Do Firms Mislead Investors by Overstating Earnings Before … · 2017-09-28 · Do Firms Mislead...
Transcript of Do Firms Mislead Investors by Overstating Earnings Before … · 2017-09-28 · Do Firms Mislead...
Do Firms Mislead Investors by Overstating Earnings Before Seasoned Equity Offerings?
Lakshmanan Shivakumar*
London Business School
October 5, 2000
* I have benefited from the comments of Ray Ball, Ronnie Barnes, Dick Brealey, J.S. Butler, Paul Chaney, BillChristie, Paul Healy, Debra Jeter, S.P. Kothari (editor), Craig Lewis, Maureen McNichols, Pat O’Brien, HenriServaes, Chitaranjan Sinha, an anonymous referee and seminar participants at Vanderbilt University, Universityof California (Riverside), London Business School, ESSEC, Virginia Tech, University of Toronto, University ofAlberta and the 1997 American Accounting Association annual meeting. I am especially grateful to RonaldMasulis for his valuable suggestions and for generously providing the data on equity offerings. Financial supportfrom the Financial Markets Research Center at Vanderbilt University and from the Deans’ Research Fund atLondon Business School are gratefully acknowledged.Address for correspondence: L. Shivakumar, London Business School, Regent’s Park, London NW1 4SA, UK.Tel: +44 20 7262 5050, Fax: +44 20 7724 7875, Email: [email protected]
Do Firms Mislead Investors by Overstating Earnings Before Seasoned Equity Offerings?
Abstract
I examine earnings management around seasoned equity offerings and, consistent with Rangan
(1998) and Teoh et al. (1998), find evidence of earnings management around the offerings.
However, in contrast to their conclusions, I show that investors infer earnings management
and rationally undo its effects at equity offering announcements. The investor naïveté
conclusion of Teoh et al. (1998) and Rangan (1998) appears to be due to test misspecification.
I conclude that seasoned equity issuers’ earnings management may not be designed to mislead
investors, but may merely reflect the issuers’ rational response to anticipated market behavior
at offering announcements.
Key Words: Corporate Finance; Accruals; Earnings Management; Seasoned Equity
Offerings; Offering Announcements.
JEL Classification: G14; M41
1
1. Introduction
Earnings management around firm-specific events has received considerable attention
from researchers in recent years.1 These studies typically examine managers’ reporting
behavior around specific corporate events, and conclude that evidence of earnings
management is consistent with managerial opportunism. However, relatively little is known
about investor response to earnings management, particularly following firm-specific news
releases that should alert investors to such earnings management. This paper examines both
managerial reporting behavior and investors’ response around public offerings of common
stock. The results suggest that earnings management is explained by a rational expectations
model at least as well as by managerial opportunism.
I hypothesize that managers overstate earnings before announcing seasoned equity
offerings, and that an offering announcement reveals this overstatement to market participants.
Thus, on the announcement of an equity offering, investors lower their assessments of prior
earnings surprises, and rationally discount firm value. The average price drop at the
announcements of seasoned equity offerings is consistent with this investor conditioning
process.
At first glance, the above hypothesis appears paradoxical. Why would issuing firms
engage in earnings management if investors undo its effects at offering announcements? I
argue that earnings management before equity offerings is not intended to mislead investors,
but is instead the issuers’ rational response to anticipated market behavior at offering
announcements. Since issuers cannot credibly signal the absence of earnings management,
investors treat all firms announcing an offering as having overstated prior earnings, and
consequently discount their stock prices. Anticipating such market behavior, issuers rationally
2
overstate earnings prior to offering announcements, at least to the extent expected by the
market. Earnings management by issuers and the resulting discounting by investors is a
unique Nash equilibrium in a prisoner’s dilemma game between issuers and investors. I refer
to this argument for earnings management as the “Managerial Response” hypothesis.
A secondary objective of this paper is to reexamine the evidence presented in two
recent studies by Rangan (1998) and Teoh et al. (1998). Rangan (1998) and Teoh et al. (1998)
investigate whether earnings management before seasoned equity offerings causes the poor
long-run stock performance following equity offerings, which originally appears in Loughran
and Ritter (1995) and Spiess and Affleck-Graves (1995). Both Rangan and Teoh et al.
hypothesize that investors fail to recognize earnings management at the time of equity
offerings and naively extrapolate pre-offering earnings increases. Consistent with their
hypothesis, Rangan and Teoh et al. report a negative relation between pre-offering abnormal
accruals and post-offering abnormal stock returns, measured either as market-adjusted returns
or as prediction errors from the Fama-French three-factor model. However, statistical tests
based on these measures of abnormal returns are severely mis-specified owing to, among other
factors, skewness in long-horizon returns data (see Kothari and Warner, 1997 and Barber and
Lyon, 1997). Further, Kothari et al. (2000) show that such skewness combined with data
attrition, either because of firms’ survival or deletion of extreme observations, can induce
spurious association between ex ante forecast variables (such as abnormal accruals) and ex
post security returns. Hence I reexamine the evidence in Rangan’s and Teoh et al. using
alternative methodologies that previous research suggests are relatively well specified.
My paper is closely linked to a recent study by Erickson and Wang (1999), who
investigate earnings management around stock for stock mergers and show that acquiring
1For example, DeAngelo (1986), Perry and Williams (1994) (around management buy outs), Aharoney et al.
3
firms overstate earnings in the pre-merger quarters. Erickson and Wang present several
alternative explanations for their findings, including a rational expectations argument that
acquirers overstate earnings before merger agreements because the target firms anticipate it
and adjust for the anticipated earnings management when negotiating on purchase price.
However, Erickson and Wang do not test the hypothesis. I formalize their rational
expectations argument and empirically examine the argument, albeit in the context of
seasoned equity offerings rather than mergers.
The paper’s major results are as follows: Consistent with earnings management,
accruals are abnormally high before equity offerings, and they predict subsequent declines in
net income. Also, consistent with the market learning about this earnings management through
offering announcements, investors’ response to unexpected earnings are significantly weaker
following offering announcements. Further, the pre-announcement abnormal accruals predict
the two-day negative price reaction to an offering announcement, which supports the view that
investors rationally correct for earlier earnings management at offering announcements.
Finally, the negative relation between pre-offering abnormal accruals and post-offering stock
performance, documented by Rangan and Teoh et al., depends crucially on the choice of
testing procedures and is not robust to methodologies that research shows are relatively well
specified. These findings support the Managerial Response hypothesis for earnings
management.
This paper makes several contributions to the literature. First, it proposes a non-
opportunistic motive for earnings management, and challenges the frequently articulated view
that earnings management around corporate events is synonymous with managerial
opportunism. Second, by establishing a specific cause for the average price drop observed at
(1993) (around initial public offerings), and Erickson and Wang (1999) (around stock-for-stock mergers).
4
offering announcements, it explains investors’ response to such announcements. Third, the
paper sheds new light on the debate on long-run stock price performance following equity
offering announcements by reexamining the relation between earnings management and post-
issue stock price performance. Finally, by showing a weaker price response to earnings
following an equity offering announcement, the paper enhances our understanding of the
relation between accounting earnings and equity valuation.
The next section develops the hypotheses examined in this study. Section 3 discusses
the measurement of earnings management. Section 4 presents and interprets the empirical
results. Section 5 provides the summary and conclusions.
2. Hypothesis development
2.1 Managerial Response hypothesis
I model earnings management before equity offerings as the outcome of a rational
expectations model: Investors expect firms announcing equity offerings to manage earnings
and, consistent with this expectation, issuers overstate earnings before announcing their
offerings. To formalize this argument, consider the game depicted in Figure 1. If investors can
directly observe the level of earnings overstatements in reported earnings, then the co-
operative outcome given by box (1,1) in Figure 1 will be an equilibrium. In this equilibrium,
only the unmanaged earnings (i.e., reported earnings excluding earnings management) will be
priced at earnings releases, and offering announcements will not cause revisions in the stock
price or in investors’ beliefs about prior earnings. Moreover, the new shares in this equilibrium
will be issued at their fundamental value.
However, in reality, investors cannot perfectly discern the amount of earnings
management. In this situation, issuing firms have an incentive to deviate from the above
equilibrium. By overstating earnings, a manager might attempt to fool investors and issue new
5
shares at an artificially high price. Given this incentive, investors assume that firms
announcing equity offerings have previously managed earnings upward, and therefore
discount these firms’ stock prices.2 In this circumstance, even issuers who have not previously
overstated earnings will suffer stock price declines at offering announcements, resulting in
artificially low offer prices. Hence, it is only rational for issuers to overstate earnings before
announcing an equity offering, at least to the extent expected by the market. This leads to box
(2,2) in Figure 1 as the equilibrium outcome, which is in fact a unique Nash equilibrium in
pure strategies.3 However, this equilibrium is Pareto inefficient as long as there are nonzero
costs to earnings management.
The above explanation also extends the Myers-Majluf (1984) adverse selection model.
In the Myers-Majluf model, managers prefer to issue equity when their private valuation is
lower than that of the market. However, investors rationally infer this managerial preference
and lower the seasoned equity offerer’s valuation. While the Myers-Majluf model allows
managers to have information superior to that of the market, and allows for firms to be
temporarily overvalued, Myers and Majluf do not identify the sources of informational
advantage and mispricing. I hypothesize that earnings management before equity offerings is a
means of temporary overvaluation of offering firms. Thus, in this extended Myers-Majluf
model, managers choose to overstate earnings before equity offerings, and at the offering
announcements investors infer this reporting choice, causing the downward valuation of
announcing firms. This extension indicates that the asymmetry in information between
investors and managers, assumed to be exogenous in the original model, may actually have an
endogenous dimension.
2 The expected level of earnings management will depend on investors’ perception of managerial incentives, andon their ability to overstate earnings.3 The proof is obvious once we recognize that the game depicted in Figure 1 is the prisoner’s dilemma.
6
The above rational expectations model is also consistent with the model in Stein
(1989), who argues that managers may be trapped into making myopic economic decisions
simply because investors expect them to do so. Although Stein (1989) focuses primarily on
economic or real managerial decisions, his model can be interpreted to explain accrual
manipulations as well. For instance, earnings management by issuers can be viewed as a
specific case of myopic behavior where managers overstate current earnings by borrowing
from future income.
2.2 Alternative hypotheses
A frequently advanced alternative to the Managerial Response hypothesis is that
managers overstate earnings before seasoned equity offerings because of opportunism or
hubris. By overstating earnings before an offering, managers seek to mislead investors and
issue stocks at inflated prices. If investors fail to understand the transitory nature of earnings
management, then when subsequent earnings decline unexpectedly, investors will be
disappointed, and will lower the firm’s assessed value.4 This argument, in contrast to the
Managerial Response hypothesis, does not predict negative returns at the time of the seasoned
equity offerings, because investors do not decipher the earnings management signal in the
offering announcement.
Although earnings management may benefit offering firms’ managers and
shareholders, it also has potential costs that can reduce the incentives for and ability of firms
to manage earnings. First, if earnings overstatement causes losses to investors, there is
increased likelihood of litigation from disgruntled investors and the attendant financial losses.
Second, earnings management that violates generally accepted accounting principles (GAAP)
7
may result in a qualified audit report adversely affecting the firm’s reputation. Third, as shown
by Dechow et al. (1996), firms identified by the Securities and Exchange Commission (SEC)
as violators of financial reporting requirements face an increase in their future costs of capital.
Finally, if illegal manipulation is discovered, managers face loss of reputation, loss of
position, and criminal penalties.
3. Estimating earnings management
Following previous research, I use abnormal accruals in the quarters around an equity
offering announcement as a measure of managerial discretion in reported earnings figures
(Teoh et al., 1998; Rangan, 1998; Defond and Jiambalvo, 1994; Jones, 1991; DeAngelo, 1986;
Healy, 1985). Abnormal accruals (ABNACCit) are defined as the difference between actual and
expected accruals, where expected accruals are estimated using the Jones model (Jones,
1991).5 In the Jones model, expected accruals are estimated after controlling for changes in a
firm’s economic environment. More specifically, the model includes the change in revenues
and gross property, plant and equipment as explanatory variables to control for the portion of
accruals relating to less-discretionary changes in working capital accounts and depreciation
expense. Expected accruals in this model are:
E(accit/ait–1) = β1 (1/ait–1) + β2 (∆revit/ait-1) + β3 (gppeit/ait–1) (1)
where ∆revit is the change in revenues in period t from period t–1; gppeit is the gross property,
plant and equipment at the end of period t; and ait–1 is the book value of total assets at the end
4 There is a growing body of evidence indicating that stock prices do not fully reflect information contained incurrent earnings and accruals (e.g., Bernard and Thomas, 1989; Ball and Bartov, 1996; Sloan, 1996.)5 Total accruals (accit) are defined as ∆Receivablesit (103) + ∆Inventoryit (104) + ∆Accounts payable and accruedliabilitiesit (105) – ∆Taxes payableit (106) + ∆Other current assets and liabilitiesit (107) – Depreciationit (77).Since the above definition uses data from cash flow statements that are available only from 1987 onwards, I usedata from the balance sheet and income statement when data items are unavailable in the cash flow statement.Cash from operations is defined as earnings before extraordinary items (8) less total accruals (accit); quarterlyCOMPUSTAT data items are indicated parenthetically.
8
of period t–1. For each firm-quarter of interest, the model parameters β1, β2 and β3 are
estimated from the following OLS regression, using contemporaneous data of non-offering
firms in the same two-digit SIC industry as the sample firm:
accit/ait–1 = b1 (1/ait–1) + b2 (∆revit/ait–1) + b3 (gppeit/ait–1) + εi (2)
where accit is the actual accruals of firm i in period t, and b1, b2, and b3 are the OLS estimates
of β1, β2, and β3 respectively. In order to have meaningful parameter estimates, I require the
estimation sample to have at least 20 observations and exclude extreme observations using the
DFFITS procedure in SAS.6
Although the paper primarily focuses on the empirical results obtained from the Jones
model, these results remain qualitatively unchanged when abnormal accruals are estimated
using either the modified Jones model developed by Dechow et al. (1995) or the extended
Jones model presented by Jeter and Shivakumar (1999). Moreover, decomposing total accruals
into short-term and long-term accruals as in Teoh et al. does not materially alter any of the
conclusions.
4. Empirical results
4.1 Sample description
The initial sample of equity offerings is obtained from Securities Data Corporation,
and consists of 2,995 seasoned underwritten primary and secondary offerings between January
1983 and December 1992. I exclude firms making shelf offerings and combination offerings
of equity and other securities. Further, utilities and financial companies are excluded as, owing
6 Errors in COMPUSTAT and some infrequent transactions (such as acquisitions and asset sales) affect accrualsor cash flow estimates for firms in the estimation sample. Parameters estimated using these outliers substantiallyincrease the noise in the estimates of abnormal accruals, reducing the power of the tests. When the influentialobservations are not deleted, the statistical significance of the results is weakened but the tenor of the results isunaffected.
9
to greater regulation for these firms, their ability to engage in earnings management might
differ from those of industrial firms. For inclusion in the final sample, the issuing firms must
have accounting data on the 1998 quarterly COMPUSTAT database.
For each equity offering in the sample, the first public announcement of the offering is
obtained by searching the Dow Jones Text and LEXIS online services. To avoid confounding
effects, issuers with another major news release occurring between the day preceding and the
day following the offering announcement day are excluded. The data for stock returns are
from the University of Chicago’s Center for Research in Security Prices (CRSP) daily files.
Finally, as a matter of notation, I define event quarter –1 as the fiscal quarter for which
earnings were last announced before an equity offering announcement. All other quarters are
identified relative to this quarter.
Dechow et al. (1996) suggest that managers of firms that require frequent external
financing will report earnings conservatively to create a positive reputation in the market, from
which they can benefit in subsequent offerings. Since the incentives of firms frequently raising
external capital may differ from those of other firms, frequent issuers are excluded from the
analyses. Frequent issuers are defined as firms with more than one public offering of seasoned
common stock in a two-year period. This definition is based on the belief that an offering
made within two years of another offering may be anticipated at the time of the earlier
offering, which may change a manager’s incentives to engage in earnings management. The
sample of frequent offerings consists of 437 offerings by 191 firms.7
Panel A of Table 1 presents the distribution of the non-frequent offerings across
various years. There are 1,222 equity offerings in the sample. Four of the sample years (1983,
Following Jeter and Shivakumar (1999), I exclude observations with absolute values of DFFITS greater than 2.0.7 Since frequent issuers are not known ex-ante, excluding these firms from the analyses could induce a forward-looking bias. Hence, I repeated the tests with the first offering by each frequent issuer included in the sample of“non-frequent” offerings. This inclusion has no significant impact on the qualitative results.
10
1986, 1991 and 1992) are very active, and contain more than 10% of the sample. The highest
level of equity issuance is in the hot issue market of 1983, which accounts for nearly 24% of
the sample. Panel B of this table summarizes the issue and issuer characteristics. The mean
and median equity capitalization prior to the offerings are $534.8 million and $121.4 million
respectively, indicating the presence of a few large firms in the sample. For the median firm,
an equity offering raises $27 million and results in a 20% increase in the number of shares
outstanding.
4.2 Offering firms’ operating performance
To investigate whether firms overstate earnings around equity offerings, I initially
analyze the abnormal net income of issuers in the eight quarters preceding and the eight
quarters following the offering announcement. Then, given the evidence of earnings
improvements in pre-announcement quarters, I examine the time profile of accruals and cash
flows to evaluate their relative contribution to the increases in earnings.
To control for possible seasonalities, I compute abnormal net income using a seasonal
random walk model. Thus, abnormal net income is defined as the change in net income from
the corresponding quarter of the previous year. Throughout the paper I standardize all
accounting variables by the beginning of period book value of total assets.
Columns 2–4 of Table 2 present the median abnormal net income and the associated
sign tests in event quarters –8 through +7. In addition, Figure 2 plots the median abnormal net
income. I report the medians as they are not likely to be influenced by extreme observations,
although the qualitative results remain unchanged if I consider means rather than medians.
The number of observations for this analysis varies from 745 to 1,126, and tends to be higher
in the post-event quarters because of the improved coverage and greater data availability on
COMPUSTAT in the more recent years. The results show that offering firms experience
11
significant improvements in their earnings in quarters –3 through 0. The median abnormal
earnings vary from 0.12% to 0.18% of total assets during this period. Prior to quarter –3,
median abnormal earnings tend to be insignificant and less than 0.10% in magnitude. The
increase in earnings around the offering is temporary, however, with earnings declining from
quarter +2. From quarter +2, abnormal earnings are all significantly negative. The general
pattern of pre-offering increases in earnings, followed by subsequent earnings declines, is in
line with the findings of Hansen and Crutchley (1990), Loughran and Ritter (1997), and Teoh
et al. (1998).8 This earnings profile is consistent with managers borrowing income from
future periods to enhance earnings immediately around equity offerings. However, this
finding can also be interpreted as managers timing equity issues following periods of
unusually good financial performance, which reflect long-run mean reversion in earnings
(Fama and French, 2000).
Next I examine the relative contributions of cash flow from operations and accruals to
the observed pattern in net income. Columns 5–7 of Table 2 report the median abnormal cash
flows measured using a seasonal random walk model. The results indicate that the net income
profile is not mirrored by cash from operations. The median abnormal cash flows are
insignificant in most event quarters and, if any, are actually negative immediately around the
offering announcement. Therefore, new issues appear to occur when cash flows from
operations are declining and not when they are at a peak. Consequently, the observed
improvements in net income must be driven by the accrual component of earnings. Also, these
findings suggest that, if managers are overstating earnings around equity offerings, they do not
seem to use cash flows to do so.
8 Though not reported, the abnormal sales measured as the change in sales from the corresponding quarter of theprevious year exhibit a similar profile.
12
Columns 8-10 of Table 2 report the median abnormal accruals estimated using the
Jones model. The median abnormal accruals in quarters –8 through –5 are all positive, but not
statistically significant. Between quarters –4 and +4 the medians are significantly positive, and
vary from 0.33% to 1.08% of total assets. The abnormal accruals tend to be much higher
during this period than during all other periods. From quarter +4 onwards, the median
abnormal accruals are insignificantly different from zero.9 This pattern of abnormal accruals
is comparable to those reported in Rangan (1998) and Teoh et al. (1998). Rangan finds
median abnormal accruals to be positive in quarters –4 to +4 around an offering
announcement, although the medians are significant only in the two quarters immediately
following the announcement.10 Similarly, but using annual data, Teoh et al. report
significantly positive abnormal accruals in years –3 to +2 around an offering.
The abnormal increase in accruals around equity offering announcements are
consistent with issuers employing accruals to deliberately overstate earnings. However, neither
the above analysis nor those in Teoh et al (1998) and Rangan (1998) identify reversal of
abnormal accruals in the post-offering period. Although one might expect abnormal accruals
resulting from earnings management to reverse in the post-offering period, the above tests
have low power in detecting it, as the reversals need not happen within a focussed time-period.
Section 4.3 provides corroborative evidence for this argument.
The positive abnormal accruals in quarters 0 through 4 suggest that issuing firms
possibly overstate earnings even after offering announcements. This, to some extent, should be
expected given managerial incentives to avoid reporting reduced earnings immediately after an
offering. For instance, the prospect of litigation following an offering heightens the pressure
9 When issuers are sorted into two equal groups based on their market capitalization, the median abnormalaccruals are significantly positive in quarters –6 through +5 for small firms, while they are significantly positivein quarters –3 through +4 for the large firms.
13
on management to meet the market’s anticipated earnings, particularly if these earnings were
influenced by managerial actions before the offering (that is, by optimistic earnings forecasts
or postponement of bad news announcements). Also, underwriters of an issue may play a role
in earnings management by encouraging the management to report favorable earnings after an
offering in order to maintain their reputation in the market and to maintain their goodwill with
clients.
4.3 Test of abnormal accruals as earnings management proxies
The abnormally high accruals around equity offerings are not unique to the hypothesis
of managers overstating earnings. An alternative, but not mutually exclusive, interpretation is
that managers time equity offerings to follow periods of unexpectedly high accruals. To
identify whether the abnormal accruals are merely the consequence of a timing decision or
represent earnings management, I initially examine whether abnormal accruals systematically
vary across audited and unaudited quarters. I then examine whether abnormal accruals before
offering announcements predict the post-announcement changes in net income.
Unlike intermediate quarters, fourth-quarter results are reported after discussions with
auditors and after audit of the firm’s annual report. The scrutiny of auditors before announcing
fourth-quarter results reduces managerial discretion, making earnings management less likely
in these quarters. In the long run, for the same reason, managers may find it harder to avoid
reversals of earnings management in fourth quarters. In contrast, managers have very little
incentive to reverse earnings management in intermediate quarters. These arguments suggest
systematic differences in the pattern of managed accruals across fiscal quarters. However,
such differences are not expected if abnormal accruals are unrelated to earnings management.
10 Differences in statistical significance between this study and Rangan (1998) are possibly attributable to
14
To test the above predictions, I analyze abnormal accruals in event quarters classified
by fiscal quarter. Event quarters corresponding to a firm’s fourth quarter are classified as
audited quarters, and those corresponding to intermediate quarters are classified as unaudited
quarters.11 The result from this analysis, presented in Table 3, shows that the profile of
abnormal accruals in unaudited quarters is statistically indistinguishable from that reported in
Table 2 for all quarters. The median abnormal accruals are significantly positive in event
quarters – 5 through +5 for this group. In contrast, with the exception of quarter –1, the
median abnormal accruals are insignificant in all the audited quarters before the offering
announcement. The evidence does not suggest earnings management in the audited quarters,
particularly of the magnitude and frequency observed in unaudited quarters. Furthermore, the
median abnormal accruals for the audited group are significantly negative in quarters +4 and
+7, which is consistent with reversal of abnormal accruals from earlier quarters.12 These
results are consistent with managers using accruals to overstate earnings around equity
offerings.
As an additional check for whether abnormal accruals measure earnings management
before offering announcements, I examine the relation between pre-announcement abnormal
accruals and post-announcement changes in net income. If offering firms use abnormal
accruals to borrow income from the future, then a negative relation is expected between
abnormal accruals around the offering and subsequent earnings changes. I test this prediction
by regressing changes in net income on lagged abnormal accruals across the sample firms. In
this regression, I use annual measures of abnormal accruals and earnings changes since the
precise quarters in which earnings management occur are unknown. The annual measures are
differences in sample size. Rangan’s sample consists of 230 firms.11 I checked the issue prospectus of a few randomly selected firms to see whether interim quarters prior to equityofferings are audited. I did not find any evidence of this.
15
constructed by aggregating quarterly data, and year –1 is defined to consist of quarters –4
through –1. Thus, the abnormal accruals in year –1 (ABNACCi(Yr –1)) are computed by
cumulating abnormal accruals in quarters –4 through –1. Finally, the post-offering-
announcement changes in net income are computed relative to net income in year –2 to avoid
spurious correlations that may arise if changes are measured relative to net income in year –1.
Table 4 presents the regression results. The t-statistics for cross-sectional tests in Table
4 and elsewhere in the paper are based on White’s heteroskedasticity consistent estimator for
standard errors. When change in net income in year 0 is the dependent variable, the
coefficients on abnormal accruals are insignificant. However, when change in net income in
year 1 is the dependent variable, the coefficients on both ABNACCi(Yr –1) and on
ABNACCi(Yr 0) are negative and statistically significant. The ability of abnormal accruals in
years –1 and 0 to predict the changes in net income in year +1 is consistent with these
abnormal accruals reversing in year +1. Moreover, the evidence of reversals occurring only in
year +1 and not in year 0 supports the findings reported in Table 3 on accrual reversals
occurring beyond quarter +3. These results are also in line with Rangan (1998) and Teoh et
al. (1998), who document a negative relation between pre-offering abnormal accruals and
post-offering changes in net income.
Overall, the results are consistent with abnormal accruals representing earnings
management. The evidence presented in this section, combined with the positive abnormal
accruals reported in Table 2, supports earnings management around equity offerings and is
consistent with the findings of Rangan (1998) and Teoh et al. (1998). The remainder of the
paper focuses on the investors’ response to this earnings management.
12 The average abnormal accruals are lower for the audited group, relative to the unaudited group in eventquarters –3 to +7, although the difference is not significant in quarter +3.
16
4.4 Market reaction to earnings announcements around equity offerings
In order to examine whether offering announcements signal earnings overstatements, I
analyze investors’ price response to earnings released around offering announcements. Before
an offering announcement, investors would be unaware of the heightened incentives of issuers
to overstate earnings, and as such would price unexpected earnings of all firms, both issuers
and non-issuers, as if they included an average amount of discretion. This suggests that the
earnings overstatements by issuers will lead to positive earnings surprises and positive market
reaction at earnings releases before an offering announcement.
However, an offering announcement can signal the increased incentives for issuers to
overstate earnings. This will aid market participants to better forecast subsequent earnings and
may dampen their price response to earnings surprises in the post-announcement quarters.
This suggests that the price reaction to earnings releases will be systematically lower in the
post-announcement quarters. In contrast, a finding of little change in investors’ response after
offering announcements would support the notion that investors are naive and fail to recognize
earnings management by issuers. This section tests these predictions by analyzing the earnings
announcement returns and the earnings response coefficients in the quarters before and after
an offering announcement.
4.4.1 Earnings Announcement Returns
Earnings announcement returns for quarter k are defined as the cumulative abnormal
returns during a six-day window (days –1 through +4) around the earnings announcement date
(day 0). The six-day window for earnings announcement returns is intended to capture the
delay in the market’s response to earnings announcements, and is based on the findings of
17
Bernard and Thomas (1989) that a disproportionately large percentage of the price response
occurs within five days of an earnings announcement.
The abnormal returns are computed as the difference in returns between the sample
firm and a growth-matched firm. This return-metric is motivated by the fact that a larger
proportion of issuers tend to be growth firms when compared with non-issuers and growth
firms are known to have smaller average returns, particularly around earnings announcements
(Lakonishok et al., 1994; La Porta et al., 1997).13 The growth-matched firm is chosen
amongst all firms that have market capitalization at the end of quarter –1 within 30% of the
sample firm and whose growth rate of sales (measured over quarters –8 to –1) is closest to that
of the sample firm.
The results presented in Table 5 show earnings announcement returns to be
significantly positive in quarters –5 to –1. This is consistent with average unanticipated
earnings being positive before offering announcements and with investors pricing the
unanticipated earnings. However, in almost all quarters following the offering announcement,
the mean earnings announcement returns are insignificantly different from zero and there is no
discernable pattern. Even when earnings announcement returns are averaged across post-
announcement quarters, the returns remain indistinguishable from zero. Finally, these results
are robust across both large and small firms and across firms sorted on ABNACCi(Yr –1).
The significantly positive market reactions to earnings released before an offering
announcement, followed by insignificant market reactions following the offering
announcement, suggest that the offering announcement conveys information to investors
regarding future earnings. These results do not support the view that investors are disappointed
with earnings released after an offering announcement, and are consistent with the findings of
18
Brous et al. (1999) and Hansen and Sarin (1998). In analyses similar to those reported here,
but using alternative benchmarks for earnings announcement returns, Brous et al. (1999) show
that investors do not suffer systematic losses at earnings releases following an equity issue.
Further, Hansen and Sarin (1998) show that analysts’ forecasts for issuers are not overly
optimistic compared with similar growth firms.14
Overall, the findings are consistent with offering announcements signaling earnings
overstatement to investors and causing investors to revise their beliefs about future earnings.
There is no evidence to suggest that investors fail to undo earnings management by issuers or
are systematically disappointed at earnings releases following an offering announcement.
4.4.2 Earnings Response Coefficients
As an alternative test of whether investors change their response to earnings released
after an offering announcement, I analyze the earnings response coefficients around offering
announcements. The earnings response coefficients are estimated from a pooled regression of
raw earnings announcement returns (measured as cumulative returns in the six-day earnings
announcement period) on abnormal net income in quarters –4 through +3.15 This regression
includes interactive dummy variables for the post-announcement quarters so as to identify any
changes in the response coefficients associated with the offering announcement. The
regression also includes the logarithm of market capitalization (SIZEi) as a control variable.
13 The median issuer in the sample has sales growth of 39% in the two years prior to the offering announcement,while the median two-year-growth rate of all firms on COMPUSTAT is only 16% during the sample period.14 I also analyze earnings announcement returns measured either as market-adjusted returns or as size-and-book-to-market adjusted returns. The size-and-book-to-market-adjusted returns are computed as the difference inreturns between the issuer and a non-issuer matched by size and book-to-market ratio. Consistent with Jegadeesh(1999) and Denis and Sarin (1999), earnings announcement returns based on these metrics are significantlynegative in the post-offering period. However, consistent with Loughran and Ritter (1995), similar results areobtained even during non-earnings-announcement periods. This indicates a potential bias in these return-metricsfor studies of market response to earnings announcements.15 The remainder of the paper is based on raw earnings announcement returns, although qualitatively identicalresults are obtained when alternative measures for earnings announcement returns are used.
19
Table 6 presents the regression results. The earnings response coefficient in the pre-
announcement quarters is about 0.15 and is statistically significant, indicating that investors
price abnormal earnings in these quarters. However, in the post-announcement quarters the
response coefficient decreases significantly, and is only about 0.06: a decline of more than half
from the pre-announcement quarters. This decline in earnings response coefficients indicates
that investors react less to unexpected earnings released after offering announcements and
suggests that investors perceive earnings in this period to contain less price-relevant
information. This finding is consistent with investors inferring earnings management by
issuers, and accordingly reducing their response to unexpected earnings in the post-
announcement quarters.
4.5 Earnings management and the market reaction to equity-offering announcements
Given that announcements of seasoned equity offerings cause investors to rationally
change their beliefs about subsequently reported earnings figures, this section investigates
whether the offering announcements also cause investors to correct misvaluations caused by
earlier earnings management. Announcement of an SEO may cause investors to revise upward
the probability that prior earnings numbers were overstated and, as a consequence, to lower
the firm’s stock price. This argument suggests a negative relation between the market’s price
reaction to offering announcements and pre-announcement earnings management. I test this
prediction by regressing the price reaction at offering announcements on proxies for the
earnings management. The specific proxies I consider are the earnings announcement returns
and abnormal accruals in year –1.
The price reaction to the offering announcement, EQRETi, is computed as the
cumulative returns on the day of and the day preceding the announcement. The mean two-day
price reaction to SEO announcements in the sample is –2.1%. The raw earnings announcement
20
returns (SDRETi(Yr –1)) and the abnormal accruals (ABNACCi(Yr –1)) in year –1 are
computed by summing the corresponding quarterly variables in quarters –4 through –1. To
control for issue and issuer characteristics, the regressions include the ratio of number of
shares offered to the number of shares outstanding before the offering (OFFSIZEi), the
individual stock returns (RUNUPi), and the market returns (MRUNUPi) in the 60 trading days
before the offering announcement.16 If investors correct earlier mispricing at offering
announcements, then a negative coefficient is expected both for the earnings announcement
returns and for abnormal accruals in these regressions.
Table 7 presents the regression results. Regression I, which uses only SDRETi(Yr –1)
as an explanatory variable, shows that the coefficient estimate for SDRETi(Yr –1) is –0.02
with a t-statistic of –2.01. The significantly negative coefficient is consistent with investors
reversing the price impact of earlier earnings surprises at offering announcements. Moreover,
the results in Regression II show that the coefficient for SDRETi(Yr –1) continues to be
significantly negative even when control variables are included in the regression. Furthermore,
the coefficient on RUNUPi is found to be significantly negative in Regression II, which is
consistent with the findings of Masulis and Korwar (1986).
In regressions III and IV, which include ABNACCi(Yr –1) as an independent variable,
the coefficient on ABNACCi(Yr –1) is about –0.03 and statistically significant. This result is
consistent with results based on SDRETi(Yr –1), and shows that, irrespective of the proxy used
to measure earnings management, a significant negative relation is observed between market
price reaction to the offering announcements and prior earnings management. These findings
16 I also consider size, measured as logarithm of equity capitalization, and book-to-market ratio as additionalcontrol variables. These variables have insignificant coefficients in the regression, and their inclusion has noqualitative effect on the reported results.
21
support the notion that investors rationally infer earnings management from offering
announcements and undo the effects of the earnings management at offering announcements.17
4.6 Earnings management and post-offering stock underperformance
In contrast to the above findings, Rangan and Teoh et al., based on their analysis of
pre-offering abnormal accruals and post-offering stock returns, document evidence consistent
with investors failing to recognize earnings management by issuers. Rangan and Teoh et al.
argue that investors naively extrapolate pre-issue earnings increases, resulting in overvaluation
of offering firms. Further, they argue that when earnings management reverses in the post-
offering period and earnings decline, investors are disappointed and correct their valuation
errors. Consistent with their arguments, they find pre-issue abnormal accruals to predict the
abnormal returns in the one to four years following equity offerings.
One possible explanation for the apparently contradictory evidence between this paper
and those of Rangan and Teoh et al. may be that my finding on market correction at offering
announcements indicates only a partial correction and the finding of an insignificant earnings
announcement return following offering announcements is due to issuers pre-releasing
unfavorable earnings information. However, an alternative explanation, attributable to the
criticisms leveled by Kothari and Warner (1997), and Barber and Lyon (1997) on studies of
long-horizon event studies, is that the evidence documented in Rangan and Teoh et al. is
influenced by the choice of return metric and methodology. This is notwithstanding the fact
that recent studies (e.g., Fama, 1998; Mitchell and Stafford, 2000; Brav et al., 2000; Eckbo et
17 When the regressions are estimated separately for firms sorted into two groups based on their marketcapitalization at the end of quarter –1, negative coefficients are observed on SDRETi(Yr –1) and on ABNACCi(Yr–1) for both groups. However, the statistical significance of the results is weakened due to fewer observations inthese regressions.
22
al., 2000) cast doubts about the very existence of stock underperformance following seasoned
equity offerings, which Rangan and Teoh et al. attempt to explain.
Rangan (1998) and Teoh et al. (1998) use long-run abnormal returns measured either
as the market-adjusted returns or as the prediction errors from the Fama-French model.
However, Barber and Lyon (1997) and Kothari and Warner (1997) show that statistical tests
based on these measures of abnormal returns are severely mis-specified owing to, among other
factors, skewness in the long-horizon returns data. Furthermore, Kothari et al. (2000) show
that the skewness in long-horizon returns combined with data attrition, either because of firms’
survival or deletion of extreme observations, can induce a statistically significant association
between ex ante forecast variables (such as abnormal accruals) and ex post security returns.
Hence, I test the robustness of the results in Rangan and Teoh et al. to alternative
methodologies that are less susceptible to these biases.
This section initially replicates the Rangan and Teoh et al. results in event time and
then tests the sensitivity of these results to the use of three alternative methodologies: (1) a
control-firm approach, (2) a calendar-time portfolio approach, and (3) a Fama-MacBeth panel
procedure. To be consistent with Teoh et al., the event quarters in this section are defined
relative to the offering date rather than to the offering announcement date.
4.6.1 Event-time regressions
In order to investigate whether pre-offering earnings management causes poor post-
offering stock performance, Rangan and Teoh et al. estimate regressions of the post-offering
stock returns on pre-offering abnormal accruals in event time. Following their methodology, I
regress post-offering stock returns, measured either as raw returns or as market-adjusted
returns, on abnormal accruals in year –1, ABNACCi(Yr –1). As before, year –1 is defined as
consisting of event quarters –4 through –1, and the annual values for the variables are
23
computed by cumulating the constituent quarterly values. The post-offering stock returns are
computed as the buy-and-hold returns over one-, two- and four-year horizons, beginning the
day after the offering date. If a sample firm is delisted, its stock returns are set to zero for the
rest of the period. The results are robust to setting the returns of delisted firms equal to the
value-weighted market returns, as well as to truncating, rather than filling, the returns of
delisted firms. Finally, the regressions include logarithm of market capitalization (SIZEi) and
the book-to-market ratio (BMi) before the offering as control variables.
The estimates from the cross-sectional regressions are presented in Table 8. Focusing
on columns (1) to (6), I find the coefficient estimate on abnormal accruals to be insignificant
in regressions of the one-year-ahead returns. However, this result changes when returns are
measured over two or four years following the offering. The coefficients on ABNACCi(Yr –1)
are significantly negative for both raw returns and the market-adjusted returns. This is
consistent with the findings of Rangan and Teoh et al. Moreover, the coefficient on SIZEi is
significantly positive, which is consistent with the finding of Brav et al. (2000) that the post-
offering stock underperformance is concentrated in the group of smallest issuers.
However, as noted earlier, long-horizon tests using raw returns and market-adjusted
returns are known to be mis-specified. To control for this mis-specification, I initially use the
control-firm approach suggested by Barber and Lyon (1997). Under this approach, abnormal
returns are defined as the return of the sample firm less the return on a control firm, matched
by size and book-to-market ratio. The control firm is chosen amongst all firms that have
market values of equity between 70% and 130% of that of the sample firm and whose book-to-
market ratio is closest to that of the sample firm.18 All variables for this matching are
18 The qualitative results remain unchanged when the control firms are chosen by matching issuers with non-issuers based on size and growth in sales in the two-years before the offering.
24
measured as of quarter –1. Barber and Lyon (1997) show that long-horizon tests based on this
abnormal return metric are well-specified.
The results from this control firm approach are presented in columns (7) to (9) of Table
8. Irrespective of the duration over which the dependent variable is measured, the coefficient
on ABNACCi(Yr –1) is insignificant. Moreover, the magnitude of the coefficient on
ABNACCi(Yr –1) is much lower in this regression compared with coefficients from
regressions using raw or market-adjusted returns. This indicates that the relation between pre-
offering abnormal accruals and post-offering stock returns is sensitive to the abnormal return
metric used in the regression.
One possible explanation for the lack of significance under the control firm approach is
the lack of power in the test. I investigate this argument further by analyzing firms sorted on
market-capitalization, as Teoh et al. find the strongest relation between pre-offering accruals
and post-offering returns in their subsample of smallest issuers. The results from analyzing
firms sorted into two equal-sized groups are presented in columns (10) and (11) of Table 8.19
The results for both small and large firms are qualitatively similar to those discussed above.
For both subsamples, there is no significant relation between ABNACCi(Yr –1) and post-
offering stock performance under the control-firm approach. 20
4.6.2 Calendar-time portfolio approach
As an alternative to the control-firm approach, I use the calendar-time portfolio method
advocated by Fama (1998) and employed in previous research by Jaffe (1974), Mandelker
(1974), Loughran and Ritter (1995), and Brav and Gompers (1997) among others. Under this
19 To conserve space, only the regression results for four-year-ahead returns are reported. However, qualitativelyidentical results are obtained from using one-year- and two-year-ahead returns in the analysis.20 I repeat the regression analyses using annual compustat data, as done in Teoh et al. The use of annual dataincreases the number of observations in the analyses by about 40%, but leaves the conclusions unaltered.
25
approach, quintiles are formed for each month from March 1983 to October 1996 by sorting
all sample firms that have issued equity in the previous 48 months based on ABNACCi(Yr –
1).21 The equal-weighted returns are then computed for each quintile and examined for
systematic variation across the quintiles. A monotonic increase in the risk-adjusted returns
from the highest to the lowest quintile provides evidence consistent with the ability of
ABNACCi(Yr –1) to predict post-offering returns.22
Relative to the event-time regressions, this approach has at least two advantages. First,
as pointed out by Fama (1998), this approach corrects for correlation of returns across events
that are not absorbed by the model used to adjust for expected returns. Second, skewness is
not an issue here, as this approach uses monthly returns of portfolios.
To set the stage, I initially analyze the average raw returns, size (measured as
logarithm of equity capitalization), and book-to-market ratio across the abnormal accrual
quintiles. The size and the book-to-market ratio are measured at the end of quarter –1. For
each month, the quintile values are obtained by averaging the variables across constituent
firms.
Panel A of Table 9 presents the time-series averages of raw returns, size, and book-to-
market ratios for each quintile. The average return for the lowest abnormal accrual quintile is
0.87%, while it is only 0.16% for the highest quintile. However, the returns do not
monotonically decrease across the quintiles. Quintile 2 has the highest average return, at
1.35%, and the returns for both quintiles 3 and 4 are about 0.9%. Nonetheless, the difference
in returns between the extreme quintiles, which can be viewed as the payoff from a strategy of
buying stocks in the lowest quintile and shorting stocks in the highest quintile, is economically
21 The average month contains 161 observations. The two months before March 1983 and after October 1996 areomitted, as these months do not have sufficient observations to form quintiles.
26
large (0.71%) and statistically significant. Moreover, the payoffs to this strategy are positive in
nearly 58% of the months.
Looking at the average size and book-to-market ratios for the quintiles, it becomes
apparent that the above strategy of buying quintile 1 and shorting quintile 5 may not be risk-
free. Relative to firms in quintile 1, the firms in quintile 5 tend to be smaller and to have a
lower book-to-market ratio. To control for differences in the risk exposure across the
quintiles, I regress the monthly quintile returns (in excess of risk-free rate) on Fama-French
factors (namely, the excess market return, SMB and HML) and then analyze the intercept from
these regressions.23 The regressions are carried out separately for each quintile.
Table 9 reports the estimated intercepts with the associated t-statistics. The intercepts
from the Fama-French model are insignificant for quintiles 1 to 4 and no pattern is discernible
across these quintiles. However, the intercept is significantly negative for the highest
abnormal accrual quintile. Also, the difference in the intercepts across extreme quintiles is a
significant 0.0066. This suggests that the strategy of buying quintile 1 and shorting quintile 5
yields a monthly excess profit of 0.66% after controlling for risk exposures associated with
Fama-French factors. This evidence is consistent with the arguments of Teoh et al (1998)
and Rangan (1998) that earnings management prior to stock offerings leads to stock
underperformance in the post-offering period and supports managerial opportunism as a
reason for the earnings management.
However, this result needs to be interpreted with caution for at least two reasons. First,
the above regressions assume that loadings on the Fama-French factors are constant over time.
This assumption is untenable given that the composition and number of firms in the quintiles
22 Loughran and Ritter (2000) argue that the calendar-time approach reduces the power of tests of long-run stockperformance, since it weights each month equally, and ignores the possibility that the number of equity issues in amonth could be related to the level of stock misvaluations.23 I thank Gene Fama for providing the data on Fama-French factors.
27
change constantly. Secondly, firms in the highest abnormal accrual quintile are dominated by
small and low book-to-market firms. For instance, over 70% of the firms in the highest
abnormal accrual quintile are also in the lowest Fama-French (1993) book-to-market
quintile.24 Fama-French (1993, 1996) show their model to be mis-specified for these firms and
to yield average excess returns of as much as –0.45% per month. Although the samples are
not directly comparable across this study and Fama and French (1993, 1996), their results
indicate that the negative intercept in Table 9 could, at least partly, reflect this mis-
specification.25
Due to the above limitations, I use an alternative approach to test whether differences
in raw returns across extreme abnormal accrual quintiles persists after controlling for
differences in firm characteristics. Specifically, I regress the return differential (Rt,low-high)
across the extreme quintiles on differences in their size ((SIZEt,low-high) and book-to-market
ratios (BMt,low-high). Since the composition and number of firms in the event quintiles change
through time, the precision of the quintile returns will also vary over time. Hence, to increase
the signal-to-noise ratio, I use the information contained in the cross-sectional standard errors
of monthly quintile returns and estimate a generalized least-square regression assuming
independence across quintile returns.26 The coefficient estimates from the regression are as
follows (t-statistics in parentheses below coefficient estimates)
Rt,low-high = 0.002 + -0.12 (SIZEt,low-high) + 0.056 (BMt,low-high) (3)(0.58) (-0.79) (1.92)
24 The percentage is obtained by comparing book-to-market ratios of sample firms with the book-to-marketvalues presented in Table 1 of Fama-French (1993). Fama-French (1993) present the mean book-to-market fortheir 25 size-and-book-to-market sorted firms. This mean is then averaged across neighboring portfolios toobtain the upper and lower limits for each portfolio and these limits are used to categorize sample firms into thevarious Fama-French portfolios.25 23% of firms in the highest abnormal accrual quintile are in the Fama-French’s smallest size and lowest book-to-market quintiles. For the lowest abnormal accrual quintile, 58% of the firms are in the lowest Fama-French(1993) book-to-market quintile and 13% are in the smallest size and lowest book-to-market quintile. Forquintiles 2 to 4, the percentages are similar to those for the lowest abnormal accrual quintile.26 The result from ordinary least-square regression is qualitatively similar.
28
The insignificant intercept in the regression shows that the returns across the abnormal
accrual quintiles are not significantly different from each other after controlling for differences
in firm characteristics. Moreover, the coefficient on the book-to-market ratio is significantly
positive, suggesting that the return differential observed across the extreme quintiles is
attributable to differences in their book-to-market ratios. This also suggests that the strategy of
buying the lowest quintile and shorting the highest quintile has a significant exposure to the
risk factor associated with book-to-market ratio. Finally, the coefficient on size in this
regression is negative, but insignificant.27 These results indicate that pre-offering abnormal
accruals do not predict the post-offering risk-adjusted returns.28
4.6.3 Fama-MacBeth panel procedure
As a final robustness test of the results in Rangan and Teoh et al., I use the Fama-
MacBeth procedure to investigate the relation between pre-offering abnormal accruals and
post-offering stock returns. Under this approach, I estimate cross-sectional regressions of
monthly returns of firms that have issued equity in the prior 48 months on abnormal accruals
in Year –1 (ABNACCi(Yr –1)), logarithm of market capitalization (sizei), and book-to-market
ratio (bmi) at the end of quarter –1. The regressions are estimated separately for each month in
the period March 1983 through October 1996. The time series of coefficients from these
monthly regressions is then analyzed to identify the relation between ABNACCi(Yr –1) and the
stock returns following equity offerings.
27 The insignificant coefficient on size is consistent with the absence of size premium in the post-1983 period(Jegadeesh and Titman, 2000).28 I repeat the above analyses by value-weighting (rather than equally-weighting) firms in each quintile. Thequalitative results remain unaffected by this change. For instance, when firms are value-weighted, the differencein raw returns across extreme quintiles is 1.20% (t-stat=3.25) and the difference in intercepts from Fama-Frenchmodel is 0.01 (t-stat=2.61). Finally, in the regression of return differentials on differences in firm characteristics,the intercept is –0.005 (t-stat=-0.80) and the coefficient on BMtlow-high is 0.031 (t-stat=2.26).
29
An advantage of the Fama-MacBeth procedure is that it does not assume time-invariant
risk premiums, an assumption implicit in Regression (3) under the calendar-time portfolio
approach. Moreover, as with the calendar-time portfolio approach, cross-correlation of returns
across events and skewness in returns are not of much concern here.
Table 10 presents the averages of the coefficients from the monthly cross-sectional
regressions, together with the t-statistics computed as the mean coefficient estimate divided by
its time-series standard error. The average coefficients on all the explanatory variables are
insignificant, with the exception of the book-to-market ratio. The book-to-market ratio has a
statistically positive coefficient of 0.006, which is consistent with previous studies (Fama and
French, 1992). The insignificant coefficient on ABNACCi(Yr –1) shows that pre-offering
abnormal accruals do not predict future returns above and beyond book-to-market-ratio and
size. Further, for an average issuer with ABNACCi(Yr –1) of 5%, the coefficient of –0.012
implies an economically insignificant excess return of –0.06% per month. These findings are
consistent with the results reported using the calendar-time portfolio.
Similar to the above regressions, Teoh et al. also estimate monthly cross-sectional
regressions of post-offering returns on lagged abnormal accruals. However, in contrast to the
above findings, Teoh et al. report (in their table 8) a significantly negative coefficient for
abnormal accruals from their analysis.29 The difference in the findings between the two studies
is attributable to a bias in Teoh et al.’s computation of t-statistics, which is effectively
computed as the average t-statistic across the monthly regressions, multiplied by the square
root of the number of monthly regressions (√N). Multiplying the t-statistics by √N biases the
magnitude of their t-statistics upward. Adjusting for this bias reveals the magnitude of the
mean t-statistic to be less than 1.0 in their regressions.
30
Overall, the evidence in this study indicates that issuers overstate earnings before
equity offerings, and that investors unravel this earnings management well before the actual
offering. These findings are consistent with the Managerial Response hypothesis and suggest
that earnings management before seasoned equity offerings is more likely the issuers’
response to investors’ expectations than indicative of managerial opportunism..
5. Summary and conclusions
This paper analyzes whether firms overstate earnings before seasoned equity offerings
and whether, at offering announcements, investors recognize and undo the effects of such
earnings management. Consistent with earnings management, net income and accruals are
abnormally high around equity offerings and pre-offering abnormal accruals predict
subsequent declines in net income. However, investors appear to rationally infer this earnings
management at equity offerings announcements and, as a result, reduce their price response to
unexpected earnings released after offering announcements. Also, at the offering
announcement, investors seem to correct the price impact of earlier earnings management, as
evidenced by a negative relation between pre-announcement abnormal accruals and the stock
price reaction to the offering announcement.
In contrast to the above findings, Rangan (1998) and Teoh et al. (1998) document a
negative relation between pre-offering abnormal accruals and post-offering abnormal stock
returns, which they interpret as suggesting failure on the part of investors to recognize
earnings management causing post-offering stock underperformance. However, the statistical
tests based on the abnormal return metrics used in these studies have been shown to be biased
(Kothari and Warner, 1997; Barber and Lyon, 1997). Hence this paper reexamines the
29 The coefficient on lagged abnormal accruals in Teoh et al.’s regressions vary from –0.693 to –5.738, which
31
evidence presented in Rangan and in Teoh et al. and finds their results to depend crucially on
their choice of abnormal return metrics. Their results are not robust to alternative
methodologies that are known to be relatively well specified.
Overall, the results presented here indicate that investors unravel earnings management
well before an equity offering, which at first glance seems to suggest that earnings
management by issuers is wasteful on average. However, using a rational expectations
framework, this paper shows that earnings management by issuers, rather than being intended
to mislead investors, may actually be the rational response of issuers to anticipated market
behavior at offering announcements. In a world with managerial discretion over accounting
numbers, earnings management by issuers and subsequent price reversal by investors appears
to be the unfortunate outcome.
translates to –0.007 to –0.057 when the dependent variable is not expressed in percentage, as done in this study.
References
Aharoney, J., Lin, C., Loeb, M., 1993. Initial public offerings, accounting choices and earnings
management. Contemporary Accounting Research 10, 61–81.
Ball, R., Bartov, E., 1996. How naïve is the stock market’s use of earnings information? Journal
of Accounting and Economics 21, 319-37.
Barber, B., Lyon, J., 1997. Detecting long-run abnormal stock returns: the empirical power and
specifications of test statistics. Journal of Financial Economics 43, 341–372.
Bernard, V., Thomas, J., 1989. Post-earnings-announcement drift: Delayed price response or risk
premium? Journal of Accounting Research 27, 1–33.
Brav, A., Gompers, P., 1997. Myth or reality? The long-run underperformance of initial public
offerings: Evidence from venture and non-venture capital-backed companies. Journal of
Finance 52, 1791-1821.
Brav, A., Geczy, C., Gompers, P., 2000. Is the abnormal return following equity issuances
anomalous? Journal of Financial Economics 56, 201-249.
Brous, P., Datar, V., Kini, O., 1999. Is the market optimistic about the future earnings of
seasoned equity offerings? Working paper, Seattle University, July.
DeAngelo, L., 1986. Accounting numbers as market valuation substitutes: a study of
management buyouts of public stockholders. The Accounting Review 61, 400–420.
Dechow, P., Sloan, R., Sweeney, A., 1995. Detecting earnings management. The Accounting
Review 70, 193–225.
Dechow, P., Sloan, R., Sweeney, A., 1996. Causes and consequences of aggressive financial
reporting. Contemporary Accounting Research 13, 1–36.
Defond, M., Jiambalvo, J., 1994. Debt covenant violation and manipulation of accruals. Journal
of Accounting and Economics 17, 145–176.
Denis, D., Sarin, A., 1999. Is the market surprised by poor earnings realizations following seasoned equity
offerings? Working paper, Purdue University, December.
Eckbo, B., Masulis, R., Norli, O., 2000. Conditional long-run performance following security
offerings: Is there a new issues puzzle? Journal of Financial Economics 56, 251-291.
Erickson, M., Wang, S., 1999. Earnings management by acquiring firms in stock for stock
mergers. Journal of Accounting and Economics 27, 149-176.
Fama, E., 1998. Market efficiency, long-run returns and behavioral finance. Journal of Financial
Economics 49, 283–306.
Fama, E., French, K., 1992. The cross-section of expected stock returns. Journal of Finance 47,
427-466.
Fama, E., French, K., 1993. Common risk factors in the returns on stocks and bonds. Journal of
Financial Economics 33, 3-56.
Fama, E., French, K., 1996. Multifactor explanations of asset pricing anomalies. Journal of
Finance 51, 55-84.
Fama, E., French, K., 2000. Forecasting profitability and earnings. Journal of Business 73, 161-
176.
Hansen, R., Crutchley, C., 1990. Corporate earnings and financings: an empirical analysis.
Journal of Business 63, 347–371.
Hansen, R., Sarin, A., 1998. Are analysts overoptimistic around seasoned equity offerings?
Working paper, Virginia Tech, November.
Healy, P., 1985. The effect of bonus schemes on accounting decisions. Journal of Accounting and
Economics 7, 85–107.
Jaffe, J., 1974. Special information and insider trading. Journal of Business 47, 411-428.
Jegadeesh, N., 1999. Long-run performance of seasoned equity offerings: Benchmark errors and
biases in expectations. Working paper, University of Illinois at Urbana Champaign, March.
Jegadeesh, N., Titman, S., 2000. Profitability of momentum strategies: An evaluation of alternative
explanations. Working paper, National Bureau of Economic Research, June.
Jeter, D., Shivakumar, L., 1999. Cross-sectional estimation of abnormal accruals using quarterly
and annual data: Effectiveness in detecting event-specific earnings management.
Accounting and Business Research 29, 299-319.
Jones, J., 1991. Earnings management during import relief investigations. Journal of Accounting
Research 29, 193–228.
Kothari, S., Warner, J., 1997. Measuring long-horizon security price performance. Journal of
Financial Economics 43, 301–339.
Kothari, S., Sabino, J., Zach, T., 2000. Implications of data restrictions on performance
measurement and tests of rational pricing. Working paper, Massachusetts Institute of
Technology, August.
Lakonishok, J., Shleifer, A., Vishny, R., 1994. Contrarian investment, extrapolation and risk.
Journal of Finance 49, 1541-1578.
La Porta, R., Lakonishok, J., Shleifer, A., Vishny, R., 1997. Good news for value stocks: Further
evidence on market efficiency. Journal of Finance 52, 859-874.
Loughran, T., Ritter, J., 1995. The new issues puzzle. Journal of Finance 52, 23–51.
Loughran, T., Ritter, J., 1997. The operating performance of firms conducting seasoned equity
offerings. Journal of Finance 52, 1823–1850.
Loughran, T., Ritter, J., 2000. Uniformly least powerful tests of market efficiency. Journal of
Financial Economics 55, 361-389.
Mandelker, G., 1974. Risk and return: The case of merging firms. Journal of Financial
Economics 1, 303-335.
Masulis, R., Korwar, A., 1986. “Seasoned equity offerings: An empirical investigation. Journal of
Financial Economics 15, 91–118.
Mitchell, M., Stafford, E., 2000. Managerial decisions and long run stock price performance.
Journal of Business 73, 287-320.
Myers, S., Majluf, N., 1984. Corporate financing and investment decisions when firms have
information that investors do not have. Journal of Financial Economics 13, 187–221.
Perry, S., Williams, T., 1994. Earnings management preceding management buyout offers.
Journal of Accounting and Economics 20, 293–316.
Rangan, S., 1998. Earnings before seasoned equity offerings: Are they overstated? Journal of
Financial Economics 50, 101–122.
Sloan, R., 1996. Do stock prices fully reflect information in accruals and cash flows about future
earnings? Accounting Review 71, 289-315.
Spiess, D., Affleck-Graves, J., 1995. Underperformance in long-run stock returns following
seasoned equity offerings. Journal of Financial Economics 38, 243–267.
Stein, J., 1989. Efficient capital markets, inefficient firms: A model of myopic corporate
behavior. Quarterly Journal of Economics 104, 655-669.
Teoh, S., Welch, I., Wong, T., 1998. Earnings management and the post-issue underperformance
of seasoned equity offerings. Journal of Financial Economics 50, 63–100.
Table 1Descriptive statistics on equity offerings and offering firms
Panel A Distribution by calendar year
Year No. ofofferings
Frequency
1983 288 23.61984 66 5.41985 119 9.71986 154 12.61987 117 9.61988 46 3.81989 68 5.61990 52 4.31991 162 13.31992 150 12.3
Panel B Issuer and issue characteristics
Totalassetsa
Marketvaluea
Bookvaluea
Offeringamount
Offeringsizeb
Mean 801.1 534.8 316.3 45.5 0.24Median 97.8 121.4 42.9 27.0 0.20Standarddeviation
3453.7 2278.6 1999.6 64.3 0.18
The sample consists of 1222 underwritten seasoned offerings of common stock by industrial firms overthe period 1983 through 1992. The sample excludes offerings made within two years of each other.a The total assets, book value of equity and the market value of equity of the firms are measured at the endof the quarter –1.b Offering size is computed as the number of shares offered divided by the number of shares outstandingbefore the offering.
Table 2Median abnormal net income, cash from operations and accruals around equity offering announcements
Abnormal net income Abnormal cash from operations Abnormal AccrualsEvent quarter No. of
obs.Median
(percent)Sign test(p-value)
No. ofobs.
Median(percent)
Sign test(p-value)
No. ofobs.
Median(percent)
Sign test(p-value)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)-8 745 0.01 0.88 540 -0.44 0.25 494 0.21 0.34-7 776 -0.12 0.02 573 0.31 0.40 508 0.05 0.89-6 801 -0.03 0.24 578 0.06 0.84 557 0.43 0.15-5 835 -0.02 0.37 620 0.22 0.07 591 0.27 0.18-4 859 0.01 0.73 647 0.30 0.10 645 0.42 0.03-3 903 0.12 0.00 688 0.03 0.91 698 0.33 0.04-2 970 0.18 0.00 720 0.10 0.53 751 0.76 0.00-1 1027 0.13 0.00 778 -0.04 0.86 767 1.08 0.000 1073 0.14 0.00 815 -0.52 0.02 765 0.63 0.001 1103 0.01 0.76 866 -0.41 0.10 770 0.94 0.002 1125 -0.14 0.00 890 -0.20 0.24 782 0.58 0.003 1126 -0.16 0.00 908 0.11 0.53 779 0.83 0.004 1110 -0.26 0.00 904 -0.21 0.48 792 0.47 0.005 1091 -0.19 0.00 898 -0.11 0.40 807 0.30 0.176 1068 -0.21 0.00 876 0.38 0.02 808 0.21 0.227 1063 -0.20 0.00 885 0.06 0.69 782 0.16 0.33
The abnormal part of net income and cash from operations is estimated using a seasonal random walk model. Abnormal accruals are estimatedusing the Jones model for expected accruals
Jones model: ABNACCi(Qtr t) = accit/ait–1 – [β1 (1/ait–1) + β2 (∆revit/ait–1) + β3 (gppeit/ait–1)]where ABNACCi(Qtr t) is the abnormal accruals for firm i in quarter t; accit is accruals for firm i in quarter t; ait–1 is the book value of total assets atthe end of quarter t–1; ∆revit is the change in sales from quarter t–1 to quarter t; and gppeit is gross property, plant, and equipment at end of quartert. For each sample firm and event quarter, the parameters β1 through β3 are estimated across all firms in the same two-digit SIC as the sample firm,and have the data available for the corresponding quarter. All variables are standardized by the book value of total assets at the beginning of thequarter.
Table 3Median abnormal accruals for quarters that are audited and for quarters that are unaudited
Unaudited AuditedEvent quarter No. of
obs.Median
(percent)Sign test(p-value)
No. ofobs.
Median(percent)
Sign test(p-value)
-8 402 0.22 0.40 92 0.08 0.75-7 372 0.47 0.11 136 -0.59 0.22-6 381 0.20 0.39 176 -0.31 0.26-5 423 0.34 0.04 168 0.02 1.00-4 542 0.43 0.04 103 0.27 0.69-3 521 0.38 0.04 177 0.18 0.65-2 523 0.94 0.00 228 0.15 0.95-1 532 1.41 0.00 235 0.69 0.030 612 0.77 0.00 153 -0.29 0.421 568 1.22 0.00 202 -0.05 0.942 542 0.98 0.00 240 -0.16 0.953 545 1.04 0.00 234 0.27 0.274 639 0.69 0.00 153 -0.71 0.055 588 0.69 0.00 219 0.05 1.006 552 0.39 0.14 256 -0.15 0.427 541 0.28 0.21 241 -0.43 0.04
For each event quarter the sample firms are classified into audited and unaudited groups depending on whether the financial statements are for thefourth (financial) quarter or not. The abnormal accruals are computed using the Jones model (see footnotes to Table 2).
Table 4Regression of change in net income on prior abnormal accruals
Dependentvariable
Intercept ABNACCi(Yr –1) ABNACCi(Yr 0) Adj. R2 N
∆Net income 0.008 0.040 0.00 448(year 0) (1.68) (1.28)
∆Net income -0.002 -0.080 -0.076 0.06 399(year +1) (-0.28) (-1.94) (-1.74)
ABNACCi(Yr 0) is the abnormal accruals in year 0 and is computed by cumulating abnormal accruals in quarters 0 through +3. The abnormalaccruals in each event quarter is estimated using the Jones model. ABNACCi(Yr –1) is similarly defined for year –1. For years 0 and +1, ∆Netincome is computed by subtracting the net income in year –2 from that year’s net income. The t-statistics are shown in parentheses.
Table 5Mean six-day returns at earnings announcement periods around an equity offering announcement
Event quarterNo. ofobs.
Mean(percent)
t-statistic
(1) (2) (3) (4)
-8 520 0.02 0.04-7 551 0.36 0.86-6 571 -0.47 -1.09-5 611 0.85 2.02-4 649 1.11 2.58-3 668 1.06 2.39-2 701 2.07 5.06-1 733 1.20 2.960 724 -0.44 -1.261 710 0.17 0.452 697 -0.45 -0.933 684 -0.51 -1.214 656 -0.17 -0.395 640 0.04 0.106 619 0.00 0.007 602 -0.65 -1.40
Average(qtr –4 to –1)All firms 744 0.82 4.37
Average(qtr 0 to qtr 7)All firms 738 -0.21 -1.30Small firms 347 -0.28 -1.02Large firms 391 -0.15 -0.81Low-accrual-firms 197 -0.12 -0.43High-accrual-firms 193 -0.23 -0.67
Earnings announcement period consists of day –1 to +4 around the compustat earnings announcement date (day0). The earnings announcement returns are computed by subtracting from raw returns, the returns of a matchednon-issuing firm that has market capitalization within 30% of the sample firm and has the closest sales growth inprior two-years to the sample firm. The average returns across quarters are obtained by first computing theaverage returns for each firm, and then averaging these across firms. Small (large) firms are firms with marketcapitalization below (above) the median market capitalization. Low (high) accrual-firms are firms withabnormal accruals in year –1, ABNACCi(Yr –1), below (above) the median ABNACCi(Yr –1).
Table 6Pooled regression of earnings announcement returns on abnormal net income.
SDRETi(Qtr t) = α0 + α1 ABNNIi(Qtr t) + α2 POSTi × ABNNI i(Qtr t) + α3 SIZEi(Qtr t)
α0 α1 α2 α3 Adj. R2 N
Regression I 0.009 0.152 -0.089 0.01 7699(9.01) (3.50) (-1.91)
Regression II 0.027 0.158 -0.093 -0.004 0.02 7632(7.31) (3.58) (-1.97) (-5.05)
Earnings announcement returns for quarter t, SDRETi(Qtr t), are computed as the cumulative returns indays –1 through +4. Day 0 is the COMPUSTAT earnings announcement date. The regression is carriedout across event quarters –4 through +3. POSTi is an indicator variable that takes the value 1 for post-offering announcement quarters and 0 for all other quarters. The abnormal net income, ABNNIi(Qtr t), iscomputed from a seasonal random walk model for expectations. SIZEi(Qtr t) is the logarithm of marketcapitalization at the end of quarter t. The t-statistics are presented within parentheses.
Table 7Regression of two-day market reaction to equity offering announcements on prior earnings announcementreturns and on lagged abnormal accruals
I II III IVINTERCEPT -0.013a -0.013a -0.014a -0.015a
(-8.20) (-5.77) (-7.36) (-5.29)
SDRETi(Yr –1) -0.019b -0.017c
(-2.01) (-1.86)
ABNACCi(Yr –1) -0.031b -0.029c
(-1.98) (-1.85)
OFFSIZEi* -0.128 0.153
(-0.19) (0.19)
RUNUPi -0.010c -0.013c
(-1.74) (-1.77)
MRUNUPi 0.021 0.032c
(1.51) (1.71)
Adj R2 0.11 0.11 0.11 0.12N 844 844 534 534* (*100)
a Significant at the 1% level based on a two-tailed test.b Significant at the 5% level based on a two-tailed test.c Significant at the 10% level based on a two-tailed test.
The price reaction to equity offering announcement (EQRETi) is measured as the cumulative returns in theday of the and the day preceding the first public announcement of the offering. The earningsannouncement return SDRETi(Yr –1) and the abnormal accruals ABNACCi(Yr–1) for year –1 arecomputed by summing the corresponding quarterly variables in quarters –4 through –1. The quarterlyearnings announcement returns are the six-day cumulative returns in days –1 through +4 around theCOMPUSTAT earnings announcement dates. The quarterly abnormal accruals are estimated from theJones model. The regressions include the ratio of shares offered to shares outstanding before the offering,OFFSIZEi, and the daily compounded individual stock and market returns in the 60 trading days beforeoffering announcement, RUNUPi and MRUNUPi, as control variables. Small (large) firms are firms withmarket capitalization below (above) the median market capitalization. The t-statistics are shown inparentheses.
Table 8Regression results of long-run abnormal returns following seasoned equity offerings on pre-offering abnormal accruals
Raw returns Market-adjusted returns Control-firm adjusted returnsSmallfirms
Largefirms
One-yearreturns
(1)
Two-yearreturns
(2)
Four-yearreturns
(3)
One-yearreturns
(4)
Two-yearreturns
(5)
Four-yearreturns
(6)
One-yearreturns
(7)
Two-yearreturns
(8)
Four-yearreturns
(9)
Four-yearreturns(10)
Four-yearreturns(11)
INTERCEPT -0.217a -0.374a -0.237 -0.322a -0.520a -0.620a -0.036 0.004 0.766 -1.142 0.300(-2.75) (-2.68) (-1.17) (-4.18) (-4.20) (-2.70) (-0.35) (0.02) (1.54) (-0.73) (0.36)
ABNACCi(Yr –1) 0.023 -0.372c -0.575c -0.077 -0.454b -0.636c -0.002 -0.321 0.006 0.228 -0.138(0.17) (-1.75) (-1.72) (-0.61) (-2.46) (-1.77) (-0.01) (-1.35) (0.01) (0.18) (-0.11)
SIZEi 0.053a 0.075a 0.093a 0.049a 0.061a 0.100a 0.016 -0.002 -0.003 0.595 -0.015(3.96) (3.30) (3.07) (3.80) (3.04) (3.06) (0.89) (-0.09) (-0.04) (1.44) (-0.14)
BMi 0.004 0.235c 0.350b 0.019 0.194c 0.175 -0.121c 0.061 -0.544 -1.421b 0.238(0.07) (1.94) (1.99) (0.33) (1.80) (0.83) (-1.73) (0.54) (-1.54) (-2.07) (0.64)
N 551 551 551 551 551 551 536 536 536 268 268Adj. R2 0.03 0.05 0.09 0.04 0.08 0.05 0.00 0.00 0.01 0.03 0.01a Significant at the 1% level based on a two-sided test.b Significant at the 5% level based on a two-sided test.c Significant at the 10% level based on a two-sided test.
The raw returns are computed as the daily compounded returns over one, two or four-year period, beginning the day after the offering.ABNACCi(Yr –1) is the abnormal accruals from the Jones model in year –1, and is computed by summing quarterly abnormal accruals over eventquarters –4 through –1. The market-adjusted returns are computed by subtracting the return on the value-weighted market index from raw (buyand hold) returns. The control-firm adjusted returns are computed by subtracting the returns of a matched non-issuing firm from raw returns. Thematched firm is chosen from all firms that have market capitalization within 30% of the sample firm, and has the closest book-to-market ratio tothe sample firm. All regressions include log of firm size, SIZEi and the book-to-market ratio BMi before the offering. Small (large) firms are firmswith market capitalization below (above) the median market capitalization. The t-statistics are provided within parentheses.
Table 9Calendar-month returns and descriptive statistics for portfolios formed based on abnormal accruals in year –1
Quintiles1
Low2 3 4 5
HighLow - High
Returns (%)Mean 0.0087 0.0135 0.0092 0.0091 0.0016 0.0071
t-stat 1.72 2.87 2.06 1.76 0.31 2.13% > 0 61.59 61.59 59.15 60.98 51.22 57.93Sign test (p-value) 0.00 0.00 0.02 0.01 0.81 0.05
SizeMean 4.99 5.40 5.59 4.90 4.62 0.38
t-stat 250.71 172.80 246.35 263.40 213.41 11.76
Book-to-market ratioMean 0.52 0.58 0.55 0.46 0.39 0.13
t-stat 33.71 50.85 79.70 101.04 62.89 7.02
Intercepts from Fama-French model -0.0029 0.0013 -0.0025 -0.0021 -0.0095 0.0066
(-1.15) (0.74) (-1.60) (-0.93) (-4.08) (1.93)
Each month from March 1983 to October 1996, all sample firms that have made a seasoned equityoffering in the previous 48 months are sorted into quintiles based on their abnormal accruals in year –1.The equally-weighted portfolio returns are computed for each month. The table presents the time-seriesaverages and t-statistics for the returns. The row titled “%>0” gives the percentage of months in which thequintile returns are positive. For each quintile, the table also reports the average size (measured aslogarithm of equity capitalization) and the book-to-market ratio, both measured at the end of year –1 andthe intercepts from the Fama-French model. The intercepts are obtained from the following time-seriesregression of excess returns (Rit – Rf ) on the Fama-French factors:Rit – Rf = αi + bi (RM – Rf) + si SMB + hi HML
Table 10Time-series averages of monthly cross-sectional regressions of returns on stock characteristics andabnormal accruals in year –1
Mean t-stat % >0 Sign test(p-value)
INTERCEPT 0.011 1.36 0.51 0.81
ABNACCi(Yr –1) -0.012 -1.15 0.47 0.48
BMi 0.006 1.76 0.54 0.39
SIZEi -0.001 -0.94 0.55 0.24
Each month from March 1983 to October 1996, the following cross-sectional regression is run across allfirms that have made a seasoned equity offering in the previous 48 months:
Ri = γ0 + γ1 ABNACCi(Yr –1) + γ2 BMi + γ3 Sizei + εi
where, Ri is the monthly stock return of firm i, ABNACCi(Yr –1) is the abnormal accruals in year –1, Sizeiis the logarithm of equity capitalization and BMi is the book-to-market ratio. Both Sizei and BMi aremeasured at end of quarter –1. This table presents the time-series averages and t-statistics for thecoefficients estimates. The t-statistics is the mean coefficient estimate divided by the time-series standarderror of the coefficient estimate. The column titled “%>0” gives the percentage of months in which theestimated coefficient is positive. The last column provides the p-value from the sign test of whether thepercentage of months with positive coefficients is different from 50%.
Figure 1
Payoffs from earnings management game between offering firms and market participants.
Before offering announcement
(1)Firms do not overstate earnings
(2)Firms overstate earnings
At offering announcement:
(1) Investors do no believe prior earnings to beoverstated
(0, 0) (H, –H)
(2) Investors believe prior earnings to be overstated (–H, H) (–C, –C)
Offering firms have two strategies. They can either overstate or not overstate their earnings prior to offering announcements. Themarket participants also have two strategies. They either believe or do not believe that earnings before offering announcements wereoverstated. If they believe prior earnings to be overstated, they revise stock prices downward at the announcement of an equityoffering. The first entry in each box is the payoff to offering firms, while the second entry represents the payoff to market participants.H stands for a positive payoff and C stands for the costs of earnings management.
Figure 2Median abnormal net income around equity offering announcements.
Abnormal net income is computed as the change in net income from the corresponding quarter of the previous year.
-0.30
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Event quarters
Med
ian
abno
rmal
net
inco
me