380 0.00975 6.13 <.0001 0.05891 0.23795 0.02083 0.00441 0.03784 0.36910 Sell 727 0.00052 0.19 0.8456 0.07241 0.29163 0.03811 0.00216 0.03695 0.29014 Difference 0.01027 3.50 0.0005

Investor Sophistication (H2) Low (buy) 817 0.01121 5.55 <.0001 0.05761 0.20182 0.01918 0.00525 0.03558 0.36910 High (buy) 515 0.00771 3.14 0.0018 0.05563 0.17311 0.02253 0.00418 0.04000 0.24008 Difference 0.0035 1.09 0.2748 Low (sell) 355 0.00531 1.25 0.2125 0.08015 0.32563 0.03538 0.00138 0.03925 0.32668 High (sell) 348 0.00500 1.32 0.1862 0.07040 0.26424 0.04081 0.00320 0.03315 0.29014 Difference 0.01031 1.81 0.0702

Market Condition (H3) Bear (buy) 706 0.01730 6.93 <.0001 0.06628 0.21905 0.01956 0.01585 0.05115 0.36910 Bull (buy) 674 0.00079 0.46 0.6465 0.04443 0.17311 0.02219 0.00296 0.02010 0.28661 Difference 0.01651 5.45 <.0001 Bear (sell) 382 0.00018 0.05 0.9634 0.07828 0.27913 0.04154 0.00200 0.04354 0.30622 Bull (sell) 345 0.00147 0.38 0.7071 0.07280 0.32563 0.03655 0.00668 0.02800 0.32668 Difference 0.00129 0.23 0.8187 Information Asymmetry (H4) High (buy) 418 0.00814 2.82 0.0051 0.05901 0.21905 0.02329 0.00384 0.03738 0.23128 Low (buy) 941 0.00998 5.39 <.0001 0.05675 0.20182 0.20127 0.00463 0.03811 0.36910 Difference -0.00184 -0.55 0.5857 High (sell) 314 0.00065 0.14 0.8926 0.08501 0.27913 0.04153 0.00092 0.04115 0.32668 Low (sell) 398 0.00044 0.13 0.8973 0.06728 0.32563 0.03488 0.00235 0.03519 0.29490 Difference -0.00109 -0.18 0.8534

105

Table 4 (continued)

Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Earnings Quality (H5) Low (buy) 417 0.00806 3.20 0.0015 0.05136 0.18856 0.01885 0.00494 0.03440 0.18379 High (buy) 490 0.00835 3.71 0.0002 0.04981 0.15993 0.01934 0.00445 0.03444 0.21624 Difference -0.00029 -0.09 0.9314 Low (sell) 207 0.00677 1.33 0.1858 0.07332 0.32563 0.04007 0.00922 0.03252 0.23237 High (sell) 190 0.00716 1.45 0.1496 0.06821 0.27913 0.02680 0.00101 0.03900 0.32668 Difference -0.01393 -1.95 0.0514

CAR is the cumulative abnormal return over the information activism event window (0,+1). Sentiment indicates whether the information activism event contained a buy or sell recommendation. Investor Sophistication is measured using the percentage of institutional holdings. Market Condition indicates whether the information activism event occurred during the Bull or Bear period. Information Asymmetry is measured using the mean bidask spread. Earnings Quality is measured using the Dechow Dichev (2002) model (see equation 9). Ttests are used to analyze the difference in means.

106

Table 5 Univariate Analysis – CAR Cramer Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Overall (H1a) 1,663 0.00650 4.15 <.0001 0.06383 0.29336 0.02478 0.00359 0.03837 0.29490

Sentiment (H1c) Buy 1,025 0.01119 6.29 <.0001 0.05693 0.20547 0.01888 0.00628 0.03945 0.29490 Sell 638 0.00105 0.36 0.7170 0.07300 0.29336 0.03840 0.00222 0.03512 0.29014 Difference 0.01224 3.61 0.0003

Investor Sophistication (H2) Low (buy) 615 0.01213 5.44 <.0001 0.05526 0.20269 0.01665 0.00611 0.03723 0.28974 High (buy) 383 0.00926 3.30 0.0011 0.05498 0.16814 0.02166 0.00699 0.04050 0.18314 Difference 0.00287 0.80 0.4241 Low (sell) 302 0.00535 1.18 0.2396 0.07897 0.32125 0.03349 0.00065 0.03826 0.31697 High (sell) 314 0.00465 1.15 0.2531 0.07195 0.26424 0.04154 0.00292 0.03517 0.29014 Difference 0.01000 1.64 0.1012

Market Condition (H3) Bear (buy) 474 0.02069 7.02 <.0001 0.06420 0.20269 0.01442 0.19752 0.05391 0.22134 Bull (buy) 551 0.00235 1.22 0.2221 0.04512 0.16024 0.02095 0.00165 0.02161 0.28973 Difference 0.01834 5.35 <.0001 Bear (sell) 345 0.00047 0.11 0.9109 0.07768 0.27546 0.04002 0.00181 0.04360 0.30259 Bull (sell) 293 0.00195 0.46 0.6478 0.07300 0.32125 0.03526 0.00775 0.02829 0.31700 Difference 0.00148 0.25 0.8052 Information Asymmetry (H4) High (buy) 288 0.01236 3.63 0.0003 0.05783 0.15369 0.01987 0.00701 0.04154 0.22134 Low (buy) 721 0.01076 5.26 <.0001 0.05490 0.20269 0.01769 0.00611 0.03963 0.28973 Difference 0.0016 0.41 0.6805 High (sell) 269 0.00137 0.27 0.7861 0.08277 0.27546 0.04035 0.00138 0.04077 0.31698 Low (sell) 357 0.00009 0.02 0.9809 0.06952 0.32125 0.03502 0.00251 0.03524 0.29102 Difference -0.00146 -0.23 0.8154

107

Table 5 (continued)

Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Earnings Quality (H5) Low (buy) 309 0.00859 3.13 0.0019 0.04826 0.16814 0.01769 0.00508 0.03282 0.18672 High (buy) 382 0.01175 4.58 <.0001 0.05010 0.16024 0.01512 0.00962 0.03820 0.21624 Difference -0.00316 -0.84 0.4011 Low (sell) 183 0.00581 1.09 0.2780 0.07219 0.32125 0.04007 0.00249 0.03608 0.15970 High (sell) 176 0.00515 1.05 0.2958 0.06510 0.27546 0.02678 0.00069 0.03832 0.31698 Difference -0.01096 -1.51 0.1325 CAR is the cumulative abnormal return over the information activism event window (0,+1). Sentiment indicates whether the information activism event contained a buy or sell recommendation. Investor Sophistication is measured using the percentage of institutional holdings. Market Condition indicates whether the information activism event occurred during the Bull or Bear period. Information Asymmetry is measured using the mean bidask spread. Earnings Quality is measured using the Dechow Dichev (2002) model (see equation 9). Ttests are used to analyze the difference in means.

108

Table 6 Univariate Analysis – CAR Blog Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Overall (H1a) 444 0.00466 1.55 0.1226 0.06343 0.23376 0.02740 0.00192 0.03341 0.36910 Sentiment (H1c) Buy 355 0.00500 1.52 0.1301 0.06203 0.23377 0.02586 0.00206 0.03319 0.36910 Sell 89 0.00330 0.45 0.6531 0.06908 0.21669 0.03664 0.00066 0.03665 0.22831 Difference 0.0017 0.23 0.8221 Investor Sophistication (H2) Low (buy) 202 0.00780 1.79 0.0755 0.06203 0.18592 0.02586 0.00184 0.03375 0.36910 High (buy) 132 0.00248 0.51 0.6115 0.05597 0.16784 0.02544 0.00115 0.03053 0.22848 Difference 0.00532 0.80 0.4268 Low (sell) 53 0.00299 0.24 0.8078 0.08886 0.23212 0.04005 0.00121 0.04040 0.25410 High (sell) 34 0.00825 0.90 0.3740 0.05338 0.14784 0.04153 0.00714 0.01950 0.11231 Difference 0.01124 0.74 0.4636 Market Condition (H3) Bear (buy) 232 0.00932 2.09 0.0374 0.06781 0.21666 0.02459 0.00388 0.04064 0.36910 Bull (buy) 123 0.00593 1.66 0.0994 0.03963 0.16784 0.02802 0.00855 0.01429 0.10851 Difference 0.01525 2.67 0.0079 Bear (sell) 37 0.00132 0.09 0.9259 0.08592 0.14784 0.06938 0.00169 0.04043 0.23481 Bull (sell) 52 0.00186 0.19 0.8530 0.07205 0.23212 0.03917 0.00467 0.02080 0.25409 Difference -0.00318 -0.19 0.8501 Information Asymmetry (H4) High (buy) 130 0.00200 0.38 0.7057 0.06025 0.21667 0.03300 0.00440 0.02745 0.22987 Low (buy) 220 0.00693 1.72 0.0864 0.05966 0.14933 0.02354 0.00151 0.03341 0.36910 Difference -0.00893 -1.35 0.1786 High (sell) 45 0.00096 0.06 0.9491 0.09994 0.23212 0.06733 0.00915 0.04648 0.25410 Low (sell) 41 0.00376 0.58 0.5645 0.04146 0.07842 0.01975 0.00422 0.02459 0.11231 Difference -0.00280 -0.17 0.8634

109

Table 6 (continued)

Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Earnings Quality (H5) Low (buy) 108 0.00541 0.98 0.3276 0.05716 0.18592 0.03027 0.00398 0.04210 0.13813 High (buy) 108 0.00398 0.87 0.3850 0.04742 0.14304 0.02795 0.00438 0.02180 0.15013 Difference 0.00939 1.31 0.1904 Low (sell) 24 0.01517 0.91 0.3738 0.08193 0.23212 0.04155 0.02420 0.00833 0.23482 High (sell) 14 0.03224 1.25 0.2329 0.09642 0.07842 0.03200 0.00500 0.06554 0.25409 Difference -0.04741 -1.61 0.1156 CAR is the cumulative abnormal return over the information activism event window (0,+1). Sentiment indicates whether the information activism event contained a buy or sell recommendation. Investor Sophistication is measured using the percentage of institutional holdings. Market Condition indicates whether the information activism event occurred during the Bull or Bear period. Information Asymmetry is measured using the mean bidask spread. Earnings Quality is measured using the Dechow Dichev (2002) model (see equation 9). Ttests are used to analyze the difference in means.

110

Table 7 Univariate Analysis – ABVOL All Events Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Overall (H1b) 2,107 22.29 25.72 <.0001 39.74 0.00 4.57 10.86 22.67 493.24

Investor Sophistication (H2) Low (buy) 817 18.82 16.30 <.0001 32.95 0.00 3.66 9.88 19.80 311.91 High (buy) 515 20.59 13.93 <.0001 33.52 0.00 4.93 10.80 23.14 359.25 Difference -1.77 -0.95 0.3438 Low (sell) 355 30.74 9.17 <.0001 63.17 0.00 4.78 11.61 25.26 620.99 High (sell) 348 25.54 13.61 <.0001 35.02 0.00 6.43 12.56 28.11 214.77 Difference 5.20 1.35 0.1769 Market Condition (H3) Bear (buy) 706 20.67 17.60 <.0001 31.16 0.00 4.88 11.87 22.35 359.25 Bull (buy) 674 18.04 13.60 <.0001 34.38 0.00 3.61 8.65 19.29 345.10 Difference 2.63 1.49 0.1376 Bear (sell) 382 26.06 13.57 <.0001 37.53 0.00 5.96 13.88 28.95 351.07 Bull (sell) 345 29.38 8.82 <.0001 61.85 0.00 4.57 10.54 23.59 620.99 Difference -3.32 -0.86 0.3883 Information Asymmetry (H4) High (buy) 418 20.24 12.40 <.0001 33.32 0.00 4.55 10.67 23.50 359.25 Low (buy) 941 19.13 17.91 <.0001 32.71 0.00 4.19 9.96 20.80 345.10 Difference 1.11 0.57 0.5663 High (sell) 314 26.97 9.40 <.0001 50.84 0.00 5.94 11.59 25.19 620.99 Low (sell) 398 28.08 11.19 <.0001 50.07 0.00 4.85 12.12 28.57 526.57 Difference -1.11 -0.29 0.7694

111

Table 7 (continued)

Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Earnings Quality (H5) Low (buy) 417 17.67 14.09 <.0001 25.59 0.00 5.22 10.89 22.89 345.10 High (buy) 490 13.08 12.69 <.0001 22.80 0.00 3.14 7.11 15.02 302.98 Difference 4.59 2.83 0.0048 Low (sell) 207 29.37 7.40 <.0001 57.10 0.35 6.75 12.54 28.78 620.99 High (sell) 190 17.76 6.94 <.0001 35.29 0.24 4.14 8.63 16.66 371.58 Difference 11.61 2.46 0.0144

ABVOL is the cumulative abnormal trading volume over the information activism event window (0,+1). Sentiment indicates whether the information activism event contained a buy or sell recommendation. Investor Sophistication is measured using the percentage of institutional holdings. Market Condition indicates whether the information activism event occurred during the Bull or Bear period. Information Asymmetry is measured using the mean bidask spread. Earnings Quality is measured using the Dechow Dichev (2002) model (see equation 9). Ttests are used to analyze the difference in means.

112

Table 8 Univariate Analysis – ABVOL Cramer Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Overall (H1b) 1,663 20.57 29.90 <.0001 28.05 0.00 4.76 11.03 23.57 193.88

Investor Sophistication (H2) Low (buy) 615 17.03 18.79 <.0001 22.47 0.00 3.98 9.43 19.78 163.30 High (buy) 383 19.04 17.22 <.0001 21.64 0.00 4.88 11.23 23.83 120.59 Difference -2.01 -1.39 0.1638 Low (sell) 302 26.70 10.84 <.0001 42.79 0.00 4.81 12.06 25.48 267.35 High (sell) 314 25.09 13.55 <.0001 32.82 0.00 6.52 12.92 28.37 204.19 Difference 1.61 0.52 0.6014 Market Condition (H3) Bear (buy) 474 19.97 19.10 <.0001 22.76 0.27 5.17 12.03 23.76 121.09 Bull (buy) 551 16.05 17.54 <.0001 21.47 0.00 3.70 8.66 20.80 163.30 Difference 3.92 2.84 0.0047 Bear (sell) 345 25.16 14.19 <.0001 32.93 0.00 6.43 13.99 29.95 210.67 Bull (sell) 293 25.70 10.38 <.0001 42.36 0.00 4.81 10.95 23.59 267.35 Difference -0.54 -0.18 0.8603 Information Asymmetry (H4) High (buy) 288 20.73 14.32 <.0001 24.56 0.48 4.99 12.29 25.23 163.30 Low (buy) 721 16.78 21.57 <.0001 20.88 0.00 4.23 9.55 20.30 121.09 Difference 3.95 2.41 0.0165 High (sell) 269 26.33 10.91 <.0001 39.59 0.75 6.59 12.93 26.26 267.35 Low (sell) 357 24.45 13.11 <.0001 35.24 0.00 4.78 11.63 27.34 220.12 Difference 1.88 0.62 0.5388

113

Table 8 (continued)

Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Earnings Quality (H5) Low (buy) 309 16.57 17.14 <.0001 17.00 0.00 4.89 10.67 23.26 97.40 High (buy) 382 12.42 13.49 <.0001 18.00 0.20 3.17 6.68 15.02 145.49 Difference 4.15 3.09 0.0021 Low (sell) 183 28.15 9.56 <.0001 39.83 0.50 7.14 13.06 29.32 267.35 High (sell) 176 17.24 7.82 <.0001 29.25 0.24 4.09 8.26 16.38 220.12 Difference 10.91 2.97 0.0032

114

Table 9 Univariate Analysis – ABVOL Blog Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Overall (H1b) 444 41.40 5.67 <.0001 153.71 0.00 3.85 10.94 22.08 1,879.67

Investor Sophistication (H2) Low (buy) 202 33.09 5.15 <.0001 91.32 0.00 3.49 12.31 22.06 727.00 High (buy) 132 30.68 2.71 0.0077 130.27 0.00 4.82 9.87 18.48 1,425.78 Difference 2.41 0.18 0.8537 Low (sell) 53 243.48 1.99 0.0518 890.59 0.33 5.01 12.53 53.64 4,944.03 High (sell) 34 84.88 1.57 0.1261 315.39 1.04 3.14 12.24 25.53 1,833.94 Difference 158.60 1.19 0.2397 Market Condition (H3) Bear (buy) 232 34.93 4.29 <.0001 124.01 0.00 4.40 11.89 21.69 1,425.78 Bull (buy) 123 23.26 4.82 <.0001 53.50 0.00 2.79 8.78 15.24 415.45 Difference 11.68 1.23 0.2181 Bear (sell) 37 129.99 2.24 0.0312 352.56 0.83 5.19 15.49 33.01 1,833.94 Bull (sell) 52 211.60 1.71 0.0929 890.98 0.33 3.42 11.45 38.08 4,944.03 Difference -81.61 -0.60 0.5517 Information Asymmetry (H4) High (buy) 130 23.47 3.52 0.0006 75.97 0.00 2.85 8.45 14.42 727.00 Low (buy) 220 35.75 4.42 <.0001 120.00 0.00 4.60 12.38 23.18 1,425.78 Difference -12.28 -1.17 0.2422 High (sell) 45 260.38 1.81 0.0770 964.60 0.33 3.51 9.59 32.63 4,944.03 Low (sell) 41 99.58 2.15 0.0376 296.49 0.60 6.89 21.96 49.82 1,833.94 Difference 160.8 1.06 0.2920

115

Table 9 (continued)

Hypothesis N Mean t-statistic p-value Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum Earnings Quality (H5) Low (buy) 108 16.18 8.48 <.0001 19.84 0.00 5.69 11.35 20.27 172.58 High (buy) 108 12.11 8.09 <.0001 15.55 0.03 2.88 8.62 14.94 122.19 Difference 4.07 1.68 0.0945 Low (sell) 24 12.89 5.18 <.0001 12.19 0.35 2.99 8.72 22.89 37.36 High (sell) 14 13.42 5.64 <.0001 8.90 1.70 6.35 10.90 21.00 28.03 Difference -0.53 -0.14 0.8882 ABVOL is the cumulative abnormal trading volume over the information activism event window (0,+1). Sentiment indicates whether the information activism event contained a buy or sell recommendation. Investor Sophistication is measured using the percentage of institutional holdings. Market Condition indicates whether the information activism event occurred during the Bull or Bear period. Information Asymmetry is measured using the mean bidask spread. Earnings Quality is measured using the Dechow Dichev (2002) model (see equation 9). Ttests are used to analyze the difference in means.

116

TABLE 10 Descriptive Statistics for Multivariate Analysis Variables

Variable N Mean Std Dev Minimum Q1 (25%) Median Q3 (75%) Maximum

CAR 1186 0.070211 0.438130 1.373550 0.292420 0.078425 0.160720 1.231880 ABVOL 1186 0.672004 1.338644 0.028690 0.165600 0.305475 0.605530 10.557690 BEAR 1298 0.531587 0.499194 0 0 0 1 1 PRG# 1298 1.607088 1.987163 0 0 0 1 16 BLG# 1298 0.451464 0.684357 0 0 0 1 9 PRG%B 1298 0.371255 0.459603 0 0 0 1 1 BLG%B 1298 0.288153 0.449189 0 0 0 1 1 SOPH 1157 0.710700 0.286928 0.026690 0.570720 0.791390 0.912690 1.278010 INS 1158 0.001604 0.001269 0.000488 0.000956 0.001276 0.001729 0.057878 EQ 722 0.031502 0.031787 0.002590 0.011130 0.019400 0.040700 0.160770 PRESS 1298 0.104777 0.306388 0 0 0 0 1 SIZE 1226 8.237916 1.822144 1.505188 6.943933 8.215766 9.591028 12.937411 EXCHG1 1299 0.625866 0.484085 0 0 1.000000 1.000000 1.000000 EXCHG2 1299 0.008468 0.091667 0 0 0 0 1.000000 EXCHG3 1299 0.285604 0.451876 0 0 0 1.000000 1.000000 CAR = monthly cumulative abnormal return over the 3month Bull/Bear period ABVOL = monthly abnormal trading volume in 000’s over the 3month Bull/Bear period BEAR = dummy variable for market condition set to (1) for Bear period and (0) for Bull period PRG# = the number of times a firm is mentioned on the investment news program ( Cramer ) during the 3month Bull/Bear period BLG# = the number of times a firm is mentioned on the financial blog ( SeekingAlpha ) during the 3month Bull/Bear period PRG%B = the percentage of buy recommendations to total mentions on the investment program ( Cramer ) during the 3month Bull/Bear period BLG%B = the percentage of buy recommendations to total mentions on the financial blog ( SeekingAlpha ) during the 3month Bull/Bear period SOPH = investor sophistication measured as the percentage of institutional holdings INS = Information asymmetry measured as bidask spread

117

Table 10 (continued) EQ = earnings quality measured using the Dechow Dichev (2002) model (see equation 9) PRESS = dummy variable for press set to (1) if the firm had significant news, earnings announcement, or SEC filing during the 3month Bull/Bear period; a control variable SIZE = firm size measured as natural log of the market value of the common stock; a control variable EXCHG = dummy variable scheme for the stock exchange that the firm is traded on; EXCHG1 is set to (1) if traded on NYSE, (0) otherwise; EXCHG2 is set to (1) if traded on AMEX, (0) otherwise; EXCHG3 is set to (1) if traded on NASDAQ, (0) otherwise; if EXCHG1, EXCHG2, and EXCHG3 are all set to (0) then stock is traded on a Regional exchange

118

TABLE 11 Pearson & Spearman Correlation Matrix for Regression Equations Variables CAR ABVOL PRG # BLG# PRG%B BLG%B SOPH BEAR INS EQ PRESS SIZE CAR 1.0000 0.5913 0.03100 0.00351 0.01098 0.00305 0.07157 0.09253 0.08163 0.09576 0.01808 0.01515 (0.0417) (0.2862) (0.9039) (0.7057) (0.9164) (0.0166) (0.0014) (0.0055) (0.0110) (0.5339) (0.6041) ABVOL 0.02777 1.00000 0.09883 0.5386 0.07737 0.09569 0.10418 0.13467 0.02479 0.15341 0.00176 0.08347 (0.3392) (0.0007) (0.0637) (0.0077) (0.0010) (0.0005) (<.0001) (0.3999) (<.0001) (0.9517) (0.0042) PRG# 0.04081 0.02359 1.00000 0.48759 0.53789 0.41936 0.01942 0.06773 0.30344 0.14142 0.32149 0.40034 (0.1601) (0.4171) (<.0001) (<.0001) (<.0001) (0.5093) (0.0147) (<.0001) (0.0001) (<.0001) (<.0001) BLG# 0.01443 0.09168 0.07016 1.00000 0.29686 0.81722 0.08409 0.09365 0.05840 0.11354 0.06275 0.01674 (0.6196) (0.0016) (0.0115) (<.0001) (<.0001) (0.0042 (0.0007) (0.0469) (0.0022) (0.0238) (0.5583) PRG%B 0.00581 0.14506 0.44114 0.20550 1.00000 0.25563 0.01501 0.09802 0.21820 0.14315 0.22649 0.28855 (0.8415) (<.0001) (<.0001) (<.0001) (<.0001) (0.6101) (0.0004) (<.0001) (0.0001) (<.0001) (<.0001) BLG%B 0.00577 0.04652 0.18750 0.70138 0.24872 1.00000 0.07151 0.16449 0.07667 0.05468 0.01950 0.08012 (0.8426) (0.1093) (<.0001) (<.0001) (<.0001) (0.0150) (<.0001) (0.0091) (0.1421) (0.4827) (0.0050) SOPH 0.03485 0.09889 0.03348 0.05130 0.04479 0.05827 1.00000 0.08705 0.01499 0.06322 0.04590 0.05961 (0.2438) (0.0009) (0.2552) (0.0811) (0.1279) (0.0475) (0.0030) (0.6207) (0.0964) (0.1187) (0.0433) BEAR 0.04917 0.02228 0.04966 0.08912 0.09657 0.16429 0.08241 1.00000 0.05332 0.01214 0.05228 0.14005 (0.0905) (0.4434) (0.0737) (0.0013) (0.0005) (<.0001) (0.0050) (0.0697) (0.7446) (0.0597) (<.0001) INS 0.10305 0.01185 0.24110 0.05517 0.20934 0.06140 0.25273 0.06681 1.00000 0.19187 0.23649 0.71914 (0.0005) (0.6875) (<.0001) (0.0605) (<.0001) (0.0367) (<.0001) (0.0230) (<.0001) (<.0001) (<.0001) EQ 0.04627 0.07610 0.07771 0.10190 0.11880 0.07020 0.03103 0.00909 0.26769 1.0000 0.11871 0.31550 (0.2198) (0.0434) (0.0368) (0.0061) (0.0014) (0.0594) (0.4147) (0.8074) (<.0001) (0.0014) (<.0001) PRESS 0.02482 0.04074 0.39056 0.11651 0.23409 0.01771 0.01419 0.05228 0.14755 0.07263 1.00000 0.26128 (0.3932) (0.1609) (<.0001) (<.0001) (<.0001) (0.5239) (0.6298) (0.0597) (<.0001) (0.0511) (<.0001) SIZE 0.03041 0.10157 0.39443 0.08104 0.29825 0.06262 0.00939 0.13760 0.60903 0.31379 0.27034 1.00000 (0.3021) (0.0005) (<.0001) (0.0045) (<.0001) (0.0283) (0.7503) (<.0001) (<.0001) (<.0001) (<.0001) Spearman Correlation coefficients are reported above the diagonal and Pearson correlations below the diagonal. Prob > |r| under Ho: Rho = 0 CAR = monthly cumulative abnormal return over the 3month Bull/Bear period ABVOL = monthly abnormal trading volume over the 3month Bull/Bear period PRG# = the number of times a firm is mentioned on the investment news program (Cramer) during the 3month Bull/Bear period BLG# = the number of times a firm is mentioned on the financial blog (SeekingAlpha) during the 3month Bull/Bear period PRG%B = the percentage of buy recommendations to total mentions on the investment news program (Cramer) during the 3month Bull/Bear period. BLG%B = the percentage of buy recommendations to total mentions on the financial blog (SeekingAlpha) during the 3month Bull/Bear period SOPH = investor sophistication measured as % of institutional holdings BEAR = dummy variable for market condition set to (1) for Bear period and (0) for Bull period INS = Information asymmetry measured as bidask spread EQ = earnings quality measured using the Dechow Dichev (2002) model (see equation 9) PRESS = dummy variable for press set to (1) if the firm had significant news, earnings announcement, or SEC filing during the 3month Bull/Bear period; a control variable SIZE = firm size measured as the natural log of the market value of the common stock; a control variable 119

TABLE 12 INTENSITY Regression Results - CAR

CAR = α + β1PRG# + β2BLG# + β3SOPH + β4BEAR + β5INS + β6EQ + β7PRG#*SOPH + β8PRG#*BEAR + β9PRG#*INS + β10 PRG#*EQ + β11 BLG#*SOPH + β12 BLG#*BEAR + β13 BLG#*INS + β14 BLG#*EQ + β15 PRESS + β16 SIZE + ε (5)

CAR = α + β1PRG# + β2BLG# + β3SOPH + β4BEAR + β5INS + β6PRG#*SOPH + β7PRG#*BEAR + β8PRG#*INS + β9BLG#*SOPH + β10 BLG#*BEAR + a β11 BLG#*INS + β12 PRESS + β13 SIZE + ε (5.1)

CAR = α + β1PRG# + β2BLG# + β3SOPH + β4BEAR + β5INS + β6PRG#*SOPH + β7PRG#*BEAR + β8PRG#*INS + β9BLG#*SOPH + β10 BLG#*BEAR + b β11 BLG#*INS + β12 PRESS + β13 SIZE + β14 EXCHG1 + β15 EXCHG2 + β16 EXCHG3 + ε (5.2) Equation (5) Equation (5.1)a Equation (5.2)b Variable Hypothesis Expected Sign Coefficient p-value d Coefficient p-value d Coefficient p-value d Intercept ? 0.8651 <.0001 0.5926 <.0001 0.5209 0.0001 PRG# H1a + 0.0478 0.1284 0.0622 0.0280 0.0606 0.0328 BLG# H1a + 0.1528 0.1633 0.0800 0.2507 0.0678 0.3319 SOPH 0.2719 0.0129 0.1325 0.1055 0.1551 0.0609 BEAR 0.1152 0.0104 0.0893 0.0239 0.0934 0.0183 INS 83.9811 0.0019 87.5608 <.0001 92.6719 <.0001 EQ 1.3495 0.0683 PRG#*SOPH H2 0.0165 0.6238 0.0235 0.4500 0.0202 0.5155 PRG#*BEAR H3 + 0.0158 0.3029 0.0201 0.1267 0.0206 0.1179 PRG#*INS H4 + 20.214 0.1870 19.9792 0.1062 21.1081 0.0895 PRG#*EQ H5 + 0.1873 0.5485 BLG#*SOPH H2 0.1291 0.3067 0.0653 0.4106 0.0648 0.4141 BLG#*BEAR H3 + 0.0756 0.0954 0.0671 0.0920 0.0633 0.1127 BLG#*INS H4 + 0.2959 0.9683 2.0068 0.7814 2.0653 0.7753 BLG# *EQ H5 + 0.6968 0.2710 PRESS 0.0242 0.5322 0.0435 0.2111 0.0394 0.2584 SIZE 0.0749 <.0001 5.8600 <.0001 0.0684 <.0001 EXCHG1 0.1615 0.0408 EXCHG2 0.1367 0.4901 EXCHG3 0.1364 0.0814 Fstatistic 3.34 <.0001 4.91 <.0001 4.26 <.0001 Adjusted R 2 0.0529 0.0446 0.0458 Nc 672 1,089 1,089 120

Table 12 (continued) CAR = monthly cumulative abnormal return over the 3month Bull/Bear period PRG# = the number of times a firm is mentioned on the investment news program (Cramer) during the 3month Bull/Bear period BLG# = the number of times a firm is mentioned on the financial blog ( SeekingAlpha ) during the 3month Bull/Bear period SOPH = investor sophistication measured as % of institutional holdings BEAR = dummy variable for market condition set to (1) for Bear period and (0) for Bull period INS = Information asymmetry measured as bidask spread EQ = earnings quality measured using the Dechow Dichev (2002) model (see equation 9) PRG#*SOPH = interaction term based on PRG# x SOPH PRG#*BEAR = interaction term based on PRG# x BEAR PRG#*INS = interaction term based on PRG# x INS PRG#*EQ = interaction term based on PRG# x EQ BLG#*SOPH = interaction term based on BLG# x SOPH BLG#*BEAR = interaction term based on BLG# x BEAR BLG#*INS = interaction term based on BLG# x INS BLG#*EQ = interaction term based on BLG# x EQ PRESS = dummy variable for press set to (1) if the firm had significant news, earnings announcement, or SEC filing during the 3month Bull/Bear period; a control variable SIZE = firm size measured as the natural log of the market value of the common stock; a control variable EXCHG = dummy variable scheme for the stock exchange that the firm is traded on; EXCHG1 is set to (1) if traded on NYSE, (0) otherwise; EXCHG2 is set to (1) if traded on AMEX, (0) otherwise; EXCHG3 is set to (1) if traded on NASDAQ, (0) otherwise; if EXCHG1, EXCHG2, and EXCHG3 are all set to (0) then stock is traded on a Regional exchange a Equation (5.1) excludes the earnings quality variable (EQ)and interaction terms to allow for a larger sample size. b Equation (5.2) excludes the earnings quality variable (EQ) and interaction terms and adds the variable for stock exchange (EXCHG), an additional control variable. c The number of firms in the multivariate analysis increased from 676 to 1,089 after dropping the EQ variable which resulted in a lower sample size due to missing data. d pvalues are based on twotailed tests.

121

TABLE 13 INTENSITY Regression Results - ABVOL

ABVOL = α + β1PRG# + β2BLG# + β3SOPH + β4BEAR + β5INS + β6EQ + β7PRG#*SOPH + β8PRG#*BEAR + β9PRG#*INS + β10 PRG#*EQ + β11 BLG#*SOPH + β12 BLG#*BEAR + β13 BLG#*INS + β14 BLG#*EQ + β15 PRESS + β16 SIZE + ε (6)

ABVOL = α + β1PRG# + β2BLG# + β3SOPH + β4BEAR + β5INS + β6PRG#*SOPH + β7PRG#*BEAR + β8PRG#*INS + β9BLG#*SOPH + β10 BLG#*BEAR + a β11 BLG#*INS + β12 PRESS + β13 SIZE + ε (6.1)

ABVOL = α + β1PRG# + β2BLG# + β3SOPH + β4BEAR + β5INS + β6PRG#*SOPH + β7PRG#*BEAR + β8PRG#*INS + β9BLG#*SOPH + β10 BLG#*BEAR + b β11 BLG#*INS + β12 PRESS + β13 SIZE + β14 EXCHG1 + β15 EXCHG2 + β16 EXCHG3 + ε (6.2) Equation (6) Equation (6.1) a Equation (6.2) b Variable Hypothesis Expected Sign Coefficient p-value d Coefficient p-value d Coefficient p-value d Intercept ? 0.6442 0.0022 3.4687 <.0001 6.3142 <.0001 PRG# H1b + 0.2113 <.0001 0.3017 0.0006 0.2146 0.0020 BLG# H1b + 0.2034 0.1207 0.0046 0.9829 0.3532 0.0391 SOPH 0.3183 0.0149 1.4291 <.0001 0.7088 0.0005 BEAR 0.1301 0.0154 0.1817 0.1380 0.0552 0.5686 INS 16.0461 0.6190 254.0795 <.0001 79.0393 0.0763 EQ 1.0040 0.2561 PRG#*SOPH H2 0.2468 <.0001 0.3817 <.0001 0.2737 0.0003 PRG#*BEAR H3 + 0.0360 0.0491 0.0122 0.7649 0.0184 0.5683 PRG#*INS H4 + 84.6825 <.0001 73.9697 0.0537 83.2803 0.0062 PRG#*EQ H5 + 0.3920 0.2936 BLG#*SOPH H2 0.3852 0.0109 0.4831 0.0499 0.4882 0.0120 BLG#*BEAR H3 + 0.0326 0.5472 0.1261 0.3069 0.0364 0.7092 BLG#*INS H4 + 13.207 0.1382 2.7221 0.9033 0.1094 0.9951 BLG#*EQ H5 + 1.1595 0.1255 PRESS 0.0680 0.1415 0.0760 0.4602 0.0474 0.5782 SIZE 0.0294 0.0674 0.1982 <.0001 0.0625 0.0235 EXCHG1 4.8423 <.0001 EXCHG2 5.0186 <.0001 EXCHG3 4.7885 <.0001 Fstatistic 8.71 <.0001 6.39 <.0001 49.30 <.0001 Adjusted R 2 0.1553 0.0605 0.4153 Nc 672 1,089 1,089

122

Table 13 (continued) ABVOL = monthly abnormal trading volume in the 000’s over the 3month Bull/Bear period PRG# = the number of times a firm is mentioned on the investment news program (Cramer) during the 3month Bull/Bear period BLG# = the number of times a firm is mentioned on the financial blog (SeekingAlpha) during the 3month Bull/Bear period SOPH = investor sophistication measured as % of institutional holdings BEAR = dummy variable for market condition set to (1) for Bear period and (0) for Bull period INS = Information asymmetry measured as bidask spread EQ = earnings quality measured using the Dechow Dichev (2002) model (see equation 9) PRG#*SOPH = interaction term based on PRG# x SOPH PRG#*BEAR = interaction term based on PRG# x BEAR PRG#*INS = interaction term based on PRG# x INS PRG#*EQ = interaction term based on PRG# x EQ BLG#*SOPH = interaction term based on BLG# x SOPH BLG#*BEAR = interaction term based on BLG# x BEAR BLG#*INS = interaction term based on BLG# x INS BLG#*EQ = interaction term based on BLG# x EQ PRESS = dummy variable for press set to (1) if the firm had significant news, earnings announcement, or SEC filing during the 3month Bull/Bear period; a control variable SIZE = firm size measured as the natural log of the market value of the common stock; a control variable EXCHG = dummy variable scheme for the stock exchange that the firm is traded on; EXCHG1 is set to (1) if traded on NYSE, (0) otherwise; EXCHG2 is set to (1) if traded on AMEX, (0) otherwise; EXCHG3 is set to (1) if traded on NASDAQ, (0) otherwise; if EXCHG1, EXCHG2, and EXCHG3 are all set to (0) then stock is traded on a Regional exchange a Equation (6.1) excludes the earnings quality variable (EQ)and interaction terms to allow for a larger sample size. b Equation (6.2) excludes the earnings quality variable (EQ) and interaction terms and adds the variable for stock exchange (EXCHG), an additional control variable. c The number of firms in the multivariate analysis increased from 676 to 1,089 after dropping the EQ variable which resulted in a lower sample size due to missing data. d pvalues are based on twotailed tests.

123

TABLE 14 SENTIMENT Regression Results – CAR

CAR = α + β1PRG%B + β2BLG%B + β3SOPH + β4BEAR + β5INS + β6EQ + β7PRG%B*SOPH + β8PRG%B*BEAR + β9PRG%B *INS + β10 PRG%B*EQ + β11 BLG%B*SOPH + β12 BLG%B*BEAR + β13 BLG%B*INS + β14 BLG%B*EQ + β15 PRESS + β16 SIZE + ε (7)

CAR = α + β1PRG%B + β2BLG%B + β3SOPH + β4BEAR + β5INS + β6PRG%B*SOPH + β7PRG%B*BEAR + β8PRG%B *INS + β9BLG%B*SOPH + β10 BLG%B*BEAR + a β11 BLG%B*INS + β12 PRESS + β13 SIZE + ε (7.1)

CAR = α + β1PRG%B + β2BLG%B + β3SOPH + β4BEAR + β5INS + β6PRG%B*SOPH + β7PRG%B*BEAR + β8PRG%B *INS + β9BLG%B*SOPH + β10 BLG%B*BEAR + b β11 BLG%B*INS + β12 PRESS + β13 SIZE + β14 EXCHG1 + β15 EXCHG2 + β16 EXCHG3 + ε (7.2) Equation (7) Equation (7.1) a Equation (7.2) b Variable Hypothesis Expected Sign Coefficient p-value d Coefficient p-value d Coefficient p-value d Intercept ? 0.7778 <.0001 0.6046 <.0001 0.5183 <.0001 PRG%B H1c + 0.0736 0.6164 0.0536 0.6416 0.0391 0.7343 BLG%B H1c + 0.0678 0.5923 0.1460 0.1069 0.1258 0.1675 SOPH 0.2100 0.0371 0.1792 0.0166 0.2000 0.0080 BEAR 0.0994 0.0434 0.0569 0.1739 0.0580 0.1655 INS 98.3140 <.0001 96.2421 <.0001 101.3113 <.0001 EQ 1.5756 0.0313 PRG%B*SOPH H2 0.1110 0.4269 0.0054 0.9624 0.0225 0.8451 PRG%B*BEAR H3 + 0.0094 0.8870 0.0135 0.8185 0.0147 0.8035 PRG%B*INS H4 + 6.3659 0.9044 36.1708 0.2730 36.5194 0.2687 PRG%B*EQ H5 + 0.3805 0.7180 BLG%B*SOPH H2 0.0337 0.8085 0.1245 0.2134 0.1173 0.2422 BLG%B*BEAR H3 + 0.0284 0.6891 0.0937 0.1361 0.0944 0.1339 BLG%B*INS H4 + 5.4972 0.4380 10.4818 0.1314 10.2726 0.1413 BLG%B*EQ H5 + 1.6227 0.0370 PRESS 0.0356 0.3415 0.0692 0.0392 0.0654 0.0516 SIZE 0.0664 <.0001 0.0543 <.0001 0.0602 <.0001 EXCHG1 0.1664 0.0355 EXCHG2 0.1451 0.4679 EXCHG3 0.1480 0.0604 Fstatistic 3.16 <.0001 4.40 <.0001 3.86 <.0001 Adjusted R 2 0.0490 0.0390 0.0403 Nc 672 1,089 1,089 124

Table 14 (continued) CAR = monthly cumulative abnormal return over the 3month Bull/Bear period PRG%B = the percentage of buy recommendations to total mentions on the investment news program (Cramer) during the 3month Bull/Bear period. BLG%B = the percentage of buy recommendations to total mentions on the financial blog (SeekingAlpha) during the 3month Bull/Bear period SOPH = investor sophistication measured as % of institutional holdings BEAR = dummy variable for market condition set to (1) for Bear period and (0) for Bull period INS = Information asymmetry measured as bidask spread EQ = earnings quality measured using the Dechow Dichev (2002) model (see equation 9) PRG%B*SOPH = interaction term based on PRG%B x SOPH PRG%B*BEAR = interaction term based on PRG%B x BEAR PRG%B*INS = interaction term based on PRG%B x INS PRG%B*EQ = interaction term based on PRG%B x EQ BLG%B*SOPH = interaction term based on BLG%B x SOPH BLG%B*BEAR = interaction term based on BLG%B x BEAR BLG%B*INS = interaction term based on BLG%B x INS PRESS = dummy variable for press set to (1) if the firm had significant news, earnings announcement, or SEC filing during the 3month Bull/Bear period; a control variable BLG#*EQ = interaction term based on BLG# x EQ SIZE = firm size measured as the natural log of the market value of the common stock; a control variable EXCHG = dummy variable scheme for the stock exchange that the firm is traded on; EXCHG1 is set to (1) if traded on NYSE, (0) otherwise; EXCHG2 is set to (1) if traded on AMEX, (0) otherwise; EXCHG3 is set to (1) if traded on NASDAQ, (0) otherwise; if EXCHG1, EXCHG2, and EXCHG3 are all set to (0) then stock is traded on a Regional exchange a Equation (7.1) excludes the earnings quality variable (EQ)and interaction terms to allow for a larger sample size. b Equation (7.2) excludes the earnings quality variable (EQ) and interaction terms and adds the variable for stock exchange (EXCHG), an additional control variable. c The number of firms in the multivariate analysis increased from 676 to 1,089 after dropping the EQ variable which resulted in a lower sample size due to missing data. d pvalues are based on twotailed tests.

125

TABLE 15 SENTIMENT Regression Results - ABVOL

ABVOL = α + β1PRG%B + β2BLG%B + β3SOPH + β4BEAR + β5INS + β6EQ + β7PRG%B*SOPH + β8PRG%B*BEAR + β9PRG%B *INS + β10 PRG%B*EQ + β11 BLG%B*SOPH + β12 BLG%B*BEAR + β13 BLG%B*INS + β14 BLG%B*EQ + β15 PRESS + β16 SIZE + ε (8)

ABVOL = α + β1PRG%B + β2BLG%B + β3SOPH + β4BEAR + β5INS + β6PRG%B*SOPH + β7PRG%B*BEAR + β8PRG%B *INS + β9BLG%B*SOPH + β10 BLG%B*BEAR + a β11 BLG%B*INS + β12 PRESS + β13 SIZE + ε (8.1)

ABVOL = α + β1PRG%B + β2BLG%B + β3SOPH + β4BEAR + β5INS + β6PRG%B*SOPH + β7PRG%B*BEAR + β8PRG%B *INS + β9BLG%B*SOPH + β10 BLG%B*BEAR + b β11 BLG%B*INS + β12 PRESS + β13 SIZE + β14 EXCHG1 + β15 EXCHG2 + β16 EXCHG3 + ε (8.2) Equation (8) Equation (8.1) a Equation (8.2) b Variable Hypothesis Expected Sign Coefficient p-value d Coefficient p-value d Coefficient p-value d Intercept ? 0.2509 0.2385 2.9708 <.0001 5.9783 <.0001 PRG%B H1c + 0.1671 0.3668 1.2194 0.0007 0.7688 0.0063 BLG%B H1c + 0.0875 0.5835 0.1197 0.6698 0.4763 0.0320 SOPH 0.1951 0.1241 0.9872 <.0001 0.3629 0.0482 BEAR 0.1811 0.0036 0.1061 0.4137 0.0722 0.4786 INS 5.4059 0.8598 211.6831 <.0001 49.0356 0.2194 EQ 1.4938 0.1089 PRG%B*SOPH H2 0.1664 0.3451 0.7890 0.0268 0.3345 0.2331 PRG%B*BEAR H3 + 0.0432 0.6033 0.1357 0.4578 0.1341 0.3505 PRG%B*INS H4 + 154.3639 0.0212 149.7871 0.1434 136.7488 0.0893 PRG%B*EQ H5 + 0.3937 0.7669 BLG%B*SOPH H2 0.1245 0.4784 0.1328 0.6685 0.3276 0.1800 BLG%B*BEAR H3 + 0.1078 0.2288 0.1587 0.4154 0.1186 0.4394 BLG%B*INS H4 + 7.3113 0.4133 15.4039 0.4746 9.6407 0.5709 BLG%B*EQ H5 + 1.2441 0.2041 PRESS 0.0206 0.6621 0.0664 0.5234 0.1859 0.0232 SIZE 0.0084 0.6217 0.1420 <.0001 0.0117 0.6676 EXCHG1 4.9130 <.0001 EXCHG2 4.8620 <.0001 EXCHG3 4.8467 <.0001 Fstatistic 3.49 <.0001 5.79 <.0001 49.61 <.0001 Adjusted R 2 0.0561 0.0541 0.4169 Nc 672 1,089 1,089

126

Table 15 (continued) ABVOL = monthly abnormal trading volume in the 000’s over the 3month Bull/Bear period PRG%B = the percentage of buy recommendations to total mentions on the investment news program (Cramer) during the 3month Bull/Bear period. BLG%B = the percentage of buy recommendations to total mentions on the financial blog (SeekingAlpha) during the 3month Bull/Bear period SOPH = investor sophistication measured as % of institutional holdings BEAR = dummy variable for market condition set to (1) for Bear period and (0) for Bull period INS = Information asymmetry measured as bidask spread EQ = earnings quality measured using the Dechow Dichev (2002) model (see equation 9) PRG%B*SOPH = interaction term based on PRG%B x SOPH PRG%B*BEAR = interaction term based on PRG%B x BEAR PRG%B*INS = interaction term based on PRG%B x INS PRG%B*EQ = interaction term based on PRG%B x EQ BLG%B*SOPH = interaction term based on BLG%B x SOPH BLG%B*BEAR = interaction term based on BLG%B x BEAR BLG%B*INS = interaction term based on BLG%B x INS BLG%B*EQ = interaction term based on BLG%B x EQ PRESS = dummy variable for press set to (1) if the firm had significant news, earnings announcement, or SEC filing during the 3month Bull/Bear period; a control variable SIZE = firm size measured as the natural log of the market value of the common stock; a control variable EXCHG = dummy variable scheme for the stock exchange that the firm is traded on; EXCHG1 is set to (1) if traded on NYSE, (0) otherwise; EXCHG2 is set to (1) if traded on AMEX, (0) otherwise; EXCHG3 is set to (1) if traded on NASDAQ, (0) otherwise; if EXCHG1, EXCHG2, and EXCHG3 are all set to (0) then stock is traded on a Regional exchange a Equation (8.1) excludes the earnings quality variable (EQ) and interaction terms to allow for a larger sample size. b Equation (8.2) excludes the earnings quality variable (EQ) and interaction terms and adds the variable for stock exchange (EXCHG), an additional control variable. c The number of firms in the multivariate analysis increased from 676 to 1,089 after dropping the EQ variable which resulted in a lower sample size due to missing data. d pvalues are based on twotailed tests.

127

128

129