ORI GIN AL ARTICLE
Stock market anomalies: what can we learnfrom repurchases and insider trading?
John E. Core Æ Wayne R. Guay ÆScott A. Richardson Æ Rodrigo S. Verdi
� Springer Sciecne+Business Media, Inc. 2006
Abstract We examine whether managers’ trading decisions (both at a firm and personal
level) are correlated with trading strategies suggested by the operating accruals and the
post-earnings announcement drift (SUE) anomalies. We discuss advantages and disad-
vantages of the use of managerial trading activity to infer managers’ private valuation
about their own securities. Our results provide corroborative evidence for the accruals
anomaly, i.e., managers’ repurchase and insider trading behavior varies consistently with
the information underlying the operating accruals trading strategy. On the other hand, we
do not find corroborative evidence for the SUE anomaly.
Keywords Anomalies Æ Accruals Æ Post-earnings announcement drift
1. Introduction
An extensive literature has developed challenging the assumption of market efficiency (for
a survey of this literature see Kothari (2001) and Lee (2001)). Collectively these papers
document empirical regularities consistent with investor under- or over-reaction to publicly
available information. Although these anomalies appear to have withstood a barrage of
robustness tests, the findings are not without criticism. Schwert (2003) notes that many
anomalies occur during specific time periods or within particular selected samples that
cannot be readily generalized or implemented on an ex ante basis. Similarly, Fama (1998)
documents that long-term return anomalies are sensitive to empirical methods, in that the
abnormal returns are sensitive to models for expected (normal) returns and statistical
techniques. Ball (1992) and Kothari (2001) also discuss the robustness of return anomalies.
This paper adds to this literature by examining whether managers’ trading decisions
(both at a firm and personal level) are correlated with trading strategies suggested by the
operating accruals and the post-earnings announcement drift (SUE) anomalies. We
J. E. Core Æ W. R. Guay Æ S. A. Richardson Æ R. S. Verdi (&)Accounting Department, The Wharton School, University of Pennsylvania,1300 Steinberg Hall-Dietrich Hall, 3620 Locust Walk, Philadelphia, PA, 19104-6365, USAe-mail: [emailprotected]
123
Rev Acc Stud (2006) 11:49–70DOI 10.1007/s11142-006-6395-y
investigate these two anomalies because they seem especially suited for the purposes of
this paper. Given the role that managers play in the financial reporting process, they are
uniquely informed about financial reporting and may be in the best position to observe
pricing deviations from fundamental value. The accruals anomaly, in particular, hinges on
the investor’s inability to recognize the differential persistence of accruals and cash flows,
and managers are at an advantaged position to understand the firm’s accruals process.
Our research design is based on empirical evidence that shows executives are well-
informed about their firms’ future expected cash flows and the cost of capital, and use this
information, at least at the margin, in their decisions about share repurchases and indi-
vidual trading. For example, Ikenberry, Lakonishok, and Vermaelen (1995) and D’Mello
and Shroff (2000) find that repurchase decisions are influenced by managers’ perceptions
about stock price undervaluation; Seyhun (1992) finds that insider trading is consistent
with managers taking advantage of deviations between stock prices and fundamental
values; and Rozeff and Zaman (1998) and Beneish and Vargus (2002) link insider trading
to misvaluations based on the market-to-book ratio and operating accruals.
Consistent with this evidence, we assume that executives, as insiders, are well-informed
about whether their own equity is mispriced, i.e., whether current stock price diverges from
fundamental value, and that they use this information at the margin in making share
repurchase and insider trading decisions. Further, we assume that the information set
available to insiders encompasses any mispricing that can be identified by anomalous trading
strategies. Therefore, if the trading strategies based on operating accruals and post-earnings
announcement drift do, in fact, identify equity mispricing, then we expect that insiders’ share
repurchase and individual trading decisions will be correlated with the positions recom-
mended by these trading strategies.1 We note that our tests do not require the executives to
be aware of, or follow, any specific anomalous trading strategy. For our purposes, it is
sufficient that the executives are skilled at identifying mispricings in their equity (regardless
of the source of the mispricing), and that mispricing influences marginal trading behavior.
The use of managerial trading activity to infer managers’ private valuation of their own
securities is a unique feature of our research design. We are interested in the extent to
which the information managers use to generate abnormal returns is correlated with the
‘‘anomalous’’ public information used to form trading strategies according to the accruals
and SUE anomalies. If we find evidence consistent with managers’ repurchase and insider
trades coinciding with trades suggested by these anomalies, this supports the mispricing
explanations offered for the anomalies. If, on the other hand, we find that the repurchase or
insider trading activity is uncorrelated with the trading suggested by the anomalies, then
the inferences we draw are constrained by the joint hypothesis nature of our research
design. That is, failure to reject the null hypothesis could be due to (1) managers lacking
the necessary sophistication to understand when their stock is mispriced, (2) the anomalies
examined do not in fact measure mispricing, or (3) managers perceive prohibitive trading
costs or risk associated with trading on the relevant mispricing. For example, if managers
buy stocks based on the anomalies, they would have to take on additional firm-specific risk
beyond the large firm-specific risk they already bear through their human capital and equity
investment in the firm. When they do this, they are outside the typical diversified hedge-
portfolio approach proposed by trading strategies. Distinguishing between these
1 For example, suppose that a firm with very low accruals is identified by the accruals trading strategy asbeing undervalued and that executives of this firm use their inside knowledge of expected future cash flowsto estimate the true equity value. If the executives believe the stock is undervalued, at the margin, they willbuy shares on personal and firm accounts more heavily, thereby corroborating the accruals trading strategy.
50 Rev Acc Stud (2006) 11:49–70
123
alternatives is difficult, and ultimately the interpretation of a non-result depends on the
readers’ priors with respect to managers’ knowledge of mispricing, trading costs, and the
existence of the anomalies.
Using measures of operating accruals (Sloan, 1996) and unexpected earnings (Bernard
and Thomas, 1989, 1990), we group firms into 10 equal-sized portfolios consistent with the
trading strategies suggested by the anomalies literature. We then examine the level of
abnormal share repurchases and insider trading for these portfolios in the period leading up
to, and following, the quarter in which we measure operating accruals and unexpected
earnings. We use abnormal levels of stock repurchases and insider trading in order to
mitigate omitted correlated variables concerns arising from the influence of expected factors
affecting repurchase and/or insider trading decisions.
Our results provide corroborative evidence for the accruals anomaly, i.e., managers’
repurchase and insider trading behavior varies consistently with the information underlying
the operating accruals trading strategy. We document that low (high) accruals firms
repurchase more (less) shares, and managers of low (high) accruals firms buy more (less)
shares on their personal accounts. On the other hand, we do not find corroborative evidence
for the SUE anomaly.
The fact that we find corroborating evidence for the accruals anomaly but not the SUE
anomaly suggests that under the assumption that the SUE anomaly truly captures mis-
pricing, trading on information of the type embedded in SUE is more costly to managers
than trading on information of the type embedded in accruals. Furthermore, given that our
research design has sufficient power to corroborate an anomaly that relies on financial
statement information (i.e., accruals), future research could exploit this design to examine
the validity of other financial-statement-based ‘‘anomalies.’’
Our findings on firms’ repurchase choices are related to a literature that examines how
market mispricing affects managers’ real investment and financing decisions. For example,
Polk and Sapienza (2004) find a positive relation between real investment and equity
overvaluation as measured by accounting accruals. Ang and Cheng (2003), Dong, Hir-
shleifer, Richardson, and Teoh (2006) and Rhodes-Kropf, Robinson, and Viswanathan
(2003) find that firm level investment activity (and in particular M&A activity) varies with
the extent of firm level misvaluation. In particular, Dong et al. (2006) focus on pre-
takeover ratios of residual income model value to price, and find that misvaluation of both
bidders and targets influences the means of payment chosen, the mode of acquisition, the
premia paid, target hostility to the offer, the likelihood of offer success, and bidder and
target announcement period stock returns. Finally, Baker and Wurgler (2002) provide
evidence that firms’ capital structure are consistent with cumulative attempts to time the
equity market and take advantage of market misvaluations.
Although these papers and our paper test similar joint hypotheses about misvaluation
and managers’ understanding of this misvaluation, our paper differs with respect to the
research design. For example, because transactions costs associated with share repur-
chases and managerial trading are likely to be smaller than those for investment, we
expect that managers can more easily adjust their repurchase and insider trading activity
to take advantage of transient mispricing. On the other hand, constraints on insider
trading activities, like restrictions on insider trades prior to major news events, and
reputation costs may reduce the power of tests such that a failure to reject the null
hypothesis may render the results inconclusive. In order to mitigate this last concern, we
measure insider trading activity before and after the accounting information is made
public, such that any trading activity based on public information would make violations
of insider trading laws less likely.
Rev Acc Stud (2006) 11:49–70 51
123
The remainder of the paper proceeds as follows. Section 2 summarizes the anom-
alies we investigate. Section 3 develops expectation models for share repurchase and
insider trading activity. Section 4 describes the sample selection and research design.
Section 5 presents our main results and Section 6 provides a summary and concluding
remarks.
2. Operating accruals and post-earnings announcement drift
In this section we briefly summarize the prior literature that established the operating
accruals and the post-earnings announcement drift anomalies we investigate.
2.1. Accruals anomaly
Sloan (1996) finds that stock prices do not properly reflect the information about future
earnings contained in current earnings. His results suggest that investors fail to understand
that the accruals component of earnings is less persistent than the cash flow component of
earnings. In particular, a portfolio strategy that takes a long (short) position in firms
reporting low (high) levels of operating accruals relative to cash flow, generates size-
adjusted abnormal returns of 10.4% on an annual basis (for a large sample of firms over the
1963–1991 period).
The accruals anomaly is robust to different specifications such as the use of investing
and financing accruals (Richardson, Sloan, Soliman, and Tuna, 2005), different countries
(Pincus, Rajgopal, and Venkatachalam, 2003), and the use of quarterly data (Collins and
Hribar, 2000), among others. Overall, the results are robust and consistent with Sloan’s
(1996) findings and suggest that future stock returns could be predicted due to investors’
overreaction to the persistence of the accruals component of earnings.
2.2. Post-earnings announcement drift (SUE) anomaly
The SUE anomaly, first reported by Ball and Brown (1968) and subsequently docu-
mented in many other studies, suggests that stock prices continue to rise in the year
following a positive earnings surprise and continue to fall in the year following a
negative earnings surprise.2 Bernard and Thomas (1989, 1990) investigate competing
explanations for the SUE effect and find results consistent with a delayed response of
stock prices to new information in earnings announcements. They find that an arbitrage
portfolio that takes a long position in firms reporting high standardized unexpected
earnings (SUE) and a short position in firms reporting low SUE generates abnormal
returns of approximately 8–9% in the first quarter after the earnings announcement
(Table 5, Bernard and Thomas, 1990). Likewise, Collins and Hribar (2000) find an
average two-quarter abnormal return of about 6.88% for the SUE hedge portfolio
strategy. The results suggest that the market underreacts to the earnings announcement
and responds only gradually to new information.
2 See Jones and Litzenberger (1970), Foster, Olsen, and Shevlin (1984), Chan, Jegadeesh and Lakonishok(1996), and Ball and Bartov (1996), Bushee and Raedy (2003), among others.
52 Rev Acc Stud (2006) 11:49–70
123
3. Measures of share repurchases and insider trading
In this section we define our measures of both repurchase and insider trading activity, and
develop models for the expected levels of these activities. As we note below, a substantial
body of evidence documents that share repurchases and insider trading varies with firm
characteristics. To increase the power of our tests, we control for cross-sectional variation
in managers’ trades due to these economic determinants, and use the resulting estimates of
unexpected or residual insider trading and share repurchases in our tests. As a standard
caveat, our expectation models may reflect mis-specification in both the functional form
(linear) and the set of included/excluded independent variables. We also note a potential
concern that some of the determinants of repurchases and insider trading may be correlated
with the proxies for mispricing identified by the anomalies. Controlling for these deter-
minants may affect our ability to find a relation between the anomalies and managers’
trades. Therefore, for robustness, we run all of our tests using unadjusted share repurchases
and insider trading variables.
3.1. Share repurchases
Managers generally cite undervaluation as a reason for repurchasing shares (Brav, Graham,
Harvey, and Michaely, 2005), and empirical evidence documents that managers repurchase
more heavily when they believe the stock is undervalued, and fewer shares when they
perceive the stock to be overvalued (Ikenberry et al., 1995; Ikenberry, Lakonishok, and
Vermaelen, 2000). Therefore, given our assumption that managers’ information set sub-
sumes the information set academics use to structure anomalous trading strategies, we
expect that if a particular anomaly represents the deviation of stock price from fundamental
value, firms will adjust their repurchase programs to benefit from the temporary mispricing
inherent in the anomaly.
Actual values of share repurchases are difficult to measure precisely for U.S. firms
(Stephens and Weisbach, 1998). We measure Repurchase as the percentage change in
split-adjusted shares outstanding over a six-month period as reported by CRSP, multiplied
by )1 so that the measure is increasing in the number of shares repurchased. For example,
if a firm XYZ has 50 shares outstanding at the end of month m)4, where month m is the
ending month of a given fiscal quarter t, and 45 shares outstanding at the end of month
m+2, Repurchase equals +10% (i.e., )1 * (45)50)/50 * 100). Negative values of this
measure indicate that new shares were issued during the six-month period around quarter t.
One concern with the Repurchase measure is that it is inclusive of all transactions that
decrease or increase the number of outstanding shares. As such, it includes follow-on
offerings that require share registrations and are costly and time-consuming, and thereby
unlikely to be used to take advantage of temporary equity mispricing. To increase the
power of our tests, we would like to limit our analysis to share repurchases and share re-
issuances out of treasury that the firm can execute inexpensively and on short notice.
Unfortunately, it is not possible to accurately identify the component of the change in
shares outstanding that relates to follow-on offerings. Instead, we attempt to identify and
reduce the influence of these offerings by examining the distribution of the six-month
change in treasury stock. We do this by computing the two-quarter change in the dollar
amount of shares held in treasury (data #98) and deflating it by the market value of total
Rev Acc Stud (2006) 11:49–70 53
123
shares outstanding at the start of the period for all firm-quarter observations in our sample.3
The first percentile of this distribution is )3.85% suggesting that the large re-issues out of
treasury are about 4% of outstanding shares. We then winsorize 10.83% of our sample
where Repurchase is more negative than )3.85%, i.e., if the actual value is less than
)3.85% of shares outstanding, then we replace it with )3.85%. This allows us to remove
follow-on equity offerings from our net repurchase measure yet still retain re-issues from
treasury.
We estimate abnormal net share repurchases as the residual from an OLS prediction
model that includes proxies for the economic determinants of share repurchases largely
based on Dittmar (2000). We expect firms with high cash balances to repurchase more
shares, and firms with high capital expenditures and high debt to repurchase fewer shares.
Cash is the ratio of cash and cash equivalents (data #36) to the book value of total assets
(data #44) measured at the end of quarter t)2.4 Cap. Expenses is the ratio of capital
expenditures (data #90) to the book value of total assets, and Debt is the ratio of long-term
debt (data #51) plus the long-term debt included in current liabilities (data #45) to the book
value of total assets, all measured in quarter t)2. We also include Dividend Yield as the
dividend-per-share (data #16) divided by the mean closing price in the quarter ((data
#12 + data #13 + data #14)/3). It is calculated for the 6-month period ending at the end of
quarter t)2. We do not predict the sign for this variable because it is uncertain whether
dividends and share repurchases are complements or substitutes (Grullon and Michaely,
2002).
We control for Size using the natural logarithm of the book value of total assets (data
#44) at the end of quarter t)2. We expect that small firms have greater information
asymmetries and should have more opportunities to repurchase shares for misvaluation
purposes. However, Dittmar (2000) finds that large firms are more likely to repurchase
shares, so we do not make any prediction about the sign of the estimated coefficient for this
variable. We also include fiscal quarter and year dummies to control for seasonality in firm
share repurchase volume. We include industry indicator variables based on firms’ two-digit
SIC code.5 Finally, we include a measure of lagged Repurchase because we expect that
repurchase patterns are persistent.6
Repurchaset ¼b0 þ b1�Repurchaset�2 þ b2�Sizet�2 þ b3�Casht�2 þ b4�Debtt�2
þ b5�Cap. Expensest�2 þ b6�Dividend Yieldt�2 þ Rbi�Yeari
þ Rbj�Quarterj þ Rbm�Industrym þ e: ð1Þ
3.2. Insider trading
We obtain insider trading data from Thomson Financial Securities Data. We focus on open
market transactions and, following Beneish and Vargus (2002), our measure of insider
trading is limited to transactions of the top five executives (CEO, CFO, COO, President,
and Chairman of board corresponding to relationship codes of ‘‘CB’’, ‘‘CEO’’, ‘‘CFO’’,
3 All data items refer to the Compustat quarterly files.4 As we discuss in more detail below, we measure the 6-month period to end 2 months after a quarter t.Accordingly, financial variables relevant to the repurchase decision are measured at quarter t)2, which ends2 months prior to the computation of repurchase (and insider trading) data.5 We delete firms in two-digit SIC codes that have an average of less than two observations per quarter.6 Our inferences are unaffected if we do not control for lagged repurchases.
54 Rev Acc Stud (2006) 11:49–70
123
‘‘CO’’ and ‘‘P’’ on the Thomson Financial database) because previous papers document
that these top executives are more likely to possess private information (Seyhun, 1986).
Following Rozeff and Zaman (1998) and Piotroski and Roulstone (2005), we measure
Insider Trading using the firm’s purchase ratio, defined as
Insider tradingi;t ¼ Buyi;t=ðBuyi;t þ Selli;tÞ ð2Þ
where Buyi,t (Selli,t) is the number of shares purchased (sold) by the top five executives of
firm i during the 6-month period ending 2 months after quarter t (i.e., the same 6-month
period examined for stock repurchases).
Similar to the share repurchases model, we estimate abnormal insider trading as the
residuals from an OLS prediction model that includes variables we expect to be correlated
with insider trading but not correlated with the variables underlying the anomalies. Prior
findings suggest that small firms have relatively more purchases than sales (Seyhun, 1986;
Rozeff and Zaman, 1988), and we control for Size with the natural logarithm of the book
value of total assets (data #44) at the end of quarter t)2. We include fiscal quarter and year
dummies to address seasonality in insider trading behavior, and two-digit SIC industry
dummies to control for industry effects. As with the repurchase model we also include a
measure of lagged Insider trading to capture the normal propensity of insiders at the firm to
sell/buy stock.
Insider tradingt ¼b0 þ b1�Insider tradingt�2 þ b2�Sizet�2 þ Rbi�Yeari
þ Rbj�Quarterj þ Rbm�Industrym þ e: ð3Þ
Our main hypothesis is to test whether shares repurchase and insider trading activity are
associated with trading strategies suggested by the operating accruals and SUE anomalies.
Thus we use the residuals from Equations 1 and 3 as our measures of Ab. Repurchase and
Ab. Insider trading, and we correlate these variables with our measures of operating
accruals and SUE.
4. Sample description, variable measurement and research design
There are 81,505 NYSE and AMEX firm-quarter observations on the Compustat Quarterly
Research and Active Files for the sample period 1989–2001. We limit ourselves to the
post-1988 period as our key variables are derived from statement of cash flow data.
NASDAQ firms are excluded from our analysis for the purpose of comparing our results to
prior research (e.g., Collins and Hribar, 2000). Elimination of firms with insufficient data to
compute accruals and SUE reduces the sample to 62,972 firm-quarters. Further elimination
of firms with insufficient data to compute share repurchases, insider trading, future stock
returns and additional explanatory variables for our expectation models produces the final
sample of 58,030 firm-quarters. For our tests examining repurchase and insider trading in
the following 6 months (i.e., POST-period), the sample drops to 51,437 firm-quarters.
Insider trading data is recorded by Thomson Financial as missing when insiders do not
trade in a period. To minimize data deletions, we assign a purchase ratio to observations with
missing trading data based on the mean insider purchase ratio for a portfolio of similar-sized
firms in the same calendar quarter (we match by size quintiles ranked on assets).7 We interpret
7 Specifically, for the five size quintiles (LOW to HIGH), the ‘‘no information’’ insider trading value is .60,.52, .45, .38 and .28, respectively.
Rev Acc Stud (2006) 11:49–70 55
123
this mean insider purchase ratio as the ratio that provides no information about insiders’
valuation beliefs. Our results are robust to assigning the unconditional mean value of the
insider purchase ratio to the firms with missing trade data, as well as to using the restricted
sample with non-missing insider trading data.
Our research design is based on measures of operating accruals and SUE using quarterly
data. Earnings are generally announced a few weeks after the end of the fiscal quarter, with
financial statement information necessary to compute accruals being disclosed a few weeks
after the earnings announcement. To ensure that investors have sufficient information
about earnings and accruals, we begin the return measurement period for the accruals
anomaly and SUE 2 months after the end of the fiscal quarter. In other words, for every
quarter t that ends with month m, we rank firms according to the accruals and SUE
anomalies, and we measure abnormal returns to these anomalies beginning at the start of
month m+3 relative to the end of quarter t.8
We examine repurchase and insider trading activity in the 12-month period straddling
the anomaly measurement period (i.e., from the start of month m)3 through to the end of
month m+8 relative to the end of quarter t). We label the 6-month period leading up to the
abnormal return window (i.e., the 6 months ending at month m+2) as the ‘‘PRE’’-period.
Likewise, we label the 6-month period at the beginning of the abnormal return window
(i.e., from the start of month m+3 to the end of month m+8) as the ‘‘POST’’-period.
Figure 1 depicts the variable measurement windows for our tests. We examine both the
PRE- and POST-periods because it is not clear when management will trade on infor-
mation that is correlated with the anomalies. For example, the accruals anomaly suggests
that firms reporting low levels of accruals experience improved stock returns going for-
ward, and these returns persist for the duration of the following fiscal period (with some
concentration around subsequent earnings announcement dates). A manager seeking to
take advantage of mispricing that is correlated with extreme accruals periods may trade
toward the end of an extreme accruals period or in the following fiscal periods as these
abnormal returns start to be realized. Similar arguments can be made for the SUE anomaly.
We note that our examination of trading during the PRE-period assumes that managers
have private information about firm valuation and possibly accounting information that
allows managers to perceive mispricings prior to the return measurement periods for the
anomalies. However, regulation on insider trading activity prohibits managers from trading
in periods prior to major news events, including earnings announcements, which will limit
managers’ ability to take advantage of the mispricing during the PRE-period. If this is the
case, then the POST-period may be considered a more powerful setting.
We measure operating accruals and SUE as per Collins and Hribar (2000). We compute
accruals as the difference between earnings and cash flows from operations using data from
the statement of cash flow,
Accrualst ¼ Earningst � CFOt ð4Þ
where Earningst is earnings from continuing operations (data #8) in quarter t and CFOt is
cash flow from operations (data #108) in quarter t. Accruals is then deflated by the average
book value of total assets (data #44) in quarter t.
8 Our choice to begin compounding 6-month returns 2 months after the quarter-end is consistent with that ofCollins and Hribar, who compound starting 18 trading days after the earnings announcement. Our com-pounding strategy will start about 42 trading days after the quarter-end, where theirs will typically startroughly 48 trading days after the quarter-end (assuming that earnings announcements are usually madeabout 30 trading days after the quarter (Easton and Zmijewski, 1993)).
56 Rev Acc Stud (2006) 11:49–70
123
We compute SUE as the difference between net income in quarter t and net income in
quarter t)4 deflated by market value of equity in quarter t)4,9
SUEt ¼ ðNIt � NIt�4Þ=MV Equityt�4 ð5Þ
where SUEt is the standardized unexpected earnings in quarter t, NIt is net income in
quarter t (data #69), and MV_Equityt-4 is market value of equity computed as the closing
share price (data #14) times the number of shares outstanding (data #61) at the end of
quarter t)4.
For each anomaly, we compute size-adjusted returns during the future return accumu-
lation period by subtracting the value-weighted average return for all firms in the same
size-matched decile, where size is measured as the market capitalization at the beginning
of the fiscal period. Returns are calculated for a 6-month period beginning 2 months after
the end of the fiscal quarter. If a security delists, we calculate the return by reinvesting the
remaining proceeds in the CRSP size-matched index until the end of the period. Consistent
Quarter t-2 Quarter t-1 Quarter t Quarter t+1 Quarter t+2
Month m-4 Month m+2 Month m+8PRE-Period POST-Period
Share repurchases Pre-Period Share repurchases Post-Period
Insider Trading Pre-Period Insider Trading Post-Period
Return Accumulation Period
Accruals t
SUE t
Fig. 1 Timeline of variable measurement. Repurchase is the percentage of shares repurchased in the6-month period ending 2 months after quarter t. We calculate Repurchase as )1 multiplied by thepercentage change in shares outstanding as reported by CRSP. Thus, the measure is increasing in the numberof shares repurchases. We winsorize Repurchase to be no less than )3.85%, i.e., if the actual repurchasevalue is less than )3.85% of shares outstanding, then we replace it with )3.85%. This number represents the1st percentile in our sample of the distribution of a two-quarter change in treasury stock from Compustat(data #98) deflated by the market value of total shares outstanding at the start of the period. Insider trading isthe insider purchase ratio measured as the ratio of the number of shares purchased by insiders in the 6-monthperiod ending 2 months after quarter t, scaled by total number of shares traded (purchases plus sales) byinsiders during the same period. Insider trading is limited to open market transactions of the top fiveexecutives coded as ‘‘CB’’, ‘‘CEO’’, ‘‘CFO’’, ‘‘CO’’, and ‘‘P’’ by Thomson Financial First Call InsidersData. Accruals are computed as the difference between earnings from continuing operations (data #8) inquarter t and cash flows from operations (data #108) in quarter t using data from the statement of cash flowdivided by average total assets (data #44) at the end of the quarters t)1 and t. SUE is computed as thedifference between the net income (data #69) in the quarter t and the net income (data #69) in the quarter t)4deflated by the market value of equity (data #14 * data #61) in the quarter t)4
9 Bernard and Thomas (1989) and Collins and Hribar (2000) scale unexpected earnings by the standarddeviation of the unexpected earnings. However, Bernard and Thomas (1990) show that the two approachesprovide similar results.
Rev Acc Stud (2006) 11:49–70 57
123
with previous research, we expect that firms with low (high) accruals and high (low) SUE
will exhibit high (low) returns in the post-formation period.
Our research design creates over-lapping observations. Specifically, because we
examine accruals and SUE in every quarter t along with repurchase and insider trading
activity for the two-quarter PRE- and POST-periods, there is a one-quarter over-lap in our
PRE- and POST-dependent variables. We correct for this autocorrelation using Huber
White robust standard errors, which are a generalization of the White (1980) standard
errors that are robust to both serial correlation and heteroskedasticity (Rogers, 1993).
Table 1 presents descriptive statistics of the variables. All variables are winsorized at
the 1% and 99% levels (i.e., for each variable we re-assign its value if it is less (greater)
than the 1st (99th) percentile to the value of the 1st (99th) percentile). The mean (median)
value for share repurchases is ).17% ()0.09%), and untabulated statistics show 26.11%
(15,149 observations) of our repurchase observations are positive (a positive value indi-
cates that the firm repurchased shares in the two quarters measured). The mean (median)
purchase ratio for insider traders is 43.15% (45.33%), and untabulated results document
that for 19.45% (11,287 observations) of firms the insiders are only sellers (insider’s
purchase ratio equals zero), whereas for 11.93% (6,923 observations) the insiders are only
buyers (insider’s purchase ratio equal one). The mean (median) cash to total asset ratio is
Table 1 Descriptive statistics
Variables n Mean S.D. Min Median Max
Repurchase (%) 58,030 ).17 2.35 )3.85 ).09 11.43Ins. trading (%) 58,030 43.15 29.33 .00 45.33 100.00Size 58,030 6.30 1.84 2.04 6.30 10.53Cash 58,030 .08 .12 .00 .04 .64Debt 58,030 .27 .19 .00 .26 .89Cap. expenses 58,030 .02 .02 .00 .01 .11Dividend yield 58,030 .01 .01 .00 .00 .05
Repurchase is the percentage of shares repurchased in the 6-month period ending 2 months after quarter t.We calculate Repurchase as )1 multiplied by the percentage change in shares outstanding as reported byCRSP. Thus, the measure is increasing in the number of shares repurchases. We winsorize Repurchase to beno less than )3.85%, i.e., if the actual repurchase value is less than )3.85% of shares outstanding, then wereplace it with )3.85%. This number represents the 1st percentile in our sample of the distribution of a two-quarter change in treasury stock from Compustat (data #98) deflated by the market value of total sharesoutstanding at the start of the period
Insider trading is the insider purchase ratio measured as the ratio of the number of shares purchased byinsiders in the 6-month period ending 2 months after quarter t, scaled by total number of shares traded(purchases plus sales) by insiders during the same period. Insider trading is limited to open market trans-actions of the top five executives coded as ‘‘CB’’, ‘‘CEO’’, ‘‘CFO’’, ‘‘CO’’, and ‘‘P’’ by ThomsonFinancial First Call Insiders Data
Size is the log of total assets (data #44) measured at the end of the quarter t)2
Cash is the ratio of cash and cash equivalents (data #36) to total assets (data #44) at the end of the quartert)2
Debt is the ratio of long-term debt (data #51) plus the debt included in current liabilities (data #45) to totalassets (data #44) at the end of the quarter t)2
Cap. expenses is the ratio of capital expenditures (data #90) to the average total assets (data #44). Thisvariable is measured at the end of the quarter t)2
Dividend yield is computed as the dividend-per-share (data #16) divided by the mean closing price in thequarter ((data #12+data #13+data #14)/3). It is calculated for the 6-month period ending in quarter t)2
58 Rev Acc Stud (2006) 11:49–70
123
.08 (.04) and the mean (median) debt to asset ratio is .27 (.26). Firms have mean (median)
capital expenditures of.02 (.01) of total assets and a mean (median) dividend yield per
quarter of .01 (.00).
Table 2 presents Pearson and Spearman correlations for the variables in the analysis.
We observe that net share repurchases are positively correlated with size and dividend
yield. The latter supports the argument that repurchases and dividends are complements
(Grullon and Michaely, 2002). As expected, net share repurchases are negatively correlated
with debt and capital expenditures. Finally, we observe that the insider trading measure is
negatively correlated with size, supporting previous findings that managers of large firms
are more likely to sell shares.
Table 2 Correlation matrix
Variables Repurchase (%) Ins. trading (%) Size Cash Debt Cap. expenses Dividend yield
Repurchase (%) ) .05 .06 ).01 ).06 ).05 .08<.0001 <.0001 .129 <.0001 <.0001 <.0001
Ins. trading (%) .07 – ).37 .04 .04 ).06 ).03<.0001 <.0001 <.0001 <.0001 <.0001 <.0001
Size .04 ).55 – ).29 .17 .06 .23<.0001 <.0001 – <.0001 <.0001 <.0001 <.0001
Cash .00 .06 ).21 ).37 ).08 ).11.87 <.0001 <.0001 <.0001 <.0001 <.0001
Debt ).07 ).03 .20 ).42 – .01 ).02<.0001 <.0001 <.0001 <.0001 .17 <.0001
Cap. expenses ).04 ).15 .18 ).08 ).01 – ).05<.0001 <.0001 <.0001 <.0001 .03 <.0001
Dividend yield .16 ).16 .35 ).14 ).03 .07 –<.0001 <.0001 <.0001 <.0001 <.0001 <.0001
Pairwise Correlation Matrix—Pearson correlations are reported above the diagonal and Spearman corre-lations below the diagonal. p-values are shown in italics below correlations
Repurchase is the percentage of shares repurchased in the 6-month period ending 2 months after quarter t.We calculate Repurchase as )1 multiplied by the percentage change in shares outstanding as reported byCRSP. Thus, the measure is increasing in the number of shares repurchases. We winsorize Repurchase to beno less than )3.85%, i.e., if the actual repurchase value is less than )3.85% of shares outstanding, then wereplace it with )3.85%. This number represents the 1st percentile in our sample of the distribution of a two-quarter change in treasury stock from Compustat (data #98) deflated by the market value of total sharesoutstanding at the start of the period
Ins. trading is the insider purchase ratio measured as the ratio of the number of shares purchased by insidersin the 6-month period ending 2 months after quarter t, scaled by total number of shares traded (purchasesplus sales) by insiders during the same period. Insider trading is limited to open market transactions of thetop five executives coded as ‘‘CB’’, ‘‘CEO’’, ‘‘CFO’’, ‘‘CO’’, and ‘‘P’’ by Thomson Financial First CallInsiders Data
Size is the log of total assets (data #44) measured at the end of the quarter t)2
Cash is the ratio of cash and cash equivalents (data #36) to total assets (data #44) at the end of the quartert)2
Debt is the ratio of long-term debt (data #51) plus the debt included in current liabilities (data #45) to totalassets (data #44) at the end of the quarter t)2
Cap. expenses is the ratio of capital expenditures (data #90) to the average total assets (data #44). Thisvariable is measured at the end of the quarter t)2
Dividend yield is computed as the dividend-per-share (data #16) divided by the mean closing price in thequarter ((data #12+data #13+data #14)/3). It is calculated for the 6-month period ending in quarter t)2
Rev Acc Stud (2006) 11:49–70 59
123
5. Results
5.1. Expectation model for share repurchases and for insider trading
Table 3 presents the results of our expectation models for share repurchases and for insider
trading.10 The models include fiscal quarter, year, and industry dummies (not tabulated).
Column I presents the results of our OLS model of share repurchases. All explanatory
variables, except cash, are statistically significant in the predicted direction consistent with
previous empirical research (e.g., Dittmar, 2000).11 There is considerable serial correlation
in stock repurchases at the firm level, which is consistent with firms entering the open
market to buy back shares over a number of fiscal periods. Large firms repurchase more
shares in accordance with Dittmar’s (2000) findings and in contrast to the asymmetric
information hypothesis. Firms with high leverage and high levels of capital expenditures
repurchase fewer shares, and firms that pay high dividend yields repurchase more shares
suggesting that dividends and share repurchases are complements in our sample. Column II
presents the results of our OLS model for insider trading. As predicted, managers from
large firms sell more shares than managers of small firms, and insider trading is serially
correlated. We use the residuals from the regressions in Table 3 as our measures of Ab.
Repurchase and Ab. Insider trading.
5.2. Portfolio analysis
We begin our analysis by replicating earlier findings on the abnormal returns that could be
earned by making investments according to the strategies underlying the accruals and SUE
anomalies. We assign firms into 10 portfolios in each of the 52 sample quarters based on
accruals and SUE.12 We compute abnormal returns for each portfolio in each of the 52
quarters in the sample. Table 4 reports the average of the 52 six-month size-adjusted
returns for each portfolio in both the PRE- and POST-periods, and for the hedge portfolio
using the extreme portfolios. We form hedge portfolios taking corresponding long and
short positions in firms from the extreme decile rankings. For accruals, the hedge portfolio
takes a long position in firms reporting low accruals and a short position in the high
accruals firms. For the SUE anomaly, the hedge strategy takes a long position in the high
SUE firms and a short position in the low SUE firms. We report a t-statistic, based on the
average and standard deviation of the 52 6-month returns, which tests whether the hedge
portfolio return is statistically different from zero.
The second column (i.e., labeled ‘‘Abnormal Returns POST’’) of panel A replicates
findings for the accruals anomaly. Consistent with earlier studies, the hedge portfolio
generates an abnormal return of 4.27% (t=4.63) in the 6 months after the start of the
portfolio formation period. These results are smaller in magnitude but comparable to the
10 As discussed in Section 3.1, our repurchase measure is censored at )3.85% to avoid the influence ofsecondary equity issuances. We note however that we obtain similar results if we use an uncensored measureof repurchases or if we censor the value of repurchase to zero to retain only decreases in shares outstanding(i.e., net repurchases only).11 If we censor negative values of repurchase to zero as an alternative computation of repurchases, then wefind that cash is positively related to repurchase consistent with previous literature such as Dittmar (2000).12 We use data from 1989 to 2001, so we have 52 overlapping 6-month periods (ending every fiscal quarter)for each firm.
60 Rev Acc Stud (2006) 11:49–70
123
5.56% abnormal return spread found by Collins and Hribar (2000) for the shorter 1988 to
1997.
The second column of panel B replicates previous findings for the SUE anomaly. The
hedge portfolio generates an abnormal return of 5.97% (t=6.02) in the 6 months following
Table 3 Expectation model for repurchase and insider trading
Indep. variables Repurchase (%) Ins. trading (%)
Predicted sign I Predicted sign II
Intercept ).33 6.83***().91) (13.78)
Repurchaset-2 + .15***(18.87)
Ins. tradingt-2 + .32***(40.83)
Size � .05*** ) )4.21***(4.64) )(34.66)
Cash + .02(.10)
Debt ) ).81***()7.88)
Cap. expenses ) )5.07***()5.92)
Dividend yield � 12.02***(6.20)
R-square (%) 6.69 24.02Observations 58,030 58,030
Column I presents the results of an OLS model examining the determinants of Repurchase. Column IIpresents the results of an OLS model examining the determinants of Ins. trading. Firm-quarter observationsare drawn from 1989 to 2001. Fiscal quarters, year and industry dummies are included in the model but nottabulated in the results. t-Statistics are presented in parenthesis below coefficient estimates based on HuberWhite robust standard errors. *, **, and *** indicate two-tailed statistical significance at 10, 5, and 1 levels,respectively
Repurchase is the percentage of shares repurchased in the 6-month period ending 2 months after quarter t.We calculate Repurchase as )1 multiplied by the percentage change in shares outstanding as reported byCRSP. Thus, the measure is increasing in the number of shares repurchases. We winsorize Repurchase to beno less than )3.85%, i.e., if the actual repurchase value is less than )3.85% of shares outstanding, then wereplace it with )3.85%. This number represents the 1st percentile in our sample of the distribution of a two-quarter change in treasury stock from Compustat (data #98) deflated by the market value of total sharesoutstanding at the start of the period
Ins. trading is the insider purchase ratio measured as the ratio of the number of shares purchased by insidersin the 6-month period ending 2 months after quarter t, scaled by total number of shares traded (purchasesplus sales) by insiders during the same period. Insider trading is limited to open market transactions of thetop five executives coded as ‘‘CB’’, ‘‘CEO’’, ‘‘CFO’’, ‘‘CO’’, and ‘‘P’’ by Thomson Financial First CallInsiders Data
Size is the log of total assets (data #44) measured at the end of the quarter t)2
Cash is the ratio of cash and cash equivalents (data #36) to total assets (data #44) at the end of the quartert)2
Debt is the ratio of long-term debt (data #51) plus the debt included in current liabilities (data #45) to totalassets (data #44) at the end of the quarter t)2
Cap. expenses is the ratio of capital expenditures (data #90) to the average total assets (data #44). Thisvariable is measured at the end of the quarter t)2
Dividend yield is computed as the dividend-per-share (data #16) divided by the mean closing price in thequarter ((data #12 + data #13 + data #14)/3). It is calculated for the 6-month period ending in quarter t)2
Rev Acc Stud (2006) 11:49–70 61
123
Table 4 Portfolio analysis
Accruals/SUERanks
Obs Ab. Returns (%) Ab. Repurchase (%) Ab. Ins. trading (%)
Pre Post Pre Post Pre Post
Panel A—Portfolio strategy for the accruals anomaly1 (Low) 5,779 )1.34 4.32 .07 .04 .93 ).272 5,811 .85 3.15 .07 .06 .17 .093 5,808 .82 1.91 .09 .06 ).02 .344 5,805 .89 2.14 ).02 .07 .57 .135 5,799 .52 1.48 .03 .05 .39 .286 5,819 1.11 1.39 .00 ).04 .07 ).257 5,808 1.60 1.20 ).06 .00 ).13 .678 5,805 1.38 .65 ).01 ).04 ).70 ).409 5,814 2.17 ).03 ).01 .00 ).25 .0110 (High) 5,782 2.75 .05 ).15 ).08 )1.01 )1.10
Hedge: 1–10 )4.09 4.27 .22 .12 1.94 .84t-test )3.83 4.63 5.46 2.82 3.75 2.14Pred. Sign + + + + +
Obs 58,030 58,030 58,030 51,437 58,030 51,437
Panel B—Portfolio strategy for the SUE anomaly1 (Low) 5,779 )10.33 1.05 .04 ).01 4.13 4.172 5,811 )8.51 ).44 .17 .19 3.41 2.663 5,808 )7.29 ).69 .23 .13 2.71 1.724 5,805 )4.60 ).14 .24 .20 .52 .085 5,799 )1.37 .56 .13 .18 ).76 )1.566 5,819 2.73 1.04 ).01 ).02 )2.93 )2.687 5,808 5.51 1.82 ).17 ).09 )3.21 )2.538 5,805 7.81 3.10 ).20 ).15 )2.60 )1.759 5,814 10.34 2.94 ).25 ).16 )1.59 ).6910 (High) 5,782 16.45 7.03 ).17 ).14 .37 .33
Hedge: 10–1 26.78 5.97 ).22 ).14 )3.76 )3.84t-test 21.24 6.02 )4.38 )2.85 )6.69 )6.40Pred. Sign + + + + +
Obs 58,030 58,030 58,030 51,437 58,030 51,437
Stock returns are measured using compounded buy-hold returns, inclusive of dividends and other distri-butions. Ab. Returns t is the size-adjusted returns calculated by deducting the corresponding value-weightedreturn for all available firms in the same size-matched decile, where size is measured using market capi-talization as of the beginning of the year. Returns are calculated for two 6-month intervals. The PRE periodbegins at the start of month m)3 and ends with month m+2 (where month m is the end of fiscal quarter). ThePOST period begins 2 months after the end of the fiscal quarter t (i.e., from the start of month m+3 to m+8).For firms that delist during our future return window, we calculate the remaining return by reinvesting anyremaining proceeds in the value-weighted market portfolio size-adjusted abnormal return ending 2 monthsafter the end of the quarter t. Ab. Repurchase t is the abnormal volume of share repurchases estimated as theresidual of the regression in Table 3 and Ab. Ins. trading t is the abnormal volume of insider tradingestimated as the residual of the regression in Table 3. Hedge 1–10 (10–1) represents the net return generatedby taking a long position in the Low (High) portfolio and an equal sized short position in the High (Low)portfolio. Pre corresponds to the two-quarter period (quarters t and t)1) ending 2 months after the end ofquarter t. Post corresponds to the two-quarter period (quarters t+1 and t+2) starting 2 months after the end ofquarter t. The t-statistic tests whether hedge is statistically different from zero. *, **, and *** indicate two-tailed statistical significance at 10, 5, and 1% levels
Repurchase is the percentage of shares repurchased in the six-month period ending 2 months after quarter t.We calculate Repurchase as )1 multiplied by the percentage change in shares outstanding as reported byCRSP. Thus, the measure is increasing in the number of shares repurchases. We winsorize Repurchase to be
62 Rev Acc Stud (2006) 11:49–70
123
the portfolio formation period. This result is very close to Collins and Hribar’s (2000)
results of average abnormal returns of 6.88%.
In the next three sub-sections, we examine whether abnormal share repurchase
volume and abnormal insider trading patterns are consistent with managers attempting
to benefit from the mispricing predicted by the accruals and the SUE anomaly. For
each anomaly, we examine abnormal share repurchases and abnormal insider trading
during the 6-month period concurrent with the anomaly formation period (PRE-period),
and during the six-month period following the anomaly formation period (POST-period).
Using raw share repurchases and raw insider trading does not affect the results
qualitatively.
5.2.1. Accruals anomaly
The last four columns in Table 4, panel A show a greater abnormal volume of share
repurchases and a higher proportion of insider buy trades for firms reporting low levels of
accruals compared to firms reporting high levels of accruals. These findings hold for
repurchases and insider trading during the 6-month period concurrent with the anomaly
portfolio formation period (PRE) and the following 6-month period (POST).
These findings suggest that managers’ repurchase and personal trading decisions are
consistent with the trading strategy underlying the accruals anomaly. Our results with
insider trading corroborate Beneish and Vargus’ (2002) findings that insider trading and
accruals are correlated. The strong pattern between repurchase activity and operating
accruals is also related to the external financing measure developed in Richardson and
Sloan (2003). Specifically, our finding of a contemporaneous correlation between repur-
chase activity in the PRE-period and operating accruals is related to the link established in
Richardson and Sloan between aggregate external financing and change in net operating
assets (a broader measure of accruals).
5.2.2. SUE anomaly
In panel B of Table 4, we do not observe that managers adjust the volume of share
repurchases or insider trades according to the SUE anomaly. Moreover, we document
abnormal repurchase volume and insider trading in the direction opposite to what would be
predicted by the SUE anomaly. Overall, the SUE results suggest that share repurchase
volume and managerial trading behavior are not consistent with the mispricing predicted
no less than )3.85%, i.e., if the actual repurchase value is less than )3.85% of shares outstanding, then wereplace it with )3.85%. This number represents the 1st percentile in our sample of the distribution of a two-quarter change in treasury stock from Compustat (data #98) deflated by the market value of total sharesoutstanding at the start of the period
Ins. trading is the insider purchase ratio measured as the ratio of the number of shares purchased by insidersin the 6-month period ending 2 months after quarter t, scaled by total number of shares traded (purchasesplus sales) by insiders during the same period. Insider trading is limited to open market transactions of thetop five executives coded as ‘‘CB’’, ‘‘CEO’’, ‘‘CFO’’, ‘‘CO’’, and ‘‘P’’ by Thomson Financial First CallInsiders Data
Accruals are computed as the difference between earnings from continuing operations (data #8) in quarter tand cash flows from operations (data #108) in quarter t using data from the statement of cash flow divided byaverage total assets (data #44) at the end of the quarters t)1 and t
Table 4 Continued
Rev Acc Stud (2006) 11:49–70 63
123
by the SUE anomaly. We discuss the possibility that this latter result is due to portfolio
rebalancing by insiders in the next section.
5.3. Regression analysis
Table 5 presents multiple regression results that capture correlation across the accruals and
the SUE anomalies.13 We include two transformations of the anomaly variables. First, we
use the anomalies’ ranks (coded as a discrete variable taking values from 1 to 10) as an
explanatory variable. This variable assumes a linear relation between the returns generated
by the anomalies and the decile rankings. Because previous findings document that the
anomaly returns are stronger in the extreme portfolios, we also define a dummy variable
coded ‘)1’ if the firm belongs to the first decile (low accruals and low SUE), ‘0’ if the firm
belongs to the intermediate portfolios (deciles 2–9) and ‘1’ if the firm is included in the
highest decile (high accruals and high SUE). The Table 5 regressions include the same
control variables employed in the repurchase and insider trading expectation models
reported in Table 3.
We also include Past_returns in the model. The rationale behind this variable is that the
SUE anomaly predicts that insiders should sell less stock and repurchase fewer shares
following positive earnings surprises. However, it can be argued that insiders will rebal-
ance their equity portfolios by selling their own firm stock after positive earnings surprises
because they are normally associated with stock price run-ups (see evidence in
Table 4—panel B). This behavior potentially confounds our interpretation of the SUE
tests.
Panel A of Table 5 presents the results of six regression specifications for share
repurchases. In Columns I to III, both the anomalies variables and share repurchases are
measured during the two quarters of the PRE-period. As in Table 4, we also include results
(Columns IV to VI) in which we measure share repurchases during the two quarters of the
POST-period.14
Consistent with the portfolio analysis in Table 4, the estimated coefficients for
Accruals_ranks are negative and significant for share repurchases in both the PRE- and
POST-periods. The coefficients on the Accruals_dummy are also negative in both the
PRE- and POST-periods, and robust to the inclusion of Past_returns in the model. We
find in panel B of Table 5 a consistently negative relation between insider trading and
accruals in the PRE but no relation between insider trading and the accruals variables in
the POST-period. This latter result may be due to the conservative estimate of the
POST-period. Because information in accruals is available to managers before the
POST-period, managers do not need to wait two months to trade on the accruals
information. Overall, the results suggest that marginal trading decisions of management
(both repurchases and insider trading) are consistent with the mispricing explanation
for accruals.
The estimated coefficients on the SUE variables in panel A of Table 5 are opposite the
predicted sign in both the PRE- and the POST-periods (with the exception of the SUE_-
dummy in the POST-period). For the insider trading regressions (panel B), the coefficients
13 Our inference is unaffected if we instead run separate regressions for each anomaly.14 In unreported tests we change our PRE- and POST-intervals to match the fiscal quarter. Specifically, thePRE-period ends at the end of the fiscal quarter and the POST-period starts in the beginning of the followingquarter. Using these alternate windows we find consistent results with the accruals anomaly for both sharerepurchases and insider trading in both periods. For the SUE anomaly, the results are similar to the onespresented in the paper and inconsistent with the mispricing explanation.
64 Rev Acc Stud (2006) 11:49–70
123
Tab
le5
Mu
ltip
lere
gre
ssio
ns
Ind
.v
aria
ble
Pre
d.
sig
nR
epu
rch
ase—
pre
-per
iod
Rep
urc
has
e—po
st-p
erio
d
III
III
IVV
VI
Pa
nel
A-A
naly
sis
of
sha
rere
pu
rcha
ses
Inte
rcep
t.0
1)
.33
).2
6)
.15
).3
8)
.33
(.0
2)
().9
2)
().7
2)
().3
8)
().9
5)
().8
3)
Rep
urc
has
e t-2
+.1
5**
*.1
5**
*.1
5**
*.1
6*
**
.16
**
*.1
6**
*(1
8.4
5)
(18
.81)
(18
.42
)(1
7.5
1)
(17
.83)
(17
.43)
Siz
e�
.05
**
*.0
5**
*.0
5**
*.0
5*
**
.05
**
*.0
5**
*(4
.46
)(4
.60
)(4
.59
)(3
.98
)(3
.98
)(3
.98
)C
ash
+.0
0.0
2.0
6.1
0.1
0.1
4.0
1.1
2.3
2.4
5.4
6.6
4D
ebt
))
.80
**
*)
.82
**
*)
.83
**
*)
.87
**
*)
.85
**
*)
.86
**
*()
7.8
2)
()7
.92)
()8
.01
)()
7.9
6)
()7
.81)
()7
.86)
Cap
.ex
pen
ses
))
5.7
7*
**
)5
.32*
**
)5
.31*
**
)5
.09
**
*)
4.9
5*
**
)5
.06*
**
()6
.79)
()6
.20)
()6
.24
)()
5.5
6)
()5
.35)
()5
.51)
Div
iden
dy
ield
�1
0.9
7*
**
11
.85
**
*1
0.3
9*
**
11
.31
**
*1
2.2
1*
**
10
.73
**
*(5
.68
)(6
.10
)(5
.36
)(5
.56
)(5
.96
)(5
.25
)A
ccru
als_
ranks
))
.01
**
*)
.01
**
()3
.68)
()2
.47)
SU
E_
ran
ks
+)
.05
**
*)
.04
**
*()
12
.86
)()
8.3
9)
Acc
rual
s_dum
my
))
.10
**
*)
.08
**
*)
.05
*)
.04
*()
4.3
3)
()3
.39
)()
1.9
1)
()1
.75)
SU
E_
du
mm
y+
).1
0**
*)
.07
**
*)
.06
**
.01
()4
.23)
()2
.80
)()
2.3
8)
(.6
1)
Pas
t_re
turn
s)
).0
1**
*)
.01
**
*()
17
.66
)()
16
.91
)R
-sq
uar
e(%
)7
.14
6.7
77
.49
7.3
17
.12
7.8
9O
bse
rvat
ions
58,0
30
58,0
30
58,0
30
51,4
37
51,4
37
51,4
37
Rev Acc Stud (2006) 11:49–70 65
123
Ta
ble
5C
on
tin
ued
Ind
.v
aria
ble
Pre
d.
sign
Rep
urc
has
e—pre
-per
iod
Rep
urc
has
e—p
ost
-per
iod
III
III
IVV
VI
Pa
nel
B-A
na
lysi
sfo
rin
sid
ertr
ad
ing
Inte
rcep
t6
4.9
3*
**
60
.90
**
*6
1.9
7*
**
66
.08
**
*6
3.3
2*
**
64
.26
**
*(1
4.7
4)
(13
.82
)(1
3.9
6)
(13
.88)
(13
.29)
(13
.35)
Ins.
trad
ing
t-2
+.3
2**
*.3
2**
*.3
1*
**
.31
**
*.3
2*
**
.31
**
*(4
0.6
0)
(40
.79
)(3
9.5
2)
(37
.99)
(38
.44)
(37
.20)
Siz
e)
)4
.27*
**
)4
.23*
**
)4
.29
**
*)
4.3
2*
**
)4
.29
**
*)
4.3
4*
**
()3
5.1
0)
()3
4.7
7)
()3
5.1
1)
()3
3.6
7)
()3
3.4
5)
()3
3.7
7)
Acc
rual
s_ra
nk
s)
).0
9**
.01
()2
.23)
(.1
9)
SU
E_
ran
ks
+)
.66
**
*)
.51
**
*()
14
.91
)()
11
.00
)A
ccru
als_
du
mm
y)
).7
6**
*)
.63
**
).1
7()
.63
))
.15
().5
8)
()3
.05
)()
2.5
2)
().6
3)
().5
8)
SU
E_
du
mm
y+
)1
.80*
**
)1
.56
**
*)
1.9
0*
**
)1
.44
**
*()
7.0
6)
()6
.17)
()6
.92)
()5
.25)
Pas
t_re
turn
s)
).0
4*
**
).0
4*
**
()1
0.2
9)
()9
.76)
R-s
qu
are
(%)
24
.46
24
.12
24
.31
24
.57
24
.41
24
.60
Ob
serv
atio
ns
58
,03
05
8,0
30
58
,03
05
1,4
37
51
,43
75
1,4
37
Pan
elA
pre
sen
tsth
ere
sult
so
fan
OL
Sm
od
elex
amin
ing
the
det
erm
inan
tso
fR
epu
rch
ase
and
the
effe
cts
of
the
anom
alie
svar
iable
s.P
anel
Bpre
sents
the
resu
lts
of
anO
LS
model
exam
inin
gth
edet
erm
inan
tso
fIn
s.tr
ad
ing
and
the
effe
cts
of
the
anom
alie
svar
iable
s.F
irm
-quar
ter
obse
rvat
ions
are
dra
wn
from
1989
to2001.
Pre
-Per
iod
corr
esp
on
ds
toth
etw
o-q
uar
ter
per
iod
(qu
arte
rst
and
t)1
)en
din
g2
mo
nth
saf
ter
the
end
of
qu
arte
rt.
Po
st-P
erio
dco
rres
po
nd
sto
the
two
-qu
arte
rp
erio
d(q
uar
ters
t+1
and
t+2
)st
arti
ng
2m
on
ths
afte
rth
een
do
fq
uar
ter
t.F
isca
lq
uar
ters
,y
ear
and
ind
ust
ryd
um
mie
sar
ein
clu
ded
inth
em
od
elb
ut
no
tta
bu
late
din
the
resu
lts.
t-S
tati
stic
are
pre
sen
ted
inp
aren
thes
isbel
ow
coef
fici
ent
esti
mat
esbas
edon
Huber
Whit
ero
bust
stan
dar
der
rors
.*
,**,
and
***
indic
ate
two-t
aile
dst
atis
tica
lsi
gnifi
cance
at10,
5,
and
1%
lev
els
Rep
urc
ha
seis
the
per
centa
ge
of
shar
esre
purc
has
edin
the
6-m
onth
per
iod
endin
g2
month
saf
ter
quar
ter
t.W
eca
lcu
late
Rep
urc
ha
seas
)1
mult
ipli
edby
the
per
centa
ge
chan
ge
insh
ares
ou
tsta
nd
ing
asre
po
rted
by
CR
SP
.T
hu
s,th
em
easu
reis
incr
easi
ng
inth
en
um
ber
of
shar
esre
pu
rch
ases
.W
ew
inso
rize
Rep
urc
ha
seto
be
no
less
than
)3
.85
%,
i.e.
,if
the
actu
alre
pu
rch
ase
val
ue
isle
ssth
an)
3.8
5%
of
shar
eso
uts
tan
din
g,
then
we
rep
lace
itw
ith
)3.8
5%
.T
his
num
ber
repre
sents
the
1st
per
centi
lein
our
sam
ple
of
the
dis
trib
uti
on
of
atw
o-q
uar
ter
chan
ge
intr
easu
ryst
ock
from
Com
pu
stat
(dat
a#
98
)d
eflat
edb
yth
em
ark
etv
alu
eo
fto
tal
shar
eso
uts
tand
ing
atth
est
art
of
the
per
iod
Ins.
trad
ing
isth
ein
sid
erp
urc
has
era
tio
mea
sure
das
the
rati
oo
fth
en
um
ber
of
shar
esp
urc
has
edb
yin
sid
ers
inth
e6
-mo
nth
per
iod
end
ing
2m
on
ths
afte
rq
uar
ter
t,sc
aled
by
tota
ln
um
ber
of
shar
estr
aded
(pu
rch
ases
plu
ssa
les)
by
insi
der
sd
uri
ng
the
sam
ep
erio
d.
Insi
der
trad
ing
isli
mit
edto
op
enm
ark
ettr
ansa
ctio
ns
of
the
top
fiv
eex
ecu
tiv
esco
ded
as‘‘
CB
’’,
‘‘C
EO
’’,
‘‘C
FO
’’,
‘‘C
O’’
,an
d‘‘
P’’
by
Th
om
son
Fin
anci
alF
irst
Cal
lIn
sider
sD
ata
66 Rev Acc Stud (2006) 11:49–70
123
Siz
eis
the
log
of
tota
las
sets
(dat
a#
44
)m
easu
red
atth
een
do
fth
eq
uar
ter
t)2
Ca
shis
the
rati
oo
fca
shan
dca
sheq
uiv
alen
ts(d
ata
#3
6)
toto
tal
asse
ts(d
ata
#4
4)
atth
een
do
fth
eq
uar
ter
t)2
Deb
tis
the
rati
oo
flo
ng
-ter
md
ebt
(dat
a#
51
)p
lus
the
deb
tin
clu
ded
incu
rren
tli
abil
itie
s(d
ata
#4
5)
toto
tal
asse
ts(d
ata
#4
4)
atth
een
do
fth
eq
uar
ter
t)2
Ca
p.
Exp
ense
sis
the
rati
oo
fca
pit
alex
pen
dit
ure
s(d
ata
#9
0)
toth
eav
erag
eto
tal
asse
ts(d
ata
#4
4).
Th
isv
aria
ble
ism
easu
red
atth
een
do
fth
eq
uar
ter
t)2
Div
iden
dyi
eld
isco
mp
ute
das
the
div
iden
d-p
er-s
har
e(d
ata
#1
6)
div
ided
by
the
mea
ncl
osi
ng
pri
cein
the
qu
arte
r((
dat
a#
12
+d
ata
#1
3+
dat
a#
14
)/3
).It
isca
lcu
late
dfo
rth
e6
-m
on
thp
erio
den
din
gin
qu
arte
rt)
2
Acc
ruals
are
com
pu
ted
asth
ed
iffe
ren
ceb
etw
een
earn
ing
sfr
om
con
tin
uin
go
per
atio
ns
(dat
a#
8)
inq
uar
ter
tan
dca
shfl
ow
sfr
om
op
erat
ion
s(d
ata
#1
08
)in
qu
arte
rt
usi
ng
dat
afr
om
the
stat
emen
tof
cash
flow
div
ided
by
aver
age
tota
las
sets
(dat
a#44)
atth
een
dof
the
quar
ters
t)1
and
t
SU
Eis
com
pu
ted
asth
ed
iffe
ren
ceb
etw
een
the
net
inco
me
(dat
a#
69
)in
the
qu
arte
rt
and
the
net
inco
me
(dat
a#
69
)in
the
qu
arte
rt)
4d
eflat
edb
yth
em
ark
etv
alu
eo
feq
uit
y(d
ata
#1
4*
dat
a#
61
)in
the
qu
arte
rt)
4
Acc
ruals
_ra
nks
isa
dis
cret
ev
aria
ble
cod
ed1
–1
0d
epen
din
go
nw
hic
hp
ort
foli
oth
efi
rmb
elo
ng
sto
inth
ecu
rren
tq
uar
ter
rank
edb
yo
per
atin
gac
cru
als
atq
uar
ter
t
SU
E_ra
nks
:a
dis
cret
ev
aria
ble
cod
ed1
–1
0d
epen
din
go
nw
hic
hp
ort
foli
oth
efi
rmb
elon
gs
toin
the
curr
ent
qu
arte
rra
nk
edb
yS
UE
atq
uar
ter
t
Acc
ruals
_d
um
my:
ad
um
my
var
iab
leco
ded
‘)1’
ifth
efi
rmbel
ong
toth
efi
rst
dec
ile
(low
accr
ual
s),
‘0’
ifth
efi
rmbel
ong
toth
ein
term
edia
teport
foli
os
(dec
iles
2–9)
and
‘1’
ifth
efi
rmis
incl
uded
inth
ehig
hes
tdec
ile
(hig
hac
crual
s)at
quar
ter
t
SU
E_d
um
my:
ad
um
my
var
iab
leco
ded
‘)1
’if
the
firm
bel
on
gto
the
firs
td
ecil
e(l
ow
SU
E),
‘0’
ifth
efi
rmb
elo
ng
toth
ein
term
edia
tep
ort
foli
os
(dec
iles
2–
9)
and
‘1’
ifth
efi
rmis
incl
ud
edin
the
hig
hes
td
ecil
e(h
igh
SU
E)
atq
uar
ter
t
Pa
st_
retu
rns
isth
e6
-mo
nth
per
iod
stock
retu
rnen
din
gat
the
end
of
the
fisc
alq
uar
ter
t
Ta
ble
5C
on
tin
ued
Rev Acc Stud (2006) 11:49–70 67
123
on the SUE variables are negative and statistically significant. These findings corroborate
the portfolio results in Table 4 that suggest that share repurchase volume and insider
trading are not consistent with the mispricing predicted by the SUE anomaly.
Overall, the regression results corroborate the portfolio analysis suggesting that man-
agers’ repurchase and insider trading behavior varies consistently with the information
underlying the operating accruals trading strategy. On the other hand, we do not find
corroborative evidence for the SUE anomaly.15
6. Summary and conclusions
In this paper, we examine whether managers’ trading decisions (both at a firm and personal
level) are correlated with trading strategies suggested by the operating accruals and the
post-earnings announcement drift (SUE) anomalies. These two anomalies are especially
suited for the purposes of this paper because managers are uniquely informed about
financial reporting and are in the best position to observe pricing deviations from funda-
mental value.
The use of managerial trading activity to infer managers’ private valuation of their own
securities is a unique feature of our research design, and we discuss advantages and
disadvantages of this methodology in the paper. We argue that if managers are skilled at
identifying equity mispricing and if an anomaly accurately captures situations where stock
prices deviate from firms’ fundamental value, then managers’ marginal trading decisions
through repurchase programs and on personal accounts will be correlated with the anomaly
trading strategy. On the other hand, restrictions on the managers’ ability to take advantage
of temporary mispricing due to insider trading restrictions, prohibitive trading costs, or risk
associated with trading strategy may render the interpretation of null results inconclusive.
We provide corroborative evidence for the accruals anomaly, i.e., managers’ repurchase
and insider trading behavior varies consistently with the information underlying the
operating accruals trading strategy. Low (high) accruals firms repurchase more (less)
shares, and managers of low (high) accruals firms buy more (less) shares on their personal
accounts. On the other hand, we do not find corroborative evidence for the SUE anomaly.
The fact that we find corroborating evidence for the accruals anomaly but not the SUE
anomaly suggests that under the assumption that the SUE anomaly truly captures mis-
pricing, trading on information of the type embedded in SUE is more costly to managers
than trading on information of the type embedded in accruals. Furthermore, given that our
research design has sufficient power to corroborate an anomaly that relies on financial
statement information (i.e., accruals), future research could exploit this design to examine
the validity of other financial-statement-based ‘‘anomalies.’’
Acknowledgements We appreciate the comments and suggestions of Peter Easton (editor), JosephGerakos, Theodore Goodman, Jeffrey Ng, Jonathan Rogers, Tjomme Rusticus, Sarah Zechman, twoanonymous referees, and seminar participants at the University of Pennsylvania and 2004 EuropeanAccounting Association Meeting. We appreciate financial support from The Wharton School. Rodrigo Verdiis also grateful for financial support from the Deloitte and Touche Foundation.
15 Firms without share repurchase programs may find it too costly to start a repurchase program simply totake advantage of transient mispricing. Therefore, we repeat our analysis using 43,391 firms with somehistory of repurchasing shares. For this reduced sample we find results very similar to those reported in thepaper.
68 Rev Acc Stud (2006) 11:49–70
123
References
Ang, J. S., and Cheng, Y. (2003). Direct evidence on the market-driven acquisitions theory. Working paper,Florida State University
Baker, M., and Wurgler, J. (2002). Market timing and capital structure. Journal of Finance, 57(1), 1–32Ball, R. (1992). The earnings–price anomaly. Journal of Accounting and Economics, 15, 319–345Ball, R., and Bartov, E. (1996). How naive is the stock market’s use of earnings information? Journal of
Accounting and Economics, 21, 319–337Ball, R., and Brown, P. (1968). An empirical evaluation of accounting income numbers. Journal of
Accounting Research, 6, 159–177Beneish, M., and Vargus, M. (2002). Insider trading, earnings quality, and accruals mispricing. The
Accounting Review, 77(4), 755–791Bernard, V., and Thomas, J. (1989). Post-earnings-announcement drift: Delayed price response or risk
premium? Journal of Accounting Research, 27, 1–48Bernard, V., and Thomas, J. (1990). Evidence that stock prices do not fully reflect the implications of current
earnings for future earnings. Journal of Accounting and Economics, 13, 305–340Brav, A., Graham, J., Harvey, C., and Michaely, R. (2005). Payout policy in the 21st century. Journal of
Financial Economics, 77(3), 483–527Bushee, B., and Raedy, J. (2003). Factors affecting the implementability of stock market trading strategies.
Working paper, University of PennsylvaniaChan, K., Jegadeesh, N., and Lakonishok, J. (1996). Momentum strategies. Journal of Finance, 51, 1681–
1713Collins, D., and Hribar, P. (2000). Earnings-based and accrual-based market anomalies: One effect or two?.
Journal of Accounting and Economics 29, 101–123D’Mello, R., and Shroff, P. (2000). Equity undervaluation and decisions related to repurchase tender offers:
An empirical investigation. Journal of Finance, 55(5), 2399–2424Dittmar, A. (2000). Why do firms repurchase stock? Journal of Business, 73(3), 331–355Dong, M., Hirshleifer, D., Richardson, S., and Teoh, S. (2006). Does investor misvaluation drive the
takeover market? Journal of Finance, 61(2), 725–762Easton, P., and Zmijewski, M. (1993). SEC form 10K/10Q reports and annual reports to shareholders:
Reporting lag and squared market model prediction errors. Journal of Accounting Research, 31(1), 113–129
Fama, E. (1998). Market efficiency, long-term returns, and behavioral finance. Journal of Financial Eco-nomics, 49, 283–306
Foster, G., Olsen, C., and Shevlin, T. (1984). Earnings releases, anomalies and the behavior of securityreturns. The Accounting Review, 59, 574–603
Grullon, G., and Michaely, R. (2002). Dividends, share repurchases, and the substitution hypothesis. Journalof Finance, 57(4), 1649–1684
Ikenberry, D., Lakonishok, J., and Vermaelen, T. (1995). Market underreaction to open market sharerepurchases. Journal of Financial Economics, 39, 181–208
Ikenberry, D., Lakonishok, J., and Vermaelen, T. (2000). Stock repurchases in Canada: Performance andstrategic trading. Journal of Finance, 55(5), 2373–2397
Jones C., and Litzenberger, R. (1970). Quarterly earnings reports and intermediate stock price trends.Journal of Finance, 25, 143–148
Kothari, S. P. (2001). Capital markets research in accounting. Journal of Accounting and Economics,31,105–231
Lee, C. (2001). Market efficiency and accounting research: A discussion of ‘capital markets research inaccounting’ by S. P. Kothari. Journal of Accounting and Economics, 31, 233–253
Pincus, M., Rajgopal, S., and Venkatachalam, M. (2003). The accrual anomaly: International evidence.Working paper, University of Iowa
Piotroski, J., and Roulstone, D. (2005). Do insider trades reflect superior knowledge about future cash flowrealizations? Journal of Accounting and Economics, 39(1), 55–81
Polk, C., and Sapienza, P. (2004). The real effects of investor sentiment. Working paper, NorthwesternUniversity
Rhodes-Kropf, M., Robinson, D., and Viswanathan, S. (2003). Valuation waves and merger activity: Theempirical evidence.’ Working paper, Columbia University
Richardson, S., Sloan, R., Soliman, M., and Tuna, I. (2005). Accrual reliability, earnings persistence andstock prices. Journal of Accounting and Economics, 39(3), 437–485
Richardson, S., and Sloan, R. (2003). External financing, capital investment and future stock returns.Working paper, University of Pennsylvania
Rev Acc Stud (2006) 11:49–70 69
123
Rogers, W. (1993). Regression standard errors in clustered samples. Stata technical bulletin reprints,(Volume 3, pp. 83–94). College Station, Texas: Stata Press
Rozeff, M., and Zaman, M. (1988). Market efficiency and insider trading: New evidence. Journal ofBusiness, 61(1), 25–44
Rozeff, M., and Zaman, M. (1998). Overreaction and insider trading: Evidence from growth and valueportfolios. Journal of Finance, 53(2), 701–716
Schwert, G. (2003). Anomalies and market efficiency, Chapter 15. In George Constantinides, Milton Harris,and Rene Stulz (Eds.) Handbook of the economics of finance (pp. 937–972). North-Holland
Seyhun, H. (1986). Insider profits, costs of trading, and market efficiency. Journal of Financial Economics,16(6), 189–212
Seyhun, H. (1992). Why does aggregate insider trading predict future stock returns. The Quarterly Journalof Economics, 107(4), 1303–1331
Sloan, R. (1996). Do stock prices fully reflect information in accruals and cash flows about future earnings?The Accounting Review, 71, 289–316
Stephens, C., and Weisbach, M. (1998). Actual share reacquisitions in open market repurchase programs.Journal of Finance, 53, 313–334
White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heter-oskedasticity. Econometrica, 48, 817–838
70 Rev Acc Stud (2006) 11:49–70
123
