View
2
Download
0
Category
Preview:
Citation preview
Opportunity knocks but once: delayed disclosure of financial items in earnings
announcements and neglect of earnings news
Yifan Li
San Francisco State University
Alexander Nekrasov
University of Illinois at Chicago
Siew Hong Teoh
University of California, Irvine
Published in Review of Accounting Studies, 2020, 25: 159–200
Abstract
We define a delayed disclosure ratio (DD) as the fraction of 10-Q financial statement items that
are withheld at the earlier quarterly earnings announcement. We find that higher DD firms have
a greater delay in investor and analyst response to earnings surprises: (i) the fraction of total
market reaction to quarterly earnings news realized around the earnings announcement (after
the 10-Q filing) is smaller (greater), and (ii) analysts are more likely to defer issuing forecasts
from immediately after the earnings announcement to after the 10-Q filing. Consistent with our
limited attention model predictions, the response catch-up associated with DD is incomplete
even after the delayed items are fully disclosed at the 10-Q filing date, and persists until the
next earnings announcement date. The return reaction to earnings news over the entire quarter
does not vary with DD, so differences in earnings informativeness do not explain the DD effect.
Our findings suggest that, for limited attention effects to be mitigated, the timing of disclosures
must be coincident with the focal periods—at earnings announcement dates—when investors
and analysts are paying the most attention.
Keywords: Delayed disclosure, Analyst and investor underreaction, Earnings response
coefficient, Post-earnings announcement drift, Limited attention, Market efficiency
JEL Codes: G14, G18, G28, G29, G38, M41, M45
We thank Lucile Faurel, Joanna Ho, Sonya Lim, Ben Lourie, Radhika Lunawat, Mort Pincus, Devin
Shanthikumar, Terry Shevlin, Christopher Schwarz, Zheng Sun, and seminar/conference participants at
the Pennsylvania State University, the University of California-Irvine, the 2018 American Accounting
Association Meeting, FARS 2016 Midyear Conference, SCARF 2016 Conference, the University of
Illinois at Chicago 2016 Accounting Conference, and the UCLA/UCI/USC Accounting conference for
helpful comments.
1
1 Introduction
Quarterly earnings press releases are a major channel through which companies release
financial information. These releases usually precede the filings of Form 10-Q with the
Securities and Exchange Commission (SEC) by more than two weeks. Because the releases are
timelier than the filings, investors and analysts react more strongly to earnings announcements
than to the filing date disclosures.1
Nevertheless, evidence indicates that investors and analysts underreact to earnings
news even at the announcement date. The resulting post-event delayed reaction, commonly
referred to as the post-earnings announcement drift (PEAD), is robust to various research
design choices and sample periods (Ball and Brown 1968; Foster, Olsen, and Shevlin 1984;
Bernard and Thomas 1989; Livnat and Mendenhall 2006; Ben-Rephael, Da, and Israelsen
2017). The literature has thus concluded that investors and analysts are inefficient processors
of information contained in current earnings news about future earnings (Bernard and Thomas
1990; Abarbanell and Bernard 1992; Zhang 2008).
In this paper, we examine the role of delayed disclosure of financial statement items in
the quarterly earnings press release and the role of limited attention of investors and analysts
in explaining this inefficiency. Specifically, we define the delayed disclosure ratio, DD, as the
ratio of the number of 10-Q financial statement items that are withheld in the quarterly earnings
press release to the total number of distinct financial statement items reported in the 10-Q filing.
We predict that a larger DD is associated with less investor and analyst response to earnings
news at the announcement date and greater PEAD. If some investors update beliefs at
1 In our sample, market reactions and analyst revisions at the earnings announcement are almost twice as large as
at the filing date. Specifically, the mean absolute market-adjusted abnormal return is 6.54% around earnings
announcements and 3.77% around 10-Q filings. The mean absolute analyst earnings forecast revision is 0.50%
around earnings announcements and 0.32% around 10-Q filings (untabulated). The weak investor reaction at filing
dates is consistent with the work of Easton and Zmijewski (1993), who examine a pre-EDGAR sample. Griffin
(2003) finds a significant reaction around 10-Q/K for a post-EDGAR sample. However, Li and Ramesh (2009)
report that the significant reaction is driven by 10-Q/K filings that coincide with earnings releases.
2
infrequent, discrete intervals, we also predict that the underreaction at the earnings
announcement date associated with DD is not eliminated at the 10-Q filing date but persists to
the next time investors update their beliefs. In other words, if the financial statement
information supporting the earnings news does not “strike while the iron is hot,” its effect could
be suppressed until the next earnings announcement. We use a limited attention model to derive
these surprising implications for investor reactions to earnings news.
Delayed disclosure of supporting financial statement items in earnings announcements
has been a matter of concern to regulators, information intermediaries, and professional
investors. The SEC’s Committee on Improvements in Financial Reporting (commonly called
the Pozen committee, 2008) and the Chartered Financial Analyst (CFA) Institute (2007) have
called for fuller financial statement disclosures in earnings press releases. 2 The National
Investor Relations Institute’s (NIRI) statement on the Standards of Practice for Investor
Relations, issued in 2008 and updated in 2013, specifically urges firms to include—in their
entirety—income statement, balance sheet, and statement of cash flows in the earnings release.
Consistent with the perspective above, these agencies argue that full financial statement
disclosures in the earnings announcements reduce information acquisition and attention costs,
which lessens the informational disadvantage for less sophisticated participants and enhances
the informational efficiency of the capital market. The greater financial statement disclosure at
the earnings announcement also helps market participants obtain earlier access to critical
information for valuation.
Fully rational investors and analysts should react immediately to earnings surprises,
regardless of whether some of the financial statement items supporting the earnings news are
2 The importance of the timing of financial information dissemination was also raised in an SEC Concept Release
in April 2016. The Commission has raised questions about the costs and benefits of quarterly reporting, including:
“Should we revise or eliminate our rules requiring quarterly reporting?” The SEC mandated quarterly reports in
1970. However, even prior to the SEC’s adopting the quarterly Form 10-Q, more than 70 percent of public
companies produced quarterly reports (SEC Concept Release, April 2016).
3
missing from the earnings press releases. If information is missing, a rational observer uses the
available information to make inferences about the missing items and future earnings.
Conditional upon the earnings surprise at the announcement date, the forecasts of a rational
observer are equal to the conditional expectation of earnings. So investor expectations and
analyst forecasts (setting aside agency problems) will be unbiased, and the equity market will
be efficient.3
Extensive empirical literature, including evidence of PEAD, suggests that investors are
imperfectly rational. Hirshleifer, Lim, and Teoh (2011) offer a limited attention theory to
explain PEAD. Building on this approach, we provide a limited attention model with
predictions about how the availability of supporting information from financial statement items
affects investors’ delayed reaction. We also provide intuitive arguments about how delayed
disclosure will affect analyst reactions.
The key difference between our paper and previous studies is that we examine the
importance of supporting information arriving when the earnings surprise signal is being
processed. This paper thus provides insight into the effects of the release of financial
information in installments rather than all at once, and highlights the importance and relevance
of matching disclosure timing to when observers are maximally attentive, such as at periodic
earnings announcements.
To derive the implications of delayed disclosure for investor reactions to earnings news,
we modify the limited attention model of Hirshleifer, Lim, and Teoh (2011). The model in
Section 2 analyzes how delayed disclosure of financial statement items affects the
announcement date price reaction to earnings surprises and PEAD. It does so by comparing the
3 Studies suggest that more detailed financial statement item disclosure represents higher disclosure quality (Chen,
Miao, and Shevlin 2015; Francis, Schipper, and Vincent 2002; Collins, Li, and Xie 2009) and enhances the
credibility of earnings by limiting managers’ degrees of freedom to manage earnings (Hirst, Koonce, and
Venkataraman 2007; D’Souza, Ramesh, and Shen 2010). While disclosure quality or credibility shifts the weight
that fully attentive rational investors put on earnings news, these factors do not imply that delayed disclosure
induces a systematic delayed reaction, or PEAD, unless inattention is also present.
4
equilibrium prices, conditional on the arrival of news in two regimes: in the timely disclosure
regime (TD-Regime), the firm’s earnings and financial statement information are revealed at
the earnings announcement date; in the delayed disclosure regime (DD-Regime), only earnings
are disclosed at the earnings announcement date, and other financial statement information is
not revealed until the filing date.
A prediction from this model is that the sensitivity to earnings news at the
announcement date is lower and PEAD is higher in the DD-Regime than in the TD-Regime.
Intuitively, under the DD-Regime, attentive investors face higher uncertainty about firm value.
Assuming investors are risk averse, higher uncertainty results in higher arbitrage costs, so
attentive investors trade less aggressively to exploit inattentive investors’ underreaction to
earnings news. Therefore, the DD-Regime underreaction is more severe.
Under the further assumption that the fraction of attentive investors is lower at the filing
date than at the earnings announcement date, the model offers the surprising implication of
incomplete catch-up of price reactions at the filing date in the DD-Regime. The full catch-up
occurs only at the next earnings announcement date, when investors again pay greater attention
to the firm.
Delayed, incomplete catch-up is consistent with the maxim that “opportunity knocks
but once.” A firm has a special opportunity to convey value-relevant information to investors
at the earnings announcement date, when the investors’ attention is high. If the firm misses this
opportunity, either inadvertently or intentionally, by disclosing value-relevant financial
information in installments, the earnings information disclosed at the earnings announcement
date will not be impounded into investor valuations to the same extent as if the other financial
information had been disclosed at the same time. Furthermore, under our hypotheses, even
when the missing information is revealed at the later 10-Q filing date, it will not be incorporated
into prices to the same extent as if it had been disclosed at the earnings announcement date.
5
Analysts similarly have attention constraints, and when there is a cost of analyzing firm
financial statements and issuing forecasts, they may not choose to develop forecasts both at the
earnings-announcement and filing dates. Given the advantage of timeliness, we expect analysts
to issue forecasts more often after the earnings announcement than after the filing date, all else
equal. However, when disclosure of value-relevant information is delayed, making forecasting
at the earnings announcement date more challenging, some analysts may choose to revise only
after the filing date. So we expect fewer forecasts after earnings announcements and more
forecasts after filing dates when supporting financial information is delayed. Furthermore,
assuming the costs and differences in interpretation of earnings news are higher when
supporting financial information is withheld, we expect greater average forecast underreaction
to earnings surprises, lower forecast accuracy, and greater forecast dispersion when delayed
disclosure is high.
Two forces could lead to results contradicting our predictions. First, firms may disclose
more financial statement items other than earnings when bottom-line earnings are less
informative.4 In this case, the delayed disclosure ratio would be low, but investor and analyst
reaction to earnings news might also be weak because earnings are less important. Second,
investors and analysts may need more time to analyze information from expanded disclosures.
This idea contrasts sharply with our above argument that having less information makes
forecasting more difficult for analysts. If bottom-line earnings are a good summary statistic for
expected future cash flows, disclosure of the full set of financial statement items could be
counterproductive. Hirshleifer, Lim, and Teoh (2009) find that the arrival of competing signals
distracts market participants and slows reaction to earnings news. Arif et al. (2017) find that
the reaction to annual earnings news is muted and PEAD is higher when the earnings
4 For example, Chen, DeFond, and Park (2002) find that companies disclose balance sheets in earnings press
releases when current earnings are likely to be less informative or when future earnings are more uncertain.
6
announcement is made concurrently with the release of the 10-K filing, especially when the
concurrent 10-K is longer and more difficult to process. The distraction idea contrasts with our
argument. Thus, whether delayed disclosure is associated with delayed investor and analyst
reaction to earnings news remains an empirical question.
In the empirical analysis, we use the Compustat Preliminary History database to
identify financial statement items that are disclosed in quarterly earnings press releases, and
the Compustat Quarterly database to identify items that are reported in the 10-Q filing.5 All of
our tests use firm- and time-fixed effects to control for potential omitted variables.
Consistent with our predictions, we find that delayed disclosure, DD, is negatively
associated with the immediate market reaction, positively associated with PEAD, and not
significantly associated with the reaction to earnings news over the entire quarter. So DD relates
to the timing of market reactions to earnings news but not the informativeness of earnings
information. When supporting information is delayed, the market reaction to earnings news is
partly delayed from immediately around the earnings announcement to after the 10-Q filing.
This suggests that the market is slow to incorporate earnings news until corroborating
information arrives, which in our model reduces the risk, for attentive investors, of trading
aggressively based on public earnings information. In terms of the magnitude of DD effect, we
find that an increase in DD from zero to one corresponds approximately to a decrease in ERC
of 37% and an increase in PEAD of 42%. Most interestingly, the delayed reaction persists even
after the 10-Q filing and is fully corrected only at the next earnings announcement. This test
confirms our model’s implication that “opportunity knocks but once.” As robustness checks,
we confirm that similar results hold using a change specification, using residual DD from the
regression of DD on an expanded set of explanatory variables, and using delayed disclosure of
5 We focus on quarterly earnings announcements to avoid the issue of audit completeness at the announcements
of annual earnings (Schroeder 2016; Marshall, Schroeder, and Yohn 2019). In an additional analysis, we examine
the effect of delayed disclosure in the annual earnings announcements.
7
industry peers as an instrument for the firm’s delayed disclosure. Analyses of delayed
disclosure separately for the income statement, the balance sheet, and the statement of cash
flows reveal that delayed disclosure of each of the three statements is negatively associated
with the immediate reaction to earnings news. In contrast, the positive association with the
post-earnings announcement drift is largely due to the delayed disclosure of balance sheet items.
We next examine the association between delayed disclosure and analyst reactions to
earnings news. We find that delayed financial statement disclosure is associated with a greater
delay in analyst forecasts following the earnings announcement. Specifically, for firms with
high DD, the portion of earnings forecasts made within two trading days of the earnings
announcement date is low, while the portion of earnings forecasts made after the 10-Q filing is
high. For high DD firms, the delay in issuance of earnings forecasts is not fully resolved within
a few days of the additional financial information being disclosed at the 10-Q filing. The
delayed reaction instead persists to the next quarter’s earnings announcement. We also find that
a higher DD is associated with a larger magnitude of analyst underreaction to the earnings
surprise, higher analyst forecast dispersion, and lower analyst forecast accuracy. DD
differences are not driven by differences in the number of financial items in the 10-K/Q. Finally,
we find that the relation between DD and delay in analyst forecasts is more pronounced for less
experienced analysts, consistent with past evidence that these analysts are more strongly
affected by forecasting difficulty (Bradshaw, Lee, and Peterson 2016).
There are three caveats regarding our method and inferences. First, in testing our
predictions, we try to use a simple and objective measure of overall delay in financial statement
disclosure. Yet in some industries and economic conditions, particular financial items may be
relatively more important, so a measure that assigns greater weights to these items could have
higher power. Second, we assume that greater disclosure of financial statement information is
useful to outsiders (a similar assumption can be found in Chen et al. [2002], Hirst et al. [2007],
8
Hewitt [2007], and D’Souza et al. [2010]). However, in some cases, information overload may
have a negative effect. Finally, as in most empirical studies, endogeneity is challenging to
control for. Although our results are robust to using firm fixed effects, residual DD, and
instrumented DD, we advise caution in drawing causal inferences.
Our study contributes to the research on the effects of financial statement disclosures
in earnings press releases. The three papers that are most closely related to the current paper
are Levi (2008), Louis, Robinson, and Sbaraglia (2008), and Miao, Teoh, and Zhu (2016).
Similar to our study, these papers examine how investor reactions are affected by the financial
statement items that are disclosed in the earnings press releases versus only in the 10-Q filings.
The key difference is that these papers are confined to the accrual anomaly and the disclosure
of accrual items in the earnings press releases. In contrast, our study is more general in that it
compares investor reactions at various dates—the initial earnings press release date, the 10-Q
filing dates, and subsequent earnings press release dates—conditional on the relative degree of
delayed disclosure of all financial statement items.6
Given the wider scope for potential mispricing of financial statement items beyond
accrual-related items, it is important to examine how the fraction of 10-Q financial statement
items whose disclosure is delayed can affect how quickly the information is incorporated into
stock prices. Unlike accruals, earnings news is always announced in the earnings press release,
so the financial statement items disclosed in the earnings press release play a different role in
our study—they reduce information uncertainty for sophisticated investors. Given the opposing
views on the effects of financial statement disclosures (Chen et al. 2002; Arif et al. 2017),7 the
6 The samples in Levi (2008) and Louis, Robinson, and Sbaraglia (2008) are also smaller: six quarters from 2001
to 2002 in the former, and four years from 1999 to 2002 in the latter. Because we are focused on the potential
mispricing of general financial statement items, we are also able to sidestep the thorny difficulties of estimating
discretionary accruals, a common problem for accrual anomaly studies. 7 Chen et al. (2002) show that financial statement disclosures can indicate lower informativeness of earnings,
which suggests a lower ERC. Arif et al. (2017) find that concurrent disclosure of annual earnings news and the
annual financial statement filings leads to information overload, lower response to earnings news, and a greater
PEAD.
9
answer to our research question is not obvious, and so requires empirical investigation.
2 Limited attention predictions and hypotheses
Our key variable of interest is delay in financial statement disclosure from the earnings
press release date to the 10-Q filing date. We use a limited attention model to derive predictions
for how delayed disclosure, DD, affects investor reaction to earnings news, and use other
intuitive arguments to predict how DD may affect the timing, accuracy, and other properties of
analyst forecasts in reaction to earnings news.
2.1 A limited attention model of delayed price reactions to earnings news
We begin with the limited attention models of Hirshleifer and Teoh (2003) and
Hirshleifer, Lim, and Teoh (2011), and consider a setting in which investors are risk averse, a
fraction f of investors is inattentive to the earnings and financial statement disclosures, and the
remaining fraction 1 – f of investors is attentive to all publicly available information.8 All
investors participate in a market for a single risky security (stock), which can be exchanged for
riskless asset (cash).
The timeline for the model is shown in Figure 1. Let P0 denote the stock price at date 0
before the earnings announcement. At date 1, which corresponds to the earnings announcement
date, investors receive public information about the terminal value of the stock. At date 2,
which corresponds to the filing date, investors may receive further public information about
the terminal value of the stock. At date 3, stockholders receive the terminal payoff of the stock.
Following Hirshleifer and Teoh (2003), we assume for modeling convenience that the stock is
in zero net supply,9 so the equilibrium stock price at date 1 equals a weighted average of the
8 It does not matter for the analysis whether the attentiveness of any given investor at a given date is random or is
the outcome of a deterministic process. What matters is the fraction of investors who are attentive at each date.
Inattentive investors do not know there is a signal that they are ignoring, i.e., they do not know that they are
inattentive. 9 The zero net supply assumption implies a zero net market risk premium, which eliminates the nuisance risk
premium term from the pricing equation. Investor risk aversion still plays a role by determining relative weights
10
beliefs of the two investor groups conditional on information available at date 1:
1 3 3[ | ] (1 ) [ | ]I AP E P E P , (1)
where the superscript I or A denotes the belief of inattentive or attentive investors, respectively,
and is an increasing function of f, the fraction of inattentive investors:
3
3 3
( )
( ) (1 ) ( )
I
I A
f P
f P f P
, (2)
where A,I(P3) = 1/varA,I(P3) is the perceived precision about firm value of attentive and
inattentive investors. Equations (1) and (2) follow from the market-clearing condition that the
date 1 total demand for the risky asset by attentive and inattentive investors equals the
outstanding supply. Because investors are risk averse, the weights of attentive and inattentive
investors (1- and , respectively) are increasing functions of their perceived precision about
the firm value.
Consider two alternative regimes. In the TD-Regime, the date 1 information set ϕ
consists of earnings e and an additional financial statement item . In the DD-Regime, ϕ
consists of earnings e only, and is withheld at date 1 and only reported at date 2. Attentive
investors update their priors upon the arrival of the information at date 1, but inattentive
investors ignore the information. Therefore, the price movement at date 1 reflects only the
Bayesian updates of attentive investors. In TD-Regime, the price at date 1 is
1 3 3 3 3( , ) [ | , ] (1 ) [ | , ] [ ] (1 ) [ | , ]I AP e E P e E P e E P E P e , (3)
where the first expectation is the prior expectation for the terminal value P3 and the second
expectation is the fully rational Bayesian update for P3, conditional on observing both pieces
of information e and , and the effective weight on the inattentive beliefs in the pricing equation
is given by:
by attentive and inattentive investors in the pricing equation.
11
3
3 3
( )
( ) (1 ) ( | , )
I
I A
f P
f P f P e
. (4)
In Case 2, the price at date 1 when only e is observed is given by
1 3 3 3 3( ) [ | ] (1 ) [ | ] [ ] (1 ) [ | ]I AP e E P e E P e E P E P e . (5)
As before, the first expectation for the inattentive investors is just the prior expectation of
terminal value P3, and the second expectation is the fully rational Bayesian update for the
terminal value based only on e. The effective weight for the beliefs of inattentive investors is
given by
3
3 3
( )
( ) (1 ) ( | )
I
I A
f P
f P f P e
. (6)
Since precision A(P3|e) < A(P3|e,), the weight ω on inattentive investors’ beliefs in setting
price at date 1 is higher in the DD-Regime than the weight on their beliefs in the TD-Regime.
Intuitively, attentive risk-averse investors trade less aggressively in the DD-Regime, where
uncertainty about firm value is higher (precision A(P3|e) < A(P3|e,)). So their weight on
equilibrium price is lower, and the weight of inattentive investors is higher. The price reaction
at date 1 upon the arrival of earnings news e e depends on the expectation of price at date 1,
conditional on e e as follows.
TD-Regime:
1 0 3 3 0[ ( , ) | ] [ ] (1 ) [ | ]E P e P e e E P E P e P . (7)
DD-Regime:
1 0 3 3 0[ ( ) | ] [ ] (1 ) [ | ]E P e P e e E P E P e P . (8)
Differentiating (7) and (8) with respect to earnings news e e , we obtain the earnings response
coefficients (ERCs) in the two cases.
TD-Regime:
12
3
1 0 3,
[ ( , ) | ] [ | ](1 ) (1 )
( )P e
dE P e P e e dE P e
d e e de
. (9)
DD-Regime:
3
1 0 3,
[ ( ) | ] [ | ](1 ) (1 )
( )P e
dE P e P e e dE P e
d e e de
. (10)
The term 3,P e is the ERC if all investors were fully attentive and f, , and ω were zero.
Comparing (9) and (10), the actual ERCs differ between the two cases by the difference
between and ω. As explained above, since A(P3|e) < A(P3|e,) and ω > , the ERC in
the DD-Regime (10) is less than the ERC in the TD-Regime (9). This leads to the following
proposition.
Proposition 1: The ERC to earnings news is lower when financial statement disclosure is
delayed (DD-Regime) than when it is not (TD-Regime).
To derive implications for post-earnings announcement drift (PEAD), we calculate the
price change from the earnings announcement, date 1, to the terminal payoff, date 3. In the TD-
Regime, where both e and are disclosed, the expected price change conditional on the
earnings news e e is
3 1 3 3[ ( , ) | ] ( [ | ] [ ])E P P e e e E P e E P . (11)
Differentiating (11) with respect to earnings news e - �̅�, the PEAD coefficient is
3
3 3 3,
( [ | ] [ ]) [ | ]
( )P e
d E P e E P dE P e
d e e de
. (12)
If all investors were attentive, would be zero, so PEAD would be zero. When some
investors are inattentive, PEAD is positive, even in this case of timely disclosure. As in the
work of Hirshleifer, Lim, and Teoh (2011), the magnitude of PEAD in our study increases with
the fraction f of inattentive investors, since increases with f.
13
Consider the DD-Regime, in which only e is disclosed at the earnings announcement
date and the additional financial statement item is disclosed only at the filing date. The
expectation for the price change from the earnings announcement to the terminal payoff,
conditional on the earnings news, is
3 1 3 3[ ( ) | ] ( [ | ] [ ])E P P e e e E P e E P . (13)
Differentiating (13) with respect to earnings news, we obtain the prediction for the PEAD
coefficient as
3
3 3,
( [ | ] [ ])
( )P e
d E P e E P
d e e
. (14)
Comparing (12) and (14), PEAD is larger in the DD-Regime than in the TD-Regime because
the effective weight of inattentive investors is larger in the former than in the latter, ω > .
Proposition 2: PEAD is larger when financial statement disclosure is delayed (DD-Regime)
than when it is not (TD-Regime).
Now consider the price reaction at the filing date, date 2. In the TD-Regime, the reaction
of both inattentive and attentive investors will remain unchanged at date 2. Inattentive investors
ignore information at the earnings announcement date and continue to ignore information at
the filing date. Since both pieces of information were disclosed at the earnings announcement
(date 1), the attentive investors have already reacted fully to the information, so there is no
additional reaction at the filing date (date 2).
In contrast, in the DD-Regime, there is an additional reaction at date 2. Recall that f is
the fraction of inattentive investors at date 1. We denote f’ as the fraction of inattentive investors
at the filing date. Given a fixed cost of attending and the fact that the earnings announcement
is more timely than the filing, it is reasonable to assume that f’ > f, so there are more inattentive
14
investors at the filing date (date 2) than at the earnings announcement (date 1). To simplify the
derivation, we assume that investors who are inattentive at date 1 remain inattentive at date 2
and that investors who are attentive at date 1 either remain attentive (fraction 1 – f’) or become
inattentive (fraction f’ – f) at date 2. This assumption is not necessary to arrive at the conclusion.
Appendix A shows that the price at the filing date, P2, is a weighted average of beliefs by (i)
inattentive investors (fraction f), (ii) investors attentive to earnings (fraction 1– f), and (iii)
investors attentive to both earnings and the additional financial statement item (fraction f’ –
f).
The difference in the total price change from date 0 to date 2 between the TD-Regime
and the DD-Regime depends on the difference in the expectation of date 2 price E[P2|e,] in
the two cases. In the former, this expectation is the same as the expectation at date 1, because
both e and are available at date 1. This expectation is formed by the fraction f of inattentive
investors I setting price expectations at the prior date, as they do not update, and the fraction 1
– f of attentive investors that update based on both pieces of information.
In the DD-Regime, the fraction of inattentive investors remains at f. The fraction of
investors attending to e remains the same as in the TD-Regime, 1 – f . However, the fraction of
investors attending to is 1 – f’, which is smaller than in the TD-Regime because f’ > f. This
difference drives the higher conditional variance for this subgroup in the DD-Regime; therefore,
the effective weight on investors who are fully attentive to both e and is lower than in the
TD-Regime. Thus the total reaction to earnings news from date 0 to date 2 is smaller in the
DD-Regime than in the TD-Regime. Because the sensitivity of total investor response to
earnings news over dates 1 and 2 is lower in the former than in the latter, there is a greater
sensitivity to earnings news in the catch-up investor reaction after the filing date in the DD-
Regime than in the TD-Regime.10 (At the extreme, if all investors ignore filing date information,
10 Risk aversion prevents the attentive investors at the filing date from being able to fully arbitrage away the
15
f’ = 1, then in the DD-Regime there is no catch-up reaction at the filing date, and all of the
catch-up reaction occurs at the terminal payoff, date 3.) This leads to the following proposition.
Proposition 3: The sum of the earnings response coefficients at the earnings announcement
date and at the filing date is lower in the DD-Regime than in the TD-Regime. PEAD after the
filing date is higher in the DD-Regime than in the TD-Regime.
2.2 Hypotheses for investor immediate and delayed reaction to earnings news
Propositions 1 and 2 show that the market reaction to earnings news, ERC, is more
muted, and that post-earnings announcement drift, PEAD, is more pronounced, when the
disclosure of the financial statement item is delayed—that is, when DD is high. Intuitively,
delaying financial statement disclosure increases arbitrage risk for attentive investors by
increasing the uncertainty about firm value at the earnings announcement date.11 As a result,
attentive investors trade less aggressively to exploit inattentive investors’ underreaction to
earnings news, which leads to a lower immediate reaction to earnings news and a greater PEAD.
Thus we have the following testable hypotheses.
H1a: Stock returns in the short announcement window are less sensitive to earnings news when
companies delay financial statement disclosure in earnings press releases. In other words,
firms with higher DD have smaller ERCs.
H1b: The post-earnings announcement return reaction (PEAD) is higher when companies
delay financial statement disclosure in earnings press releases. In other words, firms with
higher DD have larger PEAD.
Studies have found that the market reacts to news faster and more efficiently when
investors are more sophisticated (Bartov, Radhakrishnan, and Krinsky 2000), when investor
attention is greater (Hirshleifer, Lim, and Teoh 2009), and when earnings information is
mispricing induced by the inattentive investors at the earlier earnings announcement date. 11 This prediction is consistent with the Francis et al. (2007) finding that information uncertainty, proxied by the
noise in how accruals map into cash flows, positively relates to PEAD.
16
presented more saliently (Huang, Nekrasov, and Teoh 2018).12 Here, we test for a different
attention effect, one that relates to the mismatch in the timing of when firms disclose supporting
information and when investors are most attentive.
To test (H1a) and (H1b), we estimate the following regressions of cumulative abnormal
return over the earnings announcement window, CAR_EA, and over the post-earnings
announcement window, CAR_DRIFT.
CAR_EAjt = α + β1 RSUEjt + β2 DDjt + β3 RSUEjt * DDjt
+ Controls + RSUEjt * Controls +εjt, (15a)
CAR_DRIFTjt = α + β4 RSUEjt + β5 DDjt + β6 RSUEjt * DDjt
+ Controls + RSUEjt * Controls +εjt, (15b)
where CAR_EA is the cumulative abnormal return over the three-day window centered on the
earnings announcement date. CAR_DRIFT is the cumulative abnormal return over the period
starting two days after the earnings announcement for quarter t and ending one day after the
earnings announcement for quarter t+1 [+2, Next EA+1].13 RSUE is the decile rank of the
standardized unexpected earnings. We use firm size (SIZE) and analyst following
(ANA.FOLLOW) to control for information environment. We include market-to-book ratio
(MTB) to control for firm growth opportunity; past stock returns (RET) to control for changes
in the expected future performance; and institutional ownership (INST.OWN) to control for
investor base. We use the decile rank of the number of same-day earnings announcements by
other firms as a distraction proxy, as suggested by Hirshleifer, Lim, and Teoh (2009) (NRANK).
To control for potential omitted variables, we add firm fixed effects and time (year-quarter)
fixed effects to all regressions. Definitions of all variables are provided in Appendix B. If
delayed disclosure results in delayed investor response to earnings news, we predict that the
12 Huang et al. (2018) study the effects of presentation format, whereas our study examines the timing of the
disclosure of the detailed financial statement items. 13 We include the return over the next earnings announcement window since research shows that much of the
delayed investor reaction is concentrated in that period (Bernard and Thomas 1990).
17
coefficient on RSUE * DD is negative, β3 < 0, in regression (15a) and positive, β6 > 0, in
regression (15b).
We next examine the relation between DD and the market reaction to earnings news at
the 10-Q filing date and the price drift after the filing date to the next earnings announcement.
In Section 2, we derived the price at the filing date, assuming that the fraction of inattentive
investors at the filing date, f’, is greater than the fraction of inattentive investors at the earnings
announcement date: f’ > f. The assumption is consistent with the evidence in past research that
investor attention to news is higher when the news is disclosed more prominently or when
investor expectations of news arrival is higher.14 We showed that, in the DD-Regime, investors
do not fully catch up at the filing date, so there is a post-filing drift. This happens because
financial statement disclosure at the time of relatively low investor attention (the filing date) is
less effective in reducing arbitrage risk than disclosure at the time of relatively high investor
attention (the earnings announcement date). Proposition 3 of the model predicts that DD is
positively associated with a post-filing drift. Thus we have the following hypothesis:
H2: The post-filing reaction to earnings news is higher when the firm delays financial statement
disclosure. In other words, firms with higher DD have higher post-filing drift in returns.
To test H2, we estimate investor reaction between the 10-Q filing and the next earnings
announcement and around the next earnings announcement date.
14 Past empirical studies find that investor attention is influenced by, among other factors, the coverage of the
announcement in prominent media outlets (Klibanoff, Lamont, and Wizman 1998), the day of the week
(DellaVigna and Pollet 2009), the number of same day other firm earnings announcements (Hirshleifer, Lim, and
Teoh 2009), the dissemination of the announcement on social media (Blankespoor, Miller, and White 2014), and
the salience of the earnings announcement headline (Huang, Nekrasov, and Teoh 2018). Additionally, the Wall
Street Journal publishes anticipated dates when earnings will be released but not when financial statements will
be filed with the SEC, so investors are more likely to be attentive on earnings announcement dates than on filing
dates. You and Zhang (2011) make a similar argument that investors pay greater attention to earnings
announcements than 10-K filings. In the model, we assume that an information signal, in the form of financial
statement items, is more likely to be processed by investors when it is disseminated more prominently or when
investor attention is higher, i.e., in the earnings press release.
18
2.3 Hypotheses for analyst response to earnings news
Analysts expend time and effort to form earnings forecasts. When there are costs of
analyzing firm financial statements and issuing forecasts, analysts with limited resources may
devote analytical attention to the firm only at either the earnings announcement date or the
filing date each quarter, whereas less constrained analysts may analyze the firm at both event
dates. The majority of analysts—around 71% of our sample (untabulated)—issue earnings
forecasts for a given firm only once a quarter.
Given the advantage of timeliness, analysts are likely to issue forecasts as soon as
possible after the earnings announcement, all else equal. When disclosures of valuation-
relevant financial statement items are delayed until the subsequent 10-Q filing date, issuing
forecasts earlier has the advantage of timeliness, but the forecasts will be noisier and require
more effort (for estimating missing items). Delaying forecasts to after the 10-Q filing, when
the information items used in the forecast derivation are available, offers the benefits of less
effort and greater accuracy. So we expect that, when the disclosure of financial statement items
is delayed until the 10-Q filing, analysts are more likely to delay their forecasts to after the
filing date. This leads to the following hypotheses.
H3a: When companies delay financial statement disclosure, the proportion of analyst earnings
forecasts made immediately after the earnings announcement declines.
H3b: When companies delay financial statement disclosure, the proportion of analyst earnings
forecasts made after the 10-Q filing rises.
To test these hypotheses, we estimate the following regressions.
%FORECASTSjt = α1 + β1 DDjt + Controls + εjt, (16)
where %FORECASTS is the proportion of earnings forecasts issued in a specific time
subwindow, relative to the total number of earnings forecasts issued in the period starting on
the day of the earnings announcement and ending two days before the next earnings
19
announcement. We calculate %FORECASTS in four subwindows: around the earnings
announcement [0,+1] (we refer to this measure as %FORECASTS_EA), between the earnings
announcement and the 10-Q filing [+2, Filing Date-2], around the 10-Q [Filing Date-1, Filing
Date +1], and between the 10-Q and the next earnings announcement [Filing Date+2, Next EA-
2]. The relative proportion of forecasts in different time windows is designed to capture the
timing, rather than the number, of analyst earnings forecasts.
Control variables include the decile rank of standardized unexpected earnings (RSUE),
firm size (SIZE), the market-to-book ratio (MTB), past stock returns (RET), analyst following
(ANA.FOLLOW), institutional ownership (INST.OWN), and the decile of the number of same-
day earnings announcements by other firms (NRANK). We predict that the coefficient on the
delayed disclosure ratio, DD, is negative (β1 < 0) for %FORECASTS in the earnings
announcement window (H3a) and positive (β1 > 0) for %FORECASTS between the 10-Q and
the next earnings announcement (H3b).
We turn next to the hypotheses for forecast underreactions and accuracy. Consistent
with limited analyst attention (or possibly with agency issues), studies show that analysts
underreact to earnings news contained in earnings announcements (Mendenhall 1991;
Abarbanell, and Bernard 1992). Specifically, research finds that analyst earnings forecast errors
for quarter t+1 are positively correlated with earnings surprises in quarter t. Analyst
underreaction to earnings news is higher when it is more costly for analysts to obtain the
additional information needed for more efficient forecasts, that is, when the information is on
small firms and firms with lower analyst coverage.
When firms delay financial disclosure in earnings announcements, we expect the lack
of relevant items at the earnings announcement to impede the development of earnings
forecasts, leading to lower forecast accuracy and higher forecast dispersion across analysts.
Therefore, we hypothesize that delayed disclosure will be associated with (i) greater analyst
20
underreaction to earnings news, (ii) lower forecast accuracy, and (iii) higher forecast dispersion.
H4a: Analyst underreaction to earnings surprises is higher when companies delay financial
statement disclosure in earnings press releases.
H4b: Analyst forecast accuracy is lower when companies delay financial statement disclosure
in earnings press releases.
H4c: Analyst forecast dispersion is higher when companies delay financial statement
disclosure in earnings press releases.
To test these hypotheses, we estimate the following regression equations.
FUTURE.RSUEjt+1 = α + β1 RSUEjt + β2 DDjt + β3 RSUEjt*DDjt
+ Controls + RSUEjt * Controls + εjt; (17a)
ABS.FEjt+1 = α + β1 DDjt + β2 ABS.FEjt + β3 MISSjt + β4 ABS.FEjt*MISSjt
+ Controls + εjt; (17b)
STD.ESTjt+1 = α + β1 DDjt + β2 ABS.FEjt + β3 MISSjt + β4 ABS.FEjt*MISSjt
+ Controls + εjt; (17c)
where FUTURE.RSUEjt+1 is the decile rank of the analyst earnings forecast error for quarter
t+1 (Actual EPSjt+1 minus Forecasted EPSjt+1), scaled by the stock price at the end of quarter
t, and Forecasted EPSjt+1 is the median of forecasts issued either (i) between the earnings
announcement and the 10-Q filing, or (ii) between the 10-Q filing and the next earnings
announcement. ABS.FEjt+1 is the absolute value of analyst earnings forecast error for quarter
t+1 (Actual EPSjt+1 minus Forecasted EPSjt+1), scaled by the stock price at the end of quarter
t. STD.ESTjt+1 is the standard error of analyst earnings forecasts for quarter t+1, scaled by the
stock price at the end of quarter t. MISSjt equals one if quarter t earnings miss the analyst
forecast and zero otherwise. Controls is the same set of control variables as in equation (1). We
expect the coefficient on RSUEjt*DD in equation (17a), the coefficient on DD in equation (17b),
and the coefficient on DD in equation (17c) to all be positive.
21
3 Sample, data, and empirical results
3.1 Sample and data
We obtain financial statement information included in quarterly earnings
announcements from the Compustat Preliminary History database (Preliminary), company
financial data in 10-Q filings from Compustat, stock returns and prices from CRSP, and
analysts’ earnings forecasts from I/B/E/S. Following the literature, we exclude utilities and
financial services firms (SIC codes 4900–4999 and 6000–6999). Our sample covers all years
available on the Preliminary History database from 1990 through 2013, excluding 1999, when
Compustat collected only limited items from earnings announcements.15 We drop observations
for the fourth fiscal quarter (Section 4.8 examines them), observations with less than three
trading days between the earnings announcement date and the 10-Q filing date, and
observations with less than five financial statement items in any of the three financial
statements (income statement, balance sheet, and statement of cash flows) in the 10-Q filing.
The final sample for the market reaction tests consists of 121,807 firm-quarter observations.
For the analyst response tests, we also require at least one forecast for the next quarter earnings
issued after the current quarter’s earnings announcement. The final sample for the analyst
reaction tests consists of 66,763 observations. Appendix C describes the details of the sample
selection.
3.2 Measure of delayed financial statement disclosure
To measure the delay of financial statement disclosure in earnings press releases, we
calculate the delayed disclosure ratio, DD, which equals the number of financial statement
15 To check the accuracy of the Prelim data, we select a random sample of 50 firms with available observations
before and after 1999. We then manually identify financial statement items disclosed in earnings press releases
and compare the delayed disclosure ratio based on the manually collected data and the ratio based on the Prelim
data. The Spearman correlation between the two ratios exceeds 0.90 both before and after 1999, suggesting that
the delayed disclosure measure based on the Prelim data is reasonably accurate.
22
items that are not disclosed in the earnings press release but reported in the corresponding 10-
Q filing scaled by the number of financial statement items reported in the 10-Q filing. The
Preliminary database is used to identify which financial statement items are disclosed for a
given firm-quarter earnings announcement.16 We calculate DD as a simple fraction of delayed
items rather than as a dollar value-weighted ratio because market participants do not know the
value of delayed items at the time of the earnings announcement. Also, it is not clear whether
items with relatively large magnitudes, such as cost of goods sold and goodwill, are more
important than items with relatively small magnitudes, such as operating income and
inventory.17
In calculating DD, our goal is to construct a simple and objective measure that captures
overall delay of financial statement information and to relate this measure to the delayed
reaction to earnings news. An alternative approach would be to analyze delayed disclosure of
individual items or groups of items with distinct functional roles. Our approach has two
advantages. First, which items are more (less) important likely depends on the industry, the
economic conditions, the disclosure of related items, and other factors. Second, firms that delay
disclosure of some items are also more likely to delay disclosure of other items. The overlap
makes it empirically difficult to classify delayed items into categories with distinct roles. This
also suggests that DD is a reasonable construct of overall delay in financial statement disclosure.
3.3 Descriptive statistics
Table 1 presents descriptive statistics. The average fraction of financial statement items
16 Delayed disclosure is measured by comparing the disclosure in the earnings press release with the disclosure in
the same quarter 10-Q filing. As a robustness check, we also calculate DD by comparing items disclosed in the
earnings press release with items reported in the 10-Q filing for the same quarter last year. The correlation between
this DD and our main measure is 0.9982. All our results are qualitatively and quantitatively similar when using
this alternative DD measure (untabulated). 17 Nevertheless, most of the results hold when using dollar value–weighted DD. Higher dollar value–weighted
DD is negatively (positively) associated with the fraction of analyst forecasts issued immediately after the earnings
announcement (after the 10-Q filing) and the market reaction to earnings news immediately around the earnings
announcement (after the 10-Q filing) (untabulated).
23
that are delayed till the 10-Q filing, DD, is 57.2%.18 The average number of analysts issuing
earnings forecasts immediately after the earnings announcement, FORECASTS_EA, is 3.992.
The average proportion of such forecasts relative to all forecasts, %FORECASTS_EA, is 39.6%.
The average proportion of earnings forecasts issued after the 10-Q, %FORECASTS_[Filing+2,
Next EA-2], is 39.2%. As in other studies that use analyst forecasts, firms in our sample are
relatively large, with a mean (median) market capitalization, MV, of $2,301.3 ($295.9) million.
DD exhibits significant variation with an interquartile range of 0.430 to 0.803. Variation
in DD comes mainly from the variation in the number of items withheld in the earnings
announcement, rather than the variation in the number of items in the 10-Q filing. The
correlation between DD and the number of items withheld in the earnings announcement is
0.992, whereas the correlation between DD and the number of items in the 10-Q filing—that
is, the denominator of the DD ratio—is -0.070 (untabulated). Most of the variation in DD is
due to cross-sectional rather than within-firm variation. Individual firms typically do not make
big period-to-period changes in the amount of information that is withheld at the earnings
announcement. The correlation between DD in the current quarter and DD in the same quarter
last year is 0.71 (untabulated). All our tests control for firm fixed effects, so our research focus
is on within-firm changes in DD. In additional analysis, we also estimate tests using a change
specification.
Table 2 reports correlations among variables. Consistent with H3a and H3b, DD is
negatively associated with the proportion of earnings forecasts issued immediately after the
earnings announcement, %FORECASTS_EA, and positively associated with the proportion of
earnings forecasts issued after the 10-Q filing (correlations -31.7% and 20.5%, respectively).
DD is also negatively associated with firm size, analyst following, and institutional
18 The average percentage of delayed income statement, balance sheet, and cash flow statement items is 24.7%,
51.5%, and 86.4%, respectively (untabulated). The descriptive statistics are comparable with those of D’Souza et
al. (2010).
24
ownership.19
Figure 2 shows graphically %FORECASTS for firms in the bottom decile of DD (the
solid blue line) and firms in the top decile of DD (the dashed red line). The horizontal axis
shows time intervals between the current and next earnings announcement. The vertical axis
shows the portion of earnings forecasts issued by the indicated point in
time: %FORECASTS_[0, τ], where τ = +1, Filing Date-2, Filing Date+1, or Next EA-2. As the
figure reveals, the portion of earnings forecasts issued within two trading days is notably lower
when DD is high (14.1% for high DD versus 53.0% for low DD). Conversely, the portion of
analyst forecasts delayed until after the 10-Q filing is greater when DD is high (56.6% for high
DD versus 32.8% for low DD).20
3.4 Main results
Our first test examines whether investor immediate reaction to earnings news is weaker
for firms that delay financial statement disclosure (H1a). Table 3 shows the results of the
regression of the firm stock return on RSUE*DD and control variables. To check whether total
reaction to earnings news is associated with DD, we first examine the investor reaction over
the window that covers both the announcement and post-announcement windows (columns 1
and 2). The coefficient on RSUE*DD is insignificant (-0.014, p = 0.215), suggesting that the
informativeness of earnings news does not relate significantly to DD. The remaining four
columns show results for investor immediate reaction. Results in columns 3 and 4 establish the
baseline for the average ERC in our sample. Consistent with the results of other studies,
announcement window returns are positively associated with earnings surprise (coefficient =
0.082, p <0.001). Consistent with H1a, the coefficient on RSUE*DD is negative and significant
19 Together, firm size, institutional ownership, and analyst following explain 12.1% of the variation in DD based
on regression R2 (untabulated). 20 While Figure 2 also suggests that there is a positive association between DD and %FORECASTS_[EA+2, Filing
Date -2], the results in the next section show that this association is negative and insignificant in the multivariate
analysis.
25
(-0.036, p <0.001), indicating that the immediate price reaction to earnings surprises is weaker
when firms delay financial statement disclosure. An increase in DD from zero to one
corresponds to a decrease in ERC of approximately 44% (-0.036/0.082=-44%). Importantly,
the total reaction is not significantly associated with DD, which suggests that the lower
immediate reaction for these firms is due to delayed response to earnings news rather than
differences in earnings informativeness.
Table 4 directly tests the relation between DD and investor delayed reaction to earnings
news (H1b). If firms with higher DD experience greater delay in investor reaction to earnings
news, we expect a stronger association between earnings surprise, RSUE, and abnormal return
after the earnings announcement, CAR_DRIFT. Panel A reports results of the regression of
CAR_DRIFT on RSUE*DD and control variables. The first two columns establish the
benchmark for the magnitude of the average delayed reaction in our sample. Consistent with
the post-earnings announcement drift, the coefficient on RSUE is positive and significant
(0.054, p <0.001). Consistent with H1b, the coefficient on RSUE*DD in the last two columns
is positive and significant (0.025, p =0.032), suggesting that investor delayed reaction to
earnings news is greater when DD is high. An increase in DD from zero to one corresponds to
an increase in PEAD of approximately 46% (0.025/0.054=46%).21
We next examine, in greater detail, when investor delayed reaction is corrected. If
inattentive investors are not focusing on the firm at the 10-Q filing date, then the delayed
reaction associated with DD will not be eliminated immediately at the filing date (H2). Instead,
the association between DD and the post-earnings announcement drift will continue even after
the remaining information becomes publicly available.
21 We add further robustness checks. In addition to year fixed effects, we add a control for a time trend in the
immediate and delayed response to earnings news by including an interaction between RSUE and a time trend
variable. The time trend variable is the natural logarithm of (year – 1988), and year is the observation’s fiscal year
from 1989 to 2013 (Collins, Li, and Xie 2009). The untabulated results are consistent with those in Tables 3 and
4. As another robustness check, we use earnings surprise based on a seasonal random walk instead of analyst
forecasts, and find similar results.
26
Table 4 Panel B shows the results of the return tests for the following subwindows:
between the earnings announcement and the 10-Q filing (columns 1 and 2), around the 10-Q
(columns 3 and 4), between the 10-Q and the next earnings announcement (columns 5 and 6),
and around the next earnings announcement (columns 7 and 8). Consistent with the delayed
reaction to earnings news stemming from the delay of supporting financial information till the
10-Q filing, there is no correction before the 10-Q. Instead, the relation between DD and
investor reaction is negative and insignificant in the window prior to the filing date (-0.004, p
= 0.308). The results in columns 3 and 4 indicate that there is also no significant immediate
correction at the time of the 10-Q filing, which is consistent with the idea that investors do not
pay as much attention to 10-Q filings as they do to earnings announcements. The correction of
the delayed reaction happens only after the 10-Q filing. The coefficient on RSUE*DD is
positive after the 10-Q filing (0.023, p = 0.010) and around the next earnings announcement
(0.005, p = 0.044). The results indicate that the relation between DD and investor reaction to
earnings news extends beyond the time of arrival of the delayed information.
Figure 3 shows graphically the timing of investor reaction to earnings news for firms
with DD=0 (the solid blue line) and firms with DD=1 (the dashed red line).22 Consistent with
the results in Tables 3 and 4, the delayed reaction starts at the earnings announcement. The
delayed reaction is not corrected before or immediately at the time of the 10-Q filing but
persists beyond the filing date and is corrected only after the next earnings announcement. As
a result, firms that have a greater portion of financial disclosure delayed till the 10-Q filing also
have a greater portion of the total market reaction delayed until after the filing. Specifically,
54.2% (27.4%) of reaction is delayed until after the filing when DD=1 (DD=0).
We next test the association between delayed disclosure and analyst delayed response
to earnings news (H3a and H3b). Table 5 columns 1 and 2 show the results for the portion of
22 The notes to Figure 3 provide details on how the graph is constructed.
27
forecasts issued in the earnings announcement window, %FORECASTS_EA. Consistent with
H3a, the coefficient on DD is negative and significant (-0.019, p =0.006). This indicates that,
when firms delay financial statement disclosure, the portion of timely analyst forecasts
decreases.
The remaining columns show the results for the proportion of delayed analyst forecasts
issued in three post-announcement subwindows: between the earnings announcement and the
10-Q filing (columns 3 and 4), around the 10-Q (columns 5 and 6), and between the 10-Q and
the next earnings announcement (columns 7 and 8). There is no significant relation between
DD and %FORECASTS between the earning announcement and the 10-Q (-0.007, p =0.292).
The results in columns 5 and 6 indicate that there is no significant immediate response to the
10-Q filing (0.001, p =0.331), while the positive and highly significant coefficients in the last
two columns indicate that analysts respond to the arrival of information in the 10-Q filing with
a delay (0.025, p <0.001).
Overall, the results in Table 5 are consistent with (H3a) and (H3b). When firms delay
financial statement disclosure, analysts are less likely to issue earnings forecasts following the
earnings announcement and more likely to do so after the 10-Q. The relation between DD and
the timing of earnings forecasts does not disappear immediately when the delayed information
arrives in the 10-Q filing but persists beyond the filing date.
We next test whether delayed disclosure is associated with the magnitude of analyst
underreaction to earnings news (H4a), analyst forecast accuracy (H4b), and analyst forecast
dispersion (H4c). Table 6 Panel A presents the results of the regression of future earnings
forecast error, FUTURE.RSUE, on current earnings surprise, RSUE, the interaction variable of
interest, RSUE*DD, and control variables. To establish the baseline for the magnitude of
average analyst underreaction in our sample, we first estimate a regression of FUTURE.RSUE
and the 10-Q filing on RSUE and control variables (columns 1 and 2). Consistent with the prior
28
research, current earnings surprise is significantly positively associated with future earnings
forecast error (coefficient = 0.218, p <0.001). Consistent with H4a, the coefficient on
RSUE*DD is positive and significant (0.066, p = 0.010). When compared to the average analyst
underreaction, the magnitude of the coefficient on RSUE*DD indicates that an increase in DD
from zero to one corresponds to an increase in the analyst underreaction of approximately 30%
(0.066/0.218= 30%). The last two columns show the results for forecasts issued after the 10-Q
filing. The coefficient on RSUE*DD is positive and significant (0.059, p = 0.033), indicating
that the underreaction persists even after the arrival of supporting information in the 10-Q.
These findings corroborate the conclusion from the market tests that there is immediate
underreaction and that the underreaction persists after the filing date.
The results of the regression of absolute analyst forecast errors on DD reveal a similar
pattern (Panel B of Table 6). Consistent with H4b, the coefficient on DD is positive and
significant (0.003, p < 0.001), suggesting that analyst forecast accuracy is lower when firms
delay financial statement disclosure. The coefficient on DD is also significant after the filing
date (0.003, p < 0.001).
In Panel C, we find a positive association between DD and analyst forecast standard
errors (0.001, p < 0.001), which is consistent with disclosure delay increasing forecast
dispersion (H4c). DD effects again persist after the filing of 10-Q, as evidenced by a positive
coefficient on DD in the post-filing window (0.001, p < 0.001).
4 Robustness tests and additional analyses
4.1 Adding further controls and using residual delayed disclosure ratio
To examine whether the relation between DD and investor and analyst response to
earnings news is incremental to other factors that may affect the choice of DD, we perform the
following two-stage analysis. In the first, we regress DD on an expanded set of explanatory
29
variables and calculate residual DD, RES.DD. In the second, we re-estimate the investor and
analyst response tests using RES.DD. The first-stage regression is as follows.
DDjt = α + β1 RSUEjt + β2 SIZEjt + β3 MTBjt + β4 RETjt + β5 ANA.FOLLOWjt
+ β6 INST.OWNjt + β7 EA.LAGjt + β8 SMOOTHjt + β9 JMOBjt
+ β10 PRO.FORMAjt + β11 SPEC.ITEMjt + β12 EARN.VOLjt + β13 LOSS jt
+ β14 NEG.SUE jt + β15 HIGH.TECHjt + β16 MERGERjt + β17 YOUNGjt
+ β18 RET.VOLjt + β19 BIG4jt + εjt.
(18)
In addition to the base set of control variables in the main analysis (SIZE, MTB, RET,
ANA.FOLLOW, INST.OWN, and NRANK), we use numerous additional control variables. We
add a control for the timing lag between the fiscal quarter end and the earnings announcement
date (EA.LAG), since the fraction of financial statement items disclosed in the earnings
announcement is likely to be negatively associated with the earnings announcement lag
(Schroeder 2016). We add a proxy of income smoothing (SMOOTH) and an indicator of the
firm just meeting or beating earnings expectations (JMOB), to proxy for managerial incentives
to intervene in the earnings reporting process (D’Souza et al. 2010). We also include an
indicator for the use of pro forma earnings (PRO.FORMA), since Regulation G requires
companies to disclose items excluded from pro forma earnings. Following Chen et al. (2002),
we also include variables to capture settings in which current earnings are less likely to be a
sufficient valuation metric: earnings volatility (EARN.VOL), an indicator of negative earnings
(LOSS), an indicator of negative earnings surprise (NEG.SUE), a high-tech industry indicator
(HIGH.TECH), an indicator of mergers and acquisitions (MERGER), an indicator of special
items (SPEC.ITEM), a young firm indicator (YOUNG), and an indicator of high return volatility
(RET.VOL). To proxy for the effect of the auditor, we include an indicator of a Big Four auditor
(BIG4). All variables are defined in Appendix B.
The results of the estimation of equation (18) are reported in Table 7 Panel A. DD tends
to be lower for large firms as well as for those with greater analyst following, institutional
30
ownership, and Big Four auditors, suggesting that firms with richer information environments
disclose more financial statement items at earnings announcements. DD is lower for firms that
use pro forma earnings, consistent with the prediction that the disclosure of items excluded
from pro forma earnings reduces delayed disclosure. DD also tends to be lower for firms with
special items, higher earnings volatility, and losses, as well as for high-tech firms and young
firms, suggesting that managers provide financial statement disclosures when future earnings
are more uncertain. Table 7 Panel B reports the results of investor and analyst response tests
using the residual DD variable. The results are similar to those in Tables 3–6, indicating that
our results are robust to adding further controls and using the residual DD measure.23
4.2 Instrumental variable approach
We employ an instrumental variable approach to further mitigate the concern about
potential omitted correlated variables influencing the degree of delayed disclosure. Our
approach follows Holzman, Marshall, Schroeder, and Yohn (2019), who use disclosure by
industry peers to instrument the firm’s disclosure in earnings press releases. The measure is
likely to be a reasonable instrument since the industry peers’ disclosure is not chosen by the
firm but is a significant predictor of the firm’s disclosure. Specifically, our instrument for DD
is the median delayed disclosure ratio in the same four-digit SIC industry in the previous quarter.
The instrument for DD*RSUE is the interaction of the instrument for DD and RSUE. The results
of this analysis, not tabulated for brevity, are consistent with the findings from our main tests.
23 We conduct two sets of robustness tests. First, we add earnings announcement lag, special items indicator, loss
indicator, indicator of negative earnings surprise, and their interactions with earnings surprise as additional
controls to the main regressions in Tables 3-6. The results of the key independent variables—RSUE*DD in Tables
3-4 and Table 6 (Panel A) and DD in Table 5 and Table 6 (Panels B and C)—are robust. Next, regarding the
residual DD variable in the 2SLS regressions in Table 7, we also add the additional controls in the second stage
regression. The results for the key independent variables relating to residual DD remain robust. Since residual DD
is orthogonal to the control variables used in the first stage, the two-stage approach avoids the need to include, in
the second stage, a large number of interactions, RSUEjt*Controljt, that are highly correlated with each other, and
thus increases the reliability of the empirical inferences.
31
4.3 Robustness tests using year-over-year changes
Our main tests use firm fixed effects to isolate within-firm variation in delayed
disclosure and control for potential omitted variables. We re-estimate our tests using the change
specification as an alternative. Specifically, for each dependent and independent variable, we
calculate the year-over-year change and estimate regression using these changes. Untabulated
results are consistent with the results using firm fixed effects.
4.4 Management guidance
Many firms provide earnings guidance during earnings announcements. To ascertain
whether DD effect is incremental to investor and analyst response to earnings guidance, we
add a control for guidance surprise. Similar to earnings surprise, we measure guidance surprise
as the decile rank of the difference between earnings guidance and consensus forecast. We also
include an indicator variable that equals one if the firm issues earnings guidance and zero
otherwise. Untabulated results show that the results are robust to controlling for earnings
guidance.
4.5 Conference calls
Companies often hold conference calls around their earnings announcements.
Kimbrough (2005) finds that analyst and investor underreaction to earnings news is lower after
firm-initiated conference calls. To examine the incremental effect of DD, we calculate an
indicator variable that equals one if there is a conference call during the earnings announcement
window [-1,+1] and zero otherwise. We also try a conference call initiation indicator that equals
one after the firm initiates earnings calls and zero otherwise. Untabulated results show that our
results are robust to controlling for conference calls.
4.6 Early and late subperiods
The financial reporting environment changed over the sample period. In particular,
32
Regulation G (Reg G), adopted by the Securities and Exchange Commission in 2003, requires
firms that supplement GAAP earnings by non-GAAP earnings to quantitatively reconcile the
two measures. Since items excluded from pro forma earnings must be disclosed, the
reconciliation can reduce delayed disclosure if the firm did not disclose those items before.
Consequently, we test the effect of DD before and after Reg G. Untabulated results show that
the effect of DD is significant in both subperiods.
4.7 Annual earnings announcements
Our analyses are based on quarterly earnings announcements for the first three quarters.
Since annual financial statements are audited, whether the audit has been completed may affect
financial statement disclosure in annual earnings announcements and the market response to
annual earnings news (Schroeder 2016; Marshall, Schroeder, and Yohn 2019).24 We examine
the effect of delayed disclosure on the market reaction to annual announcements where we
control for the audit completion. Similar to the findings for quarterly announcements, delayed
disclosure is significantly negatively associated with the immediate market reaction. In contrast,
the association between delayed disclosures and delayed reaction to annual announcements is
not statistically significant, which may be partially explained by the Rangan and Sloan (1998)
finding that the post-earnings-announcement drift is relatively smaller after annual
announcements than after quarterly announcements.25
24 Schroeder (2016) finds that completed audits at the earnings announcement date enhance the information
content of additional financial statement items at the annual earnings announcement. Marshall et al. (2019) find
that the market places more weight on good annual earnings news announced after audit completion. Our study
differs from these studies in that we study quarterly earnings that do not have audit-related issues. In addition, our
limited attention approach affords predictions for post-event investor reactions, whereas the credibility effect from
completed audits in these studies predicts only earnings announcement period reactions and not post-event
reactions. 25 The lower magnitude of PEAD after annual announcements is likely to result in a lower test power. We find
that even though the magnitude of the coefficient on RSUE*DD might suggest an economically meaningful effect
when compared to the average PEAD after annual announcements, the coefficient is not statistically significant.
33
4.8 Delayed disclosure for income statement, balance sheet, and statement of cash flows
Our findings indicate that delayed disclosure of financial statement information has, on
average, a negative effect on how the market processes earnings news. However, the effect of
delayed disclosure may differ across financial statements. For example, Holzman et al. (2019)
show that investors do not always benefit from greater disclosure of income statement items.
We thus examine how the effect of delayed disclosures on investor reaction to earnings news
varies for the income statement, balance sheet, and statement of cash flows. Table 8 reports the
results of estimating regressions (15a) and (15b) using the delayed disclosure separately for the
income statement (DD.IS), the balance sheet (DD.BS), and the statement of cash flows
(DD.SCF). The results show that the delayed disclosure of each of the three statements is
negatively associated with the immediate reaction to earnings news. In contrast, the positive
effect on the post-earnings announcement drift is largely attributable to the delayed disclosure
of balance sheet items.26
4.9 Analyst experience
Bradshaw, Lee, and Peterson (2016) suggest that inexperienced analysts are more likely
to experience forecasting difficulties. We therefore examine whether delayed disclosure has a
stronger effect on inexperienced analysts. We obtain the number of prior quarters for which the
analyst has issued earnings forecast for the firm, partition the analysts in the bottom (top)
quartile as the low (high) experience group, and rerun the analyst reaction tests separately for
the two groups. Untabulated results show that the negative relation between DD and the
proportion of timely forecasts is more pronounced among inexperienced analysts than among
experienced analysts. In contrast, the effects of DD on analyst underreaction to earnings news,
forecast dispersion, and forecast accuracy are similar for both groups. These findings suggest
26 Holzman et al. find that delayed disclosure of operating income items increases PEAD, whereas delayed
disclosure of nonoperating income items decreases PEAD, so the net PEAD effect from delayed disclosure of
items in the overall income statement is weak, which is consistent with our finding.
34
that inexperienced analysts take more time to issue forecasts when DD is high. When they do
issue forecasts, the effects of DD on their forecast quality resemble those for experienced
analysts.
5. Conclusion
A large body of research finds that investors and analysts are slow in updating
expectations about future earnings following earnings news announced in earnings press
releases. This study examines the link between the delay of investor and analyst reaction to
earnings news and delayed financial statement disclosure in earnings press releases. We show
that, for companies that delay financial statement disclosure until the 10-Q filing, investor
immediate reaction to earnings is weaker and investor delayed reaction to earnings after the
10-Q is stronger. We also show that analysts are less likely to issue earnings forecasts after the
earnings announcement and more likely to issue forecasts after the 10-Q filing. Delayed
disclosure is positively associated with analyst underreaction to earnings news and forecast
dispersion and negatively associated with forecast accuracy. Overall, the results suggest that,
when firms delay financial statement disclosure in earnings announcements, investors and
analysts have greater difficulty processing and responding to earnings news.
Our study complements work on expanded financial statement disclosure in earnings
press releases. Research has examined the role of expanded disclosure as a source of additional
information news, which can be viewed as competing with earnings when earnings are less
value relevant. Our study contributes to this literature by showing that delayed financial
statement disclosure is associated not only with a lower amount of information contained in
earnings press releases but also with a delay in investor and analyst reaction to earnings news
that continues even after the delayed financial information is disclosed in 10-Qs/Ks. The
findings are consistent with alleged benefits in the call by regulators and financial
intermediaries for greater financial statement disclosure in earnings press releases.
35
References
Abarbanell, J., & Bernard, V. (1992). Tests of analysts’ overreaction/underreaction to earnings
information as an explanation for anomalous stock price behavior. The Journal of Finance, 47,
1181–1207.
Arif, S., Marshal, N., Schroeder, J., & Yohn, T. (2017). A growing disparity in earnings
disclosure mechanisms: The rise of concurrently released earnings announcements and 10-Ks.
Working paper, Indiana University.
Ball, R., & Brown, P. (1968). An empirical evaluation of accounting income number. Journal
of Accounting Research, 6, 159–178.
Bartov, E., Radhakrishnan, S., & Krinsky, I. (2000). Investor sophistication and patterns in
stock returns after earnings announcements. The Accounting Review, 75, 43–63.
Ben-Rephael, A., Da, Z., & Israelsen, R. (2017). It depends on where you search: Institutional
investor attention and underreaction to news. Review of Financial Studies, 30, 3009–3047.
Bernard, V. L., & Thomas, J. K. (1989). Post-earnings-announcement drift: Delayed price
response or risk premium? Journal of Accounting Research, 27, 1–36.
Bernard, V. L., & Thomas, J. K. (1990). Evidence that stock prices do not fully reflect the
implications of current earnings for future earnings. Journal of Accounting and Economics, 13,
305–340.
Blankespoor, E., Miller, G., & White, H. (2014). The role of dissemination in market liquidity:
Evidence from firms' use of Twitter™. The Accounting Review, 89, 79–112.
Bradshaw, M., Lee, L., & Peterson, K. (2016). The interactive role of difficulty and incentives
in explaining the annual earnings forecast walkdown. The Accounting Review, 91, 995–1021.
Chen, S., DeFond, M., & Park, C. (2002). Voluntary disclosure of balance sheet information
in quarterly earnings announcements. Journal of Accounting and Economics, 33, 229–251.
Chen, S., Miao, B., & Shevlin, T. (2015). A new measure of disclosure quality: The level of
disaggregation of accounting data in annual reports. Journal of Accounting Research, 53,
1017–1054.
Collins, D., Li, O., & Xie, H. (2009). What drives the increased informativeness of earnings
announcements over time? Review of Accounting Studies, 14, 1–30.
DellaVigna, S., & Pollet, J. (2009). Investor inattention and Friday earnings announcements.
Journal of Finance, 64, 709–49.
D’Souza, J., Ramesh, K., & Shen, M. (2010). Disclosure of GAAP line items in earnings
announcements. Review of Accounting Studies, 15, 179–219.
Foster, G., Olsen, C., & Shevlin, T. (1984). Earnings releases, anomalies, and the behavior of
security returns. The Accounting Review, 59, 574–603.
36
Francis, J., Schipper, K., & Vincent, L. (2002). Expanded disclosures and the increased
usefulness of earnings announcements. The Accounting Review, 77, 515–546.
Francis, J., LaFond, R., Olsson, P., & Schipper, K. (2007). Information uncertainty and post-
earnings-announcement-drift. Journal of Business Finance and Accounting, 34, 403–433.
Givoly, D., Li, Y., Lourie, B. & Nekrasov, A. (2019). Key performance indicators as
supplements to earnings: Incremental informativeness, demand factors, measurement issues,
and properties of their forecasts. Review of Accounting Studies, 24, 1147–1183.
Griffin, P. A. (2003). Got information? Investor response to form 10-K and form 10-Q EDGAR
filings. Review of Accounting Studies, 8, 433–446.
Easton, P. D., & Zmijewski, M. E. (1993). SEC form 10-K/10-Q reports and annual reports to
shareholders: Reporting lags and squared market model prediction errors. Journal of
Accounting Research, 31, 113–129.
Hewitt, M. (2009). Improving investors’ forecast accuracy when operating cash flows and
accruals are differentially persistent. The Accounting Review, 84, 1913–1931.
Hirshleifer, D., & Teoh, S. H. (2003). Limited attention, information disclosure, and financial
reporting. Journal of Accounting and Economics, 36, 337–386.
Hirshleifer, D., Lim, S., & Teoh, S. H. (2009). Driven to distraction: Extraneous events and
underreaction to earnings news. Journal of Finance, 64, 2287–2323.
Hirshleifer, D., Lim, S., & Teoh, S. H. (2011). Limited investor attention and stock market
misreactions to accounting information. Review of Asset Pricing Studies, 1, 35–73.
Hirst, D., Koonce, L., & Venkataraman, S. (2007). How disaggregation enhances the credibility
of management earnings forecasts. Journal of Accounting Research, 45, 811–837.
Holzman, E., Marshall, N., Schroeder, J., & Yohn, T. (2019). Is all disaggregation good for
investors? evidence from earnings announcements. Working paper, Indiana University.
Huang, X., Nekrasov, A., & Teoh, S. H. (2018). Headline salience, managerial opportunism,
and over- and underreactions to earnings. The Accounting Review, 93, 231–255.
Kim, J., Nekrasov, A., Shroff, P. & Simon, A. (2019). Valuation implications of unconditional
accounting conservatism: Evidence from analysts' target prices. Contemporary Accounting
Research, 36, 1669–1698.
Kimbrough, M., (2005). The effect of conference calls on analyst and market underreaction to
earnings announcements. The Accounting Review, 80, 189–219.
Klibanoff, P., Lamont, O., & Wizman, T. A. (1998). Investor reaction to salient news in closed-
end country funds. Journal of Finance, 53, 673–699.
Levi, S. (2008). Voluntary disclosure of accruals in earnings press releases and the pricing of
accruals. Review of Accounting Studies, 13, 1–21.
37
Li, E., & Ramesh, K. (2009). Market reaction surrounding the filing of periodic SEC reports.
The Accounting Review, 84, 1171–1208.
Livnat, J., & Mendenhall, A. (2006). Comparing the post–earnings announcement drift for
surprises calculated from analyst and time series forecasts. Journal of Accounting Research,
44, 177–205.
Louis, H., Robinson, D., & Sbaraglia, A. (2008). An integrated analysis of the association
between accrual disclosure and the abnormal accrual anomaly. Review of Accounting Studies,
13, 23–54.
Marshall, N., Schroeder, J., & Yohn, T. (2019). An incomplete audit at the earnings
announcement: Implications for financial reporting quality and the market’s response to
earnings. Contemporary Accounting Research, forthcoming.
Mendenhall, R. R. (1991). Evidence on the Possible Underweighting of Earnings-Related
Information. Journal of Accounting Research, 29, 170–179.
Miao, B., Teoh, S. H., & Zhu, Z. (2016). Limited attention, statement of cash flow disclosure,
and the valuation of accruals. Review of Accounting Studies, 21, 473–515.
Petersen, M. (2004). Information: Hard and soft. Working paper, Northwestern University.
Petersen, M. (2009). Estimating standard errors in finance panel data sets: Comparing
approaches. Review of Financial Studies, 22, 435–480.
Rangan, S., & Sloan, R. (1998). Implications of the integral approach to quarterly reporting for
the post-earnings-announcement drift. The Accounting Review, 73, 353–371.
Schroeder, J. (2016). The impact of audit completeness and quality on earnings announcement
GAAP disclosures. The Accounting Review, 91, 677–705.
Trueman, B. (1994). Analyst forecasts and herding behavior. Review of Financial Studies, 7,
97–124
You, H., & Zhang, X. (2011). Limited attention and stock price drift following earnings
announcements and 10-K filings. China Finance Review International, 1, 358–387.
Zhang, Y. (2008). Analyst responsiveness and the post-earnings-announcement drift. Journal
of Accounting and Economics, 46, 201–215.
38
Appendix A: Price at the filing date for delayed disclosure regime
This appendix describes date 2 price in the delayed disclosure regime (DD-Regime) in the
model of Section 2. Date 1 is the earnings announcement. Date 2 is the 10-Q filing. And date
3 is the terminal payoff. The fraction of inattentive investors is f at date 1 and f’ at date 2. Given
the fixed cost of attending and the greater timeliness of the earnings announcement compared
to the 10-Q filing, it is reasonable to assume that f’ > f, so there are more inattentive investors
at date 2 than at date 1. Inattentive investors I at date 1 will continue to be inattentive at date 2.
Investors paying attention at date 1 observe e but not (as it is unavailable), and investors
paying attention at date 2 observe both e and . Thus a fraction 1 – f’ of investors pay attention
to both e and , and an additional fraction f’ – f of investors pay attention only to e. We use
superscripts I, e, and (e,) to denote the three investor groups.
The date 2 price is determined by market clearing, so that the total demand for the risky
asset by the three investor groups equals its outstanding supply. The pricing equation at date 2
will be a weighted average of beliefs of these three groups of investors, and the weights depend
on f and f’ as well as the respective precision ((P3) = 1/var(P3)) of the terminal value P3 as
perceived by them.
],|[]|[][),( 3,,
332 ePEePEPEeP eeeeII , (1A)
where 3,
3 3 3
( )
( ) ( ' ) ( | ) (1 ') ( | , )
II
I e e
f P
f P f f P e f P e
(2A)
3,
3 3 3
( ' ) ( | )
( ) ( ' ) ( | ) (1 ') ( | , )
ee
I e e
f f P e
f P f f P e f P e
(3A)
,
, 3,
3 3 3
(1 ') ( | , )
( ) ( ' ) ( | ) (1 ') ( | , )
ee
I e e
f P e
f P f f P e f P e
(4A)
39
Appendix B: Variable definitions
Variable Definition
ABS.FE The absolute value of earnings forecast error for quarter t+1
(Actual EPSjt+1 minus Forecasted EPSjt+1) scaled by the stock price
at the end of quarter t. Forecasted EPSjt+1 is the median of all
forecasts issued after the quarter t earnings announcement and
before the earnings announcement for quarter t+1. Forecast errors
in two subwindows are also examined: before the 10-Q filing, [0,
Filing Date -2], and after the 10-Q filing, [Filing Date, Next EA-2].
ANA.FOLLOW Analyst following, measured as the natural logarithm of the number
of analysts following the firm.
BIG4 The indicator variable that equals one if the firm’s annual financial
statement was audited by Deloitte, PricewaterhouseCoopers, Ernst
& Young, or KPMG.
CAR[window] The cumulative abnormal return over a given time window. Five
subwindows are used: around the earnings announcement [-1,+1],
between the earnings announcement and the 10-Q filing [+2, Filing
Date -2], around the 10-Q [Filing Date -1, Filing Date +1], between
the 10-Q and the next earnings announcement [Filing Date +2, Next
EA -2], and around the next earnings announcement [Next EA -1,
Next EA +1]. The abnormal return is calculated as the raw stock
return minus the return on the respective Fama-French reference
portfolio based on size and book-to-market ratio (six portfolios are
formed based on size and book-to-market, 2 x 3).
CAR_DRIFT CAR for the period starting two days after the earnings
announcement for quarter t and ending one day after the earnings
announcement for quarter t+1 [+2, Next Earnings Announcement
Date +1].
CAR_EA CAR for the earnings announcement window [-1,+1].
DD The delayed disclosure ratio, calculated as the number of financial
statement items that are missing in the earnings announcement but
reported in the corresponding 10-Q filing for a given firm-quarter
scaled by the number of nonmissing financial statement items
reported in the corresponding 10-Q filing.
DD.IS, DD.BS, and DD.SCF are the delayed disclosure ratios of
income statement items, balance sheet items, and cash flow
statement items, respectively.
EA.LAG The natural logarithm of the number of days between the fiscal
quarter end and the earnings announcement date for quarter t.
40
EARN.VOL The standard deviation of earnings before extraordinary items
scaled by total assets, calculated over past eight quarters.
%FORECASTS_[window] The number of analyst earnings forecasts issued in a given
subwindow divided by the total number of earnings forecasts issued
in the window starting on the day of the earnings announcement
and ending two days before the earnings announcement for quarter
t+1. Four subwindows are used: around the earnings announcement
[0,+1], between the earnings announcement and the 10-Q filing
[+2, Filing Date -2], around the 10-Q filing [Filing Date -1, Filing
Date +1], and between the 10-Q filing and the next earnings
announcement [Filing Date +2, Next EA -2].
%FORECASTS_EA %FORECASTS for the earnings announcement window [0,+1].
FUTURE.RSUE The decile rank of the analyst earnings forecast error for quarter
t+1 (Actual EPSjt+1 minus Forecasted EPSjt+1) scaled by the stock
price at the end of quarter t, where Forecasted EPSjt+1 is the median
of all forecasts issued after the earnings announcement for quarter
t and before the earnings announcement for quarter t+1. Forecast
errors in two subwindows are also examined: before the 10-Q filing
[0, Filing Date -2] and after the 10-Q filing [Filing Date, Next EA-
2].
HIGH.TECH The indicator variable that equals one for firms in the following
industries: drugs (SIC code 2833–2836), R&D services (8731–
8734), programming (7371–7379), computers (3570–3577),
electronics (3600–3674), and precise measurement instruments
(3810–3845).
INST.OWN Institutional ownership, measured as the sum of all institutional
holdings of the stock scaled by the number of shares outstanding.
The measure is based on the most recent institutional ownership
data available in month t-1.
JMOB The indicator variable that equals one if the current earnings
surprise is between 0 and 3 cents per share and zero otherwise.
LOSS The indicator variable that equals one if the firm reports a negative
net income before extraordinary items in the current quarter and
zero otherwise.
MERGER The indicator variable that equals one if the firm engages in
mergers and acquisitions in the current quarter and zero otherwise.
MISS The indicator variable that equals one if quarter t earnings are
below the analyst forecast and zero otherwise.
MTB The market-to-book ratio measured at the end of the fiscal quarter
t.
41
NEG.SUE The indicator variable that equals one if quarter t earnings are less
than the consensus earnings forecast (or less than earnings for the
same quarter last year when analyst consensus forecast is not
available), and zero otherwise.
NRANK The decile rank of the number of same-day earnings
announcements by other firms.
PRO.FORMA The indicator variable that equals one if IBES actual earnings per
share is not equal to earnings per share excluding extraordinary
items (Compustat item EPSPXQ) and zero otherwise.
RET The stock returns prior to the earnings announcement measured
over the window [-60,-3].
RET.VOL The indicator variable that equals one if the standard deviation of
daily stock returns over prior 250 days is above industry median
and zero otherwise.
RSUE The decile rank of standardized unexpected earnings, SUE, where
SUE is calculated as the difference between announced quarterly
earnings as reported by I/B/E/S and the consensus earnings
forecast, scaled by stock price at the end of previous fiscal quarter.
When consensus analyst forecast is not available, SUE is calculated
as announced earnings minus earnings for the same quarter last
year, scaled by stock price at the end of previous fiscal quarter.
SIZE The natural logarithm of market value of equity at the end of the
fiscal quarter t.
SMOOTH The correlation between change in total accruals and change in
operating cash flows over past eight quarters, multiplied by -1.
SPEC.ITEM The indicator variable that equals one if the special item variable
(SPIQ) in Compustat is non-zero, and zero otherwise.
STD.EST The standard error of analyst forecasts for quarter t+1 EPS that are
issued after the quarter t earnings announcement and before the
earnings announcement for quarter t+1, scaled by the stock price at
the end of quarter t. Standard errors of forecasts in two subwindows
are also examined: before the 10-Q filing, [0, Filing Date -2], and
after the 10-Q filing, [Filing Date, Next EA-2].
YOUNG The indicator variable that equals one if the firm’s age is below the
two-digit SIC industry median and zero otherwise. Firm age is the
number of years from initial public listing to the current fiscal
quarter.
42
Appendix C: Sample selection
Firm-quarter observations from the Compustat Preliminary History database over
1990–2013, excluding year 1999 486,742
minus: Utilities and financial services firms (SIC codes 4900–4999 and 6000–
6999) (116,878)
minus: Observations with less than three days between the earnings announcement
date or with the earnings announcement date before fiscal quarter end (10,138)
minus: 10-Q/K financial statements have less than five data items (7,746)
minus: No preliminary data (64,416)
minus: Missing stock return data from CRSP (116,503)
minus: The 4th fiscal quarters (49,254)
Sample for Market Reaction Tests 121,807
minus: No analyst earnings forecasts for the next quarter after the current quarter’s
earnings announcement (55,044)
Sample for Analyst Reaction Tests 66,763
43
Date 2
Filing date
Date 1
Earnings announcement
Shareholders receive
terminal payoff, P3
Date 3
Terminal date
Fraction of inattentive
investors, f’ > f Fraction of inattentive
investors, f Price, P0
Firm releases earnings, e,
and financial statement item, δ
Firm releases earnings, e Firm releases financial
statement item, δ
Figure 1: Timeline for the model
Timely disclosure regime
(TD-Regime):
Delayed disclosure regime
(DD-Regime):
Date 0
Prior to earnings
announcement
44
0%
20%
40%
60%
80%
100%
PER
CEN
TAG
E O
F A
NA
LYST
S FO
REC
AST
S
TIME
Low DD
High DD
Next EA10-Q/K FilingEA
Figure 2: Analyst delayed response to earnings announcements
This figure shows the timing of analyst earnings forecasts for firms with DD in the bottom decile (the
solid blue line) and firms with DD in the top decile (the dashed red line). The horizontal axis shows
time intervals between the current earnings announcement and next earnings announcement. The
vertical axis shows the fraction of total forecasts that are issued by the indicated point in
time: %FORECASTS_[0, τ], where τ = +1, Filing Date-2, Filing Date+1, or Next EA-2. All variables
are defined in Appendix B.
EA-1 EA+1 Filing-2 Filing+1 Next EA-2
Low DD 0.0% 53.0% 65.3% 67.2% 100.0%
High DD 0.0% 14.1% 38.4% 43.4% 100.0%
45
0%
20%
40%
60%
80%
100%P
ERC
ENTA
GE
OF
REA
CTI
ON
CO
MP
LETE
D
TIME
DD=0
DD=1
Next EA10-Q/K FlingEA
Figure 3: Investor delayed response to earnings announcements
This figure shows the timing of investor reaction to earnings news for firms with DD=0 (the solid blue
line) and firms with DD=1 (the dashed red line). The vertical axis shows the percentage of total reaction
completed. The horizontal axis shows time intervals between the current earnings announcement and
next earnings announcement. The graph is constructed the following way: We run regressions CAR = α
+ β1 RSUE + β2 DD + β3 RSUE * DD + Controls + ε with CAR over different windows [-1, τ], where τ
= +1, Filing Date-2, Filing Date+1, Next EA-1, or Next EA+1. All variables are defined in Appendix
B. For firms with DD=0, the reaction up to time τ, relative to the total reaction, is β1[-1, τ]/β1[-1, Next
EA+1]. For firms with DD=1, the reaction up to time τ, relative to the total reaction, is (β1[-1, τ]+ β3[-
1, τ])/(β1[-1, Next EA+1]+β3[-1, Next EA+1]).
EA-1 EA+1 Filing-1 Filing+1 Next EA-1 Next EA+1
DD=0 0.0% 60.7% 69.6% 72.6% 97.8% 100.0%
DD=1 0.0% 38.3% 45.0% 45.8% 93.3% 100.0%
46
Table 1 Descriptive statistics
Variable Mean Median Std.Dev. P25 P75
DD 0.572 0.566 0.254 0.430 0.803
SUE 0.001 0.000 0.042 -0.001 0.003
CAR_EA 0.003 0.001 0.088 -0.040 0.046
CAR_DRIFT -0.007 -0.011 0.241 -0.138 0.116
#FORECASTS_EA 3.992 2.000 4.700 1.000 6.000
%FORECASTS_EA 0.396 0.375 0.324 0.067 0.667
%FORECASTS_[Filing+2, Next EA-2] 0.392 0.375 0.298 0.143 0.600
MV 2,301.3 295.9 6,918.9 75.9 1,209.2
MTB 2.975 2.086 3.154 1.285 3.485
RET 0.022 0.014 0.231 -0.102 0.135
#Analysts 5.044 3.000 5.434 0.000 7.000
INST.OWN 0.490 0.494 0.279 0.251 0.721
#EA 231.3 220.0 134.0 120.0 336.0
The table reports descriptive statistics of key variables over the sample period, 1990–2013.
#FORECASTS_EA is the number of forecasts issued within two trading days of the earnings
announcement date [0,+1]. %FORECASTS_[Filing +2, Next EA -2] is %FORECASTS for the window
between the 10-Q and the next earnings announcement. MV is the market value of equity at the end of
the fiscal quarter, in millions of dollars. #Analysts is the number of analysts following the firm. #EA is
the number of same date earnings announcements. All other variables are defined in Appendix B.
47
Table 2 Correlations among variables
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
(1) DD
(2) SUE 0.005
(3) CAR_EA 0.002 0.126
(4) CAR_DRIFT -0.021 0.051 0.049
(5) % FORECASTS[0,+1] -0.317 0.037 0.003 0.055
(6) % FORECASTS[Filing +2, Next EA -2] 0.205 -0.017 -0.003 -0.087 -0.652
(7) SIZE -0.268 -0.038 -0.022 -0.043 0.143 -0.027
(8) MTB -0.041 0.011 -0.025 0.003 0.017 -0.003 0.275
(9) RET 0.017 0.105 -0.026 0.002 -0.000 0.001 -0.023 0.107
(10) ANA.FOLLOW -0.268 -0.058 -0.010 -0.024 0.138 -0.026 0.741 0.150 -0.080
(11) INST.OWN -0.349 -0.043 -0.004 -0.031 0.271 -0.139 0.645 0.075 -0.054 0.598
(12) NRANK -0.136 0.012 -0.012 0.017 0.064 -0.069 0.085 0.059 -0.019 0.078 0.087
The table reports Pearson correlations among key variables. Correlations significant at the 10% level or lower are reported in bold. All the variables are as
defined in Appendix B.
48
Table 3 Total and immediate market reaction to earnings news
CAR_TOT or CAR_EA = α + β1 RSUEjt + β2 DDjt + β3 RSUEjt*DDjt
+ Controls + RSUEjt * Controls + εjt (15a)
Total Reaction Immediate Reaction
[0, Next EA+1] Pred. [-1,+1] [-1,+1]
Coeff. p-value sign Coeff. p-value Coeff. p-value
(1) (2) (3) (4) (5) (6)
RSUE 0.195 *** <0.001 + 0.082 *** <0.001 0.110 *** <0.001
RSUE*DD -0.014 0.215 H1a: - -0.036 *** <0.001
DD -0.001 0.845 0.015 *** <0.001
SIZE -0.082 *** <0.001 -0.011 *** <0.001 -0.007 *** <0.001
MTB 0.004 *** <0.001 0.000 * 0.096 -0.000 * 0.056
RET -0.115 *** <0.001 -0.030 *** <0.001 -0.037 *** <0.001
ANA.FOLLOW -0.007 *** 0.003 0.005 *** <0.001 -0.008 *** <0.001
INST.OWN -0.013 0.174 -0.002 0.360 -0.022 *** <0.001
NRANK 0.016 *** 0.003 -0.000 0.974 0.007 *** <0.001
RSUE*SIZE -0.015 *** <0.001 -0.009 *** <0.001
RSUE*MTB -0.002 * 0.088 0.001 *** <0.001
RSUE*RET 0.090 *** <0.001 0.012 *** 0.005
RSUE*ANA.FOLLOW 0.031 *** <0.001 0.026 *** <0.001
RSUE*INST.OWN 0.008 0.552 0.037 *** <0.001
RSUE*NRANK -0.011 0.194 -0.013 *** <0.001
Firm FE Yes Yes Yes
Year-Quarter FE Yes Yes Yes
#obs 121,807 121,807 121,807
R2 0.176 0.169 0.177
The table reports results of the regression to test the association between the delayed disclosure ratio and total and immediate market reaction to earnings news.
All variables are defined in Appendix B. *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively, based on two-tailed tests.
49
Table 4 Delayed investor reaction to earnings news
Panel A: Overall delayed investor reaction
CAR_DRIFTjt = α + β1 RSUEjt + β2 DDjt + β3 RSUEjt*DDjt
+ Controls + RSUEjt * Controls + εjt (15b)
Predicted [+2, Next EA+1] [+2, Next EA+1]
sign Coeff. p-value Coeff. p-value
(1) (2) (3) (4)
RSUE + 0.054 *** <0.001 0.076 *** <0.001
RSUE*DD H1b: + 0.025 ** 0.015
DD -0.017 *** 0.008
SIZE -0.077 *** <0.001 -0.074 *** <0.001
MTB 0.002 *** <0.001 0.004 *** <0.001
RET -0.037 *** <0.001 -0.078 *** <0.001
ANA.FOLLOW 0.003 ** 0.019 0.001 0.682
INST.OWN -0.007 0.284 0.008 0.420
NRANK 0.011 *** 0.001 0.008 0.102
RSUE*SIZE -0.006 ** 0.012
RSUE*MTB -0.003 *** 0.002
RSUE*RET 0.079 *** <0.001
RSUE*ANA.FOLLOW 0.004 0.250
RSUE*INST.OWN -0.027 ** 0.034
RSUE*NRANK 0.004 0.590
Firm FE Yes Yes
Year-Quarter FE Yes Yes
#obs 121,807 121,807
R2 0.144 0.146
50
Panel B: Delayed investor reaction within subwindows
[+2, Filing Date -2] [Filing Date -1,
Filing Date +1]
[Filing Date +2,
Next EA -2]
[Next EA -1,
Next EA +1]
Coeff. p-value Coeff. p-value Coeff. p-value Coeff. p-value (1) (2) (3) (4) (5) (6) (7) (8)
RSUE 0.012 ** 0.015 0.006 ** 0.040 0.040 *** <0.001 0.012 *** <0.001
RSUE*DD -0.004 0.308 -0.003 0.238 0.023 ** 0.010 0.005 ** 0.044
DD 0.001 0.824 0.002 0.290 -0.013 ** 0.025 -0.004 ** 0.028
SIZE -0.011 *** <0.001 -0.003 *** <0.001 -0.050 *** <0.001 -0.004 *** <0.001
MTB 0.000 0.611 0.000 * 0.083 0.003 *** <0.001 0.000 *** 0.009
RET -0.032 *** <0.001 -0.002 0.255 -0.045 *** <0.001 0.002 0.277
ANA.FOLLOW -0.000 0.941 0.001 0.123 0.001 0.727 -0.001 0.344
INST.OWN 0.001 0.868 0.005 *** 0.009 -0.002 0.809 0.003 0.177
NRANK 0.003 * 0.080 0.002 0.101 0.003 0.456 0.002 0.214
RSUE*SIZE -0.001 0.182 0.000 0.447 -0.002 0.361 -0.002 *** <0.001
RSUE*MTB -0.000 0.225 -0.000 0.249 -0.002 * 0.087 -0.000 ** 0.047
RSUE*RET 0.023 *** <0.001 -0.002 0.397 0.060 *** <0.001 -0.000 0.925
RSUE*ANA.FOLLOW 0.005 *** 0.001 -0.000 0.841 -0.002 0.625 0.000 0.881
RSUE*INST.OWN 0.016 *** 0.001 -0.006 ** 0.034 -0.029 ** 0.010 -0.002 0.484
RSUE*NRANK -0.003 0.386 0.001 0.466 0.008 0.254 -0.002 0.197
Firm FE Yes Yes Yes Yes
Year-Quarter FE Yes Yes Yes Yes
#obs 121,807 121,807 121,807 121,807
R2 0.093 0.090 0.137 0.084
The table reports results of the regression to test the association between the delayed disclosure ratio and delayed investor reaction to earnings news. The
dependent variable in Panel A is the cumulative abnormal return over the period starting two days after the earnings announcement for quarter t and ending one
day after the earnings announcement for quarter t+1. In Panel B, the dependent variable is the cumulative abnormal return over the indicated subwindow. All
variables are as defined in Appendix B. *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively, based on two-tailed tests.
51
Table 5 Analyst delayed response to earnings announcements
%FORECASTSjt = α + β1 DDjt + Controls + εjt
(16)
Time Windows
Pred.
[0, +1] Pred.
[+2, Filing Date -2] Pred.
[Filing Date -1,
Filing Date +1] Pred.
[Filing Date +2,
Next EA -2]
Sign Coeff. p-value Sign Coeff. p-value Sign Coeff. p-value Sign Coeff. p-value
(1) (2) (3) (4) (5) (6) (7) (8)
DD H3a: - -0.019 *** 0.006 ? -0.007 0.299 ? 0.001 0.758 H3b: + 0.025 *** 0.001
RSUE 0.018 *** <0.001 0.006 0.111 -0.002 0.251 -0.022 *** <0.001
SIZE -0.009 *** <0.001 -0.018 *** <0.001 -0.003 *** <0.001 0.030 *** <0.001
MTB -0.002 *** <0.001 0.001 *** 0.003 0.000 0.929 0.001 ** 0.045
RET 0.003 0.519 0.008 * 0.091 0.002 0.404 -0.013 ** 0.018
ANA. FOLLOW 0.018 *** <0.001 0.007 *** 0.004 -0.001 0.369 -0.024 *** <0.001
INST.OWN 0.041 *** <0.001 -0.005 0.569 0.004 0.242 -0.040 *** <0.001
NRANK -0.005 0.265 0.018 *** <0.001 -0.001 0.490 -0.012 ** 0.016
Firm FE Yes Yes Yes Yes
Year-Quarter FE Yes Yes Yes Yes
#obs 66,763 66,763 66,763 66,763
R2 0.520 0.308 0.202 0.347
The table reports results of the test of analyst delayed response to earnings news. The dependent variable is the proportion of analyst earnings forecasts made
in the indicated time window, relative to the total number of forecasts made between the current earnings announcement and next earnings announcement. All
variables are defined in Appendix B. *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively, based on two-tailed tests.
52
Table 6 Analyst underreaction to earnings news and forecast accuracy
Panel A: Analyst underreaction to earnings news
FUTURE.RSUEjt+1 = α + β1 RSUEjt + β2 DDjt + β3 RSUEjt*DDjt + Controls +
RSUEjt * Controls + εjt (17a)
Pred.
[0, Filing Date -2] [Filing Date,
Next EA -2]
Sign Coeff. p-value Coeff. p-value Coeff. p-value
(1) (2) (3) (4) (5) (6)
RSUE + 0.218 *** <0.001 0.023 0.465 0.047 0.189
RSUE*DD H4a: + 0.066 *** 0.010 0.059 ** 0.033
DD -0.043 *** 0.004 -0.037 ** 0.024
SIZE -0.033 *** <0.001 -0.048 *** 0.000 -0.031 *** 0.000
MTB 0.002 *** <0.001 0.004 *** 0.001 0.001 0.601
RET 0.130 *** <0.001 0.094 *** 0.000 0.035 ** 0.035
ANA.FOLLOW -0.009 ** 0.019 -0.004 0.554 -0.010 0.247
INST.OWN -0.031 ** 0.015 -0.030 0.116 -0.049 ** 0.017
NRANK 0.013 ** 0.033 0.019 * 0.079 0.018 0.127
RSUE*SIZE 0.032 *** 0.000 0.011 * 0.055
RSUE*MTB -0.003 0.131 -0.000 0.892
RSUE*RET 0.072 *** 0.004 0.057 ** 0.047
RSUE*ANA.FOLLOW -0.008 0.531 0.026 * 0.084
RSUE*INST.OWN -0.005 0.852 -0.007 0.827
RSUE*NRANK -0.012 0.516 -0.003 0.886
Firm FE Yes Yes Yes
Year-Quarter FE Yes Yes Yes
#obs 60,808 60,808 54,472
R2 0.213 0.214 0.211
53
Panel B: Analyst forecast accuracy
ABS.FEjt+1 = α + β1 DDjt + β2 ABS.FEjt + β3 MISSjt +β4 ABS.FEjt*MISSjt +Controls+εjt (17b)
Pred. [0, Filing Date -2] [Filing Date , Next EA -2]
Sign Coeff. p-value Coeff. p-value (3) (4) (5) (6)
DD H4b: + 0.003 *** <0.001 0.003 *** <0.001
ABS.FE 0.002 *** <0.001 0.003 *** <0.001
MISS -0.005 *** <0.001 -0.005 *** <0.001
ABS.FE*MISS 0.017 *** <0.001 0.015 *** <0.001
SIZE -0.007 *** <0.001 -0.006 *** <0.001
MTB 0.000 0.986 0.000 0.274
RET -0.006 *** <0.001 -0.004 *** <0.001
ANA.FOLLOW 0.002 *** <0.001 0.002 *** <0.001
INST.OWN -0.009 *** <0.001 -0.007 *** <0.001
NRANK -0.001 0.204 -0.000 0.487
Firm FE Yes Yes
Year-Quarter FE Yes Yes
#obs 60,808 54,472
R2 0.601 0.608
Panel C: Analyst forecast dispersion
STD.ESTjt+1 = α + β1 DDjt + β2 ABS.FEjt + β3 MISSjt + β4 ABS.FEjt*MISSjt +
Controls + εjt (17c)
Pred. [0, Filing Date -2] [Filing Date , Next EA -2]
Sign Coeff. p-value Coeff. p-value (3) (4) (5) (6)
DD H4c: + 0.001 *** <0.001 0.001 *** <0.001
ABS.FE 0.001 *** <0.001 0.001 *** <0.001
MISS -0.002 *** <0.001 -0.002 *** <0.001
ABS.FE*MISS 0.005 *** <0.001 0.006 *** <0.001
SIZE -0.002 *** <0.001 -0.002 *** <0.001
MTB 0.000 ** 0.036 -0.000 0.102
RET -0.001 *** <0.001 -0.002 *** <0.001
ANA.FOLLOW 0.001 *** <0.001 0.001 *** <0.001
INST.OWN -0.003 *** <0.001 -0.003 *** <0.001
NRANK 0.000 0.233 -0.000 0.609
Firm FE Yes Yes
Year-Quarter FE Yes Yes
#obs 50,296 41,078
R2 0.668 0.631
Panel A reports results of the test of the association between the delayed disclosure ratio and analyst
underreaction to earnings news. Panel B reports results of the test of the association between the delayed
disclosure ratio and analyst forecast accuracy. Panel C reports results of the test of the association
between the delayed disclosure ratio and analyst forecast dispersion. The sample sizes in Panel C are
smaller than in Panel A/B due to the requirement of at least two analyst forecasts to compute forecast
dispersion. All variables are as defined in Appendix B. *, **, *** denote significance at the 10%, 5%,
and 1% levels, respectively, based on two-tailed tests.
54
Table 7 Additional analyses: using residual delayed disclosure
Panel A: First-stage estimation
DDjt = α + β1 RSUEjt + β2 SIZEjt + β3 MTBjt + β4 RETjt + β5 ANA.FOLLOWjt
+ β6 INST.OWNjt + β7 EA.LAGjt + β8 SMOOTHjt + β9 JMOBjt
+ β10 PRO.FORMAjt + β11 SPEC.ITEMjt + β12 EARN.VOLjt
+ β13 LOSSjt + β14 NEG.SUEjt + β15 HIGH.TECHjt
+ β16 MERGERjt + β17 YOUNGjt + β18 RET.VOLjt + β19 BIG4jt + εjt
(18)
Predicted Coeff. p-value
sign (1) (2)
RSUE ? -0.010 ** 0.016
SIZE - -0.003 *** <0.001
MTB ? 0.001 ** 0.021
RET ? 0.002 0.381
ANA.FOLLOW - -0.015 *** <0.001
INST.OWN - -0.061 *** <0.001
EA.LAG + 0.057 *** <0.001
SMOOTH + 0.003 0.312
JMOB + -0.009 *** <0.001
PRO.FORMA - -0.004 *** 0.002
SPEC.ITEM - -0.014 *** <0.001
EARN.VOL - -0.280 *** <0.001
LOSS - -0.009 *** <0.001
NEG.SUE - -0.010 *** 0.005
HIGH.TECH - -0.005 *** <0.001
MERGER - -0.002 0.284
YOUNG - -0.028 *** <0.001
RET.VOL - -0.006 *** <0.001
BIG4 - -0.019 *** <0.001
Industry FE Yes
Year FE Yes
#obs 111,034
R2 0.346
55
Panel B: Second-stage estimation
Pred.
Time Window Pred
. Time Window
Sign Coeff. p-value Sign Coeff. p-value
Immediate and Delayed Market Reaction to Earnings News
CAR = α + β1 RSUEjt + β2 RES.DDjt + β3 RSUEjt*RES.DDjt
+ Controls + RSUEjt * Controls + εjt [0, +1] [+2, Next EA+1]
RES.DD*RSUE - -0.052 *** <0.001 + 0.042 *** <0.001
[+2, Filing Date -2] [Filing Date-1, Filing Date +1]
RES.DD*RSUE - -0.008 ** 0.031 ? -0.001 0.719
[Filing Date +2, Next EA -2] [Next EA -1, Next EA +1]
RES.DD*RSUE + 0.035 *** <0.001 + 0.013 *** 0.001
Analyst Delayed Response to Earnings Announcements
%FORECASTSjt = α + β1 RES.DDjt + Controls + εjt
[0, +1] [+2, Filing Date -2]
RES.DD - -0.020 *** 0.004 ? -0.001 0.847
[Filing Date -1, Filing Date +1] [Filing Date +2, Next EA -2]
RES.DD ? 0.000 0.874 + 0.020 *** 0.008
Analyst Underreaction to Earnings News
FUTURE.RSUEjt = α + β1 RSUEjt + β2 RES.DDjt + β3 RSUEjt*RES.DDjt
+ Controls + RSUEjt * Controls + εjt [0, Filing Date -2] [Filing Date, Next EA -2]
RES.DD*RSUE + 0.059 ** 0.021 ? 0.049 * 0.076
Analyst Forecast Errors
ABS.FEjt+1= α + β1 RES.DDjt + Controls + εjt
[0, Filing Date -2] [Filing Date, Next EA -2]
RES.DD + 0.003 *** <0.001 ? 0.002 *** <0.001
Analyst Forecast Dispersion
STD.ESTjt+1 = α + β1 RES.DDjt + Controls + εjt
[0, Filing Date -2] [Filing Date, Next EA -2]
RES.DD + 0.001 *** <0.001 ? 0.001 *** <0.001
Panel A reports the results of the regression of delayed disclosure ratio, DD, on a set of explanatory
variables. Panel B reports the results of the tests using the residual delayed disclosure ratio, RES.DD,
estimated from the regression in Panel A. Panel B regressions include the base set of control variables
(SIZE, MTB, RET, ANA.FOLLOW, INST.OWN, and NRANK), their respective interactions with RSUE,
and firm and year-quarter fixed effects. All variables are as defined in Appendix B. *, **, *** denote
significance at the 10%, 5%, and 1% levels, respectively, based on two-tailed tests.
56
Table 8 Additional analyses: the effect of delayed disclosure across the income
statement, balance sheet, and statement of cash flows
Coeff. p-value Coeff. p-value Coeff. p-value
Immediate Market Reaction to Earnings News
Y = CAR_EA
RSUE*DD.IS -0.028 *** <0.001
RSUE*DD.BS -0.021 *** <0.001
RSUE*DD.SCF -0.016 *** <0.001
Delayed Market Reaction to Earnings News
Y = CAR_DRIFT
RSUE*DD.IS 0.021 0.146
RSUE*DD.BS 0.017 *** 0.008
RSUE*DD.SCF 0.007 0.420
Delayed Market Reaction in Sub-windows
Y = CAR [+2, Filing Date-2]
RSUE*DD.IS -0.004 0.490
RSUE*DD.BS -0.003 0.280
RSUE*DD.SCF -0.001 0.695
Y = CAR [Filing Date-1, Filing Date+1]
RSUE*DD.IS -0.004 ** 0.009
RSUE*DD.BS -0.001 0.079
RSUE*DD.SCF 0.002 * 0.251
Y = CAR [Filing Date+2, Next EA-2]
RSUE*DD.IS 0.027 ** 0.031
RSUE*DD.BS 0.016 *** 0.004
RSUE*DD.SCF 0.003 0.696
Y = CAR [Next EA-1, Next EA+1]
RSUE*DD.IS 0.003 0.335
RSUE*DD.BS 0.003 ** 0.046
RSUE*DD.SCF 0.003 0.176
CONTROLS Yes Yes Yes
RSUE*CONTROLS Yes Yes Yes
Firm FE Yes Yes Yes
Year-Quarter FE Yes Yes Yes
#obs 121,807 121,807 121,807
This table reports the results of estimating regressions (15a) and (15b) using the delayed disclosure ratio
of income statement items (DD.IS), balance sheet items (DD.BS), or cash flow statement items
(DD.SCF). All regressions include the base set of control variables (SIZE, MTB, RET, ANA.FOLLOW,
INST.OWN, and NRANK), their respective interactions with RSUE, and firm and year-quarter fixed
effects. All variables are as defined in Appendix B. *, **, *** denote significance at the 10%, 5%, and
1% levels, respectively, based on two-tailed tests.
Recommended