The Sound of Silence: What Do We Know When Insiders Do Not Trade?* George P. Gao and Qingzhong Ma Cornell University This draft: October 24, 2013 * We are grateful to Ferhat Akbas, Warren Bailey, Sanjeev Bhojraj, Walter Boudry, Linda Canina, Matt Clayton, Zhi Da, Frank D'Souza, Julia D. D'Souza, David Feldman, Vyacheslav Fos, Pengjie Gao, Ross Garon, John Griffin, Levon Goukasian, Yaniv Grinstein, Omrane Guedhami, Michael Johnson, G. Andrew Karolyi, Jan Katz, Hyunseob Kim, Heather Knewtson, Paul Koch, Piotr Korczak, Dima Leshchinskii, Claire Liang, Bob Libby, Crocker Liu, Edith Liu, Peng Liu, Xiaomeng Lu, Ken Merkley, Roni Michaely, Pam Moulton, David Ng, Dominik Rösch, Gideon Saar, Rick Sias, Richard Smith, Mike Sturman, Andrey Ukhov, Abhishek Varma, Samuel Weaver, David Whidbee, Jinghua Yan, Eric Yeung, Xiaojing Yuan, Kelvin Chunhui Zhang, Wei Zhang, and seminar participants at Cornell University (Johnson School accounting), Cornell finance brownbag seminars, the 2012 Northern Finance Association conference, SAC Capital quant seminar, the 2013 Midwest Finance Association conference, the 2013 Applied Finance Conference at St. John’s University, the 2013 SFS Finance Cavalcade, and the 2013 Financial Management Association (FMA) annual meeting in Chicago (a top ten session) for helpful suggestions, and Irene Kim and Douglas Skinner for help with constructing the litigation risk data. All errors are our own. Gao: 314 Sage Hall, Samuel Curtis Johnson Graduate School of Management, Cornell University, Ithaca, New York, 14853. Email: [email protected], phone: (607) 255–8729. Ma (corresponding author): 435 B Statler Hall, Cornell University, Ithaca, New York, 14853. Email: [email protected], phone: (607) 255–8140.
50
Embed
The Sound of Silence: What Do We Know When Insiders Do … ANNUAL MEETINGS/2014-… · Levon Goukasian, Yaniv Grinstein, Omrane Guedhami, Michael Johnson, ... Ken Merkley, Roni Michaely,
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
The Sound of Silence:
What Do We Know When Insiders Do Not Trade?*
George P. Gao and Qingzhong Ma
Cornell University
This draft: October 24, 2013
* We are grateful to Ferhat Akbas, Warren Bailey, Sanjeev Bhojraj, Walter Boudry, Linda Canina, Matt Clayton, Zhi
Da, Frank D'Souza, Julia D. D'Souza, David Feldman, Vyacheslav Fos, Pengjie Gao, Ross Garon, John Griffin,
Levon Goukasian, Yaniv Grinstein, Omrane Guedhami, Michael Johnson, G. Andrew Karolyi, Jan Katz, Hyunseob
Kim, Heather Knewtson, Paul Koch, Piotr Korczak, Dima Leshchinskii, Claire Liang, Bob Libby, Crocker Liu,
Edith Liu, Peng Liu, Xiaomeng Lu, Ken Merkley, Roni Michaely, Pam Moulton, David Ng, Dominik Rösch,
Gideon Saar, Rick Sias, Richard Smith, Mike Sturman, Andrey Ukhov, Abhishek Varma, Samuel Weaver, David
Whidbee, Jinghua Yan, Eric Yeung, Xiaojing Yuan, Kelvin Chunhui Zhang, Wei Zhang, and seminar participants at
Cornell University (Johnson School accounting), Cornell finance brownbag seminars, the 2012 Northern Finance
Association conference, SAC Capital quant seminar, the 2013 Midwest Finance Association conference, the 2013
Applied Finance Conference at St. John’s University, the 2013 SFS Finance Cavalcade, and the 2013 Financial
Management Association (FMA) annual meeting in Chicago (a top ten session) for helpful suggestions, and Irene
Kim and Douglas Skinner for help with constructing the litigation risk data. All errors are our own.
Gao: 314 Sage Hall, Samuel Curtis Johnson Graduate School of Management, Cornell University, Ithaca, New
York, 14853. Email: [email protected], phone: (607) 255–8729. Ma (corresponding author): 435 B Statler Hall,
Cornell University, Ithaca, New York, 14853. Email: [email protected], phone: (607) 255–8140.
This paper examines stock returns following insider silence, periods of no insider
trading. We hypothesize that, for fear of litigation risk, rational insiders do not sell
own-company shares when they withhold bad news; neither would they buy,
given unfavorable prospects; thus they keep silent. By contrast, insiders sell
shares when they do not anticipate significant bad news. Consistent with this
hypothesis, future returns are significantly lower following insider silence than
following insider net selling. Further, the silence-sell return difference is wider
among firms with worse information environment and firms at higher litigation
risk. In sum, insider silence sounds bad news.
Keywords: Insider trading, Insider silence, Information environment, Litigation
risk
JEL Classification: G12, G14, G18
1
1. Introduction
It is well documented that managers tend to withhold bad news about their firms.1 This
paper examines managers’ insider trading decisions and subsequent stock returns when they
withhold bad news. We hypothesize that, when managers (and insiders in general) withhold bad
news, their rational decision is not to sell own-company shares on the bad news, as doing so
involves high levels of litigation risk. This is because shareholders launch securities class action
lawsuits following large stock price drops on the basis of 10b-5, mostly alleging that insiders
knew the adverse information but failed to disclose it to the market.2 Insider selling prior to the
disclosure of the information would be taken as evidence that insiders knew the information.3 By
contrast, lack of insider selling undercuts plaintiffs’ allegation that insiders knew the
information. Thus, for fear of shareholder litigation, rational insiders do not sell own-company
shares while possessing significant negative information. Neither would they buy, given the
unfavorable prospects. Hence, insider silence, a period of no insider trading, is associated with
bad news.4 On the other hand, insiders selling shares indicates that they do not expect as bad
news in the future.5 This litigation risk hypothesis thus implies that insider silence is associated
with significant negative future returns, and insider net selling is associated with not-so-negative
1 See, for example, empirical evidence in Kothari, Shu, and Wysocki (2009) and survey evidence in Graham,
Harvey and Rajgopal (2005). Sletten (2012) and Tse and Tucker (2010) provide empirical evidence of clustering in
bad news announcements, which is predicted if firms withhold bad news (Dye and Sridhar, 1995; Acharya,
DeMarzo, and Kremer, 2011); Rogers, Shrand, and Zechman (2013) provide evidence of managers tacitly
withholding industry-wide bad news. Theories on managers withholding bad news include Dye (1985), Jung and
Kwon (1988), Dye and Sridhar (1995), Shin (2003, 2006), and Acharya, DeMarzo, and Kremer (2011). 2 See Arshadi (1998) and Bainbridge (2007) for overviews of insider trading laws and enforcement strategies. For a
recent example of securities class-action lawsuits following large stock price declines, see the case involving
Expedia, http://securities.stanford.edu/1050/EXPE00_01/index.html. 3 See, e.g., Johnson, Nelson, and Pritchard (2007), Rogers (2008), Rogers, Van Buskirk, and Zeckman (2011).
4 One might argue that insider silence could also be related to good news, such as a takeover offer. Such a possibility
adds noise to the relation between insider silence and negative future returns and biases against us in empirical tests.
Section 4.2 examines this possibility in greater detail and shows that, with a six-month window to measure insider
trading activity, takeover cases do not significantly contaminate insider silence as a negative signal. 5 Numerous studies show that insider sales are inferior as a return predictor (Jaffe, 1974; Finnerty, 1976; Seyhun,
1986; Rozeff and Zaman, 1988; Lin and Howe, 1990; Seyhun, 1998; Noe, 1999), or insider sales have no predictive
ability (Chowdhury, Howe, and Lin, 1993; Lakonishok and Lee, 2001; Jeng, Metrick, and Zeckhauser, 2003).
(or non-negative) future returns. That is, the spread in future returns between insider silence and
insider net selling, or the silence-sell spread, is negative.
Further, the hypothesis has two cross-sectional implications. First, when the information
environment of a firm is worse, it is more likely that insiders possess important information that
outsiders do not know. Thus, we expect the silence-sell spreads are wider among firms with
worse information environment. Second, if litigation risk drives insiders’ decision not to sell on
negative information, the silence-sell spreads are wider among firms at higher litigation risk.
We test these implications in a sample of NYSE/ASE/NASDAQ common stocks from
1990 to 2011. Each month we form portfolios of stocks based on their cumulative insider trading
activity over the trailing six months. Stocks whose insiders do not trade over the past six months
form the “silence” portfolio; stocks that insiders net buy and net sell over the past six months
form the “buy” and “sell” portfolios, respectively. We then examine their future returns.
We report the following findings. First, the silence portfolio on average earns a
significant negative abnormal return (adjusted by size and book-to-market ratio) over the
subsequent 12-month period, which is significantly lower than the “sell” portfolio. The silence-
sell spread is a significant -2.50%. This result is consistent with the litigation risk hypothesis.
Second, the silence-sell spread in future returns varies systematically in the cross-section.
It is wider for growth stocks, stocks of younger firms, and more volatile stocks. These results are
consistent with the hypothesis that this litigation risk effect is stronger among firms with worse
information environment. Using a composite measure of litigation risk developed by Kim and
Skinner (2012), industry membership (following Francis, Philbrick, and Schipper, 1994), and
sales growth as proxies for litigation risk, we find that the silence-sell spread is wider among
firms with higher ex-ante litigation risk, further supporting the litigation risk hypothesis.
3
The main findings survive numerous robustness checks. The significant negative silence-
sell spread survives alternative estimation methodologies (size and B/M adjusted returns, Fama
and French three-factor alphas); the silence-sell spread also arises in subsequent quarterly
earnings announcement period abnormal returns. The silence-sell spread is not explained by the
underperformance of firms that have just conducted initial public offering (IPO); it is not
explained by a broad list of anomaly variables that predict future returns in the cross section; it is
not explained by insider trading activity prior to the six-month measuring period. Furthermore,
while insiders might also abstain from trading prior to releasing extremely positive news (e.g., a
takeover offer), there is no evidence that insiders abstain from trading over the six-month period
prior to takeover announcements. As a result, takeover cases do not materially contaminate
insider silence as a negative signal. Last, the silence-sell spread does not reverse in the long term
and in fact it continues to widen for at least another two years.
The paper contributes to the insider trading literature. The existing literature focuses
exclusively on insiders’ buying and selling activities and no study examines the phenomenon of
the lack of insider trading. In addition, while it is well known that litigation risk reduces insider
trading during information rich events (e.g., Seyhun, 1992; Jagolinzer and Roulstone, 2009;
Piotroski and Roulstone, 2008), there is no systematic study on how the reduction in insider
trading is related to stock returns. Our paper fills this gap. We show that insider silence, periods
of no insider trading, signals bad news as it predicts significant negative future returns, which are
even lower than when insiders net sell. We further show that this effect is systematically related
to litigation risk and information environment. Our findings offer a potential explanation for the
puzzling results documented in the literature that insider sales, on average, are not informative.
In recent work, Cohen, Malloy, and Pomorski (2012) find that some insider trades (the
4
opportunistic ones) are informative for up to six months. Our paper is related to theirs in that we
find another significant information signal from insider trading activity: the lack of it.
This paper is also broadly related to several recent papers on the price implication of lack
of activity. Bagnoli, Kross, and Watts (2002) show that delays in earnings reports precede price
drops. Marin and Olivier (2008) report that lower insider selling volume over the recent past and
heavy insider selling volume in the distant past predicts higher probability of price crash. Akbas
(2013) reports that stocks experiencing unusually low trading volume over a week before
earnings announcements tend to have more unfavorable earnings surprises. Giglio and Shue
(2013) find that the passage of time after merger announcement is related to the probability of
merger completion. Our paper shows that insider silence, the lack of insider trading, sounds bad
news.
We describe the sample and data in Section 2, report main findings in Section 3, discuss
alternative hypotheses and robustness checks in Section 4, and conclude in Section 5.
2. Sample and data
The sample is based on all NYSE/Amex/NASDAQ common stocks (share code 10 or 11)
covered in CRSP/Compustat merged database from January 1990 to December 2011, a total of
264 year/month cross-sections. The sample starts from 1990 when insider trading data is
available and ends in 2011 so that we have at least one-year future return data. We exclude
stocks whose prior month-end price is lower than $5 and stocks that would be classified into the
lowest NYSE market capitalization decile.6 We also exclude firms with missing or non-positive
book value of equity. Firms are required to be at least one year old.7
6 These two restrictions eliminate low-priced and small stocks, which are likely to have low past returns. As a result,
the average trailing 12-month return of the remaining stocks is 25.5% (see Table 1). Without these restrictions the
average would be around 15.5%. Further, since insiders’ buying and selling activities are more informative among
small firms (e.g., Lakonishok and Lee, 2001; Sias and Whidbee, 2010), excluding these firms reduces the return
5
Stock return data are obtained from the Center for Research in Security Prices (CRSP) at
the University of Chicago and accounting data from Compustat. We follow Fama and French
(1992) to construct firm size and B/M ratio, and Jegadeesh and Titman (2001) to calculate past
returns. Insider trading data are from Thomson Reuters Insider Filing Data Feed. The Securities
and Exchange Commission (SEC) mandates that officers and directors, large shareholders (those
who own 10% or more of the outstanding shares), and affiliated shareholders report their
transactions to the SEC by the 10th
of the month following the transactions (prior to August
2002) or within two days (since August 2002). The database cleaning process largely follows
recent studies (e.g., Lakonishok and Lee, 2001; Piotroski and Roulstone, 2005; Sias and
Whidbee, 2010).8 Only open-market transactions are considered. Defined in equation (1), the net
insider demand (NID) for month j is the total number of shares insiders buy minus the total
number of shares insiders sell over the past six months, normalized by the total number of shares
outstanding at the end of month j-1.
(1)
Portfolios are formed on past insider trading activity. Stocks with no insider trading
activity reported over the past six months form the “silence” portfolio; stocks with positive and
non-positive NID form the “buy” and “sell” portfolios, respectively. Future returns start from
predictability following insiders’ buying and selling activities. In our main analysis we choose to exclude these
stocks to focus on economically significant stocks. Our main findings hold as strongly when low-priced and small
stocks are included (see Section 4.5). 7 For firms that just conducted initial public offerings, this filter allows a six-month period for insiders of these firms
to make insider trading decisions if the IPO lockup period is shorter than six months; for IPO firms with a lockup
period longer than 180 days (see Brav and Gompers, 2003), this filter is not sufficient. We address this issue in
robustness section 4.1. 8 We follow the literature (e.g., Lakonishok and Lee, 2001; Sias and Whidbee, 2010) to “clean” the insider trading
data. Specifically, we use the following filters. We delete duplicate and amended records and records with cleanse
code of “S” or “A” are deleted. Transaction price must be available, and we delete records if the number of shares in
a transaction is below 100. The transaction code is either “P” or “S” for stock transactions. We delete transactions
that involve more than 20% of total shares outstanding, and delete records if the transaction price is outside the
80%–120% range of the CRSP end-of-day stock price.
6
month j. It is worth noting that since we use report date to measure past insider trading activity
(see Lakonishok and Lee, 2001, p.97), such information is publicly available by the time the
portfolios are formed.9
Figure 1 presents, month by month from January 1990 to December 2011, the proportion
of firms with insider net selling, net buying, and insider silence over the trailing six-month
period. The proportion of insider silence is around 35% during early years and generally declines
over time, except for a spike around the end of 2010. The average proportion of insider silence is
24%.10
Figure 1 also shows that insider net selling is more frequent than net buying.
[Insert Figure 1 about here]
Table 1 shows summary statistics of the whole sample and portfolios formed on past
insider trading. The sample average NID over the trailing six months is -0.418%, consistent with
insiders on average being net sellers. The average six-month NID is also consistent with the
literature.11
Compared to stocks in the overall sample, stocks in the silence group tend to be
smaller, have higher B/M ratio, have lower institutional ownership, are less likely followed by
analysts, have fewer analysts following, are slightly younger, and are a bit more volatile. There is
no apparent difference in terms of past stock issuance. Stocks in the silence group tend to have
lower past return, lower accruals, lower past asset growth, and are less profitable. The silence
group also has lower values of KS, FPS, and sales growth, proxies we use for litigation risk (see
section 3.2.2).
9 Our main findings are robust to alternative choices in defining net insider demand. Detailed are in Section 4.5.
10 By definition, firms that fall in the “silence” group in a month are more likely to fall in the “silence” group over
the subsequent months. That is, the “silence” portfolio membership is sticky over time. Indeed, with a probability of
48%, a firm in the “silence” group will remain in the “silence” group six months later. This number indicates that on
average over half of the “silence” portfolio members are replaced during a half-year period. 11
The average quarterly NID (unreported) for our sample is -0.246, slightly more negative than the corresponding
number reported in Sias and Whidbee (2010, p.1551), which is -0.145%. Because we exclude the very small stocks,
which tend to have more positive (or less negative) net insider demand, the average NID of our sample is larger (in
magnitude) than that reported in Sias and Whidbee (2010).
7
Note that for all stocks, the average return over the trailing 12 months is 25.5%. This
average return is higher than one would expect, mainly because the sampling procedure
eliminates low-priced and small stocks, which are likely to have low past returns.12
Stocks that
insiders net buy have lower returns than stocks that insiders net sell, consistent with the literature
that the contemporaneous correlation between net insider demand and stock returns is negative
(e.g., Sias and Whidbee, 2010). In addition, stocks in the “buy” portfolio are smaller with higher
B/M ratios, consistent with the notion that insiders of larger firms tend to sell and insiders are
contrarian (Seyhun, 1986; Rozeff and Zaman, 1998; Piotroski and Roulstone, 2005).
[Insert Table 1 about here]
3. Results
3.1. Insider silence and future returns
We take three approaches to examining future returns following insider trading decisions.
First is the cumulative returns over a holding period of one, six, and 12 months, both raw returns
and cumulative abnormal returns adjusted by size and B/M; second is the abnormal returns based
on Fama and French (1993) three-factor regressions; and the third is the abnormal returns during
subsequent earnings announcement periods.
3.1.1. Cumulative returns, both raw and adjusted
We first examine the subsequent raw returns and returns adjusted by size and B/M over
the one-, six-, and 12-month holding periods. The methodology of calculating the cumulative
abnormal returns are described in the Appendix. Table 2 presents the raw (Panel A) and adjusted
(Panel B) returns for the silence, buy, sell portfolios, as well as the silence-sell and buy-sell
12
Without excluding the low-priced small stocks, the average 12-month return for the period of 1990-2011 is
15.5%. Our main conclusion is robust with or without excluding these stocks. See Section 4.5 for detail.
8
spreads. All statistical tests (this table and all other tables throughout this paper) are based on
Newey-West standard errors.
In Panel A of Table 2, the silence portfolio on average earns raw returns of 0.90%,
5.34%, and 11.08% over the one-, six-, and 12-month holding period, respectively. The sell
portfolio earns raw returns of 1.01%, 6.26%, and 13.10% over the one-, six-, and 12-month
holding period, respectively. The silence-sell spreads are all negative over these holding periods,
and significantly so for the longer holding periods. There is some evidence that the buy portfolio
earns higher returns than the sell portfolio in the one-month holding period.
Because firm characteristics are different among these portfolios, it is more informative
to examine abnormal returns. Panel B of Table 2 presents such results, where we adjust the
returns by size and B/M. Later in calendar-time portfolio regressions and Fama-MacBeth
regressions we also control for momentum and other potentially relevant variables. Panel B
shows that the silence portfolio earns abnormal returns of -0.15% (t=-2.41), -0.87% (t=-3.60),
and -1.74% (t=-4.29) over the one-, six-, and 12-month holding periods, respectively. By
comparison, the sell portfolio earns 0.00% (t=0.01), 0.30% (t=0.87) and 0.75% (t=1.26) over the
one-, six-, and 12-month holding periods, respectively.13
As a result, the silence-sell spreads are -
0.15% (t=-2.77), -1.17% (t=-4.19), and -2.50% (t=-4.10) over the one-, six- and 12-month
holding periods, respectively. These results support the litigation risk hypothesis.
The two panels also present the returns following insider net buying and the buy-sell
spreads. In the very short holding period (one month), returns are higher (with marginal
13
Note it is well documented in the literature that insider sales do not predict significant negative returns. For
example, Lakonishok and Lee (2001, p.102), in their comprehensive study of insider trades, report a small one-year
abnormal return of -0.5% for low net purchase ratio (a measure similar to net insider demand in our paper); the
abnormal return following low net purchase ratio is a positive 1.6% for larger firms; even more telling, Lakonishok
and Lee (2001, p.106-7) find that a strong insider selling signal actually predicts positive, instead of negative, future
returns. Similarly, Jeng, Metrick, and Zeckhauser (2003, p.456) report that there is no noticeable abnormal returns
over the 100 days after insider selling; Sias and Whidbee (2010, 1575) find that the 6-month abnormal returns
following net insider selling are 0.717% for medium-cap stocks and 0.045% for large-cap stocks.
9
statistical significance) following insider net buying than following net selling. For longer
holding periods, however, returns following insider net buying and following net selling are
largely indistinguishable.14
Note that in our sample the very small stocks are excluded. In
unreported analysis when these (many) stocks are added back in the sample, we find that the
buy-sell spreads are positive and more significant. These results are consistent with the prior
literature that insider trading activities predict future returns mostly among small firms (e.g.,
Lakonishok and Lee, 2001; Sias and Whidbee, 2010).15
[Insert Table 2 about here]
3.1.2. Monthly alphas
Alternatively, we examine the future returns following insider trading decisions using
monthly alphas from asset pricing models such as the one by Fama and French (1993).
Specifically, Each month j from January 1989 to December 2012, portfolios are formed on
preceding insider trading activity over the trailing six months. Stocks with no insider trading
activity over the prior six-month period form the “silence” portfolio; stocks with positive and
non-positive net insider demand (NID) over the prior six-month period form the “buy” and “sell”
portfolios, respectively. The portfolios are held over months j+1 to j+k (k=1, 6, or 12). Portfolio
returns are equal weighted (value weighted) across their constituent stocks. The overall portfolio
return for month j is the equal-weight average month-j returns of the strategy implemented in the
prior month and strategies formed up to k (k=1, 6, or 12) months earlier. These monthly
portfolios (from January 1990 to December 2012) are then regressed on the Fama and French
14
Forming finer portfolios within the net selling and net buying groups does not reveal any significant differences
either. The lack of significance of the buy-sell spread is largely due to our sampling procedure in which we exclude
the very small stocks. Untabulated analysis shows that adding back these stocks into the sample does generate a
positive and marginally significant buy-sell spread in adjusted returns over the one-year period (0.83% with t=1.52). 15
It is worth noting that a silence-sell return spread of 2.50% does not appear attractive for an aggressive trading
strategy. In this paper we simply point out that insider silence has information content, which is negative. For
potential trading strategies that utilize this phenomenon, we refer readers to contemporaneous work by Ma (2013)
and Ma and Ukhov (2013).
10
(1993) three-factor model. Table 3 presents the equal-weight (Panel A) and value-weight (Panel
B) monthly alphas for the holding period of one, six, and 12 months, respectively. Adding the
momentum factor, Pastor and Stambaugh (2003) liquidity factor, and Sadka (2006) liquidity
factor generates similar results and are thus not reported.
Panel A shows that the silence portfolio generates equal-weight monthly alphas of -
0.15% (t=-2.54), -0.19% (t=-2.95) and -0.21% (t=-2.77) for the strategy of holding one, six, and
12 months, respectively. By contrast, the monthly alphas of the sell portfolio are much smaller
and not significant. The silence-sell spreads are -0.16% (t=-2.62), -0.24% (t=-4.23) and -0.26%
(t=-4.89) for the strategy of holding one, six, and 12 months, respectively. These results are
consistent with those presented in Table 2 based on the size and B/M adjusted cumulative
abnormal returns. The value-weight results are statistically weaker. At the value-weight base, the
silence-sell spreads are -0.13% (t=-1.42), -0.14% (t=-1.78) and -0.13% (t=-1.85) for the strategy
of holding one, six, and 12 months, respectively. Since value-weight results are dominated by the
largest stocks, the results suggest that the silence-sell spread is weaker among the largest stocks.
On the other hand, in both equal- and value-weight bases, the buy-sell spreads are not significant
in any holding period. This result is also consistent with that in Table 2.
[Insert Table 3 about here]
3.1.3. Earnings announcement abnormal returns
The results so far suggest that investors fail to fully incorporate the information contained
in preceding insider trading activity. To further investigate whether investors fail to incorporate
the negative information in insider silence and whether they are systematically surprised when
the relevant information is subsequently disclosed to the market, we examine the earnings
announcement period abnormal returns. Specifically, we extract quarterly earnings
11
announcement dates from Compustat and calculate three-day announcement period abnormal
returns adjusted by CRSP equal-weight daily market returns (i.e., an event window [-1, +1]
covering one trading day before to one day after the earnings announcement date). Within an
insider trading portfolio, we then calculate the average three-day abnormal returns of its
constituent firms for the first four quarterly earnings announcements following portfolio
formation. Table 4 presents the time-series average of the abnormal returns for the whole sample
and the three portfolios (“silence,” “buy,” and “sell”) for each of the four quarterly
announcements and the four-quarter average.16
Due to data overlapping, tests on the time-series
average earnings announcement abnormal returns are Newey-West adjusted with two lags for
quarterly data and 11 lags for the four-quarter average.
Table 4 shows that the silence portfolio is associated with negative earnings
announcement returns during each of the four announcements, with an average return of -0.08%,
which is significant at the 5% level. By contrast, the sell portfolio is associated with positive
announcement period abnormal returns, with an average of 0.06%. As a result, the silence-sell
spreads are negative and significant for all four quarters, with an average of -0.15% (t=-5.06).
In sum, the results in Table 4 provide further support for the litigation risk hypothesis that
insider silence is associated with negative information. Further, the evidence suggests that
investors fail to incorporate the insider silence information into stock prices.
[Insert Table 4 about here]
3.2. Cross-sectional variation
The litigation risk hypothesis posits that when insiders possess significant negative
information, for fear of shareholder litigation they keep silent by abstaining from trading shares.
16
We follow the methodology developed in Chopra, Lakonishok, and Ritter (1992), which is used in Jegadeesh and
Titman (1993), La Porta, Lakonishok, Shleifer, and Vishny (1997), and Titman, Wei, and Xie (2004), among others.
12
The hypothesis has two cross-sectional implications. First, when the information environment of
a firm is worse, it is more likely that insiders possess important information that outsiders do not
know. Thus, we expect the silence-sell spreads are wider among firms with worse information
environment. Second, if litigation risk drives insiders’ decision not to sell on negative
information, the silence-sell spreads are wider among firms at higher litigation risk. This section
examines these two cross-sectional implications. Here we focus on the one-year cumulative
abnormal return.
3.2.1. The role of information environment
We choose book-to-market (B/M) ratio, firm age, and stock volatility as proxies for
information environment (e.g., Zhang, 2006; Choi and Sias, 2012). Information about firms with
low B/M is more ambiguous (Daniel and Titman, 1999); younger firms have worse information
environment; and volatile stocks have worse information environment. We do not use firm size,
institutional ownership, or the number of analysts following as proxies for information
environment here, because these firm characteristics are also related to litigation risk and because
there are opposite implications on the silence-sell spread. For instance, on one hand, smaller
firms have a worse information environment, implying a wider silence-sell spread. On the other
hand, smaller firms are at lower litigation risk (e.g., Kim and Skinner, 2012), implying a
narrower silence-sell spread. These two opposite effects partially cancel out. Thus, the cross-
sectional implication on these variables is unclear ex ante and becomes an empirical issue.
To explore the role of information environment on the silence-sell spread, we first form
terciles on one of the proxies (B/M, age, and volatility), and within each tercile we form
portfolios on preceding insider trading activity and examine their future returns. Table 5 reports
the results of cumulative abnormal returns over a one-year holding period. Results with one- or
13
six-month holding periods lead to the same conclusions. Panel A shows that the silence-sell
spreads are -4.29% (t=-4.85) among growth stocks (the bottom 1/3 stocks based on B/M) and -
1.30% (t=-2.73) among value stocks (the top 1/3 stocks based on B/M). The growth-value
difference between the two spreads is -2.98% (t=-4.30). Note all t-statistics throughout this paper
are based on Newey-West standard errors. To the extent that lower B/M indicates worse
information environment, the result in Panel A is consistent with our hypothesis that the silence-
sell spread is wider among firms with worse information environment.
In Panel B of Table 5 we form subsamples on firm age. As mentioned earlier, younger
firms have worse information environment. Thus we expect wider silence-sell spreads among
younger firms. Indeed, Panel B shows that the silence-sell spreads are -3.58% (t=-4.77) among
younger firms (lower 1/3 based on firm age) and -1.19% (t=-2.17) among the older firms (top 1/3
based on firm age), leading to a significant young-old difference of -2.39% (t=-3.82). Similarly,
Panel C shows that the silence-sell spreads are -4.76% (t=-4.83) among the most volatile stocks
(top 1/3 on stock volatility) and -1.49% (t=-3.47) among the least volatile stocks (bottom 1/3 on
stock volatility), leading to a significant high-low difference of -3.27% (t=-3.70). To the extent
that firm age and volatility proxy for information environment, these results are consistent with
the hypothesis that firms with worse information environment have a wider silence-sell spread.17
[Insert Table 5 about here]
3.2.2. The role of litigation risk
17
In untabulated analysis, we examine the silence-sell spreads among subsamples formed on firm size. Stocks with
market cap below the NYSE 20th
percentile form the “small” group; stocks with market cap above the NYSE 50th
percentile form the “large” group; all other stocks form the “medium” group. The silence-sell spreads are -1.95%
(t=-2.23) among small firms, -1.69% (t=-2.08) among large firms, and -3.54% (t=-5.46) among medium-sized firms.
Consistent with the conjecture that the relation is ex-ante ambiguous between firm size and the silence-sell spread in
future returns, there is no significant difference between the small and large groups (0.26% with a t=0.29).
14
We use three proxies for litigation risk: KS, FPS, and sales growth. Recent literature on
litigation risk suggests a composite measure developed by Kim and Skinner (2012, KS).18
Note
that the KS composite (see Appendix for details) is built upon FPS (following Francis, Philbrick,
and Schipper, 1994), firm size, sales growth, past stock returns, return skewness, volatility, and
trading volume, some of which are also related to information environment. We then choose two
of the components (FPS and sales growth) as relatively “pure” proxies for litigation risk.19
If
litigation risk is behind the spread between insider silence and net selling, one expects that the
spread is wider among firms with higher level of litigation risk.
Table 6 presents the results. In Panel A, KS is the proxy for litigation risk. Among firms
with low KS (bottom 1/3) the silence portfolio earns abnormal returns of -2.05% (t=-1.91) while
the return for the sell portfolio is 0.04% (t=0.03), resulting in a silence-sell spread of -2.09% (t=-
4.94). Among firms with high KS (top 1/3) the silence-sell spread is -3.91% (t=-4.07). The high-
low difference is a significant -1.82% (t=-2.30). Thus, silence-sell spread is wider among firms at
higher litigation risk as measured by KS, consistent with the litigation risk hypothesis.
In Panel B the litigation risk proxy is FPS (1994), which is based on industry
membership. Among firms that do not belong to the industries with higher litigation risk, the
silence-sell spread is -1.96% (t=-2.94), while the spread is -3.88% (t=-4.18) among firms that
belong to the industries with higher litigation risk. The difference between the two groups is a
significant -1.92% (t=-1.96). In Panel C the proxy for litigation risk is sales growth. Firms with
low sales growth (bottom 1/3) have a silence-sell spread of -1.66% (t=-2.56) while firms with
high sales growth (top 1/3) have a silence-sell spread of -4.12% (t=-4.21), resulting in a high-low
18
We gratefully appreciate the advice from Professors Kim and Skinner on constructing this variable (KS). 19
We do not choose the others because they are derived from past stock returns and proxy for the size of damage in
shareholder litigation. Such measures are sensitive to how the measuring window overlaps with the class period. By
contrast, FPS is simply industry membership and sales growth is based on past two years’ sales data, both relatively
less sensitive to how closely their measuring window overlaps with the class period.
15
difference of -2.45% (t=-2.69). This result provides further evidence that firms at higher
litigation risk are associated with a wider silence-sell spread.
[Insert Table 6 about here]
3.2.3. Fama-MacBeth regressions
This section presents Fama-MacBeth regressions. Compared to portfolio analysis, linear
regressions are more flexible to control for other firm/stock characteristics. The dependent
variable is the one-year abnormal returns following portfolio formation. The main independent
variable of interest is a binary variable “silence,” which is equal to one if for a stock/month
insiders have kept silent over the past six months, and zero otherwise. We control for firm size,
book-to-market, and past stock returns, in addition to a binary variable “buy,” which is equal to
one if insiders of the firm have net bought over the past six months. The benchmark case is when
insiders net sell. Thus, the coefficient of the “silence” variable represents the silence-sell spread
after controlling for other variables.
Table 7 presents seven regression models. Each column shows the time-series averages
of 264 cross-sectional regression coefficients with t-statistics presented in parentheses, which are
based on Newey-West standard errors. Model 1 shows a significant coefficient of -2.69% (t=-
4.66) on the “silence” variable, representing the silence-sell spread after controlling for size,
B/M, and past return. The coefficient on “buy” variable is not statistically significant.20
These
results reconfirm the portfolio results as presented in Table 2.
In Model 2, we add the interaction term between “silence” and B/M. The variable B/M
itself is already in the list of control variables. In this and all subsequent regressions in this table
that have interaction terms with the “silence” variable, all the interacted variables are demeaned
20
In an unreported regression, in which a continuous variable of net insider demand is added as a control variable ,
the “silence” variable has a coefficient of -2.68% (t=-4.46), similar to that in Model 1 of Table 7.
16
so that the coefficient on “silence” itself means the silence-sell spread when the interacted
variable is at the sample mean. In Model 2, the coefficient on this interaction term represents
how the silence-sell spread changes with B/M. The coefficient is a significant positive 1.70%
(t=6.14). This result further confirms that the silence-sell spread is wider (more negative) for
firms with lower B/M ratio (growth firms), a result consistent with the hypothesis that the
silence-sell spread is wider among firms with worse information environment. This conclusion is
further supported in Models 3 and 4, where we add the interaction term between “silence” and
firm age (model 3) or volatility (model 4). Both coefficients on the interaction terms are
significant with the expected signs. That is, the silence-sell spread is wider (more negative) for
younger firms and for stocks that are more volatile. In sum, Models 2 to 4 reconfirm the portfolio
results in Table 5 that the silence-sell spread is wider with worse information environment.
In Model 5, we add an interaction term between “silence” and KS, the composite proxy
for litigation risk. If the silence-sell spread is due to insiders’ fear for shareholder litigation risk,
this spread should be wider for firms with higher levels of ex-ante litigation risk. That is, the
interaction term should have a significant negative coefficient. Model 5 shows that the
interaction term silence * KS indeed has a significant negative coefficient of -0.53 (t=-2.17). In
Models 6 and 7, we replace KS with FPS and sales growth, respectively, and find that both
interaction terms have significant negative coefficients. In sum, the regression models 5 to 7
reconfirm that the silence-sell spread is wider among firms at higher litigation risk.21
21
Note that among the three proxies for information environment and three proxies for litigation risk, some are
closely related. For example, the composite measure of litigation risk (KS) has a component of volatility, a
component of FPS, and a component of sales growth. Thus, a regression model with all interaction terms included
could be difficult to interpret. Indeed, when all interaction terms are included (together with the six proxy variables
themselves), the coefficient on KS becomes positive, which is weakly significant. All other interaction terms still
have the expected signs, and the coefficients on silence*volatility and silence*ln(B/M) remain statistically
significant. Unreported for brevity, these results are available upon request.
17
Note that in all seven models, the coefficients on “silence” are all significant at the 1%
level, ranging from -2.52% to -2.85%. These results suggest that when the proxy variables take
the sample mean values, the silence-sell spreads are all significant. Overall, Table 7 provides
further results, in a regression setting, that the silence-sell spread is negative, which
systematically varies with proxies for information environment and proxies for ex-ante litigation
risk. These results are all consistent with the litigation risk hypothesis that, for fear of litigation
risk, insiders abstain from trading on negative information.
[Insert Table 7 about here]
4. Alternative hypotheses and robustness checks
4.1. Other information signals
The evidence so far indicates that insider silence is a signal of negative information that
predicts future returns. It is important to check whether it is actually just a proxy for other
information signals. In this section we consider three types of signals.
First, we check whether insider silence is a proxy for underperformance of firms that
have just conducted initial public offering (IPO). Note that we require sample firms at least one-
year old, so that most sample firms have passed the IPO lockup period (see Brav and Gompers,
2003, p.7) and insiders of the sample firms have at least six months to make insider trading
decisions without the IPO lockup restriction (see section 2). For the remaining IPO firms that
have lockup periods longer than six months, their insiders are still restricted to sell shares during
our six-month window to measure insider trading. Thus, it is important to check whether our
results are driven by the underperformance of these IPO firms (e.g., Ritter, 1991). To account for
this possibility, we add in the Fama-MacBeth regressions an IPO dummy variable which is equal
18
to one if the stock is younger than three years (relative to its first appearance in the CRSP
database).
Second, we check whether it is likely that insiders keep silent in response to some
publicly available signals that predict future returns. For example, firms that have issued stocks
in the past tend to underperform (Loughran and Ritter, 1995), and firms that have higher levels
of accruals tend to underperform (Sloan, 1996). If insiders abstain from trading as a response to
past stock issuance or accruals, then insider silence should not have any information content after
controlling for these other public information signals. We choose the five return anomaly
variables examined in Fama and French (2008) as proxies for the public information signals.
They are net stock issues, momentum (already in our original regressions), accruals, asset
growth, and profitability.
Third, in our hypothesis we implicitly assume that insiders have shares available to sell
but, for fear of shareholder litigation, choose not to. It is possible, however, that the outcome of
“no trading” is due to portfolio constraints as discussed in Marin and Olivier (2008). That is,
insiders cannot short sell their own company shares, insiders have concerns about corporate
control, or the shares insiders hold are restricted and not available for sale. In the framework of
Marin and Olivier (2008), uninformed investors, when observing reduced insider trading
following a period of heavy selling, infer that insiders possess bad news and update their beliefs
on expected payoffs and risk, leading to a price crash.22
To address this possibility, we control
for the net insider demand over the six-month period prior to the current time period. That is, we
control for net insider demand over months -12 to -7 relative to portfolio formation.
22
Strictly speaking, the insider silence phenomenon we examine is fundamentally different from the reduced insider
selling volume studied in Marin and Olivier (2008). In addition, Marin and Olivier (2008) model the probability of
price crash only but do not predict subsequent stock returns, which are the focus of our paper.
19
The Fama-MacBeth regression results are presented in Table 8. Based on the first model
in Table 7, Model 1 in Table 8 adds a dummy variable “IPO firm,” which is equal to one if the
stock is younger than three years. This variable has a significant negative coefficient of -2.49%
(t=-2.43), consistent with IPO underperformance. The coefficient on “silence” remains negative
and significant, with a magnitude similar to that in the first model of Table 7. Thus, Model 1 of
Table 8 suggests that IPO underperformance does not explain the insider silence effect.
In Model 2, we add the anomaly variables: net stock issues (NS), accruals (Ac/B), asset
growth (dA/A), and profitability (Y/B) in the regression. Note that momentum is already in the
list of control variables. Consistent with the literature, firms that issue stock in net underperform
(Ikenberry, Lakonishok, and Vermaelen, 1995; Loughran and Ritter, 1995; Daniel and Titman,
2006; Pontiff and Woodgate, 2008; McLean, Pontiff, and Watanabe, 2012); higher accruals are
associated with lower returns (Sloan, 1996); firms that grow faster in the past tend to earn lower
returns (Fairfield, Whisenant, and Yohn, 2003; Titman, Wei, and Xie, 2004; Cooper, Gulen, and
Schill, 2008); and higher profitability is associated with positive returns (Haugen and Baker,
1996; Cohen, Gompers, and Vuolteenaho, 2002). These results are largely consistent with Fama
and French (2008). Nevertheless, the coefficient on the “silence” dummy remains negative and
significant at -2.77 (t=-4.53), even after controlling for these anomaly variables. Although the
five anomaly variables (with past return included) do not represent all publicly available
information signals, the evidence in Model 2 is clear that these important publicly available
informational signals do not explain the underperformance associated with insider silence.
In Model 3, we add past trades by insiders (NID over months -12 to -7). The coefficient
on past trades is not significant. By contrast, the coefficient on “silence” is virtually unchanged.
Thus, insider silence contains new information beyond what is contained in earlier insider
20
trading activities. Model 4 adds all aforementioned variables in the regression. The coefficient on
“silence” remains negative and significant with a similar magnitude (-2.69% with t=-4.43).
Further, in an untabulated analysis, we add short interest as a control and find that the coefficient
on “silence” remains significant and negative (-2.11% with t=-3.70) and that the short interest
variable carries a significant negative coefficient, consistent with the literature.
In sum, the evidence in Table 8 suggests that the information contained in insider silence
is not explained by IPO underperformance, a broad list of anomaly variables, and insider trading
activity of earlier time period.
[Insert Table 8 about here]
4.2. Insider silence and positive information: the takeover cases
We motivate the phenomenon of insider silence from the perspective of shareholder
litigation, which represents the primary enforcers of insider trading regulations. Because
shareholders are more likely to launch lawsuits following large price drops, the litigation risk is
higher when insiders sell on negative news than when insiders buy on positive news.23
This
argument establishes that insider silence is more likely associated with negative information. The
empirical evidence presented so far confirms this hypothesis. There are, however, situations in
which insiders also abstain from trading when anticipating large stock price jumps. After all, the
regulation on insider trading, in particular 10b-5, applies to both buying and selling on
information. Thus, it is important to consider the potential relation between insider silence and
positive informational events. Of interest is the insider trading decisions prior to takeover offers,
as stock prices on average experience significant jumps at the public announcement. Indeed,
23
See Bettis, Coles, and Lemmon (2000, p.208), Ke, Huddart, and Petroni (2003, p.316), Cheng and Lo (2006, p.
821), Piotroski and Roulstone (2008, p.410), Rogers (2008, p. 1269), Lee, Lemmon, Li, and Sequeira (2012),
Cohen, Malloy, and Pomorski (2012, p. 1040, model 4, Panel B of Table IX), among others.
21
Agrawal and Nasser (2012) document that insiders of takeover targets scale back from selling
shares prior to the public announcement.
In this section we reconcile the average negative relation between insider silence and
future returns as we have shown so far in this paper with a possible positive relation in the case
of takeover offers. Specifically, we examine how frequent takeovers are, how likely insiders
abstain from trading over a six-month period prior to takeover announcements, and the extent to
which takeover events contaminate the average negative signal in insider silence. We first
classify every firm/month as either targeted or non-targeted. A firm/month is targeted if this firm
receives a takeover offer over the subsequent 12-month period; otherwise, it is non-targeted. We
then form portfolios based on both preceding insider trading (silence, buy, and sell) and
subsequent takeover status (targeted vs. non-targeted).
Table 9 presents the average numbers of stocks in the portfolios and their future one-year
returns. For convenience we also present the whole sample data. For the whole sample of
average 2,487 stocks, 141 (5.7%) are classified as targeted during the subsequent 12-month
period. These targeted firms receive an annual return of 20.58% on average, which is consistent
with the literature on mergers and acquisitions (e.g., Andrade, Mitchell, and Stafford, 2001).
Accordingly, the non-targeted portfolio earns a significant negative return of -1.18% (t=-2.64).
Interestingly, among the firms that are targeted, insiders net sell in more than half (74/141) of the
firms over the past six months, net buy in about 20% (28/141) of the firms, and keep silent in
only 28% (39/141) of them. Thus, there is no evidence that insiders predominantly abstain from
trading own-company shares over the six-month period prior to the public announcement.24
After
24
This result is probably related to how long it takes for a takeover deal from deal initiation to the public
announcement. The merger literature seems to use a two-month (Schwert, 1996) or three-month (e.g., Boone and
Mulherin, 2007) period to measure the price runup. If the start of the stock price runup period is somewhere near the
start of the time point insiders refrain from buying shares, it is not surprising that insider silence does not emerge as
22
excluding these targeted firms, the silence portfolio earns a return of -3.10% over the subsequent
12-month period, and the sell portfolios earns -0.39%, resulting in a silence-sell spread of -2.72%
(t=-4.42). This return spread is comparable to the whole sample spread of -2.50% (t=-4.10).
The evidence in Table 9 suggests that for the sample we examine in the paper, the
probability of receiving a takeover offer over a one-year period is not trivial but relatively low
(5.7%); more importantly, there is no evidence that insiders predominantly keep silent over the
six-month period prior to the public announcement; the silence-sell spreads are comparable with
or without excluding these takeover cases. Therefore, the conclusion is in line with our
hypothesis: Insider silence on average signals negative insider information.
[Insert Table 9 about here]
4.3. Long term performance
The results so far suggest that insider silence is associated with negative returns over the
future one, six, and 12 months. In this section we study the long-term stock performance
following preceding insider trading activity. We have two goals in mind. First, following the
litigation risk hypothesis, if insiders keep silent because they possess negative information, a
priori it is unclear whether the negative information is completely incorporated into stock prices
over a one-year period. An investigation of the long-term performance indicates how long it
takes for the negative information to be incorporated into prices. Second is to exclude the
possibility that the results are just by chance. If the relation between insider silence and negative
future returns is by chance, it is expected that the negative returns reverse over the long term.
Table 10 presents the returns of the silence and sell portfolios over the 2nd
& 3rd
years
following portfolio formation. The silence firms earn an average of -1.65% (t=-4.53) and -0.64%
an overwhelming phenomenon during a six-month period prior to the public announcement. After all, the insider
trading pattern of the targeted firms over the first three or four months of the six-month period is no different from
that of a typical firm.
23
(t=-1.86) over the 2nd
and 3rd
subsequent years, respectively. Thus, there is no reversal over the
long term, and it takes more than a year for the negative information to be incorporated into
prices. By comparison, the sell portfolio earns positive returns over each of the two subsequent
years. As a result, the silence-sell spreads are negative and significant over the two years. The
evidence clearly rejects the possibility that the main results in the paper are by chance.
It is worth noting that the long-term returns following insider net selling are not negative.
These results are consistent with the existing literature. For example, Lakonishok and Lee (2001,
p.102) report abnormal returns to a portfolio of low net purchase ratio (more likely net selling)
are 1.0% and 1.2% for the second and third years, respectively, following portfolio formation.
By contrast, the corresponding returns to a portfolio of high net purchase ratio (more likely net
buying) are 0.1% and -0.1% over the second and third years, respectively. Similarly, Sias and
Whidbee (2010, p.1575) report a two-year abnormal return of 1.93% and 0.28% for mid- and
large-cap stocks, respectively, following net insider selling.
The result that insider net selling is associated with non-negative (or even positive) long-
term returns appears counterintuitive to conventional wisdom, which suggests that insider net
selling is motivated, at least partly, by bad news and thus predicts negative future returns. If one
takes into account the litigation risk involved in insider selling, however, it is not surprising non-
negative long-term returns are associated with insider net selling. Despite the high-profile cases
that some insiders commit illegal insider trading by selling on significant negative news, the
result indicates that, on average, insiders do not sell on significant negative information. The fact
that a small number of insiders who committed apparently illegal insider trading were
subsequently exposed to the media and prosecuted by legal forces also helps remind all corporate
24
insiders the litigation risk involved in insider trading.25
Such legal actions help maintain the
equilibrium outcome in which corporate insiders as a whole do not sell shares on significant
negative information. Indeed, Cohen, Malloy, and Pomorski (2012, p. 1039) show that
opportunistic insider trading, which predicts returns up to six months, is sensitive to the number
of SEC sanctions and that opportunistic trades (especially sales) attract subsequent SEC activity.
The results on insider trading in this paper suggest that corporate insiders on average sell shares
while anticipating not so negative (or mildly positive) returns in the future and abstain from
trading while anticipating significant negative future returns.
[Insert Table 10 about here]
4.4. Database quality
In our analysis we classify a firm to the “silence” group if we do not observe any insider
trading activity as recorded in the database. The validity of this definition depends on the quality
of the database, Thomson Reuters. Several other possibilities could lead to observing no insider
trading. First, insiders have traded but failed to report to the SEC or the SEC’s record of insider
trades is incomplete.26
Thus, a researcher could mistakenly assign stocks to the “silence”
portfolio while they actually belong to the “buy” or “sell” portfolio. If these unfiled or
unrecorded trades are not driven by private information, the incompleteness of the database
simply adds random noise to the data and works against our empirical testing.
Second, insiders have traded on private information but decided not to report to the SEC
due to fear of regulatory and legal action. Seyhun and Bradley (1997, p. 200) discuss this
25
For recent legal cases against such apparent illegal insider trading, see Pulliam and Barry (2012), “Executives’
good luck in trading own stock,” The Wall Street Journal, November 22, 2012. 26
The insider trading data we use are made available through Wharton Research Data Services (WRDS). The
WRDS web has the following description of the database: “The Insider Filing Data Feed (IFDF) is designed to
capture all U.S. insider activity as reported on Forms 3, 4, 5, and 144 in line-by-line detail.” (See http://wrds-