The current issue and full text archive of this journal is available on Emerald Insight at:

www.emeraldinsight.com/1475-7702.htm

RAF
18,3 Do firms use early guidance to
disclose the effect of conservatism
on future earnings?
432 Carlo D’Augusta
Department of Accounting, Jennings A. Jones College of Business,
Received 27 September 2018 Middle Tennessee State University, Murfreesboro, Tennessee, USA, and
Revised 29 December 2018
Accepted 3 February 2019
Giulia Redigolo
Department of Economics, Finance and Accounting, ESADE Business School,
Universitat Ramon Llull, Barcelona, Spain

Abstract
Purpose – By deferring profits and anticipating losses, conservatism makes earnings increases more
persistent and earnings declines more likely to revert. Therefore, the level of conservatism in current earnings
has implications for future earnings expectations. Past research shows that outsiders can fail to understand
these implications. This paper aims to investigate whether firms help outsiders by voluntarily disclosing their
expectations about how conservatism will affect future earnings trends.
Design/methodology/approach – The authors examine the likelihood and content of “early” earnings
guidance – i.e. guidance about future earnings that is released around or before the announcement of current
earnings. The sample is made of 8,820 annual earnings announcements, 62 per cent of which are combined
with early guidance.
Findings – The authors find that the more conservative current earnings, the higher: the likelihood that the
firm releases early guidance; the likelihood that the firm predicts a positive change in earnings; and the
difference between the forecasted earnings and current earnings. The authors also find such guidance to be
relevant to analysts, who use it to update their forecasts.
Practical implications – By showing that firms use early guidance to disclose the effect of conservatism
on future earnings, the study is interesting to users and preparers because it shows that analysts need and use
such disclosure; and regulators because it alleviates concerns about the information consequences of
conservatism.
Originality/value – The findings show that firms do not refrain from committing to positive early
guidance to disclose the earnings effects of conservatism. This is interesting in light of the difficulty of
predicting such effects, the manager incentives to keep expectations low and the cost of committing to
positive guidance instead of less risky qualitative disclosure alternatives. In this way, the authors contribute
to the literature on the interrelation between voluntary disclosure and conservatism in financial reports.
Keywords Voluntary disclosure, Accounting conservatism, Earnings guidance
Paper type Research paper

1. Introduction
We investigate whether firms use early earnings guidance to voluntarily disclose their
expectations about how conservative accounting will affect next year’s earnings. We define
Review of Accounting and
Finance
guidance to be “early” if it is released on or before the current year’s earnings announcement
Vol. 18 No. 3, 2019
pp. 432-455
day. Following Weil et al. (2012), we define conservative accounting as the implementation
© Emerald Publishing Limited of accounting methods that, in the absence of certainty, tend to anticipate expenses and
1475-7702
DOI 10.1108/RAF-09-2018-0203 losses but to defer revenues and gains. Because of the properties of the double-entry system,
conservatism has two main effects. On the balance sheet, it results in the understatement of Conservatism
accumulated net assets because of the higher (lower) valuation of liabilities (assets): this type and early
of effect has been referred to as “balance sheet conservatism” (Sunder et al., 2018). On the
income statement, conservatism results in lower current earnings because of the faster
guidance
(slower) recognition of economic expenses (revenues). In this paper, we focus on the income
statement effects of conservatism. We define current earnings to be “conservatively
reported” if they are lower due to the application of conservative accounting methods; in a
given year, we define a firm to be “more conservative” than another firm if it reports current 433
earnings more conservatively[1]. As noted by Weil et al. (2012, p.754), the effects of
conservatism are temporary: “over long-enough time spans, income is cash-in less cash-out.
If a (conservative) reporting method shows less income in early periods, it must show higher
income in some later period”. In other words, higher conservatism on the current year’s
income statement will reduce current earnings because of the faster recognition of expenses
and losses and/or the deferral of uncertain profits. However, it will also increase future
earnings when the recognition of the deferred profits eventually begins and the recognition
of expenses and losses stops or declines. Therefore, conservatism in the current income
statement affects the earnings time-series in a way that makes current increases (decreases)
in earnings more (less) persistent (Basu, 1997). Thus, the level of conservatism in current
earnings has implications for future earnings expectations: if in a given year two firms
report the same current change in earnings but, in that year, one firm has been more
conservative than the other, the future change in earnings will likely be higher (i.e. more
positive or less negative) for the more conservative firm relative to the less-conservative one.
The literature shows that outsiders may fail to fully understand these implications,
which can have negative repercussions on the usefulness of income statement numbers
(Chen et al., 2014; Penman and Zhang, 2002; Mensah et al., 2004; Louis et al., 2014; Barth
et al., 2017). This has led regulators to question the desirability of conservatism in financial
statements (Watts, 2003; FASB, 2010). Compared to outsiders, managers likely have
superior information about the effects of conservatism on future earnings. For instance, they
could be able to predict how much of the deferred revenues and gains will meet recognition
requirements the next year or how much of the current expenses and losses will not persist.
In this paper, we investigate whether managers share this information with outsiders. More
specifically, we study whether managers, when current earnings have been reported
conservatively, voluntarily disclose their expectations of higher future earnings.
We focus on earnings guidance because it is arguably the most direct way to convey
information about future earnings. In addition, the nature of our research question requires
that we do not consider all the earnings guidance released by a firm but that instead we
discriminate based on its timing. Absent any guidance, outsiders may have to wait until
future earnings are announced to learn that such earnings are more favorable than what
past earnings suggested. Releasing guidance later during the year, perhaps just before
future earnings are announced, would be of little benefit, as outsiders would be left for most
of the period with an incomplete information set. On the other hand, releasing guidance
together with (or before) the current earnings announcement would ensure that outsiders
can combine the information about the current earnings with managers’ predictions. Our
research question is, therefore, whether firms release early guidance that incorporates the
higher (lower) expected persistence of positive (negative) earnings components so as to
predict a future earnings trend that is more positive – or less negative – than what outsiders
would infer based on current (conservative) earnings numbers.
As earnings guidance is voluntary, the answer to this question is not clear ex ante. One
reason to disclose the earnings effects of conservatism is represented by managers’
RAF asymmetric payoff function. As their compensation and tenure are tied to earnings
18,3 performance, managers benefit (suffer) from revealing positive (negative) earnings trends
(Burgstahler and Dichev, 1997; Graham et al., 2005). It is, therefore, in managers’ interest to
inform outsiders of the higher (lower) persistence of profits (losses) whose recognition has
been slowed down (accelerated) due to conservatism. Based on this argument, we make
three hypotheses: the more conservative current earnings, the more likely the firm to release
434 early guidance; the more conservative current earnings, the more likely the firm to predict a
positive change in earnings; and regardless of the sign of the predicted change in earnings,
such a change will be more favorable (i.e. more positive or less negative) when current
earnings are more conservative.
Several considerations add tension to these hypotheses. An early commitment to higher
future earnings is risky for firms, due to litigation and reputation concerns and audit costs
(Francis et al., 1994; Krishnan et al., 2012; Lee et al., 2012). They may prefer to wait in
cautious silence until conservatism’s effects on earnings are known with more certainty.
Also, firms could simply be unable to predict such effects so long in advance. Finally, they
may prefer to use less risky qualitative disclosure venues or simply prefer to keep
expectations down to a level that is easier to beat. It is, therefore, an empirical question of
whether firms use early guidance to disclose their expectations about the earnings effects of
conservatism.
We build our sample using annual earnings announcements made by industrial
companies listed on the NYSE, AMEX or NASDAQ and managers’ annual earnings forecast
data relative to the fiscal years 1996 to 2011. The propensity to release guidance varies
across firms, many of which do not release any forecast at all. Prior literature suggests that
two reasons why firms issue guidance are to reduce information asymmetry and to mitigate
litigation risk (Coller and Yohn, 1997; Skinner, 1994). Because both can be attenuated by
conservatism, conservative firms may be less inclined to release forecasts (Hui et al., 2009).
Our goal is not to examine how conservatism affects firms’ motivations to become
forecasters or to release frequent guidance throughout multiple years, which is examined by
other studies (Hui et al., 2009; Jaggi and Xin, 2014; Li, 2008). Rather, our question is whether
a forecaster is willing to use early guidance to reveal the effects of conservatism on future
earnings. Therefore, we only include in the sample firms that have released one or more
forecast of the current year’s earnings (i.e. we exclude non-forecasters) and control for
the number of those forecasts. The final sample is made of 8,820 earnings announcements.
We label “early guidance” the management forecast estimate about next year’s earnings that
is outstanding when current earnings are announced.
Thus, the early guidance estimate will be either the most recent forecast released before
the earnings announcement; or a forecast bundled with the earnings announcement. If early
guidance is outstanding at the time of the announcement, we define the announcement to be
“combined with early guidance”. Approximately 62 per cent of the earnings announcements
in our sample are combined with early guidance.
Because our focus is on the income statement, we use conservatism measures that reflect
the downward effect of conservative reporting on current earnings: the annual build-up of
conservative reserves (Penman and Zhang, 2002), the recording of negative non-operating
accruals, the recording of quarterly earnings that are negatively skewed relative to cash
flows (Givoly and Hayn, 2000) and the extent that current earnings incorporate economic
shocks to performance (Callen et al., 2010). Therefore, the more current earnings have been
lowered by conservatism, the higher the values of the conservatism measures.
The results support our hypotheses. Using a binomial logit model we find that an
earnings announcement is more likely to be combined with early guidance when current
earnings have been reported conservatively. In a multinomial logit setting, we find that this Conservatism
effect is driven by conservatism increasing the probability of positive guidance. In an and early
Ordinary Least Squares (OLS) regression, we find the predicted change in earnings to be
significantly positively associated with the level of conservatism of current earnings.
guidance
In additional analyses, we examine earnings guidance that is released after the
announcement of current earnings (i.e. “later guidance”). We find that conservatism is not
significantly associated with the likelihood of later guidance, suggesting that, when firms
want to disclose the earnings effects of conservatism, they do it in a timely fashion. We also 435
examine the usefulness of early guidance to analysts. Our results suggest that analysts need
early guidance to understand the effect of conservatism on future earnings; and use such
guidance to revise their outstanding forecasts upward.
In the remainder of the paper, we review the literature and develop the hypotheses in
Section 2, we describe the research design in Section 3, we discuss the results in Sections 4
and 5 and we present concluding remarks in Section 6.

2. Literature review and hypothesis development

2.1 Valuation implications of the earnings effects of conservatism
The literature shows that information users may fail to fully understand how conservatism
affects current and future earnings, which, in turn, can have negative repercussions on the
usefulness of earnings information. For instance, Penman and Zhang (2002, p. 238) show
that, when investments are growing, conservative accounting makes current earnings a low-
quality indicator of future earnings by temporarily reducing the former in favor of the latter.
Their results suggest that investors fail to appreciate this effect as they “fixate” on the
reported earnings number “unaware that conservative accounting may lead to reported
earnings of doubtful quality”. Mensah et al. (2004, p. 161) show that conservatism, by
affecting the timing of earnings “with bad news reflected in earnings earlier than good
news”, increases earnings volatility, and thus, makes it harder for analysts to forecast future
earnings. Narayanamoorthy (2006) suggests that investors struggle to fully appreciate the
implications of conservatism on future earnings changes, which causes price drifts.
Separately analyzing conditional and unconditional conservatism, Chen et al. (2014, p. 240)
show that both forms lead to lower earnings persistence and smaller valuation multiples,
suggesting that uncertainty about the timing and magnitude of the effects of conservatism
on earnings “may cause investors to misestimate (and misprice) the impact of conservative
earnings”. In line with Chen et al.’s results, Louis et al. (2014) find that only sophisticated
analysts are able to incorporate the earnings effects of conditional conservatism into their
long-horizon forecasts and Barth et al. (2017) find that conditional conservatism can make
investors take a longer time to understand the valuation implications of current earnings. In
line with the aforementioned concerns, standard setters have taken a stance against
conservatism and in favor of neutral reporting (Watts, 2003; FASB, 2010)[2].

2.2 Conservatism and guidance

The literature has explored some aspects of the relationship between conservatism and
earnings guidance. For instance, Hui et al. (2009) hypothesize that conservatism mitigates
litigation risk and information asymmetry, which, in turn, reduces the need for frequent
guidance. Accordingly, they find that firms that were more conservative between 1992 and
1996 were less likely to release earnings or sales forecasts between 1997 and 2002. By
showing that past conservatism decreases firms’ likelihood to release forecasts over multiple
future years, Hui et al. suggest that conservatism and guidance are linked by an indirect
negative association mediated by lower litigation risk and information asymmetry. We
RAF investigate the existence of a different, additional link. Our question is whether, after
18,3 controlling for differences in firms’ forecasting behavior and other firm-specific confounds,
the fact that current earnings have been reported conservatively makes firms more likely to
release positive guidance when or before current earnings are announced[3].
Different from Hui et al. (2009) and Li (2008) analyzes a later sample period and finds
evidence that conservatism makes firms more likely to issue downward guidance. She
436 argues that, after the passing of Regulation Fair Disclosure, analysts may be unable to
anticipate whether earnings that are soon to be announced will be depressed by
conservative charges. This prompts managers to release negative, short-horizon guidance as
the earnings announcement approaches, to walk analysts’ expectations of current earnings
down to levels that are easier to meet. Different from Li (2008a), we do not aim to re-examine
Hui et al. (2009) argument on the association between conservatism and forecast frequency,
which we rather control for. Also, different from Li (2008), we investigate whether firms
issue upward, long-horizon guidance to predict how conservatism, having lowered current
earnings, will result in a positive future earnings trend. Li (2008) arguments add tension to
our hypotheses because they suggest that conservative managers have incentives to keep
analyst expectations down. Therefore, managers could be reluctant to reveal the positive
effects of conservatism on future earnings.
Re-examining the association between conservatism and forecast frequency on a period
different from those analyzed by Hui et al. (2009) and Li (2008a), Jaggi and Xin (2014) find it
to be negative but weakly significant. They also document that the substitution effect exists
for informative forecasts and especially for bad news ones, but not for opportunistic and
good news forecasts. Therefore, prior literature has examined the relationship between
conservatism and firms’ incentives to release frequent (if any) forecasts, finding mixed
results. As mentioned earlier, our paper is different from the papers discussed above in that
we do not examine whether (or under what conditions) conservatism is associated with
forecast frequency.

2.3 Hypothesis development

When earnings are reported conservatively, revenues and gains are more persistent because
their recognition is deferred and spread over time, while expenses and losses are less
persistent because their recognition is anticipated. As a result, conservatism makes earnings
increases more persistent and earnings declines more likely to revert. If two firms report the
same change in current earnings (e.g. DEt = Et
Et-1 = þ$100), a more-conservative firm
(hereafter MC) is more likely to expect the positive earnings trend to continue than a less
conservative firm (hereafter LC). That is, the future change in earnings forecasted by the
firm (hereafter FDEt), which is defined as the difference between the expected future
earnings and current earnings (Etþ1
Et), will be more favorable for MC than for LC (i.e.
FDEtMC > FDEtLC).
Similarly, if both firms report a current earnings decline (e.g. DEt = Et
Et-1 =
$100),
MC is more likely to expect that the current trend will attenuate or revert, whereas LC is
more likely to expect the negative trend in reported earnings to continue. As a consequence,
the forecasted earnings trend will be more favorable (i.e. less negative or more positive) for
MC than for LC (i.e. again, FDEtMC > FDEtLC).
Research shows that outsiders may fail to understand the effects of conservatism on current
and future earnings trends – i.e. fail to understand the likelihood that FDEtMC > FDEtLC even
though DEtMC = DEtLC. However, managers are arguably better informed about the earnings
effects of conservatism. Managers are also likely to benefit from communicating positive
earnings trends: if a company’s prospects look better than its competitors’, its shareholders are
less likely to sell the stock or fire the managers; similarly, its creditors are less likely to cut the Conservatism
credit supply or charge higher interest rates. Therefore, it is in MC’s managers’ interest to help and early
outsiders distinguish MC from LC.
Based on this argument, we make three empirical predictions. First, conservatism will be
guidance
associated with the likelihood of early guidance. Releasing early guidance is costly: if they fail
to meet their own predictions, managers face a significant risk of litigation (Francis et al., 1994;
Kellog (1984), termination of employment (Lee et al., 2012) and adverse market reaction
(Skinner and Sloan, 2002; Mindak et al., 2016). Even when not missed, positive guidance 437
is associated with higher audit costs (Krishnan et al., 2012), as managerial pressure to
meet higher targets can increase audit risk (Desai, 2016). As managers benefit from
explaining the effects of current conservatism on future earnings, the more
conservative current earnings, the more likely the benefits of releasing guidance to
exceed the costs, all else equal. As a consequence, conservatism will increase the
likelihood of a firm releasing early guidance rather than remaining silent. Therefore, we
express our first hypothesis as follows:

H1. The more conservative current earnings, the more likely the current earnings
announcement to be combined with early guidance.
Second, conservatism will be associated with the sign of early guidance. We argue that
MC uses early guidance to disclose their expectations of favorable earnings trends (i.e.
trends that are more positive or less negative than those of LC, all else equal). It follows
that conservatism’s effect on the probability of early guidance, predicted by H1, and
should be driven by conservatism increasing the probability of positive guidance (i.e.
FDEt > 0) over the probability of no guidance. Therefore, we express our second
hypothesis as follows:

H2. The more conservative current earnings, the more likely the current earnings
announcement to be combined with early guidance that predicts a positive change
in earnings.
Finally, if economic circumstances make both MC and LC predict a positive (negative) change
in earnings, then MC will predict a more positive (less negative) change in earnings than LC
because MC expects the future earnings increase (decline) to be augmented (attenuated) by the
effect of current conservatism on future earnings. As a result, conservatism will be positively
associated with the difference between the forecasted earnings and the current earnings (i.e.
FDEtMC > FDEtLC). Therefore, we express our third hypothesis as follows:

H3. The more conservative current earnings, the higher the difference between the
earnings forecasted by early guidance and current earnings.

3. Research design
3.1 Measurement of variables
Our measure of conservatism (CONS) is a composite measure that is equal to the average of
the decile ranks of four different proxies[4]. The first one is the earnings skewness (SKEW)
proxy proposed by Givoly and Hayn (2000), which is calculated as the difference between
the skewness in operating cash flows and earnings. The second proxy, also based on Givoly
and Hayn (2000), is equal to
1 multiplied by the magnitude of non-operating accruals
(NOACCR). The third proxy is the year-over-year change in the balance sheet reserves
created by conservative accounting, estimated following Penman and Zhang (2002) (DRES).
RAF These three measures are all accounting-based: i.e. they measure conservatism based on its
18,3 downward effect on current earnings. We consider using accounting-based proxies to be an
appropriate decision in the context of our hypotheses. Because of the way they are
constructed, accounting-based proxies are likely to capture how conservatism has decreased
current earnings in a way that makes expected earnings trends more favorable. This is
what, we hypothesize, motivates firms to release positive early guidance in a specific year.
438 Different from the first three proxies, the fourth one is based on both accounting earnings
and market returns (i.e. it is both accounting- and market-based). It is the conservatism ratio
(CRATIO) developed by Callen et al. (2010), which exploits the Vuolteenaho (2002) market-
returns decomposition model to estimate the percentage of a shock to current and future
cash flows that is incorporated in current earnings. By capturing how more-conservative
firms will recognize a larger (smaller) portion of a negative (positive) shock, CRATIO, as the
other three proxies, likely captures how conservatism depresses current earnings in favor of
future ones.
As noted by Callen et al. (2010, p. 146), CRATIO is “in the spirit” of the model developed
by Basu (1997), which is based on the intuition that conservatism results in an asymmetric
association between earnings and stock returns. We acknowledge that the coefficient
estimate of the asymmetry in the Basu (1997) model is arguably one of the most common
market-based measures of conservatism adopted in the literature. However, we believe that
it is not appropriate for our specific research design. Because it is based on estimating the
asymmetry in the earnings-return relationship across many firm-year observations, the
Basu (1997) model does not yield a firm-year specific proxy. For this reason, the model can
be used to test hypotheses where conservatism is the dependent construct (and therefore, the
asymmetric timeliness coefficient is the dependent variable). However, when conservatism
is the independent variable, a firm-year specific proxy is necessary to use it as a variable in a
regression model. While one could use a rolling window of previous earnings and returns to
estimate a firm-year coefficient of asymmetric timeliness, this is not a feasible solution in the
context of our hypotheses: the resulting measure would, in fact, only be available for firms
with a long availability of time-series data, and would end up measuring the firm’s
conservatism in its (sometimes distant) past. As explained in the previous section, our focus
is on analyzing the link between early guidance and current conservatism in earnings[5].
While CRATIO and SKEW have been used in the literature to measure the conditional
form of conservatism (Garcìa Lara et al., 2011; Kim and Pevzner, 2010), NOACCR and DRES
have often been used to capture the unconditional form (Chen et al., 2014; Kim et al., 2018).
As our hypotheses do not differentiate between the two forms, throughout the paper, we
mainly present and discuss the results of the composite measure CONS and report the
results of the other proxies in Section 5.1.
We define early guidance as to the management forecast estimate about the earnings of
the year t þ 1 that is outstanding (i.e. most recently released) when the earnings of the year t
are announced. Therefore, the early guidance estimate will be either a forecast bundled with
the earnings announcement; or an earlier forecast, if no bundled forecast is released. We
measure the change in earnings predicted by the early guidance (FDE) as the difference
between the guidance estimate (or midpoint, if it is a range forecast) and the current
earnings per share, scaled by price[6].
Several control variables are added to all test models. We control for the current earnings
change (DE) and also for a binary variable equal to one if DE is negative (NEGDE). We add the
logarithm of the firm’s market value of equity (MVAL), financial leverage (LEV) and the
market to book ratio (MB) to control for the firm’s size, external scrutiny and growth
opportunities. To control for past earnings levels (PAST_E), we include the average earnings
per share in the years t-1, t-2 and t-3, scaled by stock price and a binary variable equal to one if Conservatism
PAST_E is negative (NEGPAST_E). To control for firm-specific uncertainty, we include the and early
standard deviation of daily returns (SDRET) and the standard deviation of quarterly earnings’
changes (SDEARN). Because a firm’s choice to release guidance can depend on the quality of
guidance
the information provided by analysts, we add several variables related to analysts’ forecasts:
the number of analysts following the firm (FOLLOW), the forecast dispersion (DISP) and the
absolute magnitude of the analyst forecast error (AFE). We also include MISSEXP, a dummy
variable equal to one when current earnings fall below analysts’ expectations and 0 otherwise. 439
As firms that frequently issue forecasts are less likely to remain silent, we add the total number
of forecasts issued by the firm relative to the earnings of year t as a control for forecast
frequency (FREQ). Finally, to capture the firm’s ability in forecasting the earnings of year t, we
include the average management forecast error (F_BIAS) and its absolute value (F_INACCUR).
All our models also include industry and year fixed effects[7].

3.2 Sample construction

The sample is composed of annual earnings announcements made by industrial firms listed on
the NYSE, AMEX or NASDAQ, obtained from the CRSP/Compustat data set[8]. We merge this
sample with managers’ forecast data relative to the fiscal years 1996 to 2011, obtained from
First Call. The initial data set comprises 68,670 announcements. We delete 1,338 observations
not denominated in US$ and 2,961 observations whose share price is lower than $1 (illiquid
stocks). Forecast data are not available for 52,158 firm-year observations. It could be that these
firms never issue forecasts, whether early or not. We cannot include them in our sample
because key variables related to forecasting characteristics (e.g. FREQ, F_BIAS, F_INACCUR)
cannot be computed[9]. We also delete observations lacking data to compute CONS or any of
the control variables that are included in all models.
After the adjustments described above, the sample consists of 8,820 earnings
announcements, 5,502 of which are combined with early guidance (62 per cent of observations).
For around 88 per cent of these 5,502 observations, the early guidance estimate is a bundled
forecast, while for the other observations it is an earlier, stand-alone forecast that was on
average released 110 days before the announcement. Of the 3,318 observations with no early
guidance, 1,174 (i.e. around 35 per cent) released at least one forecast at a later point in time. On
average, the first forecast released by these 1,174 firms was issued 142[10] days after the
earnings announcement. Table I reports the descriptive statistics (all continuous variables are
winsorized at the 1st and 99th percentiles). The values of the four conservatism proxies are in
line with those reported by other studies (Kim and Pevzner, 2010; Kim et al., 2013; Biddle et al.,
2016; Aier et al., 2014; Mensah et al., 2004; Qiang, 2007).
Table II shows the pairwise correlations. In line with prior research (Hui et al., 2009; Khan
and Watts, 2009; LaFond and Watts, 2008), CONS is positively correlated with variables
related to information asymmetry and uncertainty (e.g. LEV, SDRET, SDEARN, DISP) and
negatively with variables related to firm size (MVAL and FOLLOW). Conservatism is also
negatively correlated with FREQ, which is consistent with Hui et al.’s (2009) argument that
more-conservative firms do not need to release guidance as frequently as less-conservative
ones because of lower information asymmetry and litigation risk concerns.

4. Main results
4.1 Conservatism and the probability of early guidance (hypothesis 1)
To test H1, we create a binary variable, EARLYGUID, which identifies earnings
announcements that are combined with early guidance. We then run the binomial logit
model specified by the following equation:
RAF Variable N Mean SD 25th percentile 50th percentile 75th percentile
18,3
SKEW 8,567 0.591 1.557
0.353 0.423 1.500
NOACCR 3,999 0.025 0.056
0.001 0.020 0.045
CRATIO 8,111
0.514 2.002
0.509
0.213 0.056
RES 1,188 0.064 0.192 0.002 0.018 0.077
DRES 960
0.007 0.110
0.006 0.000 0.007
440 FDE 5,502 0.024 0.052 0.003 0.008 0.022
DE 8,820
0.015 0.123
0.013 0.005 0.017
NEGDE 8,820 0.380 0.485 0.000 0.000 1.000
FREQ 8,820 3.634 2.318 2.000 4.000 5.000
F_BIAS 8,820 0.029 0.070
0.002 0.003 0.032
F_INACCUR 8,820 0.035 0.072 0.002 0.008 0.034
MVAL 8,820 21.187 1.622 20.039 21.048 22.242
MB 8,820 3.199 2.954 1.518 2.332 3.746
LEV 8,820 0.796 1.001 0.208 0.460 0.981
SDRET 8,820 0.027 0.013 0.018 0.025 0.034
SDEARN 8,820 0.023 0.030 0.007 0.012 0.025
AFE 8,820 0.004 0.011 0.000 0.001 0.003
DISP 8,820 0.003 0.007 0.000 0.001 0.002
FOLLOW 8,820 9.372 6.577 4.000 8.000 13.000
MISSEXP 8,820 0.320 0.466 0.000 0.000 1.000
PAST_E 8,820 0.038 0.069 0.027 0.048 0.065
NEGPAST_E 8,820 0.123 0.328 0.000 0.000 0.000

Notes: SKEW and NOACCR, CRATIO, RES and DRES, are conservatism proxies based on, respectively,
Givoly and Hayn (2000), Callen et al. (2010) and Penman and Zhang (2002). FDE is the difference between
the early guidance estimate and current earnings. DE is the difference between the current and previous
year’s earnings. NEGDE is a binary variable equal to one if DE is negative and 0 otherwise. FREQ is the
total number of forecasts issued by the firm relative to current earnings. F_BIAS is the average error of the
forecasts issued by the firm relative to current earnings. F_INACCUR is the absolute value of F_BIAS.
MVAL is the logarithm of the total market value of equity. MB is the market-to-book ratio. LEV is the
financial leverage. SDRET is the standard deviation of daily stock returns during the year. SDEARN is the
standard deviation of quarterly earnings’ changes, computed over a rolling window of up to 16 quarters
leading up to the last quarter of the current year. AFE is the absolute magnitude of analyst forecast error
relative to the current earnings before the announcement. DISP is the analyst forecast dispersion before the
announcement. FOLLOW is the number of analysts following the firm before the announcement. MISSEXP
is a binary variable equal one if current earnings miss analyst consensus, 0 otherwise. PAST_E is the
Table I. average value of the earnings of the past three years. NEGPAST_E is a binary variable equal to one if
Descriptive statistics PAST_E is negative, 0 otherwise. Detailed definitions of all variables are provided in the Appendix

EARLYGUIDit = b 0 þ b 1CONSit þ b kCONTROLSkit þ e (1)

The results are reported in Panel A of Table III. Consistent with H1, the coefficient b 1 is
significantly positive, suggesting that reporting conservatively makes firms more likely to release
early guidance to explain the effect of conservatism on future earnings. The results are in line.
[. . .] with expectations. For instance, the coefficients of FREQ, MVAL, SDRET, SDEARN, AFE
and FOLLOW suggest that firms are reluctant to release early guidance when they are smaller,
when they do not frequently forecast or when they operate in a turbulent, unpredictable
environment
(Ajinkya et al., 2005; Chen et al., 2011; Dye, 1986; Houston et al., 2010). Also, the coefficients
of NEGDE and MISSEXP suggest that reporting negative current performance makes firms
unlikely to venture into making early predictions.
Variable 1 2 3 4 5 6 7 8 9 10

CONS 1 0.097***
0.230* 0.169*
0.046* 0.078* 0.088*
0.061* 0.014
0.001
FDE 0.043** 1.000 0.080***
0.059***
0.108*** 0.524*** 0.473***
0.106*** 0.036**
0.030*
DE
0.197*
0.052*** 1
0.716* 0.067*
0.248*
0.127* 0.071* 0.157*
0.117*
NEGDE 0.216*
0.031*
0.449* 1
0.065* 0.186* 0.136*
0.096*
0.209* 0.183*
FREQ
0.047*
0.097*** 0.073*
0.069* 1
0.142*
0.124* 0.236* 0.040* 0.053*
F_BIAS 0.067* 0.824***
0.371* 0.133*
0.137* 1 0.663*
0.109*
0.109* 0.098*
F_INACCUR 0.056* 0.751***
0.375* 0.124*
0.139* 0.945* 1
0.202*
0.192* 0.168*
MVAL
0.065*
0.119*** 0.172*
0.142* 0.254*
0.176*
0.193* 1 0.404*
0.029*
MB 0.006
0.038** 0.121*
0.163* 0.01
0.096*
0.111* 0.316* 1
0.600*
LEV 0.023* 0.069***
0.342* 0.185* 0.005 0.241* 0.289*
0.130*
0.298* 1
SDRET 0.134* 0.052***
0.257* 0.151*
0.215* 0.181* 0.205*
0.468*
0.081* 0.123*
SDEARN 0.146* 0.013
0.018 0.025*
0.095* 0.022* 0.037*
0.170* 0.110*
0.130*
AFE 0.097* 0.135***
0.460* 0.198*
0.098* 0.398* 0.438*
0.279*
0.144* 0.359*
DISP 0.110* 0.065***
0.437* 0.232*
0.107* 0.323* 0.353*
0.260*
0.161* 0.383*
FOLLOW
0.036*
0.064*** 0.037*
0.045* 0.186*
0.063*
0.076* 0.682* 0.183*
0.086*
MISSEXP 0.040* 0.038**
0.128* 0.165*
0.083* 0.116* 0.099*
0.091*
0.075* 0.108*
PAST_E
0.168* 0.077***
0.153* 0.101* 0.087* 0.078* 0.064* 0.123*
0.142* 0.133*
NEGPAST_E 0.163*
0.014 0.068*
0.052*
0.101*
0.016 0.011
0.209* 0.048* 0.005

Notes: CONS is the composite firm-year measure of conservatism. FDE is the difference between the early guidance estimate and current earnings. DE is the
difference between the current and previous year's earnings. NEGDE is a binary variable equal to one if DE is negative and 0 otherwise. FREQ is the total number
of forecasts issued by the firm relative to current earnings. F_BIAS is the average error of the forecasts issued by the firm relative to current earnings.
F_INACCUR is the absolute value of F_BIAS. MVAL is the logarithm of the total market value of equity. MB is the market-to-book ratio. LEV is the financial
leverage. SDRET is the standard deviation of daily stock returns during the year. SDEARN is the standard deviation of quarterly earnings' changes, computed
over a rolling window of up to 16 quarters leading up to the last quarter of the current year. AFE is the absolute magnitude of analyst forecast error relative to the
current earnings before the announcement. DISP is the analyst forecast dispersion before the announcement. FOLLOW is the number of analysts following the
firm before the announcement. MISSEXP is a binary variable equal one if current earnings miss analyst consensus, 0 otherwise. PAST_E is the average value of
the earnings of the past three years. NEGPAST_E is a binary variable equal to one if PAST_E is negative, 0 otherwise. * indicates significance at less than the
10% level. Detailed definitions of all variables are provided in the Appendix
(continued)

diagonal)
and early
Conservatism

correlations (upper
441

spearman
diagonal) and
Pearson (lower
Table II.
guidance
 18,3

442
RAF

Table II.
Variable 11 12 13 14 15 16 17 18

CONS 0.116* 0.144* 0.070* 0.067*
0.036* 0.040*
0.172* 0.096*

FDE 0.045*** 0.050***
0.038**
0.084*** 0.006 0.025
DE
0.060* 0.031*
0.131*
0.179* 0.002
0.145*
0.168* 0.102*
NEGDE 0.102* 0.035* 0.183* 0.243*
0.042* 0.115* 0.113*
0.036*
FREQ
0.224*
0.124*
0.095*
0.096* 0.208*
0.069* 0.078*
0.059*
F_BIAS 0.100* 0.011 0.036* 0.085*
0.034* 0.171* 0.107*
0.036*
F_INACCUR 0.211* 0.070* 0.270* 0.216*
0.108* 0.071* 0.097* 0.022*
MVAL
0.503*
0.223*
0.339*
0.243* 0.682*
0.053* 0.057*
0.114*
MB
0.173* 0.110*
0.390*
0.420* 0.239*
0.095*
0.254* 0.005
LEV
0.078*
0.235* 0.298* 0.379*
0.050* 0.087* 0.360*
0.028*
SDRET 1 0.318* 0.273* 0.256*
0.196* 0.029*
0.250* 0.148*
SDEARN 0.310* 1 0.151* 0.147*
0.095* 0.001
0.365* 0.219*
AFE 0.317* 0.101* 1 0.578*
0.232* 0 0.003 0.089*
DISP 0.326* 0.116* 0.675* 1
0.053* 0.118* 0.01 0.081*
FOLLOW
0.150*
0.041*
0.145*
0.097* 1
0.045*
0.029*
0.062*
MISSEXP 0.061* 0.006 0.155* 0.162*
0.063* 1
0.001 0.017
PAST_E
0.282*
0.508*
0.065*
0.070* 0.038* 0.01 1
0.315*
NEGPAST_E 0.309* 0.469* 0.110* 0.121*
0.099* 0.01
0.719* 1
 Panel A: Binomial logit model (H1) Panel B: Multinomial logit model (H2) Panel C: OLS model (H3)

Effect of conservatism (CONS) on the probability combined Effect of conservatism (CONS) on the probability that an Association between conservatism (CONS) and the that an
with an early positive change in earnings (CAT_FDE) = 1) announcement is combined with early guidance predicting the announcement is predicted by early guidance relative to the
relative to guidance (FDE) magnitude of the change in the probability of no early guidance probability of no early guidance (FDE)
DEPVAR: EARLYGUID DEPVAR: CAT_FDE DEPVAR: FDE
Exp. sign Coeff t-stat Exp. sign Coeff t-stat Exp. sign Coeff t-stat
CONS þ 0.034** (2.39) CONS þ 0.055*** (3.77) CONS þ 0.018*** (3.85)
DE þ 0.007** (2.39) DE þ 0.010*** (2.89) DE þ 0.0148** (5.52)
NEGDE

0.199*** (
3.25) NEGDE

0.333*** (
5.14) NEGDE

0.087*** (
4.90)
FREQ þ 0.934*** (22.53) FREQ þ 0.944*** (22.61) FREQ þ/

0.001 (
0.17)
F_BIAS þ/

0.015 (
1.05) F_BIAS þ 0.108*** (4.15) F_BIAS þ 0.158*** (16.87)
F_INACCUR þ/
0.010 (0.70) F_INACCUR þ/

0.100*** (
3.84) F_INACCUR þ/

0.002 (
0.24)
MVAL þ 0.202*** (4.15) MVAL þ 0.213*** (4.26) MVAL þ/

0.033** (
2.48)
MB þ/

0.000 (
0.00) MB þ/
0.016 (0.49) MB þ/
0.007 (0.73)
LEV þ/

0.007 (
0.19) LEV þ/

0.038 (
0.98) LEV þ/

0.032* (
1.68)
SDRET

0.091** (
2.09) SDRET

0.113** (
2.48) SDRET þ/

0.016 (
0.98)
SDEARN

0.048 (
1.44) SDEARN

0.070** (
1.97) SDEARN þ/
0.006 (0.51)
AFE

0.125*** (
3.25) AFE

0.159*** (
3.49) AFE þ/
0.010 (0.24)
DISP

0.027 (
0.70) DISP

0.058 (
1.25) DISP þ/

0.006 (
0.22)
FOLLOW

0.020*** (
3.10) FOLLOW

0.023*** (
3.45) FOLLOW þ/

0.001 (
0.89)
MISSEXP

0.212*** (
3.69) MISSEXP

0.178*** (
3.03) MISSEXP

0.051*** (
2.89)
PAST_E þ/
0.010* (1.72) PAST_E þ/
0.011* (1.73) PAST_E þ/
0.002 (0.75)
NEGPAST_E þ/
0.042 (0.37) NEGPAST_E þ/
0.043 (0.35) NEGPAST_E þ/
0.029 (0.70)
Constant þ/

4.704*** (
7.16) Constant þ/

5.024*** (
7.41) Constant þ/
0.800*** (3.49)
Industry/Year FE YES Industry/Year FE YES Industry/Year FE YES
Observations 8,820 Observations 8,820 Observations 5,502
2 2 2
Pseudo R 0.209 Pseudo R 0.203 Adj. R 0.718

Notes: All continuous variables are winsorized at the first and 99th percentiles. The t-statistics are reported in parentheses. Standard errors are heteroskedasticity robust (White 1980) and clustered
at the firm level. The dependent variable in the binomial logit model (Panel A), EARLYGUID, is a categorical variable that takes the value of one if there is early guidance, and 0 otherwise. The
dependent variable in the multinomial logit model (Panel B), CAT_FDE, is a categorical variable that takes the value of 0 if there is no early guidance; one if the announcement is combined with
positive early guidance (FDE > 0); minus one if the announcement is combined with negative early guidance (FDE < 0). The dependent variable in the OLS model (Panel C), FDE, is the difference
between the early guidance estimate and current earnings. CONS is the composite firm-year measure of conservatism. *, ** and *** indicate significance at the 10, 5 and 1%levels respectively.
Detailed definitions of all variables are provided in the Appendix

EARLYGUID ð panel aÞ; CAT_F DE ð panel BÞ or F DE ð panel C Þ ¼ b 0 þ b 1CONS þ b 2DE þ

b 3NEGDE þ b 4FREQ þ b 5F_BIAS þ b 6F_INACCUR þ b 7MVAL þ b 8MB þ b 9LEV þ
b 10SDRET þ b 11SDEARN þ b 12AFE þ b 13DISP þ b 14FOLLOW þ b 15MISSEXP þ b 16PAST_E þ
b 17NEGPAST_E þ «

future earnings?
Table III.

the effect of
and early
Conservatism

conservatism on
firms use early
Test of the
443

guidance to predict
hypotheses – Do
guidance
RAF 4.2 Conservatism and the probability that FDE > 0 (hypothesis 2)
18,3 Different from H1, H2 discriminates based on the sign of the guidance. To test this
hypothesis, we run a multinomial logit regression specified by the following equation:

CAT_FDEit ¼ b 0 þ b 1 CONSit þ b k CONTROLSkit þ e (2)

where the dependent variable is a categorical variable that can take three alternative values:
444 zero, if the announcement is not combined with guidance; one if the announcement is
combined with positive guidance (i.e. if FDE > 0); and minus one if the announcement is
combined with negative guidance. Compared to the binomial model used to test H1, which
did not distinguish between positive and negative guidance, the multinomial model allows
testing whether reporting conservatively increases the probability of positive guidance
relative to the probability of no guidance. The results are reported in Panel B of Table III.
Consistent with H2, CONS is significantly positive, suggesting that conservative firms are
more likely to release positive early guidance rather than no guidance[11]. The coefficients
of the other variables are also in line with expectations. For instance, the coefficients of
FREQ, MVAL, SDRET, SDEARN, DISP and FOLLOW are consistent with those of Panel
A. Similarly, the coefficients of NEGDE and MISSEXP indicate that firms prefer cautious
silence over predicting earnings increases when current earnings are disappointing,
probably because of litigation risk.

4.3 The association of conservatism with FDE when early guidance is released (hypothesis 3)
H3 predicts that, if early guidance is released, conservatism is positively associated with the
difference between forecasted earnings and current earnings. To test H3 we use an OLS
model and regress FDE on CONS and the other controls. The results are presented in Panel
C of Table III. Supporting H3, the coefficient of CONS is significantly positive, suggesting
that the early guidance released by firms that have reported more conservatively is more
favorable (i.e. more positive or less negative) than the early guidance of other firms[12].
Different from Panel A, the coefficients of F_BIAS is positive, which suggests that
conditional on releasing early guidance, firms with a reputation for optimistic forecasting
tend to predict positive earnings changes. Also, the coefficients of FREQ, SDRET and
SDEARN are not significant, suggesting that guidance frequency and information
uncertainty may affect a firm’s likelihood to remain silent, but not the forecast estimates if a
firm chooses to release guidance. Similar to Panel B, firms reporting disappointing earnings
seem to incorporate less good news in early guidance, perhaps waiting to assess the
situation with better precision before committing to risky predictions.

5. Additional analyses
5.1 Alternative measures
In this section, we discuss the results obtained using alternative proxies for the two main
constructs of our models, i.e. the forecasted earnings trend (FDE) and the conservatism in
current earnings (CONS).
5.1.1 The forecasted earnings trend (FDE). In the results reported so far, FDE is
measured using the actual earnings number from the First Call database. This number often
excludes items that may be considered non-operating or non-recurring, and therefore, is
more likely to reflect a measurement of the “core” earnings of the firm rather than the
reported earnings. This exclusion, however, could also be removing part of the earnings
effects of conservatism from FDE: Conservatism reduces current earnings by deferring
revenues and gains and by anticipating expenses and losses, which can be accomplished by
disseminating negative charges throughout the income statement. Some of these charges Conservatism
could be “special” items[13] – e.g. asset write-downs, goodwill impairment charges or and early
restructuring costs. Special items have various implications for future earnings trends.
Being less likely to recur than other items, they can result in a more favorable (i.e. more
guidance
positive or less negative) future earnings trend. Also, current write-downs can decrease
future core expenses such as depreciation or cost of goods sold. Moreover, special items such
as restructuring costs can reflect managerial efforts to cut unprofitable investments and
focus on the core business, which will have a favorable impact on future revenues and 445
operating profits. Nevertheless, there is evidence that the capital markets fail to understand
special items’ implications for future earnings[14], which could prompt managers to release
early guidance to disclose such effects. Therefore, it is important to repeat the tests of our
hypotheses without excluding from the measurements the effects of conservatism on
current earnings that occur through special items. We do so by calculating FDE using the
current earnings number from Compustat, which includes special items.
The results (not tabulated for brevity) are consistent with those reported in Table III. We
note that the coefficients’ magnitude grows materially (0.077 vs 0.055 in Panel B, 0.049 vs
0.018 in Panel C). This growth is consistent with FDE now capturing also the earnings
effects of conservatism that occur through special items and managers’ desire to disclose
such effects using early guidance.
5.1.2 The effects of conservatism on current earnings (CONS). In the literature, CRATIO
and SKEW have often been used to measure the conditional application of conservative
accounting (Garcìa Lara et al., 2011; Kim and Pevzner, 2010), while NOACCR and DRES
have often been used to capture the unconditional form (Chen et al., 2014; Kim et al., 2018).
To test whether either form is the main driver of our results, we analyze the two forms
separately by calculating a conditional (unconditional) proxy, labeled CCONS (UCONS), as
the average of the decile ranks of CRATIO and SKEW (of the decile ranks of NOACCR and
DRES). We then further split each of the two proxies into its individual components (i.e. the
decile rank of each proxy).
When testing H1, the results (not tabulated for brevity) appear to be somewhat stronger if
we use UCONS (0.044, t-stat 3.37) compared to CCONS (0.025, t-stat 1.87). This effect seems to
be mostly driven by NOACCR. On the other hand, the significance of CRATIO and SKEW
appears to be weak when isolating the two proxies, which support H1 only when combined. H2
and H3 are supported whether we use CCONS or UCONS. The main drivers of these results
appear to be CRATIO and NOACCR (for both H2 and H3), and DRES (for H3), while SKEW is
not significant for either hypothesis. However, SKEW significantly supports both H2 and H3
when the measurement of FDE includes special items. Overall, the main results appear to be
driven by both conditional and unconditional conservatism proxies, though using SKEW
yields results that are sensitive to the measurement of the dependent variable.

5.2 Later guidance

In this section, we look at post-announcement guidance patterns. We create a dummy
variable (LATERGUIDE) equal to one if the firm has released a forecast of the earnings of
the year t þ 1 after the three-day window centered on the day of the announcement of the
year t earnings, and 0 otherwise. We then run the binomial logit regression of equation (1)
using LATERGUIDE as a dependent variable, first on a subsample made of firms that did
not release early guidance, then on the subsample made of firms that did. In neither case we
find CONS to be statistically significant (the t-statistics are 1.41 and 0.48, respectively). This
suggests that when firms do not use early guidance to disclose the effects of conservatism
on future earnings, they are unlikely to decide to do it later; and when firms do use early
RAF guidance to disclose the effects of conservatism on future earnings, such explanation is
18,3 unlikely to require later updates[15]. Overall, these additional findings reinforce the role
played by early guidance (as opposed to any guidance) to disclose how conservatism affects
earnings.

5.3 The relevance of the early guidance that explains the earnings effects of conservatism
446 The results of the main analysis show that firms use early guidance to disclose how
conservatism is expected to affect future earnings. A different question is whether such
disclosure conveys information that is relevant to analysts (i.e. whether analysts need it and
use it). We address this question in two ways. First, we analyze whether such information is
new to analysts. If this is the case, then conservatism will increase the likelihood that and the
extent to which early guidance beats analyst expectations. To test it, we repeat the analysis
of H2 and H3 using the analyst pre-guidance consensus estimate as a benchmark, to isolate
the portion of the forecast that is new to analysts. We label the variable so constructed
FNEWS (as opposed to FDE). If analysts are able to predict all earnings effects of
conservatism, conservative firms’ guidance will, on average, contain no more news than any
other firms’, and FNEWS will be unassociated with CONS. The results reported in Table IV
suggest the opposite, indicating that analysts need early guidance to predict the effect of
conservatism on future earnings.
Second, we analyze whether analysts use the information about the earnings effects of
conservatism contained in early guidance, i.e. whether they decide to revise their
expectations upward. Specifically, we analyze the change in the analyst consensus estimate
(i.e. the change in the mean value of analyst forecasts relative to the earnings of year t þ 1)
that occurs after the release of positive early guidance (DCONSENSUS). If analysts use
(ignore) such information, positive forecast news will be associated (unassociated) with
upward revisions. Therefore, we test whether the association between the magnitude of
positive FNEWS and DCONSENSUS becomes significantly weaker among more-
conservative firms relative to other firms, which would suggest that conservative firms’
guidance contains information that analysts do not use. The results reported in Table V
suggest the opposite as the coefficient of FNEWS is not significant in the low-conservatism
subsample but becomes stronger and significant when conservatism is high. Overall, we
conclude that early guidance released by conservative firms is relevant to analysts.

6. Concluding remarks
In this paper, we show that firms resort to voluntary disclosure to reveal their expectations
about how conservatism affects future earnings trends. In particular, we show that they do
so as early as the announcement of current earnings, which allows outsiders to combine
managers’ predictions with the currently reported numbers. Our findings also suggest that
analysts need such disclosure to correctly predict conservatism’s effects on future earnings,
and use it by revising their expectations upward. Our findings have practical implications
for regulators, preparers and users. Prior literature suggests that outsiders can struggle to
understand the earnings effects of conservatism (Chen et al., 2014; Penman and Zhang, 2002;
Mensah et al., 2004; Louis et al., 2014; Barth et al., 2017). This has led regulators to question
the desirability of conservatism in financial statements (Watts, 2003; FASB, 2010). By
showing that firms use early guidance to mitigate the potential negative impact of
conservatism, our findings can alleviate regulators’ concerns. Moreover, our findings are
interesting for preparers and users of accounting information. When considering the release
of positive guidance to disclose the effect of conservatism on future earnings, managers
weigh the risks and costs of such strategy against its potential benefits: if the former
 Panel A: Multinomial logit model Panel B: OLS model
DV: CAT_FNEW Exp. sign Coeff t-stat DV: FNEWS Exp. sign Coeff t-stat

Effect of conservatism (CONS) on the probability that an announcement is combined with early guidance Association between conservatism (CONS) and the difference between the early guidance
exceeding analysts’ consensus (CAT_FNEWS = 1) rather than not combined with guidance (CAT_FNEWS estimate and analysts’ consensus before the announcement (FNEWS)
= 0)
CONS þ 0.058*** (3.49) CONS þ 0.017*** (3.16)
DE þ 0.016*** (3.92) DE þ 0.021*** (6.84)
NEGDE

0.327*** (
4.53) NEGDE

0.063** (
2.44)
FREQ þ 0.972*** (21.68) FREQ þ/
0.009 (0.71)
F_BIAS þ 0.047*** (2.81) F_BIAS þ 0.103*** (5.80)
F_INACCUR þ/

0.015 (
0.92) F_INACCUR þ/
0.042** (2.30)
MVAL þ 0.175*** (3.03) MVAL þ/

0.037 (
1.44)
MB þ/
0.031 (0.89) MB þ/
0.005 (0.32)
LEV þ/

0.029 (
0.69) LEV þ/

0.085*** (
3.46)
SDRET

0.134** (
2.55) SDRET þ/

0.076*** (
3.48)
SDEARN

0.044 (
1.11) SDEARN þ/

0.024* (
1.71)
AFE

0.086* (
1.88) AFE þ/

0.136** (
2.48)
DISP

0.153*** (
2.74) DISP þ/

0.033 (
1.03)
FOLLOW

0.019** (
2.48) FOLLOW þ/

0.003 (
1.12)
MISSEXP

0.581*** (
8.13) MISSEXP

0.168*** (
7.71)
PAST_E þ/
0.022*** (2.97) PAST_E þ/
0.007** (2.28)
NEGPAST_E þ/

0.030 (
0.21) NEGPAST_E þ/

0.054 (
1.18)
Constant þ/

5.256*** (
6.79) Constant þ/
1.226*** (3.40)
Industry/Year FE YES Industry/Year FE YES
Observations 8,680 Observations 5,362
Pseudo R2 0.182 Adj. R2 0.585

Notes: All continuous variables are winsorized at the first and 99th percentiles. The t-statistics are reported in parentheses. Standard errors areheteroskedasticity robust (White 1980) and clustered at
the firm level. The dependent variable in the multinomial logit model (Panel A) CAT_FNEWS, is a categorical variable that takes the value of zero if there is no early guidance; one if the
announcement is combined with early guidance that is higher than the pre-guidance analyst consensus (FNEWS > 0); minus one if the announcement is combined with early guidance that is not
higher than the pre-guidance analyst consensus (FNEWS < 0). The dependent variable in the OLS model (Panel B), FNEWS, is the difference between the early guidance estimate and the pre-guidance
analyst consensus. CONS is the composite firm year measure of conservatism. *, ** and *** indicate significance at the 10, 5 and 1% levels respectively. Detailed definitions of all variables are
provided in the Appendix

CAT_FNEWS ðPanel AÞ or FNEWS ðPanel BÞ ¼ b 0 þ b 1CONS þ b 2DE þ b 3NEGDE þ b 4FREQ þ

b 5F_BIAS þ b 6F_INACCUR þ b 7MVAL þ b 8MB þ b 9LEV þ b 10SDRET þ b 11SDEARN þ
b 12AFE þ b 13DISP þ b 14FOLLOW þ b 15MISSEXP þ b 16PAST_E þ b 17NEGPAST_E þ «

Table IV.
and early
Conservatism

guidance?
contained in early
of conservatism
analysts need the
447

the earnings effects

information about
guidance – Do
Relevance of early
guidance
RAF Split: high-conservatism vs. low-conservatism firms
18,3
Subsample: LOW CONS Subsample: HIGH CONS Difference Test Exp. Sign
FNEWS 0.005 (0.16) 0.091*** (2.59) 0.086** (2.03) ?
Controls and FE YES YES
Observations 1,520 1,219
Adj. R2 0.139 0.153
448
Notes: All control variables are included in the models, but not tabulated for brevity. All continuous
variables are winsorized at the 1st and 99th percentiles. The t-statistics are reported in parentheses.
Standard errors are heteroskedasticity robust (White 1980) and clustered at the firm level. DCONSENSUS is
the change in analyst consensus relative to the earnings of year t þ 1 that occurs after the forecast release.
FNEWS is the difference between the early guidance estimate and the pre-guidance analyst consensus. To
facilitate the comparison of coefficients across the panels of this table, DCONSENSUS and FNEWS are
scaled by their sample standard deviation. *, ** and *** indicate significance at the 10, 5 and 1% levels,
respectively. Detailed definitions of all variables is provided in the Appendix In the equation above, the
coefficient b 1 represents the association between the change in the mean analyst forecast (DCONSENSUS)
Table V. and the magnitude of FNEWS when the sample is limited to early guidance that beat the pre-guidance
Relevance of early consensus
guidance – Do
DCONSENSUS ¼ b 0 þ b 1FNEWS þ b 2DE þ b 3NEGDE þ b 4FREQ þ b 5F_BIAS
analysts use the new
information about
þ b 6F_INACCUR þ b 7MVAL þ b 8MB þ b 9LEV þ b 10SDRET
the earnings effects
of conservatism
þ b 11SDEARN þ b 12AFE þ b 13DISP þ b 14FOLLOW þ b 15MISSEXP
disclosed through
early guidance?
þ b 16PAST_E þ b 17NEGPAST_E þ «

outweigh the latter, they will not do it. By finding that analysts need and use the information
about conservatism that is contained in early guidance, we show evidence of the benefits of
such disclosures, which could encourage managers to pursue it.
This paper contributes to growing academic research on the relationship between
conservative financial reports and voluntary disclosure practices (Guay and Verrecchia,
2018; Hui et al., 2009; Jaggi and Xin, 2014; D’Augusta and DeAngelis, 2017; Li, 2008;
D’Augusta, 2018). Moreover, this study contributes to the literature on the consequences of
conservatism. Recent studies have focused on its effects on information users, such as
investors in the stock market (D’Augusta et al., 2016; Kim and Zhang, 2016; Kim et al., 2013;
Barth et al., 2017; Chen et al., 2014; Louis et al., 2014) or the debt market (Balakrishnan et al.,
2016; Sunder et al., 2018). By contrast, the main focus of our paper is on the actions of
preparers – i.e. their decision to release positive early guidance. A thorough analysis of the
effect of conservatism on users’ reaction to earnings guidance is beyond the scope of this
paper and we leave it to future research. Similarly, while our study focuses on US
companies, future research could expand the analysis to the international setting, as
conservatism is a phenomenon that has been empirically observed and studied in various
countries (Ding and Stolowy, 2006; Noravesh et al., 2007; Dimitropoulos and Asteriou, 2008;
Ball et al., 2008).

Notes
1. The literature distinguishes between the conditional and the unconditional application of
conservative accounting.
 As per Beaver and Ryan (2005, p. 269), “under unconditional conservatism, the book value Conservatism
of net assets is understated due to predetermined aspects of the accounting process. Under
conditional conservatism, book value is written down under sufficiently adverse
and early
circumstances, but not up under favorable circumstances.” Although the two forms of guidance
conservatism are not merely substitutes of each other as they arise for different reasons and
have different implications for contracting efficiency, the literature shows that “both forms
of conservatism bias [. . .] earnings downward” (Qiang, 2007, p. 770). Our paper does not
focus on the causes of conservatism or its contracting usefulness, but rather on its effects 449
on earnings and their relation to earnings guidance. Therefore, our analysis is not limited
to any of the two forms, but rather examines the earnings effects of conservatism whether it
is applied conditionally or unconditionally.
2. The Statement of Financial Accounting Concepts No. 8 rejects “conservatism as an aspect of
faithful representation because including [it] would be inconsistent with neutrality (BC 3.27, p. 28)”.
Specifically, standard setters’ concerns appear to stem from the undesirability of the earnings
effects of conservatism, which would impair the usefulness of earnings information relative to
neutral reporting: “An admonition to be [conservative] is likely to lead to a bias. Understating
assets or overstating liabilities in one period frequently leads to overstating financial performance
in later periods – a result that cannot be described as prudent or neutral. (BC 3.28, p. 28)”
3. The diversity between our research question and Hui’s et al. (2009) is the reason for various
differences in the research design of the two papers. For instance, Hui et al. (2009) measure
conservatism over a six-year window before the six-year window used to measure forecast
frequency and other forecast attributes. We measure conservatism in the. Year t and then
measure whether the announcement of year t earnings is combined with guidance relative to the
year t þ 1. In addition, we do not look at the probability or the number of forecasts released at
any time in a multiple-year period. We look at the probability of early guidance and whether it
predicts the effects of conservatism on future earnings.
4. To maximize the power of our tests, if a conservatism proxy is unavailable, CONS is equal to the
average of the other available proxies.
5. Similarly, we choose not to use the conditional conservatism proxy developed by Khan and
Watts (2009) (i.e. C_Score). C_Score is a firm-year specific estimate of the coefficient of
asymmetric timeliness in the Basu (1997) model based on the value predicted by the firm
market capitalization, market leverage and market-to-book ratio. Because these three
predictors tend to be very stable over time, C_Score can be an optimal choice to test
hypotheses related to the forces that demand conservatism, to a firm’s reputation for
conservatism and to how the markets price such reputation. It would be suboptimal,
however, in the context of our hypotheses: as between-firms variations in C_Score in a given
year would only reflect variations in the three predictors and not in the extent that
conservatism has depressed a given firm’s current earnings (i.e. C_Score would not capture
the difference between MC and LC in the example of section 2.3), there would be no reason to
hypothesize that C_Score is associated with positive early guidance.
6. In the main analysis, FDE is computed using current earnings obtained from First Call, which
excludes many types of non-operating or non-recurring items. In Section 5.1 we report the results
obtained when using current earnings obtained from Compustat, which includes such items.
7. In untabulated analyses we also include additional controls, such as the number and the value of
the company’s shares owned by the CEO, the tenure of the CEO, the size of the board, the
percentage of executive directors and independent auditors on the board. We do not use these
variables in the main tests because their lack of availability would greatly reduce the sample size.
Including the additional controls always leaves the coefficients of interest practically unaffected,
and sometimes even stronger in magnitude, suggesting that our main models do not suffer from
significant omitted-variable issues.
RAF 8. Financial firms, excluded from the sample, are identified according to Fama and French (1997).
18,3 Market data are obtained from the CRSP daily stock database and analysts’ forecast data from I/
B/E/S’s Detail History. All data are obtained through the Wharton Research Data Services
(WRDS) website.
9. Moreover, our study does not aim to investigate whether conservatism affects the likelihood of
being a forecaster but, rather, whether it makes a forecaster release positive guidance together
450 with or before the annual earnings announcement. Anyway, for robustness we repeat the test the
hypotheses on a broader sample (26,231 observations) that includes observations with missing
forecast data. We find the results to be qualitatively similar, and therefore, robust to alternative
sampling criteria.
10. This suggests that such “late” guidance may be of little help for the purpose of disclosing the
effects of current conservatism on future earnings: indeed, Narayanamoorthy (2006) finds that
investors’ failure to understand such effects gives rise to anomalous return patterns for
approximately 90 days after the current earnings announcements, after which the information
content of quarterly announcements begins to resolve the mispricing.
11. In an untabulated test, we run a binomial logit regression where the dependent variable is equal
to 1 for positive guidance and zero for negative guidance. Consistent with expectations, the
coefficient of CONS is significantly positive. This is also consistent with our test of H3 where
CONS is found to increase FDE: if FDE is higher, it is more likely to be positive than negative, all
else equal. We thank an anonymous reviewer for suggesting this test.
12. To ensure that the results are not influenced by observations that have very low levels of
conservatism and issue negative guidance to reduce litigation risk, we repeat the test using only
observations that are above the fifth decile of CONS. The results (not tabulated for brevity) are
qualitatively similar.
13. Chen et al. (2014, pp. 236, 244) “find that conservative reporting affects earnings persistence beyond
the effects of special items” and note that “special items represent just one possible way to incorporate
conservatism in earnings and not all special items result in more conservative reporting”. Consistent
with this notion, Frankel and Roychowdhury (2008, p. 6) show that “[more-conservative firms] are no
more likely to report significantly large negative special items. Nor are their negative special items
any larger in magnitude as a percentage of sales, on average”. However, they “predict and find that
negative special items of [less-conservative] firms are [. . .] more persistent with respect to future net
income.” Their findings suggest that conservatism’s effects on special items pertains not so much to
their frequency and magnitude, but rather to their implications for future earnings.
14. For instance, Dechow and Ge (2005, pp. 253, 256) find that special items explain future stock returns,
which is “consistent with investors misunderstanding the transitory nature of special items” and
“overweigh the probability that the firm will be unsuccessful [in turning around the current earnings
trend]”. While investors’ misunderstanding of special items could be alleviated by analyst production
of information, Louis et al. (2014, p.20) note that “there is no consensus among analysts regarding the
treatment of special items and, consequently, the earnings forecasted by analysts are not generally
fully adjusted for special items”. Investor valuation of special items can be made harder by
managerial opportunism: Frankel and Roychowdhury (2008) show that not all special items are
“equally special”. Special items reported by more-conservative firms are more predictive of favorable
earnings trends because they reflect fuller recognition of losses and managerial actions to turn the
firm around. By contrast, less-conservative firms are more likely to classify recurring operating
expenses as “special”, which makes their special items more persistent.
15. These findings should not be interpreted as conservative firms necessarily remaining silent after
the release of early guidance: they may well release additional forecasts for various reasons.
However, such reasons do not appear to include managers’ desire to disclose how the level of
conservatism in current earnings is going to affect future earnings.
References Conservatism
Aier, J.K., Chen, L. and Pevzner, M. (2014), “Debtholders’ demand for conservatism: evidence from changes and early
in directors’ fiduciary duties”, Journal of Accounting Research, Vol. 52 No. 5, pp. 993-1027.
guidance
Ajinkya, B.B., Bhojraj, S. and Sengupta, P. (2005), “The association between outside directors,
institutional investors and the properties of management earnings forecasts”, Journal of
Accounting Research, Vol. 43 No. 3, pp. 343-376.
Balakrishnan, K., Watts, R.L. and Zuo, L. (2016), “The effect of accounting conservatism on corporate 451
investment during the global financial crisis”, Journal of Business Finance and Accounting,
Vol. 43 Nos 5/6, pp. 513-542.
Ball, R., Robin, A. and Sadka, G. (2008), “Is financial reporting shaped by equity markets or by debt
markets? An international study of timeliness and conservatism”, Review of Accounting Studies,
Vol. 13 Nos 2/3, pp. 168-205.
Barth, M.E. Landsman, W.R. Raval, V. and Wang, S. (2017), “Asymmetric timeliness and the resolution
of investor disagreement and uncertainty at earnings announcements”, available at: https://
papers.ssrn.com/sol3/papers.cfm?abstract_id=2348905
Basu, S. (1997), “The conservatism principle and the asymmetric timeliness of earnings”, Journal of
Accounting and Economics, Vol. 24 No. 1, pp. 3-37.
Beaver, W.H. and Ryan, S.G. (2005), “Conditional and unconditional conservatism: concepts and
modeling”, Review of Accounting Studies, Vol. 10 Nos 2/3, pp. 269-309.
Biddle, G.C. Ma, M.L. and Song, F.M. (2016), “Accounting conservatism and bankruptcy risk”, available
at: SSRNhttp://papers.ssrn.com/sol3/papers.cfm?abstract_id=1621272
Burgstahler, D.C. and Dichev, I.D. (1997), “Earnings management to avoid earnings decreases and
losses”, Journal of Accounting and Economics, Vol. 24 No. 1, pp. 99-126.
Callen, J.L. and Segal, D. (2010), “A variance decomposition primer for accounting research”, Journal of
Accounting Auditing and Finance, Vol. 25 No. 1, pp. 121-143.
Callen, J.L., Segal, D. and Hope, O.-K. (2010), “The pricing of conservative accounting and the
measurement of conservatism at the firm-year level”, Review of Accounting Studies, Vol. 15
No. 1, pp. 145-178.
Chen, L.H., Folsom, D., Paek, W. and Sami, H. (2014), “Accounting conservatism, earnings persistence
and pricing multiples on earnings”, Accounting Horizons, Vol. 28 No. 2, pp. 233-260.
Chen, S., Matsumoto, D. and Rajgopal, S. (2011), “Is silence golden? An empirical analysis of firms that
stop giving quarterly earnings guidance”, Journal of Accounting and Economics, Vol. 51 Nos 1/2,
pp. 134-150.
Coller, M. and Yohn, T.L. (1997), “Management forecasts and information asymmetry: an examination
of Bid-Ask spreads”, Journal of Accounting Research, Vol. 35 No. 2, pp. 181.
D’Augusta, C. (2018), “Does accounting conservatism make good news forecasts more credible and bad
news forecasts less alarming?”, Journal of Accounting Auditing and Finance, forthcom.
D’Augusta, C., Bar-Yosef, S. and Prencipe, A. (2016), “The effects of conservative reporting on investor
disagreement”, European Accounting Review, Vol. 25 No. 3, pp. 451-485.
D’Augusta, C. and DeAngelis, M.D. (2017), “Does accounting conservatism discipline qualitative
disclosure? Evidence from tone management in the MD&A”, available at: https://papers.ssrn.
com/sol3/papers.cfm?abstract_id=2910403
Dechow, P.M. and Ge, W. (2005), “The persistence of earnings and cash flows and the role of special
items”, Review of Accounting Studies, Vol. 11 Nos 2/3, pp. 253-296.
Desai, N. (2016), “Impact of company level information on auditor conservatism”, Review of Accounting
and Finance, Vol. 15 No. 4, pp. 518-532.
Dimitropoulos, P.E. and Asteriou, D. (2008), “Timeliness, conservatism and financial transparent firms
under the greek accounting setting”, Review of Accounting and Finance, Vol. 7 No. 3, pp. 252-269.
RAF Ding, Y. and Stolowy, H. (2006), “Timeliness and conservatism: changes over time in the properties”,
Review of Accounting and Finance, Vol. 5 No. 2, pp. 92-107.
18,3
Dye, R. (1986), “Proprietary and nonproprietary disclosures”, The Journal of Business, Vol. 59 No. 2,
pp. 331-366.
Fama, E.F. and French, K.R. (1997), “Industry costs of equity”, Journal of Financial Economics, Vol. 43
No. 2, pp. 153-193.
452 FASB (2010), “Conceptual framework for financial reporting”, in Statement of Financial Accounting
Concepts, Vol. 8. Financial Accounting Standards Board.
Francis, J., Philbrick, D. and Schipper, K. (1994), “Shareholder litigation and corporate disclosures”,
Journal of Accounting Research, Vol. 32 No. 2, pp. 137-164.
Frankel, R.R. and Roychowdhury, S. (2008), “Are all special items equally special? The predictive role of
conservatism”, available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1001434
Garcìa Lara, J.M., Garcìa Osma, B. and Penalva, F. (2011), “Conditional conservatism and cost of
Capital”, Review of Accounting Studies, Vol. 16 No. 2, pp. 247-271.
Givoly, D. and Hayn, C. (2000), “The changing time-series properties of earnings, cash flows and
accruals: Has financial reporting become more conservative?”, Journal of Accounting and
Economics, Vol. 29 No. 3, pp. 287-320.
Graham, J.R., Harvey, C.R. and Rajgopal, S. (2005), “The economic implications of corporate financial
reporting”, Journal of Accounting and Economics, Vol. 40 Nos 1/3, pp. 3-73.
Guay, W. and Verrecchia, R.E. (2018), “Conservative disclosure”, Journal of Financial Reporting,
Houston, J.F., Lev, B. and Tucker, J.W. (2010), “To guide or not to guide? Causes and consequences of
stopping quarterly earnings guidance”, Contemporary Accounting Research, Vol. 27 No. 1,
pp. 143-185.
Hui, K.W., Matsunaga, S.R. and Morse, D. (2009), “The impact of conservatism on management
earnings forecasts”, Journal of Accounting and Economics, Vol. 47 No. 3, pp. 192-207.
Jaggi, B. and Xin, H. (2014), Accounting conservatism and management earnings forecasts, available at:
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2246168
Kellogg, R.L. (1984), “Accounting activities, security prices, and class action lawsuits”, Journal of
Accounting and Economics, Vol. 6 No. 3, pp. 185-204.
Khan, M. and Watts, R.L. (2009), “Estimation and empirical properties of a firm-year measure of
accounting conservatism”, Journal of Accounting and Economics, Vol. 48 Nos 2/3, pp. 132-150.
Kim, Y., Li, S., Pan, C. and Zuo, L. (2013), “The role of accounting conservatism in the equity market:
Evidence from seasoned equity offerings”, The Accounting Review, Vol. 88 No. 4, pp. 1327-1356.
Kim, J.B. Nekrasov, A. Pervin, S.K. and Simon, A. (2018), “Valuation implications of unconditional
accounting conservatism: evidence from analysts’ target prices”, available at: https://
onlinelibrary.wiley.com/doi/abs/10.1111/1911-3846.12465
Kim, B.H. and Pevzner, M. (2010), “Conditional accounting conservatism and future negative
surprises: an empirical investigation”, Journal of Accounting and Public Policy, Vol. 29
No. 4, pp. 311-329.
Kim, J.-B. and Zhang, L. (2016), “Accounting conservatism and stock price crash risk: Firm-level
evidence”, Contemporary Accounting Research, Vol. 33 No. 1, pp. 412-441.
Krishnan, G.V., Pevzner, M. and Sengupta, P. (2012), “How do auditors view managers’ voluntary
disclosure strategy? The effect of earnings guidance on audit fees”, Journal of Accounting and
Public Policy, Vol. 31 No. 5, pp. 492-515.
LaFond, R. and Watts, R.L. (2008), “The information role of conservatism”, The Accounting Review,
Vol. 83 No. 2, pp. 447-478.
Lee, S., Matsunaga, S.R. and Park, C.W. (2012), “Management forecast accuracy and CEO turnover”,
The Accounting Review, Vol. 87 No. 6, pp. 2095-2122.
Li, Z. (2008), “Accounting conservatism and managers’ forecasting behavior”, Doctoral Dissertation, Conservatism
University of Minnesota.
and early
Louis, H., Lys, T. and Sun, A. (2014), Conservatism, Analyst Ability, and Forecast Error: Evidence on
Financial Statement Users ‘ Ability to account for Conservatism, available at: https://papers.ssrn. guidance
com/sol3/papers.cfm?abstract_id=1031981
Mensah, Y.M., Song, X. and Ho, S.S.M. (2004), “The effect of conservatism on analysts’ annual earnings
forecast accuracy and dispersion”, Journal of Accounting, Auditing and Finance, Vol. 19 No. 2,
pp. 159-183. 453
Mindak, M.P., Sen, P.K. and Stephan, J. (2016), “Beating threshold targets with earnings management”,
Review of Accounting and Finance, Vol. 15 No. 2, pp. 198-221.
Narayanamoorthy, G.S. (2006), “Conservatism and Cross-Sectional variation in the post–earnings
announcement drift”, Journal of Accounting Research, Vol. 44 No. 4, pp. 763-789.
Noravesh, I., Dianati Dilami, Z. and Bazaz, M.S. (2007), “The impact of culture on accounting: does
gray’s model apply to Iran?”, Review of Accounting and Finance, Vol. 6 No. 3, pp. 254-272.
Penman, S.H. and Zhang, X.-J. (2002), “Accounting conservatism, the quality of earnings, and stock
returns”, The Accounting Review, Vol. 77 No. 2, pp. 237-264.
Qiang, X. (2007), “The effects of contracting, litigation, regulation, and tax costs on conditional and
unconditional conservatism: cross-sectional evidence at the firm level”, The Accounting Review,
Vol. 82 No. 3, pp. 759-796.
Skinner, D.J. (1994), “Why firms voluntarily disclose bad news”, Journal of Accounting Research, Vol. 32
No. 1, pp. 38.
Skinner, D.J. and Sloan, R.G. (2002), “Earnings surprises, growth expectations, and stock returns or
don’t let an earnings torpedo sink your portfolio”, Review of Accounting Studies, Vol. 7 No. 2/3,
pp. 289-312.
Sunder, J., Sunder, S.V. and Zhang, J. (2018), “Balance sheet conservatism and debt contracting”,
Contemporary Accounting Research, Vol. 35 No. 1, pp. 494-524.
Vuolteenaho, T. (2002), “What drives firm-level stock returns?”, The Journal of Finance, Vol. 57 No. 1,
pp. 233-264.
Watts, R.L. (2003), “Conservatism in accounting part I: explanations and implications”, Accounting
Horizons, Vol. 17 No. 3, pp. 207-221.
Weil, R.L., Schipper, K. and Francis, J. (2012), Financial Accounting: An Introduction to Concepts,
Methods and Uses, 14th ed. South-Western College.

Appendix
Variables’ definition:

 CONSit = composite firm-year measure of conservatism, equal to the average of the decile
ranks of SKEW, NOACCR, CRATIO, and DRES. To maximize the power of our tests, if a
conservatism proxy is unavailable, CONS is equal to the average of the other available proxies.
 CRATIOit = annual decile ranking of the conservatism ratio, expressing the proportion
of total earnings news that is recognized in current reported earnings. The ratio is
calculated according to Callen et al. (2010), that is, h 2i,t/Nei,t. The numerator and
denominator are estimated following Callen and Segal (2010). The ratio is multiplied by

1 if the denominator is positive, to reflect the asymmetric timeliness in the recognition
of good and bad news (Biddle et al., 2016). Decile rankings are calculated separately for
positive and negative shocks.
 SKEWit = conservatism measure developed by Givoly and Hayn (2000), equal to the
difference between the skewness in cash flows from operations and the skewness in
RAF earnings before extraordinary items (both scaled by total assets) estimated over a rolling
18,3 window of up to 20 quarters (at least nine required) ending in the fourth quarter of year t.
 NOACCRit = non-operating accruals, measured as per Givoly and Hayn (2000),
multiplied by
1.
 DRESit = difference between RESt and RESt-1, where RES is the conservatism proxy
developed by Penman and Zhang (2002) based on inventory, research and development
454 (R&D), and advertising reserves.
 EARLYGUIDit = binary variable that assumes the value of one if the announcement
of year t earnings is combined with early guidance (i.e. if the announcement is
bundled with or preceded by a management forecast) predicting year tþ1 earnings,
and 0 otherwise.
 FDEit = difference between the forecasted earnings per share (relative to the year t þ 1)
and the current earnings per share (year t), scaled by the stock price.
 CAT_FDEit = Categorical transformation of FDE, which can take three different values:
0, if the announcement is not combined with early guidance (i.e. if there is no value of
FDE); one, if the announcement is combined with early guidance predicting a positive
earnings change (i.e. if FDE > 0); minus one, if the announcement is combined with early
guidance not predicting a positive earnings change (i.e. if FDE # 0).
 DEit = difference between the current earnings per share (year t) and the previous year’s
earnings per share (year t
1), scaled by the stock price.
 NEGDEit = binary variable that assumes the value of one if DEit is negative and 0
otherwise.
 MVALit = natural logarithm of the total market value of equity at the end of the fiscal year.
 LEVit = financial leverage, measured as total liabilities scaled by the total market value
of equity at the end of the fiscal year.
 MBit = market-to-book ratio, measured as the market value of equity divided by the
book value of equity at the end of the fiscal year.
 SDRETit = standard deviation of daily stock returns during the year.
 FOLLOWit = number of analysts following the shares of firm i on the day before the
announcement.
 DISPit = standard deviation of analysts’ forecasts of the earnings per share of year t
scaled by stock price. The standard deviation is calculated on the day before the
announcement, based on forecasts that have been issued or confirmed within the
previous 65 days.
 AFEit = magnitude of analyst forecast error relative to the earnings per share of year t,
measured as the absolute difference between the actual earnings per share and the
analyst consensus estimate measured on the day before the announcement, scaled by
stock price.
 SDEARNit = standard deviation of quarterly earnings’ changes, computed over a rolling
window of up to 16 quarters before the end of year t. Earnings changes are scaled by
total assets at the beginning of the quarter.
 FREQit = number of forecasts issued by the firm relative to the earnings being currently
announced (i.e. the earnings of year t).
 F_BIASit = average error of the forecasts issued by the firm relative to the earnings
being currently announced (i.e. the earnings of year t), scaled by stock price. The error is
calculated as forecasted earnings less actual earnings.
  F_INACCURit = absolute value of F_BIASit. Conservatism
 PAST_Eit = average value of the earnings per share of the years t-1, t-2 and t-3, scaled and early
by stock price. guidance
 NEGPAST_Eit = binary variable that assumes the value of one if PAST_Eit is negative
and 0 otherwise.
 MISSEXPit = binary variable that assumes the value of one if the earnings per share of
firm i in year t is below the analyst consensus estimate measured on the day before the 455
announcement, and 0 otherwise.
 FNEWSit = difference between the forecasted earnings per share and the average of
analyst estimates relative to next year’s earnings, measured five days before the early
guidance release, scaled by stock price.
 CAT_FNEWSit = Categorical transformation of FNEWS, which can take three different
values: 0, if the announcement is not combined with early guidance (i.e. if there is no
value of FNEWS); one, if the announcement is combined with early guidance exceeding
the analyst consensus estimate (i.e. if FNEWS > 0); minus one, if the announcement is
combined with early guidance not exceeding the analyst consensus estimate (i.e. if
FNEWS # 0).
 DCONSENSUSit = change in the analyst consensus estimate relative to the earnings of
year t þ 1 that occurs after the earnings announcement of year t. We calculate the
difference between the average of all outstanding analysts’ forecasts 5 days after the
announcement and the average 5 days before the announcement.

Corresponding author
Carlo D’Augusta can be contacted at: [email protected]

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]