Journal of Accounting and Economics 60 (2015) 18–35
Contents lists available at ScienceDirect
Journal of Accounting and Economics
journal homepage: www.elsevier.com/locate/jae
CEO opportunism?: Option grants and stock trades around
stock splits$
Erik Devos a,n, William B. Elliott b, Richard S. Warr c
a
b
c
College of Business Administration, University of Texas at El Paso, USA
Boler School of Business, John Carroll University, USA
Poole College of Management, North Carolina State University, USA
a r t i c l e in f o
abstract
Article history:
Received 22 October 2013
Received in revised form
12 February 2015
Accepted 27 February 2015
Available online 14 March 2015
Decades of research confirm that, on average, stock split announcements generate positive
abnormal returns. In our sample, 80% of CEO stock option grants are timed to occur on or
before the split announcement date. With the average market-adjusted announcement
return of 3.1%, awarding the grant before the split announcement results in an average
gain per CEO-grant of $451,748. We find additional evidence consistent with timing of CEO
stock trading around the split announcement. In the case of CEO stock sales, about twothirds occur after the split announcement, resulting in an average gain of $345,613.
& 2015 Elsevier B.V. All rights reserved.
JEL classification:
G14
J33
Keywords:
Managerial incentives
Executive compensation
Stock splits
1. Introduction
A number of studies document the apparent timing of various corporate events relative to the granting of executive stock
and option plans. For example, Aboody and Kasznik (2000) analyze stock prices around earnings forecasts and present
evidence consistent with the hypothesis that CEOs opportunistically time voluntary corporate disclosures relative to grants
of compensation plans, in an attempt to maximize their personal wealth. Our study contributes to this literature by
examining the timing of option grants to CEOs around the announcement of a stock split. Our approach allows us to more
directly link the actions of the CEO and Board of Directors (BOD) to a major corporate event (the announcement of a stock
split) and the timing of CEO option grants. Further, we are able to provide accurate estimates of opportunistic gains to the
average CEO and show that these gains are economically significant. We provide corroborating evidence of apparent
opportunism by examining CEO trading activity around the split announcement. Understanding patterns consistent with
☆
We thank Michelle Hanlon (editor), Terry Shevlin (referee), Suzanne Bellezza, Elizabeth Devos, Randall Heron, Srini Krishnamurthy, Todd Perry, Walter
Reinhart, Dan Rogers, Charles B. Ruscher, Richard Smith, Brian Young (FMA discussant), and seminar participants at the FMA Meetings (Chicago), Cleveland
State University, John Carroll University, Kent State University, Loyola University - Maryland, Portland State University, and the University of Texas at El Paso
for comments. Kim Duong, He Li, and Sayan Sarkar provided research assistance. We retain responsibility for any remaining errors.
n
Corresponding author. Tel.: þ1 915 747 7770; fax: þ 1 915 747 6282.
E-mail addresses: [email protected] (E. Devos), [email protected] (W.B. Elliott), [email protected] (R.S. Warr).
http://dx.doi.org/10.1016/j.jacceco.2015.02.004
0165-4101/& 2015 Elsevier B.V. All rights reserved.
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
19
opportunistic behavior is important, as these behaviors may reflect inadequate governance or governance breakdowns (e.g.,
Bebchuk et al., 2010; Collins et al., 2009; Morgenson, 2004).
The effect of stock split announcements on share prices has been studied for over 40 years, and one of the consistent
results of this scrutiny is the significantly positive abnormal stock return found around the announcement of the stock split.1
This result is perplexing because, economically speaking, stock splits are cosmetic – they merely adjust the price of the
minimum unit of trade. There are however, non-trivial costs related to splits, and yet their use has remained over a long
period of time as firms persist in splitting their shares. Split announcement returns are not just positive on average, they are
positive for approximately three-quarters of all splits. For the typical firm, a stock split announcement results in a fairly
certain share price increase without changing the underlying characteristics of the firm or altering the timing or content of
other corporate disclosures.
In this paper, we examine whether the granting of stock options to CEOs appears to be timed in relation to the split
announcement. Further, we test whether there is any apparent opportunistic timing of the CEO's personal sales or purchases
of the firm's stock around split announcements, so as to maximize personal wealth. Our empirical evidence is consistent
with the idea that CEOs behave in an opportunistic fashion around stock split announcements.
For stock splits that also have option grants within a 21-day window centered on the split announcement, we find that
80% of these grants occur prior to or on the announcement day. This is statistically different from the 20% of CEO stock
options granted after the split announcement. The announcement of a stock split is a discretionary event, so it is not
surprising that the result holds for scheduled, unscheduled, and unclassified option grants (as it is possible to time a split to
follow an option grant, regardless of whether that grant is scheduled or unscheduled). We estimate that such timing results
in an average increase in the value of the options awarded by a splitting firm to the CEO, of $451,748 (with a median of
$39,234).
Insider trading data provides further evidence consistent with the notion that CEOs behave opportunistically around split
announcements. We find significantly higher levels of selling immediately after compared to before the split annoncement.
In addition we find the opposite effect for stock buying activity by CEOs. The overall amount of trading around splits is four
times greater than the trading in the prior year for the same calendar time period. As a result of this trading behavior, on
average CEOs are $345,613 (with a median of $48,188) better off from selling after the split rather than before. Together with
the granting behavior; we interpret this evidence as consistent with our hypothesis that, on average, CEOs behave
opportunistically around the announcement of stock splits.
Although we have made every effort to carefully analyze the data, some qualifications remain. First, while we can observe
the benefits of timing option grants around the split announcement, we cannot say definitively that the CEO's gains are
directly at the expense of shareholders. Typically, only highly successful firms engage in stock splits. Therefore, the apparent
opportunistic timing of option grants could simply be another way in which the BOD rewards and retains a high performing
CEO. We investigate this possibility by examining the total compensation as well as the amount of options granted in the
prior year. In general, total compensation and the number of option grants are significantly higher in the year of the split,
relative to the prior year. Second, we do not address whether the ability to time options relative to splits results in more
splits occurring – in other words we assume that the decision to split is exogenous. Examining whether option-based
compensation leads to abnormal splitting behavior is an interesting question for future research. A third caveat relates to the
economic gain of the CEO. The economic gain from timing option grants is a “paper” gain only and we only estimate this for
the current option grant. Any prior option grants would also gain in value and may even be vested at the time of the split
announcement (a portion of this gain may be observed in the insider trading analysis), however given the available data, we
are unable to obtain more definitive estimates of economic gain. Finally, while our evidence is consistent with opportunistic
behavior by CEOs, it is not possible to directly test CEO intent.
The rest of the paper proceeds as follows: Section 2 provides a brief literature review while in Section 3 we develop our
hypotheses. Section 4 presents a description of the data, whereas Section 5 presents our results. Section 6 discusses the
economic significance of our findings, and Section 7 concludes the paper.
2. Literature reviewand institutional detail
Our paper is related to two distinct areas of the literature. First, the area of opportunistic managerial behavior and insider
trading that examines managerial actions around corporate events. Second, the literature on stock splits that documents the
mechanics of a stock split and the associated stock price effects of stock splits.
2.1. Opportunistic managerial behavior and insider trading
A growing body of literature finds evidence consistent with opportunistic timing of option awards. Using a sample of
Fortune 500 firms, Yermack (1997) finds strong evidence that CEO stock option awards are timed to occur immediately
before the release of positive corporate news. He argues that this timing is not the result of insider trading, leakage of the
award announcement, or manipulation of the news release. He reports that the mean (median) abnormal increase in the
1
See Fama et al. (1969).
20
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
ESO value for the CEO is $30,000 ($11,100) based on the 20 trading days following the option award date. Aboody and
Kasznik (2000) examine a sample of firms that award compensation on a fixed date annually. They find that these firms
appear to delay the release of value increasing news and speed up the release of value decreasing news around the
compensation award date. While these two papers examine different sets of firms, the effect on CEO wealth is similar. CEOs
enjoy a substantial gain in the value of their compensation packages shortly after the award. Because the announcement of
stock splits is to some extent at the discretion of the firm, our study is similar to Aboody and Kasznik's, in that fixing a
compensation award date will not prevent managers from acting opportunistically.2 We differ from Aboody and Kasnick in
that we observe two events (the stock split and option grant) that must be approved by the BOD and given the infrequency
of board meetings, these two events are likely decided upon at the same board meeting. Thus we believe that the BOD and
or CEO are likely to have both events in mind as they determine the timing of each, making it less likely that our results are
driven by statistical aberration.
Subsequent articles reinforce the premise that CEOs (or managers in general) attempt to manipulate the award of options
for their private benefit. In particular, a number of papers examine the backdating of executive stock options.
Lie (2005) hypothesizes that executives adjust the timing of the grant of unscheduled stock options after they have
observed stock prices over the quarter. His evidence and subsequent research on the subject confirms the “backdating”
hypothesis. In addition to backdating, Narayanan and Seyhun (2008) investigate whether managers engage in “forwarddating” – i.e., waiting until after the decision date of the BOD to see which direction the stock price moves. Forward-dating
benefits the CEO when the firm's stock price drops in the future and the CEO then selects this future date as the grant date.
Dhaliwal et al. (2009) find that CEOs also backdate the exercise of their options to minimize their personal tax expenses (to
the detriment of the corporate tax deduction). They report that this activity resulted in an average (median) tax savings of
$96,000 ($7,000).
In addition to studies on the timing and backdating of option grants, numerous researchers examine insider trading
around corporate events. Noe (1999) for example, studies insider stock trading around management earnings forecasts. He
finds that managers tend to trade their firm's stock after the disclosure, when there is a reduction in information asymmetry
and when managers are less likely to run afoul of insider-trading regulation. This does not mean that their trading behavior
is not self-serving, as he finds net insider selling after positive earnings forecasts and net insider buying after negative
forecasts.
2.2. The mechanics and price effects of stock splits
There are a number of important dates related to the announcement of a stock split. The decision date when the BOD
authorizes or approves the split, is an internal decision and not immediately observable by the market. The date of the board
meeting in which this decision is made is sometimes disclosed after the stock split announcement (usually in the 10-K or 10Q that follows the split).3 After the BOD authorizes a stock split, the firm discloses its intention to split (i.e., on the
announcement date or split date).4 Consistent with the existing literature (e.g., Louis and Robinson, 2005; Nayar and Rozeff,
2001, among others) we use the CRSP “declaration date” as the announcement date. We find an average (median) of 1.6 (0)
trading days between the date of the board meeting at which the stock split was authorized and the declaration date. Hence,
it seems that decisions to split are announced relatively quickly to the public.
When a firm announces a stock split, there is typically a positive price reaction on the announcement day. For example,
Ikenberry et al. (1996) report a mean abnormal announcement day return of 3.38% for 1,275 two-for-one stock splits
initiated by NYSE and Amex firms during the 1975–1990 period, although they also report a declining trend in the
magnitude of these returns. Lin, Singh, and Yu (2009) also report an announcement day abnormal return of just over 3%.5
The substantial announcement return provides an attractive opportunity for an immediate gain on an option grant and is
the focus of our study.6
Following the announcement, and of much less importance for our study, are the record, pay, and ex-dates. Shares begin
trading on a “when-issued” basis on the record date (unlike other types of distributions, the rights to the when-issued
shares continue to transfer to new owners if a sale occurs after this date). Finally, the pay date and ex-date are the two
consecutive days when the additional shares are first issued and then begin to trade in the market.
Nayar and Rozeff (2001) discuss in detail the spacing of these four dates relating to a split. As they note, the number of
days between these dates has changed over time as the settlement process has changed. In their Table 1, they note that for
2
Aboody and Kasznik do not explicitly identify the type of news events (other than earnings announcements) and so it is possible that stock split
announcements are part of their sample. Our sample of 276 stock split announcements coincides with 32 earnings announcements.
3
For our sample of 276 splits, we are able to identify 152 decision dates. In these cases we find specific mention of the board meeting at which the
decision to split was made, or specific mention of the date on which the board declared or authorized the stock split.
4
Each major exchange has specific rules relating to stock splits. On the New York Stock Exchange (NYSE) section 703.2 of its Listed Company Manual
describes the process. For the NASDAQ, rule 5250(e)(6) outlines the procedures for registering and issuing a stock split (i.e., no later than 10 days after the
record date).
5
A more complete discussion of the theories and explanations for why firms split their stock is beyond the scope of this paper. An excellent review of
the literature is provided by Weld et al. (2009).
6
It is important to note that option grants are split protected. We verified this fact by examining a random sample of proxy statements. A typical
statement includes a phrase along the lines of: “all number of shares granted and exercise prices have been adjusted to reflect our stock splits”.
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
21
shares traded on the NYSE, the mean (median) number of trading days between the declaration and record date is 24.2
(17.0), while the number of days from the record to pay date is 14.9 (14.0). The corresponding figures for NASDAQ stocks are
16.0 (12.0) and 13.4 (12.0), respectively. For our sample firms there are 16.6 (11.0) days between the declaration and record
date and 11.3 (10.0) between the record and the pay date. Our sample appears to be qualitatively similar to that of Nayar and
Rozeff. Nayar and Rozeff find a negative (although not large) price reaction on the record date and a positive abnormal
return on the pay date. For our sample of 276 splits we find a mean (median) cumulative 3-day market-adjusted return of
3.1% (2.1%), 0.0% (0.2%), and 0.8% (0.4%) around the declaration, record, and pay dates, respectively.7
A critical element of our experimental design has to do with the decision making process around both stock splits and
CEO option grants. Conditioned upon sufficient authorized, but not outstanding shares, the firm's BOD may authorize a stock
split without shareholder approval. Likewise, the compensation committee of the BOD determines the timing of any option
grant for the CEO. In other words, if the firm has sufficient authorized shares, shareholders need not be consulted for either
the split decision or the compensation grant. This lack of shareholder involvement yields an occasion for the CEO and BOD to
act in an opportunistic fashion. It is likely that if both events are announced in close proximity to one another, the decisions
for both the split announcement and the CEO option grant were made during the same meeting of the BOD.8 Because there
is no mandatory reporting period after the decision to undertake a stock split, the public announcement of the split is
completely at the discretion of management and the BOD. This discretion provides some flexibility in announcing/reporting
both the stock split and the CEO option grant and leads to our hypotheses.
3. Hypothesis development
We begin with two stylized facts: the timing of stock split announcements is at the discretion of the firm (i.e., the
management team and the BOD); and the majority of split announcements result in a positive price effect. Our null
hypothesis predicts that there is no difference in the pattern of option grants or CEO stock trading around the time of a stock
split announcement.
Since option grants are typically granted with an exercise price equal to the closing price on the day of the grant (Hall and
Murphy, 2000; Narayanan and Seyhun, 2008), this pattern of “grant-and-split” would yield an immediate increase in the
ESO value for split announcements with a positive reaction. Because of the leverage inherent in options, the paper gain to
the CEO is substantially greater than the percentage rise in share price. To illustrate this point, we use a fairly typical
example from our data set. On April 3, 2007, Lawrence Kingsley, CEO of IDEX Corp., received grants on a total of 75,300
shares. The options had a maturity of 10 years and a strike price of $51.05. On April 4, 2007, IDEX announced a 3-for-2 stock
split, resulting in a price increase from $51.05 (one day prior to the announcement) to $51.86 (one day after the split
announcement). Assuming an annualized volatility of 20% (estimated from CRSP) and a risk free rate of 4.55%, the value of a
call option on one share of stock on the grant day was $22.30. After the price increase, the value of the same call was $22.99,
a gain of $0.69 per option. The CEO's total gain was $51,871 (75,300 options granted multiplied by the $0.69 gain per option).
Obviously, this is a paper gain and may not translate into tangible wealth immediately, but a paper gain of $51,871 is not
trivial. Further, since the price paths are identical, regardless of whether the grant was made before or after the split
announcement, at a given price level the option with the lower exercise price will always have a higher value.9
Our first alternative hypothesis concerns the timing of option grants relative to the stock split announcement date and is
stated as follows:
H1. Option grants to the CEO are more likely to occur before a split to take advantage of the stock price appreciation that
normally accompanies a stock split.
Although evidence in favor of the above hypothesis would be consistent with the notion that CEOs are behaving
opportunistically, we also explore corroborating evidence. Specifically, we investigate insider trades by the CEO, as this
action is completely at the discretion of the CEO and requires no BOD approval. Given that stock split announcements
usually are accompanied by a positive stock price reaction we hypothesize that:
H2. CEO stock sales will occur more frequently after a split while CEO stock purchases will occur more frequently before
a split.
7
Note that all abnormal returns are cumulative market-adjusted abnormal returns from day 1 to day þ1, using the value weighted index.
We are unable to determine with certainty when the option grant is approved, as this is not publicly reported However, for our sample of 276
splitting firms, the BOD meets on average (median) every 1.7 (2) months. Thus it seems likely that the full board authorization of the compensation award
and the decision to split occur at the same meeting.
9
Initially, it might appear that granting an at-the-money option at a lower price level is advantageous to the firm, because it results in a lower expense
at the time of the grant. This is due to the fact that, ceteris paribus, the value of an at-the-money option is simply the share price (S0) minus the discounted
value of the exercise price (K). Since the discount on a higher K is a larger dollar value, the difference between S0 and K is larger as well. This means there is
a direct relation between the price level of the at-the-money grant and the value of the option. Ergo, granting at a lower price level results in a lower
compensation expense for the firm. However, the total cost to the firm is not only the expense of the initial grant but also the eventual exercise of the
option at a lower exercise price. Even after accounting for the time value of money and the lower initial expense, in the end, this is more costly to the firm.
8
22
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
While the argument can be made that anyone can choose to sell after the split, the reason why post-split CEO selling
might be opportunistic is because the CEO knew in advance to delay his/her trading based on private information (as in Noe
(1999)). Uninformed market participants do not possess this private information and would not know that they should
delay their sales.
4. Data
4.1. Sample
Our initial sample comprises all firms in the CRSP database that implemented at least a 3:2 split and no more than a 3:1
split (with CRSP distribution code 5523 or 5533) from January 1, 1992 to December 31, 2012.10 During this period there are
5,657 splits announced by a total of 3,533 unique firms. We capture stock prices and returns for a 21-day window, centered
on the stock split announcement date.
We use the Thomson Reuter Insider Filing Data Feed (IFDF) to obtain both stock trades and option grants. We only keep
transactions related to the person identified as the CEO. The steps we take to filter each data set are discussed below.
4.1.1. Option grant data
For the option grant data, we follow Lie (2005) and Heron and Lie (2007) and keep transactions with cleanse codes equal
to R, H, or C. We also include transactions with the additional codes L, W and Y.11 We only keep observations with codes of
CALL, EMPO, ISO, NONQ, WTS, OPTNS, or SAR, as these grants are all some form of call option. We drop all grants where the
exercise price is missing. We sum all grants made on the same day, to the same CEO and treat these as a single grant. For the
purposes of our economic significance estimates, we compute a share weighted average of the expiration date and strike
price for these grants.
Using Lie's (2005) method, we identify scheduled, unscheduled, and unclassified option grants. If the option grant
occurred within a week of the one-year anniversary of the prior year's grant, we consider the option to be a scheduled grant.
If the option was granted outside of the anniversary week, we classify the option grant as unscheduled. All other grants,
such as those for which there were no prior year observations for that particular firm, are designated as unclassified. The
initial sample contains 53,436 option grants.
4.1.2. Stock trading data
For the stock trading data, we drop filings that have a transaction code equal to 8, as these represent static holdings.
We also exclude trades that are classified as derivative related in order to more accurately reflect CEO intent. We collapse
multiple trades made on the same day by a CEO into a single trade. The final stock trading data set yields 196,222
observations.
4.1.3. Final sample
We merge the split sample separately with each of the two datasets described above. Our final two datasets are: (1) a set
of option grants made to the CEO and (2) a set of stock trades made by the CEO. Both grants and trades occurred within a 21day window centered on the stock split announcement.12 The option grant sample contains 290 option grants centered on
276 stock splits and the stock trade sample contains 1,291 days with stock trades centered on 587 stock splits.
Of the 290 grants in the option grant sample, 262 represent grants to unique CEOs. Thus, there are a small number of
cases where multiple grants were made to a CEO during the event window. Because these multiple grants occurred on
different days, collapsing them into a single observation could have a material effect on our analysis. We therefore leave
these observations in the data set (a choice that does not qualitatively alter the results). For the stock trade sample, the 1,291
stock trade days were made by 527 unique CEOs, and 90% of the CEOs had 10 or less trade days in our sample. In untabulated results, we examine the temporal distribution of grants by day of the week and month, as well as month of the
year and find no apparent clustering.
We show the temporal distribution of stock splits in Panel A of Table 1, and as can be seen from the table, the number of
splits is reasonably well distributed over time. For the option grant sample and stock trade sample, the average per year is
16.2 and 28 splits, respectively. Note that we do not have option grant data prior to 1996, so the option grant sample only
spans 17 years compared to the 21 years for the stock trade sample. In addition, the temporal distribution of the splits in
both samples is similar to the distribution of all CRSP stock splits.
10
CRSP variable “facpr” between 0.5 and 2. Our results are qualitatively similar if we use only 2:1 splits.
Cleanse code R is “data verified through the cleansing process,” H is “cleansed with a very high level of confidence,” C is “a record added to nonderivative table or derivative table in order to correspond with a record on the opposing table.” L is “one or more cleansing actions were undertaken but
secondary sources were unavailable for complete verification”. The W and Y codes both refer to issues regarding holdings. Given that our study is focused
on transactions and not the options held, excluding these codes would reduce our sample without providing increased accuracy. Our main results are not
qualitatively affected by adding these codes, but our sample is larger.
12
In order to maximize the sample size of each of our two datasets, the samples are necessarily composed of different sets of firms.
11
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
23
Table 1
Sample distribution and selected firm and split characteristics.
Panel A: Splits by year
Year
Option grant sample
—
—
—
—
17
36
29
36
25
16
10
17
28
22
20
8
3
3
0
2
4
276
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
Total
Stock trade sample
All CRSP splits
3
4
0
8
38
48
48
66
80
28
35
36
44
38
34
30
7
2
13
13
12
587
400
472
313
436
516
509
438
420
374
166
181
219
280
296
221
151
62
13
60
70
60
5,657
Panel B: Selected firm and split characteristics
Splits with grants
Sales (millions)
Total assets (millions)
Market-to-Book
Price
ROA
$3,417
$7,799
3.36
$53.36
15.43%
Splits with stock trades
$3,233
$5,475
4.49
$63.18
7.67%
Proportion listed on:
NYSE %
AMEX %
NASDAQ %
46.4%
4.4%
49.3%
40.9%
1.6%
57.5%
Proportion of split factors (facpr)
0.5
1
2
31.9%
61.2%
6.9%
38.3
59.1
2.6
This table presents the number of splitting firms that engaged in 3:2, 2:1, or 3:1 stock splits, by year (Panel A). In Panel B we show selected firm and split
characteristics.
From Panel B of Table 1, splitting firms with grants in the 21-day window, on average, had sales (Compustat item SALE) of
$3.4 billion, total assets (AT) of $7.8 billion, and a market-to-book ratio of 3.4 ((AT CEQþPRCC_F n CSHO)/AT). Prior to the
split announcement, firms in the stock trade sample, on average, had sales of $3.2 billion, total assets of $5.5 billion and a
market-to-book ratio of 4.5. Share price (PRCC_F) and ROA (OIBDP/AT) were $53.4 and 15.4% for the grant sample and $63.2
and 7.7% for the stock trade sample, respectively. Finally, slightly more of both the grant- and trade-sample firms were listed
on the NASDAQ (49.3% and 57.5%, respectively) rather than the NYSE (46.4% and 40.9%, respectively). About 60% of the splits
were 2-for-1, about 35% were 3-for-2, and the residual were 3-for-1 splits.
4.2. Price effects around splits and grants
Our hypotheses depend on whether splits lead to positive returns around the announcement date. To confirm the return
effect, we calculate mean daily returns around the split announcement day for the option grant sample. In Fig. 1 we plot the
cumulative raw and market-adjusted returns (calculated using CRSP value-weighted returns, including dividends) for our
276 stock split-grant sample firms from day 10 to day þ10, centered on the split announcement day. Consistent with the
prior literature, the figure clearly shows a price run-up prior to the announcement, substantially higher returns on the day
of and two days following the split announcement, followed by a continued drift upward. In Panel A of Table 2 we report the
daily returns (and statistical significance, if any) for each day of the 21-day window. For the grant-sample, the marketadjusted returns on the split announcement day and days þ1 to þ2 (1.89%, 1.30% and 0.59%, respectively) are all
24
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
Fig. 1. Cumulative returns centered on the split announcement. Cumulative returns are calculated using the CRSP value-weighted return (with dividends)
as the market return. The dashed line denotes raw cumulative returns and the solid line denotes market-adjusted cumulative returns. The sample consists
of 276 stock splits.
significantly different from zero at the 1% level. A similar pattern holds for the trade-sample, with market-adjusted returns
equal to 1.62%, 2.10%, and 0.40% for days 0 through þ2. For the trade sample, the market-adjusted returns for all ten days
leading up to the announcement are positive and significantly different from zero. These daily returns are similar in
magnitude to the 4,860 non-sample CRSP splits that were announced during our sample period. Panel B of Table 2 shows
the 2-day (day 0 to þ1) and 3-day (day 1 to þ1) market-adjusted and raw cumulative abnormal returns (CAR) for the
grant-sample. The 3-day market-adjusted mean CAR is 3.14% and is significantly different from zero at the 1% level. A similar
result holds for the 2-day window. These figures are slightly lower than reported by Ikenberry et al. (1996), who find a mean
of 3.38% for 1,275 two-for-one stock splits initiated by NYSE and Amex firms in the years during 1975–1990.13 However, they
report a declining trend in announcement returns and we include 3:2 splits in our sample which may lower our returns.14
It is also noteworthy that 70% (80%) of sample firms experienced 3-day (2-day) non-negative returns. Panel C shows the
CARs for the stock trade sample. The 3-day (2-day) CAR is 4.29% (3.71%) and both are significantly different from zero at the
1% level. Seventy-five percent of both the 3-day and 2-day CARs are non-negative.
5. Results
5.1. Main results
Hypothesis 1 states that option grants to the CEO are more likely to occur before a split to take advantage of the stock
price appreciation. To test this assertion we examine the number of grants made during a 21-day window centered on the
split announcement.15 Unlike many other announcements made by firms (e.g., earnings and dividends), a split announcement is unscheduled. As a result, there should be no difference in the relative timing of scheduled or unscheduled option
grants around the split announcement. However, we present each separately (scheduled, unscheduled, and unclassified
grants) as verification. In all four plots from Fig. 2, the number of grants that occurred before the split announcement is
substantially greater than the number of grants that occurred after the announcement. In the plot of all grants, the number
of grants on days 2, 1 and 0 is considerably higher than for any other days during the 21-day period. Just over 57% (166
grants) of the grants in our sample occurred during the 2 days prior to and the day of the split announcement.
13
Lin et al. (2009) also report an announcement return of over 3% for the period of 1975–2004.
We find an abnormal return around the split date ( 1, 1) of 1.6% for 3:2 splits and an abnormal return of 3.8% for all other splits in our sample.
15
We include grants made on day 0, the day of the stock split announcement, in the “before announcement” period. Because executive stock options
are typically granted with an exercise price equal to the closing price on the day of the grant, it is unclear whether the day 0 grants benefit from the entire
price rise resulting from the split announcement (this is most likely the case when the stock split is announced prior to the close of the market). However,
stock split announcements are not always made before the close of the market. When they are announced after the close, the closing price should not
reflect the stock split announcement. Seventy-three of the grants in our sample occurred on the same day (i.e., day 0) as the split announcement. To
determine whether the stock split is announced before or after the close of the market, we use the Dow Jones' Factiva database to obtain the exact time
stamp of the first mention of the stock split. Forty of the 73 splits (55%) are announced after the market close. In those cases, the price change related to the
split announcement actually occurred on day þ 1. Twenty-nine splits (40%) are announced during the day, so the price change resulting from the split
announcement was impounded in the day 0 closing price. For these 29 grants, the exercise price was set after the split announcement. However, even
though these 29 grants did not capture the day 0 split announcement return, the average day þ 1 abnormal return is also statistically and economically
significant (about 1.14%). [Note: no time stamps were available for 4 of the splits.] Notwithstanding this discussion, even when the day 0 observations are
removed from the sample, the results remain qualitatively unchanged. Using Fig. 2, this result can be confirmed visually by simply ignoring the day zero
grants.
14
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
25
Table 2
Split date announcement returns.
Panel A. Daily returns for the 21-day window surrounding the stock split announcement
Relative
day
10
9
8
7
6
5
4
3
2
1
0
1
2
3
4
5
6
7
8
9
10
Option grant sample N ¼276
Stock trade sample N ¼587
Mean market adjusted
returns
Mean market adjusted
returns
0.13%
0.17%
0.68%nn
0.43%nn
0.20%
0.30%
0.45%
0.51%nn
0.05%
0.05%
1.89%nn
1.30%nn
0.59%nn
0.19%
0.01%
0.38%n
0.28%
0.39%n
0.12%
0.04%
0.32%n
Mean raw
returns
0.50%nn
0.49%nn
0.85%nn
0.82%nn
0.56%nn
0.58%nn
0.41%nn
0.65%nn
0.48%nn
0.55%nn
1.62%nn
2.10%nn
0.40%nn
0.02%
0.20%
0.23%
0.18%
0.11%
0.24%n
0.13%
0.21%
0.03%
0.35%n
0.68%nn
0.49%nn
0.26%
0.31%
0.44%
0.51%nn
0.06%
0.00%
2.02%nn
1.29%nn
0.60%nn
0.08%
0.03%
0.47%nn
0.28%
0.47%n
0.16%
0.05%
0.40%n
All CRSP splits (non-granting and non-trades)
N ¼ 4,860
Mean raw
returns
Mean market adjusted
returns
Mean raw
returns
0.48%nn
0.64%nn
0.97%nn
0.88%nn
0.56%nn
0.64%nn
0.47%nn
0.66%nn
0.52%nn
0.56%nn
1.71%nn
2.16%nn
0.46%nn
0.09%
0.23%
0.34%n
0.27%
0.07%
0.30%n
0.14%
0.19%
0.38%nn
0.35%nn
0.34%nn
0.27%nn
0.37%nn
0.28%nn
0.35%nn
0.37%nn
0.33%nn
0.38%nn
1.70%nn
1.18%nn
0.44%nn
0.25%nn
0.12%nn
0.17%nn
0.14%nn
0.13%nn
0.14%nn
0.12%nn
0.07%
0.43%nn
0.43%nn
0.43%nn
0.34%nn
0.42%nn
0.37%nn
0.40%nn
0.43%nn
0.39%nn
0.43%nn
1.75%nn
1.19%nn
0.47%nn
0.28%nn
0.17%nn
0.21%nn
0.17%nn
0.18%nn
0.19%nn
0.15%nn
0.12%nn
Panel B. Option grant sample: cumulative stock split announcement returns
Market adjusted returns
Relative
day
(0, þ1)
( 1,þ 1)
Mean
3.19%nn
3.14%nn
Median
Raw returns
Percent 40
2.07%nn
2.09%nn
80.07%
69.93%
Mean
Median
Percent40
3.31%nn
3.31%nn
2.54%nn
2.36%nn
75.73%
73.19%
Panel C. Stock trade sample: cumulative stock split announcement returns
Market adjusted returns
Relative
day
(0, þ1)
( 1,þ 1)
Mean
3.71%nn
4.29%nn
Median
2.54%nn
2.86%nn
Raw returns
Percent 40
75%
75%
Mean
Median
Percent40
3.86%nn
4.46%nn
2.63%nn
2.89%nn
75%
76%
Market-adjusted returns are raw returns minus the CRSP value-weighted index. ** and * indicate significantly different from zero at the 1% and 5% level,
respectively.
Panel A of Table 3 presents the data in tabular form. There are a total of 290 grants issued during the event window.
If there was no timing of option grants with respect to the split announcement (day 0), we expect to observe approximately
152 grants during the preceding 10 days and the day of the split announcement (this 11 day window is about 52.4% of the 21
day period), with the balance of grants occurring after the announcement. However, it is clear from the data that there were
substantially more grants before the announcement (231 or 80%) than after (59 or 20%). The difference in these proportions
(from 52.4% and 47.6%) is significant at the 1% level (using a one-sided binomial test with Clopper-Pearson Exact estimation
of the confidence levels). There were 65, 76, and 90 (accounting for 86%, 75%, and 80%, of grants), scheduled, unscheduled,
and unclassified grants, respectively, before the split announcement. The differences between the pre- and post-split
proportions from their hypothesized proportions, for all three types of grants are statistically significant at the 1% level.
In addition, the average number of grants per day is 21.0 before and on the split announcement day and only 5.9 for
subsequent days. In untabulated robustness checks we calculate the number and percentage of grants that occurred on or
before and after the split announcement during a 101-day window.16 Similar to the shorter event window, a greater
16
There were substantially more split announcements when using the longer window. This is because we only include a split announcement in the
sample if the same firm also grants an option to its CEO within the specified event window. These results are available upon request.
26
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
Fig. 2. Option grants surrounding stock splits. Daily scheduled, unscheduled, unclassified, and all option grants during the 21-day window centered on a
stock split announcement.
proportion of grants occurred before the split announcement (about 61% of all grants). The pre- and post-split proportions
differ significantly from their hypothesized proportions. Further, the average number of grants per day is 11.2 before the split
announcement and 7.3 after. These results are consistent with Hypothesis 1.
Panel B of Table 3 uses a share-weighted measure. On a share-weighted basis, 80% (significantly different from 52.4% at
the 1% level) of the grants occurred on or before the day of the split announcement.17 To ensure that our findings are not a
statistical or temporal aberration, we investigate granting behavior of the splitting firms during the 21 trading days exactly
one year prior to the stock split. These results are reported in Panel C and show no statistical differences between the preand post-split periods, relative to their hypothesized proportions of 52.4% and 47.6%. For the overall sample, there is an equal
number of grants in the pre- and post-split periods. A similar pattern exists for the three different types of grants.18
A broader question is whether the number of option grants was unusually high around stock split announcements?
To answer this question, we attempt to develop a better understanding of typical granting activity. At the simplest level, one
can observe the distribution of grants relative to split announcements. Since we use a 21 trading-day window, in a given
year of 251 trading days we are capturing 21/251¼8.4% of all trading days. We would therefore expect to see at least 8.4% of
grants occurring in our sample of splits. Our sample begins with 5,657 splits, however only 4,036 are in the 1996–2012
period that overlaps the data availability for the option grant data. Of these 4,036 splits, only 1,336 occurred in years where
the firm also had an option grant. In other words, it seems reasonable to assume that 1,336 splitting firms also regularly
grant options. As our basic split data set has 276 split events that also had grants, we are observing 276/1336¼20.6%.
Therefore the occurrence of grants around splits is nearly 2.5 times our unconditional expectation.
17
The Thomson Reuter Insider Filing Data Feed reliably documents the information contained on Form 4. During a portion of our sample period,
insiders were allowed up to about 40 days to report a transaction. As a result, some of the reports in our sample were filed after the payment date of the
stock split. Such reports are stated in post-split shares and exercise price (e.g., a grant of 100,000 shares at an exercise price of $50, filed after the pay date
of a 2-for-1 stock split, would be reported on Form 4 as a grant of 200,000 shares at an exercise price of $25). In order to avoid any misstatements of shares
granted, we analyze any grants where the exercise price differs from the closing price on the day of the grant by more than $3. We utilize primary source
documents (proxy statements and Form 4s from EDGAR or the Ownership Reporting System) to determine the correct number of shares that were granted
on the actual day of the grant, so that all shares used in our computations are on a pre-pay date basis.
18
It is important to note that the grants in Panel C are classified independently of the grants in Panel A. For example, a grant in the year prior to the
split is the basis for labeling a grant in the split year as either scheduled or unscheduled (but not unclassified). If we classify prior year grants based on the
split year classification our results do not qualitatively change.
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
27
Table 3
Option grants by type during the 21-day window centered on a stock split announcement.
Panel A: Number of grants
Pre-split (including day 0)
Scheduled
Unscheduled
Unclassified
All grants
Grants per day
Post-split
Total
Number
Percent
Number
Percent
76
102
112
290
65
76
90
231
21.0
86%nn
75%nn
80%nn
80%nn
11
26
22
59
5.9
14%nn
25%nn
20%nn
20%nn
Panel B: Total number shares underlying options granted
Pre-split (including day 0)
Total
Scheduled
Unscheduled
Unclassified
All grants
6,184,157
16,581,182
19,512,298
42,277,638
Number
5,348,523
14,163,243
14,395,723
33,907,489
Post-split
Percent
nn
86%
85%nn
74%nn
80%nn
Number
Percent
14%nn
15%nn
26%nn
20%nn
835,634
2,417,940
5,116,575
8,370,149
Panel C: Prior year number of grants
Pre-split (including day 0)
Total
Scheduled
Unscheduled
Unclassified
All grants
Grants per day
34
45
43
122
Number
20
18
23
61
5.5
Post-split
Percent
Number
Percent
59%
40%
47%
50%
14
27
20
61
6.1
41%
60%
53%
50%
CEO stock option grants are classified using the method from Heron and Lie (2007). We use a one-tailed binomial test with the Clopper-Pearson Exact
estimation of the confidence levels to assess the differences between the number of pre- and post-split option grants. Differences denoted with a ** or * are
significant at the 1% or the 5% level, respectively.
5.2. Characteristics of opportunistic and non-opportunistic firms
To investigate whether there are differences between firms that granted options prior to the split date and those that granted
after the split date, we compare firm characteristics and other factors that may have influenced the potential managerial
opportunism. For this analysis we categorize firms that granted options before the split announcement (on day 0 or before) as
being opportunistic and the others as non-opportunistic. Table 4 presents the univariate statistics for both groups of firms.
While the non-opportunistic firms on average were nominally larger, in terms of sales ($4.1 billion versus $3.3 billion)
they were smaller in terms of total assets ($6.2 billion versus $8.1 billion). They were also marginally less profitable in terms
of return on assets (15.1% versus 15.5%) although none of these differences are statistically significant. The differences in the
average number of institutions who own the stock (206 for non-opportunistic and 193 for opportunistic firms) as well as the
number of analysts that follow the firm (10.0 for non-opportunistic and 9.8 for opportunistic firms) are also not statistically
different. However, percent institutional ownership was significantly higher for opportunistic firms compared to nonopportunistic firms (64.9% vs. 55.2%). There is no statistical difference in the percent ownership of insiders and directors.
Also, when we analyze the number of (total, insider, and outsider) block holders (minimum of 5%) and the total percent
ownership of these block holders we find no significant differences.19
To further compare the strength of governance between opportunistic and non-opportunistic firms, we compute
the Gompers et al. (2003) G-Index and the Bebchuk et al. (2009) BCF-Index. The G-Index is defined as the sum of 24
anti-takeover provisions. A higher number represents weaker corporate governance, that is, less shareholder rights and
more CEO power. The BCF-Index measures entrenchment and consists of the six most effective anti-takeover measures.
Comparison of these indices for the two sub-samples reveals no significant differences. Finally, we compare CEO duality
between the two sub samples. CEO duality is a binary variable that indicates whether the CEO is also the Chair of the BOD.
Again, we find no significant difference. On the whole, we find little evidence that the opportunistic firms were different
from the non-opportunistic firms.
19
Institutional ownership data is from the Thomson Reuters 13F database. Number of analysts is from I/B/E/S and is the number of analysts making a
one-year ahead earnings forecast for the firm. Insider and director ownership and the block holder data are collected from 10-Ks Governance data is from
IRRC Risk Metrics.
28
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
Table 4
Characteristics of opportunistic and non-opportunistic firms.
Opportunistic
Sales (million)
Total assets (million)
Market-to-book
ROA
Share price
Number of institutions holding stock
Institutional ownership
Number of analysts
Insider ownership
Number of 5% owners
Blockholder (5%) ownership
Number of inside blockholders (5%)
Insider blockholder ownership
Number of outside blockholders (5%)
Outside blockholder ownership
BCF Index
G-index
Duality
Non-opportunistic
Mean
(Median)
N
$3,275
($609)
$8,140
($736)
3.36
(2.09)
15.49%
(15.62%)
$54.33
($48.03)
193
(142
64.91%
(70.22%)
9.82
(8.00)
12.42%
(6.88%)
2.62
(3.00)
31.88%
(25.93%)
2.49
(0.00)
18.49%
(12.20%)
2.13
(2.00)
27.81%
(22.18%)
1.82
(2.00)
8.92
(9.00)
0.65
(1.00)
226
226
225
223
226
203
183
199
201
201
188
201
76
201
165
55
116
119
Mean
(Median)
$4,081
($353)
$6,196
($756)
3.36
(2.36)
15.12%
(14.61%)
$48.80
($39.75)
206
(288)
55.16%
(58.25%)
9.95
(8.00)
14.32%
(7.30%)
2.47
(2.00)
30.36%
(24.00%)
0.72
(0.00)
17.41%
(13.29%)
1.74
(1.00)
25.56%
(21.40%)
1.67
(2.00)
9.86
(9.50)
0.76
(1.00)
Diff.
N
T
(Z)
48
–0.64
(0.33)
0.53
( 0.04)
0.00
( 0.37)
0.23
(0.72)
1.18
(1.29)
0.30
( 1.04)
2.21n
(2.05n)
0.11
( 0.21)
0.76
( 0.75)
0.52
(0.96)
0.32
(0.02)
1.41
(1.16)
0.35
( 0.61)
1.45
(1.77)
0.46
(0.02)
0.25
(0.01)
1.33
( 1.44)
1.45
( 1.44)
48
48
48
48
45
41
43
43
43
37
43
19
43
31
6
16
43
We define opportunistic firms as those where the CEO received option grants prior to the split. Non-opportunistic firms are those where the option grant
occurred after the split. The BCF Index is the Bebchuk et al. (2009) governance index. The G-Index is the Gompers et al. (2003) governance index. CEO
duality is a dummy variable with a value of 1 if the CEO is also the Chair of the BOD. We use a two-sided t-test (z-test) to ascertain any difference between
the two groups. All differences denoted with a ** or * are significant at the 1% or the 5% level, respectively.
5.3. The legality of timing options around splits
Our evidence is consistent with the hypothesis that some option grants to CEOs appear to be timed relative to stock splits
and that this timing can benefit the CEO. However, the legality of this behavior is a different question. Timing of option
grants around a stock split announcement could fall into one of two broad practices, backdating and spring-loading, that
have received considerable attention from the courts and legal scholars.
5.3.1. Backdating
The practice of option grant backdating is well documented in the literature, since the original work of Lie (2005) and
Heron and Lie (2007). Briefly, backdating involves retroactively setting the grant date of an employee stock option to
coincide with a day (usually in the most recent quarter) when the firm's stock price was low. By granting the option on a day
with a low price, (and setting the exercise price at that day's closing market price) the option is more likely to be in-themoney at the time of vesting.
Backdating by itself is not illegal. However, legal problems can occur if the grant is backdated and if the firm fails to fully
disclose the tax or accounting implications of the grant.20
Lie (2005) and Heron and Lie (2007) document the existence of backdating by showing that the observed incidence of
option grants on a day with a locally low price far exceeds the unconditional probability of that number of grants occurring
20
The legality of backdating was clarified by Judge William Alsup in the US District Court of Northern California Federal Court on April 11, 2007.
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
29
on that day. They argue that the only feasible way to explain their empirical evidence is for the firm to have backdated the
grant date at the end of the quarter. In our data set, repeating such a test is not as straightforward because of the
complicating effect of the stock split. Typically, stock splits are preceded by a sustained run up in the stock price and as a
result, the lowest price of the quarter (or any shorter time period) is likely to be at the start of the period under examination.
In untabulated results, we confirm such a run-up in our sample.
An alternative method to test for the possibility of backdating is to look for differential behavior by firms whose split
announcement results in a positive price reaction compared to those with a negative price reaction. If CEOs were
backdating, it makes more sense to backdate for splits with a positive price reaction than for those where the price declined
(which represent about 25% of our sample – see Table 2, Panel B, [ 1, þ1] day market-adjusted return). We examine the
distribution of grants before and after the introduction of the Sarbanes Oxley Act of 2002 (SOX). Prior to SOX, option grants
had to be reported no later than the 10th day of the month following the grant (this allowed for a window of up to 40 days).
After the introduction of SOX on April 29, 2002, a grant was required to be reported by the end of the second business day
following the grant day. This narrower reporting window limits the ability of the firm to backdate.
In Table 5 we divide the sample based upon the direction of the price reaction to the split announcement and whether or
not the split took place before or after SOX. Our discussion will focus on unscheduled and unclassified grants because these
are most likely to be subject to backdating. If backdating was present and if SOX limited backdating, we would expect to see
more grants prior to splits with positive abnormal returns than prior to splits with negative abnormal returns. Further we
would expect that this difference is more pronounced before SOX than after SOX. In Panel A, we observe that 76% of
unscheduled and 83% of unclassified grants occurred before the split announcement in the pre-SOX period when the split
price reaction was positive. In Panel B, (pre-SOX, negative market response), we observe slightly fewer unscheduled and
unclassified options being granted pre-split (75% and 65%, respectively). These results, while in the direction predicted, are
not significantly different from those in Panel A.
In Panels C and D we examine the grants around positive and negative market responses in the post SOX period. In Panel
C (post-SOX, positive market response), we observe that 68% of unscheduled and 88% of unclassified grants occurred before
the split. In Panel D, (post-SOX, negative market response), these proportions are 82% and 70%, respectively. No clear pattern
emerges between Panels C and D. Furthermore, the introduction of SOX appears to have had little impact on the behavior of
our sample firms. Four possible conclusions can be drawn. First, there may have been little or no backdating in the sample.
Second, although SOX requires reporting within the second business day, in theory, this still provides the opportunity to
backdate by two days. Third, because SOX reduced other areas of manipulation, it might be that post-SOX, backdating by a
couple of days was one of the few measures left available to firms. And finally, by dividing our sample into four subsets, we
have reduced the power of our tests sufficiently so that we cannot detect any significant effect. However, it seems reasonable
to assume that if there was any backdating, it was fairly limited in scope.
5.3.2. Spring-loading
Spring-loading occurs when an option grant is knowingly timed to occur prior to the release of good news that will, in all
likelihood, result in a positive stock price reaction (Bebchuk et al. (2010) test for this practice more broadly). If only the CEO
is in possession of this material information, but not the BOD, a case for illegal insider trading may be made, although courts
have been reluctant to apply insider trading laws to option grants (see Anabtawi (2004) and Sen (2009)). More generally the
message is mixed. In a rigorous review of case and regulatory law, Tompkins (2008) argues that “the practice of springloading stock option grants seems to pass muster for both regulatory and interstitial bodies of law…” In 2006, the SEC
promulgated a new rule designed to reduce the information asymmetry surrounding executive compensation. In part, this
new disclosure rule requires that firms reveal any coordination between executive option grants and any “material nonpublic information that is likely to result in an increase in its stock price, such as immediately prior to a significant positive
earnings or product development announcement”.21 At the same time, SEC Commissioner, Paul Atkins stated that in
approving a grant, directors could adjust the number of options granted to yield the same economic effect after taking
account of the price reaction.22 While such an adjustment seems unlikely in the case of stock splits, given the less than
certain price reaction, we will discuss the broader possibility of substituting opportunistically timed option grants for other
compensation in the next section.
To conclude, based on the analysis and discussion above, while the evidence we find is consistent with opportunistic
behavior of the CEO, without the presence of clear insider trading intent by the CEO, it is very unlikely that option grant
timing around stock splits would be considered illegal.
5.4. Are opportunistically timed options a substitute for other compensation?
An argument could be made that opportunistically timed option grants simply represent part of the overall compensation package of the CEO and that a rationally acting BOD would adjust the CEO's other compensation to offset the gains
21
SEC final rule release number 33-8732A, Executive Compensation and Related Person Disclosure (conforming amendments). Available on the SEC's
website at: http://www.sec.gov/rules/final/2006/33-8732a.pdf.
22
http://www.sec.gov/news/speech/2006/spch070606psa.htm/.
30
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
Table 5
Option grants by type and conditioned on the market response to the stock split announcement, before and after SOX.
Panel A: Pre SOX – Positive market response
Pre-split (including day 0)
Scheduled
Unscheduled
Unclassified
All grants
Grants per day
Post-split
Total
Number
Percent
Number
Percent
23
42
59
124
19
32
49
100
9.1
83%nn
76%nn
83%nn
81%nn
4
10
10
24
2.4
17%nn
24%nn
17%nn
24%nn
Panel B: Pre SOX – Negative market response
Pre-split (including day 0)
Total
Scheduled
Unscheduled
Unclassified
All grants
Grants per day
12
24
17
53
Number
Post-split
Percent
nn
11
18
11
40
3.6
92%
75%n
65%
75%nn
Number
Percent
1
6
6
13
1.0
8%nn
25%n
35%
25%nn
Panel C: Post SOX – Positive market response
Pre-split (including day 0)
Total
Scheduled
Unscheduled
Unclassified
All grants
Grants per day
30
25
26
81
Number
26
17
23
66
6.0
Percent
87%nn
68%nn
88%nn
81%nn
Post-split
Number
Percent
4
8
3
15
1.5
13%nn
32%nn
12%nn
19%nn
Panel D: Post SOX – Negative market response
Pre-split (including day 0)
Scheduled
Unscheduled
Unclassified
All grants
Grants per day
Post-split
Total
Number
Percent
Number
Percent
11
11
10
32
9
9
7
25
2.3
82%n
82%n
70%
78%nn
2
2
3
7
0.7
18%n
18%n
30%
22%nn
CEO stock option grants are classified using the method of Heron and Lie (2007). We use a one-tailed binomial test with the Clopper-Pearson Exact
estimation of the confidence levels to assess the differences between the number of pre- and post-split option grants. All differences denoted with a ** or *
are significant at the 1% or the 5% level, respectively.
related to the split announcement. The BOD would not have to forecast the price reaction of the split, instead it could simply
wait until after the split, observe the price reaction, and then adjust the remaining CEO compensation package. To explore
this possibility, we use the ExecuComp database to gather data on the total compensation for CEOs of splitting and granting
firms. Because ExecuComp only covers the S&P 1500, we only obtain data for 110 of our sample firms. We examine the “Total
Compensation” variable, which is defined as the sum of Salary, Bonus, Other Annual Compensation, Restricted Stock Grants,
Long term incentive plans, Payouts, Other, and the Value of Option Grants. In the year prior to the split, the mean (median)
Total Compensation was $3.782 million ($1.938 million). In the year of the split, Total Compensation is nearly double that of
the prior year, at $7.501 million ($3.451 million) and in the year following the split, Total Compensation fell to $6.049 million
($3.324 million). Clearly, in aggregate, the total compensation increased substantially in the split year over the pre-split year
and remained high in the post-split year. This result is not particularly surprising, since splits are typically preceded by a
price run-up, presumably the result of improved firm performance. In the split year, the CEO benefits from the
opportunistically timed options, but also from any other elements of his/her incentive package that are related to the
stock price performance.
Based on the testimony of SEC commissioner Paul Atkins, the BOD could offset the impact of the price reaction to the
stock split by reducing the number of options granted. Such an adjustment would require that the BOD be able to anticipate
the stock price rise resulting from the split announcement. Given that only 75% of splits result in an actual price increase,
this forecast would clearly be made with non-trivial error. However, we can create a very rough estimate of whether this
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
31
occurred by examining whether firms changed the number of options granted during the split year, relative to the previous
and subsequent years.23 For firms with grants in all three years, 56% either increased the options granted or made no change
in the split year, relative to the number of options granted in the previous year. Compared to the subsequent year, the
number of options granted in the split year was either higher or the same for 53% of the firms.24 For the majority of firms, it
does not appear that the number of options granted during the split year was decreased to offset the potential gain resulting
from the split announcement.
Because we include all ESOs granted in a given year, it is possible that we are capturing some grants awarded for events
unrelated to the stock split. To reduce the potential that this affects our results, we limit the window to just the days in the
pre-split period (i.e., days 10 to 0; see Panel B of Table 3) and compare these to any shares granted in the same period,
one year earlier (i.e., Panel C of Table 3). In the stock split year, we find an average of 146,786 (33,907,489/231) ESOs per
CEO-grant. If this number were adjusted downward, in anticipation of the likely positive stock split announcement return,
one would expect this average to be low relative to the prior year. However, when we calculate the number of shares
granted per CEO-grant for that period, the average is 101,838 (6,212,104/61) ESOs.
The collective evidence presented above does not seem to be consistent with the conjecture that the BOD was actively
reducing the number of options granted to offset the split-associated price gains.
5.5. Insider trading
So far our tests have focused on the timing of the decision to split and grant stock options – two decisions that may be
initiated and implemented by the BOD without any managerial input. Although, both decisions are likely made in
consultation with the CEO. As a means of more directly determining CEO intent, we now study CEO stock trading activity
around stock split announcements.
Hypothesis 2 states that there will be discernible patterns in CEO stock trading behavior. Since on average, stock
split announcements are associated with positive stock price reactions, we expect CEOs to buy (sell) their firm's stock
before (after) the stock split. Similar to our earlier analyses we investigate the 21-day window surrounding stock split
announcements.
To test Hypothesis 2, we focus only on open market stock trades. We exclude any trades that are made as a result of
exercising an option. In Fig. 3 we plot CEO trades aggregated by the day relative to the split day. We group trades into buys
and sells. Consistent with Hypothesis 2, Fig. 3 shows a large increase in selling activity on the days immediately following
the split announcement. The figure also indicates a smaller clustering of buying prior to the split announcement, although
the magnitude of the buying activity is very small relative to the selling.
In Table 6 we present the same data in tabular form. For consistency, we follow the format of Table 3. Under Hypothesis 2,
we expect to see more selling activity after the split and more buying activity prior to the split. Panel A of Table 6 shows that
a total of 1,291 stock trades occurred during the 21 days around the stock split announcement. Of these, 1,222 were sell
transactions, while 69 were buys (recall that we aggregate all trades made by a CEO on a single day into a single trade). On
the whole, CEOs sold more during the post-split period (65% of the total) compared to the pre-split period (35% of the total).
This difference is statistically and economically significant. In dollar terms, about $3.3 billion of stock was sold after the split
compared to about $2.1 billion sold before the split. These results are consistent with our hypothesis that CEOs delayed
trading until after the split.
We see a similar pattern with purchases. Of the 69 stock buys, 64% occurred before the split while 36% were post-split.
Again these results are consistent with CEOs buying at the pre-split price which 75% of the time is lower than the post-split
price. For both sales and purchases, these patterns are consistent with opportunistic behavior on the part of CEOs.
The pattern of trades in Panel A could be related to the specific time of year rather than the split announcement. In an
effort to control for this possibility, in Panel B of Table 6 we analyze CEO trading activity one year earlier. Panel B contributes
two important results to our analysis. First, the total number of trades in the prior year was less than one-third of that
traded around the split announcement (reported in Panel A). Second, neither buys nor sells exhibit a significant difference
between the pre- and post-split proportions (to make this distinction, we use the split announcement date minus one year).
Thus, at least when compared to the prior year, the trading around the split announcement appears unusual. This evidence
is consistent with our earlier conclusion that CEOs appear to make opportunistic trades around the announcement of stock
splits.
5.5.1. Insider trading: multiple split announcements
Some of the firms in the insider-trading sample had more than one split announcement. In the cases where the CEO
remained the same, we look at the trading behavior of these CEOs around these multiple splits.25 The distribution of split
frequencies by CEO is as follows: Of the 527 unique CEOs in the sample, 47 were CEOs of firms that split twice, 6 were CEOs
of firms that split 3 times, and 1 was a CEO of a firm that split 4 times. The remaining 471 CEOs presided over only 1 split,
23
On a calendar year basis, we aggregate all grants during the year of the split, the previous year, and the subsequent year. This sample is comprised of
147 firms that had non-zero grants at anytime during each of all three years.
24
The mean (median) number of shares granted in year 1, 0, and þ1, are 85,895 (30,000), 128,554 (35,000), and 160,998 (40,000), respectively.
25
Theoretically, we could perform a similar analysis for the option grant sample; however there are only 26 splits with available data in this sample.
32
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
Fig. 3. Daily number of stock trades around split announcement. This figure plots the daily number of non-derivative related stock trades (buy and sell)
around the stock split date. Multiple trades by a CEO on a single day are treated as a single sale or buy. All dates are aligned relative to the split date.
Table 6
Stock trades during the 21-day window centered on a stock split announcement.
Panel A: Number of trades
Pre-split (including day 0)
Sells
Buys
Total
Number
1,222
69
425
44
Post-split
Percent
Number
35%nn
64%nn
Percent
65%nn
36%n
797
25
Panel B: Prior year number of trades
Pre-split (including day 0)
Sells
Buys
Post-split
Total
Number
Percent
Number
278
19
133
10
48%
53%
145
9
Percent
52%
47%
CEO trades are aggregated on a daily basis by as buys or sells for all trades that are not the result of a derivative transaction (such as an option exercise). We
use a one-tailed binomial test with the Clopper-Pearson Exact estimation of the confidence levels to assess the differences between the number of pre- and
post-split option grants. All differences denoted with a ** or * are significant at the 1% or the 5% level, respectively.
resulting in the total of 587 splits.26 For CEOs that repeatedly split the stock of their firm, the proportion of sales that
occurred after the split around these repeated splits was 52% (not statistically significant different from the sales before the
split). If we look at the behavior of these CEOs in their first split in our sample, we find that 61% sold after the split
(significantly different from the proportion of sales after the split, at the 5% level). If we compare this frequency with the
overall sample average (in Table 6) of 64% sold after the split, it does not appear that managers that engage in multiple splits
are more opportunistic than the overall sample average.27
5.5.2. Insider trading: tax based
Since many of the grants in our sample are scheduled, it is likely that option grants from prior years may vest near the
time of the split announcement. Following Dhaliwal et al. (2009), we examine the pattern of derivative related trades
around the split announcement. A CEO that wished to “exercise and sell” her vested options would likely do so after the split
announcement, in order to take advantage of the expected price rise as a result of the split.28 Likewise, a CEO who wanted to
26
There are no cases in our sample where a CEO presides over splits in two or more distinct firms (i.e., the CEO moves).
A related question is whether the observed opportunism is a CEO or firm effect. Ideally, to examine this question we would need multiple cases
where a CEO presided over multiple splits in different companies. Such instances are quite rare and prevent us from robustly answering this question.
However, our evidence regarding stock trading behavior indicates that CEOs are certainly very aware of the benefits of timing trades around splits.
28
We thank Terry Shevlin, the referee, for suggesting this line of inquiry.
27
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
33
“exercise and hold” would likely do so before the split announcement, so as to minimize the gain that would be taxed at her
ordinary tax rate.
To examine tax-based exercise behavior around splits, we closely follow the method of Dhaliwal et al. (2009) and capture
CEO option exercises from SEC Form 4 (designated with the transaction code “M”). We group multiple exercises on a single
day into a single observation. Using the same data, we identify stock sales (designated with transaction codes “D”, “F”,
and “S”). We classify an exercise as an “exercise and sell” if an exercise is followed by a code “D”, “F”, or “S” sale within
30 days of the exercise, otherwise it is classified as an “exercise and hold.” Table 7 presents the frequency of the “exercise
and sell” as well as the “exercise and hold” transactions vis-à-vis the stock split announcement. We also present these trades
graphically in Fig. 4.
Out of 1,436 “exercise and sells”, we find that 1,406 of the sells occurred on the day of or day after the exercise. Therefore
we restrict our analysis to these immediate “exercise and sells”, given that an exercise before the split followed by a sell after
would confound our analysis. Like Dhaliwal et al., we exclude transactions where the CEO exercised options on multiple
days of a given month. Because we are using a month as the time range, we examine the window of 30 days before and 30
days after the split.
Of the 1,406 “exercise and sell” transactions that occurred near the split and were concluded within a day, about 59%
(826) occurred after the split announcement compared to 41% (580) before. Both these percentages are statistically different
from their hypothesized values. This result appears to be consistent with our expectation. It is possible that a majority of the
options do not vest until after the split announcement and thus the trades may not be opportunistic. However, we do not
have sufficient information to test this premise. There are only 275 “exercise and hold” trades, as defined by Dhaliwal et al.,
in our dataset. There is no significant difference between the 131 that occur before the split announcement and 144 that
occur after.
We conclude that there is little evidence consistent with CEOs opportunistically exercising stock options after the split to
minimize their tax liabilities.
6. Economic significance
The evidence presented above suggests that option grants are timed vis-à-vis stock split announcements. To estimate a
CEO's economic gain, we only consider the grants awarded prior to or on the day of the stock split announcement.
We calculate the value of each option on day 1, relative to the stock split, and then multiply that value by the number of
shares granted to determine the aggregate value of the ESO.29 For the 73 options granted on day zero, we calculate the value
of the ESO on day 0 rather than day 1 (i.e., the day of the stock split announcement).30
In a similar way, we calculate the value of each ESO using the stock price on day þ1 relative to the split announcement.
The difference between the two values is the net gain or loss to the CEO as a result of the stock split announcement.
On average, the gain per CEO-grant is $451,748 (based on aggregating all the grants to a single CEO on a single day).
The median gain is $39,234. These estimates are based upon only the grants that were made on or before the stock split
announcement day. However, because approximately a quarter of the raw announcement returns are non-positive, a quarter
of the CEOs experience either no gain or a loss (and are included in the above estimates). Clearly, these amounts are nontrivial. Aboody and Kasznik (2000) report an average and median gain of $92,500 and $18,500, respectively; Yermack (1997)
reports a mean gain of $30,000 and a median gain of $11,000; and Dhaliwal et al. (2009) report a mean gain of $96,000 and a
median gain of $7,000.
Our economic significance estimates rely on the option being held until maturity. However, if the option is exercised
prior to maturity, our estimates may be biased upward (e.g., Huddart and Lang, 1996). Nearly three quarters of the grants in
our sample (of the 247 observations where vesting dates are available) become vested (i.e., exercisable by the grantee)
within 1–5 years. Therefore, we estimate the economic significance assuming that CEOs hold options for only 1 year and 5
years. The average economic significance, assuming all grants mature in 1 year, is $353,609 (with a median of $29,342).
Assuming a maturity of 5 years for all grants, the average (median) economic gain is $424,804 ($35,993) for 5 years'
maturity.
We also estimate the economic gain to CEOs from their trading activities around the split announcement. As with the
option grants, not all trades appear to be opportunistic. We only account for sales that occurred after, and buys that occurred
before the split announcement as opportunistic. By deferring the sale, the CEO captures the stock price increase associated
with the split in the proceeds of his/her sale. To estimate the economic gain from delaying a trade until after the
announcement,
29
We use the Black Scholes non-dividend option pricing model and assume the following inputs: time to maturity is the time to expiration of the
option, volatility is the annualized standard deviation of returns over the prior 251 trading days, the risk free rate is the constant maturity 5-year Treasury
Rate and the strike price is the exercise price reported in the Insider Filing Data Feed (or as corrected, see Footnote 17 for further discussion). We assume
that the volatility is unaffected by the stock split, however, Ohlson and Penman (1985) show that on average, stock volatility increases after a split.
By ignoring any possible increase in volatility, we are biasing our estimates downward. We combine options granted on the same day and estimate
weighted average exercise prices and maturities.
30
In the primary analysis, these 73 grants are categorized as “opportunistic.” They are also included in this analysis. When these grants are excluded,
our results are qualitatively similar. Without the day zero grants, the average gain per CEO-grant is $336,283 and the median gain is $68,948.
34
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
Table 7
Tax-based option exercises during the 30-day window centered on a stock split announcement.
Pre-split (including day 0)
Total
Exercise and sell
Exercise and hold
1,406
275
Number
580
131
Post-split
Percent
nn
41%
48%
Number
Percent
826
144
59%nn
52%
“Exercise and sell” trades are defined as an option exercise followed by a stock sale within 1 day. “Exercise and hold” trades are an option exercise with no
stock sale in the next 30 days. The method of classification closely follows Dhaliwal et al. (2009). We use a one-tailed binomial test with the ClopperPearson Exact estimation of the confidence levels to assess the differences between the number of pre- and post-split option grants. All differences denoted
with a ** or * are significant at the 1% or the 5% level, respectively.
Fig. 4. Daily number of option exercises and sells or holds around split announcement. This figure plots the daily number of option exercises that are
followed by an immediate sale “Exercise and Sell”, or where the shares and held more than 30 days “Exercise and Hold”. Multiple transactions by a single
CEO on a single day are treated as a single trade. All dates are aligned relative to the split date.
we compare the dollar amount of the actual post-announcement sale to a dollar amount after removing the effect of the day
0 to day þ1 stock split announcement return. We find that the average CEO gained $345,613 by delaying a stock sale until
after the split announcement. The median gain was $48,188.
We also examine the effect of a purchase that occurs before the split announcement. To estimate the economic gain,
we take the dollar value of those trades in the 10 days before the split and multiply each by the cumulative stock price
return from day 0 to day þ1. The average gain is $5,279 per split event (with a median of just $1,668). The rather modest
average gain is a reflection of the considerably lower level of buying in general.31
Because the grant and trade samples are not identical (although there is considerable overlap between them), we cannot
simply add together the economic gains. Nevertheless, it is clear that the gains are substantial.
7. Conclusion
We show that firms appear to manipulate the timing of stock split announcements vis-à-vis CEO option grants in order
to increase CEO compensation, potentially at the expense of shareholders. In our sample of 276 stock split announcements
where an option grant occurred within 10 days on either side of the split announcement, the average market-adjusted 3-day
announcement return is 3.14%.
We find evidence consistent with the hypothesis that option grants to CEOs are not randomly distributed around the
time of a split announcement. Since by convention, the exercise price of most ESO grants is set to the closing price on the
day the options are granted, grants made before a split announcement benefit from a lower exercise price relative to any
grants made after the split announcement. We find that 80% of the grants occurred before or on the day of the stock split
announcement. Granting prior to the split results in an average (median) gain per CEO-grant of $451,748 ($39,234).
While we find evidence consistent with the notion of opportunism, we find no compelling evidence for the existence of
backdating as an explanation for the pattern of option grants. Although, we note, that in our sample, typical tests for
31
See Table 6 for total number of sales and purchases.
E. Devos et al. / Journal of Accounting and Economics 60 (2015) 18–35
35
backdating are hampered by the steady run up in stock prices that usually occurs before splits. However, the pattern of
“grant and split” is an example of option spring-loading.
Because the timing of the split announcement and CEO option grant is determined by the Board of Directors (or jointly
determined by the CEO and the board), it is not completely clear whether the CEO is complicit in this decision. We provide
additional evidence consistent with the notion that CEOs are aware of and take actions that benefit them around the split
announcement. Specifically, we examine CEO stock trading activity around the split announcement. For stock sales
(purchases), 65% (64%) occurred after (before) the split announcement. Opportunistic trading results in an average gain to
the CEO of $345,613 for sales (and about $5,279 for purchases). Furthermore, we find little evidence consistent with CEOs
exercising options (granted in previous years) around the split announcement in a manner that minimizes their tax liability.
Although our analyses support the conjecture that the timing of option grants and CEO trading is consistent with
opportunistic behavior, we are unable to directly test this hypothesis. In addition, it is not clear whether shareholders are
necessarily harmed by this apparent option grant timing, as it is possible that this is just another way by which the BOD
attempts to reward and retain a high performing CEO. Finally, we do not test whether the existence of option-based
compensation is a determinant of stock splits. We leave this last question for future work.
References
Aboody, D., Kasznik, R., 2000. CEO stock option awards and the timing of corporate voluntary disclosures. Journal of Accounting and Economics 29 (1),
73–100.
Anabtawi, I., 2004. Secret compensation. North Carolina Law Review 82 (835).
Bebchuk, L., Cohen, A., Ferrell, A., 2009. What matters in corporate governance? Review of Financial Studies 22, 783–827.
Bebchuk, L.A., Grinstein, Y., Peyer, U., 2010. Lucky CEOs and lucky directors. Journal of Finance 65 (6), 2363–2401.
Collins, D.W., Gong, G., Li, H., 2009. Corporate governance and backdating of executive stock options. Contemporary Accounting Research 26 (2), 403–445.
Dhaliwal, D., Erickson, M., Heitzman, S., 2009. Taxes and the backdating of stock option exercise dates. Journal of Accounting and Economics 47 (1–2),
27–49.
Fama, E.F., Fisher, L., Jensen, M.D., Roll, R., 1969. The adjustment of stock prices to new information. International Economic Review 10, 1–21.
Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity prices. The Quarterly Journal of Economics 118, 107–156.
Hall, B., Murphy, K., 2000. Optimal exercise prices for executive stock options. American Economic Review 90 (2), 209–214.
Heron, R.A., Lie, E., 2007. Does backdating explain the stock price pattern around executive stock option grants? Journal of Financial Economics 83 (2),
271–295.
Huddart, S., Lang, M., 1996. Employee stock option exercises an empirical analysis. Journal of Accounting and Economics 21 (1), 5–34.
Ikenberry, D.L., Rankine, G., Stice, E.K., 1996. What do stock splits really signal? Journal of Financial and Quantitative Analysis 31 (3), 357–375.
Lie, E., 2005. On the timing of CEO stock option awards. Management Science 51 (5), 802–812.
Lin, J., Singh, A., Yu, W., 2009. Stock splits, trading continuity, and the cost of equity capital. Journal of Financial Economics 93 (3), 474–489.
Louis, H., Robinson, D., 2005. Do managers credibly use accruals to signal private information? Evidence from the pricing of discretionary accruals around
stock splits. Journal of Accounting and Economics 39 (2), 361–380.
Morgenson, G., 2004. In the timing of options, many, um, coincidences. New York Times, December 5.
Narayanan, M.P., Seyhun, H.N., 2008. The dating game: do managers designate option grant dates to increase their compensation? Review of Financial
Studies 21 (5), 1907–1945.
Nayar, N., Rozeff, M.S., 2001. Record date, when-issued, and ex-date effects in stock splits. Journal of Financial and Quantitative Analysis 36 (1), 119–139.
Noe, C.F., 1999. Voluntary disclosures and insider transactions. Journal of Accounting and Economics 27 (3), 305–326.
Ohlson, J., Penman, S.H., 1985. Volatility increases subsequent to stock splits: an empirical aberration. Journal of Financial Economics 14 (2), 251–266.
Sen, R., 2009. The returns to spring-loading. Working Paper. SSRN 〈http://ssrn.com/abstract=1102639〉.
Tompkins, J.J., 2008. Opportunity knocks, but the SEC answers: Examining the manipulation of stock options through the spring-loading of grants and Rule
10b-5. Washington University Journal of Law & Policy 26, 413–458.
Yermack, D.L., 1997. Good timing: CEO stock option awards and company news announcements. Journal of Finance 52 (3), 449–476.
Weld, W., Michaely, R., Thaler, R., Benartzi, S., 2009. The nominal share price puzzle. Journal of Economic Perspectives 23 (2), 121–142.