Accounting Conservatism and Managers’ Forecasting Behavior
Accounting
Conservatism and Managers’
Forecasting Behavior
Zining
Li
Carlson
School of Management
University
of Minnesota
Minneapolis,
MN 55455 zli1@csom.umn.edu
February
2007
I am greatly indebted to my adviser, Pervin Shroff,
for her continued encouragement and guidance on this project. I wish to thank
Frank Gigler, Chandra Kanodia, Ram Venkataraman, Raghunathan Venugopalan, and
Ivy Xiying Zhang for their invaluable input. I have also greatly benefited from
discussions with Stephanie Grimm, Huijing Fu, Alexander Nekrasov, Claire
Rosenfeld, Timothy Shields, Frank Fang Yu, and Haiwen Zhang. I gratefully
acknowledge the assistance of First Call
(Thomson Financial) in providing management forecast data. All errors are mine.
Abstract
Conservative accounting results in a
downward bias in reported earnings and net assets in times of investment
growth. Prior research has shown that
analysts do not fully understand this downward bias, which may partially
explain the observed optimism in their earnings forecasts. Managers, faced with the pressure to meet or
just beat analysts’ earnings expectations, have an incentive to issue earnings
guidance that may influence analysts to adjust their forecasts for the effect
of conservatism. This paper examines how
accounting conservatism affects managers’ forecasting behavior. First, I find that both the likelihood of a
firm issuing management earnings forecasts and the forecast frequency increase
in the level of conservatism. Second, I
investigate whether management earnings forecasts inform analysts about the
downward bias in earnings caused by accounting conservatism. I find that the news in management forecasts,
measured relative to the prevailing analyst consensus, is negatively correlated
with the level of conservatism. Further,
this correlation varies with changes in asset growth, consistent with managers
conveying the complex effect of changes in growth on conservatively reported
earnings that is likely to be misestimated by analysts. Finally, I find that, subsequent to the
issuance of management earnings forecasts, analysts revise their forecasts to
incorporate the effect of accounting conservatism conveyed by management
forecasts. Overall, the evidence
suggests that managers provide earnings guidance which conveys information
about the effect of accounting conservatism that is not fully recognized by
analysts.
1.
Introduction
Traditionally, conservatism in accounting means when
in doubt choose the solution that will be least likely to overstate assets and
income (Kieso, Weygandt, and Warfield, 2006).
Thus, conservative behavior involves choosing lower (rather than higher)
revenues, higher (rather than lower) expenses, and recognizing unrealized
losses but not unrealized gains (Beaver, 1998).
In times of growth in investment, this leads to a downward bias in
reported earnings and net assets. If the
effect of the downward bias in conservatively reported earnings is not clearly
understood by analysts, it may lead to overly optimistic analysts’
forecasts. The predictive ability of
analysts is further hampered when a firm’s growth in investment changes,
leading to a change in the effect of conservatism on reported earnings (Penman
and Zhang, 2002). If analysts
consistently and systematically misestimate the effect of accounting
conservatism, do managers bridge this information gap by providing voluntary
disclosures? The objective of this paper
is to examine the relation between accounting conservatism and voluntary
disclosure behavior of managers.
First, I examine whether the
frequency of management forecasts increases with the degree of accounting
conservatism. Prior empirical research
has shown that firms lose a significant portion of their market value when they
miss analysts’ earnings expectations even by a small margin (Skinner and Sloan,
2002). Thus, the pressure to meet or
just beat analysts’ forecasts may provide managers with the incentive to inform
analysts about the effect of accounting conservatism on earnings. Consequently, managers would be more likely
to issue forecasts and would issue forecasts more frequently to align analysts’
forecasts with their own expectation of the forthcoming reported earnings.
Second, I examine whether managers’
earnings guidance informs analysts about the effect of accounting
conservatism. For my sample, I find that
the optimism in initial analysts’ earnings forecasts is positively correlated
with the degree of accounting conservatism, suggesting that analysts do not
fully incorporate the downward bias caused by conservatism. Thus, if managers have incentives to meet or
just beat analysts’ forecasts, they are likely to issue earnings guidance that
may influence analysts to lower the prevailing consensus to account for the
downward bias caused by conservatism.
Further, the downward bias in earnings caused by accounting conservatism
becomes larger (smaller) when investment growth increases (decreases). It is possible that analysts may not
correctly incorporate the effect of the interaction between conservatism and
change in investment growth in their earnings forecasts. Specifically, I examine whether the
difference between management earnings guidance and the prevailing analysts’
consensus is negatively correlated with the level of accounting conservatism
and whether the correlation varies with changes in investment growth.
Third, I attempt to validate my
expectation that analysts may not fully understand the effect of conservatism,
and that management earnings guidance informs analysts about its effect on
earnings. If managers succeed in
conveying the effect of conservatism, I expect analysts to revise their
forecasts downward subsequent to the issuance of management earnings
guidance. Specifically, I examine
whether subsequent analysts’ revisions incorporate the news relating to
accounting conservatism that is reflected in the management guidance.
I use several alternative measures to
capture accounting conservatism: (i) the sum of R&D reserves, advertising
reserves and LIFO reserves (Penman and Zhang, 2002), (ii) accumulated negative
non-operating accruals (Givoly and Hayn, 2000), (iii) the difference between
the skewness of operating cash flows and the skewness of earnings (Givoly and
Hayn, 2000), and (iv) a composite measure that equals the sum of the decile
ranks of the above three measures.
I examine the
cross-sectional relation between accounting conservatism and the frequency of
management forecasts after controlling for other factors that prior research
has found to influence a firm’s voluntary disclosure decision. My empirical results show that managers issue
more forecasts when the firm’s financial accounting is relatively more
conservative. I obtain consistent
results when I employ a Poisson regression to take account of the fact that
management forecasts represent count data.
Results of a probit regression also suggest that the likelihood of a
firm issuing at least one forecast during the sample period increases with the
level of conservatism.
I next examine
whether managers issue guidance to lower analysts’ expectations when their
accounting is conservative. I find that
the difference between management forecast and the prevailing analysts’
consensus has a significant negative correlation with the level of accounting
conservatism, and the negative correlation is stronger when investment growth
is increasing. Additionally, consistent
with management forecasts informing analysts about the effects of accounting
conservatism, I find that analysts’ revisions are positively correlated with
the conservatism-related news that is reflected in the management
guidance.
Overall, my
results suggest that managers provide guidance to analysts which informs them
about the bias caused by accounting conservatism and analysts respond to this
information by revising their forecasts to incorporate the bias. This paper contributes to the literature on
voluntary disclosure by analyzing how accounting conservatism affects the
decisions of managers to disseminate information. My results suggest that accounting
conservatism can be an explanation for the findings of previous research that
analysts’ expectations start out optimistic and then are consistently revised
downward over time. These findings could
be interpreted as management guidance conveying information about the effect of
accounting conservatism on earnings and not necessarily attempting to mislead
investors.
The remainder
of the paper is organized as follows.
Section 2 describes the hypotheses.
Section 3 discusses the sample selection and research design. Empirical results are presented in Section 4,
and Section 5 concludes.
2. Hypotheses Development
Conservatism describes accounting practice which when
consistently applied understates the net assets of a firm. Examples of conservative accounting practices
include LIFO inventory valuation (in times of rising input prices), accelerated
depreciation methods, expensing R&D spending, and overestimating bad debt
allowances and warranty
liabilities. In
general, conservative accounting results in a lower net income, at the same
time creating “hidden” reserves by understating assets or overstating
liabilities.
Prior research has documented that
financial analysts do not fully understand the effect of accounting
conservatism on reported earnings. Pae
and Thornton (2003) find that analyst forecast dispersion and analysts’ forecast
errors are higher for firms reporting bad news than for firms reporting good
news. They interpret their results as
indicating that financial analysts fail to incorporate the effect of
conservatism in their earnings forecasts.
Consistent with this finding, Hui and Matsunaga (2004) document that
analysts’ initial forecast errors are
increasing in the level of accounting conservatism. While prior evidence has shown that analysts’
earnings forecasts are on average optimistic, these authors find that their
forecasts are more optimistic when a firm’s accounting is relatively more
conservative.
2.1 Accounting conservatism and frequency of
management forecasts
Managers, as insiders of the firm,
have private information about the firm’s operating, investing and financing
activities, and thus a better ability to evaluate the effect of accounting
conservatism in predicting future earnings.
I argue that firms with more conservative accounting are more likely to
issue earnings guidance. My hypothesis
is based on the premise that managers are under pressure to meet or beat
analysts’ earnings expectations (DeGeorge, Patel, and Zeckhauser, 1999, and
Skinner and Sloan, 2002). If analysts do
not correctly estimate the effect of accounting conservatism on earnings as
suggested by previous empirical research, managers may have incentives to align
analysts’ expectations with their own predictions of reported earnings through
voluntary disclosures. Thus, I test
whether managers of firms with more conservative accounting tend to make more
voluntary disclosures in the form of earnings guidance. H1: The frequency of management earnings
forecasts increases in the level of accounting conservatism.
Alternatively, Gigler and Hemmer
(2001) predict a negative relation
between accounting conservatism and managers’ voluntary disclosures in a
principal-agent setting.[1] My hypothesis differs from the implications
of the model in Gigler and Hemmer (2001).
One major difference is that their model characterizes managers as
disclosing their private information about the underlying economic income rather than forecasting the accounting earnings to be reported.
In addition, managers’ incentives for voluntary disclosure arise from
the risk-sharing benefit in their model in contrast with meeting or beating
analysts’ expectations on which I base my hypothesis.
2.2 Accounting conservatism and news in
management forecasts
Accounting conservatism in general
results in a downward bias in earnings and creates hidden reserves. The downward bias may be due to either early
recognition of expenses and losses or postponement of revenues and gains or
both. If analysts do not fully
understand the downward bias resulting from either of these two causes, they
are likely to overestimate earnings.
Because managers have incentives to meet or just beat analysts’
forecasts of forthcoming earnings, they may issue earnings guidance to
influence analysts to adjust their forecasts downward. Although the downward revision in analysts’
forecasts may lead to a negative market reaction, the findings in Bartov,
Givoly and Hayn (2002) show that the amount of this negative reaction is
significantly smaller than the premium the firm gains by meeting or just
beating earnings expectations.
Thus, given the incremental rewards for meeting or
beating expectations, I expect that management earnings guidance is likely to
inform analysts about the downward bias due to accounting conservatism to
induce them to adjust their forecasts for conservatism.
An alternative view suggests that,
in the presence of conservative accounting, managers may reveal their private
information about good news, which would result in their earnings guidance
being higher than the prevailing
analyst consensus. This is because
conservative accounting leads to an information deficiency, since revenues and
gains are not disclosed in a timely manner through regular accounting channels
(Guay and Verrechhia, 2006). Thus, managers would use other channels to reveal
good news and mitigate this information deficiency. However, under conservative accounting, good
news may not be included in the forthcoming
reported earnings which are what analysts are forecasting. Thus, I argue that, given the incentive to
meet or just beat analysts’ forecasts, managers would be more likely to convey
good news via their long-term forecasts and exclude such news from their
guidance of forthcoming earnings. Hence,
I expect that under conservative accounting managers would provide earnings
guidance informing the analyst about the downward bias due to conservatism.
H2a: The difference between
management earnings forecast and the prevailing analyst consensus is negatively
correlated with the level of accounting conservatism.
2.2.1
Accounting conservatism and investment growth
While in general accounting conservatism
leads to a downward bias in reported earnings, when a firm’s investment in
assets declines, the effect reverses to an upward bias in reported earnings and
withdrawals from hidden reserves (Penman and Zhang (2002), and Li (2006)).
Thus, the accumulation of reserves accelerates when the growth in investment is
increasing and decelerates (slows down) when the growth in investment is
decreasing. As a result, the effect of
the downward bias on earnings caused by accounting conservatism intensifies when
investment growth increases, and becomes weaker when investment growth
decreases. If analysts do not fully
understand the effect of accounting conservatism, their overestimation of
forecasted earnings will be higher when investment growth increases and lower
when investment growth declines.[2] Thus, managers’ earnings guidance to
influence analysts to adjust for conservatism will be based on the investment
growth regime that the firm faces.
H2b: The difference between management earnings
forecast and the prevailing analyst consensus is more (less) negatively
correlated with the level of accounting conservatism when investment growth
increases (decreases).
2.3 Accounting
conservatism and the effect of management guidance on subsequent analysts’
forecast revisions
It is generally believed that
forecast accuracy is one of the criteria used for ranking analysts. Consistent
with this belief, prior research has shown a positive correlation between
analyst ranking and their forecast accuracy (Stickel, 1992). Mikhail, Walther, and Willis (1999) also show
that forecast accuracy matters to analysts for retention and promotion. If analysts have incentives to achieve
accuracy, they would update their forecasts when new earnings-relevant
information becomes available. Consistent with this, prior research shows that
analysts react to management guidance by revising their forecasts. Ajinkya and Gift (1984) hypothesize and find
that managers issue forecasts to move analysts’ expectations toward management
beliefs about future earnings and subsequent analysts’ revisions are consistent
with the information conveyed by managers’ forecasts. Williams (1996) studies revisions in analyst
consensus subsequent to management earnings guidance and finds that analysts
are more likely to react to management guidance that indicates earnings will be
lower than expected. Cotter, Tuna, and
Wysocki (2006) find that analysts respond to management guidance in a timely
manner: about 52% of analysts revise their forecasts within 5 days of the
issuance of management guidance.
Prior research also studies the
incentives of managers to guide analysts’ forecasts toward the forthcoming
earnings realization. The findings of
Matsumoto (2002) suggest that firms would either manage earnings upward or
guide analysts’ forecasts downward to avoid missing expectations at the
earnings announcement. Richardson, Teoh
and Wysocki, (2004) provide evidence that managers guide analysts’ expectations
down to “beatable” earnings targets.
Cotter, Tuna, and Wysocki (2006) also find that firms that issue
specific (point and range) earnings guidance are more likely to meet or beat
analysts’ expectations. Overall, these
papers suggest that managers may be managing the expectations of analysts in
order to influence analysts to project achievable target earnings.
My analysis is not based on the
“expectation management” hypothesis. I
examine whether managers guide analysts’ forecasts to adjust for the effects of
accounting conservatism. Thus, I focus
only on the component of management guidance that relates to the effect of
accounting conservatism. Although this
component will likely be a downward adjustment in earnings expectation, the
overall management guidance can be either higher or lower than the current
expectation, depending on the nature of other news that managers convey in the
guidance. In other words, my hypothesis does
not preclude cases where managers issue guidance that may drive up analysts’
earnings expectations. Although the incentives for firms with conservative
accounting to issue earnings guidance may come from meeting or beating
analysts’ earnings expectations, I hypothesize that an objective of such
guidance is to inform analysts about their misestimation of the effect of
accounting conservatism.
If management earnings guidance
successfully informs analysts about the effect of accounting conservatism on
earnings, I expect analysts to revise their forecasts in the direction
indicated by managers’ earnings forecasts.
Specifically, I examine whether subsequent analysts’ revisions
incorporate the news relating to accounting conservatism that is reflected in the
management guidance.
H3: Analysts’ revisions in response to management
forecasts are positively correlated with the effect of accounting conservatism
on earnings indicated by the news in management forecasts.
Results consistent with this hypothesis will validate
my assumption that analysts may not fully understand the effect of
conservatism, and that management earnings guidance informs analysts about its
effect on earnings.
3.
Data, Sample selection, and Research Design
I obtain a sample of management EPS forecasts over the
period 2001-2005 from the CIG (Company Issued Guidance) file maintained by First Call Historical Data. CIG provides management forecasts of EPS with
varying levels of specificity such as point, range, open-ended and
qualitative. For the hypothesis relating
to forecast frequency, I use all types of management EPS forecasts. For the remaining hypotheses, I use point and
range EPS forecasts only. Companies’
financial information is obtained from the annual Compustat industrial database.
To compare with non-forecasting firms, I include all firms that did not
issue any management forecasts during the sample period from the universe of First Call firms.
The sample period is subsequent to
the enactment of the SEC’s Regulation Fair Disclosure (Reg FD) effective October,
2000. Reg FD prohibits firms from
disclosing earnings-related information to select analysts without
simultaneously disclosing the same information to the public. As a result, the CIG coverage of management
forecasts substantially increased after Reg FD.
Heflin, Subramanyam, and Zhang (2003) document that the number of
earnings-related disclosures made by managers more than doubled, from 1,160 in
the three quarters prior to Reg FD to 2,981 in the three quarters following Reg
FD. I exclude the effect of a change in
management forecasting behavior due to Reg FD by selecting a sample period that
is post Reg FD. Hence, this sample
provides a clean setting in which there is minimal private communication
between managers and analysts and news is mostly revealed via public channels.
3.1 Measures of Accounting Conservatism
I use four alternative measures of
accounting conservatism: (i) the sum of R&D reserves, advertising reserves,
and LIFO reserves (SUMCAP), (ii)
accumulated negative non-operating accruals (NOPACR), (iii) the difference in skewness of operating cash flows
and earnings (SKEWNESS), and (iv) a
composite measure which is the sum of decile ranks of the above three measures
(COMPOSITE).
SUMCAP is based on
the measure of accounting conservatism developed by
Penman and Zhang (2002) and is
calculated as:
SUMCAP = RNDCAP + ADVCAP + LIFOCAP
where RNDCAP
is the firm’s estimated R&D asset that would be reported on the balance
sheet if R&D expenditures were capitalized and amortized. Following Amir, Lev, and Sougiannis (2003), I
use a uniform straight-line amortization rate of 20% and calculate RNDCAP as:
RNDCAPt
= 0.9×RDt + 0.7×RDt-1 + 0.5×RDt-2 + 0.3×RDt-3 + 0.1×RDt-4 where RDt is the R&D expenditure
for year t (Compustat #46) and is assumed to be incurred evenly during the
year.[3] ADVCAP
is the estimated asset that would be reported on the balance sheet if
advertising expenditures were capitalized and amortized using the
sum-of-the-years’ digits amortization schedule with a 2-year useful life. That is,
ADVCAP = ADVt + ⅓×ADVt-1
where ADVt
is the advertising expense for year t
(Compustat #45) and is assumed to be
incurred at the end of each year.
Sum-of-the-years’ digits amortization over two years is used because
advertising has a short useful life, typically one or two years, and provides
more benefits when it is initiated (Penman and Zhang, 2002). LIFOCAP
is the LIFO reserve disclosed in a company’s financial statements (Compustat #240). RNDCAP,
ADVCAP, and LIFOCAP are scaled by beginning total assets.
SUMCAP measures “hidden” reserves
resulting from the conservative accounting treatment of specific items. To capture accounting conservatism on a more
general basis, I also use two other measures proposed by Givoly and Hayn
(2000). These are the cumulative amount of negative non-operating accruals and
the skewness of earnings relative to the skewness of operating cash flows.
Givoly and Hayn (2000) argue that
non-operating accruals, which are total accruals excluding changes in working
capital, capture the net effect of accounting items such as “loan losses and
bad debt provisions, restructuring charges, the effect of changes in accounting
estimates, gains or losses on the sale of assets, asset write-downs, the
accrual and capitalization of expenses, and the deferral of revenues and their
subsequent recognition”. They argue that a negative sign of cumulative
non-operating accruals therefore indicates conservative accounting. Following Givoly and Hayn (2000), I use
accumulated negative non-operating accruals (NOPACR) as a measure of accounting conservatism. For a given year, the amount of negative
non-operating accruals is
calculated as:
- (Total Accruals before
depreciation - Operating Accruals)
= - (Net Income + Depreciation – Cash from
Operations – Operating Accruals) where operating accruals include changes in
accounts receivable, inventories, prepaid expenses, accounts payable and income
taxes payable. Negative non-operating accruals for each year are scaled by
beginning total assets. To estimate the
cumulative amount of negative non-operating accruals for each year, NOPACR is calculated as a moving average
over a period of five years ending with the current year.
SKEWNESS
is the difference between the skewness of a firm’s operating cash flows (Compustat #308) and the skewness of its
earnings (Compustat #172) – hence, a
higher positive number indicates a higher level of conservatism. The rationale
for using this measure relates to the fact that the immediate recognition of
expenses and losses and delayed recognition of revenues and gains under
conservative accounting is likely to result in a negatively skewed distribution
of earnings (but not operating cash flows).
I calculate SKEWNESS using annual
data of earnings and operating cash flows (scaled by beginning total assets)
over a period of 10 years ending with the current year. To mitigate the measurement error problem
associated with the first three measures of conservatism (SUMCAP, NOPACR, and SKEWNESS), I use ranked variables scaled
by the number of observations. The
tables report results based on ranked measures of conservatism. Unreported results based on raw measures are
substantially the same.
The above
measures are cumulative or “balance-sheet” measures that capture the level of a
firm’s conservatism or its “hidden” reserves.
I interact these measures with the rate of asset growth to capture the
expected increase or decrease in accumulation of reserves during a given
year. The measure of differential
timeliness in recognizing losses versus gains (Basu, 1997) when measured at the
firm level does not indicate either the level of or the expected change in
reserves during a given year and therefore is not an appropriate candidate for
the purpose of my study.[4]
3.2
Accounting conservatism and frequency of management forecasts
To examine the relation between
accounting conservatism and the frequency of management forecasts, I estimate
the following cross-sectional model:
NUMFCST = a0 + a1CONSERVATISM
+ a2SIZE
+ a3 NUMANALYST
+ a4NEWCAP
+ a5LITIGATE + a6ROA + a7STDRET
+ ε (1) NUMFCST is the number of
forecasts issued by managers in a year. CONSERVATISM is one of the four measures
of accounting conservatism, namely SUMCAP,
NOPACR, SKEWNESS, and COMPOSITE. I expect a1 to be positive if firms
with a higher (lower) level of accounting conservatism issue a larger (smaller)
number of voluntary earnings forecasts.
I control for other factors that may
influence a firm’s voluntary disclosure decision as documented in the
literature. These include firm size,
analyst following, new issuance of capital, litigation risk, firm performance,
and return volatility. Large firms usually make more voluntary disclosures than
small firms since the cost of disclosure tends to decrease in firm size (Lang
and Lundholm, 1993). I include firm size
(SIZE), the logarithm of market value
of equity, as a control variable. A firm is more likely to make voluntary
disclosures when there are more analysts following the firm due to the higher
demand for information about the firm’s prospects. (Lang and Lundholm, 1993,
and Healy and Palepu, 2001). I use the
logarithm of one plus the number of analysts that issue at least one earnings
estimate for a firm as the measure of analyst following (NUMANALYST). The need to
raise capital in the form of stock or debt is also a motivation for manages to
make voluntary disclosures, since increased disclosures are likely to lower the
cost of capital (Frankel, McNichols, and Wilson, 1995). New capital issuance (NEWCAP) is calculated as the sum of the issuance of stock and
long-term debt (Compustat #108 + Compustat #111), scaled by beginning
total assets.
Legal actions against managers for
inadequate or untimely disclosures may encourage firms to increase voluntary
disclosures. Skinner (1997) and Field,
Lowry, and Shu (2005) find empirical evidence that firms with high litigation
risk are more likely to disclose poor performance on a timely basis. I proxy for litigation risk using a dummy
variable LITIGATE, where LITIGATE equals one if a firm belongs to
an industry with high litigation risk, and zero otherwise. Following Francis, Philbrick and Schipper
(1994) and Matsumoto (2002), I classify Biotechnology (SIC 2833 – 2836),
Computer Hardware (SIC 3570-3577), Computer Software (SIC 7371-7379),
Electronics (SIC 3600-3674), and Retailing (SIC 5200- 5961) as industries with
high litigation risk.
Both theoretical and empirical
evidence on the relation between firm performance and voluntary disclosures are
mixed (see Lang and Lundholm, 1993). On
the one hand, firms with poor performance are less likely to make voluntary
disclosures due to the costs associated with these disclosures. On the other hand, firms may either precommit
to a disclosure policy regardless of performance outcome, or be more likely to
disclose bad performance due to concerns about investors’ legal actions
(Skinner, 1994). I include ROA to control for firm
performance. ROA is the annual return on assets, calculated as income from
continuing operations (Compustat #18)
divided by beginning total assets.
Finally, I include return volatility calculated as the standard
deviation of a firm’s annual stock returns over the period 1996-2005 (STDRET) as an additional control
variable. Return volatility is used as
a proxy for the uncertainty associated with the firm’s future performance. Healy and Palepu (2001) argue that higher the
uncertainty, the fewer the voluntary disclosures made by managers. Alternatively, return volatility may capture
the information asymmetry between managers and investors, in which case higher
return volatility may lead to more voluntary disclosures to satisfy investors’
demand for information.
Following Nagar, Nanda, and Wysocki
(2003), all variables except STDRET
are averaged over the sample period. STDRET is calculated over the period
1996-2005 with a minimum requirement of 5 years. I first estimate model (1) using a
cross-sectional OLS regression.[5] In addition, given that the dependent
variable, i.e. the number of forecasts issued by managers, represents count
data (which are non-negative, discrete, and nondecreasing over time), I also
estimate a Poisson regression.[6] Lastly, I estimate model (1) using a probit
regression where the dependent variable equals one if a firm issued at least
one management forecast during the sample period, and zero otherwise. The probit regression tests whether the
likelihood of a firm issuing at least one forecast during the sample period
increases with the level of conservatism.
3.3 Accounting conservatism and news in
management forecasts
I study the
relation between conservative accounting and management forecasting behavior by
examining the “news” in management forecasts (NEWS). I define NEWS as the difference between management
forecasts and the prevailing analyst consensus, and examine 1) how the news in
management forecasts correlates with the level of accounting conservatism, and
2) how this correlation depends on the interaction between change in growth and
level of accounting conservatism. I focus my analysis on one-quarter-ahead
management earnings forecasts and estimate the following cross-sectional and
time-series
(panel) regression:[7]
NEWSit = β0 + β1CONSERVATISMit-1 + β2∆GROWTHit
+ β3CONSERVATISMit-1×∆GROWTHit +β4∆EPSit + ηit (2)
The dependent variable NEWSit is calculated as
one-quarter ahead management EPS forecast minus the median consensus of
analysts’ EPS forecasts for that quarter, where the analyst consensus is the
most recent consensus forecast within a period of 30 days preceding the date of
issuance of the manager’s forecast.[8] The coefficient on CONSERVATISMit-1 is expected to be negative, i.e. the
higher the level of conservatism, the lower is the management forecast relative
to the prevailing analyst consensus. Additionally, I introduce the interaction
term between the change in investment growth and the level of accounting
conservatism. When investment growth
increases, the downward-bias effect of conservative accounting will amplify;
when investment growth decreases, the downward-bias effect of conservative
accounting will attenuate. ∆GROWTHit equals one if growth rate decreases, zero
otherwise. Thus, the coefficient on the
interaction term is predicted to be negative.
Growth rate is measured as the percentage change in total assets from
year t-1 to t.[9] Besides the information about the effect of
conservatism, managers’ forecasts may include other information regarding the
forthcoming earnings which may be correlated with the conservatism
measure. I include change in quarterly
EPS of quarter t relative to quarter t-4 as a proxy for other information in
managers’ EPS forecasts.[10]
3.4 Accounting
conservatism and the effect of management guidance on subsequent analysts’
forecast revisions
The literature
on analysts’ reaction to managers’ forecasts has documented that analysts in
general respond to managers’ forecasts in a timely fashion and revise their
estimates in the same direction as indicated by the news in managers’ forecasts
(for example, Cotter, Tuna and Wysocki, 2006). The interest of my study is to
examine how much of the analysts’ reaction to managers’ forecasts relates to
the effect of accounting conservatism.
To isolate the information in managers’ forecast that only pertains to
the effect of conservative accounting on earnings, I first run a regression of
news in management forecast on the level of conservatism controlling for the
interaction between change in growth and conservatism.
NEWSit = δ0 + δ1CONSERVATISMit-1 + δ2∆GROWTHit
+
δ3CONSERVATISMit-1×∆GROWTHit + µit (3)
The fitted value of regression (3) represents the
news in management forecasts that relates to the effect of accounting
conservatism, and the residual represents other information in managers’
forecasts. I obtain the fitted value (FITNEWS) and the residual (RESIDUAL) for each firm-quarter and then
estimate the following regression:
REVISIONit = γ0 + γ1FITNEWSit + γ2RESIDUALit+ ϕit (4)
The dependent variable is the revision in analysts’
consensus EPS forecast for quarter t
immediately after the management forecast, scaled by the beginning price.
(Analysts’ consensus is calculated over a 30-day period before and after the
management forecast.) The coefficient on FITNEWS
is predicted to be positive since analysts would revise their
forecasts in the same direction as the effect of conservative accounting
inferred from management forecasts. The
coefficient on RESIDUAL is also
predicted to be positive since analysts’ revisions are expected to be
positively correlated with other information in managers’ forecasts.
4. Empirical Results
4.1 Descriptive Statistics
In Table 1, Panel A describes the firm characteristics
of all firms covered by the First Call database. Panel B describes those characteristics for
the sub-sample of firms that issued one or more management earnings forecasts
during the period 2001-2005. All
variables in Table 1 are average values for a firm over 2001-2005. There are a total of 4,049 firms in the First Call sample, with 57% of the
sample (2,305 firms) issuing at least one management forecast over the sample
period. Firms that make EPS forecasts
issue three forecasts in a year on average and are followed by more analysts
compared to the full sample of First Call firms. The mean (median) market value of equity for
the sample of firms with management forecasts is $3,972 ($610) million,
compared with $2,565 ($220) million for the full sample. Forecasting firms are also more profitable:
the mean (median) annual ROA is 1% (3%) for the forecasting sample and –3%
(2%)) for the full sample. Forecasting
firms have higher median market-to-book ratios compared to the full sample,
while the mean market-to-book ratio is similar for both groups. Both
forecasting and non-forecasting firms have similar rates of asset growth. Overall, the summary statistics suggest that
firms that issue management forecasts have a relatively larger market capitalization,
are more profitable, and have a larger analyst following.
4.1.1
Accounting conservatism and analysts’ forecast optimism
To validate my assumption that
analysts’ forecasts are more optimistic when the firm’s accounting is more
conservative, I examine the relation between accounting conservatism and the
error in the consensus analyst forecast immediately preceding the management
forecast. Table 2, panel A reports the
mean and median analysts’ forecast error across high, medium and low levels of
conservatism for the sample of firms that issue at least one management
forecast of quarterly EPS over the sample period. Conservatism is measured by COMPOSITE. Analysts’ forecast error is computed as the
most recent consensus forecast of one-quarter-ahead EPS within a 30-day period
prior to the management forecast minus the actual EPS, scaled by the beginning
price.[11] The mean (median) analysts’ forecast error
for firms with a high level of conservatism is 0.0066 (0.0011), which is
significantly higher than that of firms with a low level of conservatism,
0.0032 (0.0001). The simple correlation
reported in panel B shows that the error in the consensus analyst forecast
preceding the management forecast is positively correlated with the level of
conservatism. Taken together, these
results suggest that analysts’ forecasts appear to be optimistic at least
partly because analysts fail to fully incorporate the effect of accounting
conservatism on earnings (Pae and Thornton, 2003, and Hui and Matsunaga, 2004).
4.2 Relation between accounting conservatism and
the frequency of management
forecasts
Table 3
presents univariate analysis of the relation between the frequency of
management forecasts and accounting conservatism. Panel A reports the mean (median) level of
accounting conservatism of firms that issue (i) frequent (≥ 3), (ii) infrequent (1 or 2) and (iii)
zero management forecasts. The level of
conservatism is measured by COMPOSITE,
SUMCAP, NOPACR, and SKEWNESS,
respectively. The results indicate that
firms that forecast frequently have a higher level of accounting conservatism
relative to firms with zero or infrequent forecasts. Panel B reports the simple correlation
between the frequency of management forecasts (NUMFCST) and all four measures of accounting conservatism.
Consistent with my prediction, the frequency of management forecasts issued by
a firm is positively correlated with the firm’s level of accounting
conservatism. Multivariate regression
results on the relation between the frequency of management forecasts and
accounting conservatism are presented in Table 4. In Panel A, the level of conservatism is
measured by COMPOSITE. The results
from the crosssectional linear regression indicate that the frequency of
management forecasts increases in the level of conservatism. This result remains robust after I account
for the count-data nature of the dependent variable by using a Poisson
regression model. The probit regression
results also support the hypothesis that a firm is more likely to issue one or
more forecasts when its accounting is more conservative. The estimated coefficients on control
variables are largely consistent with the findings of previous studies. I find that large firms tend to issue more
management forecasts than small firms.
Firms with a larger analyst following exhibit a higher frequency of
management forecasts which supports the information demand hypothesis. The correlation between a firm’s financing
activities in the security market and its management forecast frequency,
however, is not significant. Consistent with Lang and Lundholm (1993) and Nagar
et al. (2003), I find a positive relation between firm performance and the
frequency of management forecasts. In
the OLS and Poisson estimation, return volatility (STDRET) is found to be negatively correlated with the frequency of
management forecasts, which implies that the higher the uncertainty, the fewer
management forecasts a firm issues. The
estimated coefficient on STDRET is
positive and insignificant in the probit regression.
I estimate the same OLS, Poisson and
probit regressions using the components of the COMPOSITE conservatism measure, that is, SUMCAP, NOPACR, and
SKEWNESS. The estimated coefficients
on the level of conservatism and control variables are consistent with the
results in Panel A. Since the Poisson
model is better specified than the OLS, only the results of Poisson and probit
estimations are reported in panel B (OLS estimation yields similar
results). The estimated coefficient on
the level of conservatism is positive and significant as expected for all
measures. Collectively, both the
univariate and multivariate results support H1, that a higher level of accounting
conservatism is associated with higher frequency of management earnings
forecasts.[12]
4.3 Relation between accounting conservatism and
news in management forecasts
In this section, I report the results
of the analysis testing whether managers issue forecasts to inform analysts
about the effect of accounting conservatism.
Table 5 presents results of the univariate analysis of the relation
between accounting conservatism and news in management forecasts. NEWS
is calculated as the difference between the management forecast of
one-quarter-ahead EPS and the most recent analyst consensus forecast within a
period of 30 days preceding the date of issuance of the management forecast,
scaled by the beginning price. Panel A
tabulates the mean and median NEWS
for firms in the high, medium and low conservatism groups. For firms with a high level of conservatism,
the mean (median) NEWS is 0.0003
(-0.0001), while for firms with a low level of conservatism, the mean (median) NEWS is 0.0018 (0.0000). Although on
average managers appear to provide good news relative to the prevailing
consensus, the magnitude of this news is lower for higher levels of
conservatism. Panel B presents the
simple correlation between NEWS and
the level of conservatism. The
correlation is significantly negative for all measures of conservatism.
Table 6 reports
results of the multivariate analysis. In
panel A, conservatism is measured by COMPOSITE. After controlling for other information that
managers may convey through their forecasts, the estimated coefficient on COMPOSITE is -0.0043, significant at the
1% level. This is consistent with H2a
that the higher the level of accounting conservatism, the lower should be the
difference between the management forecast and the prevailing analyst consensus. I use change in EPS for quarter t relative
be confounded by the fact that
communication was observable for some firms but not observable for others in
their sample. In addition, their sample
included other types of management forecasts besides earnings forecasts (for
example, revenue and growth) which may be affected by conservatism differently.
to quarter t-4
(DEPS)
as the proxy for other information conveyed through management forecasts. As expected, the estimated coefficient on DEPS
is significantly positive. Panel A also
reports the results of the analysis testing how the change in investment growth
(DGROWTH),
conditional on the level of conservatism (COMPOSITE),
is associated with the news in management forecasts. DGROWTH is an indicator variable that
equals one if the change in total asset growth from t-1 to t is negative and
zero otherwise. Consistent with H2b, I
find a significant and positive coefficient on the interaction between DGROWTH
and COMPOSITE, indicating that the
effect of conservatism reflected in management forecasts is smaller when
investment growth is declining. The
estimated coefficient on COMPOSITE is
significantly negative, indicating that the effect of conservatism reflected in
management forecasts is larger when investment growth is increasing. The estimated coefficient on DEPS
is significantly positive, indicating that managers also convey information
about other events that may affect the forthcoming reported earnings. Regression results without controlling for
other information in management forecasts (DEPS) are similar but obtain a much
smaller R2. The results in
panel B support the same conclusion when the level of accounting conservatism
is measured by SUMCAP, NOPACR, and SKEWNESS. Overall, the results in Tables 5 and 6
provide evidence that managers issue earnings forecasts to convey the effect of
accounting conservatism in an effort to influence analysts to adjust the
misestimation of accounting conservatism in their forecasts.
4.4 Relation between accounting conservatism
and analysts’ revision in response to management forecasts
Table 7 presents the results of the
analysis of analysts’ revisions in response to the news in management forecasts
(NEWS). Analyst’s forecast revision is the change in
analysts’ consensus calculated over a 30-day period before and after the
management forecast, scaled by the beginning price. The first column in Table 7 reports results
of the univariate regression of analysts’ revisions on NEWS. The significant and
positive coefficient indicates that analysts revise their forecasts in the same
direction as the news in management forecasts. The adjusted R2
suggests that the news in management forecasts accounts for about 24% of
analysts’ revisions. The other four
columns report the results of the relation between analysts’ revisions and the
effect of accounting conservatism reflected in the news conveyed by management
forecasts. The variables, Fitted News and Residual, are the fitted (predicted) value and the residual from a
regression of NEWS on the level of
conservatism, change in growth rate (DGROWTH) and the interaction between
conservatism level and DGROWTH. The level of conservatism is
measured by COMPOSITE, SUMCAP, NOPACR, and SKEWNESS, respectively. The estimated coefficients on Fitted News for all four measures of
conservatism are significantly positive, which suggests that analysts revise
their forecasts to incorporate the effect of accounting conservatism conveyed
by management forecasts. Thus, the results in Table 7 are consistent with
analysts’ forecasts not correctly (or fully) accounting for the effect of
conservatism before the release of management forecasts.
A consequence of analysts
incorporating the effect of conservatism is that the correlation between
analysts’ forecast error and accounting conservatism would become weaker after
analysts revise their forecasts in response to the news in management
forecasts. Untabulated results show that
the simple correlation between analysts’ forecast error, measured using the
consensus subsequent to the issuance of management forecast, and the level of
conservatism is insignificant (p-value = 0.5033). This is in contrast with a significant
positive correlation between the initial
forecast error and the level of conservatism reported in Table 2. Thus, it appears that managers successfully
convey the effect of accounting conservatism that is originally missed by
analysts.
5.
Concluding Remarks
This paper examines how accounting conservatism affects
managers’ forecasting behavior.
Conservative accounting produces a downward bias in reported earnings
and net assets in times of investment growth.
This effect of accounting conservatism intensifies when asset growth
increases and attenuates when asset growth decreases (Penman and Zhang, 2002,
and Li, 2006). If analysts do not fully
understand the downward bias caused by accounting conservatism, their earnings
forecasts are likely to appear optimistic.
Managers, faced with pressure to meet or just beat analysts’ earnings
expectations, may have an incentive to issue earnings guidance to influence
analysts to adjust their earnings forecasts for the effect of accounting
conservatism. I first examine whether the
frequency of management forecasts increases with the level of a firm’s
accounting conservatism. Next, I examine whether management earnings guidance
conveys information particularly about the effect of accounting
conservatism. Finally, I investigate whether
analysts revise their forecasts to incorporate the portion of the news in
management guidance that relates to the effect of accounting conservatism.
My empirical analyses generate the
following main results. First, I find
that both the likelihood and the frequency of management earnings forecasts
increase with the level of accounting conservatism. This result is consistent across the four
alterative measures of accounting conservatism, and remains robust when I use
Poisson estimation to accommodate the count-data nature of the frequency of
management forecasts. Second, I find that the news in management earnings
forecasts is negatively correlated with the level of accounting conservatism.
In addition, this negative correlation is stronger when asset growth is
increasing. Third, subsequent to the issuance of management forecasts, I find
that analysts’ revisions are positively correlated with the effect of
accounting conservatism that is reflected in management earnings
forecasts.
Overall, my results provide evidence
that managers issue earnings guidance that informs analysts about the downward
bias caused by accounting conservatism; analysts infer the effect of accounting
conservatism from the management forecast and revise their forecasts to
incorporate this effect. Prior research
on management forecasts primarily focuses on examining short-term managerial
incentives for issuing forecasts, for instance, incentives to avoid negative
earnings surprises, or to deter litigation.
This study provides evidence of a positive association between the use
of conservative accounting practices and the frequency of management forecasts,
which suggests that both conservative accounting and management forecasts are
part of a firm’s long-term disclosure policy.
Moreover, while prior studies have examined the factors that determine
the frequency of management forecasts, this study contributes to this
literature by investigating the specific type of information conveyed by
management forecasts and how analysts respond to this information.
This study provides an additional
explanation for why we observe analysts’ optimism diminishing over the
forecasting horizon. Evidence suggesting
that managers “talk down” analysts’ forecasts has been attributed to
“expectations management”. This study
suggests that analysts’ initial optimism may reflect their misestimation of the
effect of accounting conservatism and managers provide information to correct
their misestimation. Thus, a downward
revision in analysts’ forecasts subsequent to the issuance of management forecasts
may be due to legitimate information about the effect of conservatism conveyed
by managers and attributing it entirely to expectations management may be
premature.
References:
Aboody, D. and B. Lev. 2000.
“Information Asymmetry, R&D, and Insider Gains.” The
Journal of Finance (December): 2747-2766.
Ahmed, A., B. Billings, and M.
Stanford-Harris. 2002. “The Role of Accounting Conservatism in Mitigating
Bondholder-Shareholder Conflicts over Dividend Policy and in Reducing Debt
Costs.” The Accounting Review
(October): 867-890.
Ajinkya, B. and M. J. Gift. 1984.
“Corporate Managers’ Earnings Forecasts and Symmetrical Adjustments of Market
Expectations.” Journal of Accounting
Research 22(2): 425-444.
Amir, E., B. Lev and T. Sougannis.
2003. “Do Financial Analysts Get Intangibles?” Tel
Aviv University Working paper.
Basu, S. 1997. “The Conservatism
Principle and the Asymmetric Timeliness of
Earnings.”
Journal of Accounting and Economics
24: 3-37.
Bartov, E., D. Givoly, and C. Hayn.
2002. “The Rewards to Meeting or Beating Earnings
Expectations.”
Journal of Accounting and Economics
33: 173-204.
Beatty, A., J. Webber, and J. Yu.
2006. “Conservatism and Debt” Ohio State University
Working paper.
Beaver, W., 1998, Financial
Reporting: An Accounting Revolution, 3rd Edition, Upper
Saddle River, NJ:
Prentice-Hall.
Chen, Shuping. 2004. “Why Do
Managers Fail to Meet Their Own Forecasts?”
University of Washington Working
paper.
Clement, M. and P. Y. Davis-Friday. 2002. “The
Relation between Accounting
Conservatism and Voluntary
Management Earnings Forecast.” University of
Notre Dame Working paper
Core, J. 2001. “Discussion of
Information Asymmetry, Corporate Disclosure, and the Capital Markets: A Review
of the Empirical Disclosure Literature.” Journal
of Accounting and Economics (31): 441-456.
Cotter, J., I. Tuna, and P. Wysocki.
2006. “Expectation Management and Beatable
Targets: How Do Analysts React to Explicit Earnings
Guidance.” Contemporary Accounting
Research 23(3): 593-624.
Degeorge, F., Patel, J., Zeckhauser,
R. 1999. “Earnings Management to Exceed
Thresholds” Journal of Business 72(1):1-33.
Gigler, F. and T. Hemmer. 2001.
“Conservatism, Optimal Disclosure Policy, and the
Timeliness of Financial Reports.” The Accounting Review 76 (4): 471-493.
Givoly, D. and C. Hayn. 2000. “The
Changing Time-series Properties of Earnings, Cash Flows and Accruals: Has Financial Reporting Become More
Conservative?” Journal of Accounting and
Economics 29 (3): 287-320.
Givoly, D., C. Hayn, and A.
Natarajan. 2007. “Measuring Reporting Conservatism.” The
Accounting Review, forthcoming.
Guay, W. and R. Verrecchia. 2006.
“Discussion of an Economic Framework for Conservative Accounting and Bushman
and Piotroski (2006).” Journal of
Accounting and Economics (42): 149-165.
Field, L., M. Lowry, and S. Shu.
2005. “Does Disclosure Deter or Trigger Litigation?”
Journal of Accounting & Economics 39,
487-507
Francis J., D. Philbrick, and K.
Schipper. 1994. “Shareholder Litigation and Corporate
Disclosure.”
Journal of Accounting & Research 32(2):
137-164.
Frankel, R., M. McNichols, and G. P.
Wilson. 1995. “Discretionary Disclosure and
External Financing.” The Accounting Review 70 (1): 135-150.
Healy, P. and K. Palepu. 2001.
“Information Asymmetry, Corporate Disclosure, and the Capital Markets: A Review
of the Empirical Disclosure Literature.” Journal
of Accounting and Economics (31): 405-440.
Heflin, F., K.R. Subramanyam, and
Y. Zhang. 2003. “Regrulation FD and the Financial
Information Environment: Early
Evidenc.” The Accounting Review 78
(1): 1-37.
Hirst, E., L. Koonce and S.
Venkataraman. 2006. “Management Earnings Forecasts: A
Review and Framework.” The University of Texas at Austin Working
paper.
Hui, K. W. and S. Matsunaga. 2004.
“Empirical Evidence on the Relation between Accounting Conservatism and
Management Forecast Frequency.” Hong Kong University of Science and Technology
Working paper.
Hutton, A. and P. C. Stocken. 2005.
“Effect of Reputation on the Credibility of
Management Forecasts.” Dartmouth
College Working paper.
Kasznik, R. 1999. “On the Relation
between Earnings Management and corporate
Voluntary Disclosure.” Journal of Accounting Research (31)1:57-81.
Kasznik, R., and B. Lev. 1995. “To
Warn or Not to Warn: Management Disclosures in the Face of an Earnings
Surprise.” The Accounting Review 70
(1): 113-134.
Kasznik, R., and M. McNichols.
2002. “Does Meeting Earnings Expectations Matter? Evidence from Analyst
Forecast Revisions and Share Prices.” Journal
of Accounting Research (40)3:727-760.
Kieso, D. E., J. J. Weygandt, and T.
D. Warfield. 2006. Intermediate Accounting. 12th
Edition, John Wiley & Sons,
Inc.
Lafond, R and R. Watts, 2006. “The
Information Role of Conservative Financial
Statements.” Massachusetts Institute of Technology Working
paper.
Lang, M. and R. Lundholm. 1993.
“Cross-sectional Determinants of Analyst Ratings of
Corporate Disclosures.” Journal of Accounting Research 31:
246-271.
Lev, B. and T. Sougannis. 1996.
“The Capitalization, Amortization, and Value-Relevance of R&D.” Journal of Accounting and Economics 21:
107-138.
Li, Z. 2006. “Accounting
Conservatism, Growth, and Valuation.” the University of
Minnesota Working paper.
Matsumoto, D.A. 2002. “Management’s
Incentives to Avoid Negative Earnings
Surprise.” The Accounting Review 77 (3): 483-514.
Mikhail, M., B. Walther, and R.
Willis. 1999. “Does Forecast Accuracy Matter to
Security Analysts?” The Accounting Review 74 (2): 185-200.
Nagar, V., D. Nanda, and P. Wysocki.
2003. “Discretionary Disclosure and Stock-Based
Incentives.”
Journal of Accounting and Economics 34:
283-309.
Pae, J and D. Thornton, 2003. “Do
Analyst Earnings Forecasts Allow for Accounting
Conservatism?” Queen’s University Working paper.
Penman, S., and X. Zhang. 2002.
“Accounting Conservatism, Quality of Earnings, and
Stock Returns.” The Accounting Review 77: 237-264.
Rock, S., S. Sedo, and M.
Willenborg. 2001. “Analysts Following and Count-data
Econometrics.”
Journal of Accounting and Economics 30:
351-373.
Roger, J. and P. C. Stocken. 2005.
“Credibility of Management Forecasts.” The
Accounting Review 80: 1233-1260.
Richardson, S., S. H. Teoh, and P.
D. Wysocki. 2004. “The Walk-down to Beatable Analyst Forecasts: The Role of
Equity Issuance and Insider Trading Incentives.” Contemporary Accounting Research 21 (4): 885-924.
Skinner, D., 1994. “Why Firms
Voluntarily Disclose Bad News.” Journal
of Accounting
Research
32: 38-61.
Skinner, D., 1997. “Earnings
Disclosures and Stockholder Lawsuits.” Journal
of
Accounting and Economics 23:
249-282.
Skinner, D. and R. Sloan. 2002.
“Earnings Surprises, Growth Expectations and Stock
Returns or Don’t Let an Earnings
Torpedo Sink.” Review of Accounting
Studies 7:
289-312.
Stickel, S., 1992. “Reputation and
Performance among Analysts.” Journal of
Finance 47: 1811-1836.
Tasker, S. 1998. “Bridging the
Information Gap: Quarterly Conference Calls as a
Medium for Voluntary Disclosure.” Review of Accounting Studies 3: 137-167.
Verrecchia, R. 2001. “Essays on
Disclosure.” Journal of Accounting and
Economics 32:
97-180.
Watts, R., 2003 (a) “Conservative
in Accounting part I: explanations and implications”
Accounting Horizons, 17(3),
207-221.
Watts, R., 2003 (b) “Conservative
in Accounting part II: evidence and research opportunities” Accounting Horizons, 17(4),
287-301.
Williams, P., 1996. “The Relation
between a Prior Earnings Forecast by Management and
Analyst Response to a Current Management Forecast.” The Accounting Review 71: 103-115.
Table 1
Descriptive Statistics of firm
characteristics for the full sample of First
Call firms and the subsample of firms that issue management earnings
guidance
Panel A. Full sample
N
|
Mean
|
Median
0.20
|
1st Quartile 0.00
|
3rd Quartile 2.80
|
Std. Deviation
|
|
Frequency of management forecasts
|
4049
|
1.84
|
2.88
|
|||
Analyst following
|
4049
|
7.95
|
3.60
|
0.20
|
12.00
|
10.69
|
Market value of equity ($millions)
|
4049
|
2,564.80
|
220.11
|
5.08
|
1,066.40
|
11,908.00
|
Market-to-book ratio
|
4049
|
1.92
|
1.38
|
1.09
|
2.08
|
1.63
|
Growth in total assets
|
4049
|
0.12
|
0.07
|
0.00
|
0.17
|
0.29
|
Return on assets
|
4049
|
-0.03
|
0.02
|
-0.03
|
0.06
|
0.23
|
Panel B. Sub-sample of firms that
issued one or more management forecasts over
2001-2005
N
|
Mean
|
Median
2.20
|
1st Quartile 0.60
|
3rd Quartile 5.00
|
Std. Deviation
|
|
Frequency of management forecasts
|
2305
|
3.23
|
3.17
|
|||
Analyst following
|
2305
|
12.01
|
8.60
|
3.40
|
17.00
|
11.62
|
Market value of equity ($millions)
|
2305
|
3,971.63
|
609.65
|
172.33
|
2,101.31
|
14,995.00
|
Market-to-book ratio
|
2305
|
1.90
|
1.49
|
1.15
|
2.14
|
1.21
|
Growth in total assets
|
2305
|
0.12
|
0.08
|
0.00
|
0.18
|
0.25
|
Return on assets
|
2305
|
0.01
|
0.03
|
-0.01
|
0.07
|
0.14
|
Frequency of management forecasts is the number of
management earnings forecasts issued by a firm in a year. Analyst following is
the number of analysts that issued at least one earnings estimate for a firm in
a year. Market value of equity equals a
firm’s security price times its common shares outstanding at the end of the
year. Market-to-book ratio is the ratio of market value of equity over the book
value of equity. Growth in total assets
is the percentage change in total assets from year t-1 to t. All variables
are average values for each firm over the period 2001-2005.
Table 2
Relation between accounting
conservatism and errors in analysts’ forecasts issued prior to management
forecasts
Panel A: Analysts’ forecast errors
across high, medium, and low levels of conservatism measured by COMPOSITE
High Conservatism
|
Analysts’ forecast error
Mean Median
|
|
0.0066
|
0.0011
|
|
(0.0001)
|
(0.0001)
|
|
Medium Conservatism
|
0.0036
|
0.0000
|
(0.0001)
|
(0.0435)
|
|
Low Conservatism
|
0.0032
|
0.0001
|
(0.0001)
|
(0.0004)
|
|
High - Low
|
0.0034
(0.0001)a
|
0.0009
(0.0001)b
|
Panel B: Pearson Correlation between
analysts’ forecast error and measures of accounting conservatism
COMPOSITE
8327
|
SUMCAP
9058
|
NOPACR
8823
|
SKEWNESS
|
|
N
|
8685
|
|||
Analysts’ forecast error
|
0.0531
|
0.0557
|
0.0843
|
0.0431
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
a p-value for the t-test of difference in means
of the high versus the low conservatism groups. b p-value for the test of difference in medians
of the high versus the low conservatism groups.
P-values are reported in parentheses.
Analysts’ forecast error is calculated as the most
recent consensus forecast of one-quarter-ahead EPS within a 30-day period prior
to the management forecast, minus the actual EPS for that quarter, scaled by
the beginning price. COMPOSITE is calculated as the sum of decile ranks of
SUMCAP, NOPACR, and SKEWNESS. SUMCAP is the sum of the estimated R&D
reserve, the estimated advertising reserve, and the LIFO reserve, scaled by the
beginning total assets. NOPACR is the
accumulated negative non-operating accruals, scaled by the beginning total
assets. SKEWNESS is the difference between the skewness of operating cash flows
and the skewness of earnings, where earnings and operating cash flows are
scaled by the beginning total assets.
SUMCAP, NOPACR, and SKEWNESS are the ranks of variables scaled by the
number of observations. All conservatism
measures are averages of ranks over the sample period, 2001-2005.
Table 3
Relation between the frequency of
management forecasts and the level of accounting conservatism: Univariate
analysis
Panel A: Level of conservatism for
firms that issue (i) frequent, (ii) infrequent, and (iii) zero management
forecasts
COMPOSITE
|
SUMCAP
|
NOPACR
|
SKEWNESS
|
||||||
N
|
Mean
|
Median
|
Mean
|
Median
|
Mean
|
Median
|
Mean
|
Median
|
|
(i) Frequent
|
1155
|
0.46
|
0.47
|
0.59
|
0.58
|
0.52
|
0.53
|
0.58
|
0.60
|
(ii) Infrequent
|
1150
|
0.45
|
0.46
|
0.58
|
0.52
|
0.50
|
0.50
|
0.54
|
0.54
|
(iii) Zero
|
1744
|
0.43
|
0.43
|
0.56
|
0.41
|
0.49
|
0.46
|
0.44
|
0.40
|
(i) – (iii)
|
0.03
|
0.04
|
0.03
|
0.17
|
0.03
|
0.07
|
0.14
|
0.20
|
|
(0.0001)a
|
(0.0001)b
|
(0.0001)a
|
(0.0001)b
|
(0.0001)a
|
(0.0001)b
|
(0.0001)a
|
(0.0001)b
|
Panel B: Pearson correlation between
the frequency of management forecasts (NUMFCST) and the measures of accounting
conservatism.
COMPOSITE
0.0537
|
SUMCAP
0.0443
|
NOPACR
|
SKEWNESS
|
|
NUMFCST
|
0.0263
|
0.0628
|
||
(0.0019)
|
(0.0048)
|
(0.1198)
|
(0.0002)
|
a p-value for the
t-test of difference in means of the frequent versus the zero forecast groups. b p-value for the test of difference in medians
of the frequent versus the zero forecast groups.
All means and medians reported in panel A are
significant at the 1% level.
P-values are reported in parentheses.
Frequent forecasters are firms that issued three or
more forecasts in a year. Infrequent
forecasters are firms that issued one or two forecasts in a year. NUMFCST is the frequency of management
forecasts measured as the number of forecasts issued by managers in a year
averaged over the sample period, 2001-2005.
The measures of conservatism (defined in Table 2) are averaged over the
sample period, 2001-2005.
Table 4
Relation between the frequency of
management forecasts and the level of accounting conservatism: Multivariate
analysis
NUMFCST = a0 + a1CONSERVATISM + a2SIZE +a3 NUMANALYST
+ a4NEWCAP + a5LITIGATE
+ a6ROA + a7 STDRET + ε
(1)
In the OLS and Poisson models, the
dependent variable is the number of management forecasts (NUMFCST) issued in a
year, averaged over the period 2001-2005.
In the probit model, the dependent variable equals one if a firm issued
at least one management forecast during the period 2001-2005, zero otherwise.
Panel A: Measure of conservatism is
COMPOSITE
OLS
|
Poisson 3370
|
Probit
|
||||
N
|
3370
|
3370
|
||||
Intercept
|
-0.9462
|
-0.8540
|
-1.0008
|
|||
(0.0001)
|
(0.0001)
|
(0.0001)
|
||||
CONSERVATISM
|
+
|
1.0898
|
0.7881
|
0.8032
|
||
(0.0019)
|
(0.0001)
|
(0.0001)
|
||||
SIZE
|
+
|
0.2562
|
0.0370
|
0.0638
|
||
(0.0001)
|
(0.0544)
|
(0.0035)
|
||||
NUMANALYST
|
+
|
0.8571
|
0.5547
|
0.5790
|
||
(0.0001)
|
(0.0001)
|
(0.0001)
|
||||
NEWCAP
|
+
|
0.0354
|
0.0573
|
-0.0169
|
||
(0.7895)
|
(0.4885)
|
(0. 4220)
|
||||
LITIGATE
|
+
|
0.2952
|
0.1112
|
0.0543
|
||
(0.0031)
|
(0.0164)
|
(0.3697)
|
||||
ROA
|
+/-
|
0.8702
|
2.1449
|
0.7156
|
||
(0.0001)
|
(0.0001)
|
(0.0001)
|
||||
STDRET
|
+/-
|
-0.0488
|
-0.1129
|
0.0160
|
||
(0.0084)
|
(0.0001)
|
(0.1816)
|
||||
Adj. R2 / Pseudo R2
|
32.47%
|
38.45%
|
28.60%
|
Panel B: Measures of conservatism are SUMCAP, NOPACR,
and SKEWNESS, respectively
SUMC
|
AP
|
NOPACR
|
SKEWNES
|
S
|
|||||
Poisson
|
Probit
|
Poisson
|
Probit
|
Poisson
|
Probit
|
||||
N
|
4049
|
4049
|
3495
|
3495
|
3380
|
3380
|
|||
Intercept
|
-1.1793
|
-1.4266
|
-0.7685
|
-0.8501
|
-0.7558
|
-0.8409
|
|||
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
||||
CONSERVATISM
|
0.5090
|
0.9204
|
0.2156
|
0.1907
|
0.1663
|
0.1537
|
|||
(0.0001)
|
(0.0001)
|
(0.0115)
|
(0.0541)
|
(0.0131)
|
(0.0755)
|
||||
SIZE
|
0.0304
|
0.0366
|
0.0357
|
0.0488
|
0.0365
|
0.0498
|
|||
(0.1115)
|
(0.0785)
|
(0.0559)
|
(0.0288)
|
(0.0538)
|
(0.0222)
|
||||
NUMANALYST
|
0.6230
|
0.6472
|
0.5678
|
0.6061
|
0.5605
|
0.6017
|
|||
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
||||
NEWCAP
|
0.2505
|
0.0695
|
0.0410
|
-0.0047
|
0.1086
|
-0.1148
|
|||
(0.0008)
|
(0.3977)
|
(0.6131)
|
(0.9605)
|
(0.1669)
|
(0.1683)
|
||||
LITIGATE
|
0.1674
|
0.1217
|
0.1638
|
0.1477
|
0.1935
|
0.1422
|
|||
(0.0006)
|
(0.0785)
|
(0.0001)
|
(0.0101)
|
(0.0001)
|
(0.0143)
|
||||
ROA
|
2.2681
|
1.1096
|
2.0154
|
0.8167
|
2.1150
|
0.7894
|
|||
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
||||
STDRET
|
-0.0560
|
0.0346
|
-0.0888
|
0.0226
|
-0.0903
|
0.0214
|
|||
(0.0306)
|
(0.0152)
|
(0.0022)
|
(0.0569)
|
(0.0001)
|
(0.0982)
|
||||
Pseudo R2
|
32.47%
|
29.36%
|
38.35%
|
28.39%
|
39.33%
|
27.93%
|
P-values are reported in parentheses.
SIZE is the logarithm of a firm’s market value of
equity. NUMANALYST is the logarithm of
the number of analysts (plus one) that issue earnings estimates for a firm.
NEWCAP is the issuance of stock and long-term debt, scaled by the beginning
total assets. LITIGATE equals one if a firm belongs to a high litigation risk
industry, zero otherwise. ROA is the return on assets. STDRET is the return volatility calculated as
the standard deviation of annual returns over 1996-2005. All other variables are average values over
the period 2001-2005. Other variables are defined in Table 2.
40
Table 5
Relation between the news in
management forecasts and accounting conservatism: Univariate analysis
Panel A: The news in management
forecasts across high, medium, and low levels of conservatism measured by
COMPOSITE
High Conservatism
|
NEWS Mean Median
|
|
0.0003
|
-0.0001
|
|
(0.1226)
|
(0.0001)
|
|
Medium Conservatism
|
0.0018
|
0.0000
|
(0.0001)
|
(0.8859)
|
|
Low Conservatism
|
0.0018
|
0.0000
|
(0.0001)
|
(0.0001)
|
|
High – Low
|
-0.0016
(0.0001)a
|
-0.0001
(0.0001)b
|
Panel B: Pearson Correlation between
the news in management forecasts and measures of accounting conservatism
COMPOSITE
-0.0641
|
SUMCAP
-0.0846
|
NOPACR
-0.0224
|
SKEWNESS
|
|
NEWS
|
-0.0295
|
|||
(0.0001)
|
(0.0001)
|
(0.0386)
|
(0.0075)
|
a p-value for the t-test of difference in means
of the high versus the low conservatism groups. b p-value for the test of difference in medians
of the high versus the low conservatism groups.
P-values are reported in parentheses.
The news in management forecasts (NEWS) is calculated
as the difference between the management forecast of onequarter-ahead earnings
and the most recent consensus analyst forecast within a 30-day period prior to
the issuance of the management forecast, scaled by the beginning price. Other variables are defined in Table 2.
Table 6
Relation between the news in
management forecasts and accounting conservatism: Multivariate analysis
NEWSit = β0 + β1CONSERVATISMit-1 + β2∆GROWTHit
+ β3CONSERVATISMit-1×∆GROWTHit +β4∆EPSit + ηit (2)
Panel A: Measure of conservatism is
COMPOSITE
Dependent variable is NEWS
N
|
8045
|
8045
|
8045
|
8045
|
||||
Intercept
|
0.0033
|
0.0013
|
0.0031
|
0.0051
|
||||
(0.0001)
|
(0.0341)
|
(0.0001)
|
(0.0001)
|
|||||
CONSERVATISM
|
-
|
-0.0043
|
-0.0065
|
-0.0040
|
-0.0056
|
|||
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
|||||
DGROWTH
|
-
|
-0.0041
|
-0.0040
|
|||||
(0.0001)
|
(0.0001)
|
|||||||
CONSERVATISM×DGROWTH
|
+
|
0.0046
|
0.0036
|
|||||
(0.0107)
|
(0.0489)
|
|||||||
DEPS
|
+
|
0.0481
|
0.0453
|
|||||
(0.0001)
|
(0.0001)
|
|||||||
Adj. R2
|
3.26%
|
4.24%
|
0.40%
|
1.75%
|
Panel B: Measures of conservatism are
SUMCAP, NOPACR, and SKEWNESS
Dependent variable is NEWS
SUMCAP
|
NOPA
8481
|
CR
8481
|
SKEWNE
|
SS
|
|||||
N
|
8749
|
8749
|
8074
|
8074
|
|||||
Intercept
|
0.0023
|
0.0008
|
0.0018
|
0.0005
|
0.0021
|
0.0007
|
|||
(0.0001)
|
(0.0146)
|
(0.0001)
|
(0.2284)
|
(0.0001)
|
(0.2052)
|
||||
CONSERVATISM
|
-
|
-0.0022
|
-0.0035
|
-0.0010
|
-0.0014
|
-0.0014
|
-0.0026
|
||
(0.0001)
|
(0.0001)
|
(0.0985)
|
(0.1763)
|
(0.0340)
|
(0.0051)
|
||||
DGROWTH
|
-
|
-0.0034
|
-0.0025
|
-0.0032
|
|||||
(0.0001)
|
(0.0002)
|
(0.0001)
|
|||||||
CONSERVATISM×DGROWTH
|
+
|
0.0021
|
0.0013
|
0.0022
|
|||||
(0.0704)
|
(0.2978)
|
(0.0845)
|
|||||||
DEPS
|
+
|
0.0266
|
0.0248
|
0.0481
|
0.0449
|
0.0482
|
0.0448
|
||
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
||||
Adj. R2
|
2.24%
|
3.45%
|
3.13%
|
4.10%
|
2.93%
|
3.94%
|
P-values are reported in parentheses.
DGrowth equals one if growth rate
decreases from year t-1 to t, zero otherwise. Growth rate is measured as the percentage
change in total assets. DEPS is the change in earnings per
share for quarter t relative to
quarter t-4, scaled by the beginning
price. Other variables are defined
in Tables 2 and 5.
Table 7
Analysts’ revisions in response to
the news in management forecasts
REVISIONit = γ0 + γ1FITNEWSit + γ2RESIDUALit + ϕit
(4)
The dependent variable is the revision in analysts’
consensus EPS forecast for quarter t
immediately after the issuance of management forecasts (REVISION), scaled by the beginning price.
COMPOSITE
8027
|
SUMCAP
8701
|
NOPACR
8415
|
SKEWNESS
|
|||
N
|
8027
|
8045
|
||||
Intercept
|
-0.0013
|
-0.0014
|
-0.0015
|
-0.0017
|
-0.0016
|
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
||
NEWS
|
+
|
0.1610
(0.0001)
|
||||
FITNEWS
|
+
|
0.2270
|
0.2550
|
0.3005
|
0.2631
|
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
|||
RESIDUAL
|
+
|
0.1598
|
0.1618
|
0.1650
|
0.1579
|
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
(0.0001)
|
|||
Adj. R2
|
23.79%
|
23.87%
|
23.07%
|
24.51%
|
23.45%
|
P-values are reported in parentheses.
FITNEWS and RESIDUAL are the fitted values and
residuals estimated from the following regression:
NEWS
it = δ0 + δ1CONSERVATISMit-1 + δ2∆GROWTHit
+ δ3CONSERVATISMit-1×∆GROWTHit + µit
(3)
where NEWS equals the difference between the
management forecast and the consensus analyst forecast immediately before the
issuance of the management forecast.
Other variables are defined in Tables 2, 5, and Table 6.
[1] These authors show that a
compensation contract that induces managers’ voluntary disclosures is
beneficial since it creates risk sharing between the firm and its
managers. However, the risk-sharing
benefit decreases in the degree of conservatism. Therefore, one implication of their model is
that, when a firm’s accounting is more conservative, the firm would be less
likely to adopt such a contract, which would result in fewer voluntary disclosures
by managers.
[2] Penman and Zhang (2002)
also provide evidence that, even when a firm applies conservative accounting
consistently, short-term changes in investment growth lead to unpredictable
changes in earnings.
[3] Amir, Lev and Sougannis
(2003) point out that this amortization schedule is roughly the industry
average amortization schedule estimated in Lev and Sougannis (1996).
[4] Additionally, the
calculation of Basu’s differential timeliness measure requires that a firm must
have a certain number of years of negative market returns during the estimation
period. Such a restriction dramatically
decreases the sample size.
[5] Since a firm’s disclosure
behavior in general does not vary much over time, I estimate a cross-sectional
regression using averages of variables over the sample period for each firm
which eliminates problems due to time dependence. I conduct a robustness check by estimating
yearly regressions and obtain substantially similar results.
[6] In the context of studying
analyst following, Rock, Sedo and Willenborg (2001) argue that the Poisson is
more appropriate than the OLS regression when the dependent variable is count
data. In such a situation, the OLS regression
may yield biased and inconsistent results.
[7] I use one-quarter-ahead
management forecasts because managers may have better knowledge of the effect
of conservative accounting on the forthcoming quarter’s earnings given the
shorter forecasting horizon. I also replicate
my analysis for one-year-ahead forecasts and find similar but weaker results.
[8] 8
The calculation of NEWS
is consistent with extant literature on management forecasts, for example,
Hutton and Stocken (2005). The use of
mean instead of median analysts’ consensus yields the same results. Analysts’ forecast consensus data are from
the Summary file provided by First Call.
[9] 9 For robustness check, I
also use capital expenditure growth and sales growth as alternative proxies for
growth. Results are substantially
similar.
[10] To the extent that the
actual change in EPS of the current quarter includes the effect of accounting
conservatism, including it as an independent variable will reduce the
explanatory power of the conservatism variable.
[11] My sample of
one-quarter-ahead management EPS forecasts includes 21,681
firm-quarter-forecasts (17,371 firm-quarters).
The requirement that there be at least one consensus analyst forecast
within a 30day window before and after the issuance of the management forecast
reduces the sample to 10,996 firmquarter-forecasts (9,108 firm-quarters). Additional Compustat data requirements result in the reduction of the sample
size to 8,327 firm-quarter-forecasts for the COMPOSITE conservatism measure as
reported in Table 2.
[12] Clement and Davis-Friday
(2002) and Hui and Matsunaga (2004) find a negative correlation between
accounting conservatism and the frequency of management forecasts. Their contrary result may be due to the fact
that their sample includes the pre-FD period when private communication to
analysts may have occurred but was not observable. Thus, their tests of the frequency of
management forecasts are likely to
No comments