Activist Short-Selling by Wuyang Zhao A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Rotman School of Management University of Toronto © Copyright by Wuyang Zhao 2017
Activist Short-Selling
by
Wuyang Zhao
A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy
Rotman School of Management University of Toronto
© Copyright by Wuyang Zhao 2017
ii
Activist Short-Selling
Wuyang Zhao
Doctor of Philosophy
Rotman School of Management
University of Toronto
2017
Abstract
Activist short-selling – short-sellers publicly talking down securities to benefit their short
positions – attracts attentions from a broad range of market participants, but is largely ignored by
the prior literature. Combining information from Seeking Alpha and Activist Shorts Research, I
collect more than 6,000 activist short-selling cases against listed companies from 2006 to 2015
and conduct two large-sample studies on this controversial phenomenon.
Chapter 1 examines the determinants and consequences of activist short-selling,
highlighting the critical roles of ex-ante available firm characteristics. I find that (1) activist short-
selling leads to much larger market reactions than comparable level of short-selling without public
talking-down, (2) activist short-sellers are more likely to target firms with severe overvaluation
and uncertainty features, (3) targets’ overvaluation (uncertainty) features are increasingly
(decreasingly) important in predicting returns from the short term to the long term, (4) their
overvaluation features predict short-selling allegations that focus on valuation issues such as
“bubble,” while their uncertainty features predict allegations that sound severe such as “fraud,”
and (5) uncertainty features also predict targets’ likelihood of responding to allegations.
iii
Chapter 2 focuses on the consequences of activist short-selling on a specific group of
market players: sell-side analysts who cover target firms. Since the sell-side business model
discourages analysts from publishing negative opinions, the demand for negative information is
met by other sources, such as activist short-sellers who frequently accuse analysts’ talking-up as
the main driver of overvaluation. I find that analysts react to activist short-selling by revising their
target-price forecasts. The variation in the timeliness and direction of the reactions can be
explained by the ability and incentives of analysts as well as the initial impact of activist short-
selling. More importantly, analysts’ reputation can be severely damaged by activist short-sellers,
particularly when analysts talk-up too much previously, when short-sellers seem to be right, and
when analysts take too long to respond. However, the direction of analysts’ reactions (i.e., revising
up or down) seems unrelated to the reputation loss. Finally, analysts are more likely to move to
smaller brokerage houses if their covered firms are targeted by activist short-sellers.
iv
Acknowledgments
Frequently I ask myself: what if I had not come to Rotman for this accounting PhD? I can
think of a lot of possibilities, but under no situation could I be as happy, optimistic, and self-
fulfilling as I am now – not even close. Now this Ph.D. is coming to its end; I have a lot of people
to thank.
First of all, I thank the Chair of my advising committee: Professor Ole-Kristian Hope. As
evidenced by the several thousand emails we had in the past four years, he had substantial impact
on every single important event I have had. Here are some examples. Attending his seminar in
Norway in the summer of 2012 led to my admission into the Rotman Accounting PhD program.
Reading a book given by him (i.e., Fooling Some of the People All of the Time) ignited my interest
in short-selling. He is also a key coauthor of my two published papers. As always, I have received
his incredible support and guidance in developing this dissertation. I have been and will always be
learning from him on how to become a successful scholar: hard-working, optimism, confidence,
and care for younger generations.
I also want to thank my four committee members for their valuable support in developing
the dissertation and hunting for a job: Alex Edwards, Partha Mohanram, Baohua Xin, and
Dushyant Vyas. Interacting with them helps me to be a better researcher and a better person.
Specifically, Alex was my mentor in the first year, and since then I have learned how to be a good
colleague from him. Partha’s seminar course on valuation prepared me for the research on activist
short-selling. Also, his broad knowledge about almost everything reminds me of the importance
of being an interesting person. Baohua’s theory seminar led me to think of my dissertation topic
in a deeper level, and I have learned from him how to strike a balance between being humble and
being confident. Dushyant was among the first who made me determined to focus on activist short-
selling, mostly because this is a topic that is highly relevant to practice. He has shaped my taste on
what constitutes good research.
I am grateful to my external examiner, Prof. Michael Welker from Queens’ University, for
his encouragements and highly constructive comments. I still remember that I was talking with
Michael about the attack by Citron Research against Valeant on Oct. 21, 2015 at PCAOB/JAR
conference in D.C. That case makes me to seriously consider to focus on activist short-selling.
v
I wish to thank all the Rotman professors who helped me during the Ph.D. process. I have
discussed my dissertation with almost every faculty member individually, including Francesco
Bova, Jeff Callen, Feng Chen, Gus De Franco, Daehyun Kim, Nan Li, Scott Liao, Gord
Richardson, Christopher Small, Franco Wong, Aida Wahid, Minlei Ye, and Ping Zhang. I also
wish to thank all the fellow Ph.D. students who made my life easier during the four years, including
Muhammad Azim, Mahfuz Chy, Danqi Hu, Ross Lu, Barbara Su, Mingyue Zhang, and particularly
Stephanie Cheng, with whom I was very fortunate to start the program at Rotman.
I thank my coauthors and friends outside Rotman, especially Han Wu at HEC Paris and
Forester Wong at USC. They are always available for random chats, which not only makes the
research work less lonely but more enjoyable.
Finally, I am grateful to my family, especially to my wife and my best friend, Dr. Jessie
Yin Zhu. Equally well-educated and probably more talented, she contains her own aspiration and
ambition, but accompanies me wholeheartedly on an uncertain journey. I am grateful to her
dedication to love, her sacrifice to the family, and her unmatched trust and confidence in me. I
dedicate this doctorate thesis to her and our daughter Grace Shuman Zhao.
In addition, I want to thank people who provided comments on my dissertation, including
Pat Akey, Stefan Anchev, Karthik Balakrishnan, Mark Bradshaw, Wenjiao Cao, Shuping Chen,
Ted Christensen, Dain Donelson, Yiwei Dou, Fabrizio Ferri, Robert Freeman, Pingyang Gao,
Jonathan Glover, Amy Hutton, Ross Jennings, Alon Kalay, Bin Ke, Steve Kachelmeier, Mark Ma,
John McInnis, Stephen Penman, Shivaram Rajgopal, Scott Richardson, Sugata Roychowdhury,
Lakshmanan Shivakumar, Hun Tong Tan, Florin Vasvari, Brian White, Brady Williams, Eyub
Yegen, Yong Yu, Ronghuo Zheng, Emanuel Zur (FEA Discussant), and seminar participants at
Boston College, Columbia University, London Business School, Nanyang Technological
University, National University of Singapore, University of Texas at Austin, 2016 AAA Doctoral
Consortium, 2016 CMU Accounting Mini Conference, 2016 Conference on Financial Economics
and Accounting (FEA), and 2017 Financial Accounting and Reporting Section midyear meeting
(FARS) for comments. Last but not least, I thank Adam Kommel from Activist Shorts Research
for generously sharing his data, and Mingqi Li for Python assistance.
vi
Table of Contents
Abstract ........................................................................................................................................... ii
Acknowledgments.......................................................................................................................... iv
Table of Contents ........................................................................................................................... vi
Introduction ......................................................................................................................................1
Chapter 1 Activist Short-Selling: A Large-Sample Study on the Determinants and
Consequences .............................................................................................................................3
1 Introduction .................................................................................................................................3
2 Related Literature and Hypotheses Development .......................................................................8
2.1 The Differences between Activist and Passive Short-Selling ...............................................9
2.1.1 Short-Selling Incentives ...........................................................................................9
2.1.2 Constraints in the Equity-Loan Market..................................................................10
2.1.3 Short-Selling Risk ..................................................................................................10
2.1.4 Activist Short-Selling as a Coordination Device ...................................................11
2.2 Market Reactions to Activist Short-Selling (H1) ................................................................12
2.3 Activist Short-Selling Strategies (H2 and H3) ....................................................................12
2.4 Overvaluation and Uncertainty Features and Short- and Long-Term Returns (H4) ...........13
3 Data ...........................................................................................................................................14
3.1 Seeking Alpha .....................................................................................................................14
3.2 Activist Shorts Research .....................................................................................................15
3.3 Sample-Construction Process .............................................................................................15
4 Does Activist Short-Selling Have Larger Market Reactions Than Passive Short-Selling? ......16
5 Determinants of Being Targeted by Activist Short-Selling ......................................................18
5.1 Variables and Models ..........................................................................................................18
vii
5.1.1 Overvaluation Features ..........................................................................................18
5.1.2 Uncertainty Features ..............................................................................................19
5.1.3 The Model ..............................................................................................................19
5.2 Determinants Sample ..........................................................................................................20
5.3 Regression Results .............................................................................................................21
5.3.1 Overvaluation Features .............................................................................................21
5.3.2 Uncertainty Features ..............................................................................................21
5.3.3 Horseracing Variables and Including Both Features Together ..............................22
6 Overvaluation and Uncertainty Features and Short- and Long-Term Returns .........................23
6.1 Models and Variables ..........................................................................................................23
6.2 Sample and Regression Results ..........................................................................................23
6.3 An Ex-Ante Approach to Separate Winners vs. Losers among Activist Short-Selling ......24
7 Activist Short-Sellers’ Allegations and Firms’ Tendency to Respond .....................................25
7.1 Short-Sellers’ Allegations ...................................................................................................26
7.2 Firms’ Tendency to React to Short-Sellers’ Allegations ....................................................26
8 Supplemental Analyses and Robustness Tests ..........................................................................27
8.1 Separating ASR and SA Samples .......................................................................................27
8.2 Overvaluation and Uncertainty Features and Short-Interest ...............................................28
8.3 Overvaluation and Uncertainty Features and Analysts’ Recommendations .......................28
8.4 A Pseudo Test Using Peer Firms Matched on Firm Features .............................................29
8.5 The Interaction Effects of Overvaluation and Uncertainty Features ..................................29
8.6 The Role of Media and Short-Sellers’ Reputation ..............................................................30
8.7 Multiple Activist Short-Selling Events in the Same Firm-Quarter .....................................30
8.8 Other Robustness Checks ....................................................................................................31
9 Conclusion ................................................................................................................................31
viii
References of Chapter 1 .................................................................................................................34
Appendices of Chapter 1 ................................................................................................................38
Appendix A: Activist Short-Selling in Seeking Alpha and Activist Shorts Research ..............38
Appendix B: The Activist Short-Selling Date and Financial Reporting Data ..........................40
Appendix C: Variable Definitions ............................................................................................41
Main Tables of Chapter 1 ..............................................................................................................45
Main Figures of Chapter 1 .............................................................................................................59
Additional Robustness Tests ..........................................................................................................60
Chapter 2 Selling Financial Analysts Short: The Impact of Activist Short-Selling on
Sell-Side Analysts ....................................................................................................................83
1 Introduction ...............................................................................................................................83
2 Related Literature ......................................................................................................................89
2.1 Sell-Side Optimism .............................................................................................................89
2.2 Activist Short-Sellers as a Source of Negative Research....................................................90
2.3 Prior Research on the Information Flows between Short-Sellers and Analysts .................91
3 Do Financial Analysts React to Activist Short-Selling? ...........................................................92
4 What Determines Financial Analysts’ Reactions to Activist Short-Selling? ............................94
5 Analysts’ Reputation Loss after Activist Short-Selling ............................................................96
5.1 Main Results .......................................................................................................................96
5.2 Cross-Sectional Variation: Analysts’ View Prior to Activist Short-Selling .......................98
5.3 Cross-Sectional Variation: The Market’s Initial Reactions ................................................98
5.4 Cross-Sectional Variation: What Analysts Can Do to Avoid Reputation Loss ..................99
6 The Impact of Activist Short-Selling on Analysts’ Careers......................................................99
7 Robustness Checks and Supplemental Analyses ....................................................................102
7.1 An Omitted-Variable Problem ..........................................................................................102
7.2 Analysts React by Revising EPS Forecasts ......................................................................102
ix
7.3 Do Investors Become Less Responsive in General? .........................................................103
8 Conclusion ..............................................................................................................................104
References of Chapter 2 ...............................................................................................................105
Appendices of Chapter 2 ..............................................................................................................107
Appendix A: Variable Definitions ..........................................................................................107
Main Tables of Chapter 2 ............................................................................................................109
Main Figures of Chapter 2 ...........................................................................................................120
1
Introduction
Recent years have witnessed a financial innovation where short-sellers publicly talk down
stocks to benefit their short positions (Ljungqvist and Qian 2016), a phenomenon commonly
referred to as “activist short-selling” in the investing community. On top of commenting on
popular media and at investing conferences, the rising popularity of social media platforms has
allowed short-sellers to disseminate their short-theses to a massive crowd. Prominent activist short-
selling cases such as Citron Research’s attack on Valeant attract tremendous attention from the
public, the media, and the regulators. However, due to the lack of easily available data, the prior
literature on short-selling mainly explores its “passive” dimension, focusing on short-interest,
realized short-sales, and market-wide short-selling regulations and largely ignoring the
possibilities of short-sellers as activists. This doctoral thesis seeks to fill the void in the literature
by conducting two large-sample studies on the roles of activist short-selling in capital markets.
Combining information from Seeking Alpha (SA) and Activist Shorts Research (ASR), I collect
more than 6,000 activist short-selling cases from 2006 to 2015 in the US equity market.
Chapter 1 illustrates the differences between activist short-selling and passive short-selling
(i.e., short-selling without public talking-down) and investigates the determinants and
consequences of activist short-selling. First, activist short-selling is expected to be more
informative than its passive counterpart because it is (1) based on profit incentives, (2) largely
unconstrained by the supply in the equity-loan market, and (3) riskier. Second, higher-order beliefs
– the fact that investors infer from short-sellers’ bearish public signals how other investors are
likely to behave – can encourage panic-selling among investors. Indeed, I do find that activist
short-selling leads to bigger market reactions than comparable passive short-selling does.
The core of Chapter 1 is to highlight the crucial roles of two ex-ante firm characteristics:
overvaluation and uncertainty. I find that activist short-sellers are more likely to target firms with
egregious overvaluation (as typical short-sellers do), and firms with high uncertainty (and therefore
with an investor base that could be easily influenced). Next, I examine the return implications of
these two features in the short and long term. I find that long-term returns depend on overvaluation
features more than short-term returns do, but depend on uncertainty features less than short-term
returns do. These results are consistent with the Higher-Order Beliefs view, because short-term
2
returns are affected more by the panic of “lest everyone else get out first,” which relates to
uncertainty, but long-term returns reflect firm fundamentals captured by the overvaluation
features. These results also suggest an ex-ante way to separate winners vs. losers among activist
short-selling cases. Finally, with the rich information available in ASR, I find that overvaluation
features predict allegations on valuation issues (e.g., “bubble”), that uncertainty features predict
allegations that sound severe (e.g., “fraud”), and that firms with uncertainty features are more
likely to respond to allegations by accusing short-sellers of manipulation or by providing relevant
information to address their allegations.
Chapter 2 investigates the consequences of activist short-selling on a specific group of
market player: sell-side analysts. Note that the business model of sell-side research discourages
analysts to publish negative opinions. However, a healthy and efficient capital market requires
opinions from both positive and negative sides. Activist short-sellers are equipped with the right
incentive and abilities to publicly talk down stocks and to meet the market’s need for negative
opinions. Moreover, activist short-sellers frequently accuse analysts’ systematic talking-up as the
key driver of overvaluation. I find that analysts react to activist short-selling by revising their
target-price forecasts, and the variation in the timeliness and direction of the reactions can be
explained by the ability and incentives of analysts as well as the initial impact of activist short-
selling. When analysts’ talking-up meets with activist short-sellers’ talking-down, analysts’
reputation and future career prospects could possibly suffer. Following prior literature, I examine
how the market reacts to analysts’ EPS revisions differently after activist short-selling. I find that
analysts’ reputation can be severely damaged by activist short-sellers, particularly when analysts
talk-up too much previously, when short-sellers seem to be right, and when analysts take too long
to respond. However, the direction of analysts’ reactions (i.e., revising up or down) seems
unrelated to the reputation loss. Also, activist short-selling can influence analysts’ career prospects
such that they are more likely to move to smaller brokerage houses if their covered firms are
targeted by activist short-sellers.
3
Chapter 1 Activist Short-Selling: A Large-Sample Study on the Determinants
and Consequences
Bloomberg TV: “How concerned are you that some short-sellers may be manipulating the market
and prices of certain stocks by either publishing research or making public comments that benefit
their positions?”
Mary Jo White: “… Short-selling has a legitimate positive purpose in the marketplace. That’s very
different, though, than if you manipulate by short-selling.… The whole topic of short-selling is
something that continually gets attention from the SEC, as it does in the marketplace.” 1
1 Introduction
More than 99% of U.S. publicly traded firms (as of the end of 2015 in Compustat) have
some shares being shorted, but most short sales are conducted without public awareness due to
lack of disclosure requirements at the short-seller level. However, in some cases, short-sellers
voluntarily announce their short theses (i.e., the rationale for short-selling) to the public, as
Bloomberg TV mentioned above: “publishing research or making public comments that benefit
their positions.” The investing community refers to such actions as “activist short-selling” and the
short-sellers identify themselves as activist short-sellers.2 These cases, such as Citron Research vs.
Valeant, often create market turmoil and capture the attention from a wide range of market
participants.3 Managers and shareholders are worried that their stocks are attacked by these short-
sellers, traders try to profit from the speculative opportunities created in the process, and regulators
and the public are concerned about the possibility that activist short-sellers are manipulating the
market by creating panic. Motivated by the attentions from those market participants, I conduct a
large-sample study on activist short-selling and highlight the role of ex-ante firm characteristics in
1 Mary Jo White, the SEC Chairwoman, was interviewed on Bloomberg TV on November 10, 2015. See
http://www.bloomberg.com/news/articles/2015-11-10/sec-s-white-says-short-selling-getting-her-intense-attention-. 2 See https://www.activistshorts.com/. These terms are also used in the law literature (e.g., Lee 2013). 3 Citron Research issued multiple short-selling reports on Valeant Pharmaceuticals (NYSE: VRX) from September to
November of 2015. The most influential one was published on October 21, 2015, calling Valeant “the Pharmaceutical
Enron.” The stock price plunged as much as 40% on that day. See more details at http://www.wsj.com/articles/the-
short-who-sank-valeant-stock-1445557157.
4
attracting activist short-sellers, predicting the consequences after firms are targeted, and potentially
helping to separate manipulative short-selling from legitimate short-selling.
Prior literature on short-selling mainly explores its “passive” dimension, focusing on short-
interest, realized short-sales, and market-wide short-selling regulations (see the review by Reed
2013). Activist short-selling differs from passive short-selling in two important ways. First, activist
short-selling is likely to be more informative because (1) it is motivated by incentives betting on
price declines rather delaying tax payment or hedging risk of holding long positions, (2) it is largely
unconstrained by the supply in the equity-loan market, and (3) it is considerably riskier than
passive short-selling. Second, activist short-selling becomes a public signal when the short-seller
talks down stocks. According to Higher-Order Beliefs (HOB) theory (Morris and Shin 2002; Allen,
Morris, and Shin 2006; Gao 2008), a public signal can facilitate an investor in guessing how other
investors will behave. In the presence of an activist short-seller’s bearish public signal, an investor
will sell immediately if she is afraid that other investors will sell soon. Such panic-selling could
collapse stock prices in a stampede, and activist short-sellers are frequently accused of
manipulating stock prices in this way.4
Recognizing the above differences between passive short-selling studied in the prior
literature and activist short-selling, this paper examines the following related questions. First, does
activist short-selling lead to larger market reactions than comparable passive short-selling?
Second, what firm characteristics attract activist short-sellers? Third, how do these ex-ante
characteristics predict short-term and long-term returns for targeted firms? Fourth, do these
characteristics predict the types of allegations short-sellers make and the likelihood for targeted
firms to respond?
I construct a large sample of more than 6,000 activist short-selling cases by combining
information from Seeking Alpha (SA hereafter; www.seekingalpha.com) and Activist Shorts
4 For example, Mike Pearson, the then CEO of Valeant, accused the short-seller Andrew Left of Citron Research that
“(His) motivation is the same as one who runs into a crowded theatre and falsely yells fire. He wanted people to run.
He intentionally designed the report to frighten [emphasis added] our shareholders to drive down the price of our stock
so he could make money for his short selling.” See https://www.ft.com/content/41160372-7bf3-11e5-98fb-
5a6d4728f74e.
5
Research (ASR hereafter; www.activistshorts.com). While SA is the largest crowdsourced
investing platform appealing to non-celebrity short-sellers, ASR tracks all influential activist short-
selling events, providing a complement to the SA sample.
I first report a rapidly increasing trend of activist short-selling in the past decade, echoing
the “intense attention” from the SEC and the marketplace as Chairwoman Mary Jo White
mentioned in the opening quote of this chapter. There are substantial market reactions to activist
short-selling. For example, on average the price drops 1.56% on day 0 (the date in which the short
thesis is announced) benchmarked on the Fama-French three-factor model. To assess whether
activist short-selling leads to larger market reactions than comparable passive short-selling (H1),
I construct five benchmarks based on market reactions to short-interest announcements of either
the targeted firms or their industry peers with the closest or highest level of short-interest ratio or
increase in short-interest. Despite the existence of spillover effects (i.e., peer firms’ prices could
also be adversely affected by activist short-selling), the results show that the market reaction to
activist short-selling is much more pronounced than that to passive short-selling benchmarks.
These results are consistent with both the information view that activist short-selling is more
informative than passive short-selling and the Higher-Order Beliefs view that the bearish public
signal creates a stampede of panic-selling.
My primary hypotheses (H2 – H4) focus on the determinants and consequences of activist
short-selling. The “short-selling” nature indicates that activist short-sellers would like to target
firms with egregious overvaluation, while the “activist” nature suggests that they would like to
target firms with high uncertainty (i.e., a representative investor is not sure how precise her
information is) and therefore with an investor base that could be easily influenced. For the 12
overvaluation features I examine, activist short-sellers target 10 of them, and seven still survive if
I horserace them in one regression, including previous price run-ups, high P/B ratio relative to peer
firms, high P/V ratio, high asset growth, high net operating assets, low recent earnings, and high
accounting-manipulation probability. For the nine uncertainty features I examine, activist short-
sellers target eight of them, and six survive in horseracing, including low accounting quality,
absence of blockholders, few dedicated institutional investors, high bid-ask spreads, non–Big Four
auditor, and internal control weaknesses. I construct two aggregate measures – Overvaluation and
6
Uncertainty – by averaging the seven overvaluation features and six uncertainty features for
subsequent tests.
Next, I examine the return implications of these two features in the short and long term. I
find that overvaluation features affect long-term returns more than short-term returns, but
uncertainty features affect long-term returns less than short-term returns. These results are
inconsistent with the information view, which predicts no differential impact between the short
and long term by either overvaluation or uncertainty features. But they are consistent with the HOB
view, because long-term returns reflect firm fundamentals captured by the overvaluation features,
but short-term returns also reflect the panic of “lest everyone else get out first,” which corresponds
to uncertainty features.5 These results also suggest an ex-ante way to separate winners vs. losers
among activist short-selling cases. Specifically, target firms with overvaluation features higher
than the determinant-test sample median have a one-year cumulative abnormal return of about -
45%, while the remaining target firms do not underperform their risk-adjusted benchmark.
Finally, with the rich information available in ASR, I find that overvaluation features
predict allegations on valuation issues (e.g., “bubble”), that uncertainty features predict allegations
that sound severe (e.g., “fraud”), and that firms with uncertainty features are more likely to respond
to allegations by accusing short-sellers of manipulation or by providing relevant information to
address their allegations. These results are consistent with both the information and HOB views in
that short-sellers provide information about the actual overvaluation but they also try to create a
panic among investors facing higher uncertainty by using severe allegations.
In supplemental analyses, I first confirm that all inferences remain the same in either the
SA or the ASR sample, although ASR short-sellers, unsurprisingly, cause much larger market
reactions than SA short-sellers do. Second, I find that overvaluation and uncertainty features do
not predict short-interest ratio (or its increase), or analysts’ sell recommendations (or downgrades)
in the same way as they predict activist short-selling. Third, I conduct a pseudo test using industry
5 Christopher Cox, former Chairman of the SEC, wrote that “when an irrational panic is fueled by false rumors that
investors believe must be acted on immediately – lest everyone else get out first – market integrity is threatened.” See
Cox, “What the SEC Really Did on Short Selling.” Wall Street Journal, July 24, 2008. Available at
https://www.sec.gov/news/speech/2008/spch072408cc.htm.
7
peer firms matched with the same aggregate Overvaluation and Uncertainty values with each
activist short-selling target and find that these two features do not predict matched firms’ returns
in the same way as they predict targets’ returns. Fourth, I find some evidence that overvaluation
and uncertainty features are substitutes in attracting activist short-selling, but complements in
affecting long-term returns. Fifth, I show that activist short-selling cases covered by media or
initiated by reputable short-sellers lead to more negative returns. Finally, I conduct several other
robustness checks, and all inferences remain unchanged.
To the best of my knowledge, this is the first large-sample study on activist short-selling,
complementing two recent small-sample studies that find substantial market reactions to short-
selling reports. Specifically, Ljungqvist and Qian (2016) use 358 short-seller reports from 2006 to
2011 and find that short-selling reports lead to immediate spikes in SEC filing views, volatility,
order imbalances, realized spreads, turnover, and selling by current shareholders. Chen (2016)
examines 443 short-seller reports on 87 U.S.-listed Chinese firms between 2007 and 2014 and also
documents substantial negative market reactions to the targeted firms and their peer firms sharing
the same non-Big 4 auditors.
There are clear differences between my study and these two. First, while all three studies
find negative reactions to activist short-selling, the key finding of my dissertation is about the vital
roles of firm characteristics – a dimension largely ignored by these two studies. Specifically, I
identify two ex-ante available characteristics (i.e., overvaluation and uncertainty) that predict the
probability of being targeted, the short- and long-term returns after being targeted, the types of
allegations, and firms’ tendency to respond after being attacked. Second, while these two studies
show the information role of short-selling reports, they do not consider the possible coordination
role through which a short-seller’s bearish public signal could create panic among investors – one
major concern the regulators and the public have over activist short-selling. This dissertation
explicitly discusses such a possibility and presents evidences consistent with it. Third, I combine
SA and ASR to construct a much larger (i.e., more than 13 times) sample of activist short-selling
over a longer period (i.e., 2006 to 2015). Such a large sample enables me to investigate the
determinants of being targeted by activist short-selling.
8
This dissertation also has implications for practice. Managers can reduce firms’
attractiveness to activist short-sellers, for example, by reporting conservatively to avoid
overvaluation or improving transparency to reduce uncertainty. The takeaway for activist short-
sellers is that targeting firms with overvaluation (uncertainty) features is more profitable in the
long (short) term. Relatedly, investors with long positions and investors following short-selling
campaigns can benefit from the proposed trading strategy of separating winners vs. losers in
activist short-selling based on ex-ante available overvaluation and uncertainty characteristics.
Finally, this study can potentially inform policy debates over activist short-selling. First,
my findings suggest one approach to separate possible manipulative short-selling from legitimate
short-selling based on targets’ characteristics, which decide whether a target firm should be shorted
or should not have been shorted. In particular, regulators could pay more attention to short-selling
cases in which the target firms present few overvaluation features but many uncertainty features,
it is more likely that the short-seller is taking advantage of the high uncertainty and manipulating
the stock. Second, my findings echo the observation that “the biggest damage to stock prices is
almost always caused by the shorts who voice their thesis – those known as activist shorts,”
therefore casting doubts on the effectiveness of requiring short-position disclosure to stabilize the
market because activist shorts would not be affected as they disclose short theses voluntarily
anyway.6
2 Related Literature and Hypotheses Development
Short-sellers sell stocks that they do not own. First, they need to find a party to lend shares
to them with collateral (generally 102% of the loan’s value). After selling the shares borrowed
from the lender, they wait for the stock price to go down in order to buy shares back at a lower
price and return them to the lender. Finally, they get the collateral back from the lender. The most
frequently used matrix of short-selling activity is short-interest – the total shares sold short scaled
6 “Don’t Sell Short Sellers Short” by Michael Regan, Bloomberg Gadfly, November 16, 2015. See
https://www.bloomberg.com/gadfly/articles/2015-11-16/stop-treating-short-sellers-like-villains-despite-uproar.
9
by the total shares outstanding. Please refer to Reed (2013) for more institutional details of the
short-selling process.
2.1 The Differences between Activist and Passive Short-Selling
In the investing community, activist short-selling refers to the phenomenon that short-
sellers publicly talk down securities to benefit their short-positions, rather than wait quietly for
price declines in passive short-selling. This feature dictates the major differences between activist
and passive short-selling.
2.1.1 Short-Selling Incentives
Short-selling is conducted because of profit incentives betting on the decline of stock prices
and other incentives such as tax and hedging (Brent, Morse, and Stice 1990).7 By definition,
activist short-selling is primarily motivated by the profit incentives – activist short-sellers publicly
disclose their short theses that explain why the targets’ prices should go down. In other words,
short-sellers whose motivation is tax or hedging would be unlikely to conduct activist short-selling.
The incentives are obscure for short-selling without the “activist” element because they are
unobservable in such passive cases. As a result, when researchers rely on short-selling metrics
such as short-interest, they pool profit incentives with tax and hedging incentives. This is likely a
reason why several studies fail to find negative relations between short-selling metrics and future
stock returns (e.g., Brent et al. 1990; Figlewski and Webb 1993; Woolridge and Dickinson 1994),
despite the overwhelming evidence on short-sellers’ sophistication of identifying trading
opportunities (e.g., Pownall and Simko 2005; Desai, Krishnamurthy, and Venkataraman 2006;
Karpoff and Lou 2010; Khan and Lu 2013). In the same vein, several studies find that the
predictability of short-selling metrics varies in the incentives. For example, Aitken, Frino,
McCorry, and Swan (1998) find that market reactions to short-sale disclosure in Australia are much
7 Tax incentive refers to the term “shorting against the box” – investors take a short position in a security that they
already hold long to defer taxable gains. Hedging incentive refers to the practice that investors hedge the risk of
holding long positions by taking short positions in related securities. For example, many hedge funds use a “pairs-
trading” strategy – long one stock but short another stock in the same industry. Speculation incentive pertains to short-
sellers betting on price declines, the incentive exemplifying the unique role of short-sellers in capital markets.
10
less negative when it is related to hedging or tax incentives. Recently, Comerton-Forde, Jones, and
Putnins (2016) highlight the importance of separating two different types of short-sellers with
distinct incentives: market makers with no negative information and informed shorts with negative
information.
2.1.2 Constraints in the Equity-Loan Market
Another reason short-interest (as well as other short-selling metrics) represents a noisy
measure of profit incentives is the supply constraint in the equity-loan market (Reed 2015). If the
supply constraint is binding (not binding), short-interest reflects the supply of lendable shares
(demand of short-selling) (Beneish, Lee, and Nichols 2015).
By contrast, activist short-selling is largely unconstrained by the supply in the equity-loan
market. As Ljungqvist and Qian (2016) argue, if it is not feasible to take a large short position and
move the price by trading, activist short-sellers can move the price by disclosing their short theses
and thus encouraging long investors to sell the stock. In other words, activist short-selling reflects
the demand for short-selling rather than the supply in the equity-loan market.
2.1.3 Short-Selling Risk
Short-selling is risky mainly because of its capped upside but unlimited downside. In
addition, the equity-loan lender can recall the loan at any time or change the loan fees. These
features make both the fundamental risk and the noise trader risk substantial concerns to short-
sellers (De Long, Shleifer, Summers, and Waldmann 1990). According to Diamond and
Verrecchia (1987), short-selling is informative exactly because of its high risk – only investors
who expect short-selling profits and can compensate for the risk will decide to sell short. As
Engelberg, Reed, and Ringgenberg (2016) illustrate, stocks with more short-selling risk have lower
future returns, less price efficiency, and less short-selling.
Short-sellers can reduce such risk by taking short positions in a portfolio of stocks whose
prices are more likely to go down in the future (i.e., a valuation-based strategy). Indeed, Dechow,
Hutton, Meulbroek, and Sloan (2001) show that short-sellers target firms in stocks with low
fundamental-to-price ratios and cover their short positions as these ratios mean-revert. Another
11
way to reduce the risk is to take short positions after the price has started to decline (i.e., a
momentum-based strategy). Lamont and Stein (2004) find that short-interest on the NASDAQ in
aggregate is positively associated with the prior month’s declines in the NASDAQ index. At an
individual stock level, Savor and Gamboa-Covazos (2011) also find that short-sellers increase their
positions following prior-month price declines.
The “activist” nature of activist short-selling substantially increases the already-high short-
selling risk.8 Targeted companies often “go down fighting” with the short-sellers using a variety
of approaches (Lamont 2012). Also, activist short-sellers receive “intense attention” from the SEC
as indicated in the opening quote. Further, a reputation loss ensues if a short-seller loses a bet
publicly.9 Finally, it is not clear whether a valuation- or momentum-based strategy can reduce
activist short-selling risk for two reasons. First, activist short-sellers usually target one stock each
time by providing in-depth analyses (93% in my sample). As the risk is not diversified, the cost of
a Type I error (i.e., being short a stock whose price subsequently goes up) is considerable. So they
need to be very confident in their short theses and the overvaluation should be egregious.10 Second,
by definition, activist short-sellers should “wake the market up” rather than “wait for the market
to wake up.” Thus, a momentum-based strategy is not designed for activist short-selling.
According to Diamond and Verrecchia (1987), the nontrivial extra risk of activist short-selling is
expected to increase its informativeness. Also, this idea is consistent with Grossman and Stiglitz
(1980) that the market rewards arbitrageurs who incur costs to identify arbitrage opportunities.
2.1.4 Activist Short-Selling as a Coordination Device
Finally and rather importantly, the “activist” element of activist short-selling makes it a
public signal, affecting investors in ways above and beyond its information content. According to
HOB theory, which is initiated from Keynes’s (1936) beauty-contest analogy and later formalized
by Morris and Shin (2002) and Gao (2008), among others, a public signal has two roles: the
8 It is possible, however, that short-sellers can reduce noise trader risk by publicly talking down stocks successfully. 9 Related, since short-sellers are often portrayed as unethical, criminal, and even “un-American” (e.g., The Economist
2008), an investor’s reputation at the personal level could be affected when her short-selling is known publicly. 10 Some well-known activist shorts, such as Manuel Asensio, make it clear that overvaluation is not the primary reason
for short-selling. According to Asensio, “an overvalued company is an opinion where someone has miscalculated
future earnings; valuation is a judgment.” See Page 41, “The Most Dangerous Trade” by Richard Teitelbaum.
12
information role (e.g., the short thesis contains information regarding the target’s overvaluation
issues) and the coordination role (e.g., it creates an expectation or panic of “lest everyone else get
out first”). Indeed, the information role echoes why activist short-sellers have “a legitimate positive
purpose” (see the opening quote by Mary Jo White) in the capital market, but the coordination role
reflects the concerns that they may be manipulating the market by creating panic. The key HOB
insight is that investors tend to put too much weight on the public signal relative to the weight a
social planner would (i.e., weight based on the relative precision of the public signal and their own
private signals). In other words, investors would be affected excessively by activist short-sellers’
bearish public signal.
2.2 Market Reactions to Activist Short-Selling (H1)
The above differences between passive and activist short-selling have two major
implications. First, activist short-selling is expected to be more informative than its passive
counterpart because it is (1) based on profit incentives, (2) largely unconstrained by the supply in
the equity-loan market, and (3) riskier. Second, higher-order beliefs – the fact that investors infer
from short-sellers’ bearish public signals how other investors are likely to behave – can encourage
panic-selling among investors. Both the information view and the HOB view predict that activist
short-selling leads to larger market reactions than its passive counterpart does. Therefore, my first
hypothesis (stated in the alternative form) is as follows:
H1: Activist short-selling leads to larger market reactions than comparable passive
short-selling does.
2.3 Activist Short-Selling Strategies (H2 and H3)
Passive short-selling strategies documented in prior literature facilitate our understanding
of activist short-selling, even though they do not apply unconditionally. That is because activist
short-sellers are short-sellers in nature, therefore they benefit from declines in stock prices. If some
egregious overvaluation features identified by prior literature could reliably predict future
underperformance (i.e., small Type I error), we would expect that activist short-sellers, like passive
short-sellers with negative information, are more likely to target firms with these features.
Therefore, my second hypothesis is as follows (stated in the alternative form):
13
H2: Activist short-sellers are more likely to target firms with overvaluation features.
Activist short-sellers publicly talk down stocks because they want to engage other
investors. They are more likely to succeed if investors can be easily influenced. Clearly, if investors
are faced with high uncertainty – they are uncertain about the precision of the existing information
signals – they would be easily convinced by the short thesis (i.e., the information view), and they
would also likely be afraid that others would sell first (i.e., the HOB view).11 Therefore, both these
views lead to the following hypothesis (stated in the alternative form):
H3: Activist short-sellers are more likely to target firms with uncertainty features.
2.4 Overvaluation and Uncertainty Features and Short- and Long-
Term Returns (H4)
The two features in the determinants hypotheses also have implications for short- and long-
term price consequences. Both overvaluation and uncertainty features should be associated with
negative market returns given the impact of activist short-selling, simply because stocks with more
overvaluation features have larger downside and investors of stocks with more uncertainty features
are easier for short-sellers to convince. But based on the information view, there is no obvious
reason that the relations between these features and short-term returns would differ from the
relations between these features and long-term returns. However, the HOB view predicts
differently because higher-order beliefs are more important in the short term than in the long
term.12 As a result, short-term returns should be affected more by uncertainty features because
investors facing uncertainty would worry how other people behave and such “worry” (or panic)
matters more for returns in the short term than in the long term. By contrast, no matter whether
panic dominates in the short term, fundamentals determine the price in the long term. As a result,
11 Sophisticated investors, such as insiders and activist short-sellers, are more likely to gain information advantages
over a representative investor under higher uncertainty. For example, Huddart and Ke (2007) find that insider trading
is more profitable under weaker information environment. 12 There are at least two reasons. First, short-term investors care more about the near-term price in other people’s view
than the fundamental value that the price would converge to in the long term. Second, in the short term, investors do
not have sufficient opportunities to communicate among themselves. As Qu’s (2013) experimental study shows,
communication among investors, even cheap talk, can help avoid coordination failures.
14
to the extent overvaluation features capture fundamentals, the HOB view predicts that the
importance of overvaluation features relative to uncertainty features increases from the short term
to the long term. To facilitate interpretation, I list the null form (predicted by the information view)
and the alternative form (predicted by the HOB view) as two competing hypotheses – H4A and
H4B, respectively:
H4-A: For targeted firms, overvaluation (uncertainty) features are comparably
important in determining returns from the short term to the long term.
H4-B: For targeted firms, overvaluation (uncertainty) features are increasingly
(decreasingly) important in determining returns from the short term to the long term.
3 Data
I combine information from Seeking Alpha (SA) and Activist Shorts Research (ASR) to
construct a large sample of activist short-selling from 2006 to 2015. Whereas SA is an ideal
platform for non-celebrity shorts, ASR tracks short-selling campaigns waged by prominent traders.
3.1 Seeking Alpha
Founded in 2004 by former Wall Street analyst David Jackson, SA is the most popular
crowdsourced platform for investment research (Chen, De, Hu, and Hwang 2014), with broad
coverage of stocks, asset classes, ETFs, and investment strategies. According to the information
on the website as of December 30, 2015, SA has four million registered users (with a 48% annual
growth) and 18.5% of the audience are financial professionals.13 As a crowdsourced platform, the
articles on SA are written by contributors and reviewed by the editorial board.14 After acceptance,
the contributors can publish their articles and receive $35 per article plus further compensation
13 See http://seekingalpha.com/page/about_us. 14 The SA editorial team reviews submitted articles “for clarity, consistency and impact.” There are four main editorial
principles: (1) articles interest SA’s readership, (2) articles conform to SA’s standards of rigor and clarity, (3) articles
about a stock trading at less than $1 or with a market cap below $100 million will see extra scrutiny, and (4) authors
must agree in writing to SA’s disclosure standards. See http://seekingalpha.com/page/editorial_principles.
15
based on page views. By the end of 2015, there were 12,354 contributors in total. In 2006 SA
opened a new section called “Short Ideas,” where contributors write articles illustrating why they
are short-selling or plan to short-sell certain securities. Critical to this study in identifying activist
short-sellers, SA “contractually requires all authors to disclose positions in stocks they write
about.”15
3.2 Activist Shorts Research
Founded in 2014 by Adam Kommel, a former analyst at FactSet SharkRepellent, ASR is
an independent database dedicated to tracking activist short-selling campaigns.16 Kommel and his
colleagues analyze campaigns by summarizing the returns, allegations, short-seller tactics,
company rebuts, and actions taken by regulators. ASR complements SA because it tracks celebrity
shorts who post research reports on their own well-known platforms (such as Muddy Waters) and
who disclose their short theses by appearing on business media (such as James Chanos) or
attending investing conferences (such as David Einhorn).
3.3 Sample-Construction Process
To compile the sample, I use Python to crawl all “Short Ideas” articles from the SA website
at http://seekingalpha.com/analysis/investing-ideas/short-ideas. For each SA article, I extract
information regarding the unique article identifier, article title, author name, author self-
descriptions, the stock(s) discussed in the article, the date of posting (I use the next day if an article
is posted after 4:00 p.m. EST – the time when the market closes), and most importantly, the text
of analyses, including the disclosure by the author on whether she is short the stock(s). Panel A of
Appendix A presents a typical SA article. In Panels B through D, I illustrate three scenarios
regarding the author’s relation with the analyzed stock: (1) the author is short the stock, (2) the
author has no position in the stock but may initiate a short position in the next 72 hours, and (3)
15 See http://seekingalpha.com/article/2389665-is-it-wrong-to-take-a-position-in-a-stock-and-then-write-about-it-on-
seeking-alpha. Relatedly, according to Section 17(b) of the Securities Act of 1933, it is unlawful to give trading advice
without disclosing one’s own interest. 16 Kommel generously gave me free access to ASR from November 2015 to June 2016. In July 2016, ASR was
acquired by Activist Insight.
16
the author has no position in the stock and has no plan to initiate a short position in the next 72
hours. For about 2,000 SA articles without any statement among these three cases, I read them
through and manually check whether the authors explicitly mention their positions. I only focus
on the first case as I define activist short-selling as those SA articles in which authors have short
positions in the analyzed stocks.17 This leaves me with 5,716 articles out of 15,072 published from
February 13, 2006 to December 31, 2015, with 6,197 stock-article level observations.
By December 31, 2015, ASR had collected data on 773 campaigns by 98 short-sellers. The
campaign history of these 98 short-sellers is fully covered from 2011 onward but is only selectively
covered before 2010. I manually collected another 172 campaigns by these 98 short-sellers not
included in ASR. Out of these 945 activist short-selling cases (773 + 172), 341 are also available
on SA. For the combined 6,801 activist short-selling cases, 6,081 are matched with PERMNO and
GVKEY and comprise my final sample of activist short-selling. The detailed sample-construction
process is explained in Panel A of Table 1.
Panel B of Table 1 illustrates the distribution of activist short-selling cases by year and by
stock exchange. An increasing number of listed firms are targeted by activist short-sellers.
Specifically, the frequencies are steadily increasing but with clear surges in 2011 (50% more than
2010) and 2013 (90% more than 2012). Regarding the overall distribution across exchanges,
51.5% of the shorted stocks are listed on NASDAQ, 43.8% on NYSE, and the remaining 4.7% on
AMEX.
4 Does Activist Short-Selling Have Larger Market
Reactions Than Passive Short-Selling?
H1 hypothesizes that activist short-selling leads to larger market reactions than comparable
passive short-selling does because of its informativeness and its coordination role that could create
17 In untabulated analyses I find that other two cases (i.e., “May Short” and “No Plan”) have smaller market reactions.
17
panic among other investors. I exploit the fact that short-interest is reported on a fixed schedule
and use the market reactions to short-interest announcements to mimic comparable passive short-
selling. Specifically, the Financial Industry Regulatory Authority (FINRA) collects short-interest
in individual securities on the settlement date twice per month (once per month before September
7, 2007) and the exchanges that list stocks publish the data at 4:00 p.m. eight business days later
(Senchack and Starks 1993; Hu 2016; Kahraman and Pachare 2016).
I use five types of announcements as benchmarks. The first (second) approach focuses on
the short-interest announcements of a firm in the same month, same market cap quintiles of the
same Fama–French 48 industry, and having the closest short-interest ratio (increase in short-
interest) with each targeted firm. The third (fourth) approach focuses on the short-interest
announcements of a firm in the same month, same market cap quintiles of the same Fama–French
48 industry, and having the highest short-interest ratio (increase in short-interest) with each
targeted firm. The final approach constructs a same-firm benchmark – the last short-interest
announcement of each targeted firm five days before the activist short-selling.
Table 2 reports the daily abnormal returns (adjusted by the Fama–French three-factor
model) from three days before to three days after the events (i.e., activist short-selling in Panel A
or short-interest announcements in Panels B to F). The market reacts strongly to activist short-
selling, with the mean (median) abnormal return on day 0 being -0.0156 (-0.058) and a t-value of
-16.64. Moreover, the negative abnormal returns continue to be significant through day 9 except
day 7 (untabulated). By contrast, the market reactions to the five benchmarks are much smaller.
The most negative one is in Panel E – to the short-interest announcement by industry peers with
the highest increase in short-interest, consistent with Diamond and Verrecchia’s (1987) prediction
that unexpected increase in short-interest is bad news. But its AR(0) is about one-tenth and CAR(-
1, 1) is one-quarter of its counterpart in Panel A. These results show that activist short-selling leads
to much larger market reactions than passive short-selling does, thus supporting H1.18
18 The inference remains unchanged if I use alternative market-reaction measures such as volatility and trading
volume. It is worth noting two caveats of using market reactions to short-interest announcements to proxy for market
reactions to passive short-selling. First, to the extent there are industry spillover effects (i.e., industry peers are also
adversely affected by activist short-selling), the above approach could overestimate the impact of passive short-selling.
Second, note that short-interest is announced with a delay (i.e., eight business days after being reported by FINRA
member firms). As a result, the above approach could underestimate the impact of passive short-selling. An alternative
18
5 Determinants of Being Targeted by Activist Short-
Selling
5.1 Variables and Models
5.1.1 Overvaluation Features
Conceptually speaking, overvaluation exists when a security’s price (P) exceeds its
intrinsic value (V) and therefore the price is likely to go down in the future. I consider three broad
sets of features that are potentially associated with future underperformance. The first set only
focuses on the price. For example, Lakonishok, Shleifer, and Vishny (1994) show that the market
tends to overreact to past performance; as a result, a price run-up (PriceRunUp) is associated with
future underperformance. The second set considers both V and P using valuation multiples,
ranging from the simple P/B ratio (e.g., Liu, Nissim, and Thomas 2000) to the more complicated
P/V ratio (Frankel and Lee 1998; Li and Mohanram 2014). The third set focuses only on V by
employing nine anomaly variables from Beneish, Lee, and Nichols (2015): LowGrossProfit
(Novy-Marx 2010), AssetGrowth (Cooper, Gulen, and Schill 2008), Investment (Titman, Wei, and
Xie 2004), net operating assets (NOA) (Hirshleifer, Hou, Teoh, and Zhang 2004), Accruals (Sloan
1996), payout ratio (LowPayout%) (Daniel and Titman 2006), quarterly earnings (LowEarnings)
(Chen, Novy-Marx, and Zhang 2010), Ohlson Bankruptcy Score (OScore) (Ohlson 1980;
Stambaugh, Yu, and Yuan 2012), and the Beneish Manipulation Score (MScore) (Beneish, Lee,
and Nichols 2013). To focus on egregious overvaluation, I transform these continuous variables
into indicators equal to one if the according variables are in the top quintile of overvaluation, and
zero otherwise. Appendix C illustrates how to calculate these variables. A positive coefficient
indicates that the overvaluation feature attracts activist short-sellers, therefore supporting H2.
benchmark would be the market reactions to timely short-position disclosures, which are mandatory in several EU
countries after the recent financial crisis but are not mandatory in the US. However, as Jones, Reed, and Waller (2016)
report, the short-window reaction to large short-position disclosures in EU is very low, i.e. CAR(-1, 1) is an
insignificant -0.41%.
19
5.1.2 Uncertainty Features
As discussed, I predict that activist short-sellers prefer targeting firms whose investors face
higher uncertainty and therefore are easier to influence. I consider three such scenarios. First,
investors are likely to feel uncertain when they are not sophisticated enough to process and analyze
relevant information. In particular, individual investors are likely to feel more uncertain than
institutional investors do (LowInstOwn). Second, in a weak information environment, investors
have insufficient knowledge about what is happening and are uncertain about the precision of
information they have. This could happen either because the internal information systems are not
functioning well, such as low accounting quality (LowAccQuality), internal control weakness
(ICW), or auditor switches (AuditSwitch), or because the disagreement in opinion is severe, such
as high information asymmetry among investors (BidAskSpread), or high analyst disagreement
(AnalystDisagree). Finally, investors are uncertain about the current valuation if they have few
credible information sources, such as blockholders (NonBlock), dedicated institutional investors
(LowDedicated), or Big Four auditors (NonBig4). For the five continuous variables, I transform
them into indicators equal to one if the according variables are in the top quintile of uncertainty,
and zero otherwise. No transformation is needed for ICW, AuditSwitch, NonBlock, and NonBig4
as they are already indicators with one indicating high uncertainty. Appendix C illustrates how to
calculate these variables. Again, a positive coefficient indicates that the uncertainty feature attracts
activist short-sellers, therefore supporting H3.
5.1.3 The Model
For the two determinants hypotheses, I estimate the following Logit model in the main
analyses at the firm-quarter level. I use alternative estimation methods (OLS and Negative
Binomial) in the supplemental section and all inferences remain.
, , , 0 1 , 2 , 3 ,
4 , 5 , 6 ,
(
)
i t j j i t i t i t i t
i t i t i t i t
Target f Determinant Size Leverage Illiquidity
Volatility LnAnalyst ShortInterest IND QTR
(1)
For a given fiscal quarter (labeled as “data quarter” – further illustrated in Appendix B),
Target is one if a firm is targeted by activist short-sellers from 45 days after the fiscal end of the
20
quarter to 45 days after the fiscal end of the next quarter, and zero otherwise. Determinant refers
to one of the overvaluation or uncertainty features at the data quarter. In this way, all determinants
variables are available when activist short-sellers decide whether to target a firm or not.
I include several control variables measured at the data-quarter ends. First, I control for
two basic firm-level characteristics: firm size measured by the log of total assets (Size) and
leverage measured by the ratio of total liabilities to total assets (Leverage). Also, I control for two
market-based variables: illiquidity (Illiquidity) measured by the quarterly mean of Amihud’s
(2002) daily illiquidity measure and volatility measured by the quarterly standard deviation of
daily stock returns (Volatility). In addition, I include the log of one plus the number of analysts
forecasting earnings for the quarter (LnAnalyst). Finally, I include the short-interest ratio at the
end of the quarter to control for overall short-selling activities (ShortInterest). Fama–French 48
industry and quarter fixed effects are included to control for industry-wide and time-specific
factors. All t-statistics are based on standard errors clustered at the firm level. Appendix C provides
more detailed definitions of all variables.
5.2 Determinants Sample
Panel A of Table 3 presents the by-year distribution of the determinants sample. From
Quarter 4, 2005 to Quarter 4, 2015, 1.79% of the firm-quarter observations are targeted by activist
short-sellers. Panel B compares the firm-level characteristics between targeted and non-targeted
firm-quarters. Targeted firm-quarters have significantly larger size, higher leverage, lower
illiquidity, higher volatility, more analysts, and higher short-interest ratio. For overvaluation
features, targeted firm-quarters have higher means for all except NOA, Accruals, and
LowGrossProfit. The uncertainty features present a more mixed picture. Targeted firm-quarters
have higher means for LowAccQuality, AnalystDisagree, ICW, and BidAskSpread. These
univariate analyses are largely consistent with the bivariate correlations in Panel C. Specifically,
Target is positively correlated with all overvaluation variables except Accruals, and with all
uncertainty variables except LowInstOwn, NonBig4, and LowDedicated.
21
5.3 Regression Results
5.3.1 Overvaluation Features
Table 4 shows that activist short-sellers target firms with overvaluation features, providing
broad support to H2. Specifically, 10 out of 12 overvaluation features (except LowGrossProfit and
Accruals) attract activist short-selling.19 As all features are indicator variables, I can also compare
the economic magnitudes of these features. The coefficients range from 0.166 (Investment) to
0.666 (P/V) and 0.781 (P/B). It is not surprising that P/V and P/B have the largest statistical and
economic significances, because these two consider both the current price and the intrinsic value
and therefore potentially better capture the actual overvaluation. In terms of marginal effects, given
all other variables at the mean values, the probability of being targeted by activist short-sellers
increases from 0.66% (0.78%) to 1.44% (0.95%) when we move from stocks in the bottom four
quintiles of P/B (Investment) to stocks in the top quintile of P/B (Investment). As a benchmark, the
unconditional probability that a firm-quarter gets targeted is 1.79%. The economic significances
of all other significant overvaluation features are between those of Investment and P/B. With
respect to control variables, across the 12 columns, activist shorts are more likely to target firms
that are more liquid, more volatile, covered by more analysts, and with higher levels of short-
interest ratio.
5.3.2 Uncertainty Features
Table 5 shows that activist short-sellers target firms with uncertainty features in general,
thus supporting H3. All uncertainty features, except AnalystDisagree, attract activist short-
selling.20 The coefficients range from 0.168 (LowInstOwn) to 0.379 (ICW) and 0.391 (NonBig4),
suggesting that activist short-sellers pay particular attention to auditing-related information. In
terms of marginal effects, given all other variables at the mean values, the probability of being
targeted by activist short-sellers increases from 0.73% (0.78%) to 1.07% (0.92%) when NonBig4
19 One possible reason for the insignificant coefficient on Accruals is that the accruals anomaly has decayed in recent
years (Green, Hand, and Soliman 2011; Mohanram 2014).
20 One possible reason for the insignificant coefficient on AnalystDisagree is related to the fact that activist short-
sellers are more likely to target stocks for which analysts are mostly optimistic (i.e., smaller disagreement).
22
(LowInstOwn) increases from zero to one. These numbers indicate that the impact of uncertainty
features is smaller than that of overvaluation features. The coefficients on control variables are
similar to those in Table 4.
5.3.3 Horseracing Variables and Including Both Features Together
As Table 3, Panel C shows, many overvaluation (uncertainty) features are correlated with
each other. For this reason, I employ stepwise regression to horserace all variables and only keep
those features loading significantly at the 0.10 level. In Table 6, columns 1 and 2 indicate that
seven overvaluation features (PriceRunUp, P/B, P/V, AssetGrowth, NOA, MScore, and
LowEarnings) and six uncertainty features (BidAskSpread, NonBlock, LowDedicated,
LowAccQuality, ICW, and NonBig4) survive.21 In column 3, I put all overvaluation and uncertainty
features together and the same 13 variables survive, suggesting that these two sets of features
explain largely non-overlapping variation. Based on these 13 variables, I create two aggregate
variables: Overvaluation is the average of all seven overvaluation indicators and Uncertainty is
the average of the six uncertainty indicators, respectively.
Column 4 of Table 6 presents results using the aggregate variables and confirms the
previous inferences that H2 and H3 are supported. Specifically, the coefficients on both
Overvaluation and Uncertainty are positive and highly significant. Regarding marginal effects,
given all other variables at the mean values, the probability of being targeted by activist short-
sellers increases from 0.47% (0.59%) to 5.25% (2.17%) when Overvaluation (Uncertainty)
increases from zero to one, an 11-time (four-time) increase in the probability of being targeted.
21 Widely used in accounting and finance (e.g., Ou and Penman 1989; Klassen and Laplante 2012; Titman and Tiu
2011), stepwise regression adds and removes variables one by one based on a pre-determined p-value threshold (in
this test, 0.10). Those variables with coefficients significant at the 0.10 or better levels are kept and otherwise removed.
23
6 Overvaluation and Uncertainty Features and Short-
and Long-Term Returns
6.1 Models and Variables
To test H4-A vs. H4-B regarding price consequences in the short and long term, I estimate
model (2): CAR (cumulative abnormal returns) refers to returns with different windows such as
short-term AR(0) (abnormal return on the disclosure date), CAR(0, 1), CAR(0, 2), medium-term
CAR(1 Week), CAR(1 Month), and long-term CAR(1 Year).22 All CARs are adjusted using the
Fama–French three-factor model.
, 0 1 , 2 , 3 , 4 ,
5 , 6 , 7 , 8 ,
9 , 10 ,
11
[ 5,0] [ 5,0]
[ 5,0
i t i t i t i t i t
i t i t i t i t
i t i t
CAR Overvaluation Uncertainty Size Leverage
Illiquidity Volatility LnAnalyst ShortInterest
EarnAnnounce AnForecast
ConfCall
, ,]i t i t i tIND QTR
(2)
In addition to all control variables in model (1), I also control for three important
information sources that shape the corporate information environment (Beyer, Cohen, Lys, and
Walther 2010): mandatory disclosure, voluntary disclosure, and analyst forecasting. Specifically,
I include three indicator variables in model (2): EarnAnnounce[-5, 0], ConfCall[-5, 0], and
AnForecast[-5, 0] that indicate the existence of earnings announcements, conference calls, and
analyst forecasts in the five days prior to the activist short-selling date, respectively.
6.2 Sample and Regression Results
Panel A of Table 7 provides summary statistics of variables used in the market-reaction
tests. Specifically, AR(0) and all CARs are negative, confirming the observation in Section 4 that
the market reacts negatively to activist short-selling. Panel B presents the regression results. In
22 It is notoriously difficult to estimate long-term abnormal returns because of the “Bad-model problems” (Fama 1998).
Following Fama’s recommendation, I use CARs rather than BHARs in the main test but all inferences remain
unchanged if I calculate one-year abnormal returns using a buy-and-hold approach.
24
short-term windows, such as the abnormal return on the disclosure date (i.e., AR(0)), the coefficient
on Uncertainty is significantly negative at the 1% level. The significance decreases monotonically
as the CAR windows extend and the coefficient becomes insignificant for CAR(1 Year). By
contrast, the coefficient of Overvaluation is insignificant for AR(0), but it becomes increasingly
significant as the CAR windows extend. In particular, it is highly significant for CAR(1 Year).
Moreover, the relative magnitude of the coefficients on Uncertainty and Overvaluation
(i.e., the ratio of coefficients on these two variables) decreases dramatically as the CAR windows
extend. On the disclosure date, compared to firms with none of the uncertainty features (i.e.,
Uncertainty = 0), firms with all six uncertainty features (i.e., Uncertainty = 1) have 4.17% more
negative returns; note the mean of AR(0) is -1.5%. In contrast, the equivalent statistic for
Overvaluation is only 0.59%. The ratio of β1 (i.e., the coefficient on Overvaluation) to β2 (i.e., the
coefficient on Uncertainty) increases dramatically and monotonically from 0.141 for AR(0) (-
0.59%/-4.17%) to 0.636 (-3.54%/-5.57%) for CAR (1 Week), and finally to 8.605 (-88.6%/-10.3%)
for CAR (1 Year). These observations provide support for H4-B based on the HOB view that the
importance of overvaluation features relative to uncertainty features increases from the short to the
long term.
Note that the dependent variables across columns have very different distributions, so I
cannot directly compare coefficients of Overvaluation (Uncertainty) between different columns.
Instead, at the bottom of Panel B, I tabulate results with standardized CARs by decile ranks ranging
from 0 to 1. The coefficients of Overvaluation become monotonically more negative from -0.0344
for AR(0) to -0.2572 for CAR(1 Year), while those of Uncertainty become monotonically less
negative from -0.1685 for AR(0) to -0.0265 for CAR(1 Year). The differences in coefficients
between columns 1 and 6 on both variables are significant at the 5% level (p-value = 0.0001 for
Overvaluation and 0.023 for Uncertainty), providing further support to H4-B.
6.3 An Ex-Ante Approach to Separate Winners vs. Losers among
Activist Short-Selling
The above results have potentially important trading-strategy implications for both activist
short-sellers and other investors who pay attention to short-selling campaigns. To illustrate, I split
25
the whole sample into four groups based on whether Overvaluation or Uncertainty is higher than
the determinant-test sample median (i.e., including both targeted and non-targeted firm-quarters).
Figure 1 plots the mean cumulative abnormal returns in a window of (-60, 250) for targets in these
four groups: High Overvaluation and High Uncertainty (solid line), High Overvaluation and Low
Uncertainty (dotted line), Low Overvaluation and High Uncertainty (short dash line), and Low
Overvaluation and Low Uncertainty (long dash line). Overvaluation features clearly dominate the
return pattern in the long term: the solid and dotted lines substantially trend down while the other
two lines trend up shortly after being targeted. However, in the short term, the uncertainty features
also have some impact. For example, both the solid line and the short dash line present a sudden
drop immediately after the activist short-selling.
As a result, the ideal targets for activist short-sellers would be firms with both
overvaluation and uncertainty features – the solid line drops about 60% over the following year.
If short-sellers can bear the risk of holding short positions for a long time, they can simply ignore
the uncertainty features. By contrast, they can also choose to target firms with high uncertainty
and prepare to exit quickly. Also, Figure 1 can be interpreted as an ex-ante way to separate
legitimate vs. manipulative activist short-selling, an issue exemplified in the opening quote of this
paper. If a firm has egregious overvaluation features, short-selling against this stock is more likely
to be legitimate. By contrast, a short-selling campaign against a stock with little indication of
overvaluation but high uncertainty is more likely to be manipulative.
7 Activist Short-Sellers’ Allegations and Firms’ Tendency
to Respond
Using the rich information provided by ASR, in this section, I explore the implications of
overvaluation and uncertainty features. Specifically, I examine the relations between these features
and short-selling allegations and firms’ tendency to respond after being attacked.
26
7.1 Short-Sellers’ Allegations
ASR identifies the primary allegation for each activist short-selling campaign and classifies
them into 18 types. Table 8 tabulates the frequency, the mean CAR(0, 1), the proportion of firms
that respond to short-sellers, and the mean and median values of Overvaluation and Uncertainty
by each type. It is not surprising that those allegations explicitly focused on valuation/accounting
issues, such as Bubble, Stock Promotion, and Accounting Fraud, have relatively high
Overvaluation, suggesting that short theses provide information regarding the underlying
overvaluation of the targeted firms. Another interesting observation is that severe allegations, such
as Accounting Fraud and Major Business Fraud, have relatively higher Uncertainty and bigger
market reactions, suggesting that activist short-sellers are more likely to raise severe allegations to
create panic among investors who are relatively easier to influence. Also, it is natural to expect
that firms receiving these severe allegations are more likely to respond.
To formally test the relations among stock features, allegations, and firms’ tendency to
respond, I aggregate these five allegations (40.3% of all cases) as “Overvaluation Allegations”:
Accounting Fraud, Bubble, Misleading Accounting, Other-Overvaluation, and Stock Promotion,
and these four allegations (26.6% of all cases) as “Severe Allegations”: Accounting Fraud, Major
Business Fraud, Pyramid Scheme, and Other-Illegal. Columns 1 and 2 of Panel B confirm the
above descriptive observations: Overvaluation features are associated with the allegations
regarding valuation issues, and Uncertainty features are associated with allegations that are
perceived as severe. To the extent that the variable Overvaluation captures the actual
overvaluation, the results in column 1 are consistent with the information view that the short theses
are informative. However, column 2 suggests that short-sellers are likely overstating the problems
for firms with higher uncertainty to convince investors – consistent with the information view, and
to create panic – consistent with the HOB view.
7.2 Firms’ Tendency to React to Short-Sellers’ Allegations
Column 3 of Panel A (Table 8) indicates that 42.7% of targets respond to short-selling
allegations. Those firms usually accuse short-sellers of manipulation and provide additional
relevant information (or misinformation). For example, after being accused by several short-
27
sellers, on October 6, 2015, Valeant released “Valeant Corrects Misleading Reports,” listing
“Facts” to address each “Assertion” from short-sellers.23 It is not clear whether firms with more
overvaluation features are more likely to respond, but firms with more uncertainty features should
be more motivated to respond for at least two reasons: (1) investors of these firms are in greater
need of credible information, and (2) they are more easily influenced – by both short-sellers and
managers. Indeed, column 3 of Panel B (Table 8) indicates that firms with more uncertainty
features are more likely to react to short-selling allegations. Column 4 indicates that firms are also
more likely to respond to severe allegations. These results are consistent with both the information
view and the HOB view, as I cannot separate the possibility that firms provide information in such
responses from the possibility that they only hope to reduce panic.
8 Supplemental Analyses and Robustness Tests
8.1 Separating ASR and SA Samples
All analyses so far are conducted combining both SA and ASR samples. Since ASR covers
celebrity shorts while SA is largely about man-on-the-street shorts, it is not surprising that the
market reactions to the ASR sample are much larger. For example, the disclosure date return is -
1.1% for the SA sample but -5% for the ASR sample (untabulated). More importantly, all
inferences about the determinants and consequences of activist short-selling hold in both samples.
Specifically, Panel A of Table 9 shows that, in both samples, activist short-sellers are more likely
to target firms with overvaluation features and firms with uncertainty features. Panels B and C of
Table 9 show that the importance of overvaluation features relative to uncertainty features in
affecting returns increases from the short to the long term.
23 See https://www.sec.gov/Archives/edgar/data/885590/000119312515337744/d83478dex991.htm.
28
8.2 Overvaluation and Uncertainty Features and Short-Interest
This section examines whether activist short-sellers’ targets are just firms with high
(increase in) short-interest ratio.24 In column 1 of Panel A (Table 10), I regress the short-interest
ratio at the first settlement date 45 days after the fiscal end of the data quarter
(ShortInterest_45Days) on Overvaluation and Uncertainty, with all control variables and fixed
effects in Equation (1). I also consider the possibility that these determinants variables only predict
high short-interest ratio. Then in column 2, I construct a dependent variable named
TopShortInterest_45Days, coded as one if ShortInterest_45Days is among the top 1.79% of all
firms in the same quarter, and zero otherwise (note that 1.79% of firm-quarters are targeted by
activist short-sellers). Similarly, in columns 3 and 4, I use an indicator of the increase in short-
interest ratio in the first 45 days after the end of the data quarter (IncShortInterest) and an indicator
of whether such increase is among the top 1.79% (TopIncShortInterest) as dependent variables. In
all four columns, the coefficients on Overvaluation are significantly positive as expected (but with
much smaller magnitudes than those in Table 4), while those on Uncertainty are all negative,
suggesting that these two features do not predict short-interest ratio or its increase in the same way
as they predict activist short-selling.
8.3 Overvaluation and Uncertainty Features and Analysts’
Recommendations
Analysts’ issuances of Sell (or downgrading) recommendations resembles activist short-
selling in that both are “public talking-down” behaviors. But these two phenomena differ
profoundly because short-sellers are real investors in the analyzed stocks while analysts are not.
This section examines whether Overvaluation and Uncertainty predict Sell or Downgrading
recommendations in the same way as they predict activist short-selling. For a given fiscal quarter,
AnalystSell is one if a firm receives at least one Strong Sell or Sell rating from any of its analysts
24 Anecdotes suggest that this is not the case. For example, veteran activist short-seller Doug Kass claims he has a
strict rule that he would never short a stock with short-interest higher than 8% because “I simply don’t want to get
caught in a short squeeze.” See page 183, “The Most Dangerous Trade” by Richard Teitelbaum. A short squeeze is a
rapid increase in the price of a stock that occurs when there is a lack of supply and an excess of demand for that stock.
Short squeezes result when short-sellers cover their positions on a stock.
29
from 45 days after the end of the fiscal quarter to 45 days after the end of the next fiscal quarter,
and zero otherwise. Downgrade is one if a firm receives at least one downgrade (i.e., from a more
favorable rating to a less favorable one) from any of its analysts from 45 days after the end of the
fiscal quarter to 45 days after the end of the next fiscal quarter, and zero otherwise. Columns 5 and
6 of Panel A (Table 10) show that analysts are less likely to issue Sell and to downgrade ratings
for firms with more overvaluation and uncertainty features – exactly opposite to what activist
short-sellers do. These results also echo Drake, Rees, and Swanson’s (2011) finding that analysts
sometimes recommend stocks with features negatively associated with future returns.
8.4 A Pseudo Test Using Peer Firms Matched on Firm Features
One may argue that firms with different levels of Overvaluation and Uncertainty could
vary in subsequent returns even without being targeted by activist short-sellers. To confirm that
the results in Panel B of Table 7 are attributable to the impact of activist short-selling, I conduct a
pseudo test as follows. For each targeted observation, I match a non-targeted firm with the same
values of Overvaluation and Uncertainty, the same Fama–French 48 industry membership, and
the closest market cap in the month end before the targeted firm’s activist short-selling date. Then
I label the targeted firm’s activist short-selling date as the pseudo date for the matched firm and
rerun Equation (2) using this pseudo sample. Panel B of Table 10 presents the results. The
coefficients on Overvaluation are much smaller in magnitude than those in Table 7, while
coefficients on Uncertainty are largely positive (not even negative as for real targets). Also, the
R2s are all much smaller than their counterparts in Table 7. These results corroborate the causal
inference that activist short-selling leads to the return patterns documented in Section 6.
8.5 The Interaction Effects of Overvaluation and Uncertainty
Features
The main analyses focus on the separate impact of overvaluation and uncertainty features.
I also check the interaction of these two features in both determinants and consequences
regressions. In Table A1, I find some weak evidence of substitution effects in the determinants
model but complementary effects in the price consequences model. Specifically, as Panel A shows,
for activist short-sellers’ decision to target firms (especially for ASR shorts’), Overvaluation
30
(Uncertainty) is less important as Uncertainty (Overvaluation) increases. For price consequences
as shown in Panel B, the negative association between Overvaluation and long-term returns
becomes more negative as Uncertainty increases.
8.6 The Role of Media and Short-Sellers’ Reputation
This section explores two essential elements for activist short-sellers’ success: the publicity
of campaigns and the credibility of short-sellers. First, if a campaign attracts the attention of
mainstream media, it reaches a bigger audience. For each activist short-selling case, I manually
count Factiva Top US Newspapers articles with the names of both the short-seller and the target
company mentioned in the Headline and Lead Paragraph. Seventy-five campaigns from 23 short-
sellers are covered within 30 days after the activist short-selling. Unsurprisingly, media-covered
campaigns have much more negative returns (i.e., mean AR (0) is -8.8%). However, all inferences
remain the same if I only focus on activist short-selling cases that are not covered in mainstream
media, as Table A2 presents. Second, I use the average CAR(0, 1) of a short-seller’s previous
campaigns as a proxy for her reputation. Table A3 shows that there are strong positive associations
between past returns and returns in the current campaign. These results are consistent with both
the information view and the HOB view because reputable short sellers could have both better
information and better coordination of other investors’ beliefs (Gao 2008).
8.7 Multiple Activist Short-Selling Events in the Same Firm-
Quarter
All 6,081 activist short-selling cases target 3,344 firm-quarters, suggesting that many firm-
quarters are attacked more than once. For example, Herbalife (NYSE: HLF) was targeted 47 times
in the second quarter of 2014. In the main analyses, I treat firm-quarters targeted once the same
way to firm-quarters targeted multiple times. Table A4, Panel A shows that all inferences are the
same if I use a Negative Binomial or OLS with the number of times a firm-quarter is targeted as
the dependent variable. For price consequences, all inferences remain the same if I focus on the
first activist short-selling case for each firm-quarter (Panel B).
31
8.8 Other Robustness Checks
First, the patterns documented in this paper are stable over time: all inferences on
determinants and consequences hold in both the first and second half of each sample (Table A5).
Second, I construct the feature variables by using the decile ranks of these variables. In this
specification, NOA, LowEarnings, and LowDedicated are no longer significant at the 0.10 level.
But eight out of 12 overvaluation features and seven out of nine uncertainty features are still
significant at the 0.10 level or better in individual regressions (Table A6). Third, all results remain
quantitatively similar if I construct Overvaluation (Uncertainty) by either averaging all 12
overvaluation (nine uncertainty) features in Panel A of Table A7 or by averaging 10 overvaluation
(eight uncertainty) features that load significantly in Table 4 (Table 5) in Panel B of Table A7.
Fourth, all inferences remain the same if I calculate abnormal returns based on either the market
model or the Fama–French three-factor plus momentum model (Table A8). Fifth, all findings are
robust if I cluster standard errors differently, such as by quarter, by industry, or by firm and quarter
(Table A9). Sixth, my return results hold using only those observations without any analyst
revisions, conference calls, and earnings announcements in the five days prior to the activist short-
selling date (Table A10). Seventh, the inferences on the return tests hold even if I control for short-
seller fixed effects, indicating that the firm characteristics have substantial explaining power in
predicting returns in addition to who the short-seller is (Table A11). Eighth, all return results are
similar if I only use observations with no missing value in any CARs (Table A12). Ninth, Table
A13 emphasizes the role of liquidity. I find that the investors facing higher uncertainty are more
likely to respond to short-seller’s bearish signal only if the liquidity is low – they are afraid that if
they do not sell now, they can only sell at a much lower price later. This test provides further
support for the HOB view. Tenth, Figure A1 plots that we can separate winners vs. losers in both
SA and ASR sample.
9 Conclusion
The existing short-selling literature primarily focuses on the “passive” dimension of short-
selling, although activist short-selling attracts the most intense attention from the public and the
32
media. This dissertation fills the void in the literature by conducting a large-sample study on the
determinants and consequences of activist short-selling. Combining information from Seeking
Alpha (SA) and Activist Shorts Research (ASR), I show that activist short-selling is increasingly
frequent over the past decade, and causes much larger market reactions than passive short-selling
does. For determinants, I find that activist short-sellers are more likely to target firms with
egregious overvaluation features and uncertainty features. For the price consequences of firms
being targeted, the overvaluation (uncertainty) features become increasingly (decreasingly)
important in determining returns as the CAR windows extend from the short term to the long term.
These results are consistent with Higher-Order Beliefs (HOB) theory, which provides a rational
explanation for the commonly held allegations against activist short-sellers that they create panic
in the marketplace.
In addition, I find that stocks with more overvaluation features are more likely to receive
short-selling allegations on valuation issues (e.g., “bubble”), while stocks with more uncertainty
features are more likely to receive severe allegations (e.g., “fraud”). Finally, firms with more
uncertainty features are more likely to respond by accusing short-sellers of manipulation or
providing relevant information to address their allegations. These findings indicate that firm-level
characteristics also have implications on the interactions between short-sellers and their targets.
This study has certain caveats. First, I compile a large sample of activist short-selling by
combining information from SA and ASR. However, I cannot rule out the possibility that some
short-sellers disclose their short theses on platforms other than SA and are under the radar of ASR
(although likely not common). Second, I focus on the roles of overvaluation and uncertainty
features but hold an agnostic view on what causes these features in the first place. I leave a
thorough investigation on the causes of these features to future research. Third, except for the
broad classification of short-selling allegations, I do not dive into the detailed content of short
theses as Chen (2016) and Ljungqvist and Qian (2016) do in their much smaller samples, because
my primary focus is on ex-ante available firm characteristics. I argue that the nature of firms (i.e.,
whether they should be shorted or should not have been shorted) – not who the short-sellers are or
what they say – are the first-order factor determining whether firms are targeted and what are the
return consequences. To avoid distracting from the main messages of this paper, I leave a thorough
examination of the content of short theses to future research. Finally, I only compare two rational
33
frameworks – the information view and the HOB view – when I interpret results. However, I
cannot and do not intend to rule out the possibility that these results can be explained by alternative
theories, especially behavioral ones. For example, noise traders’ overreaction to salient news
(Shiller 1984; Daniel, Hirshleifer, and Subrahmanyam 1998; Lee and So 2014) such as activist
short-selling could provide the same empirical predictions as HOB does.
34
References of Chapter 1
Aitken, M. J., Frino, A., McCorry, M. S., & Swan, P. L. (1998). Short sales are almost
instantaneously bad news: Evidence from the Australian Stock Exchange. The Journal of
Finance, 53(6), 2205-2223.
Allen, F., Morris, S., & Shin, H. S. (2006). Beauty contests and iterated expectations in asset
markets. Review of Financial Studies, 19(3), 719-752.
Amihud, Y. (2002). Illiquidity and stock returns: Cross-section and time-series effects. Journal of
Financial Markets, 5(1), 31-56.
Beneish, M. D., Lee, C. M., & Nichols, D. C. (2013). Earnings manipulation and expected
returns. Financial Analysts Journal, 69(2), 57-82.
Beneish, M. D., Lee, C. M., & Nichols, D. C. (2015). In short supply: Short-sellers and stock
returns. Journal of Accounting and Economics, 60(2), 33-57.
Beyer, A., Cohen, D. A., Lys, T. Z., & Walther, B. R. (2010). The financial reporting environment:
Review of the recent literature. Journal of Accounting and Economics, 50(2), 296-343.
Brent, A., Morse, D., & Stice, E. K. (1990). Short interest: Explanations and tests. Journal of
Financial and Quantitative Analysis, 25(02), 273-289.
Bushee, B. J. (1998). The influence of institutional investors on myopic R&D investment
behavior. The Accounting Review, (73)3, 305-333.
Chen, H., De, P., Hu, Y. J., & Hwang, B. H. (2014). Wisdom of crowds: The value of stock
opinions transmitted through social media. Review of Financial Studies, 27(5), 1367-1403.
Chen, J., Novy-Marx, R., and Zhang, L., (2010). An alternative three-factor model. University of
Rochester working paper.
Chen, L. (2016). The informational role of internet-based short-sellers: The evidence from short
sellers’ reports on US-listed Chinese firms. Journal of Business Finance & Accounting,
43(9-10), 1444-1482.
Comerton-Forde, C., Jones, C. M., & Putniņš, T. J. (2016). Shorting at close range: A tale of two
types. Journal of Financial Economics, 121(3), 546-568.
Cooper, M. J., Gulen, H., & Schill, M. J. (2008). Asset growth and the cross‐section of stock
returns. The Journal of Finance, 63(4), 1609-1651.
Corwin, S. A., & Schultz, P. (2012). A simple way to estimate bid‐ask spreads from daily high and
low prices. The Journal of Finance, 67(2), 719-760.
Daniel, K., Hirshleifer, D., & Subrahmanyam, A. (1998). Investor psychology and security market
under‐and overreactions. Journal of Finance, 53(6), 1839-1885.
Daniel, K., & Titman, S. (2006). Market reactions to tangible and intangible information. The
Journal of Finance, 61(4), 1605-1643.
Dechow, P. M., Hutton, A. P., Meulbroek, L., & Sloan, R. G. (2001). Short-sellers, fundamental
analysis, and stock returns. Journal of Financial Economics, 61(1), 77-106.
Dechow, P. M., Sloan, R. G., & Sweeney, A. P. (1995). Detecting earnings management. The
Accounting Review, 70(2), 193-225.
De Long, J. B., Shleifer, A., Summers, L. H., & Waldmann, R. J. (1990). Noise trader risk in
financial markets. Journal of Political Economy, 703-738.
Desai, H., Krishnamurthy, S., & Venkataraman, K. (2006). Do short sellers target firms with poor
earnings quality? Evidence from earnings restatements. Review of Accounting
Studies, 11(1), 71-90.
35
Diamond, D. W., & Verrecchia, R. E. (1987). Constraints on short-selling and asset price
adjustment to private information. Journal of Financial Economics, 18(2), 277-311.
Drake, M. S., Rees, L., & Swanson, E. P. (2011). Should investors follow the prophets or the
bears? Evidence on the use of public information by analysts and short sellers. The
Accounting Review, 86(1), 101-130.
Engelberg, J., Reed, A. V., & Ringgenberg, M. (2016). Short selling risk. UNC Kenan-Flagler
working paper.
Fama, E. F. (1998). Market efficiency, long-term returns, and behavioral finance. Journal of
Financial Economics, 49(3), 283-306.
Figlewski, S., & Webb, G. P. (1993). Options, short sales, and market completeness. The Journal
of Finance, 48(2), 761-777.
Frankel, R., & Lee, C. M. (1998). Accounting valuation, market expectation, and cross-sectional
stock returns. Journal of Accounting and Economics, 25(3), 283-319.
Gao, P. (2008). Keynesian beauty contest, accounting disclosure, and market efficiency. Journal
of Accounting Research, 46(4), 785-807.
Green, J., Hand, J. R., & Soliman, M. T. (2011). Going, going, gone? The apparent demise of the
accruals anomaly. Management Science, 57(5), 797-816.
Grossman, S. J., & Stiglitz, J. E. (1980). On the impossibility of informationally efficient
markets. The American Economic Review, 70(3), 393-408.
Hirshleifer, D., Hou, K., Teoh, S. H., & Zhang, Y. (2004). Do investors overvalue firms with
bloated balance sheets? Journal of Accounting and Economics, 38, 297-331.
Hu, D. (2016). Does the public availability of market participants’ trading data affect firm
disclosure? Evidence from short sellers. Doctoral dissertation. University of Toronto,
Rotman School of Management.
Huddart, S. J., & Ke, B. (2007). Information asymmetry and cross‐sectional variation in insider
trading. Contemporary Accounting Research, 24(1), 195-232.
Jones, C. M., Reed, A. V., & Waller, W. (2016). Revealing shorts: An examination of large short
position disclosures. Review of Financial Studies, forthcoming.
Kahraman, B., & Pachare, S. (2016). The impact of increased public disclosure policies in the
shorting market. University of Oxford working paper.
Karpoff, J. M., & Lou, X. (2010). Short sellers and financial misconduct. The Journal of
Finance, 65(5), 1879-1913.
Keynes, J. M. (1936). The general theory of interest, employment and money. London: Macmillan.
Khan, M., & Lu, H. (2013). Do short sellers front-run insider sales? The Accounting Review, 88(5),
1743-1768.
Klassen, K. J., & Laplante, S. K. (2012). Are US multinational corporations becoming more
aggressive income shifters? Journal of Accounting Research, 50(5), 1245-1285.
Lakonishok, J., Shleifer, A., & Vishny, R. W. (1994). Contrarian investment, extrapolation, and
risk. The Journal of Finance, 49(5), 1541-1578.
Lamont, O. A. (2012). Go down fighting: Short-sellers vs. firms. Review of Asset Pricing
Studies, 2(1), 1-30.
Lamont, O. A., & Stein, J. C. (2004). Aggregate short interest and market valuations. American
Economic Review, 94(2), 29-32.
Lee, C. M. C., & So, E. C. (2014). Alphanomics: The informational underpinnings of market
efficiency. Foundations and Trends in Accounting, 9(2-3), 59-258.
36
Lee, J. (2013). Activist short sellers: Market manipulators or market protectors? Review of Banking
& Financial Law, 32, 274-565.
Lee, L. F., & Lo, A. K. (2016). Do opinions on financial misstatement firms affect analysts’
reputation with investors? Evidence from reputational spillovers. Journal of Accounting
Research, 54(4), 1111-1148.
Li, K. K., & Mohanram, P. (2014). Evaluating cross-sectional forecasting models for implied cost
of capital. Review of Accounting Studies, 19(3), 1152-1185.
Liu, J., Nissim, D., & Thomas, J. (2002). Equity valuation using multiples. Journal of Accounting
Research, 40(1), 135-172.
Ljungqvist, A., & Qian, W. (2016). How constraining are limits to arbitrage? Evidence from a
recent financial innovation. Review of Financial Studies, 29(8), 1975-2028.
Mohanram, P. S. (2014). Analysts’ cash flow forecasts and the decline of the accruals
anomaly. Contemporary Accounting Research, 31(4), 1143-1170.
Morris, S., & Shin, H. S. (2002). Social value of public information. The American Economic
Review, 92(5), 1521-1534.
Novy-Marx, R., (2010). The other side of value: Good growth and the gross profitability premium.
University of Chicago working paper.
Ohlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy. Journal of
Accounting Research, 18(1), 109-131.
Pownall, G., & Simko, P. J. (2005). The information intermediary role of short sellers. The
Accounting Review, 80(3), 941-966.
Ou, J. A., & Penman, S. H. (1989). Financial statement analysis and the prediction of stock
returns. Journal of Accounting and Economics, 11(4), 295-329.
Park, C. W., & Stice, E. K. (2000). Analyst forecasting ability and the stock price reaction to
forecast revisions. Review of Accounting Studies, 5(3), 259-272.
Qu, H. (2013). How do market prices and cheap talk affect coordination? Journal of Accounting
Research, 51(5), 1221-1260.
Reed, A. V. (2013). Short selling. Annual Review of Financial Economics, 5, 245-258.
Reed, A. V. (2015). Connecting supply, short-sellers and stock returns: Research
challenges. Journal of Accounting and Economics, 60(2), 97-103.
Savor, P. G., & Gamboa-Cavazos, M. (2011). Holding on to your shorts: When do short sellers
retreat? Working paper. Harvard University.
Senchack, A. J., & Starks, L. T. (1993). Short-sale restrictions and market reaction to short-interest
announcements. Journal of Financial and Quantitative Analysis, 28(2), 177-194.
Shiller, R. J. (1984). Stock prices and social dynamics. Brookings Papers on Economic
Activity, 1984(2), 457-510.
Short-selling: Nasty, brutish, and short. (2008, June 19). The Economist. Retrieved from
http://www.economist.com/node/11591349.
Sloan, R. (1996). Do stock prices fully reflect information in accruals and cash flows about future
earnings? The Accounting Review, 71(3), 289-315.
Stambaugh, R. F., Yu, J., & Yuan, Y. (2012). The short of it: Investor sentiment and
anomalies. Journal of Financial Economics, 104(2), 288-302.
Titman, S., & Tiu, C. (2011). Do the best hedge funds hedge? Review of Financial Studies, 24(1),
123-168.
Titman, S., Wei, K. J., & Xie, F. (2004). Capital investments and stock returns. Journal of
Financial and Quantitative Analysis, 39(04), 677-700.
37
Woolridge, J. R., & Dickinson, A. (1994). Short selling and common stock prices. Financial
Analysts Journal, 50(1), 20-28.
38
Appendices of Chapter 1
Appendix A: Activist Short-Selling in Seeking Alpha and Activist Shorts Research This appendix illustrates the two sources of my sample in greater detail. Panel A presents the SA “Short Ideas”
category where I identify my SA sample and a typical SA short-selling article. Panels B to D show three scenarios
regarding whether the SA article author is short the analyzed stock(s) or not.25 Panel E presents a typical newsletter
ASR sends to its subscribers. ASR tracks and summarizes all prominent short-selling campaigns in its newsletters.
Panel A: A typical article on SA “Short Ideas” category
25 These SA articles are available at http://seekingalpha.com/article/3740536-herbalife-moving-underground-
business-opportunity-pitch (Panels A and B), http://seekingalpha.com/article/3784096-mannkinds-bear-thesis-
remains-intact-woes-continue (Panel C), and http://seekingalpha.com/article/3778456-apple-needs-bring-cash-home-
sooner-better (Panel D).
39
Panel B: A typical SA article indicating that the author is short a stock
Panel C: A typical SA article indicating that the author may short a stock
Panel D: A typical SA article indicating that the author has no plan to short a stock
Panel E: An ASR newsletter sent to its subscribers
main body of
the article (this
is the end)
SA contributor’s disclosure that
s/he is short the analyzed stock
SA contributor’s
disclosure that s/he is
not short the analyzed
stock but may short in
the next 72 hours
40
Appendix B: The Activist Short-Selling Date and Financial Reporting Data
This appendix illustrates how I match financial reporting data with the activist short-selling
date. To make sure that the financial statement data are available when an activist short-seller
decides to target a firm, I match each activist short-selling case to the last fiscal quarter ended at
least 45 days prior to the activist short-selling date. That fiscal quarter is labeled as “data quarter”
of a certain activist short-selling case. Suppose A, C, E, and G are the fiscal quarter ending dates
of Q1, Q2, Q3, and Q4. The activist short-selling cases from B to D are matched with Q1 data,
from D to F are matched with Q2 data, from F to H are matched with Q3 data, and from H to I are
matched with Q4 data. There are 45 days between A and B, C and D, E and F, and G and H. If
there is no activist short-selling against one firm from B to D, then the Q1 quarter of this firm is
classified as a non-targeted observation.
41
Appendix C: Variable Definitions
Following Beneish et al. (2015), all income statement and cash flow statement variables are trailing four quarters.
Balance sheet variables are for the most recent quarter. Lagged income statement variables are for quarters t-7 to t-4.
Lagged balance sheet variables are for quarter t-4.
Variables Definitions
Variables in determinants tests: Dependent variables
Target Indicator. For a given fiscal quarter, one if a firm is targeted by activist short-
sellers at least once from 45 days after the fiscal end of the quarter to 45 days
after the fiscal end of the next quarter, and zero otherwise. As illustrated in
Appendix B, Target is one for Q1 if there is at least one activist short-selling
case from B to D.
Variables on Overvaluation Features (all 12 individual continuous variables are transformed
into 0-1 indictors such that all values in the top quintile are defined as one, and zero otherwise)
PriceRunUp The raw return in the one year ending at the fiscal end of the data quarter.
P/B The percentile rank of market cap to book value of equity in each Fama–
French 48 industry at the fiscal end of the data quarter.
P/V The ratio of price to intrinsic value at the fiscal end of the data quarter. The
estimation of intrinsic value follows the Earnings Persistent model in Li and
Mohanram (2014). Specifically, I estimate a cross-sectional forecasting model
, 0 1 , 2 , 3 , , ,*i t i t i t i t i t i tE NegE E NegE E to generate earnings
forecasts for year t+1 to year t+5, where Et+π = (ib-spi)t+π/cshot (where ib, spi,
and csho are Compustat items for income before extraordinary items, special
items, and common shares outstanding, respectively); NegE is an indicator for
negative earnings. Then I assume the abnormal earnings remain constant after
year t+5. The intrinsic value is the aggregate discounted abnormal earnings,
where the cost of capital is assumed to be 12%.
LowGrossProf
it
(-1)*(Salest – Cost of Goods Soldt) /Total assetst
AssetGrowth Total assetst /Total assetst-4
Investment (CAPEXt + Increase in inventoryt)/ Total assetst-4
NOA (Debt in current liabilitiest + Long-term debtt +Total equityt)/ Total assetst
Accruals (Net incomet – Cash from operationst)/ Total assetst
LowPayout% (-1)*Clean Surplus Relation Payoutt / MVEt-4
LowEarnings (-1)*Income before extraordinary itemst/Total assetst
OScore -0.407*size+6.03*tlta–1.43*wcta+0.0757*clca–2.37*nita–1.83*futl+
0.285*intwo–1.72*oeneg–0.521*chin–1.32.
size=ln(Total assetst);
tlta=Total liabilitiest/Total assetst;
wcta=(Current assetst – Current liabilitiest)/ Total assetst;
clca=Current liabilitiest/Current assetst;
nita=Net incomet/Total assetst;
futl=CFOt/Total liabilitiest;
chin=(NIt–NIt-4]/[Abs(NIt)+Abs(NIt-4)];
oeneg=1 if total equityt is negative and 0 otherwise;
intwo=1 if net income is negative in both of the last two years and 0 otherwise
42
MScore -4.84+0.92*dsri+0.528*gmi+0.404*aqi+0.892*sgi+0.115*depi –
0.172*sgai+4.679*tata–0.327*levi
dsri=(Receivablest/Salest)/(Receivablest-4/Salest-4);
gmi=Gross margint-1/Gross margint (Gross margin=1–COGS/Sales);
aqi=(1– (PPEt+CAt)/Assetst]/[1– (PPEt-4+CAt-4)/Assetst-4;
sgi=Salest/Salest-4; depi=[DPt-1/(DPt-1+PPEt-1)]/[DPt/(DPt+PPEt)] (DP=Depreciation);
sgai=(SGAt/Salest)/(SGAt-4/Salest-4);
tata=(IBt – CFOt)/ Total assetst ;
lvgi=Leveraget/Leveraget-4 (Leverage=debt/assets)
Overvaluation The average of the following seven variables that survive from the stepwise
regression: PriceRunUp, P/B, P/V, AssetGrowth, NOA, LowEarnings, and
MScore.
Variables on Uncertainty Features (all continuous variables are transformed into 0-1 indictors
such that all values in the top quintile are defined as one, and zero otherwise)
LowInstOwn (-1)* The institutional ownership at the fiscal end of the data quarter.
LowAccQualit
y
The standard deviation of discretionary accruals from fiscal year t-4 to fiscal
year t. Fiscal year t is the fiscal year consisting of the data quarter.
Discretionary accruals are calculated based on modified Jones model
(Dechow, Sloan, and Sweeney 1995).
AnalystDisagr
ee
The standard deviation of analyst forecast error for each analyst’s last EPS
forecast prior to the data quarter earnings announcements. Forecast error is
scaled by the closing price of the month prior to the earnings announcement.
BidAskSpread The quarterly average of daily bid/ask spread as calculated by Corwin and
Schultz (2012).
NonBIG4 Indicator. One if the auditor in the data quarter is not among the Big Four
auditors.
ICW Indicator. One if internal control weakness is identified in the previous fiscal
year.
AuditSwitch Indicator. One if there is auditor switch in the previous fiscal year.
NonBlock Indictor. One if there is no blockholder (holding >=5% shares) in the company.
LowDedicated (-1)*the number of dedicated institutional investors classified by Bushee’s
website (Bushee 1998).
Uncertain The average of the following six variables that survive from the stepwise
regression: LowAccQuality, BidAskSpread, NonBig4, ICW, NonBlock, and
LowDedicated.
Variables in market-reaction tests
AR(0) Abnormal return on the disclosure date adjusted by Fama–French three-factor
model returns. Factor loadings are estimated in a 110-day pre-event window
ending 30 trading days before the CAR window starts (the same procedure for
all CARs).
CAR(0, 1) Cumulative abnormal return in a two-day window starting from the activist
short-selling date adjusted by Fama–French three-factor model returns.
CAR(0, 2) Cumulative abnormal return in a three-day window starting from the activist
short-selling date adjusted by Fama–French three-factor model returns.
43
CAR(1 Week) Cumulative abnormal return in a five-day window starting from the activist
short-selling date adjusted by Fama–French three-factor model returns.
CAR(1 Month) Cumulative abnormal return in a 22-day window starting from the activist
short-selling date adjusted by Fama–French three-factor model returns.
CAR(1 Year) Cumulative abnormal return in a 250-day window starting from the activist
short-selling date adjusted by Fama–French three-factor model returns.
Control variables
Size The log of total assets at the fiscal end of the data quarter.
Leverage The ratio of total liabilities to total assets at the fiscal end of the data quarter.
Illiquidity The mean of Amihud’s (2002) daily illiquidity measure in the data quarter,
which is measured as the log of one plus the ratio of absolute return and the
dollar trading volume and scaled by 106.
Volatility The quarterly standard deviation of daily return in the data quarter.
LnAnalyst Log of one plus the number of analysts who provide EPS estimates for the data
quarter prior to the earnings announcements.
ShortInterest The ratio of total shares in short position to total shares outstanding based on
the last settlement date prior to the fiscal end of the data quarter.
EarnAnnounce
[-5,0]
Indicator. One if there is an earnings announcement in the period five days
prior to the activist short-selling date.
AnForecast
[-5,0]
Indicator. One if there is at least one analyst forecast in the period five days
prior to the activist short-selling date.
ConfCall
[-5, 0]
Indicator. One if there is at least one conference call in the period five days
prior to the activist short-selling date. 26
Variables used in tests regarding primary allegations and targets’ tendency to respond
Overvaluation
Allegations
Indicator (only for ASR sample). One if the primary allegation from the short-
sellers is one of the following: Bubble, Stock Promotion, Other-
Overvaluation, Misleading Accounting, and Accounting Fraud.
Severe
Allegations
Indicator (only for ASR sample). One if the primary allegation from the short-
sellers is one of the following: Accounting Fraud, Major Business Fraud,
Pyramid Scheme, and Other-Illegal.
Firms Respond Indicator (only for ASR sample). One if the firm responds specifically to the
short-selling allegations.
Variables used in supplement analyses
Target by SA Indicator. One if a firm is targeted by at least one SA activist short-seller from
45 days after the fiscal end of the data quarter to 45 days after the fiscal end
of the next data quarter. Those data quarters that are matched only with ASR
shorts are set as missing.
Target by ASR Indicator. One if a firm is targeted by at least one ASR activist short-seller
from 45 days after the fiscal end of the data quarter to 45 days after the fiscal
end of the next data quarter. Those data quarters that are matched only with
SA shorts are set as missing.
26 Conference call data are collected from http://seekingalpha.com/earnings/earnings-call-transcripts. I thank Jingjing
Wang for sharing the data.
44
ShortInterest_
45Days
The ratio of total shares in short position to total share outstanding based on
the first settlement date at least 45 days after the fiscal end of the data quarter.
TopShortIntere
st_45Days
Indicator. One if ShortInterest_45Days is among the top 1.79% of the quarter,
and zero otherwise. Note 1.79% is the unconditional probability a firm-quarter
is targeted by activist shorts in this study.
IncShortIntere
st
Indicator. One if ShortInterest_45Days is larger than ShortInterest. This
variable captures whether short-interest increases in the 45 days after the fiscal
end of the data quarter.
TopIncShortInt
erest
Indicator. One if ShortInterest_45Days – ShortInterest is among the top 1.79%
of the quarter, and zero otherwise.
AnalystSell Indicator. For a given fiscal quarter, one if a firm receives at least one Sell or
Strong Sell rating from any of its analysts from 45 days after the fiscal end of
the quarter to 45 days after the fiscal end of the next quarter, and zero
otherwise.
Downgrade Indicator. For a given fiscal quarter, one if a firm receives downgrading at
least once from any of its analysts from 45 days after the fiscal end of the
quarter to 45 days after the fiscal end of the next quarter, and zero otherwise.
Downgrading includes all changes from a more favorable rating to a less
favorable rating.
45
Main Tables of Chapter 1
Table 1: Sample of Activist Short-Selling
This table illustrates the construction steps and the distribution of the activist short-selling sample. Panel A explains
the detailed steps through which I construct the sample. In particular, I combine articles from the “Short Ideas” section
in SA, campaigns collected from ASR, and the self-collected campaigns from these ASR short-sellers. Panel B
presents the distribution of the final sample of activist short-selling by year and by stock exchange.
Panel A: Sample-selection steps
SA data (Seeking Alpha) ASR data (Activist Shorts Research)
Sample Selection Steps No. of Obs. Sample Selection Steps No. of Obs.
SA “Short Ideas” articles from
2006 to 2015
15,072 Campaigns from ASR 773
Articles identified as activist
short-selling
5,738 Campaigns hand-collected on
those ASR shorts
172
Activist short-selling cases
(short-seller – stock – date level)
6,197 ASR short-selling cases
(short-seller – stock – date
level)
945
Combining SA and ASR 6,801 (341 are in both SA and ASR)
With PERMNO and GVKEY 6,08127
Panel B: Year distribution of the activist short-selling sample
Year\Exchange NYSE AMEX NASDAQ Total % by year
2006 33 18 40 91 1.50%
2007 100 16 86 202 3.32%
2008 164 46 98 308 5.06%
2009 170 4 112 286 4.70%
2010 158 14 127 299 4.92%
2011 125 21 302 448 7.37%
2012 208 20 347 575 9.46%
2013 447 48 591 1,086 17.86%
2014 622 60 815 1,497 24.62%
2015 635 41 613 1,289 21.20%
Total 2,662 288 3,131 6,081 100%
% by Exchange 43.78% 4.74% 51.49% 100%
27 Most of the stocks missing PERMNO are on OTC or oversea markets. It is worth noting that my sample includes
almost all observations used in Chen (2016) and Ljungqvist and Qian (2016).
46
Table 2: Market Reactions to Activist Short-Selling and Passive Short-Selling28
This table reports the daily abnormal returns (Fama–French three-factor model adjusted) to activist short-selling (Panel
A) and five passive short-selling benchmarks (Panels B to F) that rely on market reactions to short-interest
announcements. Specifically, Panel B (C) reports the abnormal returns to the short-interest announcement by a firm
with the closest (increase in) short-interest ratio, in the same month, Fama–French 48 industry, and same size quintile
with each targeted stock. Panel D (E) reports the abnormal returns to the short-interest announcement by a firm with
the highest (increase in) short-interest ratio, in the same month, Fama–French 48 industry, and same size quintile with
each targeted stock. Panel F reports the abnormal returns to the latest short-interest announcement by each targeted
stock five days before the activist short-selling.
Trading Day -3 -2 -1 0 - Event 1 2 3
Panel A: Market reaction to activist short-selling (N=5,808)
Mean AR -0.0005 0.0008 -0.0015 -0.0156 -0.0054 -0.0013 -0.0016
Median AR -0.0017 -0.0018 -0.0030 -0.0058 -0.0028 -0.0018 -0.0021
t-stat -0.46 0.65 -1.19 -16.64 -7.31 -2.06 -2.52
Panel B: Benchmark 1 – industry peer with the closest short-interest ratio (N=3,061)
Mean AR -0.0006 0.0003 0.0004 -0.0009 -0.0003 0.0003 -0.0010
Median AR -0.0007 -0.0009 -0.0005 -0.0010 -0.0009 -0.0006 -0.0016
t-stat -1.09 0.49 0.83 -1.79 -0.64 0.51 -1.99
Panel C: Benchmark 2 – industry peer with the closest increase in short-interest (N=3,041)
Mean AR -0.0009 -0.0001 0.0000 -0.0005 0.0000 -0.0017 -0.0009
Median AR -0.0008 -0.0009 -0.0005 -0.0009 -0.0003 -0.0012 -0.0011
t-stat -1.70 -0.14 0.03 -1.04 -0.07 -3.34 -1.88
Panel D: Benchmark 3 – industry peer with the highest short-interest ratio (N=2,268)
Mean AR -0.0032 0.0004 0.0002 -0.0016 -0.0015 -0.0017 -0.0018
Median AR -0.0017 -0.0006 -0.0010 -0.0012 -0.0013 -0.0022 -0.0015
t-stat -2.66 0.44 0.27 -1.90 -1.93 -1.70 -1.86
Panel E: Benchmark 4 – industry peer with the highest increase in short-interest (N=2,193)
Mean AR -0.0023 -0.0006 -0.0022 -0.0016 -0.0020 0.0000 -0.0017
Median AR -0.0011 -0.0013 -0.0016 -0.0017 -0.0019 -0.0019 -0.0015
t-stat -1.91 -0.63 -2.44 -1.94 -2.52 -0.02 -1.68
Panel F: Benchmark 5 – the target’s own latest short-interest announcement (N=4,508)
Mean AR 0.0006 -0.0007 0.0003 -0.0001 0.0008 -0.0001 -0.0008
Median AR -0.0009 -0.0010 -0.0007 -0.0010 -0.0015 -0.0018 -0.0017
t-stat 0.85 -0.99 0.45 -0.17 1.00 -0.15 -0.93
28 The median short-interest ratio of Benchmark 1 (Benchmark 3) is 7.2% (26.9%), while the median ratio for the
target firms is 8.5%. The median increase in short-interest of Benchmark 2 (Benchmark 4) is 0.000% (2.103%), while
the median increase for the target firms is 0.007%. Note that the number of observations is not the same in all panels
because (1) one stock can be targeted by activist short-sellers for multiple times, and these cases would match the
same industry peer and same short-interest announcement of its own, and (2) multiple stocks can be in the same size
quintile of the same industry and they would match the same stock with the highest (increase in) short-interest ratio.
47
Table 3: Descriptive Statistics of the Sample for Determinants Tests
This table illustrates the distribution and summary statistics of the sample for determinants tests (firm – quarter level).
Note that the “quarter” is the fiscal quarter of which financial reporting data is used (labeled as “data quarter” in
Appendix B). Panel A shows the by-year distribution of the firm-quarters by whether that are targeted by activist
shorts or not; Panel B compares the determinant-test variables between the firm-quarters that are targeted by activist
shorts and the firm-quarters that are not; Panel C tabulates the Pearson correlations among variables used in the
determinants tests. The correlation coefficients in bold and italic are significant at the 0.05 level. All variables are
defined in the Appendix C.
Panel A: By-year distribution of determinants sample (firm-quarter level)
Year/Type Non-Target Target Total % by Year
2005 5,071 7 5,078 2.72%
2006 19,988 91 20,079 10.75%
2007 19,643 156 19,799 10.60%
2008 19,089 179 19,268 10.32%
2009 17,844 170 18,014 9.64%
2010 17,414 204 17,618 9.43%
2011 17,096 244 17,340 9.28%
2012 16,692 340 17,032 9.12%
2013 16,254 772 17,026 9.12%
2014 16,985 701 17,686 9.47%
2015 17,352 480 17,832 9.55%
Total 183,428 3,344 186,772 100.00%
% by type 98.21% 1.79% 100.00%
48
Panel B: Summary statistics for firm-quarters by whether they are targeted by activist short-sellers
Variables Firm-QTR not targeted Firm-QTR targeted Mean Difference
No. Obs Mean No. Obs Mean
Size 183,428 6.502 3,344 6.967 0.465***
Leverage 183,428 0.502 3,344 0.515 0.013***
Illiquidity 183,428 0.176 3,344 0.033 -0.142***
Volatility 183,428 0.031 3,344 0.033 0.002***
LnAnalyst 183,428 1.402 3,344 1.959 0.557***
ShortInterest 183,428 0.045 3,344 0.103 0.057***
PriceRunUp 183,428 0.196 3,344 0.285 0.089***
P/B 183,428 0.197 3,344 0.374 0.177***
P/V 183,428 0.131 3,344 0.255 0.124***
AssetGrowth 183,428 0.184 3,344 0.304 0.119***
Investment 183,428 0.147 3,344 0.189 0.042***
NOA 183,428 0.200 3,344 0.192 -0.008
Accruals 183,428 0.182 3,344 0.160 -0.022***
LowPayout% 183,428 0.183 3,344 0.222 0.039***
OScore 183,428 0.199 3,344 0.218 0.019***
MScore 183,428 0.175 3,344 0.231 0.056***
LowGrossProfit 183,428 0.186 3,344 0.172 -0.014**
LowEarnings 183,428 0.198 3,344 0.253 0.055***
LowInstOwn 183,428 0.203 3,344 0.179 -0.024***
NonBlock 183,428 0.186 3,344 0.183 -0.003
LowDedicated 183,428 0.263 3,344 0.195 -0.068***
LowAccQuality 183,428 0.118 3,344 0.133 0.015***
AnalystDisagree 183,428 0.130 3,344 0.188 0.058***
NonBig4 183,428 0.233 3,344 0.196 -0.037***
AuditSwitch 183,428 0.048 3,344 0.047 -0.000
ICW 183,428 0.054 3,344 0.064 0.010**
BidAskSpread 183,428 0.199 3,344 0.224 0.025***
49
Panel C: Pearson correlations among variables used in the determinants tests
1 Target 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
2 Size 0.03
3 Leverage 0.01 0.35 4 Illiquidity -0.04 -0.44 -0.03 5 Volatility 0.01 -0.40 0.03 0.44 6 LnAnalyst 0.07 0.56 0.12 -0.44 -0.27 7 ShortInterest 0.10 0.11 0.02 -0.17 -0.01 0.25 8 PriceRunUp 0.03 -0.02 -0.01 -0.09 0.00 0.02 0.04 9 P/B 0.06 -0.06 0.07 -0.09 -0.03 0.13 0.07 0.16 10 P/V 0.05 -0.12 -0.04 0.00 0.10 0.02 0.07 0.03 0.05 11 AssetGrowth 0.04 -0.02 -0.05 -0.08 0.01 0.07 0.09 0.14 0.08 0.02 12 Investment 0.02 0.00 -0.02 -0.03 0.02 0.05 0.04 0.01 0.03 0.00 0.11 13 NOA 0.00 -0.15 -0.29 0.03 0.04 -0.11 -0.03 -0.03 -0.08 0.03 0.07 0.00 14 Accruals -0.01 -0.05 -0.05 0.02 -0.01 -0.06 -0.01 0.03 -0.03 -0.03 0.10 -0.09 0.04 15 LowPayout% 0.01 -0.15 0.00 0.05 0.19 -0.09 0.00 0.09 -0.02 0.10 0.34 0.05 0.09 0.02 16 OScore 0.01 -0.34 0.35 0.25 0.34 -0.23 -0.04 -0.04 0.10 0.16 -0.02 -0.03 0.01 -0.02 0.23 17 MScore 0.02 -0.14 -0.08 0.02 0.07 -0.07 0.01 0.05 0.03 0.05 0.18 -0.01 0.08 0.20 0.11 0.03 18 LowGrossProfit 0.00 -0.03 0.07 0.03 0.11 -0.12 -0.05 -0.04 -0.05 0.11 0.06 -0.05 0.21 0.09 0.24 0.27 0.08 19 LowEarnings 0.02 -0.41 -0.03 0.20 0.35 -0.21 -0.04 -0.06 0.04 0.29 -0.02 -0.01 0.06 -0.07 0.26 0.47 0.04 0.26 20 LowInstOwn -0.01 -0.13 0.02 0.17 0.17 -0.21 -0.11 -0.04 -0.04 0.02 -0.01 -0.02 0.06 -0.02 0.11 0.16 0.05 0.08 0.16
21 NonBlock 0.01 -0.24 -0.09 0.12 0.17 -0.13 -0.02 0.01 0.02 0.06 0.07 0.05 0.00 0.04 0.10 0.12 0.06 -0.03 0.17
22 LowDedicated 0.01 -0.11 0.00 0.09 0.10 -0.11 -0.01 -0.02 -0.02 0.04 0.01 0.01 0.00 0.02 0.04 0.07 0.04 0.00 0.07
23 LowAccQuality 0.02 0.00 0.12 -0.09 0.13 0.15 0.04 -0.07 -0.07 0.15 0.01 0.01 0.03 -0.01 0.13 0.17 0.04 0.14 0.20
24 AnalystDisagree 0.01 -0.44 -0.01 0.32 0.52 -0.33 -0.09 -0.03 -0.02 0.09 -0.02 0.00 0.06 0.00 0.22 0.39 0.07 0.16 0.43
25 NonBig4 -0.01 -0.47 -0.14 0.35 0.21 -0.39 -0.13 0.00 -0.02 0.02 0.03 0.03 0.09 0.08 0.08 0.17 0.07 0.02 0.15
26 AuditSwitch 0.00 -0.10 -0.02 0.08 0.06 -0.10 -0.02 -0.01 -0.01 0.03 0.02 0.00 0.01 0.01 0.03 0.05 0.04 0.00 0.05
27 ICW 0.00 -0.05 -0.02 0.17 0.07 -0.25 -0.10 0.00 -0.01 -0.07 0.00 0.01 0.02 0.02 0.04 0.05 0.05 0.00 0.03
28 BidAskSpread -0.02 -0.32 -0.02 0.34 0.25 -0.35 -0.17 -0.03 -0.04 0.04 -0.01 -0.01 0.07 0.02 0.11 0.21 0.09 0.08 0.20
20 21 22 23 24 25 26 27 21 NonBlock 0.08
22 LowDedicated 0.06 0.08 23 LowAccQuality 0.08 0.04 0.02 24 AnalystDisagree 0.21 0.19 0.10 0.13 25 NonBig4 0.09 0.17 0.12 -0.06 0.25 26 AuditSwitch 0.04 0.05 0.13 0.00 0.07 0.13 27 ICW 0.01 0.04 0.04 -0.06 0.10 0.14 0.03 28 BidAskSpread 0.64 0.13 0.10 0.04 0.29 0.24 0.08 0.10
50
Table 4: The Determinants of Being Targeted by Activist Short-Sellers: Overvaluation Features
This table tests H2 that activist short-sellers are more likely to target firms with overvaluation features. Each column reports Logit regression results regarding one
overvaluation feature. All continuous variables are transformed into 0-1 indictors such that all values in the top quintile are defined as one, and zero otherwise. All
variables are defined in the Appendix C. t statistics in parentheses are based on standard errors clustered by firm. * p < 0.1, ** p < 0.05, *** p < 0.01 (two-sided
tests)
DV=Target (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Overvaluation
Feature =
Price
Runup
P/B P/V Low
Gross
Profit
Asset
Growth
Investment NOA Accruals Low
Payout%
OScore MScore Low
Earnings
Overvaluation 0.354*** 0.781*** 0.666*** 0.147 0.452*** 0.166** 0.125* -0.026 0.203*** 0.258*** 0.407*** 0.389***
Feature (7.46) (11.71) (9.52) (1.56) (6.77) (2.32) (1.78) (-0.39) (3.42) (3.16) (6.73) (5.70)
Size 0.023 0.087*** 0.037 0.018 0.029 0.023 0.019 0.019 0.023 0.037 0.029 0.043
(0.81) (3.08) (1.31) (0.63) (1.03) (0.81) (0.67) (0.65) (0.80) (1.26) (1.02) (1.49)
Leverage -0.184 -0.396*** -0.202 -0.183 -0.138 -0.169 -0.127 -0.178 -0.184 -0.372** -0.148 -0.242*
(-1.40) (-2.79) (-1.55) (-1.39) (-1.05) (-1.29) (-0.96) (-1.35) (-1.40) (-2.45) (-1.13) (-1.85)
Illiquidity -1.822*** -1.738*** -1.819*** -1.907*** -1.815*** -1.905*** -1.903*** -1.909*** -1.893*** -1.935*** -1.861*** -1.919***
(-7.57) (-7.64) (-7.50) (-7.61) (-7.49) (-7.62) (-7.62) (-7.63) (-7.57) (-7.63) (-7.57) (-7.51)
Volatility 25.125*** 26.822*** 23.887*** 24.694*** 25.088*** 25.163*** 25.025*** 25.149*** 24.413*** 24.268*** 24.738*** 23.030***
(20.08) (21.31) (19.14) (19.49) (20.04) (20.11) (20.04) (20.02) (19.59) (19.25) (19.78) (17.75)
LnAnalyst 0.393*** 0.277*** 0.344*** 0.393*** 0.368*** 0.380*** 0.392*** 0.388*** 0.391*** 0.389*** 0.391*** 0.382***
(8.59) (6.73) (8.14) (8.59) (8.40) (8.44) (8.56) (8.51) (8.55) (8.55) (8.56) (8.45)
ShortInterest 3.336*** 3.308*** 3.298*** 3.360*** 3.220*** 3.345*** 3.363*** 3.361*** 3.344*** 3.346*** 3.300*** 3.364***
(15.32) (15.53) (15.23) (15.34) (14.63) (15.21) (15.33) (15.29) (15.20) (15.11) (14.75) (15.30)
FF 48 FE YES YES YES YES YES YES YES YES YES YES YES YES
QTR FE YES YES YES YES YES YES YES YES YES YES YES YES
Constant -8.131*** -8.566*** -8.111*** -8.010*** -8.163*** -8.040*** -8.058*** -7.996*** -8.037*** -8.057*** -8.160*** -8.123***
(-13.45) (-14.57) (-13.46) (-13.38) (-13.51) (-13.47) (-13.49) (-13.39) (-13.46) (-13.50) (-13.60) (-13.61)
Observations 181,76729 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767
Pseudo R2 0.145 0.154 0.149 0.143 0.146 0.143 0.143 0.143 0.144 0.144 0.145 0.145
29 The number of observations is smaller than that in Panel B of Table 3, because Logit model with industry FE only uses those industries with at least one
observation being targeted. This explanation also applies to subsequent tables with Logit models.
51
Table 5: The Determinants of Being Targeted by Activist Short-Sellers: Uncertainty Features
This table tests H3 that activist short-sellers are more likely to target firms with uncertainty features. Each column reports Logit regression results regarding one
uncertainty feature. All continuous variables are transformed into 0-1 indictors such that all values in the top quintile are defined as one, and zero otherwise. All
variables are defined in Appendix C. t statistics in parentheses are based on standard errors clustered by firm. * p < 0.1, ** p < 0.05, *** p < 0.01 (two-sided tests)
DV=Target (1) (2) (3) (4) (5) (6) (7) (8) (9)
Uncertainty
Feature =
Low
InstOwn
Non
Block
Low
Dedicated
Low
AccQuality
Analyst
Disagree
NonBig4 Audit
Switch
ICW BidAskSpread
Uncertainty 0.1683** 0.3502*** 0.2395*** 0.2567*** -0.0270 0.3913*** 0.2879*** 0.3790*** 0.2554***
Feature (2.30) (5.03) (3.27) (3.19) (-0.41) (4.91) (2.77) (3.91) (3.58)
Size 0.0155 0.0062 0.0215 0.0265 0.0182 0.0437 0.0198 0.0209 0.0267
(0.55) (0.23) (0.76) (0.93) (0.64) (1.49) (0.70) (0.74) (0.93)
Leverage -0.1792 -0.1450 -0.1896 -0.1823 -0.1702 -0.1594 -0.1780 -0.1860 -0.1984
(-1.36) (-1.12) (-1.44) (-1.37) (-1.28) (-1.21) (-1.35) (-1.41) (-1.50)
Illiquidity -1.9437*** -1.9655*** -1.9864*** -1.9206*** -1.9143*** -2.0243*** -1.9160*** -1.9228*** -1.9592***
(-7.64) (-7.68) (-7.70) (-7.61) (-7.65) (-7.77) (-7.65) (-7.66) (-7.57)
Volatility 24.6971*** 24.7225*** 24.6263*** 24.8857*** 25.3041*** 25.1617*** 25.1147*** 24.8520*** 22.6003***
(19.56) (19.62) (19.71) (19.76) (19.68) (20.08) (20.01) (19.73) (17.21)
LnAnalyst 0.4028*** 0.4263*** 0.4055*** 0.3889*** 0.3898*** 0.4100*** 0.3928*** 0.3958*** 0.3982***
(8.61) (9.22) (8.71) (8.55) (8.43) (8.85) (8.59) (8.68) (8.65)
ShortInterest 3.3694*** 3.3607*** 3.3981*** 3.3624*** 3.3607*** 3.3728*** 3.3456*** 3.3498*** 3.3651***
(15.21) (14.75) (15.41) (15.36) (15.30) (15.39) (15.13) (15.27) (15.25)
FF 48 FE YES YES YES YES YES YES YES YES YES
QTR FE YES YES YES YES YES YES YES YES YES
Constant -8.0300*** -8.0200*** -8.0907*** -8.0488*** -8.0039*** -8.2710*** -8.0429*** -8.0814*** -8.0323***
(-13.45) (-13.58) (-13.56) (-13.50) (-13.41) (-13.89) (-13.49) (-13.58) (-13.48)
Observations 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767
Pseudo R2 0.143 0.144 0.144 0.144 0.143 0.145 0.143 0.144 0.143
52
Table 6: The Determinants of Being Targeted by Activist Short-Sellers: Stepwise
Regressions and Including Both Features Together
This table uses stepwise regressions and considers two sets of determinants together: overvaluation features and
uncertainty features. In columns 1–3, I use stepwise Logit regressions, removing all variables that are insignificant at
the 0.10 level. Specifically, seven overvaluation features survive in column 1, six uncertainty features survive in
column 2, and the same 13 variables survive in column 3, where I put all overvaluation and uncertainty features
together. In column 4, I use two aggregate variables, which are the average of the seven overvaluation features and
six uncertainty features, respectively. All variables are defined in Appendix C. t statistics in parentheses are based on
standard errors clustered by firm. * p < 0.1, ** p < 0.05, *** p < 0.01 (two-sided tests)
DV=Target (1) (2) (3) (4)
Only
Overvaluation
Features
Only
Uncertainty
Features
Both Features:
Individual
Measures
Both Features:
Aggregate
Measures
PriceRunUp 0.1560*** 0.1619***
(3.34) (3.47)
P/B 0.6898*** 0.6967***
(11.02) (11.16)
P/V 0.5137*** 0.5465***
(7.51) (7.99)
AssetGrowth 0.3079*** 0.2748***
(4.87) (4.36)
LowEarnings 0.3098*** 0.2645***
(4.51) (3.85)
NOA 0.1328* 0.1139*
(1.92) (1.65)
MScore 0.3071*** 0.2798***
(4.94) (4.51)
BidAskSpread 0.1808** 0.1806**
(2.53) (2.57)
NonBlock 0.3036*** 0.2680***
(4.33) (3.87)
LowDedicated 0.1884** 0.2241***
(2.53) (3.08)
LowAccQuality 0.2016** 0.1414*
(2.50) (1.80)
ICW 0.3039*** 0.2751***
(3.13) (2.82)
NonBIG4 0.3288*** 0.3274***
(4.08) (4.08)
Overvaluation 2.4654***
(14.94)
Uncertainty 1.3267***
(7.79)
Control variables YES YES YES YES
FF 48 FE YES YES YES YES
QTR FE YES YES YES YES
Constant -8.9839*** -8.4275*** -9.3845*** -9.3046***
(-15.05) (-14.33) (-15.95) (-15.45)
Observations 181,767 181,767 181,767 181,767
Pseudo R2 0.164 0.148 0.168 0.164
53
Table 7: Market Reactions to Activist Short-Selling
This table presents details for the market-reaction tests. Panel A presents summary statistics for variables used in the
market-reactions tests; Panel B shows the regression results, with CAR windows extending from AR(0) in column 1
to CAR(1 Year) in column 6. At the bottom of Panel B, I briefly report the results using standardized CARs as
dependent variables. All variables are defined in the Appendix C. t statistics in parentheses are based on standard
errors clustered by firm. * p < 0.1, ** p < 0.05, *** p < 0.01 (two-sided tests)
Panel A: Market reactions to activist short-selling: summary statistics30
Variables Obs. Mean STD 1st Quartile Median 3rd Quartile
AR(0) 5,702 -0.015 0.060 -0.029 -0.005 0.010
CAR(0,1) 5,701 -0.020 0.074 -0.041 -0.008 0.014
CAR(0,2) 5,699 -0.022 0.085 -0.049 -0.010 0.017
CAR(1 Week) 5,694 -0.026 0.104 -0.064 -0.013 0.024
CAR(1 Month) 5,629 -0.042 0.199 -0.127 -0.027 0.060
CAR(1 Year) 4,474 -0.257 1.187 -0.742 -0.117 0.402
Overvaluation 5,702 0.300 0.211 0.143 0.286 0.429
Uncertainty 5,702 0.155 0.198 0.000 0.167 0.167
Size 5,702 7.251 2.249 5.665 7.247 8.775
Leverage 5,702 0.555 0.300 0.310 0.558 0.770
Illiquidity 5,702 0.025 0.127 0.000 0.001 0.006
Volatility 5,702 0.034 0.018 0.021 0.029 0.041
LnAnalyst 5,702 2.101 1.075 1.386 2.197 2.996
ShortInterest 5,702 0.133 0.177 0.021 0.075 0.187
EarnAnnounce[-5,0] 5,702 0.119 0.324 0.000 0.000 0.000
AnForecast[-5,0] 5,702 0.410 0.492 0.000 0.000 1.000
ConfCall[-5, 0] 5,702 0.014 0.116 0.000 0.000 0.000
30 I require non-missing values for AR(0) and all control variables. That explains why I have 5,702 observations in
Table 7 but 5,808 in Panel A of Table 2. Also, all inferences remain if I require non-missing values for all return
measures, i.e., from AR(0) to CAR(1Year).
54
Panel B: Market reactions to activist short-selling: regression results (1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1 Week)
CAR
(1 Month)
CAR
(1 Year)
Overvaluation (β1) -0.0059 -0.0113* -0.0237*** -0.0354*** -0.1000*** -0.8863***
(-1.16) (-1.88) (-3.65) (-4.30) (-5.44) (-5.92)
Uncertainty (β2) -0.0417*** -0.0481*** -0.0469*** -0.0557*** -0.0774*** -0.1030
(-5.74) (-5.61) (-3.80) (-3.53) (-3.17) (-0.48)
Size 0.0016** 0.0027*** 0.0026*** 0.0028** 0.0064** 0.0727***
(2.46) (3.30) (2.79) (2.26) (2.53) (3.10)
Leverage 0.0045 0.0014 -0.0009 -0.0000 0.0112 -0.0528
(1.09) (0.27) (-0.16) (-0.00) (0.66) (-0.31)
Illiquidity -0.0094 -0.0184 -0.0169 -0.0129 -0.0094 0.0090
(-0.70) (-1.53) (-1.07) (-0.69) (-0.24) (0.03)
Volatility 0.0081 -0.1568* -0.1559 -0.2835* -0.6668* -10.3406***
(0.12) (-1.66) (-1.39) (-1.90) (-1.70) (-3.27)
LnAnalyst 0.0060*** 0.0067*** 0.0086*** 0.0096*** 0.0199*** 0.0280
(4.76) (3.99) (4.46) (3.98) (3.84) (0.71)
ShortInterest 0.0053 0.0083 0.0118* 0.0202** 0.0401* 0.3076
(1.08) (1.23) (1.67) (2.06) (1.69) (1.41)
EarnAnnounce[-5,0] 0.0047 0.0000 0.0008 0.0018 -0.0055 0.0393
(1.30) (0.00) (0.13) (0.27) (-0.62) (0.64)
AnForecast[-5,0] -0.0026 -0.0028 -0.0020 -0.0053 -0.0048 -0.0666
(-1.36) (-1.13) (-0.69) (-1.37) (-0.72) (-1.57)
ConfCall[-5, 0] 0.0040 0.0138 0.0149 0.0124 0.0108 0.0385
(0.54) (1.27) (1.31) (0.99) (0.45) (0.35)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0246* -0.0301*** -0.0295*** -0.0503** -0.1027 -0.0633
(-1.85) (-3.76) (-2.87) (-2.17) (-1.57) (-0.28)
Observations 5,702 5,701 5,699 5,694 5,629 4,474
Adjusted R2 0.075 0.091 0.083 0.080 0.089 0.182
β1/ β2 0.141 0.235 0.505 0.636 1.292 8.605
Standardized CARs as dependent variables (i.e., decile ranks ranging from zero to one)
Overvaluation (β1) -0.0344 -0.0442* -0.0867*** -0.1166*** -0.1530*** -0.2572***
(-1.51) (-1.90) (-3.72) (-4.95) (-5.35) (-6.80)
Diff with Column 1 p = 0.763 p = 0.108 p = 0.012 p = 0.001 p < 0.001
Uncertainty (β2) -0.1685*** -0.1619*** -0.1165*** -0.1077*** -0.0997*** -0.0265
(-5.03) (-5.03) (-3.28) (-3.04) (-2.79) (-0.50)
Diff with Column 1 p = 0.886 p = 0.287 p = 0.212 p = 0.160 p = 0.023
β1/ β2 0.204 0.273 0.744 1.083 1.535 9.706
55
Table 8: Do Overvaluation and Uncertainty Features Predict Short-Sellers’ Primary
Allegations and Firms’ Tendency to Respond?
This table considers the relations among firm features, short-selling allegations, and firms’ responses. Panel A
tabulates the basic statistics by each type of primary allegation. Panel B reports regression results regarding how
overvaluation and uncertainty features affect short-selling allegations and firms’ tendency to respond.
Panel A: Descriptive statistics by each allegation group
Primary Allegation
N. of
Obs.
CAR
(0, 1)
Proportion of Firms
that Respond
Overvaluation Uncertainty
Mean Median Mean Median
Accounting fraud 58 -0.098 0.828 0.333 0.286 0.325 0.333
Bubble 35 -0.051 0.171 0.469 0.429 0.205 0.167
Competitive pressures 31 -0.028 0.097 0.323 0.286 0.199 0.167
Dividend cut coming 6 -0.054 0.500 0.167 0.143 0.083 0.083
Industry issues 47 -0.026 0.191 0.219 0.143 0.113 0.000
Ineffective roll-up 18 -0.089 0.556 0.262 0.286 0.120 0.167
Major business fraud 68 -0.115 0.882 0.252 0.286 0.400 0.333
Medical effectiveness 65 -0.055 0.215 0.519 0.571 0.254 0.167
Misleading accounting 53 -0.041 0.340 0.270 0.286 0.179 0.167
Other – Illegal 39 -0.063 0.692 0.352 0.286 0.239 0.167
Other – Overvaluation 74 -0.044 0.257 0.324 0.286 0.245 0.167
Over-levered 29 -0.050 0.414 0.217 0.143 0.253 0.167
Patent expiration 5 -0.136 0.400 0.514 0.714 0.200 0.167
Patent invalid 14 -0.001 0.357 0.316 0.286 0.143 0.000
Product ineffective 35 -0.062 0.314 0.367 0.286 0.352 0.333
Pyramid scheme 5 -0.112 1.000 0.200 0.143 0.000 0.000
Stock promotion 38 -0.092 0.553 0.414 0.429 0.395 0.500
Upcoming earnings miss 20 -0.011 0.000 0.257 0.286 0.217 0.167
Total 640 -0.061 0.427 0.332 0.286 0.254 0.167
Panel B: Logit regression results (1) (2) (3) (4)
Overvaluation
Allegations
Severe
Allegations
Firms Respond Firms Respond
Overvaluation 1.1222* -0.5868 0.3207 0.5991
(1.94) (-0.75) (0.44) (0.76)
Uncertainty -0.6451 1.8736** 3.4145*** 3.3159***
(-1.04) (2.52) (4.70) (4.27)
Overvaluation Allegations 0.2942
(1.03)
Severe Allegations 2.5671***
(7.14)
Control Variables Yes Yes Yes Yes
FF 48 FE Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes
Constant -0.0323 -1.7260 -2.1368 -2.8438
(-0.03) (-1.34) (-1.31) (-1.55)
Observations 601 579 561 561
Pseudo R2 0.191 0.295 0.199 0.304
56
Table 9: Comparing SA Sample and ASR Sample: Determinants and Consequences
This table presents regression results focusing on only either SA sample or ASR sample. Panel A examines the
determinants of being targeted by SA (column 1) or ASR (column 2) activist shorts. Panel B presents regression results
regarding the market reactions to SA activist short-selling. Panel C presents regression results regarding the market
reactions to ASR activist short-selling. In both panels, CAR windows extend from AR(0) in column 1 to CAR(1 Year)
in column 6. All variables are defined in Appendix C. t statistics in parentheses are based on standard errors clustered
by firm. * p < 0.1, ** p < 0.05, *** p < 0.01 (two-sided tests)
Panel A: Determinants of being targeted by SA or ASR short-sellers
(1) (2)
Target by SA Target by ASR
Overvaluation 2.4372*** 2.6786***
(13.98) (11.59)
Uncertainty 1.2659*** 1.5717***
(7.15) (6.27)
Size 0.1310*** -0.0765*
(4.51) (-1.69)
Leverage -0.0630 -0.4789**
(-0.49) (-2.24)
Illiquidity -1.6754*** -3.3598***
(-6.96) (-4.65)
Volatility 19.7637*** 9.2944***
(14.86) (3.74)
LnAnalyst 0.3904*** 0.0350
(8.65) (0.59)
ShortInterest 3.0794*** 3.6662***
(14.30) (13.24)
FF 48 FE Yes Yes
QTR FE Yes Yes
Constant -10.2821*** -7.7433***
(-13.33) (-8.37)
Observations 181,767 175,991
Pseudo R2 0.167 0.157
57
Panel B: Consequences of activist short-selling in the SA sample (1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1 Week)
CAR
(1 Month)
CAR
(1 Year)
Overvaluation (β1) -0.0038 -0.0067 -0.0187*** -0.0295*** -0.0918*** -0.8751***
(-0.84) (-1.21) (-3.13) (-3.83) (-5.09) (-5.69)
Uncertainty (β2) -0.0390*** -0.0480*** -0.0470*** -0.0543*** -0.0742*** -0.0927
(-5.14) (-5.41) (-3.48) (-3.16) (-2.89) (-0.40)
Control variables Yes Yes Yes Yes Yes Yes
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0250*** -0.0181* -0.0103 0.0041 -0.0545 -0.2656
(-3.58) (-1.94) (-0.97) (0.31) (-1.24) (-1.23)
Observations 5,243 5,242 5,240 5,235 5,177 4,117
Adjusted R2 0.067 0.081 0.076 0.075 0.080 0.178
β1/ β2 0.097 0.140 0.398 0.543 1.237 9.440
Panel C: Consequences of activist short-selling in the ASR sample (1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1 Week)
CAR
(1 Month)
CAR
(1 Year)
Overvaluation (β1) -0.0072 -0.0257 -0.0256 -0.0383 -0.1433*** -1.1476***
(-0.41) (-1.23) (-1.15) (-1.49) (-3.04) (-3.53)
Uncertainty (β2) -0.0630*** -0.0450* -0.0343 -0.0400 -0.1493** 0.0133
(-3.39) (-1.81) (-1.22) (-1.29) (-2.51) (0.04)
Control variables Yes Yes Yes Yes Yes Yes
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0113 -0.0061 -0.0154 -0.0461 0.0364 0.4931
(-0.34) (-0.20) (-0.50) (-1.10) (0.33) (1.05)
Observations 753 753 753 753 745 576
Adjusted R2 0.089 0.105 0.099 0.122 0.133 0.222
β1/ β2 0.114 0.571 0.746 0.958 0.960 -86.29
58
Table 10: Several Sets of Pseudo Analyses
This table reports several sets of pseudo analyses. Panel A presents regression results on how overvaluation and
uncertainty features predict short-interest ratio, increase in short-interest, and unfavorable analyst recommendations.
Column 1 is OLS regression with short-interest ratio as the dependent variable; columns 2 to 6 are Logit regressions
with the indicators of top short-interest ratio (top 1.79% in the quarter), of short-interest increase, of top increase in
short-interest (top 1.79% in the quarter), of Sell recommendations, and of analyst downgrading as dependent variables,
respectively. Panel B reports market reaction tests using pseudo targets matched on Overvaluation, Uncertainty, size,
and industry, with CAR windows extending from AR(0) in column 1 to CAR(1 Year) in column 6. All variables are
defined in Appendix C. t statistics in parentheses are based on standard errors clustered by firm. * p < 0.1, ** p < 0.05,
*** p < 0.01 (two-sided tests)
Panel A: Do overvaluation and uncertainty predict short-interest and analyst recommendations? (1) (2) (3) (4) (5) (6)
ShortInterest
_45Days
(OLS)
TopShort
Interest_45Days
(Logit)
IncShort
Interest
(Logit)
TopIncShort
Interest
(Logit)
Analyst
Sell
(Logit)
Down
grade
(Logit)
Overvaluation 0.0053*** 0.3003* 0.5750*** 0.8554*** -0.2820*** -0.4711***
(3.82) (1.71) (16.64) (6.12) (-3.66) (-5.31)
Uncertainty -0.0105*** -0.0116 -0.2890*** -0.5943*** -0.1450 -0.2481**
(-6.51) (-0.06) (-8.08) (-3.79) (-1.40) (-2.07)
Size 0.0006** 0.0239 0.0074* 0.1354*** 0.1892*** 0.1600***
(2.56) (0.76) (1.79) (6.37) (14.69) (11.32)
Leverage 0.0105*** 0.7121*** 0.0459** 0.4293*** -0.1268* -0.1780**
(8.71) (6.43) (2.08) (4.28) (-1.72) (-2.23)
Illiquidity 0.0014 0.4043** 0.0434*** 0.1746** -1.2684*** -1.2590***
(0.68) (2.54) (2.72) (2.06) (-6.96) (-6.17)
Volatility 0.3777*** 19.5760*** 0.8443** 22.3793*** 9.7842*** 9.2408***
(14.90) (12.48) (2.30) (18.64) (11.49) (9.82)
LnAnalyst -0.0017*** -0.3230*** 0.0484*** -0.0623* 0.8469*** 0.8556***
(-4.08) (-6.29) (6.64) (-1.79) (33.40) (28.17)
ShortInterest 0.8925*** 16.3190*** -2.4705*** 3.5261*** 2.2216*** 1.9154***
(57.52) (20.32) (-23.05) (11.81) (12.18) (11.84)
FF 48 FE YES YES YES YES YES YES
QTR FE YES YES YES YES YES YES
Constant -0.0044 -7.4074*** -0.0162 -5.2364*** -5.2678*** -6.7353***
(-0.91) (-11.75) (-0.18) (-15.43) (-21.02) (-20.66)
Observations 186,772 186,772 186,772 186,772 186,772 186,772
Adjusted/Pseudo R2 0.468 0.441 0.031 0.069 0.179 0.151
Panel B: Return consequences of pseudo targets (1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR(1 Week) CAR(1 Month) CAR (1 Year)
Overvaluation (β1) -0.0032 -0.0095*** -0.0156*** -0.0166*** -0.0666*** -0.5530***
(-1.48) (-3.34) (-4.39) (-3.12) (-4.74) (-5.10)
Uncertainty (β2) 0.0052* 0.0035 0.0076 0.0042 0.0302 -0.0346
(1.85) (0.80) (1.33) (0.53) (1.59) (-0.24)
Control variables Yes Yes Yes Yes Yes Yes
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant 0.0021 0.0026 -0.0011 0.0632 -0.0223 -0.2236
(0.44) (0.31) (-0.13) (1.08) (-0.35) (-0.79)
Observations 4,722 4,721 4,719 4,714 4,661 3,681
Adjusted R2 0.004 0.005 0.015 0.013 0.045 0.110
59
Main Figures of Chapter 1
Figure 1: Mean Cumulative Abnormal Returns for Targets with High and Low
Overvaluation and Uncertainty
This figure plots the mean cumulative abnormal returns in a window of (-60, 250) for four groups of targets with
Overvaluation or Uncertainty higher or lower than the medians of the determinant-test sample (i.e., including both
targeted and non-targeted firm-quarters). The long dash line represents targets with Low Overvaluation and Low
Uncertainty, the short dash line represents targets with Low Overvaluation and High Uncertainty, the dotted line
represents targets with High Overvaluation and Low Uncertainty, and the solid line represents targets with High
Overvaluation and High Uncertainty. The abnormal return is calculated benchmarking on the Fama–French three-
factor model.
60
Additional Robustness Tests
Table A1: The Interaction between Overvaluation and Uncertainty Features
This table presents the interaction effects of overvaluation and uncertainty features in the determinants and
consequences of activist short-selling. Panel A reports results on determinant-test models and Panel B reports results
on the return consequences.
Panel A: Determinants test
(1) (2) (3)
Target Target by SA Target by ASR
Overvaluation × Uncertainty 0.1762 0.2370 -1.6254*
(0.30) (0.38) (-1.92)
Overvaluation 2.4363*** 2.3985*** 3.0481***
(11.51) (10.78) (9.87)
Uncertainty 1.2706*** 1.1906*** 2.1584***
(4.94) (4.47) (5.35)
Size 0.1263*** 0.1310*** -0.0743
(4.46) (4.51) (-1.64)
Leverage -0.1430 -0.0612 -0.4857**
(-1.12) (-0.47) (-2.27)
Illiquidity -1.7937*** -1.6680*** -3.4761***
(-7.28) (-6.97) (-4.67)
Volatility 18.8672*** 19.7758*** 9.2793***
(14.37) (14.85) (3.75)
LnAnalyst 0.3746*** 0.3904*** 0.0296
(8.76) (8.65) (0.50)
ShortInterest 3.1826*** 3.0813*** 3.6491***
(14.89) (14.31) (13.07)
FF 48 FE Yes Yes Yes
QTR FE Yes Yes Yes
Constant -9.2984*** -10.2740*** -7.8543***
(-15.40) (-13.30) (-8.49)
Observations 181,767 181,767 175,991
Pseudo R2 0.164 0.167 0.157
61
Panel B: Return consequence test
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation 0.0276 -0.0050 -0.0325 -0.0525 -0.1314 -1.7977**
× Uncertainty (0.97) (-0.14) (-0.86) (-1.10) (-1.28) (-2.42)
Overvaluation -0.0102** -0.0105* -0.0185*** -0.0271*** -0.0792*** -0.5879***
(-2.05) (-1.68) (-2.59) (-2.91) (-3.49) (-3.35)
Uncertainty -0.0508*** -0.0465*** -0.0362** -0.0383* -0.0340 0.5047
(-4.13) (-3.24) (-2.07) (-1.68) (-0.80) (1.56)
Size 0.0016** 0.0027*** 0.0026*** 0.0028** 0.0064** 0.0728***
(2.47) (3.30) (2.78) (2.25) (2.52) (3.10)
Leverage 0.0047 0.0013 -0.0012 -0.0004 0.0103 -0.0585
(1.16) (0.26) (-0.19) (-0.05) (0.61) (-0.34)
Illiquidity -0.0080 -0.0186 -0.0186 -0.0157 -0.0162 -0.0882
(-0.59) (-1.54) (-1.17) (-0.82) (-0.42) (-0.31)
Volatility 0.0060 -0.1564* -0.1535 -0.2796* -0.6569* -10.306***
(0.09) (-1.65) (-1.37) (-1.88) (-1.68) (-3.30)
LnAnalyst 0.0061*** 0.0067*** 0.0085*** 0.0094*** 0.0195*** 0.0191
(4.85) (3.99) (4.40) (3.89) (3.75) (0.49)
ShortInterest 0.0055 0.0083 0.0115 0.0198** 0.0390 0.2889
(1.13) (1.23) (1.63) (2.02) (1.64) (1.33)
EarnAnnounce[-5,0] 0.0047 0.0000 0.0008 0.0018 -0.0054 0.0392
(1.30) (0.00) (0.14) (0.27) (-0.61) (0.64)
AnForecast[-5,0] -0.0026 -0.0027 -0.0020 -0.0053 -0.0047 -0.0645
(-1.36) (-1.13) (-0.68) (-1.36) (-0.71) (-1.52)
ConfCall[-5,0] 0.0041 0.0138 0.0148 0.0122 0.0104 0.0339
(0.55) (1.27) (1.30) (0.98) (0.44) (0.31)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0239* -0.0302*** -0.0304*** -0.0517** -0.1061 -0.1112
(-1.85) (-3.75) (-3.04) (-2.30) (-1.62) (-0.47)
Observations 5,702 5,701 5,699 5,694 5,629 4,474
Adjusted R2 0.075 0.090 0.083 0.081 0.089 0.185
62
Table A2: Excluding Activist Short-Selling Cases Covered by Mainstream Media
This table reports return consequence results using activist short-selling cases that are not covered by Top U.S.
Mainstream media. I manually check Factiva Top US Newspapers articles with the names of both the short-seller and
the target company mentioned in the Headline and Lead Paragraph.
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation -0.0033 -0.0095* -0.0206*** -0.0306*** -0.0950*** -0.8822***
(-0.77) (-1.69) (-3.33) (-3.85) (-5.18) (-5.92)
Uncertainty -0.0440*** -0.0531*** -0.0512*** -0.0581*** -0.0695*** -0.0361
(-5.86) (-5.81) (-4.60) (-4.09) (-2.74) (-0.16)
Size 0.0017*** 0.0024*** 0.0023** 0.0028** 0.0065** 0.0639***
(2.61) (3.04) (2.57) (2.26) (2.49) (2.74)
Leverage 0.0037 0.0004 -0.0019 -0.0018 0.0092 -0.0197
(0.95) (0.09) (-0.32) (-0.22) (0.59) (-0.12)
Illiquidity -0.0148 -0.0211* -0.0194 -0.0180 -0.0151 -0.0005
(-1.12) (-1.69) (-1.19) (-0.92) (-0.39) (-0.00)
Volatility -0.0067 -0.1586* -0.1334 -0.2754* -0.5013 -10.0363***
(-0.10) (-1.69) (-1.19) (-1.86) (-1.29) (-3.34)
LnAnalyst 0.0046*** 0.0057*** 0.0079*** 0.0088*** 0.0200*** 0.0282
(3.76) (3.39) (3.93) (3.52) (3.80) (0.70)
ShortInterest 0.0020 0.0066 0.0077 0.0134 0.0284 0.2455
(0.32) (0.98) (1.05) (1.49) (1.14) (1.17)
EarnAnnounce[-5,0] 0.0037 0.0000 0.0013 0.0015 -0.0046 0.0538
(1.10) (0.01) (0.22) (0.22) (-0.53) (0.91)
AnForecast[-5,0] -0.0028 -0.0033 -0.0031 -0.0060 -0.0031 -0.0583
(-1.53) (-1.41) (-1.06) (-1.56) (-0.46) (-1.40)
ConfCall[-5,0] 0.0069 0.0115 0.0084 0.0111 0.0080 0.0454
(1.12) (1.04) (0.72) (0.87) (0.32) (0.40)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0113 -0.0242** -0.0278* -0.0437 -0.0583* 0.0364
(-0.95) (-2.41) (-1.73) (-1.15) (-1.73) (0.13)
Observations 5,226 5,225 5,223 5,218 5,156 4,106
Adjusted R2 0.073 0.088 0.081 0.080 0.086 0.173
63
Table A3: The Role of Short-Seller’s Reputation
This table highlights the role of the short-seller’s reputation in affecting returns after firms are targeted. CumReturn is
the accumulated CAR(0, 1) for the same short-seller’s all previous activist short-selling cases.
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
CumReturn 0.0046*** 0.0051*** 0.0040*** 0.0040** 0.0057 0.0346
(4.41) (3.92) (2.92) (2.30) (1.22) (1.31)
Overvaluation -0.0060 -0.0119** -0.0243*** -0.0354*** -0.0968*** -0.8860***
(-1.31) (-2.10) (-3.96) (-4.46) (-5.41) (-5.87)
Uncertainty -0.0396*** -0.0468*** -0.0469*** -0.0548*** -0.0792*** -0.1017
(-5.51) (-5.52) (-3.82) (-3.49) (-3.29) (-0.47)
Size 0.0013** 0.0024*** 0.0024*** 0.0027** 0.0066*** 0.0723***
(2.05) (2.94) (2.59) (2.16) (2.62) (3.13)
Leverage 0.0032 -0.0002 -0.0017 -0.0006 0.0098 -0.0483
(0.80) (-0.04) (-0.29) (-0.07) (0.60) (-0.29)
Illiquidity -0.0093 -0.0192 -0.0181 -0.0139 -0.0086 0.0056
(-0.69) (-1.59) (-1.14) (-0.74) (-0.22) (0.02)
Volatility 0.0056 -0.1374 -0.1358 -0.2627* -0.5548 -9.8060***
(0.09) (-1.49) (-1.23) (-1.79) (-1.43) (-3.11)
LnAnalyst 0.0055*** 0.0064*** 0.0080*** 0.0090*** 0.0191*** 0.0265
(4.59) (3.99) (4.28) (3.86) (3.75) (0.68)
ShortInterest 0.0053 0.0075 0.0112 0.0204** 0.0414* 0.2965
(1.07) (1.13) (1.61) (2.15) (1.73) (1.35)
EarnAnnounce[-5,0] 0.0047 0.0004 0.0014 0.0022 -0.0048 0.0460
(1.37) (0.08) (0.25) (0.34) (-0.55) (0.78)
AnForecast[-5,0] -0.0027 -0.0029 -0.0024 -0.0057 -0.0064 -0.0692
(-1.48) (-1.21) (-0.83) (-1.48) (-0.96) (-1.63)
ConfCall[-5,0] 0.0043 0.0135 0.0150 0.0133 0.0122 0.0342
(0.58) (1.24) (1.33) (1.07) (0.51) (0.31)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0209 -0.0266*** -0.0264** -0.0484** -0.1053 -0.0777
(-1.60) (-3.27) (-2.51) (-2.06) (-1.63) (-0.35)
Observations 5,606 5,605 5,603 5,598 5,533 4,393
Adjusted R2 0.075 0.092 0.081 0.080 0.089 0.179
64
Table A4: Results are Not Driven by Firm-Quarters That Are Targeted Multiple Times
This table addresses the possibility that the results are driven by some firm-quarters targeted by multiple times. Panel
A explicitly models the number of times each firm-quarter is targeted in the determinant test. The same overvaluation
and uncertainty features survive the step-wise regressions. Panel B focuses on the first activist short-selling case for
each firm-quarter.
Panel A: Modeling the number of times a firm-quarter is targeted
(1) (2) (3) (4) (5) (6)
N_Target
(Negative
Binomial)
N_Target
(Negative
Binomial)
lnN_Target
(OLS)
lnN_Target
(OLS)
N_Target
(Negative
Binomial)
lnN_Target
(OLS)
Size 0.1908*** 0.1187*** 0.0025*** 0.0012** 0.2003*** 0.0027***
(6.76) (3.84) (4.53) (2.27) (6.87) (4.76)
Leverage -0.1172 0.0045 -0.0052** -0.0024 0.0290 -0.0019
(-0.74) (0.03) (-1.98) (-0.94) (0.19) (-0.74)
Illiquidity -1.6610*** -2.2906*** -0.0025 -0.0081*** -1.9173*** -0.0043***
(-7.17) (-8.11) (-1.61) (-6.55) (-7.39) (-2.90)
Volatility 29.5617*** 28.6861*** 0.3616*** 0.3468*** 22.9882*** 0.2793***
(16.91) (15.11) (12.11) (11.43) (13.00) (9.81)
LnAnalyst 0.2027*** 0.5023*** 0.0027*** 0.0063*** 0.3570*** 0.0047***
(4.63) (7.61) (3.77) (6.50) (6.80) (5.64)
ShortInterest 4.5082*** 5.2621*** 0.1432*** 0.1579*** 4.8448*** 0.1492***
(9.65) (9.87) (5.19) (5.55) (9.54) (5.33)
Run-up 0.1018* 0.0038***
(1.76) (3.82)
HighP/B 0.8691*** 0.0159***
(10.69) (6.75)
HighP/V 0.4794*** 0.0158***
(5.45) (5.17)
HighAssetGrowth 0.3953*** 0.0092***
(4.83) (4.02)
LowQTREarning 0.3146*** 0.0039***
(3.83) (2.67)
HighNOA 0.2001** 0.0023**
(2.44) (2.00)
HighManipulation 0.3568*** 0.0042***
(4.56) (2.91)
HighB/SSpread 0.2340** 0.0032***
(2.55) (2.64)
NonBlockholder 0.4242*** 0.0062***
(5.10) (5.45)
LowDedicated 0.2832*** 0.0035***
(3.16) (3.01)
LowEarnQuality 0.2653*** 0.0028*
(2.76) (1.72)
HighSTDFE -0.1966**
(-2.40)
NonBIG4 0.4248*** 0.0055***
(4.30) (3.97)
AuditSwitch 0.3234** 0.0038*
65
(2.51) (1.83)
IntCntrlWeak 0.2855**
(2.31)
Overvaluation 2.8292*** 0.0549***
(12.68) (7.89)
Uncertainty 1.7256*** 0.0206***
(8.16) (6.41)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -9.7738*** -9.4165*** -0.0450*** -0.0345*** -10.1642*** -0.0513***
(-15.07) (-14.42) (-5.89) (-4.84) (-15.49) (-6.40)
Observations 186,772 186,772 186,772 186,772 186,772 186,772
Adjusted/Pseudo R2 0.155 0.153 0.035 0.030 0.154 0.034
Panel B: Focusing on the first activist short-selling case of each firm-quarter
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation -0.0087 -0.0119 -0.0222*** -0.0329*** -0.0781*** -0.7040***
(-1.51) (-1.61) (-2.80) (-3.43) (-4.33) (-6.03)
Uncertainty -0.0474*** -0.0507*** -0.0537*** -0.0609*** -0.0764*** -0.1403
(-5.95) (-5.28) (-4.80) (-4.39) (-3.20) (-0.94)
Size 0.0015** 0.0021** 0.0019* 0.0015 0.0032 0.0299*
(2.06) (2.24) (1.85) (1.16) (1.36) (1.87)
Leverage 0.0035 0.0002 -0.0031 -0.0010 0.0178 -0.0564
(0.74) (0.04) (-0.48) (-0.13) (1.28) (-0.60)
Illiquidity -0.0037 -0.0121 -0.0105 -0.0147 -0.0201 0.0356
(-0.30) (-1.01) (-0.62) (-0.74) (-0.51) (0.13)
Volatility 0.0374 -0.1755 -0.1570 -0.2674 -0.7195** -10.225***
(0.41) (-1.49) (-1.07) (-1.55) (-2.11) (-4.35)
LnAnalyst 0.0067*** 0.0086*** 0.0101*** 0.0112*** 0.0245*** 0.0597*
(4.54) (4.31) (4.40) (4.06) (4.68) (1.78)
ShortInterest 0.0007 0.0004 0.0016 0.0094 0.0444 0.3478**
(0.10) (0.04) (0.14) (0.52) (1.59) (2.39)
EarnAnnounce[-5,0] 0.0066 0.0051 0.0071 0.0037 0.0079 0.0078
(1.48) (1.02) (1.20) (0.54) (0.66) (0.10)
AnForecast[-5,0] -0.0027 -0.0017 0.0004 -0.0012 -0.0034 -0.0567
(-1.14) (-0.58) (0.12) (-0.31) (-0.46) (-1.19)
ConfCall[-5,0] -0.0059 -0.0022 -0.0024 0.0067 -0.0168 -0.0837
(-0.66) (-0.24) (-0.24) (0.56) (-0.75) (-0.65)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0219 -0.0242*** -0.0234** -0.0419* -0.0962 0.1325
(-1.59) (-2.77) (-2.00) (-1.75) (-1.45) (0.77)
Observations 3,373 3,373 3,373 3,368 3,331 2,635
Adjusted R2 0.077 0.097 0.091 0.087 0.075 0.111
66
Table A5: Results are Consistent in Different Time Periods
This table shows that the key inferences hold in different sample periods. Panel A tabulates determinant results each year from 2006 to 2015; Panel B reports return
consequence results using years prior to 2012; Panel C reports return consequence results using years after 2013 (note earlier years have fewer firms targeted by
activist short-sellers.)
Panel A: Determinant test
DV = Target (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Overvaluation 2.7037*** 3.1853*** 2.9567*** 2.8194*** 2.8474*** 2.9033*** 3.0972*** 1.8111*** 2.5188*** 1.7388***
(5.23) (6.37) (5.77) (5.51) (5.66) (6.34) (8.48) (7.26) (9.35) (4.33)
Uncertainty 1.3294* 1.2298* 0.5503 0.2021 1.9401*** 2.2108*** 1.1808*** 0.8107*** 0.9925*** 0.6124
(1.70) (1.65) (0.71) (0.19) (3.33) (4.62) (2.61) (2.82) (3.64) (1.33)
Size 0.4046*** 0.3139*** 0.4018*** 0.3454*** 0.2223*** 0.2504*** 0.2248*** -0.0892** 0.0237 0.0733
(5.46) (3.96) (6.42) (4.59) (2.88) (3.69) (3.64) (-2.08) (0.50) (1.54)
Leverage -0.9332 -0.5517 0.4559 0.1955 -1.4401*** -1.3177*** -0.6708** 0.1673 0.1570 0.1773
(-1.47) (-1.15) (1.08) (0.39) (-3.00) (-3.28) (-1.96) (0.92) (0.77) (0.75)
Illiquidity -1.6046 -14.3867 -2.2347 -2.6260 -3.8858 -1.9363** -1.7580*** -1.6579*** -1.4186*** -1.3925***
(-1.47) (-1.58) (-0.97) (-0.88) (-1.26) (-2.53) (-2.95) (-4.46) (-2.87) (-3.81)
Volatility 43.7564*** 29.2849*** 4.0020 10.1460*** 20.2885*** 16.3969*** 25.8508*** 19.3989*** 18.6050*** 18.9834***
(5.13) (4.66) (1.15) (2.62) (2.82) (2.79) (6.06) (7.01) (6.38) (5.89)
LnAnalyst 0.4846*** 0.6941*** 0.5623*** 0.6209*** 0.3717*** 0.3775*** 0.2273** 0.3626*** 0.2716*** 0.4001***
(2.71) (4.99) (4.18) (3.82) (3.37) (3.13) (2.23) (5.03) (3.81) (5.15)
ShortInterest 2.8802*** 2.4476*** 2.2287*** 3.3029*** 3.5579*** 3.8716*** 3.4514*** 3.7828*** 3.6372*** 3.1547***
(4.74) (5.29) (4.75) (4.31) (6.41) (6.36) (6.68) (7.01) (8.25) (6.13)
FF 48 FE YES YES YES YES YES YES YES YES YES YES
QTR FE YES YES YES YES YES YES YES YES YES YES
Constant -21.4790*** -23.9080 -22.3262*** -8.3393*** -19.8827*** -6.0738*** -19.4499*** -4.5231*** -4.8707*** -18.5515***
(-27.63) (-27.36) (-27.46) (-6.47) (-25.53) (-5.64) (-36.81) (-5.32) (-4.70) (-35.01)
Observations 16,039 16,301 17,024 16,594 16,493 16,441 16,408 16,706 17,300 13,376
Pseudo R2 0.146 0.205 0.208 0.197 0.149 0.158 0.158 0.114 0.123 0.107
67
Panel B: Return consequences prior to year 2012
Pre 2012: (1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation 0.0065 -0.0083 -0.0258** -0.0470*** -0.1137*** -0.8269***
(0.79) (-0.84) (-2.33) (-3.37) (-3.46) (-3.91)
Uncertainty -0.0321*** -0.0434*** -0.0320* -0.0538** -0.0883* -0.2893
(-2.69) (-2.84) (-1.84) (-2.30) (-1.84) (-1.19)
Size 0.0015 0.0033** 0.0032** 0.0033* 0.0063 0.0501**
(1.51) (2.50) (2.14) (1.76) (1.47) (2.14)
Leverage 0.0120* 0.0036 0.0038 -0.0017 0.0096 -0.1067
(1.73) (0.40) (0.38) (-0.13) (0.36) (-0.70)
Illiquidity -0.0195 -0.0174 0.0030 0.0297 -0.0197 0.1909
(-0.64) (-0.76) (0.10) (1.13) (-0.29) (0.58)
Volatility -0.0330 -0.2193 -0.2465 -0.3496 0.1337 -2.3137
(-0.33) (-1.52) (-1.38) (-1.61) (0.19) (-0.65)
LnAnalyst 0.0063*** 0.0081*** 0.0109*** 0.0123*** 0.0298*** 0.1094**
(2.93) (2.90) (3.40) (3.26) (3.32) (2.27)
ShortInterest 0.0036 0.0078 0.0121 0.0020 0.0282 0.4056**
(0.46) (0.66) (0.96) (0.12) (0.85) (2.17)
EarnAnnounce[-5,0] 0.0079 0.0104 0.0113 0.0136 -0.0042 -0.0669
(1.30) (1.44) (1.43) (1.47) (-0.24) (-0.72)
AnForecast[-5,0] -0.0004 -0.0029 -0.0021 -0.0031 -0.0095 -0.0462
(-0.13) (-0.73) (-0.47) (-0.55) (-0.86) (-0.84)
ConfCall[-5,0] 0.0013 0.0039 0.0025 -0.0010 0.0088 0.0867
(0.15) (0.32) (0.21) (-0.08) (0.33) (0.72)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0317** -0.0364*** -0.0376*** -0.0470* -0.1256* -0.2892
(-2.11) (-3.08) (-2.68) (-1.93) (-1.67) (-1.07)
Observations 2,009 2,009 2,009 2,009 2,009 2,009
Adjusted R2 0.094 0.105 0.094 0.089 0.100 0.191
68
Panel C: Return consequences after year 2013
Post 2013 (1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation -0.0123* -0.0130* -0.0232*** -0.0312*** -0.0855*** -0.7042***
(-1.87) (-1.70) (-2.85) (-3.02) (-4.07) (-3.54)
Uncertainty -0.0445*** -0.0507*** -0.0559*** -0.0605*** -0.0773** 0.0446
(-4.68) (-4.44) (-3.21) (-2.79) (-2.51) (0.14)
Size 0.0014* 0.0024** 0.0017 0.0019 0.0050 0.1002***
(1.73) (2.30) (1.42) (1.18) (1.58) (3.11)
Leverage 0.0025 0.0001 -0.0025 0.0004 0.0159 0.0611
(0.49) (0.02) (-0.32) (0.04) (0.84) (0.32)
Illiquidity -0.0063 -0.0211 -0.0279 -0.0328 -0.0010 -0.2353
(-0.55) (-1.46) (-1.61) (-1.54) (-0.02) (-0.65)
Volatility 0.0395 -0.1471 -0.1699 -0.3611* -1.3410*** -15.7544***
(0.42) (-1.10) (-1.10) (-1.66) (-2.85) (-3.41)
LnAnalyst 0.0060*** 0.0056*** 0.0074*** 0.0082*** 0.0132** -0.0759
(3.86) (2.63) (3.07) (2.67) (2.09) (-1.48)
ShortInterest 0.0086 0.0101 0.0160 0.0330** 0.0518 0.0278
(1.13) (1.07) (1.53) (2.40) (1.42) (0.09)
EarnAnnounce[-5,0] 0.0041 -0.0042 -0.0036 -0.0034 -0.0053 0.1347*
(0.95) (-0.72) (-0.50) (-0.44) (-0.54) (1.67)
AnForecast[-5,0] -0.0045* -0.0029 -0.0018 -0.0056 -0.0023 -0.0807
(-1.94) (-0.97) (-0.47) (-1.12) (-0.27) (-1.31)
ConfCall[-5,0] -0.0079 0.0215 0.0665 0.0703 0.1076 0.3517
(-0.67) (0.77) (0.98) (0.95) (0.61) (0.32)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0902*** -0.0763*** -0.0682*** -0.0641*** -0.0316 0.4885
(-8.19) (-5.16) (-3.88) (-2.64) (-0.63) (1.19)
Observations 3,693 3,692 3,690 3,685 3,620 2,465
Adjusted R2 0.062 0.079 0.078 0.081 0.094 0.246
69
Table A6: Using decile ranks rather than top-quintile
This table uses overvaluation and uncertainty features that are constructed as decile ranks (rather than indicators based on top quintiles). Panel A reports results on
individual overvaluation features. Panel B reports results on individual uncertainty features.
Panel A: Individual overvaluation features
DV=Target (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Overvaluation
Features =
Price
Runup
P/B P/V Low
Gross Profit
Asset
Growth
Investment NOA Accruals Low
Payout%
OScore MScore Low
Earnings
Overvaluation 0.2898*** 1.1159*** 1.0154*** 0.0147 0.5868*** 0.2299* -0.0750 -0.2964*** 0.5622*** 0.0691 0.2661*** 0.1267
Features (3.95) (9.93) (8.04) (0.10) (5.74) (1.88) (-0.66) (-2.69) (5.06) (0.47) (2.97) (1.18)
Size 0.0202 0.0958*** 0.0739** 0.0355 0.0406 0.0601* 0.0199 0.0581* 0.0679** 0.0234 0.0125 0.0226
(0.71) (3.21) (1.99) (1.14) (1.34) (1.67) (0.70) (1.88) (2.11) (0.76) (0.41) (0.79) Leverage -0.1627 -0.2327 -0.2982* -0.0731 0.0486 -0.1430 -0.2139 -0.1234 -0.0759 -0.2341 -0.1923 -0.1997
(-1.22) (-1.54) (-1.84) (-0.52) (0.34) (-0.89) (-1.53) (-0.86) (-0.53) (-1.30) (-1.35) (-1.53)
Illiquidity -1.8373*** -1.6422*** -1.7190*** -1.7298*** -1.6429*** -1.5843*** -1.9141*** -1.7488*** -1.6371*** -1.9139*** -2.0326*** -1.9168*** (-7.55) (-7.43) (-5.88) (-6.80) (-6.59) (-5.92) (-7.66) (-6.47) (-6.39) (-7.65) (-6.70) (-7.62)
Volatility 25.9950*** 28.0915*** 24.9219*** 23.8982*** 24.5813*** 23.9133*** 25.1419*** 23.6466*** 21.8840*** 24.9587*** 26.4485*** 24.5520***
(19.96) (21.42) (15.63) (17.13) (17.74) (15.12) (20.02) (16.80) (15.83) (19.72) (19.43) (18.79) LnAnalyst 0.3921*** 0.2379*** 0.2577*** 0.4011*** 0.3901*** 0.4056*** 0.3842*** 0.3959*** 0.4009*** 0.3864*** 0.3811*** 0.3873***
(8.45) (5.49) (4.36) (7.92) (8.01) (6.78) (8.45) (7.74) (7.88) (8.41) (7.53) (8.43)
ShortInterest 3.3986*** 3.5437*** 3.7296*** 3.5068*** 3.4294*** 3.8147*** 3.3734*** 3.6329*** 3.5393*** 3.3754*** 3.3819*** 3.3849*** (15.52) (16.89) (13.40) (15.43) (14.88) (12.53) (15.26) (13.86) (15.65) (15.27) (14.62) (15.37)
FF 48 FE YES YES YES YES YES YES YES YES YES YES YES YES
QTR FE YES YES YES YES YES YES YES YES YES YES YES YES Constant -8.1955*** -9.1609*** -10.841*** -8.0927*** -8.5133*** -7.6678*** -7.9444*** -8.2962*** -8.5070*** -8.0270*** -8.2905*** -8.0480***
(-13.56) (-14.85) (-9.97) (-13.69) (-14.12) (-12.90) (-13.32) (-12.54) (-14.06) (-13.28) (-12.56) (-13.36)
Observations 181,055 181,767 124,131 168,916 169,766 134,498 181,426 165,225 167,160 181,426 164,618 181,399 Pseudo R2 0.144 0.152 0.162 0.142 0.146 0.151 0.143 0.144 0.144 0.143 0.147 0.143
70
Panel B: Individual uncertainty features
DV=Target (1) (2) (3) (4) (5) (6) (7) (8) (9)
Uncertainty
Features =
Low
InstOwn
Non
Block
Low
Dedicated
Low
AccQuality
Analyst
Disagree
NonBig4 Audit
Switch
ICW BidAskSpread
Uncertainty 0.3502*** 0.5426*** 0.1465 0.7145*** -0.1247 0.3952*** 0.3656*** 0.3647*** 1.1915***
Features (5.03) (5.63) (1.27) (5.00) (-1.14) (4.82) (3.48) (3.57) (9.04)
Size 0.0062 0.0124 0.0250 0.1262*** 0.0576* 0.0329 0.0211 0.0224 0.0982***
(0.23) (0.45) (0.87) (3.00) (1.72) (1.09) (0.69) (0.74) (3.06)
Leverage -0.1450 -0.1724 -0.1838 -0.4925** -0.1090 -0.1063 -0.0432 -0.0515 -0.2668**
(-1.12) (-1.32) (-1.39) (-2.17) (-0.72) (-0.80) (-0.31) (-0.36) (-2.00)
Illiquidity -1.9655*** -2.0045*** -1.9198*** -1.7181*** -3.7774*** -2.1292*** -1.7832*** -1.7875*** -1.8166***
(-7.68) (-7.59) (-7.63) (-5.20) (-3.63) (-7.27) (-6.55) (-6.56) (-7.11)
Volatility 24.7225*** 23.5767*** 25.0088*** 26.4366*** 28.2667*** 24.7304*** 23.3148*** 23.0639*** 16.2624***
(19.62) (18.74) (19.99) (13.27) (17.82) (18.92) (16.53) (16.21) (11.80)
LnAnalyst 0.4263*** 0.4578*** 0.4004*** 0.3924*** 0.5170*** 0.4141*** 0.4176*** 0.4189*** 0.4013***
(9.22) (9.52) (8.56) (6.36) (6.68) (8.46) (8.07) (8.11) (8.67)
ShortInterest 3.3607*** 3.4610*** 3.3671*** 3.5628*** 3.3092*** 3.4142*** 3.5149*** 3.5225*** 3.2714***
(14.75) (14.88) (15.32) (11.00) (13.63) (15.21) (14.98) (15.18) (14.71)
FF 48 FE YES YES YES YES YES YES YES YES YES
QTR FE YES YES YES YES YES YES YES YES YES
Constant -8.0200*** -8.2836*** -8.1153*** -9.0499*** -8.6273*** -8.1464*** -8.1348*** -8.1396*** -8.8326***
(-13.58) (-13.87) (-13.57) (-12.28) (-12.70) (-13.38) (-13.42) (-13.51) (-14.51)
Observations 181,767 181,767 181,767 108,724 119,398 171,416 161,381 161,381 181,748
Pseudo R2 0.144 0.145 0.143 0.155 0.132 0.144 0.142 0.142 0.147
71
Table A7: An Alternative Way of Aggregating Individual Features: Using All Individual
Features
This table reports results based on aggregate overvaluation and uncertainty measures (i.e., Overvaluation_All and
Uncertainty_All) that average all individual features (rather than only those surviving the step-wise regressions in the
main analyses). Specifically, Panel A reports the results of determinant test and Panel B reports the results of the return
consequence test.
Panel A: Determinant test
(1) (2) (3)
Target Target by SA Target by ASR
Overvaluation_All 2.5340*** 2.5121*** 2.5707***
(12.25) (11.59) (8.58)
Uncertainty_All 1.1499*** 1.1481*** 1.1780***
(5.94) (5.73) (3.99)
Size 0.1074*** 0.1139*** -0.1140**
(3.69) (3.84) (-2.48)
Leverage -0.3394*** -0.2579** -0.6932***
(-2.64) (-1.98) (-3.21)
Illiquidity -1.8439*** -1.7212*** -3.5408***
(-7.20) (-6.89) (-4.60)
Volatility 18.3106*** 19.0635*** 9.9757***
(13.80) (14.14) (3.94)
LnAnalyst 0.3887*** 0.4080*** 0.0246
(8.93) (8.87) (0.42)
ShortInterest 3.1884*** 3.0864*** 3.7486***
(14.44) (13.86) (13.54)
FF 48 FE Yes Yes Yes
QTR FE Yes Yes Yes
Constant -9.0083*** -10.0140*** -7.2495***
(-14.84) (-12.94) (-7.82)
Observations 181,767 181,767 175,991
Pseudo R2 0.158 0.161 0.148
72
Panel B: Return consequence test
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation_All -0.0074 -0.0136* -0.0280*** -0.0430*** -0.0880*** -0.8673***
(-1.25) (-1.74) (-3.27) (-3.75) (-3.72) (-4.33)
Uncertainty_All -0.0423*** -0.0471*** -0.0471*** -0.0549*** -0.0554** 0.0014
(-5.02) (-4.75) (-3.60) (-3.33) (-2.01) (0.01)
Size 0.0017*** 0.0029*** 0.0027*** 0.0029** 0.0076*** 0.0780***
(2.63) (3.40) (2.83) (2.28) (2.92) (3.27)
Leverage 0.0062 0.0038 0.0029 0.0053 0.0226 0.0464
(1.48) (0.74) (0.47) (0.62) (1.33) (0.27)
Illiquidity -0.0103 -0.0194 -0.0175 -0.0136 -0.0087 0.0342
(-0.77) (-1.61) (-1.11) (-0.73) (-0.23) (0.13)
Volatility 0.0106 -0.1567 -0.1470 -0.2706* -0.7249* -10.4125***
(0.16) (-1.63) (-1.29) (-1.80) (-1.79) (-3.21)
LnAnalyst 0.0065*** 0.0073*** 0.0090*** 0.0099*** 0.0191*** 0.0094
(5.20) (4.34) (4.63) (4.09) (3.64) (0.23)
ShortInterest 0.0053 0.0085 0.0120* 0.0206** 0.0417* 0.3108
(1.03) (1.21) (1.67) (2.13) (1.71) (1.43)
EarnAnnounce[-5,0] 0.0047 0.0000 0.0009 0.0019 -0.0053 0.0411
(1.31) (0.01) (0.15) (0.28) (-0.60) (0.66)
AnForecast[-5,0] -0.0026 -0.0028 -0.0021 -0.0054 -0.0050 -0.0708*
(-1.36) (-1.14) (-0.71) (-1.39) (-0.76) (-1.66)
ConfCall[-5,0] 0.0036 0.0133 0.0142 0.0116 0.0086 0.0285
(0.48) (1.22) (1.25) (0.93) (0.36) (0.26)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0255* -0.0316*** -0.0304*** -0.0510** -0.1186* -0.1425
(-1.84) (-3.75) (-2.99) (-2.34) (-1.85) (-0.61)
Observations 5,702 5,701 5,699 5,694 5,629 4,474
Adjusted R2 0.072 0.088 0.081 0.079 0.082 0.174
73
Table A8: Alternative Risk Adjustment
This table presents results using abnormal returns based on the Fama–French three-factor plus momentum model
rather than Fama-French three-factor model used in the main analyses.
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation -0.0056 -0.0103* -0.0233*** -0.0359*** -0.1032*** -0.9198***
(-1.11) (-1.71) (-3.64) (-4.45) (-5.61) (-6.06)
Uncertainty -0.0414*** -0.0471*** -0.0461*** -0.0544*** -0.0746*** -0.1054
(-5.64) (-5.43) (-3.71) (-3.46) (-3.04) (-0.48)
Size 0.0017*** 0.0030*** 0.0028*** 0.0029** 0.0063** 0.0690***
(2.63) (3.67) (3.04) (2.41) (2.57) (2.93)
Leverage 0.0046 0.0015 -0.0008 0.0016 0.0169 -0.0376
(1.12) (0.30) (-0.13) (0.20) (1.09) (-0.21)
Illiquidity -0.0090 -0.0197 -0.0189 -0.0138 -0.0120 -0.0618
(-0.66) (-1.61) (-1.18) (-0.73) (-0.30) (-0.23)
Volatility 0.0097 -0.1430 -0.1451 -0.2870** -0.7290* -10.8153***
(0.15) (-1.50) (-1.34) (-1.96) (-1.94) (-3.39)
LnAnalyst 0.0059*** 0.0064*** 0.0085*** 0.0095*** 0.0200*** 0.0278
(4.61) (3.82) (4.46) (4.05) (3.91) (0.70)
ShortInterest 0.0054 0.0084 0.0110 0.0203** 0.0410* 0.3525
(1.07) (1.25) (1.61) (2.11) (1.70) (1.55)
EarnAnnounce[-5,0] 0.0051 0.0004 0.0012 0.0016 -0.0076 0.0411
(1.43) (0.09) (0.20) (0.23) (-0.87) (0.68)
AnForecast[-5,0] -0.0029 -0.0030 -0.0021 -0.0050 -0.0046 -0.0596
(-1.50) (-1.22) (-0.71) (-1.31) (-0.69) (-1.42)
ConfCall[-5,0] 0.0033 0.0146 0.0170 0.0159 0.0213 0.0092
(0.44) (1.34) (1.52) (1.29) (0.87) (0.07)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0239* -0.0282*** -0.0271*** -0.0457** -0.0850 0.0622
(-1.71) (-3.48) (-2.99) (-2.24) (-1.20) (0.23)
Observations 5,702 5,701 5,699 5,694 5,629 4,474
Adjusted R2 0.075 0.090 0.085 0.083 0.089 0.180
74
Table A9: Two-way Clustering
This table reports t-stats based on standard errors that are clustered by both firm and quarter (rather than by only firm - it is technically challenging to conduct two-
way clustering in step-wise regressions). Panel A reports determinant-test results on individual overvaluation features. Panel B reports determinant-test results on
individual uncertainty features. Panel C reports results of the return consequence test.
Panel A: Determinants test based on individual overvaluation features
DV=Target (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Overvaluation
Features =
Price
Runup
P/B P/V Low
Gross
Profit
Asset
Growth
Investment NOA Accruals Low
Payout%
OScore MScore Low
Earnings
Overvaluation 0.354*** 0.781*** 0.666*** 0.147 0.452*** 0.166** 0.125* -0.026 0.203*** 0.258*** 0.407*** 0.389*** Features (7.46) (9.67) (9.35) (1.64) (6.31) (2.34) (1.70) (-0.35) (3.43) (3.71) (6.88) (5.23) Size 0.023 0.087** 0.037 0.018 0.029 0.023 0.019 0.019 0.023 0.037 0.029 0.043 (0.60) (2.18) (0.99) (0.47) (0.75) (0.60) (0.50) (0.48) (0.59) (0.94) (0.75) (1.10) Leverage -0.184 -0.396** -0.202 -0.183 -0.138 -0.169 -0.127 -0.178 -0.184 -0.372** -0.148 -0.242* (-1.28) (-2.46) (-1.39) (-1.27) (-0.97) (-1.18) (-0.90) (-1.25) (-1.28) (-2.27) (-1.05) (-1.68) Illiquidity -1.822*** -1.738*** -1.819*** -1.907*** -1.815*** -1.905*** -1.903*** -1.909*** -1.893*** -1.935*** -1.861*** -1.919*** (-7.77) (-7.74) (-7.51) (-7.68) (-7.72) (-7.68) (-7.71) (-7.68) (-7.68) (-7.66) (-7.68) (-7.55) Volatility 25.125*** 26.822*** 23.887*** 24.694*** 25.088*** 25.163*** 25.025*** 25.149*** 24.413*** 24.268*** 24.738*** 23.030*** (14.01) (14.75) (12.70) (13.03) (13.40) (13.15) (13.22) (13.16) (13.30) (12.82) (13.00) (12.30) LnAnalyst 0.393*** 0.277*** 0.344*** 0.393*** 0.368*** 0.380*** 0.392*** 0.388*** 0.391*** 0.389*** 0.391*** 0.382*** (8.03) (5.85) (7.23) (8.06) (7.75) (7.82) (8.01) (7.89) (7.97) (7.97) (7.96) (7.84) ShortInterest 3.336*** 3.308*** 3.298*** 3.360*** 3.220*** 3.345*** 3.363*** 3.361*** 3.344*** 3.346*** 3.300*** 3.364***
(14.31) (14.28) (14.48) (14.44) (13.76) (14.36) (14.45) (14.37) (14.38) (14.32) (14.03) (14.46) FF 48 FE YES YES YES YES YES YES YES YES YES YES YES YES
QTR FE YES YES YES YES YES YES YES YES YES YES YES YES
Constant -8.131*** -8.566*** -8.111*** -8.010*** -8.163*** -8.040*** -8.058*** -7.996*** -8.037*** -8.057*** -8.160*** -8.123***
(-17.27) (-19.15) (-17.50) (-17.55) (-17.37) (-17.64) (-17.76) (-17.48) (-17.61) (-17.73) (-17.80) (-17.80)
Observations 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767
Pseudo R2 0.145 0.154 0.149 0.143 0.146 0.143 0.143 0.143 0.144 0.144 0.145 0.145
75
Panel B: Determinants test based on individual uncertainty features
DV=Target (1) (2) (3) (4) (5) (6) (7) (8) (9)
Uncertainty
Features =
Low
InstOwn
Non
Block
Low
Dedicated
Low
AccQuality
Analyst
Disagree
NonBig4 Audit
Switch
ICW BidAskSpread
Uncertainty 0.1683** 0.3502*** 0.2395*** 0.2567*** -0.0270 0.3913*** 0.2879** 0.3790*** 0.2554***
Features (2.26) (4.53) (3.29) (3.52) (-0.39) (4.87) (2.32) (4.18) (3.99)
Size 0.0155 0.0062 0.0215 0.0265 0.0182 0.0437 0.0198 0.0209 0.0267
(0.41) (0.17) (0.56) (0.69) (0.47) (1.10) (0.52) (0.54) (0.70)
Leverage -0.1792 -0.1450 -0.1896 -0.1823 -0.1702 -0.1594 -0.1780 -0.1860 -0.1984
(-1.24) (-1.04) (-1.31) (-1.26) (-1.19) (-1.12) (-1.24) (-1.29) (-1.38)
Illiquidity -1.9437*** -1.9655*** -1.9864*** -1.9206*** -1.9143*** -2.0243*** -1.9160*** -1.9228*** -1.9592***
(-7.72) (-7.76) (-7.63) (-7.69) (-7.81) (-7.73) (-7.67) (-7.69) (-7.60)
Volatility 24.6971*** 24.7225*** 24.6263*** 24.8857*** 25.3041*** 25.1617*** 25.1147*** 24.8520*** 22.6003***
(12.88) (13.22) (13.25) (12.97) (13.15) (13.23) (13.19) (13.19) (12.12)
LnAnalyst 0.4028*** 0.4263*** 0.4055*** 0.3889*** 0.3898*** 0.4100*** 0.3928*** 0.3958*** 0.3982***
(8.03) (8.91) (8.22) (7.97) (7.83) (8.25) (8.01) (8.09) (8.04)
ShortInterest 3.3694*** 3.3607*** 3.3981*** 3.3624*** 3.3607*** 3.3728*** 3.3456*** 3.3498*** 3.3651***
(14.38) (13.99) (14.59) (14.48) (14.39) (14.64) (14.30) (14.45) (14.40)
FF 48 FE YES YES YES YES YES YES YES YES YES
QTR FE YES YES YES YES YES YES YES YES YES
Constant -8.0300*** -8.0200*** -8.0907*** -8.0488*** -8.0039*** -8.2710*** -8.0429*** -8.0814*** -8.0323***
(-17.68) (-18.05) (-17.84) (-17.70) (-17.59) (-18.27) (-17.49) (-17.76) (-17.76)
Observations 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767 181,767
Pseudo R2 0.143 0.144 0.144 0.144 0.143 0.145 0.143 0.144 0.143
76
Panel C: Return consequence test
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation -0.0059 -0.0113* -0.0237*** -0.0354*** -0.1000*** -0.8863***
(-0.96) (-1.65) (-3.87) (-4.22) (-5.53) (-5.68)
Uncertainty -0.0417*** -0.0481*** -0.0469*** -0.0557*** -0.0774*** -0.1030
(-7.04) (-5.53) (-4.03) (-4.30) (-3.69) (-0.68)
Size 0.0016*** 0.0027*** 0.0026*** 0.0028** 0.0064*** 0.0727***
(2.94) (2.79) (2.73) (2.18) (3.22) (3.79)
Leverage 0.0045 0.0014 -0.0009 -0.0000 0.0112 -0.0528
(1.20) (0.30) (-0.18) (-0.00) (0.62) (-0.33)
Illiquidity -0.0094 -0.0184* -0.0169 -0.0129 -0.0094 0.0090
(-0.74) (-1.76) (-1.09) (-0.70) (-0.23) (0.04)
Volatility 0.0081 -0.1568* -0.1559 -0.2835* -0.6668 -10.3406***
(0.12) (-1.71) (-1.27) (-1.73) (-1.38) (-2.85)
LnAnalyst 0.0060*** 0.0067*** 0.0086*** 0.0096*** 0.0199*** 0.0280
(5.37) (3.45) (3.69) (2.93) (3.82) (0.68)
ShortInterest 0.0053 0.0083** 0.0118*** 0.0202*** 0.0401** 0.3076
(1.56) (1.97) (2.96) (2.98) (2.25) (1.37)
EarnAnnounce[-5,0] 0.0047 0.0000 0.0008 0.0018 -0.0055 0.0393
(1.44) (0.00) (0.16) (0.31) (-0.70) (0.81)
AnForecast[-5,0] -0.0026* -0.0028 -0.0020 -0.0053 -0.0048 -0.0666**
(-1.65) (-1.42) (-0.83) (-1.47) (-0.75) (-2.01)
ConfCall[-5,0] 0.0040 0.0138 0.0149 0.0124 0.0108 0.0385
(0.50) (1.16) (1.12) (0.91) (0.44) (0.37)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0441 -0.0204 -0.0200 -0.0686 -0.1958 -0.0978
(-1.16) (-0.98) (-0.59) (-0.99) (-1.35) (-0.31)
Observations 5,702 5,701 5,699 5,694 5,629 4,474
Adjusted R2 0.075 0.091 0.083 0.080 0.089 0.182
77
Table A10: Excluding Observations with Analyst Revisions, Earnings Announcements, or
Conference Calls in the Previous Five Days
This table only uses these activist short-selling cases of which there are no analyst revisions, earnings announcements,
or conference calls in the five days prior to the activist short-selling date. The purpose is to (partially) address one
alternative explanation that the results can be driven by other informational events that coincide with activist short-
selling.
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week) CAR
(1Month) CAR
(1Year)
Overvaluation -0.0064 -0.0174** -0.0315*** -0.0461*** -0.0891*** -0.8655***
(-0.92) (-2.13) (-3.55) (-4.33) (-3.79) (-5.10)
Uncertainty -0.0550*** -0.0608*** -0.0601*** -0.0654*** -0.1081*** -0.2394
(-6.18) (-5.63) (-3.97) (-3.60) (-3.69) (-1.01)
Size 0.0019** 0.0024** 0.0018 0.0013 0.0068* 0.0637**
(2.38) (2.25) (1.53) (0.88) (1.93) (2.26)
Leverage 0.0071 0.0029 0.0006 -0.0001 0.0103 -0.0493
(1.34) (0.44) (0.08) (-0.01) (0.47) (-0.25)
Illiquidity 0.0057 -0.0076 -0.0120 -0.0169 -0.0013 -0.0788
(0.58) (-0.62) (-0.67) (-0.78) (-0.03) (-0.24)
Volatility 0.0293 -0.2070 -0.2089 -0.4220** -0.7857 -10.8708***
(0.32) (-1.59) (-1.37) (-2.14) (-1.54) (-3.16)
LnAnalyst 0.0049*** 0.0065*** 0.0094*** 0.0100*** 0.0157** 0.0204
(3.21) (3.31) (4.29) (3.66) (2.53) (0.47)
ShortInterest 0.0058 0.0070 0.0093 0.0204* 0.0266 0.0833
(0.85) (0.59) (0.92) (1.71) (0.98) (0.41)
FF 48 FE YES YES YES YES YES YES
QTR FE YES YES YES YES YES YES
Constant -0.0239 -0.0237** -0.0214* -0.0428* -0.0911 0.2022
(-1.34) (-2.19) (-1.78) (-1.72) (-1.15) (0.88)
Observations 3,271 3,270 3,268 3,263 3,229 2,576
Adjusted R2 0.095 0.115 0.102 0.095 0.088 0.182
78
Table A11: Control for Short-Seller Fixed Effects
This table presents results controlling for short-seller fixed effects. The purpose of doing so is to show that target-firm
characteristics are important conditional on the short-seller characteristics.
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation -0.0065 -0.0145* -0.0250*** -0.0362*** -0.0685*** -0.4855***
(-0.87) (-1.70) (-2.77) (-3.27) (-2.87) (-2.65)
Uncertainty -0.0343*** -0.0437*** -0.0338** -0.0440*** -0.0974*** -0.1097
(-4.01) (-4.10) (-2.48) (-2.65) (-3.17) (-0.46)
Size 0.0017* 0.0037*** 0.0040*** 0.0046*** 0.0119*** 0.1387***
(1.94) (3.31) (3.24) (2.70) (3.37) (4.29)
Leverage -0.0027 -0.0048 -0.0094 -0.0037 -0.0155 -0.0959
(-0.51) (-0.80) (-1.36) (-0.39) (-0.76) (-0.46)
Illiquidity -0.0129 -0.0220 -0.0159 -0.0066 0.0029 0.2248
(-0.78) (-1.36) (-0.81) (-0.31) (0.07) (0.83)
Volatility 0.0282 -0.1087 -0.1301 -0.3481** -0.9118* -11.3919***
(0.33) (-0.97) (-0.95) (-1.98) (-1.93) (-3.27)
LnAnalyst 0.0051*** 0.0055** 0.0085*** 0.0093*** 0.0117* -0.0173
(2.97) (2.42) (3.51) (3.09) (1.81) (-0.36)
ShortInterest 0.0148* 0.0192** 0.0251** 0.0371*** 0.0619** 0.4839**
(1.84) (2.02) (2.42) (2.74) (2.10) (2.27)
EarnAnnounce[-5,0] 0.0031 -0.0032 -0.0025 -0.0002 -0.0023 0.0272
(0.72) (-0.53) (-0.34) (-0.03) (-0.23) (0.42)
AnForecast[-5,0] -0.0020 -0.0018 -0.0007 -0.0032 -0.0030 -0.0337
(-0.87) (-0.60) (-0.20) (-0.68) (-0.38) (-0.77)
ConfCall[-5,0] 0.0026 0.0120 0.0113 0.0067 0.0240 -0.0598
(0.29) (0.89) (0.82) (0.44) (0.89) (-0.41)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Short-Seller FE Yes Yes Yes Yes Yes Yes
Constant -0.0235 0.0006 -0.0168 -0.0803 -0.1567 -0.0317
(-0.44) (0.02) (-0.32) (-0.80) (-0.73) (-0.07)
Observations 5,697 5,696 5,694 5,689 5,624 4,469
Adjusted R2 0.068 0.099 0.113 0.114 0.154 0.301
79
Table A12: Non-Missing Values for All CARs
This table presents results only using observations with non-missing values of all CARs. As a result the same
observations are used in all columns.
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation -0.0003 -0.0048 -0.0186*** -0.0315*** -0.0917*** -0.8863***
(-0.05) (-0.76) (-2.72) (-3.60) (-4.39) (-5.92)
Uncertainty -0.0425*** -0.0526*** -0.0504*** -0.0598*** -0.0743*** -0.1030
(-5.09) (-5.52) (-3.57) (-3.32) (-2.87) (-0.48)
Size 0.0016** 0.0025*** 0.0025** 0.0032** 0.0071** 0.0727***
(2.31) (2.88) (2.51) (2.34) (2.41) (3.10)
Leverage 0.0038 -0.0009 -0.0046 -0.0074 -0.0001 -0.0528
(0.89) (-0.18) (-0.76) (-0.80) (-0.01) (-0.31)
Illiquidity -0.0098 -0.0229* -0.0181 -0.0217 -0.0005 0.0090
(-0.65) (-1.71) (-1.03) (-1.03) (-0.01) (0.03)
Volatility 0.0088 -0.1826* -0.1899 -0.3945** -0.6948 -10.3406***
(0.12) (-1.74) (-1.54) (-2.49) (-1.55) (-3.27)
LnAnalyst 0.0055*** 0.0057*** 0.0070*** 0.0066** 0.0174*** 0.0280
(3.97) (3.09) (3.40) (2.56) (3.11) (0.71)
ShortInterest 0.0041 0.0056 0.0099 0.0202* 0.0319 0.3076
(0.77) (0.75) (1.26) (1.92) (1.25) (1.41)
EarnAnnounce[-5,0] 0.0063 0.0018 0.0036 0.0049 -0.0005 0.0393
(1.59) (0.32) (0.54) (0.66) (-0.05) (0.64)
AnForecast[-5,0] -0.0014 -0.0014 -0.0007 -0.0025 -0.0080 -0.0666
(-0.67) (-0.53) (-0.24) (-0.63) (-1.02) (-1.57)
ConfCall[-5,0] 0.0016 0.0103 0.0102 0.0069 0.0088 0.0385
(0.21) (0.92) (0.86) (0.53) (0.36) (0.35)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0250* -0.0255*** -0.0254** -0.0424* -0.0988 -0.0633
(-1.89) (-3.05) (-2.26) (-1.78) (-1.50) (-0.28)
Observations 4,474 4,474 4,474 4,474 4,474 4,474
Adjusted R2 0.081 0.096 0.085 0.088 0.085 0.182
80
Table A13: The Role of Illiquidity
This table provides additional suggestive evidence on Higher-Order Beliefs (HOB) view. I find that the sensitivity of
return to Uncertainty is stronger in the more illiquid sample than in the less illiquid sample. The Amihud illiquidity
measure captures the price impact of a certain trading volume. For illiquid stocks, an investor could only sell at a
much lower price when other investors sell the stocks. For liquid stocks, the same amount of shares sold by other
people do not matter that much. In other words, investors of illiquid stocks would more likely panic. In the presence
of activist short-selling, for high uncertainty firms, investors would put more weight on the short-seller’s signal, but
the weight would be even higher when they care more about other people’s exit.
Panel A: High illiquidity sample
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation -0.0064 -0.0165 -0.0309*** -0.0465*** -0.0851*** -0.6943***
(-0.71) (-1.62) (-2.80) (-3.34) (-3.21) (-3.26)
Uncertainty -0.0446*** -0.0493*** -0.0462*** -0.0576*** -0.0741*** -0.1406
(-5.23) (-4.81) (-3.42) (-3.49) (-2.62) (-0.56)
Size 0.0018 0.0031** 0.0029* 0.0013 0.0088* 0.1062***
(1.60) (1.99) (1.65) (0.56) (1.93) (2.61)
Leverage 0.0030 -0.0018 -0.0076 -0.0065 0.0113 -0.0719
(0.58) (-0.27) (-0.98) (-0.63) (0.55) (-0.35)
Illiquidity -0.0092 -0.0157 -0.0124 -0.0100 0.0115 0.1206
(-0.68) (-1.24) (-0.76) (-0.52) (0.29) (0.41)
Volatility -0.0100 -0.2301* -0.2554* -0.4810*** -1.1346** -15.3534***
(-0.11) (-1.90) (-1.74) (-2.60) (-2.32) (-4.02)
LnAnalyst 0.0069*** 0.0096*** 0.0124*** 0.0141*** 0.0244*** 0.0335
(3.62) (3.76) (4.32) (3.91) (3.33) (0.65)
ShortInterest -0.0074 -0.0028 0.0028 0.0186 0.0393 0.2369
(-0.73) (-0.19) (0.18) (0.95) (1.04) (0.63)
EarnAnnounce[-5,0] 0.0099* 0.0046 0.0065 0.0048 -0.0150 -0.0480
(1.75) (0.71) (0.91) (0.58) (-0.93) (-0.48)
AnForecast[-5,0] -0.0047 -0.0040 -0.0017 -0.0052 0.0009 -0.0626
(-1.27) (-0.93) (-0.34) (-0.84) (0.08) (-0.90)
ConfCall[-5,0] -0.0127 0.0167 0.0223 0.0359 0.0558 0.1396
(-0.67) (1.05) (1.28) (1.56) (1.22) (0.52)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0194 -0.0284** -0.0269* -0.0528* -0.1245 0.1144
(-0.99) (-2.44) (-1.75) (-1.84) (-1.26) (0.37)
Observations 2,865 2,865 2,865 2,862 2,837 2,274
Adjusted R2 0.055 0.079 0.074 0.075 0.079 0.195
81
Panel B: Low illiquidity sample
(1) (2) (3) (4) (5) (6)
AR(0) CAR(0,1) CAR(0,2) CAR
(1Week)
CAR
(1Month)
CAR
(1Year)
Overvaluation -0.0026 -0.0051 -0.0153** -0.0212** -0.0850*** -0.9142***
(-0.53) (-0.74) (-1.97) (-2.11) (-3.74) (-5.28)
Uncertainty -0.0049 -0.0061 -0.0161 -0.0103 -0.0258 -0.0198
(-0.42) (-0.45) (-1.01) (-0.46) (-0.50) (-0.03)
Size 0.0027*** 0.0033** 0.0036** 0.0054*** 0.0138*** 0.1027**
(2.84) (2.57) (2.28) (2.70) (3.54) (2.03)
Leverage -0.0015 0.0017 0.0057 0.0069 0.0076 0.2001
(-0.28) (0.25) (0.73) (0.69) (0.32) (0.78)
Illiquidity 11.4002 -1.8695 -2.5690 4.1218 106.1941** 729.6384*
(1.25) (-0.17) (-0.20) (0.23) (2.33) (1.78)
Volatility 0.0218 0.0566 0.0837 0.2087 0.2787 2.7574
(0.23) (0.39) (0.48) (0.92) (0.51) (0.59)
LnAnalyst 0.0042*** 0.0016 0.0021 0.0021 0.0230** 0.1446**
(2.74) (0.72) (0.81) (0.60) (2.47) (1.99)
ShortInterest 0.0086 0.0017 0.0062 0.0120 0.0355 0.4572**
(1.49) (0.19) (0.60) (0.91) (1.14) (2.21)
EarnAnnounce[-5,0] 0.0008 -0.0034 -0.0034 0.0002 0.0088 0.1379**
(0.18) (-0.46) (-0.38) (0.02) (1.03) (2.03)
AnForecast[-5,0] -0.0017 -0.0029 -0.0033 -0.0060 -0.0077 -0.0330
(-1.07) (-1.29) (-1.08) (-1.44) (-1.27) (-0.70)
ConfCall[-5,0] 0.0113 0.0140 0.0122 0.0015 -0.0072 -0.0251
(1.54) (0.98) (0.81) (0.09) (-0.26) (-0.20)
FF 48 FE Yes Yes Yes Yes Yes Yes
QTR FE Yes Yes Yes Yes Yes Yes
Constant -0.0367*** -0.0217 -0.0153 -0.0276 -0.1785*** -1.3369**
(-3.06) (-1.27) (-0.77) (-1.00) (-2.91) (-2.07)
Observations 2,837 2,836 2,834 2,832 2,792 2,200
Adjusted R2 0.011 0.017 0.029 0.039 0.088 0.260
82
Figure A1: Separate Winners vs. Losers in both SA and ASR
This figure plots the return patterns of targets in the window of [-60, 250] around the activist short-selling dates for
both SA (Panel A) and ASR (Panel B) sample.
Panel A: Separating winners vs. losers for in the SA sample
Panel B: Separating winners vs. losers for in the ASR sample
83
Chapter 2 Selling Financial Analysts Short: The Impact of Activist Short-
Selling on Sell-Side Analysts
1 Introduction
Financial analysts have indispensable roles in capital markets as information intermediaries.
However, the sell-side business model does not guarantee (and arguably frequently fail to provide)
objective, independent, and effective research largely because of the inherent incentive problems.
Prior research has shown that analysts are motivated to bias optimistically to generate trading
commissions (e.g., Jackson 2005; Beyer and Guttman 2011), attract underwriting business (e.g.,
Lin and McNichols 1997; Michaely and Womack 1999) or cater to management for future
favorable access of their private information (e.g., Chen and Matsumoto 2006). As the survey of
Brown, Call, Clement, and Sharp (2015) shows, private communication with management is “even
more useful to analysts than their own primary research, the firms’ recent earnings performance,
and the recent 10-K or 10-Q reports” (page 10). Meanwhile, prior research has well documented
the fact that management has strong incentives to withhold negative news (e.g., Kothari, Shu, and
Wysocki 2009). In a world where analysts rely on management for information and strategically
disseminate management-favorable message, negative information would be in short supply and
analysts’ roles as information intermediaries would be severely compromised.
Among all market participants who could meet the demand of negative information, short-
sellers are arguably best equipped with incentives and capabilities to do so (e.g., Karpoff and Lou
2010). While most short-selling is done quietly in which negative information is revealed through
the trade (i.e., short-positions), some short-sellers take an aggressive approach by publicly talking
down securities, a recent financial innovation called activist short-selling by the investing
community (Ljungqvist and Qian 2016; Zhao 2017). In this way, they can more quickly
disseminate negative information into the market and accelerate the price correction. More
importantly, activist short-sellers also frequently accuse sell-side’s systematic optimism (or
“talking-up”) as a major driver of security overvaluation. For example, almost every Citron
84
Research report makes fun of Wall Street analysts, claiming that they should have done a better
job with higher diligence and intelligence.1 Such an aggressive approach makes activist short-
sellers a unique source of negative information and usually puts analysts into the corner such that
they feel obliged to respond. Anecdotal evidence shows that analysts pay great attention to
prominent short-selling campaigns such as Citron Research’s attack on Valeant in October 2015.2
Recently, a news report mentions that “now his [Carson Block of Muddy Waters Research –
another prominent activist short-seller] reports are a regular part of Wall Street’s conversation.”3
However, prior literature largely ignores the consequences of short-selling activities (let
alone activist short-selling) on the sell-side analyst community. Several papers on the information
flow between analysts and short-sellers either examine the possibilities of analysts’ tipping short-
sellers before downgrades (Christophe, Ferri, and Hsieh 2010) and short-sellers’ predicting
analysts’ downgrades (Boehmer, Jones, and Zhang 2015), or report no evidence on analysts’
reaction to short-selling reports (Ljungqvist and Qian 2016). To fill the gap in literature, I ask four
related questions in this paper. First, do analysts react to activist short-selling allegations on their
covered firms? Second, what factors explain variation in analysts’ reactions? Third, does activist
short-selling damage analysts’ reputations? Finally, does activist short-selling affect analysts’
career paths?
I start from the large sample of 6,081 activist short-selling cases from 2006 to 2015
complied by Zhao (2017) who combines information from Seeking Alpha (SeekingAlpha.com,
SA hereafter) and Activist Shorts Research (ActivistShorts.com, ASR hereafter). SA is the largest
crowdsourced investing platform appealing to non-celebrity short-sellers, while ASR tracks all
1 For the most vivid illustration, see page 12 of this report on JCOM: http://www.citronresearch.com/wp-
content/uploads/2016/03/JCOM-final-c21.pdf. 2 For example, in a note to clients, Nomura analyst Shibani Malhotra said that the weakness in Valeant's stock offers
a buying opportunity, as the brokerage firm believes the concern raised by Citron is likely “misinformation” (see
http://www.reuters.com/article/us-valeant-citron-idUSKCN0SF22520151021). By contrast, BMO analyst Alex
Arfaei said that “we cannot defend the specialty pharmacy structure,” and downgraded Valeant’s shares to “market
perform” (see http://business.financialpost.com/investing/global-investor/valeant-pharmaceuticals-international-inc-
is-tanking-again-today-after-longtime-bull-downgrades-stock). Further, User Raffat of Evercore ISI and Gregg
Gilbert from Deutsche Bank even terminated the coverage because the price was no longer determined by value (see
http://www.streetinsider.com/Analyst+Comments/Deutsche+Bank+Suspends+Coverage+on+Valeant+Pharma+(VR
X)/11374531.html). 3 See http://uk.businessinsider.com/carson-block-told-us-he-regrets-not-having-shorted-one-stock-in-2016-2016-10.
85
influential activist short-selling events, providing a good complement to the SA sample. As many
firms are targeted by multiple times in a short period of time, I focus on the activist short-selling
case with the most negative returns in the first two days following short-sellers’ public talking-
down if a firm is targeted by multiple times in a year, leaving me 2,583 activist short-selling cases.
To start with, I confirm that analysts do react to activist short-selling: they are more likely
to lower and less likely to raise price targets if short-sellers seem to be right (i.e., initial market
reaction is negative), but they are more likely to raise and less likely to lower price targets if short-
sellers seem to be wrong (i.e., initial market reaction is positive). These findings contrast to
Ljungqvist and Qian’s (2016) conclusion that “there is little evidence that analysts consider
arb[arbitrageur]’s information in their recommendations” (page 1986) based on the finding that
the distribution of recommendation types largely remains stable after firms are targeted by short-
selling reports. Note changing recommendations is not a necessary condition to infer that analysts
are considering new information. It could be that analysts keep the Buy recommendations only
because the price has declined so much that creates a new Buy opportunity after the activist short-
selling. As a result, my approach using price targets is more powerful in examining whether
analysts react to activist short-selling.
Although analysts on average revise down their target prices when short-sellers seem to be
right (i.e., prices drop after activist short-selling) and revise up otherwise, there is still large
variation in the direction and timeliness of their revisions. I find that analysts are more likely to
delay their revisions when the short-sellers seem to be wrong or when analysts are employed by
smaller brokerage houses, have a shorter experience in the profession, have a longer history of
covering the target firm, provide bullish forecasts before targeted, or only cover fewer firms.
Similarly, analysts are more likely to raise their target prices when the short-sellers seem to be
wrong or when they have a longer history of covering the target firm, present less optimism in
previous forecasts, have fewer other analysts covering the firm, or have a style of frequent
revisions. These results suggest that analysts’ incentives (i.e., history with the target firm and the
size of coverage portfolio) and abilities (i.e., size of the brokerage house and the experience as an
analyst) as well as the initial impact of the activist short-selling determine whether and how
analysts react to activist short-selling.
86
Next I examine whether and how analysts’ reputations are damaged by short-sellers’ public
talking-down behavior. Following prior literature, I investigate how the market reacts to analysts’
EPS revisions differently after activist short-selling. Depending on the CAR windows I use, the
market-reaction spreads (i.e., the difference in market reactions to the top and bottom decile of
analyst revisions) are reduced by 26% to 30% after activist short-selling. In other words, investors
are less sensitive/responsive to those analysts’ revisions afterwards. Note this is not because
investors become less responsive to all information signals after activist short-selling. For example,
they are comparably responsive to quarterly earnings announcements before and after activist
short-selling. As a result, I label this reduction in market reaction spreads to analyst revisions as
analysts’ reputation loss.
Further, I investigate under what circumstances analysts can avoid or suffer more
reputation losses. Intuitively, if analysts are relatively conservative, they would be less humiliated
by activist short-sellers. Also, if the short-selling campaign does not lead to substantial negative
return, they could also avoid reputation loss because the market is not convinced by short-sellers.
Indeed, the reputation loss is not significant when analysts’ last target price for a given stock before
activist short-selling is below the median (i.e., relatively less bullish) or when the first two-day
return after activist short-selling is positive (i.e., short-sellers seem to be wrong). In particular, the
market is 50% less responsive to analysts’ revisions (i.e., 50% loss in reputation) if they are
relatively bullish and short-sellers seem to be right.
I also investigate whether analysts can save their reputation by reacting in a certain way
after activist short-selling. Ex ante, it is unclear whether reacting in a more timely or “humble”
manner (i.e., revising down rather than defending previous optimistic positions) could help,
because the market could interpret such actions as a lack of ability and confidence in their own
stances. I find that timely reactions help: in a quarter of cases in which analysts react most timely,
the reputation loss is not statistically significant, but for the quarter of cases in which analysts who
provide the least timely reactions, the reputation loss is highly significant. But surprisingly, the
direction of revision does not seem to matter much – analysts suffer from reputation loss no matter
whether she revises up or down.
87
Finally, I examine the career consequences of activist short-selling on analysts. Specifically,
I investigate whether an analyst moves to a smaller brokerage house one, two, and three years after
activist short-selling. I find that an analyst is more likely to move to a smaller brokerage house if
her covered firms are targeted by activist short-sellers. Further, the probability of moving down is
increasing in the number of covered firms being targeted. The pattern lasts for at least three years,
indicating that activist short-selling has a persistent impact on the analysts’ career path. Additional
analyses show that analysts whose covered firms are targeted are less likely to exit the analyst
business, suggesting that their outside option is also limited by the short-selling attacks. This is
consistent with Clement and Law (2016) who show that high-ability analysts hired under a tight
labor market are more likely to leave the profession after the condition improves.
This chapter of my dissertation makes the following contributions. First, this paper adds to
the emerging literature on the consequences of competition among information intermediaries.
Prior literature documents that biases are reduced by competition among analysts (Hong and
Kacperczyk 2010; Merkley, Michaely, and Pacelli 2017) and among credit rating agencies (Becker
and Milbourn 2011; Doherty, Kartasheva, and Phillips 2012; Xia 2014). While these papers focus
on the competition among the same type of intermediaries, I examine the consequences of
competition between different types. In this spirit, a recent study by Jame, Markov, and Wolfe
(2017) is closest to mine: they document a disciplinary role of Estimize, a crowdsourced platform
for short-term earnings forecasts, on sell-side analysts. They find that analysts reduce short-term
forecast bias after their covered firms are added to Estimize. However, they have not considered
how quick analysts react to Estimize forecasts and the reputation and career consequences of this
crowdsourced platform on the sell-side community.
Second, this paper reconciles the contradictions between real-world cases and Ljungqvist
and Qian’s (2016) conclusion that analysts largely ignore short-selling reports based on the
observation that the distribution of recommendations remains stable for a long time. More broadly,
this study adds a new perspective to the literature regarding the information flow between analysts
and short-sellers. While prior literature provides evidence in support of either the tipping
hypothesis (i.e., analysts give tips to short-sellers, who therefore can trade prior to analyst
downgrades) or the prediction hypothesis (i.e., short-selling activities predict analyst downgrades).
88
This paper raises and provides evidence for a new reaction hypothesis (i.e., analysts react to short-
sellers’ activities).
Third, this article contributes to the literature on analyst reputation. Prior research has
documented analysts’ incentives to build reputation among investors (e.g., Stickel 1992; Jackson
2005). I focus on a scenario in which analysts’ reputations are severely damaged. As a result, my
study complements a recent paper by Lee and Lo (2016) who show that analysts who provide
bullish forecasts prior to restatements suffer from loss in reputation. However, restatement is a
very different setting from activist short-selling, because analysts know that they are definitely
wrong after firms restate the prior mistakes, but it is usually unclear whether they are wrong in the
presence of activist short-selling – that’s why I also find huge variation in their reactions and
provide recommendations on how they should react to avoid reputation loss – react quickly.
Finally, this paper adds to the literature on analysts’ career concerns. Prior literature finds
that some certain actions can lead to better or worse career consequences (e.g., Hong, Kubik, and
Solomon 2000; Hong and Kubik 2003; Ke and Yu 2006). In particular, Hong and Kubik (2003)
find that controlling for accuracy, analysts who are optimistic relative to the consensus are more
likely to move to more prestigious brokerage houses. This paper shows that such an “optimism”
strategy is risky – it could eventually hurt analysts’ job prospects if the optimism attracts activist
short-sellers.
The remainder of this paper is organized as follows. Section 2 reviews related literature.
Sections 3 examines whether analysts revise their price targets after activist short-selling. Section
4 explores variation in analysts’ reactions in target-price revisions. Section 5 focuses on whether
and how analysts’ reputations are damaged by activist short-sellers. Section 6 investigates the
impact of activist short-selling on analysts’ career prospects. Section 7 presents several robustness
checks and supplemental analyses. Section 8 concludes.
89
2 Related Literature
2.1 Sell-Side Optimism
Prior research has documented the systematic optimism of sell-side outputs at least since
Brown, Foster, and Noreen (1985). Although recent research documents that analysts “walk-down”
their EPS forecasts over the year to make it easier for managers to meet or beat (e.g., Richardson,
Teoh, Wysocki 2004), overwhelming evidence shows that they bias their investing
recommendations optimistically. Several incentive-based explanations are raised to answer why
analysts have such optimism. 4 First, in the presence of short-selling constraints, optimistic
recommendations are more likely to generate trading commissions than pessimistic forecasts
(Jackson 2005). Second, analysts may be motivated to provide optimistic recommendations to
cater management for the purpose of either gaining underwriting business deals (Lin and
McNichols 1998) or maintaining the assess to management’s private information (Francis and
Philbrick 1993; Soltes 2014). Third, analysts also have incentives to promote stocks that are held
by fund families affiliated with their brokerage houses (Mola and Guidolin 2009). Note that being
optimistic seems a rational strategy for analysts from a career perspective, as Hong and Kubik
(2003) show that optimism is an important predictor of analysts’ career success (as measured by
moving to more prestigious brokerage houses) in addition to accuracy.
Such prevailing optimism probably explains why following analysts’ recommendations
could be sometimes costly. For example, Jegadeesh, Kim, Krische, and Lee (2004) show that
analysts tend to recommend stocks with quantitative characteristics that are associated with
negative future returns, such as low book-to-market, high sales growth, and high accruals. Indeed,
as long as analysts rely on management for information and have incentive to mute potentially
negative information, they will tend to over-recommend. In other words, information provided by
4 While these incentive-based explanations claim that analysts add bias to their true beliefs, prior research also suggests
alternative explanations in which analysts are reporting their true beliefs. For example, McNichols and O’Brien (1997)
discuss the possibility that analysts are more likely to report on stocks about which they have favorable views.
Similarly, Hayes (1998) builds a model to illustrate that analysts have stronger incentives to cover firms that they
expect to perform well.
90
the analysts is skewed to the positive side, creating a need for alternative source of negative
information.
2.2 Activist Short-Sellers as a Source of Negative Research
Among all market participants, short-sellers are arguably the best source of negative
information as they are best-equipped with the right incentives and capabilities. Note that short-
sellers have a unique business model of benefiting from the decline of stock prices. As a result,
they have strong incentives to collect, process, and disseminate negative information into the
marketplace. Prior research has documented that they are among the most sophisticated players in
the financial market. For example, Aitken, Frino, McCorry, and Swan (1998) use intraday data
and show that short sales are instantaneously bad news. Such sales can help impound adverse
information into stock prices within fifteen minutes. Dechow, Hutton, Meulbroek, and Sloan (2001)
confirm that short sellers are sophisticated traders in that they use fundamental analysis to exploit
the lower expected future return of firms with lower ratio of fundamentals to market values. Further,
Drake, Rees, and Swanson (2011) recommend that we follow short-sellers rather than analysts
when they disagree with each other.
Prior literature focuses on the aspect that short-sellers incorporate their negative
information into prices through trading (i.e., taking short-positions). However, as short-selling is
a risky business, short-sellers may avoid taking short-positions in the presence of analysts’ public
talking-up, which could amplify the noise-trader risk (De Long, Shleifer, Summers, and
Waldmann 1990). This dilemma can be partially solved by a financial innovation commonly
referred to as “activist short-selling” in the investing community, where short-sellers publicly talk
down stocks to benefit their short positions (Ljungqvist and Qian 2016; Zhao 2017). They usually
provide arguments and evidence on why the targeted stocks are severely overvalued and should
be shorted. As a public signal informing the market, they can not only potentially reduce the noise-
trader risk, but also create panic among existing shareholders who might try to sell as soon as
possible (Zhao 2017).
Such public talking-down exerts substantial threats to analysts. Different from the situation
where short-sellers take positions quietly, analysts are almost “forced” to join in a public debate
91
regarding the valuation of stocks. What makes the situation even more intense is that many activist
short-sellers specifically highlight analysts’ mistakes and attack their misaligned incentives and
dependence on management for information. Under much greater public scrutiny, it’s likely that
some analysts feel obliged to respond – either acknowledging the merit in the short-thesis or
defending their previous positions. However, it’s also possible that some other analysts are
concerned about taking a wrong stance in a public debate and remain silent for a long time.
2.3 Prior Research on the Information Flows between Short-Sellers
and Analysts
Prior research has documented abnormally high short-selling activities prior to analyst
downgrades. There are at least three explanations: (1) analysts provide tips to short-sellers before
they downgrade (or tipping), (2) short-sellers predict the downgrades with their sophistication (or
prediction), and (3) analysts learn from and react to the short-selling activities (or reaction).
Christophe, Ferri, and Hsieh (2010) use 670 downgrades for NASDAQ stocks in 2000 and 2001
and find abnormal levels of short-selling in the three days before downgrades are publicly
announced. They claim that their evidence is more consistent with the tipping explanation than the
prediction explanation based on two sets of analyses. First, they find that firms with higher
abnormal short-interest are not likely to report negative earnings surprise. Second, abnormal short-
interest in day (-3, -1) is significantly higher than that in day (-10, -6), where day (0) is the day
when analysts downgrade stocks.
Boehmer, Jones, and Zhang (2015) use a five-year panel of proprietary NYSE short sale
order data and find heavier shorting in the week before analyst downgrades. They argue that their
evidence cannot be fully explained by tipping, but consistent with the prediction explanation,
because short-sales have incremental predictability in stock underperformance beyond analyst
downgrades, suggesting that short-sellers have information that analysts do not know and therefore
cannot tip.
To the best of my knowledge, Ljungqvist and Qian (2016) is the only paper that discusses
the reaction hypothesis. However, they conclude that analysts are very slow in incorporating
information from the short-selling reports based on the observation that analysts’
92
recommendations stay very stable for one year. This is surprising and inconsistent with the
anecdote evidence we observe in the real world, for example, from the Valeant vs. Citron Research
case I mention earlier. They correctly conclude that their evidence is “consistent with the well-
known tendency among analysts to avoid annoying top management at companies they cover,”
but no change in recommendation does not mean that analysts are not reacting to these short-
selling reports, as I explain in the next section.
3 Do Financial Analysts React to Activist Short-Selling?
To examine whether analysts react to activist short-selling, I make use of 6,081 activist
short-selling constructed by Zhao (2017) from Seeking Alpha (SA) and Activists Shorts Research
(ASR). Whereas SA is the largest crowdsourced investing platform therefore an ideal for non-
celebrity shorts, ASR tracks short-selling campaigns waged by prominent traders. For those short
theses published on SA, Zhao (2017) defines activist short-selling as those articles in which the
authors disclose explicitly that they hold short-positions in the discussed stocks. Appendix A
explains how these cases are identified.
Note that some firms are targeted by activist short-sellers by multiple times within a short
period of time. To increase the power of the analyses, for each firm in a given year, I focus on one
activist short-selling case that leads to the biggest price decline in the first two days starting from
the activist short-selling date.5 To illustrate, suppose Firm A is targeted by short-seller I, II, and
III on February 1, April 1, and June 1, 2010, respectively and the two-day raw returns of these
three campaigns are -1%, -5%, and 1%. I only focus on the case that Firm A is targeted by the
short-seller II on April 1, 2010. All analyst forecasts prior to April 1, 2010 are defined as pre-
activist short-selling, and those after that date are defined as post-activist short-selling. This
approach leaves me 2,583 activist short-selling cases, each of which corresponds to a unique firm-
year. Figure 1 presents the distribution by year. There is a clear trend that more and more firms are
targeted by activist short-sellers in recent years.
5 All inferences remain if I either use all activist short-selling cases or only use the first case for each firm-year.
93
I start my analysis by replicating what Ljungqvist and Qian (2016) find in their much
smaller sample regarding analyst recommendations around short-selling reports. In Figure 2 I plot
the number of recommendations by type (i.e., Strong Sell, Sell, Hold, Buy, and Strong Buy) in 18
five-day windows (i.e., period) prior to and after the activist short-selling. We can find that except
there is a spike of Hold recommendations around the activist short-selling date, the numbers of
other types of recommendations remain stable, largely consistent with Ljungqvist and Qian (2016).
As a result, it is tempting to conclude that analysts do not react to activist short-selling as
Ljungqvist and Qian (2016) do.
I argue that no change in recommendation does not mean that analysts are not reacting to
these short-selling reports. First, they could incorporate the new information into targeted prices
and forecast revisions, which are largely continuous, rather than into recommendations, which are
discrete with only five ratings. Second, it’s possible that analysts keep favorable (e.g., Strong Buy)
ratings because the price has dropped so much after the short-selling. In other words, it is possible
that they incorporate new information but still remain the same rating.
I focus on target price rather than recommendations in examining whether analysts react to
activist short-selling. The following Figure 3 illustrates how many analysts are changing their price
targets in each five-day interval before and after the activist short-selling.
We can find that for those targeted firms with negative returns in (0, 1) interval, 973
analysts lowered their target prices, a 50% increase from the previous 5-day intervals of 626. By
contrast, 460 analysts raised their target prices, slightly decrease from the previous 5-day intervals
of 498. Figure 1 Panel B illustrates the case for those targeted firms that do not witness negative
raw returns. The number of analysts who raised target prices increase by 50% from 218 to 330,
but the number of analysts who lowered target prices decreases by a quarter from 241 to 187 in
the first 5-day interval benchmarking on the previous 5-day interval. Importantly, the numbers of
raising and lowering price targets prior to activist short-selling in both panels are quite parallel.
It is very difficult to interpret Figure 3 using Tipping or Prediction explanations. First, it is
unclear why short-sellers need to put up a short-thesis when they know that analysts would revise
their forecasts down – they could have simply taken short positions quietly. In so doing, they could
profit from the short-selling but avoid the risk of activist short-selling. In particular, they can avoid
94
humiliating the analysts who provide tips. Second, it is unclear how these two hypotheses explain
why the direction of analysts’ reaction depends on the initial market reaction of activist short-
selling. For example, it is puzzling why analysts even raise target prices after activist short-selling
that fails to bring down prices. Third, note that the market reactions to activist short-selling (mean
CAR(0, 1) = -2.6%) is way larger than those to forecast revisions (mean CAR(0, 1) = -0.4%) –
activist short-sellers are unlikely to rely on such revisions for profits. So the most likely
explanation would be that analysts react to activist short-selling.
4 What Determines Financial Analysts’ Reactions to
Activist Short-Selling?
In Section 3 we find that analysts react to activist short-selling but exhibits huge variation
in the timeliness and directions of such reactions. This section explores reasons that explain such
variation in the responses. As there is no explicit theoretical guidance from the prior literature
regarding what are the determinants of analysts reactions, I include several typical short-selling
campaign-level, analyst-level, firm-level, and firm-analyst level variables that capture analysts’
incentives, abilities, and their awareness of the short-selling. I focus on those analysts who provide
at least one target-price forecast in the 180 days prior to the activist short-selling and check when
and how they revise their price targets after activist short-selling. I estimate the following linear
models using OLS:
, 0 1 , 2 , 3 ,
4 , 5 , 6 ,
7 , 7 , 9 ,
10 , 11 , ,
/ i t i t i t i t
i t i t i t
i t i t i t
i t i t t t i t
Delay Raise ShortsImpact BrokSize LnCovDays
PreTalkUp AnaExperience LnCovFirm
LnCoverage TPFrequency Size
Leverage BTM Year Firm
(1)
Where Delay is defined as the log of one plus the number of days between activist short-
selling and the analyst’s first target-price revision. Raise is an indicator defined as one if the first
revision after activist short-selling is to raise the price target, and zero otherwise. In this
exploratory test, I test whether analysts reactions are affected by (1) the initial impact of the activist
95
short-selling (ShortsImpact), defined as the first two-day raw return following the activist short-
selling date multiplied by minus one; (2) the brokerage house size (BrokSize), defined as the log
of one plus the number of unique analysts providing EPS forecasts in that year; (3) analyst’s
coverage history of the firm (LnCovDays), measured as the log of one plus the number of days
since the analyst provided the first EPS forecast on the firm; (4) analysts talk-up behavior prior to
the activist short-selling (PreTalkUp), measured as the decile rank of all analysts’ last target price
forecasts on a given firm prior to activist short-selling; (5) analyst experience (AnaExperience),
measured as the log of one plus the number of years the analyst has been in IBES universe; (6)
number of covered firms (LnCovFirm), measured by the log of one plus the number of unique
firms the analyst is providing EPS forecasts in the year; (7) number of analyst covering the firm
(LnCoverage), measured by the log of one plus the number of unique analysts who are providing
EPS to the firm in the year; and (8) target price forecasting frequency (TPFrequency), measured
as the log of one plus the number of times the analyst has provided target price forecast on the firm
in the 90 days prior to the activist short-selling. In addition, I also control for three firm-level
characteristics, including the size of the firm (Size), measured by the log of one plus the total assets
at the beginning of the fiscal year, the leverage of the firm (Leverage), measured by the ratio of
total liabilities to total assets at the beginning of the fiscal year, and the book-to-market ratio (BTM),
measured by the ratio of book value of equity to the total market capitalization at the beginning of
the fiscal year. Finally, I include firm and year fixed effects to control for factors that are specific
to each firm or each quarter. As each analyst has her own style in reacting to activist short-selling,
I cluster the standard errors at the analyst level.
Table 1, Panel A presents the summary statistics of variables used in the regressions. Some
statistics are worth mentioning. For example, the mean of Delay is 3.666. In other words, on
average analysts make the first target-price revision 48 days after activist short-selling, suggesting
that many analysts are hesitated to respond. But in more than 10% of the cases analysts respond
within the first week. Surprisingly, in half of the cases, analysts revise up the target price (Raise)
to defend the company and probably their previous positions, even though the mean of
ShortsImpact is 0.026 (i.e., on average there is substantial price drop after activist short-selling).
Table 1 Panel B presents the regression results. In general, I find that analysts are likely to
delay target-price revision if the initial market reaction is not in favor of short-sellers, if the
96
analysts are from smaller brokerage house, if analysts have longer covering history with the firm,
if the analyst is previously more bullish, and if the analyst is covering fewer firms. With respect to
the direction of reaction, analysts are more likely to revise up if the market reaction is not in favor
of the short-seller, if analysts have longer history of covering the firm, if the analyst is previously
less bullish, if there are fewer analysts covering the firm,6 and if the analyst has a style of revising
frequently. These results show that the variation in analysts’ reaction to activist short-selling is
partially explained by the impact of activist short-selling (proxied by the initial market reaction),
the ability (captured by the size of brokerage house), and incentive of the analysts (captured by the
covering history with the firm).
5 Analysts’ Reputation Loss after Activist Short-Selling
5.1 Main Results
Activist short-sellers’ opportunities emerge when the sell-side analysts are over-optimistic
and ignore some material negative information. After the short-theses are published – no matter
whether short-sellers explicitly attack analysts for their lack of diligence and intelligence – it is
natural to expect that market would lose some confidence on analysts’ ability and/or incentives in
providing informative forecasts. This section examines whether and how activist short-selling
affects analysts’ reputation. Specifically, I compare how differently the market reacts to analysts’
EPS revisions before and after activist short-selling. I estimate the following model:
, 0 1 , 2 , 3 , ,
4 , 5 , 6 , ,
i t i t i t i t i t
i t i t i t t t i t
CAR RevRank Post RevRank Post
Size Leverage BTM Year Firm
(2)
Where, CAR is the cumulative abnormal return adjusted by Fama-French three factor
model. In the main analyses I use four short windows: CAR(0, 1), CAR(-1, 1), CAR(0, 2), and
6 This is consistent with Hong and Kacperczyk (2010) that more competition from other analysts leads to less biased
forecasts.
97
CAR(-2, 2) are CAR in a two-day window starting from the EPS revision date, CAR in a three-day
window centered on the EPS revision date, CAR in a three-day window starting from the EPS
revision date, and CAR in a five-day window centered on the EPS revision date, respectively.
RevRank is the decile rank of the peer firm’s analyst forecast revision magnitude. The magnitude
is the revised EPS forecast minus the previous EPS forecast by the same analyst for the same fiscal
quarter scaled by the beginning-of-quarter share price ((FEPSrevised – FEPSprevious) / PRICEbeginning-
of-quarter). Then I transform this variable into decile rank and scale the rank to make it range from 0
to 1. Post is an indicator, defined as one if the EPS revision happens after the activist short-selling,
and zero otherwise. I include several basic firm-level characteristics as control variables, including
Size, Leverage, and BTM. Consistent with model 1, I also include firm and year fixed effects and
cluster standard errors at the analyst level.
Table 2, Panel A presents the summary statistics of the main variables. Panel B shows the
regression results. Note that RevRank ranges from 0 to 1, thus it is easy to interpret that the
coefficient of RevRank represents the difference in market reactions to the analyst revisions in the
top and bottom deciles. For example, in Column 1, it means that, in the Pre-activist short-selling
period, the CAR(0,1) is 4.38% higher for the top-decile revisions than for the bottom-decile
revisions. I label this 4.38% as the spread of market reactions between the top and bottom deciles.
The key coefficient of interest is the interaction term of RevRank×Post, which captures how much
the market reaction is less sensitive to the analyst revisions after activist short-selling. Again, in
Column 1, it means that in the post period, the spread of market reactions narrows by 1.14%, or
26.0% of 4.38%. In other words, after activist short-selling, the market is much less sensitive to
analyst revisions – a symptom that the investors do not believe analysts’ credibility as much as
they do in the pre-period. I label this reduction of sensitivity as reputation loss and quantify it as
the percentage of spread decrease – 26.0% in Column 1. Interestingly, the reputation loss is highly
consistent cross four columns based on different windows. The value ranges from 26% to 30%,
indicating that activist short-selling leads to substantial reputation loss of analysts. In the next two
sub-sections, I examine circumstances in which analysts suffer from less reputation loss.
98
5.2 Cross-Sectional Variation: Analysts’ View Prior to Activist Short-
Selling
First, if the analysts are not as bullish as other analysts prior to the activist short-selling,
the short-theses and the associated price decline would echo their bearish view. In that case, those
analysts’ reputation would be less affected. For brevity I focus on CAR(-1, 1) in all cross-sectional
analyses. In Columns 1 - 2 of Table 3, Panel A, I spilt the sample into More vs. Less Bullish based
on whether an analyst’s last target-price forecast prior to the activist short-selling is higher than
the median of all analysts’ last forecasts of that firm. As expected, the coefficient of RevRank×Post
in the More Bullish sample is significantly negative at the 5% level, while the counterpart in the
Less Bullish sample is insignificant at the conventional level.
5.3 Cross-Sectional Variation: The Market’s Initial Reactions
Second, if the market does not believe the short-theses, then analysts’ reputation would not
be affected. In Columns 3 - 4 of Table 3, Panel A, I spilt the sample into Shorts Seem Wrong vs.
Right based on whether the target firms’ raw return in the first two-day window starting from the
activist short-selling date is negative or not. As expected, the coefficient of RevRank×Post in the
Shorts Seem Right sample is significantly negative at the 1% level, while the counterpart in the
Shorts Seem Wrong sample is insignificant at the conventional level. It is worth mentioning that
these two coefficients are significantly different at the 5% level.
I also combine the above two dimensions (More vs. Less Bullish and Shorts Seem Wrong
vs. Right) in Columns 5 – 8 in Table 3, Panel A. It seems that the key determinant of analysts’
reputation loss is whether short-sellers cause negative returns. Column 8 shows that analysts will
still suffer from reputation loss even if they are less bullish than the median as long as the initial
market reaction suggests that short-sellers might be right (0.0201/0.0535 = 37.6%). But
unsurprisingly, Column 6 indicates that the reputation loss would be more severe for those bullish
analysts (0.0276/0.0527 = 52.4% > 37.6%). By contrast, Columns 5 and 7 show that there is no
reputation loss as long as the market thinks the shorts are wrong, regardless how bullish the
analysts are. These results provide one explanation why analysts have incentives to defend their
previous positions and try to prevent price from collapsing after activist short-selling.
99
5.4 Cross-Sectional Variation: What Analysts Can Do to Avoid
Reputation Loss
Next, I consider what analysts can do after activist short-selling to save their reputation.
Echoing the two dimensions in Section 4, I examine whether the timeliness and direction of
analysts’ reaction affects the reputation loss after activist short-selling. In Panel B of Table 3, I
split the sample into the Bottom Quartile Delay, Middle Delay, and Top Quartile Delay, based on
the first and third quartile of Delay (i.e., the time between activist short-selling and an analyst’s
first target-price revision). I find that the reputation loss is monotonically increasing with the Delay:
for those cases in which analysts react very quickly, the reputation loss is not significant (i.e.,
Column 1), but if analysts don’t provide timely responses, the reputation loss is much worse (i.e.,
Column 3). These results suggest that it is wise for analysts to provide timely responses after
activist short-selling.
But does the direction of reactions matter? This is not as clear as the other cross-sectional
predictions. On one hand, revising down could be interpreted as open-mind; on the other hand, it
could be interpreted as a lack of ability and confidence in their own analyses. In Panel C of Table
3, I split the sample based on whether the analyst’s first target-price revision in the post-period is
to Lower vs. Raise target prices. It seems that the direction does not make huge difference: the
coefficient of RevRank×Post is significantly negative in both two subsamples. Note that those
analysts who raise target prices are more reputable in the Pre-period. So the percentage reputation
loss is even larger for those analysts who lower target prices after activist short-selling (i.e.,
0.202/0.421 = 46.9% > 0.215/0.637 = 33.8%).
6 The Impact of Activist Short-Selling on Analysts’
Careers
To the extent that activist short-selling affects analysts’ reputation, we should expect it also
has impacts on analysts’ career prospects. Following prior literature on analysts career concerns
(e.g., Hong and Kubik 2003), I focus on unfavorable future job separations – whether they have to
100
move to smaller brokerage houses. In particular, I examine this career consequence one, two, and
three years after the activist short-selling year using the following model based on a panel at the
analyst-year level.
, 0 1 , 2 ,
2 , 3 , ,
a t a t a t
a t a t t a t
MoveDown AffectedAnalyst BrokSize
LnCovFirm AnaExperience Year
(3)
Move Down is an indicator equal to one if an analyst is employed by a brokerage house
that has lower percentile ranking than the one she was at the activist short-selling date, and zero
otherwise. I use percentile ranking to (1) remove the growth of the analyst profession and (2)
increase the power of test by generating more variation in the outcome variable than focusing on
the move between top and non-top employers (Hong and Kubik 2003). The variable of interest is
AffectedAnalyst, an indictor equal one if the analyst has at least one covered firm being targeted
by activist short-sellers in that year, and zero otherwise. I also control for several analyst-level
characteristics, including BrokSize, LnCovFirm, and AnaExperience. Finally, I also include year
fixed effect and cluster standard errors at the analyst level.7 I also only focus on those affected
analysts and examine whether the number of covered firms being targeted (NumCovTargets)
matters.
Table 4, Panel A reports the summary statistics of variables used in this section. We can
find that 4.6%, 9.6%, and 13.4% analysts move to smaller brokerage houses in one, two, and three
years, respectively. In the universe of analysts, on average about one third (i.e., 34.6%) have at
least one covered firms being targeted by activist short-sellers. Among those affected analysts, on
average each has two covered firms being targeted.
Table 4, Panel B provides regression results. The first three columns find that analysts
whose covered firms are targeted by activist short-sellers are more likely to move to smaller
brokerage houses after one, two, and three years. The impact is also economically meaningful. For
example, the unconditional mean of Move Down in two years is 9.6%, but having covered firms
7 Note that the dataset is a panel at the analyst-year level. One firm could be covered by many analysts and one analyst
could cover multiple firms. As a result, it is not feasible to control for firm fixed effects in this regression.
101
being targeted by activist short-sellers would increase the probability by 1.34% (Column 2), which
is 14% of the unconditional probability of 9.6%. Further, there is cross-sectional variation within
affected analysts – some analysts are so unlucky that they have more than one firms are targeted.
Columns 4-6 show that the likelihood of moving down is increasing with the number of covered
firms being targeted. For example, having one more covered firm being targeted by activist short-
sellers would increase the probability of moving down in two years by 0.84% (i.e., 8.8% of the
unconditional probability of 9.6%). These results suggest that activist short-selling causes real
damages to affected analysts’ job prospects.
I also investigate whether analysts affected by activist short-selling are more or less likely
to exit. As the prior literature is unclear whether leaving the profession is a favorable outcome for
sell-side analysts, I caution the reader on over-interpreting results of this test. Importantly, the
traditional wisdom assumes that losers leave the profession. For example, Hong, Kubik, and
Solomon (2000) write that “The probability that an analyst may have left for a better job such as
mutual fund manager after leaving the I/B/E/S sample is remote.” However, a recent paper by
Clement and Law (2016) show that high-ability analysts hired during a tight labor market are more
likely to leave the profession than other analysts do after the market condition improves. In other
words, their findings suggest that high-ability analysts have more/better outside job opportunities.
As a result, winners – rather than losers – may be more likely to leave the profession. My results
are consistent with Clement and Law’s view. I find that analysts who have covered firms attacked
by activist short-sellers are less likely to leave the profession in one, two, and three years. Among
those affected analysts, analysts with more covered firms targeted are less likely to leave in one
and two years (but not in three years) than analysts with fewer covered firms targeted. With
aforementioned caveats, I interpret these results as evidence that activist short-selling limits
affected analysts’ outside job opportunities.
102
7 Robustness Checks and Supplemental Analyses
7.1 An Omitted-Variable Problem
One concern is that analysts could be reacting to the price movement rather than to the
activist short-selling. Note that to the extent the price movement is caused by the short-seller’s
public talking-down behavior, the analysts still react to activist short-selling – indirectly. If so the
relevant alternative explanation is whether analysts are reacting to price movement that is
unrelated to the activist short-selling. However, this is unlikely because Zhao (2017) shows that
the daily abnormal returns in several days prior to activist short-selling is not significantly negative
while the abnormal return on the event day is highly significant, indicating the arrival of negative
news prior to the activist short-selling is minimum. Further, I control for several alternative news
sources such as the occurrences of earnings announcements or conference calls in the five days
prior to the activist short-selling and all inferences remain. Finally, I find that analysts are less
likely to raise price targets after attacked by more influential short-sellers (i.e., ASR) than by less
influential short-sellers (i.e., SA), after controlling for the initial return of the activist short-selling.
To the extent that the arrival of firm-level shocks is unrelated to the identity of short-sellers, this
test provides further support that analysts are reacting to activist short-selling.
7.2 Analysts React by Revising EPS Forecasts
In Section 3 and 4 I check analysts’ target-price revisions and conclude that they react to
activist short-selling. This section provides further evidence by examining analysts’ EPS forecast
revisions around activist short-selling. Figure 4 plots the number of analysts who lower and raise
quarterly EPS forecasts for activist short-selling cases that have negative CAR(0, 1) in Panel A
and positive CAR (0, 1) in Panel B. The trend is largely consistent with that in Figure 3. In Table
5 I also re-examine the determinants of cross-sectional variation in analysts’ reactions using
quarterly EPS forecasts. The results are largely consistent with those in Panel B of Table 1. The
only exception is that the number of analyst coverage (LnCoverage) makes analysts more likely
to revise up EPS forecasts, but less likely to revise up price targets.
103
7.3 Do Investors Become Less Responsive in General?
In this paper, I interpret the finding that investors become less responsive to analyst
revisions as evidence of analysts’ reputation loss after activist short-selling. However, one
alternative explanation is that activist short-selling adds noises to all information signals, including
analyst revisions. As a result, investors would be less responsive to all information signals – not
because they believe analysts less. One cross-sectional test shows that this is unlikely the case.
Specifically, if investors become less responsive to all information signals, why the reduction in
responsiveness is stronger for analysts who were more bullish on the targeted stock prior to the
activist short-selling?
Further, if this alternative explanation is justified, we should expect that investors are also
less responsive to other major information events, such as earnings announcements. To address
this possibility, I examine whether investors become less responsive to quarterly earnings
announcements (QEAs) after activist short-selling. Using all QEAs announced between 180 days
prior to activist short-selling date and 180 days after the date, I estimate the following model:
, 0 1 , 2 , 3 , , 4 ,
5 , , 6 , ,
_ i t i t i t i t i t i t
i t i t i t t t i t
CAR QEA UnexRank Post UnexRank Post Loss
Loss Post PreQEAReturn Year Firm
(4)
Where, CAR_QEA is the three-day cumulative abnormal return centered on the quarterly
earnings announcement date (Fama-French three-factor model adjusted); UnexRank is the decile
rank (transformed to range from zero to one) of unexpected earnings, which is the difference
between EPS and the latest analyst consensus; Post is an indicator whether the QEA is prior to the
activist short-selling date; Loss is an indictor whether the current quarterly EPS is negative or not;
PreQERReturn is the raw return of the stock from the consensus date to the earnings announcement
date. Table 6 presents the results. I include neither year nor firm FE in Column 1, only year FE in
Column 2, and both FEs in Column 3. Across three columns, the coefficient of interest – the
interaction of UnexRank and Post – is insignificantly positive, indicating that investors are not less
responsive to earnings announcements after activist short-selling. This serves as a pseudo test to
rule out the alternative explanation that investors react less to all information signals.
104
8 Conclusion
The prior literature has well documented the optimism from sell-side analysts. As the
business model of this key information intermediary does not incentivize publishing negative
opinion, other market participants, particularly short-sellers, seize the opportunity and meet the
demand of negative information. Different from passive short-selling that is done quietly and
reveals information through the trade (i.e., taking short-positions), activist short-sellers publicly
talk-down stocks and sometimes explicitly accuse of the sell-side’s “talk-up” research as the major
driver of equity overvaluation. Using a large sample of activist short-selling cases from 2006 to
2015, I investigate the consequences of activist short-selling on the sell-side community.
I find that analysts react to activist short-selling by revising their target-price and EPS
forecasts on the target firms. Further, the variation in the timeliness and directions of their
responses can be explained by the initial market reaction to the activist short-selling, and the ability
and incentive of analysts. Consistent with the notion that analysts have a strong motivation to react
to short-selling allegations, I find that they on average suffer from serious reputation loss such that
the market are about 30% less sensitive to their earnings revisions. Further, I find that the
reputation loss is weaker or avoided if the analyst is not-so-bullish prior to the activist short-selling,
or the market reaction suggests that the short-seller is wrong. Also, if the reputation loss could be
largely avoided if analysts react very quickly, but either revising down or revising up does not
make a big difference in saving reputation loss. Finally, I find that analysts whose covered firms
are targeted by activist short-sellers are more likely to move to smaller brokerage houses in the
next three years.
In addition to the contribution to the literature on analysts and short-sellers, this paper can
inform the policy debate on the benefits of activist short-selling. The results from this paper show
that encouraging activist short-selling may help to discipline the sell-side community, and as a
result contributing to a more efficient capital market.
105
References of Chapter 2 Aitken, M. J., Frino, A., McCorry, M. S., & Swan, P. L. (1998). Short sales are almost
instantaneously bad news: Evidence from the Australian Stock Exchange. The Journal of
Finance, 53(6), 2205-2223.
Becker, B., & Milbourn, T. (2011). How did increased competition affect credit ratings? Journal
of Financial Economics, 101(3), 493-514.
Beyer, A., & Guttman, I. (2011). The effect of trading volume on analysts’ forecast bias. The
Accounting Review, 86(2), 451-481.
Boehmer, E., Jones, C. M., & Zhang, X. (2015). What do short sellers know? Available at SSRN
2192958.
Brown, L. D., Call, A. C., Clement, M. B., & Sharp, N. Y. (2015). Inside the “Black Box” of sell‐side financial analysts. Journal of Accounting Research, 53(1), 1-47.
Brown, P., Foster, G., & Noreen, E. (1985). Security analyst multi-year earnings forecasts and the
capital market (No. 21). American Accounting Association.
Chen, S., & Matsumoto, D. A. (2006). Favorable versus unfavorable recommendations: The
impact on analyst access to management‐provided information. Journal of Accounting
Research, 44(4), 657-689.
Christophe, S. E., Ferri, M. G., & Hsieh, J. (2010). Informed trading before analyst downgrades:
Evidence from short sellers. Journal of Financial Economics, 95(1), 85-106.
Clement, M. B., & Law, K. (2016). Labor market dynamics and analyst ability. Working paper.
University of Texas at Austin.
De Long, J. B., Shleifer, A., Summers, L. H., & Waldmann, R. J. (1990). Noise trader risk in
financial markets. Journal of Political Economy, 98(4), 703-738.
Dechow, P. M., Hutton, A. P., Meulbroek, L., & Sloan, R. G. (2001). Short-sellers, fundamental
analysis, and stock returns. Journal of Financial Economics, 61(1), 77-106.
Doherty, N. A., Kartasheva, A. V., & Phillips, R. D. (2012). Information effect of entry into credit
ratings market: The case of insurers' ratings. Journal of Financial Economics, 106(2), 308-
330.
Drake, M. S., Rees, L., & Swanson, E. P. (2011). Should investors follow the prophets or the bears?
Evidence on the use of public information by analysts and short sellers. The Accounting
Review, 86(1), 101-130.
Francis, J., & Philbrick, D. (1993). Analysts' decisions as products of a multi-task
environment. Journal of Accounting Research, 216-230.
Hayes, R. M. (1998). The impact of trading commission incentives on analysts' stock coverage
decisions and earnings forecasts. Journal of Accounting Research, 36(2), 299-320.
Hong, H., & Kacperczyk, M. (2010). Competition and bias. The Quarterly Journal of
Economics, 125(4), 1683-1725.
Hong, H., & Kubik, J. D. (2003). Analyzing the analysts: Career concerns and biased earnings
forecasts. The Journal of Finance, 58(1), 313-351.
Hong, H., Kubik, J. D., & Solomon, A. (2000). Security analysts' career concerns and herding of
earnings forecasts. The Rand journal of economics, 121-144.
Jackson, A. R. (2005). Trade generation, reputation, and sell‐side analysts. The Journal of
Finance, 60(2), 673-717.
Jame, R., Markov, S., & Wolfe, M. C. (2017). Does crowdsourced research discipline sell-side
analysts? Working paper. University of Kentucky.
106
Jegadeesh, N., Kim, J., Krische, S. D., & Lee, C. (2004). Analyzing the analysts: When do
recommendations add value? The Journal of Finance, 59(3), 1083-1124.
Karpoff, J. M., & Lou, X. (2010). Short sellers and financial misconduct. The Journal of
Finance, 65(5), 1879-1913.
Ke, B., & Yu, Y. (2006). The effect of issuing biased earnings forecasts on analysts' access to
management and survival. Journal of Accounting Research, 44(5), 965-999.
Kothari, S. P., Shu, S., & Wysocki, P. D. (2009). Do managers withhold bad news? Journal of
Accounting Research, 47(1), 241-276.
Lee, L. F., & Lo, A. K. (2016). Do opinions on financial misstatement firms affect analysts’
reputation with investors? Evidence from reputational spillovers. Journal of Accounting
Research, 54(4), 1111-1148.
Lin, H. W., & McNichols, M. F. (1998). Underwriting relationships, analysts' earnings forecasts
and investment recommendations. Journal of Accounting and Economics, 25(1), 101-127.
Ljungqvist, A., & Qian, W. (2016). How constraining are limits to arbitrage? Review of Financial
Studies, 29(8), 1975- 2028.
McNichols, M., & O'Brien, P. C. (1997). Self-selection and analyst coverage. Journal of
Accounting Research, 35, 167-199.
Michaely, R., & Womack, K. L. (1999). Conflict of interest and the credibility of underwriter
analyst recommendations. Review of Financial Studies, 12(4), 653-686.
Merkley, K. J., Michaely, R., & Pacelli, J. M. (2016). Does the scope of sell-side analyst industry
matter? An examination of bias, accuracy and information content of analyst reports. The
Journal of Finance, Forthcoming.
Mola, S., & Guidolin, M. (2009). Affiliated mutual funds and analyst optimism. Journal of
Financial Economics, 93(1), 108-137.
Park, C. W., & Stice, E. K. (2000). Analyst forecasting ability and the stock price reaction to
forecast revisions. Review of Accounting Studies, 5(3), 259-272.
Richardson, S., Teoh, S. H., & Wysocki, P. D. (2004). The walk‐down to beatable analyst forecasts:
The role of equity issuance and insider trading incentives. Contemporary Accounting
Research, 21(4), 885-924.
Soltes, E. (2014). Private interaction between firm management and sell‐side analysts. Journal of
Accounting Research, 52(1), 245-272.
Stickel, S. E. (1992). Reputation and performance among security analysts. The Journal of
Finance, 47(5), 1811-1836.
Xia, H. (2014). Can investor-paid credit rating agencies improve the information quality of issuer-
paid rating agencies? Journal of Financial Economics, 111(2), 450-468.
Zhao, W. (2017). Activist Short-Selling. Working paper. Rotman School of Management.
107
Appendices of Chapter 2
Appendix A: Variable Definitions
Variables Definitions
Delay The log of one plus the number of days from activist short-selling to the
first target price revision
Raise Indicator. One if the first target price revision after activist short-selling is
to raise the target price.
ShortsImpact
The impact of activist short-selling, measured as (-1) * CR(0, 1), where
CR(0, 1) is the raw return of the target firm in the first two-day window
after activist short-selling
BrokSize
The brokerage house size of the analyst, measured as the log of one plus the
number of unique analysts providing EPS forecasts in that year
LnCovDays
The coverage history, measured as the log of one plus the number of days
since the analyst provided the first EPS forecast on the firm
PreTalkUp
Analysts talk-up behavior prior to the activist short-selling, measured as the
decile rank of all analysts’ last target price forecasts on a given firm prior to
activist short-selling. I then transform it into ranging from 0 to 1 such that
the 10% most bullish forecasts are given the value of 1 and the 10% least
bullish forecasts are given the value of 0.
AnaExperience Analyst experience, measured as the log of one plus the number of years the
analyst has been in IBES universe.
LnCovFirm
Number of covered firms, measured by the log of one plus the number of
unique firms the analyst is providing EPS forecasts in the year
LnCoverage Number of analyst covering the firm, measured by the log of one plus the
number of unique analysts who are providing EPS to the firm in the year.
TPFrequency
Target price forecasting frequency, measured as the log of one plus the
number of times the analyst has provided target price forecast on the firm in
the 90 days prior to the activist short-selling
Size
The size of the firm, measured by the log of one plus the total assets at the
beginning of the fiscal year
Leverage
The leverage of the firm, measured by the ratio of total liabilities to total
assets at the beginning of the fiscal year
BTM
The book-to-market ratio, measured by the ratio of book value of equity to
the total market capitalization at the beginning of the fiscal year.
CAR(0, 1) Cumulative abnormal return in a two-day window starting from the EPS
revision date adjusted by Fama-French three-factor model. Factor loadings
are estimated in a 110-day pre-event window ending 30 trading days before
the CAR windows starts (the same procedure for all CARs)
CAR(-1, 1) Cumulative abnormal return in a three-day window centered on the EPS
revision date
CAR(0, 2)
Cumulative abnormal return in a three-day window starting from the EPS
revision date
CAR(-2, 2)
Cumulative abnormal return in a five-day window centered on the EPS
revision date
108
Post
Indicator. One if the EPS revision (or QEA in Supplemental Analyses)
happens after the activist short-selling, and zero otherwise
RevRank
The decile rank of the peer firm’s analyst forecast revision magnitude. The
magnitude is the revised EPS forecast minus the previous EPS forecast by
one analyst for the same fiscal quarter scaled by the beginning-of-quarter
share price ((FEPSrevised – FEPSprevious) / PRICEbeginning-of-quarter). Then I
transform this variable into decile rank and scale the rank to make it range
from 0 to 1.
More Talk-up Analysts whose last target price forecast on a certain target before activist
short-selling is higher than median. Otherwise they are defined as Less
Talk-up
Shorts Seem
Wrong
The first two-day return (i.e., CR(0, 1)) starting on the activist short-selling
date is negative. Otherwise it’s Shorts Seem Right
Bottom Quartile
Delay
I rank the time an analyst takes to respond to an activist short-selling case
attacking her covered stock (i.e., the delay). Those observations with the
bottom quartile of delay are defined as Bottom Quartile Delay. The next
two quartiles are defined as Middle Delay. The top quartile is defined as
Top Quartile Delay.
Raise Target
Price
Analysts who raise target price after activist short-selling. Otherwise it’s
defined as Lower Target Price
Move Down in
1/2/3 years
Indicator. One if an analyst moves down to a lower-ranked brokerage house
one/two/three years after the activist short-selling. I rank the brokerage
houses every year and rank them into percentiles based on their number of
analysts.
Exit in 1/2/3
years
Indicator. One if an analyst produces at least one EPS forecast in year t but
stops to produce EPS forecasts in year t+1/2/3.
AffectedAnalyst
Indicator. One if an analysts has at least one covered firm being targeted by
activist short-sellers in that year, and zero otherwise.
NumCovTarget
Number of covered firms that are targeted by activist short-sellers in that
year.
CAR_QEA
Three-day cumulative abnormal return centered on the quarterly earnings
announcement date (Fama-French three-factor model adjusted).
UnexRank
The decile rank (transformed to range from zero to one) of unexpected
earnings, which is the difference between EPS and the latest analyst
consensus.
Loss Indictor. One if the current quarterly EPS is negative, and zero otherwise.
PreQERReturn
The raw return of the stock from the consensus date to the earnings
announcement date.
109
Main Tables of Chapter 2 Table 1: The Variation in Analysts’ Reactions to Activist Short-Selling
This table examines the determinants of variation in analysts’ reactions after activist short-selling. I focus on the target
price revisions by analysts who provide at least one target price forecasts on the target firm in the 90 days before
activist short-selling. All observations are at the firm-year-analyst level.
Panel A provides summary statistics. Panel B presents regression results based on OLS models. Variables are defined
in Appendix A. t statistics in parentheses are based on standard errors clustered by analyst. * p < 0.1, ** p < 0.05, ***
p < 0.01 (two-sided tests)
Panel A: Summary statistics
Variables Obs. Mean STD 1st Quartile Median 3rd Quartile
Delay 11,227 3.666 1.079 3.091 3.892 4.443
Raise 11,227 0.495 0.500 0.000 0.000 1.000
ShortsImpact 11,227 0.026 0.067 -0.008 0.017 0.048
BrokSize 11,227 3.083 1.668 0.693 3.555 4.554
LnCovDays 11,227 6.636 1.404 5.855 6.821 7.630
PreTalkUp 11,227 0.476 0.359 0.111 0.444 0.778
AnaExperience 11,227 1.743 0.875 0.527 1.946 2.485
LnCovFirm 11,227 2.182 1.080 0.527 2.639 2.996
LnCoverage 11,227 2.023 1.084 0.693 2.303 2.996
TPFrequency 11,227 0.867 0.250 0.693 0.693 1.099
Size 11,227 8.172 1.936 6.693 8.122 9.540
Leverage 11,227 0.534 0.262 0.318 0.535 0.716
BTM 11,227 0.475 1.216 0.139 0.287 0.527
110
Panel B: Determinants of analyst reactions after activist short-selling (based on target prices)
(1) (2)
Delay Raise
ShortsImpact -2.1731*** -1.4285***
(-6.65) (-12.11)
BrokSize -0.0624*** -0.0081
(-5.33) (-1.59)
LnCovDays 0.0893*** 0.0058
(9.04) (1.60)
PreTalkUp 0.0855*** -0.1625***
(3.39) (-15.15)
AnaExperience -0.0829*** -0.0049
(-3.38) (-0.51)
LnCovFirm -0.1212*** -0.0053
(-4.58) (-0.52)
LnCoverage 0.0330*** -0.0180***
(3.07) (-3.99)
TPFrequency -0.0604 0.0497***
(-1.36) (2.83)
Size -0.0155 -0.0962***
(-0.41) (-5.62)
Leverage 0.2313 0.0798
(1.53) (1.16)
BTM 0.0951*** 0.0276***
(4.47) (2.68)
Year FE YES YES
Firm FE YES YES
Constant 4.0003*** 1.2324***
(11.24) (7.83)
Observations 11,227 11,227
Adjusted R2 0.213 0.394
111
Table 2: Selling Analysts’ Reputation Short: Main Results
This table reports whether activist short-selling damages analysts’ reputation. I focus on the market reactions to EPS
revisions made in 180 days before and after activist short-selling.
Panel provides summary statistics. Panel B presents regression results. Variables are defined in Appendix A. t statistics
in parentheses are based on standard errors clustered by analyst. * p < 0.1, ** p < 0.05, *** p < 0.01 (two-sided tests)
Panel A: Summary statistics
Variables Obs. Mean STD 1st Quartile Median 3rd Quartile
CAR(0, 1) 17,741 -0.004 0.123 -0.030 -0.001 0.026
CAR(-1, 1) 17,740 -0.003 0.105 -0.040 -0.002 0.034
CAR(0, 2) 17,740 -0.004 0.132 -0.036 -0.001 0.030
CAR(-2, 2) 17,731 -0.004 0.112 -0.051 -0.002 0.041
Post 17,741 0.387 0.487 0.000 0.000 1.000
RevRank 17,741 0.500 0.319 0.222 0.556 0.778
Size 17,741 8.561 1.897 7.184 8.661 9.966
Lev 17,741 0.555 0.258 0.350 0.557 0.724
BTM 17,741 0.480 1.117 0.166 0.322 0.613
Panel B: Selling analysts’ reputation short: regression results
(1) (2) (3) (4)
CAR(0,1) CAR(-1, 1) CAR(0, 2) CAR(-2, 2)
Post 0.0059* 0.0037 0.0055 0.0029
(1.94) (1.13) (1.63) (0.76)
RevRank 0.0438*** 0.0505*** 0.0433*** 0.0519***
(11.10) (13.44) (10.01) (12.52)
RevRank*Post -0.0114** -0.0150*** -0.0128** -0.0157**
(-2.22) (-2.83) (-2.19) (-2.50)
Size -0.0150*** -0.0186*** -0.0157*** -0.0154***
(-4.28) (-3.92) (-3.90) (-2.84)
Leverage -0.0755*** -0.0063 -0.0776*** 0.0090
(-4.02) (-0.44) (-3.88) (0.63)
BTM 0.0007 0.0011 -0.0003 0.0008
(0.51) (1.02) (-0.28) (0.60)
Year FE YES YES YES YES
Firm FE YES YES YES YES
Constant 0.1101*** 0.1071*** 0.1215*** 0.0823*
(3.89) (2.68) (3.68) (1.77)
Observations 17,741 17,740 17,740 17,731
Adjusted R2 0.193 0.208 0.188 0.225
112
Table 3: Selling Analysts’ Reputation Short: Cross-Sectional Variation
This table explores the cross-sectional variation in analysts’ reputation loss after activist short-selling. Panel A focuses on how much analysts talk-up the stock prior to the activist short-selling, and whether market’s initial reaction suggests activist short-sellers are right. Main partition variables are defined in
Appendix A. t statistics in parentheses are based on standard errors clustered by analyst. * p < 0.1, ** p < 0.05, *** p < 0.01 (two-sided tests)
Panel A: Previous talk-up and whether shorts are right
(1) (2) (3) (4) (5) (6) (7) (8)
DV=CAR(-1, 1) More
Bullish
Less
Bullish
Shorts
Seem
Wrong
Shorts
Seem
Right
More
Bullish &
Shorts
Seem
Wrong
More
Bullish &
Shorts
Seem
Right
Less
Bullish &
Shorts
Seem
Wrong
Less
Bullish &
Shorts
Seem
Right
Post 0.0060 0.0029 0.0012 0.0058 0.0050 0.0078 -0.0013 0.0063
(1.22) (0.59) (0.24) (1.31) (0.75) (1.17) (-0.17) (0.98)
RevRank 0.0531*** 0.0511*** 0.0435*** 0.0512*** 0.0455*** 0.0527*** 0.0453*** 0.0535***
(9.89) (8.90) (8.62) (9.62) (5.83) (7.66) (6.45) (6.19)
RevRank*Post -0.0192** -0.0124 -0.0020 -0.0229*** -0.0086 -0.0276** 0.0001 -0.0201**
(-2.31) (-1.63) (-0.26) (-3.16) (-0.80) (-2.35) (0.01) (-2.03)
Size -0.0241*** -0.0123*** 0.0019 -0.0274*** 0.0026 -0.0338*** 0.0033 -0.0207***
(-3.14) (-2.62) (0.23) (-5.96) (0.22) (-5.13) (0.25) (-3.10)
Leverage -0.0166 0.0034 -0.0498** 0.0058 -0.0566* -0.0084 -0.0452 0.0241
(-0.77) (0.18) (-2.19) (0.24) (-1.89) (-0.31) (-1.36) (0.64)
BTM 0.0020 0.0002 -0.0027 0.0023* -0.0036 0.0035** -0.0014 0.0007
(1.27) (0.12) (-1.36) (1.69) (-1.07) (2.09) (-0.60) (0.43)
Year FE YES YES YES YES YES YES YES YES
Firm FE YES YES YES YES YES YES YES YES
Constant 0.1574** 0.0501 0.0273 0.0735* 0.0393 0.1139* -0.0042 0.0210
(2.42) (1.18) (0.36) (1.74) (0.35) (1.94) (-0.04) (0.33)
Observations 8,542 9,198 6,241 11,499 2,994 5,548 3,247 5,951
Adjusted R2 0.250 0.175 0.278 0.210 0.285 0.268 0.293 0.166
113
Panel B: The timeliness of analysts’ reactions
(1) (2) (3)
DV = CAR(-1, 1) Bottom Quartile
Delay
Middle
Delay
Top Quartile Delay
Post 0.0039 0.0076 0.0089
(0.64) (1.28) (1.30)
RevRank 0.0553*** 0.0545*** 0.0526***
(6.33) (9.33) (6.82)
RevRank*Post -0.0164 -0.0195** -0.0290***
(-1.50) (-2.12) (-2.66)
Size -0.0399*** -0.0194*** -0.0215***
(-3.15) (-3.37) (-2.78)
Leverage -0.0297 -0.0011 0.0485*
(-0.69) (-0.06) (1.79)
BTM -0.0035 -0.0016 0.0089**
(-1.19) (-0.75) (2.13)
Year FE YES YES YES
Firm FE YES YES YES
Constant 0.2586** 0.1179** 0.1054
(2.43) (2.23) (1.61)
Observations 4,157 8,175 4,074
Adjusted R2 0.181 0.195 0.239
114
Panel C: The directions of analysts’ reactions
(1) (2) (3) (4) (5) (6)
DV = CAR(-1, 1) Lower TP Raise TP Shorts
Seem Right
& Lower TP
Shorts
Seem Right
&Raise TP
Shorts
Seem Wrong
& Lower TP
Shorts
Seem Wrong
& Raise TP
Post 0.0046 0.0089 0.0081 0.0093 -0.0017 0.0097
(1.05) (1.60) (1.40) (1.15) (-0.24) (1.37)
RevRank 0.0431*** 0.0637*** 0.0465*** 0.0652*** 0.0338*** 0.0587***
(8.96) (9.75) (6.95) (6.61) (4.78) (7.33)
RevRank*Post -0.0202** -0.0215*** -0.0294*** -0.0267** -0.0002 -0.0180*
(-2.51) (-2.71) (-2.70) (-2.33) (-0.02) (-1.68)
Size -0.0075 -0.0197*** -0.0183** -0.0310*** 0.0086 -0.0033
(-1.30) (-3.22) (-2.47) (-3.63) (0.68) (-0.27)
Leverage -0.0051 0.0196 -0.0181 0.1323*** -0.0185 -0.0697
(-0.27) (0.82) (-0.55) (2.63) (-0.44) (-1.38)
BTM -0.0042** 0.0098** -0.0036 0.0127* -0.0066*** 0.0082
(-2.13) (2.23) (-0.58) (1.79) (-2.96) (0.54)
Year FE YES YES YES YES YES YES
Firm FE YES YES YES YES YES YES
Constant 0.0079 0.1144** 0.0435 0.0033 -0.0538 0.0707
(0.15) (2.12) (0.52) (0.04) (-0.45) (0.61)
Observations 8,957 7,449 6,129 4,475 2,828 2,974
Adjusted R2 0.222 0.173 0.241 0.144 0.181 0.337
115
Table 4: Analysts’ Careers after Activist Short-Selling
This table examines whether and how activist short-selling affects analysts’ future career path. I focus on whether an
analyst moves down to a smaller brokerage house one, two, and three years after activist short-selling. Panel A
provides summary statistics. Panel B presents regression results. All variables are defined in Appendix A. t statistics
in parentheses are based on standard errors clustered by analyst. * p < 0.1, ** p < 0.05, *** p < 0.01 (two-sided tests)
Panel A: Summary statistics
Variables Obs. Mean STD 1st Quartile Median 3rd Quartile
Move Down in 1 year 26,552 0.046 0.210 0.000 0.000 0.000
Move Down in 2 years 19,927 0.096 0.295 0.000 0.000 0.000
Move Down in 3 years 15,400 0.134 0.341 0.000 0.000 0.000
AffectedAnalyst 26,552 0.346 0.476 0.000 0.000 1.000
NumCoverTargets 9,186 2.065 1.632 1.000 1.000 3.000
BrokSize 26,552 3.722 1.078 2.996 3.761 4.673
LnCovFirm 26,552 2.421 0.714 2.079 2.639 2.944
AnaExperience 26,552 1.884 0.691 1.386 1.946 2.398
Exit in 1 year 32,910 0.193 0.395 0.000 0.000 0.000
Exit in 2 years 29,867 0.318 0.466 0.000 0.000 1.000
Exit in 3 years 26,625 0.402 0.490 0.000 0.000 1.000
116
Panel B: Analysts’ careers after activist short-selling: moving down
(1) (2) (3) (4) (5) (6)
All Analysts Only Affected Analysts
DV = Move Down In 1 year In 2 year In 3 year In 1 year In 2 year In 3 year
AffectedAnalyst 0.0067** 0.0134** 0.0165**
(2.09) (2.36) (2.18)
BrokSize -0.0263*** -0.0474*** -0.0732*** -0.0294*** -0.0522*** -0.0779***
(-19.07) (-17.63) (-18.61) (-11.70) (-10.92) (-11.36)
LnCovFirm 0.0107*** 0.0152*** 0.0228*** 0.0008 -0.0068 -0.0075
(5.37) (3.76) (3.95) (0.18) (-0.72) (-0.56)
AnaExperience -0.0039* -0.0064 -0.0205*** -0.0007 -0.0067 -0.0241**
(-1.78) (-1.49) (-3.25) (-0.17) (-0.89) (-2.12)
NumCovTargets 0.0046*** 0.0084*** 0.0154***
(2.92) (2.99) (3.06)
Year FE YES YES YES YES YES YES
Constant 0.1278*** 0.2403*** 0.3849*** 0.1487*** 0.3139*** 0.4962***
(15.61) (16.03) (17.85) (8.27) (9.68) (10.60)
Observations 26,552 20,367 15,934 9,186 6,691 4,696
Adjusted R2 0.020 0.031 0.052 0.019 0.030 0.052
117
Panel C: Analysts’ career after activist short-selling: exit
(1) (2) (3) (4) (5) (6)
All Analysts Only Affected Analysts
DV = Exit In 1 year In 2 years In 3 years In 1 year In 2 years In 3 years
AffectedAnalyst -0.0311*** -0.0213*** -0.0138*
(-6.55) (-3.18) (-1.70)
BrokSize -0.0118*** -0.0173*** -0.0197*** -0.0055* -0.0109** -0.0240***
(-5.85) (-5.68) (-5.21) (-1.73) (-2.16) (-3.53)
LnCovFirm -0.1371*** -0.1576*** -0.1602*** -0.1190*** -0.1504*** -0.1741***
(-35.52) (-29.24) (-24.88) (-15.43) (-13.84) (-12.84)
AnaExperience -0.0055 -0.0440*** -0.0597*** -0.0273*** -0.0675*** -0.0853***
(-1.43) (-7.63) (-8.27) (-4.75) (-7.42) (-6.86)
NumCovTargets -0.0048*** -0.0056* -0.0014
(-2.79) (-1.88) (-0.29)
Year FE YES YES YES YES YES YES
Constant 0.5701*** 0.8403*** 0.9858*** 0.5438*** 0.8765*** 1.1073***
(44.91) (49.30) (49.88) (19.97) (24.13) (25.53)
Observations 32,910 29,867 26,625 10,488 8,720 6,813
Adjusted R2 0.085 0.098 0.097 0.062 0.078 0.084
Table 5: Cross-Sectional Variation in Analyst Reactions after Activist Short-Selling (Based
on Quarterly EPS Forecasts)
This table examines the determinants of variation in analysts’ reactions after activist short-selling using quarterly EPS
forecasts. I focus on the EPS revisions by analysts who provide at least one EPS forecast on the target firm in the 90
days before activist short-selling. All observations are at the firm-year-analyst level. Variables are defined in Appendix
A. t statistics in parentheses are based on standard errors clustered by analyst. * p < 0.1, ** p < 0.05, *** p < 0.01
(two-sided tests)
(1) (2)
Delay Raise
ShortsImpact -0.0047 -0.6137***
(-0.01) (-4.04)
BrokSize -0.0085 -0.0087*
(-0.73) (-1.85)
LnCovDays 0.0887*** -0.0042
(7.16) (-0.77)
PreTalkUp -0.0226 -0.1366***
(-0.84) (-10.90)
AnaExperience -0.0189 -0.0055
(-0.74) (-0.49)
LnCovFirm -0.1372*** -0.0108
(-5.22) (-0.91)
LnCoverage 0.3296*** 0.0588**
(5.23) (2.04)
TPFrequency -0.0663 0.0212
(-1.48) (1.05)
Size -0.1345*** -0.1446***
(-3.02) (-6.89)
Leverage 0.4881*** 0.1207
(2.87) (1.59)
BTM 0.0514*** 0.0224***
(3.83) (3.19)
Year FE YES YES
Firm FE YES YES
Constant 3.6033*** 1.5149***
(8.86) (7.95)
Observations 10,195 10,195
Adjusted R2 0.279 0.225
119
Table 6: Market Reactions to Quarterly Earnings Announcements
This table examines the whether investors become less responsive to quarterly earnings announcements after activist
short-selling. All observations are at the firm-QEA (quarterly earnings announcement) level. Variables are defined in
Appendix A. t statistics in parentheses are based on standard errors clustered by firm. * p < 0.1, ** p < 0.05, *** p <
0.01 (two-sided tests)
(1) (2) (3)
CAR_QEA CAR_QEA CAR_QEA
UnexRank 0.0775*** 0.0774*** 0.1180***
(9.79) (9.80) (10.65)
Post -0.0028 -0.0028 -0.0064
(-0.70) (-0.72) (-1.43)
UnexRank × Post 0.0005 0.0006 0.0064
(0.07) (0.09) (0.83)
Loss -0.0082 -0.0083 0.0312***
(-1.44) (-1.45) (3.49)
UnexRank × Loss -0.0200 -0.0197 -0.1352***
(-1.14) (-1.12) (-5.19)
PreQEAReturn -0.0349** -0.0366*** -0.0458***
(-2.55) (-2.66) (-2.85)
Year NO YES YES
Firm NO NO YES
Constant -0.0412*** -0.0412*** -0.0672***
(-7.90) (-7.90) (-4.33)
Observations 9,079 9,079 9,079
Adjusted R2 0.082 0.083 0.118
120
Main Figures of Chapter 2 Figure 1: The Year Distribution of the Firm-year Level Activist Short-Selling
This figure presents the year distribution of firm-year observations. Note the sample starts from Zhao’s
(2016) 6,081 activist short-selling cases from 2006 to 2015. Then for each firm-year, I focus on the activist
short-selling case that presents the lowest two-day return starting from the announcement date.
121
Figure 2: Replicating Ljungvist and Qian (2016) Results Regarding Analyst
Recommendations
This figure is to replicate the observation in Ljungqvist and Qian (2016) that the proportion of buy and sell
recommendations are largely stable before and after activist short-selling. This figure presents the number
of five types of recommendations for sample firms in each period (i.e., one period = five days). For example,
there are about 250 Hold recommendations in the first five days after activist short-selling.
122
Figure 3: Number of Analysts Raised or Lowered Target Prices in 18 Five-day Intervals
Before and After the Activist Short-Selling
This figure is to show that analysts react to activist short-selling by revising target prices. In both panels,
the blue solid line represents the number of revisions that lower target prices and the red dash line represents
the number of revisions that raise target prices in each five-day period. Panel A (B) reports the number of
revisions around activist short-selling with negative (positive) first two-day returns.
Panel A: Targeted firms of which shorts seem to be right
Panel B: Targeted firms of which the shorts seem to be wrong
123
Figure 4: Number of Analysts Raised or Lowered Quarterly EPS Forecasts in 18 five-day
Intervals Before and After the Activist Short-Selling
This figure is to show that analysts react to activist short-selling by revising EPS forecasts. In both panels,
the blue solid line represents the number of revisions that lower EPS forecasts and the red dash line
represents the number of revisions that raise EPS forecasts in each five-day period. Panel A (B) reports the
number of revisions around activist short-selling with negative (positive) first two-day returns.
Panel A: Targeted firms of which the shorts seem to be right
Panel B: Targeted firms of which the shorts seem to be wrong